search

atomgalaxy / review-executor-sendrecv

A board for reviewing the executors proposal P0440
1 stars 0 forks source link

What kind of sender `exeuction::schedule` produce #9

Open tomaszkam opened 4 years ago

tomaszkam commented 4 years ago

The concept of scheduler does not define what kind of sender, the execution::schedule is intended to produce, in particular, what will be propagated to set_value method of the receiver. From my understanding, produced sender should be void-sender (call set_value()).

If this omission in requirements, or senders produced from schedule all allowed to produce values? If the later is true, how algorithms like via, on should handle these values?

RobertLeahy commented 4 years ago

This might be covered by executors/executors#467.

tomaszkam commented 4 years ago

Yes, it is, thanks. Not sure if I should cose the issue then.

atomgalaxy commented 4 years ago

Closing, does not go into report.

tomaszkam commented 4 years ago

Why? As far as I see the issue is still not resolved in the paper, and reviewing senders was part of our job. The fact that it was found concurrently, does not change fact that issue exits.

In short - the scheduler is underconstrained, because it does not define sensible constrain for the produced sender, as specified today it nearly unusable - you cannot write receiver that would match any scheduler.

atomgalaxy commented 4 years ago

NVM, I misunderstood the status. I wasn't making an "executive decision", this was just a GC attempt gone wrong.

Thanks for watching, Tomasz!