Damien L-G
Damien L-G
> What about moving these two structs into `UnorderedMap` and removing `UnorderedMapInvalidIndex`? Users would then use: > > * `Kokkos::Unorderedmap::invalid_index` > * `Kokkos::UnorderedMap::InsertResult` > * `Kokkos::UnorderedMap::InsertOpTypes` > > I don't...
> 1. There are 2 different invalid index variables, one in the `Kokkos` namespace named `UnorderedMapInvalidIndex` and one defined in the unordered map class (`invalid_index`) and used internally. I bet...
Would you be willing to open a PR (against develop) to fix this?
> It fixes that error but there are others: > > ``` > /var/jenkins/workspace/Kokkos/install/include/Kokkos_MathematicalSpecialFunctions.hpp(1328): error: identifier "Kokkos::numbers::pi_v " is undefined in device code > /var/jenkins/workspace/Kokkos/install/include/Kokkos_Complex.hpp(645): error: identifier "Kokkos::numbers::pi_v " is...
@ingo-h did you get a chance to get back to this?
> I think things are working better now. Can you elaborate? Do you mean issue is resolved in `main`?
As far as `View` runtime versus compile-time unmanaged is concerned, we can certainly improve the documentation but changing things is rather impractical. Now regarding the return type of `Impl::as_view_of_rank_n`, AFAICT...
@aprokop plz review
Ignoring the `CUDA-11.6-NVCC-DEBUG` that is running really slowly.
> Should this have a bug fix entry to the 4.4 changelog? Looks like a fixes an issue from 4.3 (or earlier)? It's not a bug fix, it improves the...