-
Notifications
You must be signed in to change notification settings - Fork 37
2022.12.01 Meeting Notes
Philipp Grete edited this page Dec 15, 2022
·
3 revisions
- Individual/group updates
- Python modules
- "When is a PR too small?"/Changelog/PR handling
- OpenMP parallelism (outer inner) [postponed]
- non-WIP PRs
JD
- tracking down a potential bug, potentially related to flux correction
- no significant progress on face fields, we'll put this on the agenda for early Jan meeting, but PR is effectively ready for review, FG will get a first look
BR
- still working on particle packing and particle IO
LR
- sparse PR fixes got merged in
BP
- signal handling code is back in WIP because of inconsistent signal handling by MPI libs. Will no turn into file trigger
- will also remove SIGALRM (and other signals)
- updating output naming (e.g., more flexible "final" naming)
TG
- Fixed Kokkos warning handling (required by our catch2 tests), which changed in Kokkos 3.7
- Updated second order refinement to match new design
- Will merge now and then we'll update zones to check in a separate PR
- Will look at refinement loop bug next
FG
- Need to answer final comments for https://github.com/parthenon-hpc-lab/parthenon/pull/753 but basically ready to merge
- other group member are waiting for FaceFields/CT in AthenaPK
PM
- finalizing work on AthenaK in prep for moving to LANL
JS
- Now has AMR work with "5D view" in AthenaK -- is happy with performance
- Would be interested in a performance comparison (e.g., hydro Sedov similar to Athena++ paper)
- Also have z4c solver in AthenaK now (for binary BH mergers)
PG
- Will submit abstract to Cray User Group Meeting with John at OLCF
- tracking down bug (JSC/AMD GPU workshop) https://github.com/parthenon-hpc-lab/parthenon/issues/659
- BW will work on adding Ascent
- Reviewing PRs
- Parthenon currently ships two modules as part of the repo
- Q: Should we move them to a separate repo?
- Potential advantages:
- streamline CI
- easier downstream handling for phdf dependent python tools
- Potential disadvantages
- changes in two repo may add developer "overhead"
- keep/maintain backward compatibility
- need new approval for new repo?
- Result:
- BP and JD will look into approval
- BP will work on separate CI concerns (C++ from Python code) and PG will provide pointer to Github Actions for file based triggers
- Can we get a "breaks downstream" label? -> Dev are encouraged to add a label to new PRs to add extra visiility for down stream devs
- Can we separate "trivial" PR from "need more attention" PRs? May be difficult because there may be hidden side effects.
- Will introduce "trivial" branch to collect trivial changes (like typos) from everyone and regularly merge potentially without Changelog for editorial changes (PG will add doc)