`parameter_sweep.parallel.parallel_manager_factory`

Module Contents

Functions

`create_parallel_manager`([parallel_manager_class])	Create and return an instance of a ParallelManager, based on the libraries available in the
`has_mpi_peer_processes`()	Returns whether the process was run as part of an MPI group with > 1 processes.
`get_mpi_comm_process`()	Returns mpi comm world
`should_fan_out`(number_of_subprocesses)	Returns whether the manager should fan out the computation to subprocesses. This

parameter_sweep.parallel.parallel_manager_factory.create_parallel_manager(parallel_manager_class=None, **kwargs)

Create and return an instance of a ParallelManager, based on the libraries available in the runtime environment.

Allows an optional python class to be passed in as parallel_manager_class. If so, this class is instantiated and returned rather than checking the local environment.

parameter_sweep.parallel.parallel_manager_factory.has_mpi_peer_processes(): Returns whether the process was run as part of an MPI group with > 1 processes.

parameter_sweep.parallel.parallel_manager_factory.get_mpi_comm_process(): Returns mpi comm world

parameter_sweep.parallel.parallel_manager_factory.should_fan_out(number_of_subprocesses): Returns whether the manager should fan out the computation to subprocesses. This is mutually exclusive with the process running under MPI.