:py:mod:`bempp.api.utils.pool`
==============================

.. py:module:: bempp.api.utils.pool

.. autoapi-nested-parse::

   Routines to administrate a process pool.



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

Classes
~~~~~~~

.. autoapisummary::

   bempp.api.utils.pool.Pool



Functions
~~~~~~~~~

.. autoapisummary::

   bempp.api.utils.pool.worker
   bempp.api.utils.pool.as_array
   bempp.api.utils.pool.to_buffer
   bempp.api.utils.pool.from_buffer
   bempp.api.utils.pool._raise_if_not_worker
   bempp.api.utils.pool.number_of_workers
   bempp.api.utils.pool.insert_data
   bempp.api.utils.pool.is_initialised
   bempp.api.utils.pool.clear_data
   bempp.api.utils.pool.get_data
   bempp.api.utils.pool.get_id
   bempp.api.utils.pool.remove_key
   bempp.api.utils.pool.has_key
   bempp.api.utils.pool.is_worker
   bempp.api.utils.pool.create_device_pool
   bempp.api.utils.pool.create_pool
   bempp.api.utils.pool.nworkers
   bempp.api.utils.pool.map
   bempp.api.utils.pool.starmap
   bempp.api.utils.pool.execute
   bempp.api.utils.pool._assign_ids
   bempp.api.utils.pool.shutdown
   bempp.api.utils.pool._execute_function_without_arguments
   bempp.api.utils.pool._remove_key_worker
   bempp.api.utils.pool._init_device_worker
   bempp.api.utils.pool._clear_data_worker



Attributes
~~~~~~~~~~

.. autoapisummary::

   bempp.api.utils.pool._DATA
   bempp.api.utils.pool._MY_ID
   bempp.api.utils.pool._USE_THREADS
   bempp.api.utils.pool._POOL
   bempp.api.utils.pool._NWORKERS
   bempp.api.utils.pool._IN_WORKER
   bempp.api.utils.pool._BUFFER


.. py:data:: _DATA

   

.. py:data:: _MY_ID

   

.. py:data:: _USE_THREADS

   

.. py:data:: _POOL

   

.. py:data:: _NWORKERS

   

.. py:data:: _IN_WORKER
   :value: False

   

.. py:data:: _BUFFER

   

.. py:function:: worker(in_queue, out_queue, worker_id, nworkers, buf, log, log_level)

   Definition of a worker.


.. py:function:: as_array(dtype, offset, shape)

   Return part of the buffer as array.

   Parameters
   ----------
   dtype : Numpy dtype object
       The type of the array
   offset : int
       Start index in buffer
   nitems : int
       Number of items of type dtype



.. py:function:: to_buffer(*args)

   Save a number of numpy arrays to a buffer.


.. py:function:: from_buffer(arrays)

   Retrieve arrays from buffer.

   arrays is a list of tuples
   [(dtype1, shape1), (dtype2, shape2), ...],
   where dtype is the type of the array and shape
   is a shape tuple.



.. py:class:: Pool(nworkers, buffer_size=100, log=False, log_level='info')


   Bases: :py:obj:`object`

   A simple pool.

   This pool is different from the multiprocessing pool
   in that map operations are guaranteed to execute on
   all processes, and we can directly target individual
   processes.


   .. py:property:: number_of_workers

      Return number of workers.


   .. py:method:: _map_impl(fun, args, options)

      Map implementation.


   .. py:method:: map(fun, args=None)

      Map function onto workers.


   .. py:method:: starmap(fun, args)

      Map function onto workers.


   .. py:method:: shutdown()

      Shutdown all workers.



.. py:function:: _raise_if_not_worker(name)

   Raise exception if not in worker.


.. py:function:: number_of_workers()

   Return number of workers.


.. py:function:: insert_data(key, data)

   Insert data.


.. py:function:: is_initialised()

   Return true if pool is initialised.


.. py:function:: clear_data()

   Clear the data in all workers.


.. py:function:: get_data(key)

   Return data.


.. py:function:: get_id()

   Return my id.


.. py:function:: remove_key(key)

   Remove data from pool.


.. py:function:: has_key(key)

   Return if key exists in pool data.


.. py:function:: is_worker()

   Return true if called from worker process.


.. py:function:: create_device_pool(identifier, buffer_size=100, log=False, log_level='info', max_workers=-1, precision=None)

   Create a pool based on a given platform identifer.

   identifier : string
       A unique identifier that is part of the platform name.
       Used to find the correct platform.
   buffer_size : int
       Shared memory buffer size in MB
   log : Boolean
       Set to True to log workers.
   log_level : String
       Logging level. One of 'debug', 'info', 'warning', 'error', 'critical'
   max_workers : int
       Maximum number of workers. If max_workers=-1 (default)
       the maximum number of workers is identical to the number
       of devices in the pool.
   precision : string or None
       Precision for the devices in the pool. If precision is None use
       single precision for GPU devices and double precision for CPU devices.
       If precision is 'single' or 'double' use the corresponding mode for the pool.



.. py:function:: create_pool(nworkers, buffer_size=100, log=False, log_level='info')

   Create a pool.


.. py:function:: nworkers()

   Get number of workers.


.. py:function:: map(fun, args)

   Corresponds to multiprocessing map.


.. py:function:: starmap(fun, args)

   Corresponds to multiprocessing map.


.. py:function:: execute(fun, *args)

   Execute function with the same arguments on all workers.


.. py:function:: _assign_ids(nworkers)

   Assign pool ids.


.. py:function:: shutdown()

   Shutdown the pool.


.. py:function:: _execute_function_without_arguments(fun)

   Execute function without arguments.


.. py:function:: _remove_key_worker(key)


.. py:function:: _init_device_worker(identifier, precision)

   Worker to initialise device.


.. py:function:: _clear_data_worker()

   Clear worker.