Taiki Endo

Thanks for the proposal. Is the performance improvement from algorithm changes or the addition of fast-path that doesn't call CAS/RMW or both? Analyzing them would be helpful. In particular, if...

Seems reasonable to me.

Hmm. This could be a bit misleading API, since in the current implementation OptionFuture does not set the inner field to None when the future is complete. We may want...

Thanks for the PR. > Not sure why https://github.com/rust-lang/futures-rs/actions/runs/5046514587/jobs/9052136457?pr=2746#step:5:326 is failing. armv7 failure is unrelated to this PR (should be fixed by rebase). > This makes me wonder whether this...

I believe this is the behavior provided by the fact that the ReadyToRunQueue used inside the FuturesUnordered is a FIFO queue, but I would prefer not to document the exact...

Thanks for the PR! As https://github.com/rust-lang/futures-rs/issues/2541 mentioned, it might be a good idea to mention how to run them in parallel. Also, could you improve the comments on the APIs...

Hmm, I’m not sure if there is a good way to do that without depending on an executor/runtime.

> I am also a bit confused by your comment. In order to run them in parallel you would have to spawn tasks, and this feels a bit out of...

> Would it be possible to add additional helper traits/methods in futures-task for creating waker functions that don't require Arc? Since `Waker` is `Send + Sync`, I don't think we...

I think it is ok to provide this if the seek can provide clear behavior regarding how it affects the take's `limit`. Also, since `Take` already provides `get_pin_mut` method for...