1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
|
<?xml version="1.0" encoding="utf-8" ?>
<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
<html xmlns="http://www.w3.org/1999/xhtml" xml:lang="en" lang="en">
<head>
<meta http-equiv="Content-Type" content="text/html; charset=utf-8" />
<meta name="generator" content="Docutils 0.4: http://docutils.sourceforge.net/" />
<title>Introduction</title>
<link rel="stylesheet" href="html4css1.css" type="text/css" />
</head>
<body>
<div class="header">
<a class="reference" href="index.html">Prev</a> <a class="reference" href="index.html">Up</a> <a class="reference" href="processing-ref.html">Next</a>
<hr class="header"/>
</div>
<div class="document" id="introduction">
<h1 class="title">Introduction</h1>
<div class="section">
<h1><a id="threads-processes-and-the-gil" name="threads-processes-and-the-gil">Threads, processes and the GIL</a></h1>
<p>To run more than one piece of code at the same time on the same
computer one has the choice of either using multiple processes or
multiple threads.</p>
<p>Although a program can be made up of multiple processes, these
processes are in effect completely independent of one another:
different processes are not able to cooperate with one another unless
one sets up some means of communication between them (such as by using
sockets). If a lot of data must be transferred between processes then
this can be inefficient.</p>
<p>On the other hand, multiple threads within a single process are
intimately connected: they share their data but often can interfere
badly with one another. It is often argued that the only way to make
multithreaded programming "easy" is to avoid relying on any shared
state and for the threads to only communicate by passing messages to
each other.</p>
<p>CPython has a <em>Global Interpreter Lock</em> (GIL) which in many ways makes
threading easier than it is in most languages by making sure that only
one thread can manipulate the interpreter's objects at a time. As a
result, it is often safe to let multiple threads access data without
using any additional locking as one would need to in a language such
as C.</p>
<p>One downside of the GIL is that on multi-processor (or multi-core)
systems a multithreaded Python program can only make use of one
processor at a time. This is a problem that can be overcome by using
multiple processes instead.</p>
<p>Python gives little direct support for writing programs using multiple
process. This package allows one to write multi-process programs
using much the same API that one uses for writing threaded programs.</p>
</div>
<div class="section">
<h1><a id="forking-and-spawning" name="forking-and-spawning">Forking and spawning</a></h1>
<p>There are two ways of creating a new process in Python:</p>
<ul>
<li><p class="first">The current process can <em>fork</em> a new child process by using the
<tt class="docutils literal"><span class="pre">os.fork()</span></tt> function. This effectively creates an identical copy
of the current process which is now able to go off and perform some
task set by the parent process. This means that the child process
inherits <em>copies</em> of all variables that the parent process had.</p>
<p>However, <tt class="docutils literal"><span class="pre">os.fork()</span></tt> is not available on every platform: in
particular Windows does not support it.</p>
</li>
<li><p class="first">Alternatively, the current process can spawn a completely new Python
interpreter by using the <tt class="docutils literal"><span class="pre">subprocess</span></tt> module or one of the
<tt class="docutils literal"><span class="pre">os.spawn*()</span></tt> functions.</p>
<p>Getting this new interpreter in to a fit state to perform the task
set for it by its parent process is, however, a bit of a challenge.</p>
</li>
</ul>
<p>The <tt class="docutils literal"><span class="pre">processing</span></tt> package uses <tt class="docutils literal"><span class="pre">os.fork()</span></tt> if it is available since
it makes life a lot simpler. Forking the process is also more
efficient in terms of memory usage and the time needed to create the
new process.</p>
</div>
<div class="section">
<h1><a id="the-process-class" name="the-process-class">The Process class</a></h1>
<p>In the <tt class="docutils literal"><span class="pre">processing</span></tt> package processes are spawned by creating a
<tt class="docutils literal"><span class="pre">Process</span></tt> object and then calling its <tt class="docutils literal"><span class="pre">start()</span></tt> method.
