Clean up global grid parameters and grid shape#1183
Conversation
|
Hi there, this is jenkins continuous integration... |
1 similar comment
|
Hi there, this is jenkins continuous integration... |
Agent-Logs-Url: https://github.com/C2SM/icon4py/sessions/148d4123-da81-4914-a9e3-16465bc49c48 Co-authored-by: jcanton <5622559+jcanton@users.noreply.github.com>
nfarabullini
left a comment
nfarabullini
left a comment
There was a problem hiding this comment.
a couple of preliminary comments
nfarabullini
left a comment
nfarabullini
left a comment
There was a problem hiding this comment.
Some other minor comments
nfarabullini
left a comment
nfarabullini
left a comment
There was a problem hiding this comment.
Look good so far but conflicts need to be solved
…al-grid-parameters
Indeed. Note: this is still a draft, I'm not ready with the changes. |
msimberg
changed the title
|
I'm stopping here with this PR. I looked into completely removing the GridParams metadata from GridDescription, but decided I don't want to mix that into this PR. I'd like to do that if/when we refactor datatest fixtures (currently there are is e.g. no fixture for reading a grid directly from a grid file; the |
|
cscs-ci run default |
|
cscs-ci run distributed |
| case icon.GeometryType.ICOSAHEDRON: | ||
| return icon.GridParams( | ||
| params=icon.IcosahedronParams( | ||
| subdivision=icon.GridSubdivision(root=grid_root, level=grid_level), |
There was a problem hiding this comment.
@jcanton Now I remember that why the mean cell area is different from icon's serialized data. Because the original icon computes mean cell area on the global grid with the equation using grid root and level (total globe's surface area divided by num of cells).
There was a problem hiding this comment.
That also possibly means that running the standalone driver will of course produce a different result from Fortran icon.
There was a problem hiding this comment.
Not sure that if the mean cell area on the limited area may also be different.
There was a problem hiding this comment.
Yeah, I think it would be different on limited area... and yes, this always a bit of a contentious thing. It's weird the way icon fortran calculates it (in my opinion), but if we care about exactly reproducing icon fortran results then we may need to take this into consideration. RBF uses mean_characteristic_length (which is derived from the mean cell area) to scale the RBF kernels, so that could introduce additional differences.
|
cscs-ci run default |
|
Mandatory Tests Please make sure you run these tests via comment before you merge!
Optional Tests To run benchmarks you can use:
To run tests and benchmarks with the DaCe backend you can use:
To run test levels ignored by the default test suite (mostly simple datatest for static fields computations) you can use:
For more detailed information please look at CI in the EXCLAIM universe. |
|
cscs-ci run distributed |
Rename GlobalGridParams to GridParams and removes most members. GridParams now only contains the type of grid and dimensions (domain height/length for torus, radius for icosahedron; the R/B params are also stored for icosahedron but are just metadata that's not required).