Commit | Line | Data |
2605996a |
1 | =head1 NAME |
2 | |
3 | perlthrtut - tutorial on threads in Perl |
4 | |
5 | =head1 DESCRIPTION |
6 | |
53d7eaa8 |
7 | B<NOTE>: this tutorial describes the new Perl threading flavour |
8 | introduced in Perl 5.6.0 called interpreter threads, or ithreads |
bfce6503 |
9 | for short. There is another older Perl threading flavour called |
53d7eaa8 |
10 | the 5.005 model, unsurprisingly for 5.005 versions of Perl. |
bfce6503 |
11 | The old model is deprecated, and will probably be removed around release |
12 | 5.10. You are strongly encouraged to migrate any existing 5.005 threads |
13 | code to the new model as soon as possible. |
2a4bf773 |
14 | |
53d7eaa8 |
15 | You can see which (or neither) threading flavour you have by |
6eded8f3 |
16 | running C<perl -V> and looking at the C<Platform> section. |
53d7eaa8 |
17 | If you have C<useithreads=define> you have ithreads, if you |
18 | have C<use5005threads=define> you have 5.005 threads. |
19 | If you have neither, you don't have any thread support built in. |
20 | If you have both, you are in trouble. |
2605996a |
21 | |
bfce6503 |
22 | The user-level interface to the 5.005 threads was via the L<Threads> |
23 | class, while ithreads uses the L<threads> class. Note the change in case. |
24 | |
25 | =head1 Status |
26 | |
27 | The ithreads code has been available since Perl 5.6.0, and is considered |
28 | stable. The user-level interface to ithreads (the L<threads> classes) |
29 | appeared in the 5.8.0 release, and as of this time is considered stable, |
30 | although as with all new features, should be treated with caution. |
2605996a |
31 | |
c975c451 |
32 | =head1 What Is A Thread Anyway? |
33 | |
34 | A thread is a flow of control through a program with a single |
35 | execution point. |
36 | |
37 | Sounds an awful lot like a process, doesn't it? Well, it should. |
38 | Threads are one of the pieces of a process. Every process has at least |
39 | one thread and, up until now, every process running Perl had only one |
40 | thread. With 5.8, though, you can create extra threads. We're going |
41 | to show you how, when, and why. |
42 | |
43 | =head1 Threaded Program Models |
44 | |
45 | There are three basic ways that you can structure a threaded |
46 | program. Which model you choose depends on what you need your program |
47 | to do. For many non-trivial threaded programs you'll need to choose |
48 | different models for different pieces of your program. |
49 | |
50 | =head2 Boss/Worker |
51 | |
52 | The boss/worker model usually has one `boss' thread and one or more |
53 | `worker' threads. The boss thread gathers or generates tasks that need |
54 | to be done, then parcels those tasks out to the appropriate worker |
55 | thread. |
56 | |
57 | This model is common in GUI and server programs, where a main thread |
58 | waits for some event and then passes that event to the appropriate |
59 | worker threads for processing. Once the event has been passed on, the |
60 | boss thread goes back to waiting for another event. |
61 | |
62 | The boss thread does relatively little work. While tasks aren't |
63 | necessarily performed faster than with any other method, it tends to |
64 | have the best user-response times. |
65 | |
66 | =head2 Work Crew |
67 | |
68 | In the work crew model, several threads are created that do |
69 | essentially the same thing to different pieces of data. It closely |
70 | mirrors classical parallel processing and vector processors, where a |
71 | large array of processors do the exact same thing to many pieces of |
72 | data. |
73 | |
74 | This model is particularly useful if the system running the program |
75 | will distribute multiple threads across different processors. It can |
76 | also be useful in ray tracing or rendering engines, where the |
77 | individual threads can pass on interim results to give the user visual |
78 | feedback. |
79 | |
80 | =head2 Pipeline |
81 | |
82 | The pipeline model divides up a task into a series of steps, and |
83 | passes the results of one step on to the thread processing the |
84 | next. Each thread does one thing to each piece of data and passes the |
85 | results to the next thread in line. |
86 | |
87 | This model makes the most sense if you have multiple processors so two |
88 | or more threads will be executing in parallel, though it can often |
89 | make sense in other contexts as well. It tends to keep the individual |
