1462648 - 2.8 - 4.54% damp (linux64, windows10-64, windows7-32) regression on push 6d96ce0138119a0d7ef38e8e660067aeedd68bd2 (Thu May 17 2018)

Status: RESOLVED INCOMPLETE

(DevTools :: Inspector, defect, P3)

Description

[Ionuț Goldan [:igoldan]]

Talos has detected a Firefox performance regression from push:

https://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?changeset=6d96ce0138119a0d7ef38e8e660067aeedd68bd2

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

Regressions:

  5%  damp windows7-32 pgo e10s stylo     71.71 -> 74.97
  4%  damp windows10-64 opt e10s stylo    74.15 -> 77.06
  3%  damp linux64 pgo e10s stylo         62.18 -> 63.94
  3%  damp windows10-64 pgo e10s stylo    69.98 -> 71.94


You can find links to graphs and comparison views for each of the above tests at: https://treeherder.mozilla.org/perf.html#/alerts?id=13284

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]]

The Gecko profiles will be added soon.

Comment 2

[Joel Maher ( :jmaher ) (UTC -8)]

gecko profiles

before:
https://perf-html.io/from-url/https%3A%2F%2Fqueue.taskcluster.net%2Fv1%2Ftask%2FBe-OZftVRv-0V2zlhMaH5w%2Fruns%2F0%2Fartifacts%2Fpublic%2Ftest_info%2Fprofile_damp.zip

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

Comment 3

[Gabriel Luong [:gl] (ΦωΦ)]

We know enabling the 3 pane inspector was gonna regress performance due to the fact that it was loading more things in the inspector. We are working to mitigate the performance regression by working on various perf work around the inspector. 

Here is a list of that work that has already landed, but we are still doing more:
Bug 1458777 - Lazy load the ClassListPreviewer in CssRuleView. r=pbro [perf]
Bug 1458769 - Cache the SplitBox and TabBar components used in the Inspector. r=pbro [perf]
Bug 1458791 - Lazy load the InplaceEditor, EditingSession and getCssProperties in the BoxModel. r=pbro [perf]
Bug 1458758 - Lazy load Menu, MenuItem and clipboardHelper in StyleInspectorMenu. r=pbro [perf]
Bug 1458751 - Lazy load the StyleInspectorMenu, KeyShortscut and clipboardHelper in CssRuleView and CssComputedView. r=pbro [perf]
Bug 1458745 - Only get the notification box in the inspector for the debugger warning only if the debugger is paused. r=pbro [perf]

Comment 4

[Gabriel Luong [:gl] (ΦωΦ)]

145877, 1458769, 1458791, 1458758, 1458751, 1458745

Comment 5

[Ionuț Goldan [:igoldan]]

How is the work for this bug going?

Comment 6

[Gabriel Luong [:gl] (ΦωΦ)]

I think it's going well. You can see we are getting back to the original baseline of 60 days ago, but still more work needs to be done. http://firefox-dev.tools/performance-dashboard/tools/inspector.html?days=60&filterstddev=true&ignoreFlags=true

Comment 7

[Gabriel Luong [:gl] (ΦωΦ)]

That said I think it is known that the 3 pane inspector would regress perf overall because it is simply loading more than 2 pane mode. I have outfitted this bug as a meta bug to track all bugs that will help improve the overall perf of the inspector. I don't think we will get back to the original baseline before the 3 pane inspector landed in 62 since there is a lot of exploratory work to see where we can improve the performance.

Comment 8

[Dave Hunt [:davehunt] [he/him] ⌚BST]

Closing as INCOMPLETE. Some of the regression was addressed, but after two years I think it's safe to accept the expected regression especially as this doesn't affect the general population of users.