Andrey Prokopenko
Andrey Prokopenko
> ```c++ > template class... KeyProperties, class... Values> > void sort_by_key(const ExecutionSpace& exec, > const Kokkos::View& keys, > const ComparatorType& comparator, > const Values&... values); > ``` I would argue...
> I think you also need to inverse the permutation vector? You don't. This is the permutation vector you want. We do this in ArborX. You may want to look...
I assigned it to myself. I'm going to implement a basic version in the next couple weeks.
@sethrj Do you have a simple ioc example to test with Intel (outside of ForTrilinos)?
The easiest way to test this would be with MPACT, so lets wait till #161 is done.
See #218.
This would be really nice. Looking at [this](https://github.com/kokkos/kokkos/issues/5285#issue-1320161352), Kokkos supports ROCm 4.5 for 3.7. Should this PR be delayed until we require Kokkos 4.0 (assuming the performance checks out)? >...
How did you realize it wasn't necessary? Were there any compiler release notes, or did you just try it out?
> I don't have the input data to run DBSCAN (I keep losing it), so @aprokop should probably check that it gets the same results. I quickly tried to do...
## DBSCAN on Crusher, ROCm 5.2.0 ``` $ for i in 1 2 3 4 5; do ./ArborX_master_f8b9a225 --binary --filename ~/csc333_orion/hacc_497M.arborx \ --verbose --core-min-size 2 --eps 0.014 --impl fdbscan |...