Open Bug 1380675 Opened 8 years ago Updated 1 year ago

Consider to not run GC when InvokeInterruptCallback is called

Categories

(Core :: JavaScript: GC, defect, P5)

defect

Tracking

()

Tracking Status
firefox57 --- affected

People

(Reporter: smaug, Unassigned)

References

(Blocks 1 open bug)

Details

Attachments

(1 file)

Attached patch for testingSplinter Review
When running some benchmarks, like Speedometer, InvokeInterruptCallback gets called and occasionally that triggers a GC slice. That is pretty much the worst time to run GC slice in such cases. Could we do something better? However, Jonco says that running GC there helps with Octane Splay. Pushed a patch to try to get Windows builds for my testing on QF laptop. remote: Follow the progress of your build on Treeherder: remote: https://treeherder.mozilla.org/#/jobs?repo=try&revision=e3cfe0752b48b6a1fb11a8acbedaee3dc20cdae7 remote: remote: It looks like this try push has talos jobs. Compare performance against a baseline revision: remote: https://treeherder.mozilla.org/perf.html#/comparechooser?newProject=try&newRevision=e3cfe0752b48b6a1fb11a8acbedaee3dc20cdae7
Locally I get better numbers in Splay but worse in Splay latency W/O run1 Splay 12859 SplayLatency 8398 run2 Splay 10742 SplayLatency 7722 With run1 Splay 14106 SplayLatency 5855 run2 Splay 13764 SplayLatency 5958 Overall score doesn't seem to change much.
Priority: -- → P5
Severity: normal → S3
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Created:
Updated:
Size: