|
1
|
|
|
__version__ = '0.1.1' |
|
2
|
|
|
|
|
3
|
|
|
from functools import partial |
|
4
|
|
|
from itertools import starmap |
|
5
|
|
|
from typing import Optional |
|
6
|
|
|
|
|
7
|
|
|
|
|
8
|
|
|
def _starfunc(f, x): |
|
9
|
|
|
'''return f(*x) |
|
10
|
|
|
''' |
|
11
|
|
|
return f(*x) |
|
12
|
|
|
|
|
13
|
|
|
|
|
14
|
|
|
def _map_parallel_multiprocessing( |
|
15
|
|
|
f, *args, |
|
16
|
|
|
processes: Optional[int] = None, |
|
17
|
|
|
): |
|
18
|
|
|
from concurrent.futures import ProcessPoolExecutor |
|
19
|
|
|
|
|
20
|
|
|
with ProcessPoolExecutor(max_workers=processes) as process_pool_executor: |
|
21
|
|
|
return list(process_pool_executor.map(f, *args)) |
|
22
|
|
|
|
|
23
|
|
|
|
|
24
|
|
|
def _starmap_parallel_multiprocessing( |
|
25
|
|
|
f, args, |
|
26
|
|
|
processes: Optional[int] = None, |
|
27
|
|
|
): |
|
28
|
|
|
from concurrent.futures import ProcessPoolExecutor |
|
29
|
|
|
|
|
30
|
|
|
with ProcessPoolExecutor(max_workers=processes) as process_pool_executor: |
|
31
|
|
|
return list(process_pool_executor.map(partial(_starfunc, f), args)) |
|
32
|
|
|
|
|
33
|
|
|
|
|
34
|
|
|
def _map_parallel_multithreading( |
|
35
|
|
|
f, *args, |
|
36
|
|
|
processes: Optional[int] = None, |
|
37
|
|
|
): |
|
38
|
|
|
from concurrent.futures import ThreadPoolExecutor |
|
39
|
|
|
|
|
40
|
|
|
with ThreadPoolExecutor(max_workers=processes) as thread_pool_executor: |
|
41
|
|
|
return list(thread_pool_executor.map(f, *args)) |
|
42
|
|
|
|
|
43
|
|
|
|
|
44
|
|
|
def _starmap_parallel_multithreading( |
|
45
|
|
|
f, args, |
|
46
|
|
|
processes: Optional[int] = None, |
|
47
|
|
|
): |
|
48
|
|
|
from concurrent.futures import ThreadPoolExecutor |
|
49
|
|
|
|
|
50
|
|
|
with ThreadPoolExecutor(max_workers=processes) as thread_pool_executor: |
|
51
|
|
|
return list(thread_pool_executor.map(partial(_starfunc, f), args)) |
|
52
|
|
|
|
|
53
|
|
|
|
|
54
|
|
|
def _map_parallel_dask( |
|
55
|
|
|
f, *args, |
|
56
|
|
|
processes: Optional[int] = None, |
|
57
|
|
|
): |
|
58
|
|
|
from dask.distributed import Client, LocalCluster |
|
59
|
|
|
|
|
60
|
|
|
cluster = LocalCluster(n_workers=processes, dashboard_address=None) |
|
61
|
|
|
client = Client(cluster) |
|
62
|
|
|
return [future.result() for future in client.map(f, *args)] |
|
63
|
|
|
|
|
64
|
|
|
|
|
65
|
|
|
def _starmap_parallel_dask( |
|
66
|
|
|
f, args, |
|
67
|
|
|
processes: Optional[int] = None, |
|
68
|
|
|
): |
|
69
|
|
|
from dask.distributed import Client, LocalCluster |
|
70
|
|
|
|
|
71
|
|
|
cluster = LocalCluster(n_workers=processes, dashboard_address=None) |
|
72
|
|
|
client = Client(cluster) |
|
73
|
|
|
return [future.result() for future in client.map(partial(_starfunc, f), args)] |
|
74
|
|
|
|
|
75
|
|
|
|
|
76
|
|
|
def _map_parallel_mpi(f, *args, **kwargs): |
|
77
|
|
|
from mpi4py.futures import MPIPoolExecutor |
|
78
|
|
|
|
|
79
|
|
|
with MPIPoolExecutor() as mpi_pool_executor: |
|
80
|
|
|
return list(mpi_pool_executor.map(f, *args)) |
|
81
|
|
