RFR (M) JDK-8025305 Cleanup CardTableModRefBS usage in G1

Mikael Gerdin mikael.gerdin at oracle.com
Fri Sep 27 07:09:28 PDT 2013


On 09/27/2013 04:08 PM, Erik Helin wrote:
> Hi Mikael,
>
> the changes looks good, thanks for cleaning this up!

Thanks Erik.
I'm all set to push this change now.

/Mikael

>
> Erik
>
> On 2013-09-24, Mikael Gerdin wrote:
>> Hi all!
>>
>> In preparation for fixing 8014555 (the bug discussed in the "G1
>> question: concurrent cleaning of dirty cards" thread) I'd like to
>> clean up the CardTableModRefBS-code slightly.
>>
>> Summary of change:
>> * Move some g1 specific functions from CardTableModRefBS to the G1
>> specific variant. I did not change the numeric CardValues. I'm
>> afraid to change the last_reserved value since other GCs seem to
>> depend on which bits are available below it.
>> * Change the setup code for G1 to make sure that the BarrierSet is
>> of the one G1 actually requires.
>> * Remove the ModRefBarrierSet* from G1CollectedHeap (it's always the
>> same as CollectedHeap::_barrier_set)
>> * Add an accessor which performs a down-cast of
>> CollectedHeap::_barrier_set to G1SATBCardTableModRefBS.
>> * Modify G1 code which uses the *claimed or *deferred card values to
>> use the type G1SATBCardTableModRefBS.
>> * Remove the unused function CardTableModRefBS::claim_card (last
>> usage was removed in "6923991: G1: improve scalability of RSet
>> scanning" (rev 0414c1049f15)
>>
>> Webrev: http://cr.openjdk.java.net/~mgerdin/8025305/webrev.0/
>> Bug link: https://bugs.openjdk.java.net/browse/JDK-8025305
>>
>> /Mikael



More information about the hotspot-gc-dev mailing list