kokkos-kernels: constexpr declaration of OrdinalTraits::invalid()

[maartenarnst]

Enhancement

@trilinos/kokkos-kernels

What feature/addition is needed? How does this differ from what we currently have?

Templated on the ordinal type, the traits class KokkosSparse::OrdinalTraits provides a member function invalid() that allows to retrieve a specific value of this ordinal type that is considered "invalid". This invalid value is used in various ways in kokkos-kernels, tpetra, and other packages, e.g. in map objects as a return value when the ordinal is not included in the map. In kokkos-kernels, this member function is currently declared static KOKKOS_INLINE_FUNCTION.

If this "invalid" value is used multiple times throughout a code, it can be convenient to store it in a constant variable at namespace scope. However, if the "invalid" value is used in device code, storing it in a constant variable at namespace scope is subject to some restrictions, as described for the cuda case in Sec. I.4.13 "Const-qualiffied variables" in the cuda manual (https://docs.nvidia.com/cuda/cuda-c-programming-guide/index.html). In particular, this reference suggest that cuda wants such a constant variable at namespace scope to be initialised by a constant expression.

In this context, this feature request is whether you would be open to declare the member function invalid() as a static constexpr KOKKOS_INLINE_FUNCTION. If you should be open to do this, I can submit a PR.

Many thanks in advance for considering this small feature request!

[jhux2]

@trilinos/kokkos-kernels

[github-actions]

This issue has had no activity for 365 days and is marked for closure. It will be closed after an additional 30 days of inactivity.
If you would like to keep this issue open please add a comment and/or remove the MARKED_FOR_CLOSURE label.
If this issue should be kept open even with no activity beyond the time limits you can add the label DO_NOT_AUTOCLOSE.
If it is ok for this issue to be closed, feel free to go ahead and close it. Please do not add any comments or change any labels or otherwise touch this issue unless your intention is to reset the inactivity counter for an additional year.

[maartenarnst]

Closing as solved in pr kokkos/kokkos-kernels#1976.