This PR renames CCPP suites to non-meaningful bird names, as suggested by various colleagues and ChatGPT. The old names are mapped to the new names via a file "alias.json" for convenience.
Some older, unsupported suites, especially those with no entry in suite_info.py, are moved to a new directory suites_not_used/, analogous to the ufs-weather-model treatment of these suites.
Requires a PR to CCPP framework to remove the ccppprebuild requirement that all suite files start with the prefix "`suite`". This Framework update requires some changes to the CI tests to use MPI, which is required in the latest version.
Ran full suite of regression tests, all succeeded. Running manual tests on MacOS showed a few suites that are failing with segmentation faults, but this seems to be intermittent and must be related to my specific machine. One suite (SCM_GFS_v16_RRTMGP "mallard") failed all tests due to memory allocation; this appears to be a pre-existing problem, see #484.
Did a bit-for-bit comparison for all the regtest cases (arm_sgp_summer_1997_A, astex, bomex, LASSO_2016051812, twpice) between this branch and the branch for PR #482 (which updated framework and physics to same hash as this PR); aside from the failing tests the only differences were for suite "robin" (previously SCM_GFS_v17_HR3_RRTMGP), and this is expected due to a changed tracers file.
Description:
This PR renames CCPP suites to non-meaningful bird names, as suggested by various colleagues and ChatGPT. The old names are mapped to the new names via a file "alias.json" for convenience.
Some older, unsupported suites, especially those with no entry in suite_info.py, are moved to a new directory
suites_not_used/
, analogous to the ufs-weather-model treatment of these suites.Dependencies:
Framework PR: https://github.com/NCAR/ccpp-framework/pull/569
Physics PR: https://github.com/NCAR/ccpp-physics/pull/1073
New guidelines are proposed in CCPP Technical documentation PR here: https://github.com/NCAR/ccpp-doc/pull/72
Tests:
Ran full suite of regression tests, all succeeded. Running manual tests on MacOS showed a few suites that are failing with segmentation faults, but this seems to be intermittent and must be related to my specific machine. One suite (SCM_GFS_v16_RRTMGP "mallard") failed all tests due to memory allocation; this appears to be a pre-existing problem, see #484.
Did a bit-for-bit comparison for all the regtest cases (arm_sgp_summer_1997_A, astex, bomex, LASSO_2016051812, twpice) between this branch and the branch for PR #482 (which updated framework and physics to same hash as this PR); aside from the failing tests the only differences were for suite "robin" (previously SCM_GFS_v17_HR3_RRTMGP), and this is expected due to a changed tracers file.
Issue:
This resolves #476