Refactoring deal.ii with Kokkos

[Elmira]

Hello,

I studied the source code, it seems there is not any refactoring of deal.ii with Kokkos? Google search suggests only to integrate Kokkos program with deal.ii kernels. 

As I study the portability of deal.ii in various heterogeneous hardware architectures, would you please shortly let me know is there any plan to refactor the software? if yes, which kernels, algebra or mesh adaption?

Thanks a lot in advance!

Elmira

Dr. Elmira Birang

Scientist @ Bavarian Academy of Sciences and Humanities

[Daniel Arndt]

Elmira,

Kokkos is at the moment only used directly for the Portable MatrixFree
implementation and the LinearAlgebra::distributed::Vector (with
MemorySpace=MemorySpace::Default). However, all of Trilinos relies
implicitly on Kokkos and the Tpetra classes, in particular, can take
advantage of all the Kokkos backends. The TpetraWrapper classes in
deal.II are still not quite complete as compared to the
TrilinosWrapper classes (using Epetra) but should be largely usable,
also see Section 2.3 in "The deal.II library, Version 9.6", DOI
10.1515/jnma-2024-0137. Are there any particular functionalities you
would like to see ported?

Best,
Daniel

> --
> The deal.II project is located at http://www.dealii.org/
> For mailing list/forum options, see https://groups.google.com/d/forum/dealii?hl=en
> ---
> You received this message because you are subscribed to the Google Groups "deal.II User Group" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to dealii+un...@googlegroups.com.
> To view this discussion visit https://groups.google.com/d/msgid/dealii/ea1444a0-f6d9-4bf5-8c82-5bd8dd27d052n%40googlegroups.com.

[Elmira]

Dear Daniel,

Thanks a lot for detailed response, studied the paper. I work on offloading of adaptive finite element pipeline in modern hardware AMD APU and Intel CPU-GPU.

Best regards,

Elmira