mpi_array.comms.CartLocaleComms¶

class mpi_array.comms.CartLocaleComms(ndims=None, dims=None, peer_comm=None, intra_locale_comm=None, inter_locale_comm=None, cart_comm=None)[source]¶

Bases: mpi_array.comms.LocaleComms

Defines cartesian communication topology for locales. In addition to the LocaleComms communicators, defines:

cart_comm: Typically this communicator is created using the call inter_locale_comm.Create_cart(...). This communicator (and associated mpi4py.MPI.Win window) is used to exchange data between locales. Note that inter_locale_comm property over-rides the LocaleComms.inter_locale_comm property to return the cart_comm communicator.

Methods

`__init__`([ndims, dims, peer_comm, ...])	Initialises cartesian communicator for inter-locale data exchange.
`alloc_locale_buffer`(shape, dtype)	Allocates a buffer using `mpi4py.MPI.Win.Allocate_shared()` which provides storage for the elements of the locale multi-dimensional array.
`free`()

Attributes

`cart_comm`	A `mpi4py.MPI.CartComm` communicator defining a cartesian topology of MPI processes (typically one process per locale) used for inter-locale exchange of array data.
`cart_coord_to_cart_rank_map`	A `dict` of `tuple` cartesian coordinate (`mpi4py.MPI.CartComm.Get_coords()`) keys which map to the associated `cart_comm` peer_rank.
`dims`	The number of partitions along each array axis.
`have_valid_cart_comm`	Is `True` if this peer_rank has `self.cart_comm` which is not `None` and is not `mpi4py.MPI.COMM_NULL`.
`have_valid_inter_locale_comm`	Is `True` if this peer rank has `self.inter_locale_comm` which is not `None` and is not `mpi4py.MPI.COMM_NULL`.
`inter_locale_comm`	Overrides `LocaleComms.inter_locale_comm` to return `cart_comm`.
`inter_locale_rank_to_peer_rank_map`	Returns sequence, `m` say, of `int` where `m[inter_r]` is the peer rank of `self.peer_comm` which corresponds to the inter-locale rank `inter_r` of `self.inter_locale_comm`.
`intra_locale_comm`	A `mpi4py.MPI.Comm` object which defines the group of processes which can allocate (and access) MPI window shared memory (allocated via `mpi4py.MPI.Win.Allocate_shared()` if available).
`ndim`	An `int` indicating the dimension of the cartesian topology.
`num_locales`	An `int` indicating the number of locales over which an array is distributed.
`peer_comm`	A `mpi4py.MPI.Comm` which is super-set of the `intra_locale_comm` and `inter_locale_comm` communicators.
`peer_ranks_per_locale`	A `numpy.ndarray` of shape `(self.num_locales, num_peer_ranks_per_locale)` with `numpy.int64` elements.
`rank_logger`	A `peer_comm` `logging.Logger`.
`root_logger`	A `peer_comm` `logging.Logger`.
`this_locale_rank_info`	A `ThisLocaleInfo` object.