Comment by seanhunter
1 year ago
I have seen this exact same argument in python and elsewhere as well. Any time anyone writes a rant against async and suggests threads as an alternative, you know they didn’t really understand what async was for and were using it for something it’s not really for, so you know somewhere in their rant they will say the performance improvement they expected didn’t happen.
Async (in any language) is not a panacea. Async is for allowing multiple things to make progress simultaneously that would otherwise be blocked on I/O. If you thread them your threads will be independently blocked on I/O and you will have additional locking overhead. If you have an embarrassingly parallel task and you aren’t blocked on I/O of course async will be slower than pure parallelism because that’s not what it’s for. It’s almost literally so you can have one async thread consuming exactly 1 CPU doing all the I/O and it will all make good progress.
Threads are for when you COMPUTE a lot, while async/await are for when you WAIT a lot.
Sometimes you are "computing a lot" over very general dataflow graphs and want to have tasks with work stealing. Async frameworks will give you that for free, seamlessly scaling to any number of underlying threads.
interesting. I hope it's not too late to ask: could you expand a little bit, in particular, what do you mean by "over very general dataflow graphs"?
Exactly. Nice rule of thumb.