- Shaped by: Magdalena
- Appetite (FTEs, weeks): 2 FTE, 1-2 weeks
- Developers: <!-- Filled in at the betting table unless someone is specifically required here -->

## Problem
Halo exchange calls with GHEX have been added to the python diffusion module in cycle 15. There are some leftovers:
- GHEX is cloned and built manually, it cannot yet be installed via pip install
- the code uses serialized ICON data (domain decomposition information, parallelized input/output fields for diffusion, parallelized metrics and interpolation fields) this data currently still is kept locally.
- verification of the parallel diffusion run against the serialized data has not been finished.

<!-- The raw idea, a use case, or something we’ve seen that motivates us to work on this -->

## Appetite
<!-- Explain how much time we want to spend and how that constrains the solution -->
It is mostly a clean up project.

- GHEX python build ~1 week
- ICON4py branch cleanup/verification ~1 week

## Solution
<!-- The core elements we came up with, presented in a form that’s easy for people to immediately understand -->
Bring the current experiment into a state that it can be merged and is usable for other developers (no/close to no manual build steps). There will probably remain a dependency on a system boost and MPI installation.

- [x] GHEX
- [x] add a python build cmake extension to GHEX (use pybind11/nanobind as blueprint) see example in https://github.com/pybind/scikit_build_example
- [x] ICON4py
- [x] add mpi4py and GHEX dependency to requirements.txt
- [x] mpi4py
- [x] pytest-mpi
- [x] GHEX (see above)
- [x] verify parallel diffusion run
- [x] make serialized data for parallel runs available from online storage
- [x] make sure all code runs in serial mode if no MPI/GHEX is present
- [x] dummy driver
- [x] tests
- [x] documentation (README): pre requisites, run of MPI tests, dummy driver (?)
- [x] fix logging configuration

## Rabbit holes
<!-- Details about the solution worth calling out to avoid problems -->
As always verification/debugging can be much more trouble some as expected.

## No-gos
<!-- Anything specifically excluded from the concept: functionality or use cases we intentionally aren’t covering to fit the ## appetite or make the problem tractable -->