Eric Holk
This was spuriously closed by a commit message from IU's internal github.
This sounds good to me.
Jeremiah says if we pull this into region analysis, it might be interesting for the paper.
Some work towards this in fd6dd95065a4a3f50ae505981e6f658a4c4d8299 and its parents.
Is there much to this besides just surrounding kernels with an `(if (= 0 (length arg)) (vector) (kernel ...))`? Does this maybe make it harder to do further optimizations?
That seems reasonable.
We could probably do this by having `real_async_trait` generate a bridge trait, although perhaps opt-in with something like `#[real_async_trait(dyn)]`. In my example, we'd generate something like this to go with...
I don't know that we need anything super formal here (and I don't think anything proposed so far crosses into "too formal"). I imagined subcommittee VCs would be basically the...
Do you have suggestions for systems to consider instead?
@rustbot label +I-nominate +T-lang I'd appreciate some feedback. I'm also open to doing a video or chat meeting if that's helpful.