Support for pbl scheme #66
Conversation
|
@oelbert - I think we've talked about not wanting to saddle the NDSL with "namelists" for the various components if we can. Will there be a physics/pySHiELD init routine and an init routine for every parameterization? If so, shouldn't this be where the input parameters are encapsulated? |
|
|
||
| Index3D = Tuple[int, int, int] | ||
|
|
||
|
|
||
| def set_4d_field_size(n, dtype): |
There was a problem hiding this comment.
Ping @CharlesKrop, we have a helper function in NDSL to get 4D field types after this PR is merged
There was a problem hiding this comment.
for 4D array, do we still keep k as vertical direction? so the verticalLoop node is the same as before, and horizontal execution nodes contain 3d loops?
|
@bensonr the PhysicsConfig dataclass should contain all of the parameters for the physics routines, but each scheme should have a child class that contains the parameters it needs, so there will be a PBLConfig, a MicrophysicsConfig, etc. Right now NDSL still has the infrastructure for parsing the namelists into Pace/PyFV3/PySHiELD, but I'd like to get some of these pjysics schemes in before tackling #64 |
|
|
||
| Index3D = Tuple[int, int, int] | ||
|
|
||
|
|
||
| def set_4d_field_size(n, dtype): |
There was a problem hiding this comment.
for 4D array, do we still keep k as vertical direction? so the verticalLoop node is the same as before, and horizontal execution nodes contain 3d loops?
|
@xyuan Data dimensions (anything after I, J, K) are executed serially. |
Description
This PR adds features needed for the ported PBL scheme. Mostly this means adding relevant namelist switches, but also includes a few useful physics constants and some 4d-field infrastructure
How Has This Been Tested?
The new namelist has been tested with the new PBLConfig in my PySHiELD branch, the 4d field allocator has also been used in the PBL scheme testing.
Checklist: