IanBogle

I have done some work on figuring out the ambiguous overload error, and I think it's a problem in the ParallelScanHIPBase hierarchy (in https://github.com/kokkos/kokkos/blob/develop/core/src/HIP/Kokkos_HIP_Parallel_Range.hpp#L659). I've come up with a piece...

I encountered a team using roctx-rename and ROCP_RENAME_KERNEL to rename kernels in traces using roctx-region. This could serve as a workaround, except it only changes the .json output, not the...

Hi @Pivamat, I'm an AMD engineer that works on LAMMPS, and I'm looking into this performance issue. I've run a few benchmarks that use PPPM against various ROCm versions and...

@akohlmey Greatly appreciate the insight, I will take this advice. I'll admit my cursory testing thus far was mostly to make sure there wasn't some very noticeable regression in ROCm...

Understood, and I appreciate your feedback here. We're similarly resource constrained, so I appreciate your willingness to step in and save my time. If nothing else, you've outlined how I...

I'll run this test when I can, I'll be out of office next week. Really appreciate your guidance here!