RFR: JDK-8061259: ParNew promotion failed is serialized on a lock

Jungwoo Ha jwha at google.com
Thu Jan 15 22:00:53 UTC 2015


I agree with not having a flag. It seems like an obvious improvement.

On Thu, Jan 15, 2015 at 1:58 PM, Kim Barrett <kim.barrett at oracle.com> wrote:

> On Jan 15, 2015, at 12:00 PM, Jungwoo Ha <jwha at google.com> wrote:
> >
> > Apparently, my memory's been swapped out. I meant to use
> _promotion_failed field.
> >
> > diff -r a184ee1d7172
> src/share/vm/gc_implementation/parNew/parNewGeneration.cpp
> > --- a/src/share/vm/gc_implementation/parNew/parNewGeneration.cpp
> Thu Jan 08 12:08:22 2015 -0800
> > +++ b/src/share/vm/gc_implementation/parNew/parNewGeneration.cpp
> Thu Jan 15 09:00:05 2015 -0800
> > @@ -1194,8 +1194,10 @@
> >          return real_forwardee(old);
> >      }
> >
> > -    new_obj = _next_gen->par_promote(par_scan_state->thread_num(),
> > -                                       old, m, sz);
> > +    if (!_promotion_failed) {
> > +      new_obj = _next_gen->par_promote(par_scan_state->thread_num(),
> > +                                        old, m, sz);
> > +    }
> >
> >      if (new_obj == NULL) {
> >        // promotion failed, forward to self
>
> Yes, this looks right.
>
> Bengt mentioned protection with a flag from the original proposed change.
> I’m not sure there’s a need for a new flag for this version of the change.
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.openjdk.java.net/pipermail/hotspot-gc-dev/attachments/20150115/d0c10471/attachment.html>


More information about the hotspot-gc-dev mailing list