Fission's Amazon AWFY warm load time regression is mostly fixed, but doesn't match e10s with Olli's paint suppression tweaks
Categories
(Core :: DOM: Content Processes, defect, P3)
Tracking
()
Tracking | Status | |
---|---|---|
firefox-esr78 | --- | unaffected |
firefox-esr91 | --- | unaffected |
firefox92 | --- | unaffected |
firefox93 | --- | unaffected |
firefox94 | - | wontfix |
firefox95 | --- | wontfix |
People
(Reporter: cpeterson, Unassigned)
References
(Regression)
Details
(Keywords: regression)
After Nika's changes for bug 1731792 landed on 2021-09-29, Amazon AWFY warm load time improved a lot but did not reach e10s performance:
- Linux: https://arewefastyet.com/linux64/warm-page-load/overview?numDays=30&series=Firefox,Firefox-Fission
- macOS: https://arewefastyet.com/mac/warm-page-load/overview?numDays=30&series=Firefox,Firefox-Fission
- Windows: https://arewefastyet.com/win10/warm-page-load/overview?numDays=30&series=Firefox,Firefox-Fission
How much do we care? Nika improved Amazon warm load perf a lot (~470 to ~290 ms), actually beating Fission and e10s before the original regressions (~380 ms). The issue is that, in the meantime, Olli's paint suppression tweaks in bug 1731132 improved e10s's Amazon warm load perf but didn't help Fission.
Comment 1•3 years ago
|
||
My take on this would be that we should look at a past baseline for e10s against current fission here and not compare two moving targets that are both influenced by our ongoing changes. I always understood our fission performance goal as a general "things should not become noticeable worse than they were before". I'd strongly suggest to fix a (past) release we want to benchmark fission against. But in the end this seems to be a product decision, so chaining in Mike.
Updated•3 years ago
|
Updated•3 years ago
|
Comment 2•3 years ago
|
||
Both fixes are shipping in Fx94, so users upgrading from an older version of Firefox will see a net improvement of about 100ms. No reason to worry about this.
Reporter | ||
Comment 3•3 years ago
|
||
(In reply to Mike Conca [:mconca] from comment #2)
Both fixes are shipping in Fx94, so users upgrading from an older version of Firefox will see a net improvement of about 100ms. No reason to worry about this.
SGTM. I will resolve this bug as WONTFIX. We have more important performance issues to worry about.
Updated•3 years ago
|
Updated•3 years ago
|
Description
•