search

CliMA / KinematicDriver.jl

Prescribed flow models for testing cloud microphysics schemes.
Apache License 2.0
8 stars 2 forks source link

KiD refactor #157

Closed trontrytel closed 6 months ago

trontrytel commented 6 months ago

This PR tries to make KiD closer to ClimaAtmos. That means:

I tried to simplify some of the rhs computations. I'm not fully settled on the state and aux structure, yet. For example whether 2M scheme should carry q_ice and q_sno in the Y state.

Anyway, I think the tests should be passing now and the examples should be running and producing some results. I think there were some examples where the results are different, so I still need to work on the PR. But I think this is a good time to start getting some feedback about it. So I'll be very grateful if you could take a look: @sajjadazimi , @nefrathenrici @edejong-caltech

trontrytel commented 6 months ago

The tests are passing, but there is a couple of cases in the CI that have different results: Box model, 1D collision with sedimentation, 1M with snow and ice and SB2006. I'll look into it. Hopefully this is the last stretch

nefrathenrici commented 6 months ago

Sorry I didn't get to review this PR, all of the changes are clear! Have you checked if this PR improves performance significantly?

trontrytel commented 6 months ago

I didn't see anything significant. Some simulations take longer now, others are slightly shorter. But most of the KiD simulations we have in the CI take ~2-3 minutes to run, so maybe it's hard to see anything?