1613400 - 3.56 - 8.61% raptor-tp6-slides-firefox-cold fcp / raptor-tp6-wikipedia-firefox-cold fcp (linux64-shippable-qr, windows10-64-shippable-qr) regression on push 6deaa5a356b2106c4393357b365280ca54ca7d22 (Tue February 4 2020)

Status: RESOLVED WONTFIX

(Core :: Graphics: WebRender, defect, P3)

Description

[Marian Raiciof [:marauder]]

Raptor has detected a Firefox performance regression from push:

https://hg.mozilla.org/integration/autoland/pushloghtml?changeset=6deaa5a356b2106c4393357b365280ca54ca7d22

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

Regressions:

9% raptor-tp6-wikipedia-firefox-cold fcp windows10-64-shippable-qr opt 1,484.00 -> 1,611.75
7% raptor-tp6-wikipedia-firefox-cold fcp windows10-64-shippable-qr opt 1,493.38 -> 1,598.33
6% raptor-tp6-slides-firefox-cold fcp windows10-64-shippable-qr opt 1,491.62 -> 1,582.08
4% raptor-tp6-wikipedia-firefox-cold fcp linux64-shippable-qr opt 1,636.25 -> 1,695.17
4% raptor-tp6-slides-firefox-cold fcp linux64-shippable-qr opt 1,405.42 -> 1,455.42

You can find links to graphs and comparison views for each of the above tests at: https://treeherder.mozilla.org/perf.html#/alerts?id=24853

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 Raptor jobs in a pushlog format.

To learn more about the regressing test(s) or reproducing them, please see: https://wiki.mozilla.org/TestEngineering/Performance/Raptor

*** 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

[Lee Salzman [:lsalzman]]

Nical, do we care about this? It would be kind of annoying to have to pref on some shader permutations for sw-wr only, but I guess we could if really necessary?

Comment 2

[Nicolas Silva [:nical]]

I am probably not the right person to say how much we care about cold startup, so by default I'd say maybe yes? It'd be great if we could compile shaders asynchronously but GL doesn't seem to have an explicit way to do that. Otherwise, having different sets of shader permutations for sw-wr sounds a bit annoying but is fine by me.

Comment 3

[Lee Salzman [:lsalzman]]

Glenn, do you think the cold startup regression is enough of an issue to warrant backing out?

Comment 4

[Glenn Watson [:gw]]

I think we probably do care about a startup regression here, but Jeff is probably a better person to ask about here.

Comment 5

[Glenn Watson [:gw]]

Unless this is a case of that shader permutation not being added to the list of pre-cached / compiled shaders?

Comment 6

[Jeff Muizelaar [:jrmuizel]]

Do we have a theory for why this regressed startup?

Comment 7

[Lee Salzman [:lsalzman]]

It uses the fast image shader for opaque images now whereas previously it used the repeat/antialiasing shader always for opaque. The fast image shader for opaque images wasn't ever getting compiled on wikipedia previously.

Comment 8

[Jeff Muizelaar [:jrmuizel]]

I'm inclined to take this regression. It seems whether we get it or not just depends on the page content and it's only cold startup. Further, I'm hoping that when we do offline shader optimization it will make up for it.

Comment 9

[Lee Salzman [:lsalzman]]

Okay, I'm going to mark this as a resolved/wontfix. If it becomes an issue at some point in the future we can reopen/revisit it.