mpi_array.globale_ufunc.GndarrayArrayUfuncExecutor¶

class mpi_array.globale_ufunc.GndarrayArrayUfuncExecutor(array_like_obj, ufunc, method, *inputs, **kwargs)[source]¶

Bases: object

Instances execute a ufunc for a mpi_array.globale.gndarray. Takes care of creating outputs, remote fetching of required parts of inputs and forwarding call to numpy.ufunc instance to perform the computation on the locale numpy.ndarray instances.

Methods

`__init__`(array_like_obj, ufunc, method, ...)	Initialise.
`create_outputs`(outputs, result_shape, ...)	Returns list of output `mpi_array.globale.gndarray` instances.
`execute`()	Perform the ufunc operation.
`execute___call__`()
`execute_accumulate`()	Not implemented.
`execute_at`()	Not implemented.
`execute_outer`()	Not implemented.
`execute_reduce`()	Not implemented.
`execute_reduceat`()	Not implemented.
`get_best_match_input`(result_shape)	Returns the element of `inputs` whose globale shape best matches `result_shape`.
`get_input_extents`(locale_info)	Returns tuple of `(locale_extent, globale_extent)` pairs, one for each of the `inputs`.
`get_inputs_shapes`()	Returns a shape `tuple` for each element of `inputs`.
`get_numpy_ufunc_peer_rank_inputs_outputs`(...)	Returns two element tuple of `(input_arrays, output_arrays)` which are to be passed to the `numpy.ufunc` object `ufunc`.
`need_remote_data`(gndarray_outputs)	Returns `True` if any locale needs to fetch remote input data in order to compute the all elements of the outputs `gndarray_outputs`.

Attributes

`array_like_obj`	The `mpi_array.globale.gndarray` object which triggered the construction of this `GndarrayArrayUfuncExecutor` object.
`casting`	A `str` indicating the casting mode.
`inputs`	The sequence of ufunc inputs.
`inter_locale_comm`	The inter-locale `mpi4py.MPI.Comm` communicator.
`intra_locale_comm`	The intra-locale `mpi4py.MPI.Comm` communicator.
`method`	A `str` indicating the method of the `ufunc` to be executed.
`outputs`	The ufunc `mpi_array.globale.gndarray` output arrays.
`peer_comm`	The peer `mpi4py.MPI.Comm` communicator.
`ufunc`	The `numpy.ufunc` to be executed.