Use of try_join_all hurts Tokio performance significantly about async-file-benchmark HOT 2 OPEN

Ah yes - thanks for bringing this up! I have been aware of the general issues with *_join_all and coop for a while. However, I have not looked at this particular project in some time - possibly before coop was introduced. Am I right in thinking that spawn should avoid those problems? In fact, spawn would even be closer to the original scenario than try_join_all. I believe I switched to try_join_all to avoid abstracting over (and measuring the overhead of) the different spawn interfaces and implementations.

So I tried updating and fixing this and other issues, but I concluded it would be better to simply update the README and archive the project. I originally wrote this as a reproduction case for a ~40x performance difference that no longer exists, and I haven't maintained it. I sincerely doubt that it produces (or ever produced?) useful or reliable metrics for real applications.

from async-file-benchmark.

Yeah, spawn should avoid these issues. If you note the issue in the README, that seems fine to me.

It turns out that there aren't many benchmarks comparing Tokio and async-std, so this is one of the benchmarks people use as it's the only they can find.

from async-file-benchmark.