add GPU AMIP runs with albedo from file #657
Conversation
juliasloan25
force-pushed
the
js/gpu-amip-albedo
branch
3 times, most recently
from
d65b6f3 to
8b91924
Compare
juliasloan25
force-pushed
the
js/gpu-amip-albedo
branch
2 times, most recently
from
c9250dd to
a3621c7
Compare
There was a problem hiding this comment.
Good idea to bring these in, thank you, @juliasloan25 . I just had some questions on why the runs for the different albedo types have different resolutions (unless they're breaking, we should make these consistent, and I would choose the lowest possible resolution, except the target run).
| dt_save_to_sol: "1days" | ||
| dz_bottom: 30 | ||
| dz_top: 3000 | ||
| h_elem: 16 |
There was a problem hiding this comment.
Could we mimic the functional albedo resolution config, or do these break for the lower resolution?
Also, I noticed that the temporal albedo slabplanet has a higher resolution than the other two albedo runs. Is it because it was breaking for helem=4?
There was a problem hiding this comment.
Good catch! I'll lower them to 4, I think it should be fine.
The temporal albedo slabplanet has a higher h_elem because I copied these from the non-GPU slabplanet runs, where the functional and static map albedo have h_elem: 4 and temporal has h_elem: 6. I'm not sure why this is, and happy to change them all to 4.
There was a problem hiding this comment.
Hmm, maybe it is unstable. 🤔 Worth trying though.
There was a problem hiding this comment.
they all work with the lower resolution :)
juliasloan25
force-pushed
the
js/gpu-amip-albedo
branch
from
63b871e to
2eba4fb
Compare
juliasloan25
force-pushed
the
js/gpu-amip-albedo
branch
from
2eba4fb to
01db1e4
Compare
Purpose
Builds off of #573 - copies the config file introduced there and adds two new runs based on it, using albedo from static map and temporal map instead of function.
The goal of the runs introduced in this PR is to ensure that the coupled simulation can run in the AMIP setup with both types of file reading. Because of this, the simulation length of these runs is only two timesteps. More complex, physically significant setups will be added to the longruns.
closes #583
Content