Skip to content

Count GC's time waiting for a safepoint as GC time #53441

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
mbauman wants to merge 3 commits into
base: master
Choose a base branch
from
Open

Count GC's time waiting for a safepoint as GC time #53441

Changes from 1 commit
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
8b663b0
Count GC's time waiting for a safepoint as GC time
mbauman Feb 22, 2024
c481fe7
track and report waiting time separately
mbauman Feb 22, 2024
1af819d
use the threshold I intended here
mbauman Feb 23, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion base/timing.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -57,7 +57,7 @@ function GC_Diff(new::GC_Num, old::GC_Num)
new.poolalloc - old.poolalloc,
new.bigalloc - old.bigalloc,
new.freecall - old.freecall,
new.total_time - old.total_time,
(new.total_time - old.total_time) + (new.total_time_to_safepoint - old.total_time_to_safepoint),
Copy link
Copy Markdown
Member

@vchuravy vchuravy Feb 22, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I would prefer introducing a new field in GC_Diff that is the safepoint time, and break the top-level report into two sections. As an example when I run a scaling test, I want to see that total_time stays constant (e.g. the work of GC scales well with the number of threads and GC workers). If gc time starts creeping up I might use that as a decision point to use more GC threads.

On the other hand if the latency to start GC is high (time to safepoint) I can use that as a decision point to look with a profiler if I spend to much time in certain ccalls.

Munging them together yields a simple: "GC is slow" but no actionable information.

Copy link
Copy Markdown
Member Author

@mbauman mbauman Feb 22, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yeah, that's fair. In my view, "GC is slow" is very actionable information — and throwing more information at people isn't always helpful. How would you want this to appear? Perhaps we take inspiration from compilation time?

2.39% compilation time: 12% of which was recompilation

new.pause - old.pause,
new.full_sweep - old.full_sweep)
end
Expand Down