[gclist] write-barrier implementation in Java VM?

Okehee Goh ogoh@cise.ufl.edu
Tue, 7 Nov 2000 12:43:33 -0500

> just out of curiosity,  what is the size of the heap for KVM (typically)?
The default heap size is about 15K and it's changable.
> i.e.,  is it worth bothering with an incremental GC for a really tiny
> VM? (which I understand is the whole idea behind KVM, correct me if
> I am wrong)?

You are right. KVM is designed for resource limited system like memory.
I'm also wondering if it's worthwhile.
Because the experimental data measured with incremental GC in KVM
 already showed serious performance degradation.
 ( total completion time is more than 10 times when application is executed
 incremental GC even though elapsed time for just GC is not too much
 I guess it's due to extra overhead such like write-barrier, context

> I found out (the hard way...) that a plain and simple stop-the-world
> approach can be surprisingly effective with small heaps...

I am just curious if there is any known theory  or paper that explains
incremental GC
 is not proper for small heap size system.
Do you have any idea how JVM's GC can support most devices which has small
memory and requires real-time requirements?

-- Okehee

> --Sylvain
> >I'd like to ask about incremental GC of Java VM.
> >I recently implemented incremental GC on KVM, SunMicrosystem's VM for
> >embedded

 Okehee Goh        ogoh@cise.ufl.edu   http://www.cise.ufl.edu/~ogoh
 RealTime System Lab.  (T) 1-352-392-6875
 CISE Dept.  University Of Florida