RFR(S): ShenandoahCollectionSet refactor
Zhengyu Gu
zgu at redhat.com
Fri Jun 2 17:11:55 UTC 2017
On 06/02/2017 12:43 PM, Aleksey Shipilev wrote:
> On 06/02/2017 06:29 PM, Zhengyu Gu wrote:
>> http://cr.openjdk.java.net/~zgu/shenandoah/one_cset/webrev.00/
>
> Awesome. One step closer to ditching all RegionSets, and just redoing them as
> the accessors over SH bitmaps.
>
> Comments:
>
> *) ShenandoahCollectionSet::print -- I don't think there is a need in
> debug_only() there, it is not supposed to be highly performant anyway.
Well, the variable is only used for assertion ...
>
> *) Can you turn _cset_map into char*? sizeof(char) == 1 is guaranteed by
> Standard, sizeof(bool), on the other hand...
Last time I was in matrix code, I was wondering if we should use jbyte
instead? hotspot has better support jbyte type.
>
> *) Typo "Should not be":
>
> assert(!in_collection_set(region), "Humongous region should not in collection
> set");
>
Thanks,
> *) So we have removed r->set_in_collection_set(false) in cleanups, because now
> collection_set()->clear() handles the same, right?
>
Yes.
Thanks,
-Zhengyu
> Thanks,
> -Aleksey
>
>
>
More information about the shenandoah-dev
mailing list