<tt class="docutils literal"><span class="pre">processing.Process</span></tt> follows the API of <tt class="docutils literal"><span class="pre">threading.Thread</span></tt>. A
trivial example of a multiprocess program is</p>
<pre class="literal-block">
from processing import Process
def f(name):
print 'hello', name
if __name__ == '__main__':
p = Process(target=f, args=('bob',))
p.start()
p.join()
</pre>
<p>Here the function <tt class="docutils literal"><span class="pre">f</span></tt> is run in a child process.</p>
<p>For an explanation of why (on Windows) the <tt class="docutils literal"><span class="pre">if</span> <span class="pre">__name__</span> <span class="pre">==</span> <span class="pre">'__main__'</span></tt>
part is necessary see <a class="reference" href="programming-guidelines.html">Programming guidelines</a>.</p>
</div>
<div class="section">
<h1><a id="exchanging-objects-between-processes" name="exchanging-objects-between-processes">Exchanging objects between processes</a></h1>
<p><tt class="docutils literal"><span class="pre">processing</span></tt> supports two types of communication channel between
processes:</p>
<dl class="docutils">
<dt><strong>Queues</strong>:</dt>
<dd><p class="first">The function <tt class="docutils literal"><span class="pre">Queue()</span></tt> returns a near clone of <tt class="docutils literal"><span class="pre">Queue.Queue</span></tt>
-- see the Python standard documentation. For example</p>
<pre class="literal-block">
from processing import Process, Queue
def f(q):
q.put([42, None, 'hello'])
if __name__ == '__main__':
q = Queue()
p = Process(target=f, args=(q,))
p.start()
print q.get() # prints "[42, None, 'hello']"
p.join()
</pre>
<p class="last">Queues are thread and process safe. See <a class="reference" href="processing-ref.html#pipes-and-queues">Queues</a>.</p>
</dd>
<dt><strong>Pipes</strong>:</dt>
<dd><p class="first">The <tt class="docutils literal"><span class="pre">Pipe()</span></tt> function returns a pair of connection objects
connected by a pipe which by default is duplex (two-way). For
example</p>
<pre class="literal-block">
from processing import Process, Pipe
def f(conn):
conn.send([42, None, 'hello'])
conn.close()
if __name__ == '__main__':
parent_conn, child_conn = Pipe()
p = Process(target=f, args=(child_conn,))
p.start()
print parent_conn.recv() # prints "[42, None, 'hello']"
p.join()
</pre>
<p class="last">The two connection objects returned by <tt class="docutils literal"><span class="pre">Pipe()</span></tt> represent the two
ends of the pipe. Each connection object has <tt class="docutils literal"><span class="pre">send()</span></tt> and
<tt class="docutils literal"><span class="pre">recv()</span></tt> methods (among others). Note that data in a pipe may
become corrupted if two processes (or threads) try to read from or
write to the <em>same</em> end of the pipe at the same time. Of course
there is no risk of corruption from processes using different ends
of the pipe at the same time. See <a class="reference" href="processing-ref.html#pipes-and-queues">Pipes</a>.</p>
</dd>
</dl>
</div>
<div class="section">
<h1><a id="synchronization-between-processes" name="synchronization-between-processes">Synchronization between processes</a></h1>
<p><tt class="docutils literal"><span class="pre">processing</span></tt> contains equivalents of all the synchronization
primitives from <tt class="docutils literal"><span class="pre">threading</span></tt>. For instance one can use a lock to
ensure that only one process prints to standard output at a time:</p>
<pre class="literal-block">
from processing import Process, Lock
def f(l, i):
l.acquire()
print 'hello world', i
l.release()
if __name__ == '__main__':
lock = Lock()
for num in range(10):
Process(target=f, args=(lock, num)).start()
</pre>
<p>Without using the lock output from the different processes is liable
to get all mixed up.</p>
</div>
<div class="section">
<h1><a id="sharing-state-between-processes" name="sharing-state-between-processes">Sharing state between processes</a></h1>
<p>As mentioned above, when doing concurrent programming it is usually
best to avoid using shared state as far as possible. This is
particularly true when using multiple processes.</p>
<p>However, if you really do need to use some shared data then
<tt class="docutils literal"><span class="pre">processing</span></tt> provides a couple of ways of doing so.</p>
<dl class="docutils">
<dt><strong>Shared memory</strong>:</dt>
<dd><p class="first">Data can be stored in a shared memory map using <tt class="docutils literal"><span class="pre">Value</span></tt> or <tt class="docutils literal"><span class="pre">Array</span></tt>.
