Closed Bug 1838177 Opened 3 years ago Closed 3 years ago

28.55 - 3.22% cnn-nav.landing LastVisualChange / ebay loadtime + 1 more (Linux) regression on Tue June 6 2023

Categories

(Core :: Web Speech, defect)

Firefox 116
defect

Tracking

()

RESOLVED FIXED
116 Branch
Tracking Status
firefox-esr102 --- unaffected
firefox114 --- unaffected
firefox115 --- unaffected
firefox116 --- fixed

People

(Reporter: alexandrui, Assigned: gerard-majax)

References

(Regression)

Details

(Keywords: perf, perf-alert, regression)

Attachments

(1 file)

Perfherder has detected a browsertime performance regression from push 9458874d5cf930c2e5d987062ef101ea45290d18. As author of one of the patches included in that push, we need your help to address this regression.

Regressions:

Ratio Test Platform Options Absolute values (old vs new) Performance Profiles
29% cnn-nav.landing LastVisualChange linux1804-64-shippable-qr cold fission webrender 4,074.17 -> 5,237.38
26% ebay PerceptualSpeedIndex linux1804-64-shippable-qr cold fission webrender 1,337.58 -> 1,679.42 Before/After
3% ebay loadtime linux1804-64-shippable-qr fission warm webrender 286.47 -> 295.70 Before/After

Improvements:

Ratio Test Platform Options Absolute values (old vs new) Performance Profiles
94% ebay LastVisualChange macosx1015-64-shippable-qr fission warm webrender 4,301.18 -> 255.26
92% ebay LastVisualChange linux1804-64-shippable-qr fission warm webrender 4,321.07 -> 361.72 Before/After
89% ebay SpeedIndex macosx1015-64-shippable-qr fission warm webrender 1,462.02 -> 159.25
88% ebay SpeedIndex linux1804-64-shippable-qr fission warm webrender 1,494.92 -> 173.52 Before/After
86% ebay PerceptualSpeedIndex macosx1015-64-shippable-qr fission warm webrender 935.35 -> 126.53
... ... ... ... ... ...
19% ebay ContentfulSpeedIndex macosx1015-64-shippable-qr cold fission webrender 534.79 -> 431.83

Details of the alert can be found in the alert summary, including links to graphs and comparisons for each of the affected tests. Please follow our guide to handling regression bugs and let us know your plans within 3 business days, or the offending patch(es) may be backed out in accordance with our regression policy.

If you need the profiling jobs you can trigger them yourself from treeherder job view or ask a sheriff to do that for you.

For more information on performance sheriffing please see our FAQ.

Flags: needinfo?(lissyx+mozillians)

Set release status flags based on info from the regressing bug 1432719

:lissyx, are there profiler markers for speechd / webspeech at all that we can use to check if these webpages are deliberately invoking web speech APIs in some way? It seems surprising that the patch in question would have had any impact on performance...

no there are none i think, but we've already have had reports on this where pages are abusing speech synthesis and it triggers the notification. looking at the screenshots in the profile we do see them, so I guess we can just run the raptor tests without the notification enabled

Flags: needinfo?(lissyx+mozillians)

(In reply to Alexandre LISSY :gerard-majax from comment #3)

no there are none i think, but we've already have had reports on this where pages are abusing speech synthesis and it triggers the notification. looking at the screenshots in the profile we do see them,

Ah, OK, that's good confirmation then.

so I guess we can just run the raptor tests without the notification enabled

Well, that would fix the automated regressions reported, but also it makes the test not resemble what happens in the "real world", right? So real world pageload would still be slowed down.

Do we know how much of the overhead is to do with the visual notification and how much is the rest of the code? And/or would flipping the pref just turn off the visual part, but still run the rest of the web speech code?

Also, the regressions also affect macOS - is that expected? I was under the impression the primary usecase was on Linux where web speech can be unavailable for various reasons...

Flags: needinfo?(lissyx+mozillians)
Assignee: nobody → lissyx+mozillians
Status: NEW → ASSIGNED
Flags: needinfo?(lissyx+mozillians)

(In reply to :Gijs (he/him) from comment #5)

(In reply to Alexandre LISSY :gerard-majax from comment #3)

no there are none i think, but we've already have had reports on this where pages are abusing speech synthesis and it triggers the notification. looking at the screenshots in the profile we do see them,

Ah, OK, that's good confirmation then.

so I guess we can just run the raptor tests without the notification enabled

Well, that would fix the automated regressions reported, but also it makes the test not resemble what happens in the "real world", right? So real world pageload would still be slowed down.

I have been told the slowdown is more related in how browsertime interprets data when we display a notification than anything else. the notification is not shown in content itself but chrome

you can see the perf compare links above we get back to normal values

Do we know how much of the overhead is to do with the visual notification and how much is the rest of the code? And/or would flipping the pref just turn off the visual part, but still run the rest of the web speech code?

yes, we still run everything we just dont display the notification

Also, the regressions also affect macOS - is that expected? I was under the impression the primary usecase was on Linux where web speech can be unavailable for various reasons...

this was confirmed as unrelated

Pushed by alissy@mozilla.com: https://hg.mozilla.org/integration/autoland/rev/1f345f63ab5a Never display speech dispatcher missing notification during raptor tests r=perftest-reviewers,AlexandruIonescu,eeejay
Status: ASSIGNED → RESOLVED
Closed: 3 years ago
Resolution: --- → FIXED
Target Milestone: --- → 116 Branch

== Change summary for alert #38691 (as of Thu, 15 Jun 2023 00:38:44 GMT) ==

Regressions:

Ratio Test Platform Options Absolute values (old vs new) Performance Profiles
49% cnn-nav.world SpeedIndex linux1804-64-shippable-qr cold fission webrender 410.52 -> 612.24
9% cnn-nav.world PerceptualSpeedIndex linux1804-64-shippable-qr cold fission webrender 776.25 -> 848.66

Improvements:

Ratio Test Platform Options Absolute values (old vs new) Performance Profiles
49% cnn-nav.landing PerceptualSpeedIndex linux1804-64-shippable-qr cold fission webrender 4,057.97 -> 2,072.75
30% cnn-nav.landing LastVisualChange linux1804-64-shippable-qr cold fission webrender 5,336.54 -> 3,719.86
21% buzzfeed PerceptualSpeedIndex linux1804-64-shippable-qr cold fission webrender 1,299.99 -> 1,028.02 Before/After
19% buzzfeed LastVisualChange linux1804-64-shippable-qr cold fission webrender 2,258.51 -> 1,830.67 Before/After
16% expedia PerceptualSpeedIndex linux1804-64-shippable-qr cold fission webrender 1,561.68 -> 1,305.79 Before/After
... ... ... ... ... ...
11% cnn-nav.landing SpeedIndex linux1804-64-shippable-qr cold fission webrender 2,138.36 -> 1,897.20

For up to date results, see: https://treeherder.mozilla.org/perfherder/alerts?id=38691

You need to log in before you can comment on or make changes to this bug.

Attachment

General

Creator:
Created:
Updated:
Size: