1915516 - Re-enable "Fenix-Fission" and "GeckoView-Fission" Android perf tests (on mozilla-central)

Status: RESOLVED FIXED

(Testing :: Raptor, task, P1)

Description

[Chris Peterson [:cpeterson]]

We started running "Fenix-Fission" and "GeckoView-Fission" perf tests on Android two years ago (bug 1768558) to compare the performance impact of Android Fission. We later turned those Fission perf tests off because Android Fission wasn't ready to ship and we didn't need to waste resources running them in automation. But now Android Fission is closer to shipping, so we'd like to re-enable these Fission perf tests.

Comment 1

[Chris Peterson [:cpeterson]]

Sparky recommends we hold off on re-enabling these Fission perf tests until the Samsung A55 migration (bug 1915462) is complete. We'll have more automation capacity then.

Comment 2

[Chris Peterson [:cpeterson]]

To reduce the capacity requirements, maybe we could consider only running these Fission perf tests on mozilla-central.

If we can only afford to test one configuration, we should test "Fenix-Fission" (instead of "GeckoView-Fission") since Fenix is the code that we actually ship to users.

Comment 3

[Florin Bilt]

Attachment: Bug 1915516 -Re-enable "Fenix-Fission" and "GeckoView-Fission" Android perf tests (on mozilla-central) r?#perftest,kshampur

Comment 4

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/02e807e3baac
Re-enable "Fenix-Fission" and "GeckoView-Fission" Android perf tests (on mozilla-central) r=perftest-reviewers,taskgraph-reviewers,kshampur,ahal

Comment 5

[Sandor Molnar[:smolnar]]

https://hg.mozilla.org/mozilla-central/rev/02e807e3baac

Comment 6

[Florin Bilt]

(In reply to Chris Peterson [:cpeterson] from comment #2)

To reduce the capacity requirements, maybe we could consider only running these Fission perf tests on mozilla-central.

If we can only afford to test one configuration, we should test "Fenix-Fission" (instead of "GeckoView-Fission") since Fenix is the code that we actually ship to users.

Hi Chris, thanks for raising the issue, we created the following ticket to resolve this.