90 | tasks small and simple, as well as allowing some parts of the pipeline |
91 | to block (on I/O or system calls, for example) while other parts keep |
92 | going. If you're running different parts of the pipeline on different |
93 | processors you may also take advantage of the caches on each |
94 | processor. |
95 | |
96 | This model is also handy for a form of recursive programming where, |
97 | rather than having a subroutine call itself, it instead creates |
98 | another thread. Prime and Fibonacci generators both map well to this |
99 | form of the pipeline model. (A version of a prime number generator is |
100 | presented later on.) |
101 | |
102 | =head1 Native threads |
103 | |
104 | There are several different ways to implement threads on a system. How |
105 | threads are implemented depends both on the vendor and, in some cases, |
106 | the version of the operating system. Often the first implementation |
107 | will be relatively simple, but later versions of the OS will be more |
108 | sophisticated. |
109 | |
110 | While the information in this section is useful, it's not necessary, |
111 | so you can skip it if you don't feel up to it. |
112 | |
6eded8f3 |
113 | There are three basic categories of threads: user-mode threads, kernel |
c975c451 |
114 | threads, and multiprocessor kernel threads. |
115 | |
116 | User-mode threads are threads that live entirely within a program and |
117 | its libraries. In this model, the OS knows nothing about threads. As |
118 | far as it's concerned, your process is just a process. |
119 | |
120 | This is the easiest way to implement threads, and the way most OSes |
121 | start. The big disadvantage is that, since the OS knows nothing about |
122 | threads, if one thread blocks they all do. Typical blocking activities |
123 | include most system calls, most I/O, and things like sleep(). |
124 | |
125 | Kernel threads are the next step in thread evolution. The OS knows |
126 | about kernel threads, and makes allowances for them. The main |
127 | difference between a kernel thread and a user-mode thread is |
128 | blocking. With kernel threads, things that block a single thread don't |
129 | block other threads. This is not the case with user-mode threads, |
130 | where the kernel blocks at the process level and not the thread level. |
131 | |
132 | This is a big step forward, and can give a threaded program quite a |
133 | performance boost over non-threaded programs. Threads that block |
134 | performing I/O, for example, won't block threads that are doing other |
135 | things. Each process still has only one thread running at once, |
136 | though, regardless of how many CPUs a system might have. |
137 | |
138 | Since kernel threading can interrupt a thread at any time, they will |
139 | uncover some of the implicit locking assumptions you may make in your |
140 | program. For example, something as simple as C<$a = $a + 2> can behave |
141 | unpredictably with kernel threads if $a is visible to other |
142 | threads, as another thread may have changed $a between the time it |
143 | was fetched on the right hand side and the time the new value is |
144 | stored. |
145 | |
6eded8f3 |
146 | Multiprocessor kernel threads are the final step in thread |
c975c451 |
147 | support. With multiprocessor kernel threads on a machine with multiple |
148 | CPUs, the OS may schedule two or more threads to run simultaneously on |
149 | different CPUs. |
150 | |
151 | This can give a serious performance boost to your threaded program, |
152 | since more than one thread will be executing at the same time. As a |
153 | tradeoff, though, any of those nagging synchronization issues that |
154 | might not have shown with basic kernel threads will appear with a |
155 | vengeance. |
156 | |
157 | In addition to the different levels of OS involvement in threads, |
158 | different OSes (and different thread implementations for a particular |
159 | OS) allocate CPU cycles to threads in different ways. |
160 | |
161 | Cooperative multitasking systems have running threads give up control |
162 | if one of two things happen. If a thread calls a yield function, it |
163 | gives up control. It also gives up control if the thread does |
164 | something that would cause it to block, such as perform I/O. In a |
165 | cooperative multitasking implementation, one thread can starve all the |
166 | others for CPU time if it so chooses. |
167 | |
