2025.07.24 Meeting Notes

Agenda

• Individual/group updates
• Slide upload from dev meeting
• Review non-WIP PRs

Update

JM

• finished docstring PR (merged)
  • https://parthenon-hpc-lab.github.io/parthenon/develop/src/0_input_parameters_ref.html
• also made this work with toml and provenance info (not merged yet)
  • toml MR few todos left -- BP will address

PM

• continuing running Artemis sims
• noticed 6x slowdown on CPUs in non-Cart coords
• now looking into caching geometry infos

FG

• done par_unify analysis
• register pressure changes all over the place
• for specific test case, occupancy change because a single register more was being used
• better direction likely going toward providing launch bounds and/or auto tuning
• PG will provide input deck for some manual testing

PG

• did some testing for "more teams per comm buffer"
• openpmd ready for testing with non-cc vars
  • should get merged before all viz tools are in place
  • PM will test again
  • add python interface: most important features
    • diffs between files
    • movie 2d file
  • PG should check if can run out of mesh indices for very deep mesh
• heffte at 10k^3
  • have some basic power spectra capabilities in Parthenon is probably useful
  • potentially even Poisson
  • upstream with shared infrastructure for calculating wavenumbers
• UCX issue fixed (at their lib level)

Slide upload

• directly to the wiki

non-WIP PR

• #1288 PM will review
• #1255 PG will review
• #1244 PG will review
• #1192 JM will check with Luke what's missing
  • PG will check downside (perf, correctness, potentially comm at scale)
• PG will test toml PR downstream next week
• #1142 see above and then merge

Next meeting tentatively 14 Aug