1562834 - 5.72% raptor-assorted-dom-firefox (linux64-shippable) regression on push 49799d8989589e9efc34091979ab77a5d2ff53c1 (Tue June 25 2019)

Status: RESOLVED INVALID

(Core :: Audio/Video: Playback, defect)

Description

[Alexandru Ionescu (needinfo me) [:alexandrui]]

Raptor has detected a Firefox performance regression from push:

https://hg.mozilla.org/integration/autoland/pushloghtml?changeset=49799d8989589e9efc34091979ab77a5d2ff53c1

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

Regressions:

6% raptor-assorted-dom-firefox linux64-shippable opt 26.93 -> 28.47

You can find links to graphs and comparison views for each of the above tests at: https://treeherder.mozilla.org/perf.html#/alerts?id=21584

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/Performance_sheriffing/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/Performance_sheriffing/Talos/RegressionBugsHandling

Comment 1

[Alastor Wu [:alwu]]

Hi, Emilio,
As it seems to me that bug1321488 was just adding some constraints on the CSS property config file, I have no idea why it would cause performance regression..
Do you have any idea how adding constraint on pseudo element would affect the performance?
Thank you very much!

Comment 2

[Emilio Cobos Álvarez (:emilio)]

I'm very suspect of this alert, since it:

• Doesn't test anything style-system specific (just ran the assorted-dom test and it's mostly just testing DOM getter performance). We have other style system microbenchmarks that would've regressed more I would've guessed.
• Only linux64 regressed, that's a bit odd.

Alexandru, seems like the push that was blamed for in this is:

https://hg.mozilla.org/integration/autoland/pushloghtml?fromchange=4a415941bef62b74389de9520e60fc7cc5cf3962&tochange=49799d8989589e9efc34091979ab77a5d2ff53c1

How sure are we that this patch is the one to blame? Would there be any chance of trying to revert it and confirm this is a regression from that patch? There is some other canvas2d code

Bobby, you're listed as the point of contact for that test, did I read the test source correctly? That test is only testing the perf of accessing DOM properties in different flavors (worker vs. window, named access vs. property access), right? The only remotely possibility of this regression that comes to mind is that switching the property flags array from u8 to u16 changed the location in the final binaries of the webidl statics for the performance object as well, in a way that access to that particular property became slower, but that sounds quite wild...

But maybe I misread that test and I'm missing something.

Comment 3

[Alexandru Ionescu (needinfo me) [:alexandrui]]

Hi Emilio, I will double check this.

Comment 4

[Bobby Holley (:bholley)]

(In reply to Emilio Cobos Álvarez (:emilio) from comment #2)

Bobby, you're listed as the point of contact for that test, did I read the test source correctly?

This test doesn't ring any bells, but I might have just forgotten. Can you point me to the test and/or th place where I'm listed as the point of contact, since maybe then the blame for that will job my memory?

Comment 5

[Emilio Cobos Álvarez (:emilio)]

(In reply to Bobby Holley (:bholley) from comment #4)

This test doesn't ring any bells, but I might have just forgotten. Can you point me to the test and/or th place where I'm listed as the point of contact, since maybe then the blame for that will job my memory?

https://wiki.mozilla.org/Performance_sheriffing/Raptor#raptor-assorted-dom

Comment 6

[Emilio Cobos Álvarez (:emilio)]

For the test I just ran ./mach raptor-test -t raptor-assorted-dom and then poked $objdir/testing/raptor/benchmarks/assorted-dom. Not sure where the upstream source is but I could figure out if needed.

Comment 7

[Bobby Holley (:bholley)]

(In reply to Emilio Cobos Álvarez (:emilio) from comment #5)

(In reply to Bobby Holley (:bholley) from comment #4)

This test doesn't ring any bells, but I might have just forgotten. Can you point me to the test and/or th place where I'm listed as the point of contact, since maybe then the blame for that will job my memory?

https://wiki.mozilla.org/Performance_sheriffing/Raptor#raptor-assorted-dom

Hm, looks like rwood added me as the owner for that last october? https://wiki.mozilla.org/index.php?title=Performance_sheriffing%2FRaptor&type=revision&diff=1202971&oldid=1202970 . rwood, can you shed some light on that?

Comment 8

[Robert Wood [:rwood]]

Not sure why I set you as the owner :bholley, maybe because of the perf-reftests or something (if this isn't correct pls let me know and I'll update the wiki).

The Raptor assorted-dom benchmark originated from AWSY, for more details (including the original test source) please see the bug where I originally landed it:

https://bugzilla.mozilla.org/show_bug.cgi?id=1473365#c0

Comment 9

[Robert Wood [:rwood]]

(In reply to Robert Wood [:rwood] from comment #8)

Not sure why I set you as the owner :bholley, maybe because of the perf-reftests or something (if this isn't correct pls let me know and I'll update the wiki).

The Raptor assorted-dom benchmark originated from AWSY, for more details (including the original test source) please see the bug where I originally landed it:

https://bugzilla.mozilla.org/show_bug.cgi?id=1473365#c0

• AWFY sorry not AWSY

Comment 10

[Bobby Holley (:bholley)]

(In reply to Robert Wood [:rwood] from comment #8)

Not sure why I set you as the owner :bholley, maybe because of the perf-reftests or something (if this isn't correct pls let me know and I'll update the wiki).

I have removed myself as the contact on the wiki.

Comment 11

[Dan Minor [:dminor]]

(In reply to Alexandru Ionescu :alexandrui from comment #3)

Hi Emilio, I will double check this.

Hi Alexandru, could you please retrigger the Talos job with and without Alastor's patch to verify that it did indeed cause the regression seen here? Thank you.

Comment 12

[Alexandru Ionescu (needinfo me) [:alexandrui]]

I found the regression bug, it's not 1321488. Sorry for the error.

Comment 13

[Emilio Cobos Álvarez (:emilio)]

Great, thanks!