168 | Preemptive multitasking systems interrupt threads at regular intervals |
169 | while the system decides which thread should run next. In a preemptive |
170 | multitasking system, one thread usually won't monopolize the CPU. |
171 | |
172 | On some systems, there can be cooperative and preemptive threads |
173 | running simultaneously. (Threads running with realtime priorities |
174 | often behave cooperatively, for example, while threads running at |
175 | normal priorities behave preemptively.) |
176 | |
bfce6503 |
177 | =head1 What kind of threads are Perl threads? |
c975c451 |
178 | |
179 | If you have experience with other thread implementations, you might |
180 | find that things aren't quite what you expect. It's very important to |
181 | remember when dealing with Perl threads that Perl Threads Are Not X |
182 | Threads, for all values of X. They aren't POSIX threads, or |
183 | DecThreads, or Java's Green threads, or Win32 threads. There are |
184 | similarities, and the broad concepts are the same, but if you start |
185 | looking for implementation details you're going to be either |
186 | disappointed or confused. Possibly both. |
187 | |
188 | This is not to say that Perl threads are completely different from |
189 | everything that's ever come before--they're not. Perl's threading |
190 | model owes a lot to other thread models, especially POSIX. Just as |
191 | Perl is not C, though, Perl threads are not POSIX threads. So if you |
192 | find yourself looking for mutexes, or thread priorities, it's time to |
193 | step back a bit and think about what you want to do and how Perl can |
194 | do it. |
195 | |
6eded8f3 |
196 | However it is important to remember that Perl threads cannot magically |
c975c451 |
197 | do things unless your operating systems threads allows it. So if your |
bfce6503 |
198 | system blocks the entire process on sleep(), Perl usually will as well. |
c975c451 |
199 | |
200 | =head1 Threadsafe Modules |
201 | |
202 | The addition of threads has changed Perl's internals |
203 | substantially. There are implications for people who write |
6eded8f3 |
204 | modules with XS code or external libraries. However, since the threads |
205 | do not share data, pure Perl modules that don't interact with external |
c975c451 |
206 | systems should be safe. Modules that are not tagged as thread-safe should |
207 | be tested or code reviewed before being used in production code. |
208 | |
209 | Not all modules that you might use are thread-safe, and you should |
210 | always assume a module is unsafe unless the documentation says |
211 | otherwise. This includes modules that are distributed as part of the |
212 | core. Threads are a new feature, and even some of the standard |
bfce6503 |
213 | modules aren't thread-safe. |
c975c451 |
214 | |
6eded8f3 |
215 | Even if a module is threadsafe, it doesn't mean that the module is optimized |
216 | to work well with threads. A module could possibly be rewritten to utilize |
217 | the new features in threaded Perl to increase performance in a threaded |
218 | environment. |
c975c451 |
219 | |
220 | If you're using a module that's not thread-safe for some reason, you |
221 | can protect yourself by using semaphores and lots of programming |
222 | discipline to control access to the module. Semaphores are covered |
223 | later in the article. Perl Threads Are Different |
224 | |
225 | =head1 Thread Basics |
226 | |
227 | The core L<threads> module provides the basic functions you need to write |
228 | threaded programs. In the following sections we'll cover the basics, |
229 | showing you what you need to do to create a threaded program. After |
230 | that, we'll go over some of the features of the L<threads> module that |
231 | make threaded programming easier. |
232 | |
233 | =head2 Basic Thread Support |
234 | |
6eded8f3 |
235 | Thread support is a Perl compile-time option - it's something that's |
c975c451 |
236 | turned on or off when Perl is built at your site, rather than when |
237 | your programs are compiled. If your Perl wasn't compiled with thread |
238 | support enabled, then any attempt to use threads will fail. |
239 | |
c975c451 |
240 | Your programs can use the Config module to check whether threads are |
241 | enabled. If your program can't run without them, you can say something |
242 | like: |
243 | |
244 | $Config{useithreads} or die "Recompile Perl with threads to run this program."; |
245 | |