For example the following code</p>
<pre class="literal-block">
from processing import Process, Value, Array
def f(n, a):
n.value = 3.1415927
for i in range(len(a)):
a[i] = -a[i]
if __name__ == '__main__':
num = Value('d', 0.0)
arr = Array('i', range(10))
p = Process(target=f, args=(num, arr))
p.start()
p.join()
print num.value
print arr[:]
</pre>
<p>will print</p>
<pre class="literal-block">
3.1415927
[0, -1, -2, -3, -4, -5, -6, -7, -8, -9]
</pre>
<p>The <tt class="docutils literal"><span class="pre">'d'</span></tt> and <tt class="docutils literal"><span class="pre">'i'</span></tt> arguments used when creating <tt class="docutils literal"><span class="pre">num</span></tt> and <tt class="docutils literal"><span class="pre">arr</span></tt>
are typecodes of the kind used by the <tt class="docutils literal"><span class="pre">array</span></tt> module: <tt class="docutils literal"><span class="pre">'d'</span></tt>
indicates a double precision float and <tt class="docutils literal"><span class="pre">'i'</span></tt> inidicates a signed
integer. These shared objects will be process and thread safe.</p>
<p class="last">For more flexibility in using shared memory one can use the
<tt class="docutils literal"><span class="pre">processing.sharedctypes</span></tt> module which supports the creation of
arbitrary <a class="reference" href="sharedctypes.html">ctypes objects allocated from shared memory</a>.</p>
</dd>
<dt><strong>Server process</strong>:</dt>
<dd><p class="first">A manager object returned by <tt class="docutils literal"><span class="pre">Manager()</span></tt> controls a server process
which holds python objects and allows other processes to manipulate
them using proxies.</p>
<p>A manager returned by <tt class="docutils literal"><span class="pre">Manager()</span></tt> will support types <tt class="docutils literal"><span class="pre">list</span></tt>,
<tt class="docutils literal"><span class="pre">dict</span></tt>, <tt class="docutils literal"><span class="pre">Namespace</span></tt>, <tt class="docutils literal"><span class="pre">Lock</span></tt>, <tt class="docutils literal"><span class="pre">RLock</span></tt>, <tt class="docutils literal"><span class="pre">Semaphore</span></tt>,
<tt class="docutils literal"><span class="pre">BoundedSemaphore</span></tt>, <tt class="docutils literal"><span class="pre">Condition</span></tt>, <tt class="docutils literal"><span class="pre">Event</span></tt>, <tt class="docutils literal"><span class="pre">Queue</span></tt>, <tt class="docutils literal"><span class="pre">Value</span></tt>
and <tt class="docutils literal"><span class="pre">Array</span></tt>. For example:</p>
<pre class="literal-block">
from processing import Process, Manager
def f(d, l):
d[1] = '1'
d['2'] = 2
d[0.25] = None
l.reverse()
if __name__ == '__main__':
manager = Manager()
d = manager.dict()
l = manager.list(range(10))
p = Process(target=f, args=(d, l))
p.start()
p.join()
print d
print l
</pre>
<p>will print</p>
<pre class="literal-block">
{0.25: None, 1: '1', '2': 2}
[9, 8, 7, 6, 5, 4, 3, 2, 1, 0]
</pre>
<p>Creating managers which support other types is not hard --- see
<a class="reference" href="manager-objects.html#customized-managers">Customized managers</a>.</p>
<p class="last">Server process managers are more flexible than using shared memory
objects because they can be made to support arbitrary object types.
