harmony-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Leo Li" <liyilei1...@gmail.com>
Subject Re: [General VM] GC strategy:how to garbage collect short-lived objects quickly.
Date Thu, 14 Sep 2006 09:31:58 GMT
Dear Xiao-Feng:
     Thank you for your advice.
     I would like generational GC, but what I worry about whether it is
preferrable to let GC start even if there is free memory existing.  Although
the initiative gc fits in my case, I do not know the side-effect of frequent
gc, for example, to pick out gc-able objects, to merge memory and to reset
pointers to moved objects, especially on other cases. In my opinion, the
strategy of current passive gc still has its market. Is it possible to let
it configurable for application developers to choose the gc strategy?
    Smatter compiler to allocate object on stack is really a good way since
many a time an object is used as a local varaible. I think it is not so
difficult for compiler to pick out local variables and what we need is just
to let VM to allow space allocated on stack.:)
    I am not quite familiar with JIT, but it will become a powerful
supplement for static analysis.

Good luck!

On 9/14/06, Xiao-Feng Li <xiaofeng.li@gmail.com> wrote:
> Hi, Dear Leo,
> There are a couple of known approaches to collect short-lived objects.
> The most common approach is generational GC, which is designed
> specifically with the assumption that most objects die young in normal
> applications. Simply put, the objects are arranged into spaces
> according to their age, and the younger objects' spaces are collected
> more frequently. GC pause time is improved since only part of the heap
> is collected normally.
> Another way is to let JIT to free objects whenever it sees
> appropriate. The idea actually is letting JIT to insert object freeing
> code sequence in the generated jitted code, so that the mutator can
> free objects proactively. The "free-me" paper in this year's PLDI
> exprimented this approach but showed this approach helps little with a
> setting of generational GC.
> Stack allocation may help the short-lived objects collection as well,
> which requires escape analysis/detection (by compiler or hardware).
> But my experience was that synchronization removal is the main benefit
> from escape analysis, and stack allocation may not really help in our
> evaluations.
> Which approach is the best for your case may depend on the real
> application behavior. Since generational GC is well-established for
> this problem, we'd take this approach at first. GCv5 proposed is a
> generational GC. We hope it can help to solve the problem you meet.
> Stay tuned... :-)
> Thanks,
> xiaofeng
> On 9/14/06, Leo Li <liyilei1979@gmail.com> wrote:
> > Hi,all:
> >    As we all know, java objects are allocated on heap instead of stack,
> > thus there is a problem about how to garbage collect short-lived objects
> > quickly.
> >    In a recent real project I involved, a server built on java tries to
> > send thousands of messages to client per second. A lot of short-lived
> > messages is created as objects and discarded. (Although I can recycle
> these
> > memory, there is still a byte array created during per call of nio read
> and
> > write.) Since current GC strategy adopted by current RI starts to work
> only
> > when the memory allocated approaching the limit of java heap, the work
> of GC
> > is huge and will raise a wave on the server performance. Furthermore,
> after
> > a long run, although I know GC will merge memory, the operating system
> > reports there is memory fragment and in the worst case the OS will even
> > report real memory is exhausted.
> >    Of course it is possible to limit the java heap so as to force gc
> > frequently as a workround, is it preferrable to collect short-lived
> objects
> > quickly such as adopt aged-related object queues as one of the gc
> strategy?
> >   What about the VMs here, drlvm or J9?
> >
> > Leo Li
> > China Software Development Lab, IBM
> >
> >
> ---------------------------------------------------------------------
> Terms of use : http://incubator.apache.org/harmony/mailing.html
> To unsubscribe, e-mail: harmony-dev-unsubscribe@incubator.apache.org
> For additional commands, e-mail: harmony-dev-help@incubator.apache.org

Leo Li
China Software Development Lab, IBM

  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message