246 | A possibly-threaded program using a possibly-threaded module might |
247 | have code like this: |
248 | |
249 | use Config; |
250 | use MyMod; |
251 | |
252 | if ($Config{useithreads}) { |
253 | # We have threads |
254 | require MyMod_threaded; |
255 | import MyMod_threaded; |
256 | } else { |
257 | require MyMod_unthreaded; |
258 | import MyMod_unthreaded; |
259 | } |
260 | |
261 | Since code that runs both with and without threads is usually pretty |
262 | messy, it's best to isolate the thread-specific code in its own |
263 | module. In our example above, that's what MyMod_threaded is, and it's |
264 | only imported if we're running on a threaded Perl. |
265 | |
266 | =head2 Creating Threads |
267 | |
268 | The L<threads> package provides the tools you need to create new |
269 | threads. Like any other module, you need to tell Perl you want to use |
270 | it; C<use threads> imports all the pieces you need to create basic |
271 | threads. |
272 | |
273 | The simplest, straightforward way to create a thread is with new(): |
274 | |
275 | use threads; |
276 | |
277 | $thr = threads->new(\&sub1); |
278 | |
279 | sub sub1 { |
280 | print "In the thread\n"; |
281 | } |
282 | |
283 | The new() method takes a reference to a subroutine and creates a new |
284 | thread, which starts executing in the referenced subroutine. Control |
285 | then passes both to the subroutine and the caller. |
286 | |
287 | If you need to, your program can pass parameters to the subroutine as |
288 | part of the thread startup. Just include the list of parameters as |
289 | part of the C<threads::new> call, like this: |
290 | |
291 | use threads; |
bfce6503 |
292 | |
c975c451 |
293 | $Param3 = "foo"; |
294 | $thr = threads->new(\&sub1, "Param 1", "Param 2", $Param3); |
295 | $thr = threads->new(\&sub1, @ParamList); |
296 | $thr = threads->new(\&sub1, qw(Param1 Param2 $Param3)); |
297 | |
298 | sub sub1 { |
299 | my @InboundParameters = @_; |
300 | print "In the thread\n"; |
301 | print "got parameters >", join("<>", @InboundParameters), "<\n"; |
302 | } |
303 | |
304 | |
305 | The last example illustrates another feature of threads. You can spawn |
306 | off several threads using the same subroutine. Each thread executes |
307 | the same subroutine, but in a separate thread with a separate |
308 | environment and potentially separate arguments. |
309 | |
bfce6503 |
310 | C<create()> is a synonym for C<new()> |
311 | |
c975c451 |
312 | =head2 Giving up control |
313 | |
314 | There are times when you may find it useful to have a thread |
315 | explicitly give up the CPU to another thread. Your threading package |
316 | might not support preemptive multitasking for threads, for example, or |
317 | you may be doing something compute-intensive and want to make sure |
318 | that the user-interface thread gets called frequently. Regardless, |
319 | there are times that you might want a thread to give up the processor. |
320 | |
321 | Perl's threading package provides the yield() function that does |
322 | this. yield() is pretty straightforward, and works like this: |
323 | |
324 | use threads; |
325 | |
bfce6503 |
326 | sub loop { |
327 | my $thread = shift; |
328 | my $foo = 50; |
329 | while($foo--) { print "in thread $thread\n" } |
330 | threads->yield(); |
331 | $foo = 50; |
332 | while($foo--) { print "in thread $thread\n" } |
333 | } |
c975c451 |
334 | |
bfce6503 |
335 | my $thread1 = threads->new(\&loop, 'first'); |
336 | my $thread2 = threads->new(\&loop, 'second'); |
337 | my $thread3 = threads->new(\&loop, 'third'); |
c975c451 |
338 | |
339 | It is important to remember that yield() is only a hint to give up the CPU, |
340 | it depends on your hardware, OS and threading libraries what actually happens. |
341 | Therefore it is important to note that one should not build the scheduling of |
342 | the threads around yield() calls. It might work on your platform but it won't |
343 | work on another platform. |
344 | |
345 | =head2 Waiting For A Thread To Exit |
346 | |
347 | Since threads are also subroutines, they can return values. To wait |
6eded8f3 |
348 | for a thread to exit and extract any values it might return, you can |
349 | use the join() method: |
c975c451 |
350 | |
351 | use threads; |
bfce6503 |
352 | |
c975c451 |
353 | $thr = threads->new(\&sub1); |
354 | |
355 | @ReturnData = $thr->join; |
356 | print "Thread returned @ReturnData"; |
357 | |
358 | sub sub1 { return "Fifty-six", "foo", 2; } |
359 | |
