Closed Bug 1651973 Opened 4 years ago Closed 4 years ago

3.57 - 15.01% raptor-tp6-slides-firefox-cold / raptor-tp6-slides-firefox-cold loadtime (windows10-64-shippable-qr) regression on push d59f0a83a7e9ec6fda8f7327f3154d67b266ec65 (Wed July 8 2020)

Categories

(Core :: Networking: HTTP, defect, P2)

Product:
Core
Component:
Networking: HTTP
Version:
80 Branch
Type:
defect

Tracking

()

Status:
RESOLVED WONTFIX
Tracking Flags:
Tracking Status
firefox-esr68 --- unaffected
firefox-esr78 --- unaffected
firefox78 --- unaffected
firefox79 --- unaffected
firefox80 --- wontfix

People

(Reporter: Bebe, Assigned: CuveeHsu)

References

(Regression)

Details

(Keywords: perf, perf-alert, regression, Whiteboard: [necko-triaged])

Perfherder has detected a raptor performance regression from push d59f0a83a7e9ec6fda8f7327f3154d67b266ec65. As author of one of the patches included in that push, we need your help to address this regression.

Regressions:

15% raptor-tp6-slides-firefox-cold loadtime windows10-64-shippable-qr opt webrender 1,905.17 -> 2,191.17
14% raptor-tp6-slides-firefox-cold loadtime windows10-64-shippable-qr opt webrender 1,935.26 -> 2,205.08
4% raptor-tp6-slides-firefox-cold windows10-64-shippable-qr opt webrender 762.78 -> 790.05

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) will be backed out in accordance with our regression policy.

For more information on performance sheriffing please see our FAQ.

Component: Performance → Networking: HTTP
Flags: needinfo?(juhsu)
Product: Testing → Core

Set release status flags based on info from the regressing bug 1649965

status-firefox78: --- → unaffected
status-firefox79: --- → unaffected
status-firefox-esr68: --- → unaffected
status-firefox-esr78: --- → unaffected

This is interesting. It reverts the benefits from bug 1633935 (which shows in bug 1651072)
I believe bug 1649965 did not only correct some behavior but also improve some performance.
I'll take a look to see if something is unexpected. Thanks.

Assignee: nobody → juhsu
Flags: needinfo?(juhsu)
Priority: -- → P2
Whiteboard: [necko-triaged]

Here's some statistics:
// 1 Set-Cookie
// 2 document has cookie
// 4 permission
161 1 36 0 0 0 4 2

Among ~200 requests for google slides loading,
36 non-main document requests are with LOAD_DOCUMENT_NEEDS_COOKIE and without Set-Cookie.

both LOAD_DOCUMENT_NEEDS_COOKIE and Set-Cookie should block the main thread IPC before bug 1649965
but bug 1633935 failed to block for LOAD_DOCUMENT_NEEDS_COOKIE.

That is, bug 1633935 did things wrong. The benefits are bases on the wrong behavior.
Bug 1649965 do the correct things.

I believe this is an WORKFORME.
Beforehand I don't think it worth improving the performance of LOAD_DOCUMENT_NEEDS_COOKIE, and obviously it's wrong.
Thanks for alerting this, which let us know we have room to improve. I file bug 1652189 for this.

more numbers:
BBC (120 4 158 15 15 0 10 1)
CNN (75 36 21 3 0 0 3 4)

Can we close this out?

Flags: needinfo?(juhsu)

We should close the alarm since it's the correct behavior and just reverse the performance improvement based on the wrong thing.
bug 1652189 is the followup to improve.

Thanks.

Status: NEW → RESOLVED
Closed: 4 years ago
Flags: needinfo?(juhsu)
Resolution: --- → WONTFIX

bug 1652189 is landed, which should help.

Has Regression Range: --- → yes
You need to log in before you can comment on or make changes to this bug.