...
- need someone on the land side to help with aerosol deposition on snow and subsequent impacts
- bugfix on snow grain size
- find a way to state consequences of atmosphere side to land into the "requirements document" use "code review process", "request hub"
- when are the BC runs going to start.
- there might be better CLM configurations for the atmosphere model to tune with rather than the satellite phenology. Land can provide IC good enough, but not the same as the BGC, better than satellite.
- agreement to deputize a small group to look at verification datasets of mutual interest to Land/Atmosphere
Atmospere-SE speed dating
Note taken by Po-Lun Ma (not proof-read yet)
Rasch: PR done. Tests for new features.
Blueshirt: Run it and confirm it is expected, regression test.
Rasch: (1) Science verification. (2) B4B.
Burrows: Tests suite works for namelist options?
SE: Yes.
Chair: will provide an example.
Rasch: switch new features to on as default.
Blueshirt: we don’t know climate. Code review, council, group decides science.
Rasch: OK.
Taylor: regression test will fail when new features on as default.
Chair: redefine the compset.
Blueshirt: F is atmosphere? This can be changed.
Yoon: CLUBB comes with many bug fixes that are turned off.
Chair: bugfix induced climate change is your job/decision.
Taylor: scientists decide climate. Tests are just 5-day simulations.
Bluesfuirt: tests are for to preserve and track changes.
Burrows: Why 5 days?
Chair: there are some tests that run for a few timesteps.
Rasch: long enough for coupler to exchange info.
Burrows: climate-changing features. All on at once or incrementally?
Chair: science question. Not practical to track every change.
Golaz: coordinate on B4B and climate-change features.
XB: coupling frequency atmos-land, atmos-ocean, atmos-ice
Blueshirt: up to the coupled team.
Chair: Per day or per 6 hours. Atmos-land 30-min. Not our call.
Bader: Coupled team is responsible. Science / computational consideration. It may change with new ocean model. Physically it doesn’t seem to be important.
XB: BGC, diurnal cycle, etc., daily coupling enough?
Bader: diurnal cycle in atmos. Not sure how this info is used in ocean/ice.
Rasch: MPAS doesn’t need it?
Chair: Science groups need to figure out what frequency is needed, then we will make that happen.
Taylor: 1-hour coupling atmos-ocean.
Chair: okay!
Blueshirt: Integrators
Burrows: tests are not started.
Singh: I need to learn how to do tests. PRs are manageable.
Chair named list of tests. List of compsets.
Rasch: how you change the test procedure from …?
Chair: coupler history for B4B, developer and integration tests.
Rasch: production platform?
Chair: NERSC (and titan?) and mira next.
Blueshirt: Developer’s test is not integrator’s or SE’s job.
Burrows: We need info on how to do tests.
Chair: e.g., restart test. No one has done it yet.
Rasch: Any established reasonable tests?
XB: Coupler tests.
Blackshirt: We do SE tests. Beyond?
Blueshirt: B4B, noise vs climate-change. From 5-day to 5-year to see. That’s what we are at now.
Koch: when to test the full model?
Chair: When V1 freezes.
Blueshirt: what’s done (developer’s test) is to help science productivity.
Koch: next few months?
Blackshirt: test for long-term simulations science question
Koch: should be both
Chair: Advice: incremental update/test. Do test even with a new data structure that does nothing
Evans: Option to mark to “not use this” in code?
Chair: sure. A comment will do.
Rasch: test everything (even with data structure) to make sure codes don’t break.
XB: coarse res coupling
Blueshirt: That’s not SE job. Coupled team’s responsibility.
Caldwell: Short-lived branch. A latest branch doesn’t resubmit.
Chair: Currently no test for scripting.
Caldwell: top priority to make sure tests done for different platform. It even should not be on master.
Blueshirt: sometimes machine upgrads can affect.
Blueshirt is another co-lead.
...
- Discuss testing: need for new tests during the code review process
- brief discussion on making new features default.
- SE team will provide template scripts for implementing new tests,
- discussed difficulties with MIRA
- strategies for making commits to master
- need for testing of resubmit capability