Open Bug 1421966 Opened 7 years ago Updated 2 years ago

Investigate triggering compaction if the proportion of unused GC things gets too high

Categories

(Core :: JavaScript: GC, enhancement, P3)

enhancement

Tracking

()

Tracking Status
firefox59 --- affected

People

(Reporter: jonco, Unassigned)

References

(Blocks 1 open bug)

Details

(Whiteboard: [MemShrink:P2])

I just saw this line from an about:memory dump go by:

│ │ │ ├──1,464.95 MB (36.98%) ── unused-gc-things

This is pretty high.  We should compute this value after sweeping and perform compacting if it's greater than some threshold, even if compacting was not requested.
Priority: -- → P3
I see things like that frequently, as well as other site "leaks".  Random case with washingtonpost:
│  │  │  ├──1,006.19 MB (45.53%) ── unused-gc-things
after sitting for a few hours in an article.

Even forced GC and CC from about:memory don't recover it; only Minimize (of course)
Whiteboard: [MemShrink]
Whiteboard: [MemShrink] → [MemShrink:P1]
Jon, are there any plans to work on this? Do we still need it?
Flags: needinfo?(jcoppeard)
Whiteboard: [MemShrink:P1] → [MemShrink:P2]
I'm interested in working on this, Can we aim it for 67?

Jon, we can discuss the priority of this with other stuff.
Yes we should do this.

Paul, that sounds good.  We'll have to be careful we don't increase the slice overrun time too much because compacting is not very incremental right now.
Flags: needinfo?(jcoppeard)
Blocks: GCScheduling
Severity: normal → S3
You need to log in before you can comment on or make changes to this bug.