-
Notifications
You must be signed in to change notification settings - Fork 64
CCPP Framework Meeting Minutes 2019 05 23
Dom Heinzeller edited this page May 29, 2019
·
3 revisions
Regression testing framework
- hurricane supplemental not yet available
- Steve started add doctests, using pylint, test suites, code coverages, etc. as required for his work to proceed
Finalizing documentation (the big document)
- in transition to sphinx, several chapters already done
- will need to look at final documentation again
SIP meeting, status of CCPP transition to Vlab master
- presentation went well, positive feedback from many
- todo: bigger exposure of CCPP through UFS portal (http://ufscommunity.org)
- next version of GFS v16 for 2021: code freeze soon for final testing, need several physics changes - these will go in first, before the CCPP
- commit queue: https://vlab.ncep.noaa.gov/redmine/projects/comfv3/wiki
- CCPP commit will need to be split into smaller junks
- neutral review from Rusty on FV3 changes (Dom to reply to this after the meeting)
- CCPP doesn't improve physics by itself (but: many bugfixes), but needed to improve physics in the future (chicken and egg problem)
- hybrid CCPP build to be removed before the CCPP transition
- SOW includes removing the dynamic build altogether, this needs to be worked out - timeline?
Impact of delayed CCPP transition on our timeline for switching to the new metadata and cap_gen.
- Steve started working on a Fortran to template metadata file for new physics
- although there is no dependency, converting to new metadata should happen after CCPP transition to master
- public release of FV3 once it is on GitHub - this should include the CCPP, and our SOW includes supporting such a release with CCPP
- When? In the fall 2019?
- Can we do new metadata and cap_gen beforehand (less community training)? Also needs updating the documentation.
- Intermediate step to use new metadata parse (incl. Fortran files) but from ccpp_prebuild.py
- need converter from new structure to CCPP prebuild metadata structure
Make GFS physics more portable:
- remove DDTs from physics schemes
- workaround: create GFS DDTs on the fly - does this make sense? does it work?
- better handling of time-split and process-split physics (via SDF) will help to remove some interstitial schemes
- agreement on standard names across GFS, NCAR, and other physics suites
- Steve has been looking into whether we want to join NUOPC effort (https://docs.google.com/document/d/1XYi8jjo7QKI5kR3hbgNSsGvbAjLnVbAxGfwxyFpVzDg/edit#heading=h.ablr1czekyis)
Unit conversion delivery by June
- small set (even a trivial example) to demonstrate the capability
- use psychrolib and/or udunits
- length mm to m, time s to min, Pa to hPa
Musica kickoff meeting (community chemistry plug-and-play model)
- GeosCAM, ECMWF, folks from Barcelona, NCAR, NOAA
- no presentation on the actual details of the model
- chemistry will have its own community data structures (DDTs) with methods, the framework will be able to handle those (this is something different than the GFS DDTs)