360 | In the example above, the join() method returns as soon as the thread |
361 | ends. In addition to waiting for a thread to finish and gathering up |
362 | any values that the thread might have returned, join() also performs |
363 | any OS cleanup necessary for the thread. That cleanup might be |
364 | important, especially for long-running programs that spawn lots of |
365 | threads. If you don't want the return values and don't want to wait |
366 | for the thread to finish, you should call the detach() method |
bfce6503 |
367 | instead, as described next. |
c975c451 |
368 | |
369 | =head2 Ignoring A Thread |
370 | |
371 | join() does three things: it waits for a thread to exit, cleans up |
372 | after it, and returns any data the thread may have produced. But what |
373 | if you're not interested in the thread's return values, and you don't |
374 | really care when the thread finishes? All you want is for the thread |
375 | to get cleaned up after when it's done. |
376 | |
377 | In this case, you use the detach() method. Once a thread is detached, |
378 | it'll run until it's finished, then Perl will clean up after it |
379 | automatically. |
380 | |
381 | use threads; |
bfce6503 |
382 | |
6eded8f3 |
383 | $thr = threads->new(\&sub1); # Spawn the thread |
c975c451 |
384 | |
385 | $thr->detach; # Now we officially don't care any more |
386 | |
387 | sub sub1 { |
388 | $a = 0; |
389 | while (1) { |
390 | $a++; |
391 | print "\$a is $a\n"; |
392 | sleep 1; |
393 | } |
394 | } |
395 | |
396 | |
bfce6503 |
397 | Once a thread is detached, it may not be joined, and any return data |
398 | that it might have produced (if it was done and waiting for a join) is |
c975c451 |
399 | lost. |
400 | |
401 | =head1 Threads And Data |
402 | |
403 | Now that we've covered the basics of threads, it's time for our next |
404 | topic: data. Threading introduces a couple of complications to data |
405 | access that non-threaded programs never need to worry about. |
406 | |
407 | =head2 Shared And Unshared Data |
408 | |
bfce6503 |
409 | The biggest difference between Perl ithreads and the old 5.005 style |
410 | threading, or for that matter, to most other threading systems out there, |
411 | is that by default, no data is shared. When a new perl thread is created, |
412 | all the data associated with the current thread is copied to the new |
413 | thread, and is subsequently private to that new thread! |
414 | This is similar in feel to what happens when a UNIX process forks, |
415 | except that in this case, the data is just copied to a different part of |
416 | memory within the same process rather than a real fork taking place. |
c975c451 |
417 | |
418 | To make use of threading however, one usually want the threads to share |
bfce6503 |
419 | at least some data between themselves. This is done with the |
420 | L<threads::shared> module and the C< : shared> attribute: |
421 | |
422 | use threads; |
423 | use threads::shared; |
424 | |
425 | my $foo : shared = 1; |
426 | my $bar = 1; |
427 | threads->new(sub { $foo++; $bar++ })->join; |
428 | |
429 | print "$foo\n"; #prints 2 since $foo is shared |
430 | print "$bar\n"; #prints 1 since $bar is not shared |
431 | |
432 | In the case of a shared array, all the array's elements are shared, and for |
433 | a shared hash, all the keys and values are shared. This places |
434 | restrictions on what may be assigned to shared array and hash elements: only |
435 | simple values or references to shared variables are allowed - this is |
f3278b06 |
436 | so that a private variable can't accidentally become shared. A bad |
bfce6503 |
437 | assignment will cause the thread to die. For example: |
438 | |
439 | use threads; |
440 | use threads::shared; |
441 | |
442 | my $var = 1; |
443 | my $svar : shared = 2; |
444 | my %hash : shared; |
445 | |
446 | ... create some threads ... |
447 | |
448 | $hash{a} = 1; # all threads see exists($hash{a}) and $hash{a} == 1 |
449 | $hash{a} = $var # okay - copy-by-value: same affect as previous |
450 | $hash{a} = $svar # okay - copy-by-value: same affect as previous |
451 | $hash{a} = \$svar # okay - a reference to a shared variable |
452 | $hash{a} = \$var # This will die |
453 | delete $hash{a} # okay - all threads will see !exists($hash{a}) |
454 | |
455 | Note that a shared variable guarantees that if two or more threads try to |
456 | modify it at the same time, the internal state of the variable will not |
