[concurrency-interest] Re: Polling many queues with few threads

Larry Riedel larryr@saturn.sdsu.edu
25 Nov 2004 01:53:05 -0000

> Suppose that we have N queues that are polled (not pooled)
> continously, in a round-robin schedule, by only M thread,
> where M < N. [...] The question is what we should do if all
> queues are empty?

How about one semaphore where consumer threads take/acquire before
polling and the producers put/release?  Would that be inefficient?