Daniel Arndt

Supersedes #724 (basically adding to it).

Use KOKKOS_CLASS_LAMBDA

5

Requiring C++17, we can now resolve the comments mentioning "capture *this". Also, this shows that we forgot to update the GitLab CI. I would propose to always just build `Kokkos`...

Add DistributedTree query only taking a callback

3

This adds a ``` template void query(ExecutionSpace const& space, Predicates const& predicates, Callback const& callback) const; ``` overload for `DistributedTree` (and `SpatialTag) with the callback being executed on the MPI...

Cleanup KOKKOS_CONFIGURE_CORE

1

While verifying https://github.com/kokkos/kokkos/pull/6823/files#r1492824399, I noticed that we have a couple of unused variables and include files. In particular, - KOKKOS_MEMSPACE_LIST - KOKKOS_BACKEND_POST_INCLUDE_LIST never contain anything and we can thus remove...

Fixes the warning caused by using Node 16 instead of Node 20, see https://github.blog/changelog/2023-09-22-github-actions-transitioning-from-node-16-to-node-20/.

Clean up shift_right_team_impl

3

I was running SonarCloud on my fork to try it out and one of the bugs reported was https://sonarcloud.io/project/issues?resolved=false&types=BUG&id=masterleinad_kokkos&open=AY2p6YQlbq2DzqS7mbTl. We basically use a unary minus operator on an unsigned type...

multi-GPU support: Add test for all policies

2

This should be the last step for #6091. In particular, this pull request extends the test in `TestCuda_interOp_StreamsMultiGPU.cpp` to test all policies by effectively copying the test in `TestCuda_InterOp_Streams.cpp` and...

Make team-level parallel_scan interface consistent with parallel_reduce

1

Related to #6317. https://github.com/kokkos/kokkos/issues/6760 says: > Currently, the parallel_scan version for TeamThreadRange and ThreadVectorRange (btw, there is no TeamVectorRange version, but that's for another issue) and is right with this...

This is the changelog for the upcoming Release 4.3 ### Features: ### Backend and Architecture Enhancements: #### CUDA: * Removed `KOKKOS_IMPL_CUDA_CLANG_WORKAROUND` macro (not needed any more) [\#6576](https://github.com/kokkos/kokkos/pull/6576) * multi-GPU support...

SYCL: Cleanup device selection

1

Follow-up to #6758 incorporating https://github.com/kokkos/kokkos/pull/6795#issuecomment-1936514087 after #6790. With this pull request, `get_sycl_devices()` returns all devices if no GPU architecture is set. This makes it possible to clean up `SYCL::impl_initialize` with...