[concurrency-interest] ParallelArray Extension
tim at peierls.net
Wed Dec 19 17:10:37 EST 2007
On Dec 19, 2007 3:09 PM, Tim Peierls <tim at peierls.net> wrote:
> ...I think the number of Aggregate constructions in combineReduce would be
> only a constant factor (related to parallelism level) less than in
> map-reduce. I don't think it would be directly related to the number of
> threads available to the ForkJoinPool.
I'm wrong on both counts. The method that decides whether to split a
computation of size n or call it a leaf aims for # leaf tasks = 2 * the
parallelism level, and for ForkJoinPool the parallelism level is just the
pool size. It allows more tasks (dynamically) if there are few stealable
tasks, but I have no idea how often this would kick in for the two examples
in this thread.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Concurrency-interest