-
Notifications
You must be signed in to change notification settings - Fork 37
2022.02.10 Meeting Notes
Philipp Grete edited this page Feb 10, 2022
·
1 revision
- Individual/group updates
- Paper
- Load balancing/CMU collaboration
- Review non-WIP PRs
AJ
- built Phoebus and looking on profiler outputs following Josh's instructions
- load balancing not triggered yet for given input deck
- using tau to collect profiling data
JM
- trying to get Parthenon and Phoebus running on Chicoma
- Code builds, and single node runs work. Working with support for multi-node runs.
BR
- working on Phoebus. Linked Singularity in for tabulated EOS.
Daniel
- updated the paper
FG
- AGN triggering is working in AthenaPK now
- Reviewed PRs
PG
- Got additional scaling numbers on Ookami and Spock
- Issues with one-sided, async MPI from GPU buffers on Summit and Spock
- Paper almost done
- Found bug for filling buffer during LB
- LANL folks will make minor updates
- Send final reminder via mail. Then submit on 21.02.
- discussed parameter input file for test case triggering LB
- will use SERIAL backend with MPI rather than OpenMP multihreading (former combination show best performance from experience)
- will also evaluate influence of pack sizes on performance/load balancing
- PG add changelog to https://github.com/lanl/parthenon/pull/649 then merge
- FG will review https://github.com/lanl/parthenon/pull/646
- PG will update clang to 13 in https://github.com/lanl/parthenon/pull/629 to get rid of the compiler bug (?) with clang12 and cuda