The previous version of hydroPSO used makeCluster to make PSOCKclusters using the par.nnodes argument. This is very user friendly for single machine cluster users. However, this prevents multi-machine PSOCK cluster users as there is no user friendly way to provide the hostnames to ssh into to spawn the PSOCK instances on other machines.
This code adds a "parallelPSOCK" parallel-type that takes par.hostnames to determine how many nodes are being launched and launches them by hostname.
Items left to fix:
The spacing and commenting conventions used were a bit unclear to me, so I just did my best. It may be better to modify them to be consistent with the rest of the code.
The previous version of hydroPSO used makeCluster to make PSOCKclusters using the par.nnodes argument. This is very user friendly for single machine cluster users. However, this prevents multi-machine PSOCK cluster users as there is no user friendly way to provide the hostnames to ssh into to spawn the PSOCK instances on other machines.
This code adds a "parallelPSOCK" parallel-type that takes par.hostnames to determine how many nodes are being launched and launches them by hostname.
An example using localhosts only:
Items left to fix: The spacing and commenting conventions used were a bit unclear to me, so I just did my best. It may be better to modify them to be consistent with the rest of the code.