-
Notifications
You must be signed in to change notification settings - Fork 37
2022.01.13 Meeting Notes
Philipp Grete edited this page Jan 13, 2022
·
1 revision
- Individual/group updates
- Paper
- Review non-WIP PRs
JM:
- Worked on Singularity
- Will trying getting codes running on chekoma
- working with BR and LR on MC implementation
JS:
- working with PM on getting RadHydro working and initial scaling tests with AthenaK
FG:
- Working on issues related to tabular cooling.
- Trying to get AGN picture in Parthenon paper
JD:
- LR will take on sparse work to improve comm pattern.
- We should still merge the sparse PR now and then work on the fix later.
PG:
- parthenon-hydro works and used for scaling
- New PR on extending build testing, e.g., HIP/clang/ROCm
- Ran scaling tests on Summit and JUWELS Booster. Former were initially not great but that was a hardware issue (NIC firmware). Weak scaling on JUWELS booster is >97 up to 128 nodes/512 GPUs
- Filled gaps in the body of the paper
- Body almost done. PG will fill in abstract, intro and conclusions by early next week. Afterwards everyone should review/edit the text.
- Need to add paragraph on reductions.
- JM will try to get numbers on Intel GPUs on Darwin so that we can demonstrate performance portability
- Scaling data is still "WIP". Larger scaling tests waiting on queue. For strong AMR scaling will use increased base size as the original problem in the Athena++ paper now fits on a single GPU when using uniform grids.
- Author list suggestion: PG first, then alphabetical order
- PG will send first complete draft to JM/JD for LANL approval
- PG will review generalized reductions as they are: https://github.com/lanl/parthenon/pull/622
- PG will re-review particles sort https://github.com/lanl/parthenon/pull/619
- JM will review and extend build fixes for Intel https://github.com/lanl/parthenon/pull/629