Closed rogerkuou closed 2 years ago
An initial implementation on parallelizing per GPI is ready on the python_da_workflow branch. An example directory has been create on Snellius: /home/eccei339/data/SURFEX/KTEST/fran10_test_esci/LDAS_Monde_script/LDAS_Monde_slurm/da_workflow
.
In the example directory, on can submit Slurm job for multiple GPI, running a dummy loop in time per GPI.
Three parts are currently missing, which will be finished in the following work:
offline.exe
: #273
The python data assimilation part contains a spatial loop per grid. It can be taken out. Then the how workflow can be parallelized on the spatial level.