Daniel Arndt
> Thanks. Is there a roadmap page somewhere ? We normally only do release notes (https://github.com/kokkos/kokkos/issues/5439, https://github.com/kokkos/kokkos/blob/master/CHANGELOG.md).
Maybe, we could block the first stream with https://docs.nvidia.com/cuda/cuda-runtime-api/group__CUDART__EXECUTION.html#group__CUDART__EXECUTION_1g05841eaa5f90f27124241baafb3e856f until after the second one is fenced.
@janciesko I took the freedom to simplify the containers test some (with the same functionality and adding a test for rank-8 views).
We should just not forget to add tests when this is fixed with `mdspan`.
> Does not build on the GPU backends Should work now.
> @masterleinad -- would you please re-test? Damien's comment was for the CI to run again.
> OK, my concern was the failing test (perhaps part of the CI issues mentioned recently?) No, the CI was last run on this a couple of months ago but...
`GCC-8.4.0` is timing out, everything else passed.
Let's close this until we have a better idea for a solution that works on Intel GPUs and gives better results than the workgroup size chosen by the compiler.