Skip to content

CCPP Framework Meeting Minutes 2026 06 11

Jump to bottom
Courtney Peverley edited this page Jun 11, 2026 · 2 revisions

Agenda

Attendees: Michael Kavulich, Dom Heinzeller, Cheryl Craig, Courtney Peverley, Jordan Powers, RhaeSung Kim

GitHub issues, PRs, discussions

PRs

  • None new

Issues

Drafts/On Hold

Capgen project board

New items for discussion

Updates from last time

Previous notes

Other business

Meeting notes

capgen-ng updates from Dom

  • DTC discussion landed on “capgen v2” as preferred name
  • As of today, Dom can reproduce all 200 capgen / CAM-SIMA tests b4b
    • Includes constituent ordering hack in one of the CAM-SIMA tests to verify that the ordering is indeed causing the answer differences
    • Will ultimately want to change things to get qv first (because it’s the largest number)
      • Mike: should this be configurable? Maybe? Perhaps down the line.
    • Also included compatibility layer in CAM-SIMA to get it to work with new capgen-ng interfaces
  • No bug fixes from CAM-SIMA testing needed to be re-fixed for NEPTUNE/UFS
  • Dustin can build UFS caps; can build several suites
    • Has problems with 32bit and fast physics; Dom will look into it today if he has time
  • Capgen calls init routines of all the schemes as many times as the schemes appear in the suite; capgen-ng runs init routines exactly one time (init schemes should be idempotent)
  • Dom: are the CAM-SIMA tests representative of everything?
    • Courtney: yes, with the caveat that we have not tested the WRF physics. We’ll have to do that before transitioning to capgen-ng
    • Cheryl: we’d like to get Haipeng to bring capgen-ng stuff into his dev branch(es) to see if he has problems
      • Known changes: scheme_name length; finalize -> final; module_name metadata field
  • Branch maintenance & transitioning to v2 / ng
    • ccpp-physics has multiple branches we have to maintain
    • Maybe we need to get to a point when all models are ready; then can coordinate when we’re ready to switch
  • Code review
    • Dom: probably no one will be able to do a line-by-line review
    • We do have to do a higher-level review; does it seem extensible to include what we know we’ll need in the future? (ex. gpus)
    • Can start looking at code now (PR #671 in the framework)
  • When PRs come in…
    • Who do we need to tell about the transition, and what documentation can we point them to?
    • Probably need a different system than markdown files that require download and aren't easy to read
  • We will need to document!
    • Can set Claude on it, but will need to de-verbosify it
  • Performance
    • Run-time performance about on par with capgen
    • Code generation performance is faster
      • prebuild: 12s
      • capgen-ng: 10-11s
      • capgen: 230s
      • Dom expects minimal to no difference on memory side for CAM-SIMA, but no data for that currently (based on the way things were allocated before vs now)

Cheryl: Do we have a requirements document since the one from 2018?

  • Dom - Ligia and Steve might know?
    • Mike to ask at DTC meeting to see if Ligia knows
  • How much is overlapping with the technical docs? Unclear.
  • Jordan has some CCPP reference documentation stuff from CCPP review at NCAR - will send to Cheryl
    • May get some of this stuff on the wiki for future reference

We’ll meet in 2 weeks (Mike won’t be there)

CCPP Framework wiki

Clone this wiki locally