dubfi.mpi.parent_context
========================

.. py:module:: dubfi.mpi.parent_context

.. autoapi-nested-parse::

   Generic MPI helpers for MPI main/parent process.

   .. codeauthor:: Valentin Bruch, DWD <valentin.bruch@dwd.de>
   .. versionadded:: 0.2.0



Classes
-------

.. autoapisummary::

   dubfi.mpi.parent_context.MpiContext
   dubfi.mpi.parent_context.MpiObj


Functions
---------

.. autoapisummary::

   dubfi.mpi.parent_context.send_array


Module Contents
---------------

.. py:class:: MpiContext

   Singleton class keeping information about MPI setup.

   Return singleton when trying to create new instance.


   .. py:method:: setup(target_module: str = '', workers: int = -1) -> None

      Create MPI setup, create child process if necessary.



   .. py:method:: is_initialized() -> bool

      Check whether self has been initialized.



   .. py:method:: init(target_module: str, workers: int = -1) -> None

      Initialize self if it is not initialized yet.



   .. py:property:: workers
      :type: int


      Number of MPI workers.


   .. py:property:: root
      :type: int


      Number (id) of MPI root (or parent) process.


   .. py:property:: iter

      Iterator over ids of worker processes.


   .. py:property:: comm

      MPI communicator.


   .. py:method:: gen_name(prefix: str = '') -> str

      Generate a new, unique name.



   .. py:method:: send_cmd(command: str) -> None

      Send command to worker processes.



   .. py:method:: quit() -> None

      Send termination signal to MPI child processes.



   .. py:method:: __del__() -> None

      Destructor: quit child processes.



   .. py:method:: Reduce(recvbuf: numpy.ndarray, op: mpi4py.MPI.Op) -> None

      Wrap :meth:`mpi4py.MPI.Comm.Reduce` to receive data in root.



.. py:function:: send_array(array: numpy.ndarray | float) -> None

   Broadcast array to all child processes.


.. py:class:: MpiObj

   Base class for all MPI parent objects.

   Initialization generates self._name and destruction deletes the
   associated objects on worker processes.


   .. py:method:: __del__()

      Destructor: Delete associated objects on worker processes.