1929640 - Add a mode to CCGCScheduler where it can collect faster in case lots of garbage is created and there isn't too much idle time to run the slices

Status: RESOLVED FIXED

(Core :: DOM: Core & HTML, enhancement)

Description

[Olli Pettay [:smaug][[e-mail]]]

If we detect that there isn't enough idle time to run all the slices, we should still schedule (more) collection time to happen during the (possibly short) idle periods we have, since that should be still the best time to do extra work. We might need to overuse the idle time.
GC slices can be also interrupted by TaskController's interrupt mechanism, but that is mostly harmful in case we're already too slowly running the slices.

Comment 1

[Olli Pettay [:smaug][[e-mail]]]

Attachment: Bug 1929640 - Add a mode to CCGCScheduler where it can collect faster in case lots of garbage is created and there isn't too much idle time to run the slices, r=mccr8,sfink

Comment 2

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/67cd0a947b7e
Add a mode to CCGCScheduler where it can collect faster in case lots of garbage is created and there isn't too much idle time to run the slices, r=mccr8,sfink

Comment 3

[amarc]

https://hg.mozilla.org/mozilla-central/rev/67cd0a947b7e

Comment 4

[Mayank Bansal]

For whatever reason, Jetstream2-regex-dna-sp and Jetstream2-regexp really like this change.

24.3% on regex-dna-sp-worst
6.1% on regex-dna-sp-geometric
17%-23% on regexp-worst

Comment 5

[Nicolas Chevobbe [:nchevobbe]]

Perfherder has detected a devtools performance change from push 67cd0a947b7ed4a5244c465a7807d8715cacab7a.

Regressions:

Ratio	Test	Platform	Options	Absolute values (old vs new)
8%	damp source-map-loader.init.DAMP	linux1804-64-shippable-qr	e10s fission stylo webrender	138.82 -> 149.83
8%	damp source-map-loader.init.DAMP	linux1804-64-shippable-qr	e10s fission stylo webrender	138.43 -> 149.12

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

You can run these tests on try with ./mach try perf --alert 42802

For more information on performance sheriffing please see our FAQ.