How to force finalization?
Locked
 
|
To unit test some of our weak collections, we need to force the system
to garbage collect and finalize all objects that are eligible. By doing things like "ObjectMemory globalGarbageCollect" and waiting a bit, and repeating a few times, we can get it to collect most of the objects most of the time. However, we haven't yet found a way to get it to collect all of the objects all of the time. I don't know enough about how the VM collects weak references to know what else to try. Before I run off and spend a lot of time reading docs and VM sources, does anyone know offhand how this works? These are WeakArray references, not ephemerons. Thanks, -Martin |
RE: How to force finalization?
|
|
If you find a way to do it, please share. I have a few tests to ensure
that our image processing routines don't leak memory, but have yet to find a reliable way of collecting all the garbage to make them pass all the time. They pass most of the time when I do #globalGarbageCollect and #yield a few times, but its still nowhere near 100%. Cheers, -Boris -- +1.604.689.0322 DeepCove Labs Ltd. 4th floor 595 Howe Street Vancouver, Canada V6C 2T5 http://tinyurl.com/r7uw4 [hidden email] CONFIDENTIALITY NOTICE This email is intended only for the persons named in the message header. Unless otherwise indicated, it contains information that is private and confidential. If you have received it in error, please notify the sender and delete the entire message including any attachments. Thank you. -----Original Message----- From: Martin McClure [mailto:[hidden email]] Sent: Friday, January 19, 2007 3:07 PM To: VWNC Subject: How to force finalization? To unit test some of our weak collections, we need to force the system to garbage collect and finalize all objects that are eligible. By doing things like "ObjectMemory globalGarbageCollect" and waiting a bit, and repeating a few times, we can get it to collect most of the objects most of the time. However, we haven't yet found a way to get it to collect all of the objects all of the time. I don't know enough about how the VM collects weak references to know what else to try. Before I run off and spend a lot of time reading docs and VM sources, does anyone know offhand how this works? These are WeakArray references, not ephemerons. Thanks, -Martin |
|
|
I've had pretty good luck in the Glorp tests just repeating
a few times if necessary, i.e.
doesCacheSatisfy: aBlock 10 timesRepeat: [ Dialect garbageCollect. (Delay forMilliseconds: 100) wait. aBlock value ifTrue: [^true]]. ^false At 06:10 PM 1/19/2007, Boris Popov wrote: If you find a way to do it, please share. I have a few tests to ensure --
Alan Knight [|], Cincom Smalltalk Development
"The Static Typing Philosophy: Make it fast. Make it right.
Make it run." - Niall Ross
|
|
|
In reply to this post by Martin McClure
Since finalizing objects may free up other objects to be finalized this is a problem of reaching a fixedpoint. Essentially you want to repeat GCs as long as the previous GC has placed objects into the finalization queue. Whenever something is placed on the finalization queue WeakArray class>>innerFinalizationLoopWith: is called. So you can try to instrument this method (MethodWrapper/DebugProbe/your own override/...) to detect finalization 'events'. Now simply keep GC-ing until no more finilization events are detected. Finalizers runs at priority 90 so you should not need delays to allow for them to run. (I haven't tried this ;-) R - |
|
|
I couldn't resist...
I hacked a way to detect finalization without instrumenting methods, the comment illustrates how to use it to force deep finalization: WeakArray class>> finalizesWhile: aBlock "Return whether objects are finalized during excution of aBlock. Must run at priority < #lowIOPriority" "[WeakArray finalizesWhile: [ ObjectMemory verboseGlobalCompactingGC]] whileTrue:[Transcript tab; show: 'need more:']" | sem continue finalized | sem := Semaphore new. continue := true. finalized := false. [ [sem wait. continue] whileTrue: [finalized := true. FinalizationSemaphore signal]] forkAt: Processor highIOPriority. [self installFinalizationSemaphore: sem. aBlock value] ensure: [self installFinalizationSemaphore: FinalizationSemaphore. continue := false. sem signal]. ^finalized Anybody care to test this in anger? R - On Jan 20, 2007, at 12:18 PM, Reinout Heeck wrote: > > Since finalizing objects may free up other objects to be finalized > this is a problem of reaching a fixedpoint. > > Essentially you want to repeat GCs as long as the previous GC has > placed objects into the finalization queue. > > Whenever something is placed on the finalization queue WeakArray > class>>innerFinalizationLoopWith: is called. So you can try to > instrument this method (MethodWrapper/DebugProbe/your own > override/...) to detect finalization 'events'. > > Now simply keep GC-ing until no more finilization events are detected. > > > Finalizers runs at priority 90 so you should not need delays to > allow for them to run. > > > > (I haven't tried this ;-) > > > R > - > > |
|
|
I published into the open repository as package ForceFinalization. after loading you can ObjectMemory finalizeDeeply. ObjectMemory globalFinalizeDeeply. ObjectMemory verboseFinalizeDeeply. ObjectMemory verboseGlobalFinalizeDeeply. Please let me know if this works, R - On Jan 20, 2007, at 2:23 PM, Reinout Heeck wrote: > I couldn't resist... > > I hacked a way to detect finalization without instrumenting methods, > the comment illustrates how to use it to force deep finalization: > > > > > WeakArray class>> > finalizesWhile: aBlock > "Return whether objects are finalized during excution of aBlock. > Must run at priority < #lowIOPriority" > > "[WeakArray finalizesWhile: [ ObjectMemory > verboseGlobalCompactingGC]] > whileTrue:[Transcript tab; show: 'need more:']" > > | sem continue finalized | > sem := Semaphore new. > continue := true. > finalized := false. > > [ > [sem wait. > continue] whileTrue: > [finalized := true. > FinalizationSemaphore signal]] > forkAt: Processor highIOPriority. > > [self installFinalizationSemaphore: sem. > aBlock value] ensure: > [self installFinalizationSemaphore: FinalizationSemaphore. > continue := false. > sem signal]. > ^finalized > > > > > > Anybody care to test this in anger? > > > R > - > > > > > > > On Jan 20, 2007, at 12:18 PM, Reinout Heeck wrote: > >> >> Since finalizing objects may free up other objects to be finalized >> this is a problem of reaching a fixedpoint. >> >> Essentially you want to repeat GCs as long as the previous GC has >> placed objects into the finalization queue. >> >> Whenever something is placed on the finalization queue WeakArray >> class>>innerFinalizationLoopWith: is called. So you can try to >> instrument this method (MethodWrapper/DebugProbe/your own >> override/...) to detect finalization 'events'. >> >> Now simply keep GC-ing until no more finilization events are >> detected. >> >> >> Finalizers runs at priority 90 so you should not need delays to >> allow for them to run. >> >> >> >> (I haven't tried this ;-) >> >> >> R >> - >> >> > > |
«
Return to VisualWorks
|
1 view|%1 views
| Free forum by Nabble | Edit this page |