Request for review: 7052429: G1: Avoid unnecessary scanning of humongous regions during concurrent marking

Tao Mao tao.mao at oracle.com
Wed Feb 6 22:17:19 UTC 2013


A new (final) webrev is updated.
http://cr.openjdk.java.net/~tamao/7052429/webrev.01/

Added the assertion. Copyright date updated.

JohnC, please help push the changeset with me as the contributor.

Thanks.
Tao

On 2/5/2013 11:05 PM, John Cuthbertson wrote:
> Hi Tao,
>
> Changes look good. It might be a good idea to add an assert after you 
> look up the heap region:
>
> assert(!hr->isHumongous() || mr.start() == hr->bottom(), "...");
>
> so that, if the region is humongous, we verify that the start of 
> memory region is the base of the heap region.
>
> Thanks,
>
> JohnC
>
> On 2/1/2013 6:37 PM, Tao Mao wrote:
>> 7052429: G1: Avoid unnecessary scanning of humongous regions during 
>> concurrent marking
>> https://jbs.oracle.com/bugs/browse/JDK-7052429
>>
>> webrev:
>> http://cr.openjdk.java.net/~tamao/7052429/webrev.00/
>>
>> changeset:
>> Avoid wasted work of _nextMarkBitMap->iterate() when the current 
>> region is a humongous region and the first bit is NOT marked, whereas 
>> the case where a humonguous region with its first bit marked is 
>> properly handled in iterate() routine.
>



More information about the hotspot-gc-dev mailing list