[concurrency-interest] ForkJoinPool with custom number of WorkQueues

Doug Lea dl at cs.oswego.edu
Sun Nov 18 17:38:37 EST 2018


On 11/15/18 8:08 PM, Carl Mastrangelo wrote:
> On Sun, Nov 11, 2018 at 6:25 AM Doug Lea via Concurrency-interest
> <concurrency-interest at cs.oswego.edu
> <mailto:concurrency-interest at cs.oswego.edu>> wrote:
> 
>     When steady state has little variance in task execution rates, reducing
>     parallelism *is* the right solution. 

(Although this is not always actionable advice because burstinesss can
be related to choice of Garbage Collector.)

>     >
>     >     One way to fix this (I think) is to lower the number of worker
>     >     queues, but keep the same number of workers.   
> 
> Thanks for the tip.   I tried this out, (and 0x003e), and as you said it
> hardly helped out.  In my profile, most of the time was moved from
> park/unpark over to scan, making it mostly a wash.

I've been working on reducing the excessive scans and park/unparks. If
anyone experiencing this issue could approximate programs showing impact
as performance tests, please get in touch off-list so I can better check
impact.

-Doug



More information about the Concurrency-interest mailing list