Also, a single manager can be shared by processes on different
computers over a network. They are, however, slower than using
shared memory. See <a class="reference" href="manager-objects.html#server-process-managers">Server process managers</a>.</p>
</dd>
</dl>
</div>
<div class="section">
<h1><a id="using-a-pool-of-workers" name="using-a-pool-of-workers">Using a pool of workers</a></h1>
<p>The <tt class="docutils literal"><span class="pre">Pool()</span></tt> function returns an object representing a pool of worker
processes. It has methods which allows tasks to be offloaded to the
worker processes in a few different ways.</p>
<p>For example:</p>
<pre class="literal-block">
from processing import Pool
def f(x):
return x*x
if __name__ == '__main__':
pool = Pool(processes=4) # start 4 worker processes
result = pool.applyAsync(f, [10]) # evaluate "f(10)" asynchronously
print result.get(timeout=1) # prints "100" unless your computer is *very* slow
print pool.map(f, range(10)) # prints "[0, 1, 4,..., 81]"
</pre>
<p>See <a class="reference" href="pool-objects.html">Process pools</a>.</p>
</div>
<div class="section">
<h1><a id="speed" name="speed">Speed</a></h1>
<p>The following benchmarks were performed on a single core Pentium 4,
2.5Ghz laptop running Windows XP and Ubuntu Linux 6.10 --- see
<a class="reference" href="../examples/benchmarks.py">benchmarks.py</a>.</p>
<p><em>Number of 256 byte string objects passed between processes/threads per sec</em>:</p>
<table border="1" class="docutils">
<colgroup>
<col width="55%" />
<col width="16%" />
<col width="29%" />
</colgroup>
<thead valign="bottom">
<tr><th class="head">Connection type</th>
<th class="head">Windows</th>
<th class="head">Linux</th>
</tr>
</thead>
<tbody valign="top">
<tr><td>Queue.Queue</td>
<td>49,000</td>
<td>17,000-50,000 <a class="footnote-reference" href="#id2" id="id1" name="id1">[1]</a></td>
</tr>
<tr><td>processing.Queue</td>
<td>22,000</td>
<td>21,000</td>
</tr>
<tr><td>Queue managed by server</td>
<td>6,900</td>
<td>6,500</td>
</tr>
<tr><td>processing.Pipe</td>
<td>52,000</td>
<td>57,000</td>
</tr>
</tbody>
</table>
<table class="docutils footnote" frame="void" id="id2" rules="none">
<colgroup><col class="label" /><col /></colgroup>
<tbody valign="top">
<tr><td class="label"><a class="fn-backref" href="#id1" name="id2">[1]</a></td><td>For some reason the performance of <tt class="docutils literal"><span class="pre">Queue.Queue</span></tt> is very
variable on Linux.</td></tr>
</tbody>
</table>
<p><em>Number of acquires/releases of a lock per sec</em>:</p>
<table border="1" class="docutils">
<colgroup>
<col width="60%" />
<col width="20%" />
<col width="20%" />
</colgroup>
<thead valign="bottom">
<tr><th class="head">Lock type</th>
<th class="head">Windows</th>
<th class="head">Linux</th>
</tr>
</thead>
<tbody valign="top">
<tr><td>threading.Lock</td>
<td>850,000</td>
<td>560,000</td>
</tr>
<tr><td>processing.Lock</td>
<td>420,000</td>
<td>510,000</td>
</tr>
<tr><td>Lock managed by server</td>
<td>10,000</td>
<td>8,400</td>
</tr>
<tr><td>threading.RLock</td>
<td>93,000</td>
<td>76,000</td>
</tr>
<tr><td>processing.RLock</td>
<td>420,000</td>
<td>500,000</td>
</tr>
<tr><td>RLock managed by server</td>
<td>8,800</td>
<td>7,400</td>
</tr>
</tbody>
</table>
<p><em>Number of interleaved waits/notifies per sec on a
condition variable by two processes</em>:</p>
<table border="1" class="docutils">
<colgroup>
<col width="60%" />
<col width="20%" />
<col width="20%" />
</colgroup>
<thead valign="bottom">
<tr><th class="head">Condition type</th>
<th class="head">Windows</th>
<th class="head">Linux</th>
</tr>
</thead>
<tbody valign="top">
<tr><td>threading.Condition</td>
<td>27,000</td>
<td>31,000</td>
</tr>
<tr><td>processing.Condition</td>
<td>26,000</td>
<td>25,000</td>
</tr>
<tr><td>Condition managed by server</td>
<td>6,600</td>
<td>6,000</td>
</tr>
</tbody>
</table>
<p><em>Number of integers retrieved from a sequence per sec</em>:</p>
<table border="1" class="docutils">
<colgroup>
<col width="60%" />
<col width="20%" />
<col width="20%" />
</colgroup>
<thead valign="bottom">
<tr><th class="head">Sequence type</th>
<th class="head">Windows</th>
<th class="head">Linux</th>
</tr>
</thead>
<tbody valign="top">
<tr><td>list</td>
<td>6,400,000</td>
<td>5,100,000</td>
</tr>
<tr><td>unsynchornized shared array</td>
<td>3,900,000</td>
<td>3,100,000</td>
</tr>
<tr><td>synchronized shared array</td>
<td>200,000</td>
<td>220,000</td>
</tr>
<tr><td>list managed by server</td>
<td>20,000</td>
<td>17,000</td>
</tr>
</tbody>
</table>
</div>
</div>
<div class="footer">
<hr class="footer" />
<a class="reference" href="index.html">Prev</a> <a class="reference" href="index.html">Up</a> <a class="reference" href="processing-ref.html">Next</a>
</div>
</body>
</html>
|
