Re: RFC: Reduce GC root set to speed up GC...

[Bryce McKinlay <mckinlay at redhat dot com>]

Andrew Haley wrote:

> Ranjit Mathew writes:
> > David Daney wrote:
> > > Recently there have been some discussions about GC performance issues.
> > > We are also looking at some things in this area.
> > > 
> > > One thing that Hans Boehm suggested was to reduce the size of the root set.
> > > 
> > > I have been thinking about how that might be done.  Here is one idea:
> > 
> > See also:
> > 
> >   http://gcc.gnu.org/ml/java/2002-06/msg00252.html
> >   http://gcc.gnu.org/ml/java-patches/2002-q2/msg00490.html
>
> Well found.  I'd forgotten all about that patch.  Some of it looks
> like it woould still be appropriate.
>
> With regard to scanning instances of class Class, we have a plan to
> use pointer-free class metadata and create them on the fly, so we
> wouldn't need to treat them in any special way.
>  
Yeah. Also, once we have BC-ABI working for CNI, we can allocate storage 
for static class fields on the heap at initialization time (pointer and 
non-pointer fields could be allocated separately). That leaves only 
native fields as static roots. We could provide some special CNI syntax 
for those, implementing them with the .data.gc.roots approach on ELF 
systems or falling back emitting a static constructor call on non-ELF 
targets.

Regards

Bryce