457 | become corrupted. However, there are no guarantees beyond this, as |
458 | explained in the next section. |
c975c451 |
459 | |
6eded8f3 |
460 | =head2 Thread Pitfalls: Races |
c975c451 |
461 | |
462 | While threads bring a new set of useful tools, they also bring a |
463 | number of pitfalls. One pitfall is the race condition: |
464 | |
465 | use threads; |
466 | use threads::shared; |
bfce6503 |
467 | |
c975c451 |
468 | my $a : shared = 1; |
469 | $thr1 = threads->new(\&sub1); |
470 | $thr2 = threads->new(\&sub2); |
471 | |
472 | $thr1->join; |
473 | $thr2->join; |
474 | print "$a\n"; |
475 | |
bfce6503 |
476 | sub sub1 { my $foo = $a; $a = $foo + 1; } |
477 | sub sub2 { my $bar = $a; $a = $bar + 1; } |
c975c451 |
478 | |
479 | What do you think $a will be? The answer, unfortunately, is "it |
480 | depends." Both sub1() and sub2() access the global variable $a, once |
481 | to read and once to write. Depending on factors ranging from your |
482 | thread implementation's scheduling algorithm to the phase of the moon, |
483 | $a can be 2 or 3. |
484 | |
485 | Race conditions are caused by unsynchronized access to shared |
486 | data. Without explicit synchronization, there's no way to be sure that |
487 | nothing has happened to the shared data between the time you access it |
488 | and the time you update it. Even this simple code fragment has the |
489 | possibility of error: |
490 | |
491 | use threads; |
492 | my $a : shared = 2; |
493 | my $b : shared; |
494 | my $c : shared; |
495 | my $thr1 = threads->create(sub { $b = $a; $a = $b + 1; }); |
496 | my $thr2 = threads->create(sub { $c = $a; $a = $c + 1; }); |
497 | $thr1->join(); |
498 | $thr2->join(); |
499 | |
500 | Two threads both access $a. Each thread can potentially be interrupted |
501 | at any point, or be executed in any order. At the end, $a could be 3 |
502 | or 4, and both $b and $c could be 2 or 3. |
503 | |
bfce6503 |
504 | Even C<$a += 5> or C<$a++> are not guaranteed to be atomic. |
505 | |
c975c451 |
506 | Whenever your program accesses data or resources that can be accessed |
507 | by other threads, you must take steps to coordinate access or risk |
bfce6503 |
508 | data inconsistency and race conditions. Note that Perl will protect its |
509 | internals from your race conditions, but it won't protect you from you. |
510 | |
f3278b06 |
511 | =head1 Synchronization and control |
bfce6503 |
512 | |
513 | Perl provides a number of mechanisms to coordinate the interactions |
514 | between themselves and their data, to avoid race conditions and the like. |
515 | Some of these are designed to resemble the common techniques used in thread |
516 | libraries such as C<pthreads>; others are Perl-specific. Often, the |
f3278b06 |
517 | standard techniques are clumsily and difficult to get right (such as |
bfce6503 |
518 | condition waits). Where possible, it is usually easier to use Perlish |
519 | techniques such as queues, which remove some of the hard work involved. |
c975c451 |
520 | |
521 | =head2 Controlling access: lock() |
522 | |
523 | The lock() function takes a shared variable and puts a lock on it. |
bfce6503 |
524 | No other thread may lock the variable until the the variable is unlocked |
525 | by the thread holding the lock. Unlocking happens automatically |
526 | when the locking thread exists the outermost block that contains |
527 | C<lock()> function. Using lock() is straightforward: this example has |
f3278b06 |
528 | several threads doing some calculations in parallel, and occasionally |
bfce6503 |
529 | updating a running total: |
530 | |
531 | use threads; |
532 | use threads::shared; |
533 | |
534 | my $total : shared = 0; |
535 | |
536 | sub calc { |
537 | for (;;) { |
538 | my $result; |
539 | # (... do some calculations and set $result ...) |
540 | { |
541 | lock($total); # block until we obtain the lock |
542 | $total += $result |
f3278b06 |
543 | } # lock implicitly released at end of scope |
bfce6503 |
544 | last if $result == 0; |
545 | } |
546 | } |
547 | |
548 | my $thr1 = threads->new(\&calc); |
549 | my $thr2 = threads->new(\&calc); |
550 | my $thr3 = threads->new(\&calc); |
551 | $thr1->join; |
552 | $thr2->join; |
553 | $thr3->join; |
554 | print "total=$total\n"; |
c975c451 |
555 | |
c975c451 |
556 | |
557 | lock() blocks the thread until the variable being locked is |
558 | available. When lock() returns, your thread can be sure that no other |
bfce6503 |
559 | thread can lock that variable until the outermost block containing the |
