(*)Use conversion factor for masscello diagnostic#335
Conversation
Use a conversion factor to rescale the units of masscello, just like every other diagnostic. This does not change the diagnostic itself, but it changes the order of the rescaling and the vertical remapping of this diagnostic onto other coordinates (like z) or spatial averaging of this diagnostic, which can change values in the last bits for this diagnostic for Boussinesq models (but not for non-Boussinesq models, for which the conversion factor is an integer power of 2). As a result some of the diagnostics derived from masscello can differ and this commit nominally fails the TC testing for reproducibility across code versions. All solutions and primary diagnostics, however, are bitwise identical, and even the derived diagnostic calculations are mathematically equivalent.
|
This commit is expected to fail the TC tests for exact diagnostic consistency across model versions. As a result, this should not be merged into dev/gfdl until the pending voluminous PR from dev/gfdl to MOM-ocean/main has been created. |
Codecov Report
@@ Coverage Diff @@
## dev/gfdl #335 +/- ##
============================================
- Coverage 37.07% 37.07% -0.01%
============================================
Files 264 264
Lines 74353 74351 -2
Branches 13788 13788
============================================
- Hits 27569 27563 -6
- Misses 41686 41690 +4
Partials 5098 5098
... and 1 file with indirect coverage changes 📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more |
marshallward
left a comment
marshallward
left a comment
There was a problem hiding this comment.
Regression makes sense here.
|
Gaea regression: https://gitlab.gfdl.noaa.gov/ogrp/MOM6/-/pipelines/18867 ✔️ |
Update CVMix to latest commit
2 participants
Use a conversion factor to rescale the units of masscello, just like every other diagnostic. This does not change the diagnostic itself, but it changes the order of the rescaling and the vertical remapping of this diagnostic onto other coordinates (like z) or spatial averaging of this diagnostic, which can change values in the last bits for this diagnostic for Boussinesq models (but not for non-Boussinesq models, for which the conversion factor is an integer power of 2). As a result some of the diagnostics derived from masscello can differ and this commit nominally fails the TC testing for reproducibility across code versions. All solutions and primary diagnostics, however, are bitwise identical, and even the derived diagnostic calculations are mathematically equivalent.