| 1 |
1 |
# Parallel processing
|
| 2 |
2 |
|
|
3 |
import Queue
|
| 3 |
4 |
import warnings
|
| 4 |
5 |
|
| 5 |
|
class Pool:
|
|
6 |
class SyncPool:
|
| 6 |
7 |
'''A dummy synchronous Pool to use if multiprocessing is not available'''
|
|
8 |
def __init__(self, processes=None): pass
|
|
9 |
|
| 7 |
10 |
class Result:
|
| 8 |
11 |
def __init__(self, value): self.value = value
|
| 9 |
12 |
|
| ... | ... | |
| 24 |
27 |
callback(value)
|
| 25 |
28 |
return Result(value)
|
| 26 |
29 |
|
| 27 |
|
def mk_pool(cpus=None):
|
| 28 |
|
'''@return tuple(pool, cpus)'''
|
| 29 |
|
try:
|
| 30 |
|
if cpus == 0: raise ImportError('Parallel processing turned off')
|
| 31 |
|
import multiprocessing
|
| 32 |
|
import multiprocessing.pool
|
| 33 |
|
except ImportError, e:
|
| 34 |
|
warnings.warn(UserWarning('Not using parallel processing: '+str(e)))
|
| 35 |
|
pool = Pool()
|
| 36 |
|
else:
|
| 37 |
|
if cpus == None: cpus = multiprocessing.cpu_count()
|
| 38 |
|
pool = multiprocessing.pool.Pool(processes=cpus)
|
| 39 |
|
return (pool, cpus)
|
|
30 |
class MultiProducerPool:
|
|
31 |
'''A multi-producer pool. You must call pool.main_loop() in the thread that
|
|
32 |
created this to process new tasks.'''
|
|
33 |
def __init__(self, processes=None):
|
|
34 |
'''
|
|
35 |
@param processes If 0, uses SyncPool
|
|
36 |
@post The # processes actually used is made available in self.process_ct
|
|
37 |
'''
|
|
38 |
try:
|
|
39 |
if processes == 0: raise ImportError('turned off')
|
|
40 |
import multiprocessing
|
|
41 |
import multiprocessing.pool
|
|
42 |
except ImportError, e:
|
|
43 |
warnings.warn(UserWarning('Not using parallel processing: '+str(e)))
|
|
44 |
processes = 1
|
|
45 |
Pool_ = SyncPool
|
|
46 |
Queue_ = Queue.Queue
|
|
47 |
else:
|
|
48 |
if processes == None: processes = multiprocessing.cpu_count()
|
|
49 |
Pool_ = multiprocessing.pool.Pool
|
|
50 |
Queue_ = multiprocessing.Queue
|
|
51 |
|
|
52 |
self.process_ct = processes
|
|
53 |
self.pool = Pool_(processes)
|
|
54 |
self.queue = Queue_()
|
|
55 |
|
|
56 |
def main_loop(self):
|
|
57 |
'''@param pool Must be a pool returned by mk_pool()'''
|
|
58 |
try:
|
|
59 |
# block=False raises Empty immediately if the queue is empty,
|
|
60 |
# which indicates that the program is done
|
|
61 |
while True: self.queue.get(block=False)() # each elem is a function
|
|
62 |
except Queue.Empty: pass
|
|
63 |
|
|
64 |
class Result:
|
|
65 |
def get(timeout=None): raise NotImplementedError()
|
|
66 |
|
|
67 |
def wait(timeout=None): raise NotImplementedError()
|
|
68 |
|
|
69 |
def ready(): raise NotImplementedError()
|
|
70 |
|
|
71 |
def successful(): raise NotImplementedError()
|
|
72 |
|
|
73 |
def apply_async(*args, **kw_args):
|
|
74 |
def call(): self.pool.apply_async(*args, **kw_args)
|
|
75 |
self.queue.put_nowait(call)
|
|
76 |
return Result()
|
parallel.py: Changed to use multi-producer pool, which requires calling pool.main_loop()