RFR: 8144052: mark_card_deferred does not need to check g1_young_gen
Thomas Schatzl
thomas.schatzl at oracle.com
Thu Nov 26 13:27:54 UTC 2015
Hi,
On Wed, 2015-11-25 at 17:56 +0100, Erik Helin wrote:
> On 2015-11-25, Thomas Schatzl wrote:
> > Hi Erik,
>
> Hi Thomas,
[...]
> > > - GC test suite with full verification
> >
> > looks good, except that I would like to ask you to add an assertion
> > with some good message to G1SATBCardTableModRefBS::mark_card_deferred
> > that checks that we are not calling it on a young gen region.
>
> Unfortunately we can't add such an assert because UpdateRSetDeferred
> will call mark_card_deferred for cards corresponding to young regions
> because we might have gotten an evacuation failure in a young region.
>
> So, UpdateRSetDeferred will call mark_card_deferred for Old and Young
> regions (never Survivor). G1ParScanThreadState will call
> mark_card_deferred for Old regions only (never Young nor Survivor).
Okay.
Thanks,
Thomas
More information about the hotspot-gc-dev
mailing list