Aim
The integrators should work in parallel on multiple MPI nodes. This requires
- Semi-automatic load balancing. I already created a class
IntegrationLoadBalancer. The idea behind this class is that one could register integrators with this load balancer and later on call its balance() method to optimize the workload for each node.
- Proper data exchange and integration with the time integration
Aim
The integrators should work in parallel on multiple MPI nodes. This requires
IntegrationLoadBalancer. The idea behind this class is that one could register integrators with this load balancer and later on call itsbalance()method to optimize the workload for each node.