CCPP Framework Meeting Minutes 2021 01 12

Jump to bottom

Dom Heinzeller edited this page Jan 12, 2021 · 6 revisions

Agenda

capgen.py progress
Transition to capgen.py
Other business

`capgen.py` progress

Need plan yet how to handle optional arguments (Dom to reach out to developers - EMC, GSL, Robert & Dustin, Phil, NRL)
DTC needs to develop a mechanism to avoid using the field_table - use an internal field table approach, similar to the internal input file?

Transition to `capgen.py`

Updated DTC timeline and resource chart (https://docs.google.com/spreadsheets/d/1XbSF1ejYcPMrVtZtk1OSiMX-vI5srLkwEUhyAirGr8k/edit?usp=sharing)
Things that need to happen for transitioning UFS to capgen.py (https://docs.google.com/document/d/1M3mCO3ESR9OJqLyUbQf-ahJtmfmGKilZlWaBTxsfMRk/edit?usp=sharing)
Dictionary of standard names:
- https://meet.google.com/linkredirect?authuser=0&dest=https%3A%2F%2Fdocs.google.com%2Fspreadsheets%2Fd%2F1kWDC7_2sagm-PTP_tXaGyI3O3sJFK8FY9FElQbb6V3o%2Fedit%23gid%3D1778121391
timestep_init and timestep_final implementation were merged into the repository last week

Other business

CCPP governance
- Last meeting was on Jan 8, 2021
- A lot of time was spent on discussing how to test code changes, impact on operational versus research code, ...
- Implementing these solutions will be a project, not just done in one afternoon
- Physics interoperability / shared code in separate submodule
- But see Fanglin's comment on code ownership: operational code needs to be surrendered
- One solution is to have operations maintains their own fork with the necessary modifications based on a tag from the authoritative repository
- An example where it works well is RRTMGP, but we expect it to be more difficult for Thompson MP
- Other candidates: NoahMP, MYNN, ...
- How far does CCPP governance (coding standards, ...) go? Just to the CCPP entry points or further down?
  - But models have their own standards and requirements
  - External contributions to the code
  - Need code owners for physics sitting in ccpp-physics repository
  - Need sufficient test coverage and agreement who creates, maintains and runs the tests
  - Can we define what a suitable test set looks like?
  - Statistics (e.g. ensemble consistence tests) can help with accepting answer changes, but do not solve the problem