Request for review: 8007764: Wrong initialized value of max_gc_pause_sec for an instance of class AdaptiveSizePolicy

Bengt Rutisson bengt.rutisson at oracle.com
Fri Feb 8 16:17:29 UTC 2013


Hi Tao,

I think the change looks good and makes sense.

But I have a hard time estimating the implications of this change.

It seems like MaxGCMinorPauseMillis is only used by CMS and 
ParallelScavenge. For both of those the default for 
MaxGCMinorPauseMillis is the same as MaxGCPauseMillis so it should be fine.

But what happens if a customer is running with 
-XX:MaxGCMinorPauseMillis=1000 but not setting MaxGCPauseMillis on the 
command line? What are the implications of your change for such runs? I 
don't think it is very common. Just thinking.

Also, a more common case is probably to run with only 
-XX:MaxGCPauseMillis set on the command line. How will the adaptive 
sizing behave differently compared to before?

Looking closer at the code it looks like CMS and ParallelScavenge are 
actually using CMSAdaptiveSizePolicy and PSAdaptiveSizePolicy 
respectively. Both of these already pass the correct value to the 
constructor of the super class AdaptiveSizePolicy. So, this bug is 
currently benign. The code that you are changing is actually not being 
used. Is this a correct conclusion?

Thanks,
Bengt

On 2/8/13 3:42 AM, Tao Mao wrote:
> 8007764: Wrong initialized value of max_gc_pause_sec for an instance 
> of class AdaptiveSizePolicy
> https://jbs.oracle.com/bugs/browse/JDK-8007764
>
> webrev:
> http://cr.openjdk.java.net/~tamao/8007764/webrev.00/
>
> changeset:
>




More information about the hotspot-gc-dev mailing list