harmony-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Xiao-Feng Li" <xiaofeng...@gmail.com>
Subject Re: [drlvm][gcv5] finalizer design
Date Tue, 19 Dec 2006 01:47:03 GMT
On 12/19/06, Rana Dasgupta <rdasgupt@gmail.com> wrote:
> I would suggest keeping both( native and java finalizer threads )around for
> now  through a unified interface. I am not sure I followed why it is a
> requirement for GCv5 to support the work balance subsystem. But it is true
> that it cannot ultimately rely on a single high priority thread for
> finalization for it  to avoid starvation, prevent denial of service, etc.

Agree. The sheduling of the native finalizing threads need improving.

> Implementing finalizer threads in java was a DRLVM choice some time back,
> based on some boundary crossing efficiency concerns for finalizer threads.
> From testing over quite a long period, it seems stable. Unless proved
> different, or unless we do a significantly better performing VM threadpool
> implemenation around GCV5, I see no reason why it should not be reused just
> because the threads are in Java.

Agree. Java or not itself is not a problem at all. There is no reason
to get rid of the Java thread finalization. I actually think it's a
good idea to keep both implementations. We don't need decide at the
moment which one is better before we have really strong arguments or
benchmarks showing the real differences.

Thanks,
xiaofeng

>
>
>
> On 12/18/06, Pavel Afremov <pavel.n.afremov@gmail.com> wrote:
> >
> > On 12/18/06, Geir Magnusson Jr. <geir@pobox.com> wrote:
> > >
> > >
> > > But don't you then make it harder to port?
> > >
> > >
> >
> > It's makes porting a bit harder, but not very, I think. The main point now
> > to avoid regression. I think GC v5 should support Work Balance Subsystem,
> > or
> > something like it before it becomes main GC for DRLVM. Also GC v5 should
> > works with "java" implementation of finalization system correctly, in any
> > case VM and GC implementations should be independent on each other.
> >
> >
> >
> > BR
> >
> > Pavel Afremov.
> >
> >
> > On 12/18/06, Geir Magnusson Jr. <geir@pobox.com> wrote:
> > >
> > >
> > >
> > > Xiao-Feng Li wrote:
> > >
> > > [SNIP]
> > >
> > > > 3 . We think a native thread Finalizer solution is better than a Java
> > > > thread solution. Since the Java thread actually runs in a native
> > > > thread, we don't need the extra wrapper.
> > >
> > > Really?
> > >
> > > [SNIP]
> > >
> > > >
> > > > Explanations to  3 .
> > > >  - In Java finalizer thread implementation, there exists potential
> > > > circular dependence between the Java thread startup and JVM
> > > > bootstrapping. The bootstrapping issues or bugs with Java code in VM
> > > > were discussed more than once.
> > > >  - In Java finalizer thread implementation, there are rounds of
> > > > redundant steps to do finalization with Java thread. In existing Java
> > > > thread implementation, to execute the finalizers, VM native calls Java
> > > > method startFinalization to wakeup finalizer threads. The finalizer
> > > > Java threads call a native method doFinalization to excute the
> > > > finalizers. This native method accesses native queue and calls Java
> > > > finalizer method again. With a native thread finalizer, it simply
> > > > calls the Java finalizer directly without all other boundary
> > > > crossings.
> > > >  - A java finalizer thread finally maps to a native thread managed by
> > > > VM. This extra mapping is unnecessary.
> > > >  - Finalizer threads are VM internal entities. VM may want to
> > > > schedule it as it wants for load balance or helper threading. This is
> > > > much easier with the direct native threads.
> > > >  - With native thread finalizer, we can share the same thread pool
> > > > with other VM components such as GC, etc. This helps to manage the
> > > > system overall performance and scalability, and it's easier.
> > > >  - DRLVM is in written in C++, its components interact through native
> > > > interfaces. It is natural for VM core components written in native
> > > > code.
> > >
> > > But don't you then make it harder to port?
> > >
> > >
> > > geir
> > >
> >
> >
>
>

Mime
View raw message