#
##
## This file is part of pyFormex 2.0 (Mon Sep 14 12:29:05 CEST 2020)
## pyFormex is a tool for generating, manipulating and transforming 3D
## geometrical models by sequences of mathematical operations.
## Home page: http://pyformex.org
## Project page: http://savannah.nongnu.org/projects/pyformex/
## Copyright 2004-2020 (C) Benedict Verhegghe (benedict.verhegghe@ugent.be)
## Distributed under the GNU General Public License version 3 or later.
##
## This program is free software: you can redistribute it and/or modify
## it under the terms of the GNU General Public License as published by
## the Free Software Foundation, either version 3 of the License, or
## (at your option) any later version.
##
## This program is distributed in the hope that it will be useful,
## but WITHOUT ANY WARRANTY; without even the implied warranty of
## MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
## GNU General Public License for more details.
##
## You should have received a copy of the GNU General Public License
## along with this program. If not, see http://www.gnu.org/licenses/.
##
#
"""Framework for multi-processing in pyFormex
This module contains some functions to perform multiprocessing
in pyFormex in a unified way.
"""
import multiprocessing as mp
from multiprocessing import cpu_count
import numpy as np
import pyformex as pf
from pyformex.arraytools import splitar
[docs]def splitArgs(args, mask=None, nproc=-1, close=False):
"""Split data blocks over multiple processors.
Parameters
----------
args: list or tuple of :term:`array_like`
A sequence of data blocks that may need to be split for parallel
processing over multiple processors. Splitting is done along the
first dimension, which should therefore be the same for all arrays
in the sequence that need to be split.
mask: list or tuple of bool
If provided, this flags the items in ``args`` that should be split.
The list should have the same length as ``args``.
If not provided, all array type items in ``args`` will be split.
nproc: int
Intended number of processors. If negative (default), it is set equal
to the number of processors detected on the host machine.
close: bool
If True, the elements where the arrays are split are included in both
blocks delimited by the element. Thus splitting an array [1, 2, 3]
in two results in [1, 2] and [2, 3], while the default split would
be [1, 2] and [3].
Returns
-------
list of tuples.
The list contains ``nproc`` tuples and each tuple contains the same
number of items as the input ``args`` and in the same order, whereby
the (nonmasked) arrays are replaced by a slice of the array along
its first axis, and the masked and non-array items are replicated
as is.
See Also
--------
arraytools.splitar: the low level function used to do the splitting
Examples
--------
>>> splitArgs([np.arange(5),'abcde'],nproc=3)
[(array([0, 1]), 'abcde'), (array([2]), 'abcde'), (array([3, 4]), 'abcde')]
>>> for i in splitArgs([np.eye(5),'=>',np.arange(5)],nproc=3):
... print("%s %s %s" % i)
[[ 1. 0. 0. 0. 0.]
[ 0. 1. 0. 0. 0.]] => [0 1]
[[ 0. 0. 1. 0. 0.]] => [2]
[[ 0. 0. 0. 1. 0.]
[ 0. 0. 0. 0. 1.]] => [3 4]
>>> for i in splitArgs([np.eye(5),'=>',np.arange(5)],mask=[1,0,0],nproc=3):
... print("%s %s %s" % i)
[[ 1. 0. 0. 0. 0.]
[ 0. 1. 0. 0. 0.]] => [0 1 2 3 4]
[[ 0. 0. 1. 0. 0.]] => [0 1 2 3 4]
[[ 0. 0. 0. 1. 0.]
[ 0. 0. 0. 0. 1.]] => [0 1 2 3 4]
"""
if nproc < 0:
nproc = mp.cpu_count()
if mask is None:
mask = [isinstance(a, np.ndarray) for a in args]
split = [
splitar(a, nproc, close=close) if m else [a] * nproc
for a, m in zip(args, mask)
]
return list(zip(*split))
[docs]def dofunc(arg):
"""Helper function for the multitask function.
Parameters
----------
arg: tuple
The first item of the tuple is a callable. The remaining
items are its arguments.
Returns
-------
object
The value of the callable when passed the remaining items as
arguments.
Examples
--------
>>> dofunc((max,(2, 5, 3)))
5
"""
func, args = arg
return func(*args)
[docs]def multitask(tasks, nproc=-1):
"""Perform tasks in parallel.
Runs a number of tasks in parallel over a number of subprocesses.
Parameters
----------
tasks: list of tuples
Each task in the list is a tuple where the first item is a callable
and the other items are the arguments to be passed to the callable.
nproc: int
The number of subprocesses to be started. This may be
different from the number of tasks to run: processes finishing a
task will pick up a next one. There is no benefit in starting more
processes than the number of tasks or the number of processing units
available. The default will set ``nproc`` to the minimum of these two
values.
Examples
--------
>>> task1 = (int.__add__, (2, 3))
>>> task2 = (int.__mul__, (2, 3))
>>> multitask((task1, task2))
[5, 6]
"""
if nproc < 0:
nproc = min(len(tasks), mp.cpu_count())
pf.debug("Multiprocessing using %s processors" % nproc, pf.DEBUG.MULTI)
if __name__ == '__draw__':
if pf.warning(
"""..
Multiprocessing in 'script' mode
================================
You are trying to use multiprocessing while running in 'script' mode.
Multiprocessing in 'script' mode may cause pyFormex to hang indefinitely.
We strongly advice you to cancel the operation now and to run your
application in 'app' mode. Multiprocessing runs fine in 'app' mode.
""",
actions=['Cancel', 'I know the risks and insist on continuing']
) == 'Cancel':
return
pool = mp.Pool(nproc)
res = pool.map(dofunc, tasks)
pool.close()
return res
# End