Grid manager with backend

[halungge]

use GT4Py backend in GridManager:

• fields (coordinates geometry) are allocated using the backend
• the existing on_gpu flag for the IconGrid construction is determined from the backend, for backwards compatibilty
• GridManager internally explicitly uses numpy for reading

further changes and restrictions

• test_io.py: uses a backend fixed to field view embedded as a sample CPU backend.
• the IconGrid construction does not fully support the CUPY vs NUMPY import. It remains on CPU everywhere.

[halungge]

cscs-ci run default

[halungge]

cscs-ci run default

[halungge]

cscs-ci run default

[halungge]

cscs-ci run default

[nfarabullini]

I don't understand why we need this if we're using numpy afterwards

[halungge]

I added the xp only for the typing. (There is this NDArray TypeAlias) Inside the grid construction I kept this flag on_gpu as it was before for compatibility, that seems to work, but I have not yet bent my head around how it works. Rather not touched the grid construction. It think the solution there will eventually be that we read (on CPU) and then copy to device probably right before we return the grid as for the other fields.

For this we will need to do the import based on the backend I think.

[github-actions]

Mandatory Tests

Please make sure you run these tests via comment before you merge!

• cscs-ci run default
• launch jenkins spack

Optional Tests

To run benchmarks you can use:

• cscs-ci run benchmark

To run tests and benchmarks with the DaCe backend you can use:

• cscs-ci run dace

In case your change might affect downstream icon-exclaim, please consider running

• launch jenkins icon

For more detailed information please look at CI in the EXCLAIM universe.

[halungge]

cscs-ci run default