UFS-dev PR#258

[grantfirl]

SOURCE: @grantfirl

DESCRIPTION OF CHANGES:

• updated all SDFs to remove interstitial "reset" schemes, introduce new phys_ts group for time-split schemes and phys_ps group for process-split schemes
• remove reset schemes from ccpp_prebuild_config.py
• change ccpp_physics_run calls in scm.F90 and scm_time_integration.F90 to call the interstitial%X_reset subroutines before radiation and phys_ps groups; use separate calls to ccpp_physics_run for radiation, phys_ps, and phys_ts groups

ISSUE: None

ASSOCIATED PRs:

• UFS-dev PR#258 ccpp-physics#1127

TESTS CONDUCTED: CI RTs

This PR catches the NCAR:main branch up with changes from the ufs-community:ufs/dev branch.

Associated ufs/dev PR:

• Remove GFS_suite_interstitial_{phys,rad}_reset.* (reset in fv3atm CCPP_driver.F90 ufs-community/ccpp-physics#258

Associated fv3atm PR:

• Convert GFS DDTs from blocked data structures to contiguous arrays: Part Deux + CCPP updates in fv3atm/ccpp-physics: split physics in two groups, reset GFS_interstitial DDT in CCPP_driver.F90 #936 NOAA-EMC/ufsatm#932

Associated NCAR PR:

• UFS-dev PR#258 ccpp-physics#1127

---

REGRESSION TEST CHANGES: None

[dustinswales]

@grantfirl These changes look good to me. However, the results do not agree with the baselines.
Maybe they aren't up-to-date? If that's the case, then these baselines need to loaded to Mohawk, https://github.com/NCAR/ccpp-scm/actions/runs/14316552427, and we can rerun the tests @hertneky @scrasmussen @mkavulich
If they are up-to-date, then something must be off.

[grantfirl]

@grantfirl These changes look good to me. However, the results do not agree with the baselines. Maybe they aren't up-to-date? If that's the case, then these baselines need to loaded to Mohawk, https://github.com/NCAR/ccpp-scm/actions/runs/14316552427, and we can rerun the tests @hertneky @scrasmussen @mkavulich If they are up-to-date, then something must be off.

Yes, I was just checking that. The baselines are out-of-date. @hertneky @scrasmussen @mkavulich Could one of you upload the baselines from https://github.com/NCAR/ccpp-scm/actions/runs/14176036114?

[grantfirl]

@grantfirl These changes look good to me. However, the results do not agree with the baselines. Maybe they aren't up-to-date? If that's the case, then these baselines need to loaded to Mohawk, https://github.com/NCAR/ccpp-scm/actions/runs/14316552427, and we can rerun the tests @hertneky @scrasmussen @mkavulich If they are up-to-date, then something must be off.

Yes, I was just checking that. The baselines are out-of-date. @hertneky @scrasmussen @mkavulich Could one of you upload the baselines from https://github.com/NCAR/ccpp-scm/actions/runs/14176036114?

The artifacts from https://github.com/NCAR/ccpp-scm/actions/runs/14316552427 and https://github.com/NCAR/ccpp-scm/actions/runs/14176036114 should be identical. 14176036114 is where the changes occurred though.

[hertneky]

Changes look good to me.

I refreshed the baselines on mowhak and reran the RTs - looks good with no differences like before.

Thanks for taking this one @grantfirl as I would have needed help from you anyways for applying the changes from the fv3atm CCPP_driver.F90. I am guessing the same changes (adding group) are not needed for the ccpp_physics_timestep_init/finalize calls - likely not group specific for SCM.

[grantfirl]

Changes look good to me.

I refreshed the baselines on mowhak and reran the RTs - looks good with no differences like before.

Thanks for taking this one @grantfirl as I would have needed help from you anyways for applying the changes from the fv3atm CCPP_driver.F90. I am guessing the same changes (adding group) are not needed for the ccpp_physics_timestep_init/finalize calls - likely not group specific for SCM.

Thanks for the baselines refresh. The only reason that the groups need to be called individually in the run phase is that interstitial_X_reset schemes that were removed only had run phases for resetting interstitial variables. The other phases can continue to be called for the entire suite because nothing needs to be done between groups for those phases for the SCM.