RFR 8221629: Shenandoah: Cleanup class unloading logic

Roman Kennke rkennke at redhat.com
Fri Mar 29 14:17:48 UTC 2019


Looks good! Thanks!

Roman

> When ClassUnloading = true, we should not mark through weak CLDS. 
> Therefore, it does not make sense to have weak_cld_closure parameter for 
> RootProcessor::process_strong_roots().
> 
> I also collapse update_all_roots/traversal_update_all_roots methods vs. 
> template.
> 
> Bug: https://bugs.openjdk.java.net/browse/JDK-8221629
> Webrev: http://cr.openjdk.java.net/~zgu/JDK-8221629/webrev.00/index.html
> 
> Test:
>    hotspot_gc_shenandoah on Linux x64 (fastdebug and release)
> 
>    Specjvm with VM following VM options (fastdebug)
>    -XX:+UnlockExperimentalVMOptions -XX:+UseShenandoahGC
>    -XX:+ClassUnloadingWithConcurrentMark
>    -XX:ShenandoahGCHeuristics=traversal
>    -XX:ShenandoahUnloadClassesFrequency=1 -XX:+UnlockDiagnosticVMOptions
>    -XX:+ShenandoahVerify
> 
> Thanks,
> 
> -Zhengyu


More information about the hotspot-gc-dev mailing list