Refactor advection stencils to use StencilTest
#336
Conversation
|
cscs-ci run |
|
cscs-ci run |
|
cscs-ci run default |
1 similar comment
|
cscs-ci run default |
|
cscs-ci run default |
|
cscs-ci run default |
1 similar comment
|
cscs-ci run default |
|
cscs-ci run benchmark |
|
cscs-ci run default |
|
cscs-ci run default |
|
cscs-ci run default |
|
cscs-ci run default |
|
launch jenkins spack |
|
launch jenkins icon |
|
launch jenkins spack |
|
cscs-ci run default |
|
cscs-ci run default |
|
launch jenkins spack |
|
launch jenkins icon |
| @@ -30,7 +30,7 @@ def _btraj_dreg_stencil_01( | |||
| return lvn_sys_pos | |||
|
|
|||
|
|
|||
| @program | |||
| @program(grid_type="unstructured") | |||
There was a problem hiding this comment.
All dycore stencils are written like this: grid_type=GridType.UNSTRUCTURED I imagine the two are equivalent but perhaps it would be better to be uniform with the dycore. This is a detail so I leave it up to you.
There was a problem hiding this comment.
Those statements should have been added in all stencils, correct? Most of the hflx_limiter_mo/pd_stencil, hor_adv_stencil_01, rbf_intp_edge_stencil_01 don't have them.
There was a problem hiding this comment.
I am not sure if it is best practice to do this for all stencil programs. I just applied this to the stencils which failed to execute otherwise. Asking @tehrengruber what he thinks.
| PROGRAM = face_val_ppm_stencil_01 | ||
| OUTPUTS = ( | ||
| Output( | ||
| "z_slope", refslice=(slice(None), slice(None, -1)), gtslice=(slice(None), slice(1, -1)) |
There was a problem hiding this comment.
This is useful thanks. If I understood correctly gtslice is the slicing of the GridTools output and in parenthesis are the dimensions.
There was a problem hiding this comment.
Yes so refslice is the slice for the reference output, and gtslice is the slice for the gt4py output field. By default both are None which means the whole field.
There was a problem hiding this comment.
You can also still use the normal syntax: OUTPUTS = ("a", "b") but then the fields are not sliced.
| z_slope.asnumpy(), | ||
| ) | ||
| # Concatenate the NaN column to the existing array | ||
| nan_column = np.full((p_face.shape[0], 2), np.nan) |
There was a problem hiding this comment.
Why not using the output option and slice the two last columns of the GridTools output?
There was a problem hiding this comment.
Yes that makes sense, thanks for the suggestion.
|
Mandatory Tests Please make sure you run these tests via comment before you merge!
Optional Tests To run benchmarks you can use:
In case your change might affect downstream icon-exclaim, please consider running
For more detailed information please look at CI in the EXCLAIM universe. |
|
cscs-ci run default |
|
launch jenkins spack |
|
cscs-ci run default |
Refactor half of all advection stencil to use StencilTest class, and make outputs sliceable. --------- Co-authored-by: Nina Burgdorfer <Nina.Burgdorfer@meteoswiss.ch>
Refactor advection stencils to use
StencilTestclass.