Optimize Math.{min, max} in CacheIR and Warp
Categories
(Core :: JavaScript Engine: JIT, task, P2)
Tracking
()
Tracking | Status | |
---|---|---|
firefox79 | --- | fixed |
People
(Reporter: jandem, Assigned: jandem)
References
(Blocks 1 open bug)
Details
(Keywords: perf-alert)
Attachments
(3 files)
Assignee | ||
Comment 1•5 years ago
|
||
This is a small speedup on micro-benchmarks. Note that CMOV doesn't accept an
immediate so I didn't change that one.
Assignee | ||
Comment 2•5 years ago
|
||
Depends on D78513
Assignee | ||
Comment 3•5 years ago
|
||
Depends on D78514
Updated•5 years ago
|
Comment 5•5 years ago
|
||
bugherder |
https://hg.mozilla.org/mozilla-central/rev/f82b155f57ca
https://hg.mozilla.org/mozilla-central/rev/635b383bccf0
https://hg.mozilla.org/mozilla-central/rev/97c5213ee8b2
Comment 6•5 years ago
|
||
(In reply to Pulsebot from comment #4)
Pushed by jdemooij@mozilla.com:
https://hg.mozilla.org/integration/autoland/rev/f82b155f57ca
part 1 - Use a conditional move in CodeGenerator::visitMinMaxI. r=evilpie
https://hg.mozilla.org/integration/autoland/rev/635b383bccf0
part 2 - Optimize Math.{min,max} with 1-4 arguments in CacheIR and Warp.
r=evilpie
https://hg.mozilla.org/integration/autoland/rev/97c5213ee8b2
part 3 - Add a test. r=evilpie
== Change summary for alert #26179 (as of Thu, 11 Jun 2020 02:42:22 GMT) ==
Regressions:
7% dromaeo_css linux64-shippable-qr opt e10s stylo 18,265.49 -> 16,920.90
7% dromaeo_css windows10-64-shippable opt e10s stylo 17,031.40 -> 15,807.30
7% dromaeo_css windows10-64-shippable-qr opt e10s stylo 16,850.29 -> 15,755.20
6% dromaeo_css windows7-32-shippable opt e10s stylo 18,091.30 -> 16,996.94
6% dromaeo_css windows7-32-shippable opt e10s stylo 18,085.28 -> 17,004.10
For up to date results, see: https://treeherder.mozilla.org/perf.html#/alerts?id=26179
Assignee | ||
Comment 7•5 years ago
|
||
(In reply to Alexandru Ionescu (needinfo me) [:alexandrui] from comment #6)
== Change summary for alert #26179 (as of Thu, 11 Jun 2020 02:42:22 GMT) ==
Regressions:
Later results confirm the numbers went back to normal after bug 1644425 landed.
Comment 8•5 years ago
|
||
Yes, that's why I didn't open a regression bug and only commented here. Just for tracking purposes.
Updated•4 years ago
|
Description
•