RFR 8042668: GC Support for shared heap ranges in CDS (RE: JDK-8059092)
tom.benson at oracle.com
Mon Jun 8 21:48:46 UTC 2015
On this point:
> > > Performance: The GC changes had no significant impact on SpecJBB,
> > For Specjbb2005, did you check object copy/young gc times specifically?
> > They do not do a lot of GCs in the first place, so an overall
> > regression would require a very large performance regression in
> young gc
> > times to be noticed.
> I only specifically checked full GC times for the reasons I mentioned,
> on impact to benchmarks to show other impact. I can look at that as
> well with
> the new version.
I found no significant difference. I ran jbb'05 at 256 threads and a
12G heap for 4 runs of 30 mins each, with each version. This gave me a
little over 5000 young GC pauses. The average of those for the new
version was about 0.5% faster than the existing version.
More information about the hotspot-gc-dev