Skip to content

2021.06.17 Meeting Notes

Jump to bottom
Joshua S Brown edited this page Jun 17, 2021 · 3 revisions

Agenda

  • Individual/group updates
  • Coarse Buffer-amr incompatibility [538] - ran out of time
  • Paper
  • Phil performance figures - ran out of time
  • Review non-WIP PRs - ran out of time

Individual/group updates

LANL/External CS

Joshua Brown

  • Added formatting pr for python
  • Added example and documentation for Py-CGAD
  • Working on several small fixes to make the CI robust
  • Bash uploader pr ready
  • Python formatting pr merged
  • Working to get shared internal dependencies on Grizzly fixed
    • TODO weak scaling tests
    • TODO investigate HDF5 compression performance
  • TODO Setting up training with HDF5 Group
  • TODO add additional performance regression tests. Only running MPI advection right now
  • TODO add different compilers to CI tests

Jonas Lippuner

  • Close to getting Sparse interface PR merged in, will require a number of one line changes in downstream codes.
  • Started working on getting the communication layer for sparse data implemented, requires thinking about how allocation and deallocations will be handled.
  • Will probably start of by ensuring allocations are handled first, once that is done can then look at implementing the deallocation machinery probably will count the number of times that sparse data is not needed on a meshblock until it by passes a reasonable threshold before deallocation occurs.

Tanner Broaddus

  • Has been working on doing some profiling with Caliper on the advection example.
  • Suggestions - look at open pr #534, kokkos profiling tools are used in there to get some ideas about what is already being used.

LANL Physics

Ben Ryan

  • (Absent from call) but looks like he is close to getting particles pr merged thanks to Phil for final review.

Jonah Miller

  • Found a number of minor bugs in the restart files
  • Spent quite a lot of time reading through Jonas pr
  • Working on fixing AMR incompatiblity in pr 538
  • Have gotten parthenon to compile amr architecture

AthenaPK

Philipp Grete

  • Finished reviewing Ben's particle pr will be pulling the plug
  • Option to enable c++17 support pr submitted
  • Updated gnu compiler
  • Wants to add more detailed output targeting individual kernels
  • Generated a bunch of neat scaling plots for presentation
  • Looked at io performance, not concerned about performance, had to limit ranks that were writing.

Forrest Glines

  • Working on downstream code
  • Worked on a pr to help with python analysis tools PR511, the formatting pr through off some of that work

Agenda

Paper

  • After discussing decided to use the link provided by Phil instead of using Forrest's account because he is limited to 6 collaborators.
  • Phil has a nice outline of sections that need to be added.
  • How to handle showing performance in downstream codes?
    • Limit description of downstream codes to essentially one paragraph focus on the types of scaling and performance that can be expected from using parthenon as a backend.
    • Try to generalize the type of problems that could expect to see similar scaling and performance (Make reviewers happy)

TODO's

Everyone - contribute to appropriate sections in paper Josh Brown - look at Forrest pr Jonah - create issue documenting problems with coarse buffer-amr incompatibility Ben - will be addressing issues brought up in particles pr Phil - Galen needs short write up of Pheobus and Athena to cover scope of what physics will be included in paper in order to get an LA-UR to cover bases

Clone this wiki locally