pauleonix

I guess the above `thrust::fill` and similar segfaulting behavior is just past the point where one can expect a perfect implementation. Mixing tagged iterators, execution policies, fancy references and system...

Thanks for the input. I was originally looking into doing a PR myself, but I agree that these API design decisions should be carefully made by the maintainers. Looking forward...

@allisonvacanti Too bad... Maybe I will just still play around with a `BlockLoadToShared` on a fork ;)

After reading #1798 I want to clarify that I would include random number generation as well on the question side, but in terms of priority I think that BLAS (dense...

Together with executors, senders/receivers, mdspan etc these future C++ features would ideally completely replace not only these BLAS wrappers, but also frameworks like alpaka itself as I understand them. That...

With general CUDA questions you will have more luck in the Nvidia Developer Forums or on StackOverflow.

Well, in theory ~~`constant_iterator`~~ `counting_iterator` is just a `T` with some iterator-specific accessor functions and therefore trivially copyable if `T` is. But `std::is_trivially_copy_constructible` still wont agree to that in practice....

I just chose `thrust::const_iterator`, because I thought it were the most basic (now that I read/understood more of the source, [`thrust::counting_iterator`](https://github.com/NVIDIA/cccl/blob/main/thrust/thrust/iterator/counting_iterator.h) is the most basic one made up just of...

https://github.com/NVIDIA/cccl/discussions/562

What I know is that they are [working on ranges](https://github.com/NVIDIA/cccl/discussions/1333#discussioncomment-8348738), so I expect some kind of `iota_view` etc. in the not-too-far future. If this will fix this particular issue or...