ptypy.utils.parallel

Utility functions and classes to support MPI computing.

This file is part of the PTYPY package.

copyright:

Copyright 2014 by the PTYPY team, see AUTHORS.

license:

see LICENSE for details.

Functions

MPImax(a[, axis])	Compute the maximum over a list of arrays distributed over multiple processes.
MPImin(a[, axis])	Compute the minimum over a list of arrays distributed over multiple processes.
MPInoise2d(sh[, rms, mfs, rms_mod, mfs_mod])	Creates complex-valued statistical noise in the shape of sh consistent across all nodes.
MPIprod(a[, axis])	Compute the product over a list of arrays distributed over multiple processes.
MPIrand_normal([loc, scale, size])	Wrapper for np.random.normal for same random sample across all nodes.
MPIrand_uniform([low, high, size])	Wrapper for np.random.uniform for same random sample across all nodes.
MPIsum(a[, axis])	Compute the sum of list of arrays distributed over multiple processes.
allgather_dict(dct)	Allgather dict in place.
allreduce(a[, op])	Wrapper for comm.Allreduce, always in place.
allreduceC(c)	Performs MPI parallel allreduce with a sum as reduction for all Storage instances held by Container c
barrier()	Wrapper for comm.Barrier.
bcast(data[, source])	Wrapper for comm.bcast and comm.Bcast.
bcast_dict(dct[, keys, source])	Broadcasts or scatters a dict dct from rank==source.
gather_dict(dct[, target])	Gathers broadcasted or scattered dict dct at rank target.
gather_list(lst, length, indices)	gathers list lst of all processes to a list of length length according to order given by indices.
receive([source, tag])	Wrapper for comm.Recv.
send(data[, dest, tag])	Wrapper for comm.Send and comm.send.
useMPI([do])	Toggle using MPI or not.

Classes

LoadManager()