RFR(S/M): 7112743: G1: Reduce overhead of marking closure during evacuation pauses

John Cuthbertson john.cuthbertson at oracle.com
Tue Nov 22 17:23:50 UTC 2011


Hi Everyone,

I have a new webrev for this change based upon some review comments from 
Tony. The new webrev can be found at: 
http://cr.openjdk.java.net/~johnc/7112743/webrev.2/

Thanks,

JohnC

On 11/18/11 10:18, John Cuthbertson wrote:
> Hi Everyone,
>
> Can I have a couple of volunteers look over these changes? The webrev 
> can be found at: http://cr.openjdk.java.net/~johnc/7112743/webrev.1/
>
> Summary:
> We have a workload where the serial "complete marking in collection 
> set" closure was taking a realy long time and lengthening the pause 
> times. Since this code might be going away at some point (see CR 
> 6888336:  G1: avoid explicitly marking and pushing objects in survivor 
> spaces) it was decided that a short term solution would be to 
> parallelize the existing code and have the GC workers execute the code 
> rather than the VM thread. Also included are some minor tweaks to the 
> PrintGCDetails output to print out the mark closure time and the 
> collection set freeing time (since some of the original code was moved 
> there).
>
> Testing: GC test suite (with 12 GC threads, makeing threshold of 5%, 
> marking verification enabled, and heap verification after GC enabled); 
> GC test suite on my workstation (4 GC threads, same flags as before); 
> Kitchensink; specjbb2005 configured to have at leat 60% heap utilization.
>
> Thanks,
>
> JohnC




More information about the hotspot-gc-dev mailing list