1586202 - 5.63 - 25.41% tart / tscrollx (linux64-shippable-qr, windows10-64-shippable, windows10-64-shippable-qr, windows7-32-shippable) regression on push 057b59fdadad75e888a739e85a683b2ff7bfc62e (Wed October 2 2019)

Status: RESOLVED WONTFIX

(Core :: Graphics: WebRender, defect)

Description

[Florin Strugariu [:Bebe]]

Talos has detected a Firefox performance regression from push:

https://hg.mozilla.org/integration/autoland/pushloghtml?changeset=057b59fdadad75e888a739e85a683b2ff7bfc62e

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

Regressions:

25% tart windows10-64-shippable opt e10s stylo 2.68 -> 3.36
23% tart windows7-32-shippable opt e10s stylo 2.67 -> 3.29
11% tscrollx windows10-64-shippable-qr opt e10s stylo 0.87 -> 0.96
6% tart windows10-64-shippable-qr opt e10s stylo 2.34 -> 2.48
6% tscrollx linux64-shippable-qr opt e10s stylo 1.17 -> 1.23

Improvements:

7% tp5o_scroll linux64-shippable-qr opt e10s stylo 2.50 -> 2.33
6% tp5o_scroll windows10-64-shippable-qr opt e10s stylo 1.71 -> 1.60

You can find links to graphs and comparison views for each of the above tests at: https://treeherder.mozilla.org/perf.html#/alerts?id=23322

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/TestEngineering/Performance/Talos

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

Comment 1

[Glenn Watson [:gw] | PTO until 29th April]

I'll take a look at this next week (Monday is a public holiday here).

Comment 2

[Glenn Watson [:gw] | PTO until 29th April]

I took a look at the graphs, but I'm a bit unsure of how to interpret them. Could you take a look at the comments below and see if those are reasonable interpretations?

---

tscrollx win/linux - These tests are measured in milliseconds and are already very low (~1ms of our 16.67 ms frame budget). These tests do tend to move around a bit when we land significant WR changes, since any small change appears to be a large percentage change. So these changes are OK.

tp5o win/linux - These are improvements, so nothing to look at here.

tart win7/win10 - These look like major regressions. However, looking at the graphs, they appear to have had a major improvement on Oct 5, which makes them faster than they were before the regression. They're also not running -qr jobs (which means that WebRender is not running) - so I think these changes are a result of something other than my patch?

tart win10 stylo - This is a regression on a -qr job, but is the same test as the regressions above on non-webrender jobs. So I wonder if that might be related to a different patch too?

---

Could you let me know if the above sounds correct, and if so, which tests you think still need to be investigated as a performance regression related to webrender changes?

Comment 3

[Florin Strugariu [:Bebe]]

you are correct on this...
As the tart win7/win10 appear to have had a major improvement on Oct 5 and the other are minor regressions we can close this as wontfix