RFR: 8133051: Concurrent refinement threads may be activated and deactivated at random

Kim Barrett kim.barrett at oracle.com
Thu Apr 14 23:23:40 UTC 2016


> On Apr 14, 2016, at 6:51 PM, Derek White <derek.white at oracle.com> wrote:
> On 4/14/16 5:38 PM, Kim Barrett wrote:
>> But we don't want N to be so large that if we have approached that
>> overage when a pause occurs, that it will push us significantly over
>> the update_rs budget.  We base N on ParallelGCThreads because any
>> overage will be distributed among the parallel worker threads at
>> update_rs time.
>> 
> 
> I was going to ask about the use of ParallelGCThreads instead of G1ConcRefinementThreads here. This makes sense. Does the explanation hold for line 145 too?
>  142 static size_t calc_init_green_zone() {
>  143   size_t green = G1ConcRefinementGreenZone;
>  144   if (FLAG_IS_DEFAULT(G1ConcRefinementGreenZone)) {
>  145     green = ParallelGCThreads;
> 
>  146   }
> 
>  147   return MIN2(green, max_green_zone);
> 
>  148 }

Yes.



More information about the hotspot-gc-dev mailing list