|
|
|
82
|
|
|
|
|
83
|
|
|
def _starmap_parallel_mpi(f, args, **kwargs): |
|
84
|
|
|
from mpi4py.futures import MPIPoolExecutor |
|
85
|
|
|
|
|
86
|
|
|
with MPIPoolExecutor() as mpi_pool_executor: |
|
87
|
|
|
return list(mpi_pool_executor.starmap(f, args)) |
|
88
|
|
|
|
|
89
|
|
|
|
|
90
|
|
|
_map_parallel_func = { |
|
91
|
|
|
'multiprocessing': _map_parallel_multiprocessing, |
|
92
|
|
|
'multithreading': _map_parallel_multithreading, |
|
93
|
|
|
'dask': _map_parallel_dask, |
|
94
|
|
|
'mpi': _map_parallel_mpi, |
|
95
|
|
|
} |
|
96
|
|
|
|
|
97
|
|
|
|
|
98
|
|
|
_starmap_parallel_func = { |
|
99
|
|
|
'multiprocessing': _starmap_parallel_multiprocessing, |
|
100
|
|
|
'multithreading': _starmap_parallel_multithreading, |
|
101
|
|
|
'dask': _starmap_parallel_dask, |
|
102
|
|
|
'mpi': _starmap_parallel_mpi, |
|
103
|
|
|
} |
|
104
|
|
|
|
|
105
|
|
|
|
|
106
|
|
|
def map_parallel( |
|
107
|
|
|
f, *args, |
|
108
|
|
|
processes: Optional[int] = None, |
|
109
|
|
|
mode: str = 'multiprocessing', |
|
110
|
|
|
): |
|
111
|
|
|
'''equiv to `map(f, *args)` but in parallel |
|
112
|
|
|
|
|
113
|
|
|
:param str mode: backend for parallelization |
|
114
|
|
|
- multiprocessing: using multiprocessing from standard library |
|
115
|
|
|
- multithreading: using multithreading from standard library |
|
116
|
|
|
- dask: using dask.distributed |
|
117
|
|
|
- mpi: using mpi4py.futures. May not work depending on your MPI vendor |
|
118
|
|
|
- serial: using map |
|
119
|
|
|
:param int processes: no. of parallel processes |
|
120
|
|
|
|
|
121
|
|
|
(in the case of mpi, it is determined by mpiexec/mpirun args) |
|
122
|
|
|
''' |
|
123
|
|
|
if processes is None or processes > 1: |
|
124
|
|
|
try: |
|
125
|
|
|
return _map_parallel_func[mode](f, *args, processes=processes) |
|
126
|
|
|
except KeyError: |
|
127
|
|
|
pass |
|
128
|
|
|
return list(map(f, *args)) |
|
129
|
|
|
|
|
130
|
|
|
|
|
131
|
|
|
def starmap_parallel( |
|
132
|
|
|
f, args, |
|
133
|
|
|
processes: Optional[int] = None, |
|
134
|
|
|
mode: str = 'multiprocessing', |
|
135
|
|
|
): |
|
136
|
|
|
'''equiv to `starmap(f, args)` but in parallel |
|
137
|
|
|
|
|
138
|
|
|
:param str mode: backend for parallelization |
|
139
|
|
|
- multiprocessing: using multiprocessing from standard library |
|
140
|
|
|
- multithreading: using multithreading from standard library |
|
141
|
|
|
- dask: using dask.distributed |
|
142
|
|
|
- mpi: using mpi4py.futures. May not work depending on your MPI vendor |
|
143
|
|
|
- serial: using map |
|
144
|
|
|
:param int processes: no. of parallel processes |
|
145
|
|
|
|
|
146
|
|
|
(in the case of mpi, it is determined by mpiexec/mpirun args) |
|
147
|
|
|
''' |
|
148
|
|
|
if processes is None or processes > 1: |
|
149
|
|
|
try: |
|
150
|
|
|
return _starmap_parallel_func[mode](f, args, processes=processes) |
|
151
|
|
|
except KeyError: |
|
152
|
|
|
pass |
|
153
|
|
|
return list(starmap(f, args)) |
|
154
|
|
|
|
ickc /
python-map_parallel
