Vivek Kale

> > You'll have to fix the indentation. > > Thanks! Applying clang-format today. Done.

Yes, I was thinking to say the same as @masterleinad from looking at your error more closely yesterday. One other question, which I assume you have done based on what...

> Here is my reproduced case that lead to error. I attached the source code and CMake file. > > [mpitest.zip](https://github.com/user-attachments/files/15539740/mpitest.zip) > > command line for cmake > > `SCOREP_WRAPPER=off...

> > > Error: Unable to load KokkosP library: /opt/scorep/lib/libscorep_adapter_kokkos_event.so > > > dlopen(/opt/scorep/lib/libscorep_adapter_kokkos_event.so, RTLD_NOW | RTLD_GLOBAL) failed with /opt/scorep/lib/libscorep_adapter_kokkos_event.so: undefined symbol: scorep_kokkos_subsystem_id > > > > > > This...

I am noting a corresponding GitHub issue in Kokkos Tools is being opened from the standpoint of documentation of the use of ScoreP using Kokkos Tools. It is here: https://github.com/kokkos/kokkos-tools/issues/255...

> @wrwilliams Thanks for the quick reply. I re-built score-P with the following configure line: ../configure --enable-shared --with-mpi=openmpi --enable-debug > > also attached the config.log [config.log](https://github.com/user-attachments/files/15547125/config.log) > > Kokkos: 4.3.x...

> > I am noting a corresponding GitHub issue in Kokkos Tools is being opened from the standpoint of documentation of the use of ScoreP using Kokkos Tools. It is...

> I have tried to reproduce this locally, both with Score-P 8.3 and our 9.0 prerelease version, and have not been able to generate this crash. OpenMPI is 4.1.6, Kokkos...

> Kokkos Serial is not spawning threads new threads as far as I know and can tell from the code. > > ``` > start of Kokkos parallel_reduce > waiting...

> @vlkale BTW, a quick and dirty fix for this is to change this line: > > https://github.com/kokkos/kokkos/blob/2d7715239700f50169bc50a96a234b05c28c9a2e/core/src/Cuda/Kokkos_Cuda_Parallel_MDRange.hpp#L114 > > to: > ``` > const dim3 block(m_rp.m_tile[0], m_rp.m_tile[1], > ((m_rp.m_tile[2]...