lib.mpi.MPIPool¶

class lib.mpi.MPIPool(mpicomm=None, check_tasks=False)¶

Bases: object

A processing pool that distributes tasks using MPI. With this pool class, the master process distributes tasks to worker processes using an MPI communicator.

This implementation is inspired by @juliohm in this module and was adapted from schwimmbad.

initialize MPIPool.

Parameters

mpicomm (MPI communicator, default=None) – Communicator. Defaults to current communicator.
check_tasks (bool, default=False) – Check that same tasks are input for all processes, if not, raises a ValueError.

Methods

`close`	Tell all the workers to quit.
`is_master`	Is the current process the master process? Master is responsible for distributing the tasks to the other available ranks.
`is_worker`	Is the current process a valid worker? Workers wait for instructions from the master.
`map`	Evaluate a function or callable on each task in parallel using MPI.
`wait`	Tell the workers to wait and listen for the master process.

close()¶: Tell all the workers to quit.

is_master()¶: Is the current process the master process? Master is responsible for distributing the tasks to the other available ranks.

is_worker()¶: Is the current process a valid worker? Workers wait for instructions from the master.

map(function, tasks)¶

Evaluate a function or callable on each task in parallel using MPI. The callable, worker, is called on each element of the tasks iterable. The results are returned in the expected order.

Parameters

function (callable) – A function or callable object that is executed on each element of the specified tasks iterable. This should accept a single positional argument and return a single object.
tasks (iterable) – A list or iterable of tasks. Each task can be itself an iterable (e.g., tuple) of values or data to pass in to the worker function.

Returns

results – A list of results from the output of each function call.

Return type

list

wait()¶: Tell the workers to wait and listen for the master process. This is called automatically when using MPIPool.map() and doesn’t need to be called by the user.