Jun 02 2025 - kokkos/kokkos-comm GitHub Wiki
Attds: Stephen, Carl, Evan, Nicole, Gabriel, Vivek, Matthew
General Topics
-
Recording of KokkosComm talk at HPSFCon 2025 is available on YouTube: https://youtu.be/p0c-T-JDHh0
-
Possibly changing meeting time
- Monday at 10 am MT / 18:00 CET (?)
- Tuesday at 10 am MT / 18:00 CET (?)
-
TN Tech visited SNL two weeks ago
- Nicole & exemplars for demonstrating/testing KokkosComm
- Kokkos-FFT doesn't seem like a good fit (besides, CEA may want to lead this vis-a-vis Gysela)
- Still considering VPIC
- Trying to make sure Tpetra + Kokkos Comm is tractable for the summer
- Trilinos + MPI Advance is a parallel / tangentially related / separate effort not currently related to Kokkos Comm
- Nicole & exemplars for demonstrating/testing KokkosComm
-
CEA (Gabriel)
- CExA engineer (Adrien Taberner) is trying to do distributed matrix-vector product examples (1D/2D partitioning)
- Carl will have a look (possibly waiting on revision from CExA)
- May look into NCCL PR if enough time
- Address some review comments first
- Test compilation locally and write CI pipeline for it
- Then CI at SNL (?)
- CExA engineer (Adrien Taberner) is trying to do distributed matrix-vector product examples (1D/2D partitioning)
-
CI @ Sandia is working
-
In Progress Stuff
- persistent: https://github.com/kokkos/kokkos-comm/pull/156
- Nicole: Some comments from Cedric
- NCCL: https://github.com/kokkos/kokkos-comm/pull/128
- Gabriel: addressing some comments from Cedric
- persistent: https://github.com/kokkos/kokkos-comm/pull/156
Round Table
- Vivek
- LDMSCON last week of June
- technologies for monitoring / analysis / tuning of HPC jobs
- LDMSCON last week of June