RFR (S) 8240868: Shenandoah: remove CM-with-UR piggybacking cycles

Roman Kennke rkennke at redhat.com
Wed Mar 11 12:37:12 UTC 2020


Hi Aleksey,

Very nice!

I see you haven't touched conc-mark. While updating-on-mark is still
used by full-GC, there should be a couple of paths that are unused now
(e.g. in the init-mark parts), do you intend to (carefully) remove them
in a follow-up?

Also, the block here:

http://hg.openjdk.java.net/jdk/jdk/file/e50512f91026/src/hotspot/share/gc/shenandoah/shenandoahHeap.cpp#l1467

is not needed anymore, either. It's only there to conclude the GC cycle,
in the case where the cycle officially (and awkwardly) ends at final-mark.

(We'll probably find more orphaned little blocks related to this in the
future.)

Other than that, it's good.

Roman

> RFR:
>   https://bugs.openjdk.java.net/browse/JDK-8240868
> 
> See the rationale in description.
> 
> Webrev:
>   https://cr.openjdk.java.net/~shade/8240868/webrev.01/
> 
> Testing: hotspot_gc_shenandoah {fastdebug,release}
> 



More information about the shenandoah-dev mailing list