Skip to content

2021.12.02 Meeting Notes

Jump to bottom
Philipp Grete edited this page Dec 2, 2021 · 2 revisions

Agenda

  • Individual/group updates
  • Paper
  • Kokkos dependency
  • Sparse update
  • Review non-WIP PRs

Individual/group updates

LANL/External CS

GS: new job ad out for scientist 2/3/4

LANL Physics

JM:

  • PR merged for unified coords object (between MeshBlockData and MeshData
  • Got fairly large sim (GRMHD with static spacetime) running on Darwin in prep for paper. Will run on Sierra next.
  • May potentially include self-gravity (1D gravity solver on CPU)

JD:

  • Need to demonstrate block structured performance soon. Sparse is crucial to reduce memory footprint.
  • Working on downstream code (turning Riot to latest parthenon develop incl MeshData and sparse)
  • Identified new requirements for sparse: will collect info in https://github.com/lanl/parthenon/issues/607 , e.g.,
    • separating is_allocated in MeshBlockPack
    • always allocating buffers (future)
  • Talk about "Parthenon Enhancement Proposal" structure for future major changes
  • Need CG solver working in <few months

BR:

  • Fixed downstream issues.
  • Will start working on particle infrastructure again (e.g., need to be able to work on all particles per cell)
  • Need little more time to get rad transport test case in paper

AthenaPK

PG:

  • Attended and presented Parthenon/AthenaPK at the IAU Symposium 362. Created an updated movie (https://pgrete.de/dl/IAUS-362_Grete_AthenaPK-AMR.mp4). Personal impression: well received and people are interested
  • Trying to track down bug in sparse PR, where some components of a vector are not properly initialized/contain garbage. Created a reproducer (see comment in sparse PR). (done as of this meeting)

JS:

  • Rad. transport with geodesic meshes in test code about to enter production

Paper

  • PG: For the scaling data in the paper, I am working on using parthenon-hydro to replicate the hydro blast wave test in the Athena++ method paper, and, similarly, the MHD blast wave test (including strong scaling) for AthenaPK with AMR. Similarly, will use MHD linear wave test for weak scaling and add "selected" data points for mesh refinement performance, similar to current Fig. 3. Given the increased interest in this overall topic, other projects also preparing publications, and soon to be submitted downstream code and application papers, PG will make this paper his top priority in Dec.

  • Use Parthenon hydro for performance and scaling. Add code description and pretty pictures for other codes.

  • Timeframe

    • Add additional content until next call in two weeks
    • "feature/text"-freeze end of the year
    • Target submission mid January

  • Josh will determine by next week on state of sparse

Kokkos dependency

See https://github.com/lanl/parthenon/issues/615 -- currently not a priority

Review non-WIP PRs

Clone this wiki locally