Open Bug 881914 Opened 11 years ago Updated 11 months ago

GC seems to trigger too late when juggling huge structured clones

Categories

(Core :: JavaScript Engine, defect, P3)

Product:
Core
Component:
JavaScript Engine
Type:
defect

Tracking

()

Status:
REOPENED

People

(Reporter: sfink, Unassigned)

References

(Blocks 1 open bug,
URL
)

Details

(Whiteboard: [MemShrink:P2])

In bug 868700, the Broadway h.264 decoder is passing smallish views (~350KB) of large ArrayBuffers (~50MB) between the main and worker threads. Since bug 789593, that results in very rapid memory usage, and about:memory shows over a gigabyte of memory growth. So Broadway should really avoid doing this, but it seems like we should only have a working set size of a few hundred MB. Afaict, the allocations should be one or two 50MB allocations between spins of the event loop.
Whiteboard: [MemShrink]
Whiteboard: [MemShrink] → [MemShrink:P2]
Assignee: general → nobody
Severity: normal → S3
Status: NEW → RESOLVED
Closed: 11 months ago
Resolution: --- → INCOMPLETE

Actually, I should have looked at the profile first. It does still seem we grow memory quite a bit here.

Blocks: GC.performance
Status: RESOLVED → REOPENED
Priority: -- → P3
Resolution: INCOMPLETE → ---
You need to log in before you can comment on or make changes to this bug.