Closed
Bug 748454
Opened 14 years ago
Closed 14 years ago
Use a multi-reporter for jemalloc stats
Categories
(Toolkit :: about:memory, defect)
Tracking
()
RESOLVED
WONTFIX
People
(Reporter: justin.lebar+bug, Unassigned)
References
Details
Calling jemalloc_stats() is not cheap -- it locks the whole heap for a bit. Right now, we call jemalloc_stats once for each relevant memory reporter.
This isn't important for the purposes of about:memory, but it is important inasmuch as we report numbers from jemalloc_stats() numbers in telemetry.
| Reporter | ||
Comment 1•14 years ago
|
||
I timed jemalloc stats, and one call is about .5us on my machine. So I don't think we need to worry about this.
| Reporter | ||
Comment 2•14 years ago
|
||
(In reply to Justin Lebar [:jlebar] from comment #1)
> I timed jemalloc stats, and one call is about .5us on my machine. So I
> don't think we need to worry about this.
(That's with a 400mb heap.)
| Reporter | ||
Updated•14 years ago
|
Status: NEW → RESOLVED
Closed: 14 years ago
Resolution: --- → WONTFIX
Comment 3•14 years ago
|
||
(In reply to Justin Lebar [:jlebar] from comment #1)
> I timed jemalloc stats, and one call is about .5us on my machine. So I
> don't think we need to worry about this.
What about if we're paging?
| Reporter | ||
Comment 4•14 years ago
|
||
I thought jemalloc had to iterate through each chunk in order to collect stats, but it actually iterates over all the arenas, and in our case, that's just one. So it's actually touching very little memory. I suspect it wouldn't be pathologically bad if we were paging.
You need to log in
before you can comment on or make changes to this bug.
Description
•