Michael Schellenberger Costa
Michael Schellenberger Costa
The convention is that `std::any` is a type erased thing and resource is what we want, so combining them would yield `any_resource`
Regarding the question about unique_ptr, that should be doable, but what we also want is something that lives on the stack and not the heap
@Artem-B I believe this relates to us making `thrust::tuple` an alias of `cuda::std::tuple` Can you check whether it works with https://github.com/NVIDIA/cccl/pull/1643
Oh I also need to look at `__swallow` :(
> @zygoloid suggests that `using std:...` may be OK: > > > I don't think this would be depending on implementation details. These function templates are required to be present...
It would solve the problem for `device_iterator` It would not solve it for `std::vector` EDIT: it would solve the problem, but we would need to add those to every sequence
pre-commit.ci autofix
I would advice to not spend too much time with `thrust::complex` The only reason it is still there are numerical issues in our `cuda::std::complex` implementation. It will be gone soon...
/ok to test
I adopted this in #1685