> However, I think the M:N task is very difficult and might not be achievable > without some other sacrifices on the actual project (speed ? Compile time ? > drastic change in API usage ? safety ? ). But I really don't know.
I have done so many threadpools now, I can do them with closed eyes. Here is a very simple, decent performance threadpool: <https://github.com/status-im/nim-taskpools> As it was written for very high-stakes (billions of dollars secured) and to run 24/7, I intentionally kept it extremely simple and easy to audit and maintain. I use std/tasks here - <https://github.com/status-im/nim-taskpools/blob/d4c4313/taskpools/tasks.nim#L58-L62> but if you can replace by your coroutines, you have easy M:N. The core of a M:N scheduler is there: * 20 lines event loop: <https://github.com/status-im/nim-taskpools/blob/d4c4313/taskpools/tasks.nim#L58-L62> * 40 lines awaiting a future: <https://github.com/status-im/nim-taskpools/blob/d4c4313/taskpools/tasks.nim#L58-L62> * Then there is lots of logic to create a future but it's the same as for a coroutine or creating a closure. Ideally Nim closures are more flexible regarding allocation and you wouldn't need this: <https://github.com/status-im/nim-taskpools/blob/d4c4313/taskpools/tasks.nim#L58-L62> The current Weave-IO is at the moment just a copy paste of Constantine's threadpool: <https://github.com/mratsim/constantine/tree/master/constantine/threadpool> which is an evolution of nim-taskpools for improvement in **compute** (parallel-for). The task system has been significantly improved (single allocation per task instead of 3): <https://github.com/mratsim/weave-io/blob/8672f1cc/weave_io/crossthread/tasks_flowvars.nim#L46-L68> The core has extra bells and whistles but it's not much: * 40 lines event loop: <https://github.com/mratsim/weave-io/blob/8672f1c/weave_io/threadpool.nim#L691-L730> * 100 lines waiting for a future: <https://github.com/mratsim/weave-io/blob/8672f1c/weave_io/threadpool.nim#L691-L730> Also as mentioned here in the changed required for Weave-IO to achieve its vision: <https://github.com/mratsim/weave-io/issues/1> the goal is to be usable with scheduler agnostic async frameworks like <https://github.com/yglukhov/yasync>
