Closed Sbozzolo closed 3 months ago
I think this looks fine! I am just wondering if any unit tests need to change to reflect these additions. How are we determining that the behavior doesnt change with this PR?
This is a good question. The PR only moves code around, so it things were previously properly tested, they will stay properly tested. I added unit tests to check that things should be consistent.
One other comment is that I think at this point we dont need to improve the bucket performance on its own, but should make changes that improve the performance of all models (e.g. allocations in
turbulent_fluxes
,net_radiation
, etc)
I agree!
This PR preallocates some of the Fields that are created when computing the bucket tendency.
New time: 3.6s
https://buildkite.com/clima/climaland-benchmark/builds/355#018fef98-5edd-4f39-9cbd-a11a529eba6f