RFR (M) 8149013: Remove unused and dead code from G1CollectorPolicy

Jon Masamitsu jon.masamitsu at oracle.com
Thu Feb 4 18:41:05 UTC 2016



On 02/04/2016 05:02 AM, Mikael Gerdin wrote:
> Hi all,
>
> Here's a cleanup of dead and unused code from the G1 collector policy.
>
> Some unused member variables can be removed:
> _no_of_gc_threads (only getter/setter)
> _parallel_gc_threads (only used locally in constructor)
>
>
> I'd also suggest to remove functionality related to 
> Trace{Young,Old}GenTime from G1. The statistics collected are usually 
> too coarse (avg. pause times over a complete run) to be useful and 
> they have not been kept in sync with the time tracking done through 
> the phase times tracking object.

Mikael,

One of the  things that we don't have at the moment is an easy
way for non-GC engineers to measure GC pauses.  Yes, the gclogapp
and the new aurora reading of the jfr events for pause time is a way
but not as easy a way as having the GC dump the statistics at the
end of the run.   Should we consider fixing and retaining this the
Trace{Young,Old}GenTime to help others help us?  Or is there
something we should implement instead?

Jon

>
> Removing this functionality from G1 leads to removal of:
>
> _stop_world_start (only used by Trace*GenTime)
> record_concurrent_pause()
> print_tracing_info()
> record_stop_world_start()
> etc.
>
> Bug: https://bugs.openjdk.java.net/browse/JDK-8149013
> Webrev: http://cr.openjdk.java.net/~mgerdin/8149013/webrev.0/
> Testing: Local build and local GCBacsher
>
> Thanks
> /Mikael



More information about the hotspot-gc-dev mailing list