On Wed, 30 Jan 2019 at 20:07, 'Bryan Mills' via golang-nuts < golang-nuts@googlegroups.com> wrote:
> The code to sequence the results using a channel is not much more verbose. > > That not only avoids the library dependency, but also makes the peak > memory consumption for the results O(runtime.NumCPU()), instead of O(N) > with the number of tasks, and allows the output to be streamed instead of > buffered to a slice. > > https://play.golang.org/p/zkBjxlcvESe > Nice! In practice though, I've usually found that I do want to keep the results around or I don't care about the order at all, so the parallel package works OK. I'd point out one down side to the sequencing approach - one very slow work item can block the others. For example, NProc is 4, the first item takes 200 milliseconds to process and all the others take 1 millisecond, then the first 4 workers have started, none more will be started until the first has, so the overall time will be quite a bit longer (249ms) than if they were all allowed to proceed irrespective of order (200ms). -- You received this message because you are subscribed to the Google Groups "golang-nuts" group. To unsubscribe from this group and stop receiving emails from it, send an email to golang-nuts+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/d/optout.
