| | 110 | |
| | 111 | == Setup of parallelization in user code == |
| | 112 | |
| | 113 | The configuration of the parallelization using `PDAF3_init_parallel` is convenient because it is performed by PDAF. Thus on the user side, one is not concerned with how to implement this configuration, which involves the creating of MPI communicator (see below about the required communicators). However, there can be cases where it is not possible or convenient to let PDAF do the parallel configuration. For example, there can be cases where the ensemble setup of the parallelization was already done in the model code and one want to use this configuration to be used by PDAF. |
| | 114 | |
| | 115 | In PDAF V3.0 and before, the configuration of the parallelization was performed in the user code and then provided by PDAF. This possibility is still present. the template code file `templates/online/init_parallel_pdaf.F90.PDAF3_set_parallel' show an example for this approach. In this example, the necessary communicators are generated in the user code and then the routine `PDAF_set_parallel` is called to provide the parallelization variables to PDAF. The required communicators are explained below. Further, see the [wiki:PDAF3_set_parallel Page on PDAF_set_parallel] for documention on this routine. |