Closed
Bug 657032
Opened 13 years ago
Closed 9 years ago
about:memory sawtooth memory usage pattern under odd circumstances
Categories
(Core :: General, defect)
Tracking
()
RESOLVED
WONTFIX
People
(Reporter: jruderman, Unassigned)
References
Details
(Keywords: memory-leak, Whiteboard: [ironic])
Attachments
(1 file)
273.02 KB,
patch
|
Details | Diff | Splinter Review |
(2011-05-13 nightly on Mac) 1. about:memory?verbose 2. javascript:setInterval(sendHeapMinNotifications, 0); void 0 Result: several entries increase steadily forever: heap-used/tjit-data/allocators/main heap-used/other Expected: memory use should be reasonably steady
Updated•13 years ago
|
Whiteboard: [ironic]
Reporter | ||
Comment 1•13 years ago
|
||
If I turn tracejit off in about:config, tjit-data becomes stable (of course) and heap-used/other bounces around within 1MB. So I bet the leak is a tjit leak. (But it would be nice to know what causes the "other" bouncing.)
Comment 2•13 years ago
|
||
Jesse, you are a sadist :P I'll take a look next week.
Comment 3•13 years ago
|
||
AFAICT this is not a true leak. On my Linux64 box, the tjit-data count started at < 1MB, got up to about 12MB before dropping back again to < 1MB, whereupon the cycle was repeated. I did a Massif run, here's the graph: MB 44.75^ # | : : : : # | : :::: ::::::::# | : : :::::: ::::::::::# | : :: ::::::::::::::::: ::::::::::# | :: : :::::::::::::: :::::::::::: ::::::::::# | :::: ::::::::::::::::::: :::::::::::: ::::::::::# ::::::: | :::::::::::: : ::::::::::::::::: :::::::::::: ::::::::::#:::::::::::: | ::::::::::::: : ::::::::::::::::: :::::::::::: ::::::::::#:::::::::::: | ::::::::::::: : ::::::::::::::::: :::::::::::: ::::::::::#:::::::::::: | ::::::::::::: : ::::::::::::::::: :::::::::::: ::::::::::#:::::::::::: | ::::::::::::: : ::::::::::::::::: :::::::::::: ::::::::::#:::::::::::: | :::::::::::::: : ::::::::::::::::: :::::::::::: ::::::::::#:::::::::::: | :::::::::::::: : ::::::::::::::::: :::::::::::: ::::::::::#:::::::::::: | :::::::::::::: : ::::::::::::::::: :::::::::::: ::::::::::#:::::::::::: | :::::::::::::: : ::::::::::::::::: :::::::::::: ::::::::::#:::::::::::: | :::::::::::::: : ::::::::::::::::: :::::::::::: ::::::::::#:::::::::::: | :::::::::::::: : ::::::::::::::::: :::::::::::: ::::::::::#:::::::::::: | :::::::::::::: : ::::::::::::::::: :::::::::::: ::::::::::#:::::::::::: | :::::::::::::: : ::::::::::::::::: :::::::::::: ::::::::::#:::::::::::: 0 +----------------------------------------------------------------------->s 0 786.2 The Massif snapshots show a range of tracer allocation points, nothing terribly surprising. Not surprisingly, if you disable javascript.options.tracejit.content the sawtooth pattern disappears.
Updated•13 years ago
|
Summary: about:memory memory leak → about:memory sawtooth memory usage pattern under odd circumstances
Reporter | ||
Comment 4•13 years ago
|
||
So I guess the tracer just isn't listening for memory-pressure notifications like it should? This makes it a little tricky to use about:memory to tell which *other* actions are causing memory use to rise permanently.
Comment 5•13 years ago
|
||
(In reply to comment #4) > So I guess the tracer just isn't listening for memory-pressure notifications > like it should? Quite possibly. I'm not aware of anywhere in the JS engine where these notifications are listened for. (But I might just not know about them.) CC'ing some JS guys who might know.
Reporter | ||
Comment 6•13 years ago
|
||
The memory pressure observer in nsJSEnvironment.cpp tells the JS engine to GC, so it's probably a good place to tell the JS engine to do other kinds of memory use reduction. http://hg.mozilla.org/mozilla-central/annotate/5ffdf4967dec/dom/base/nsJSEnvironment.cpp#l197
Comment 8•9 years ago
|
||
The trace JIT is long gone. I don't think anything will come from this bug.
Status: NEW → RESOLVED
Closed: 9 years ago
Resolution: --- → WONTFIX
You need to log in
before you can comment on or make changes to this bug.
Description
•