The mpi_array.locale Module

Defines LndarrayProxy class and factory functions for creating multi-dimensional arrays where memory is allocated using mpi4py.MPI.Win.Allocate_shared() or mpi4py.MPI.Win.Allocate().

Classes

lndarray

Sub-class of numpy.ndarray which requires buffer to be specified for instantiation.

LndarrayProxy	Proxy for lndarray instances.
PartitionViewSlices	Stores multiple tuple-of-slice objects indicating the slice (tile) of the lndarray on which a intra_locale_comm rank MPI process operates.

Factory Functions

empty([shape, dtype, comms_and_distrib, ...])	Creates array of uninitialised elements.
empty_like(ary[, dtype])	Return a new array with the same shape and type as a given array.
zeros([shape, dtype, comms_and_distrib, order])	Creates array of zero-initialised elements.
zeros_like(ary, *args, **kwargs)	Return a new zero-initialised array with the same shape and type as a given array.
ones([shape, dtype, comms_and_distrib, order])	Creates array of one-initialised elements.
ones_like(ary, *args, **kwargs)	Return a new one-initialised array with the same shape and type as a given array.
copy(ary)	Return an array copy of the given object.

Utilities

NdarrayMetaData(offset, strides, order)

Encapsulates, strides, offset and order argument of LndarrayProxy.__new__().