2021.10.28 Meeting Notes

Agenda

• Individual/group updates
• Squashing commits
• Restart versus hdf outputs
• C++17
• Paper
• Review non-WIP PRs

Individual/group updates

LANL/External CS

Jonas Lippuner

• Took care of sparse PR comments.
• PR is ready to go
• Found bug with new signaling
• Leave still start next Monday: Who's going to take over LANL CI pipeline (token, triggering)? -> Jonah

Galen Shipman

• Working on getting more sparse coverage from LANL

Daniel Holladay

• Reviewed sparse and building Parthenon
• Potentially autotuning, TeamPolicy

LANL Physics

Jonah Miller

• Focused on downstream development
• Has sth working in spherical coordinates with fixed SpaceTime
• Will use for numbers in paper

Benjamin Ryan

• Focused on downstream development

AthenaPK

Philipp Grete

• Signaling
• HDF5 output tuning
• First benchmarking of AthenaPK
  • pack size is important
  • A100 are nice

Forrest Glines

• Downstream development
• Still runs into segfaults with deep hierarchies when writing HDF5 output
• Simulations don't restart correctly. "Missing sparse info" in restart file.

Squashing commits

• Decision: Squash merges
• TODO(PG): edit commit skeleton and doc to reflect this change

Restart versus hdf outputs

• Current differences:
  • variables that go into the files
  • Xdmf written or not
  • Forced double prec. versus chosen prec.
• Decision-> streamline to single format with some trigger for restart (e.g. reserved variable name)

C++17

• XLclang should now support C++17
• Galen will follow up with LLNL to see if this works/is available on Sierra
• Jonah will also follow up with Josh
• TODO(pgrete): Create an issue what needs to be done before we can move to C++17, e.g.,
  • builds on El Cap precursor, Sierra, Summit, Frontier, Nvidia devkits, AA64

Paper

• Goal: Have some prelim. plot (not at scale) in two weeks to discuss overall state of what we want to show and how.

Review non-WIP PRs