c975c451 |
560 | lock exits. |
561 | |
562 | It's important to note that locks don't prevent access to the variable |
563 | in question, only lock attempts. This is in keeping with Perl's |
564 | longstanding tradition of courteous programming, and the advisory file |
565 | locking that flock() gives you. |
566 | |
567 | You may lock arrays and hashes as well as scalars. Locking an array, |
568 | though, will not block subsequent locks on array elements, just lock |
569 | attempts on the array itself. |
570 | |
bfce6503 |
571 | Locks are recursive, which means it's okay for a thread to |
c975c451 |
572 | lock a variable more than once. The lock will last until the outermost |
bfce6503 |
573 | lock() on the variable goes out of scope. For example: |
574 | |
575 | my $x : shared; |
576 | doit(); |
577 | |
578 | sub doit { |
579 | { |
580 | { |
581 | lock($x); # wait for lock |
582 | lock($x): # NOOP - we already have the lock |
583 | { |
584 | lock($x); # NOOP |
585 | { |
586 | lock($x); # NOOP |
587 | lockit_some_more(); |
588 | } |
589 | } |
590 | } # *** implicit unlock here *** |
591 | } |
592 | } |
593 | |
594 | sub lockit_some_more { |
595 | lock($x); # NOOP |
596 | } # nothing happens here |
597 | |
598 | Note that there is no unlock() function - the only way to unlock a |
599 | variable is to allow it to go out of scope. |
600 | |
601 | A lock can either be used to guard the data contained within the variable |
602 | being locked, or it can be used to guard something else, like a section |
603 | of code. In this latter case, the variable in question does not hold any |
604 | useful data, and exists only for the purpose of being locked. In this |
605 | respect, the variable behaves like the mutexes and basic semaphores of |
606 | traditional thread libraries. |
c975c451 |
607 | |
bfce6503 |
608 | =head2 A Thread Pitfall: Deadlocks |
c975c451 |
609 | |
bfce6503 |
610 | Locks are a handy tool to synchronize access to data, and using them |
c975c451 |
611 | properly is the key to safe shared data. Unfortunately, locks aren't |
f3278b06 |
612 | without their dangers, especially when multiple locks are involved. |
bfce6503 |
613 | Consider the following code: |
c975c451 |
614 | |
615 | use threads; |
bfce6503 |
616 | |
c975c451 |
617 | my $a : shared = 4; |
618 | my $b : shared = "foo"; |
619 | my $thr1 = threads->new(sub { |
620 | lock($a); |
bfce6503 |
621 | threads->yield; |
c975c451 |
622 | sleep 20; |
bfce6503 |
623 | lock($b); |
c975c451 |
624 | }); |
625 | my $thr2 = threads->new(sub { |
626 | lock($b); |
bfce6503 |
627 | threads->yield; |
c975c451 |
628 | sleep 20; |
bfce6503 |
629 | lock($a); |
c975c451 |
630 | }); |
631 | |
632 | This program will probably hang until you kill it. The only way it |
bfce6503 |
633 | won't hang is if one of the two threads acquires both locks |
c975c451 |
634 | first. A guaranteed-to-hang version is more complicated, but the |
635 | principle is the same. |
636 | |
bfce6503 |
637 | The first thread will grab a lock on $a, then, after a pause during which |
638 | the second thread has probably had time to do some work, try to grab a |
639 | lock on $b. Meanwhile, the second thread grabs a lock on $b, then later |
640 | tries to grab a lock on $a. The second lock attempt for both threads will |
641 | block, each waiting for the other to release its lock. |
c975c451 |
642 | |
643 | This condition is called a deadlock, and it occurs whenever two or |
644 | more threads are trying to get locks on resources that the others |
645 | own. Each thread will block, waiting for the other to release a lock |
646 | on a resource. That never happens, though, since the thread with the |
647 | resource is itself waiting for a lock to be released. |
648 | |
649 | There are a number of ways to handle this sort of problem. The best |
650 | way is to always have all threads acquire locks in the exact same |
651 | order. If, for example, you lock variables $a, $b, and $c, always lock |
652 | $a before $b, and $b before $c. It's also best to hold on to locks for |
653 | as short a period of time to minimize the risks of deadlock. |
654 | |
f3278b06 |
655 | The other syncronization primitives described below can suffer from |
bfce6503 |
656 | similar problems. |
657 | |
c975c451 |
658 | =head2 Queues: Passing Data Around |
659 | |
660 | A queue is a special thread-safe object that lets you put data in one |
661 | end and take it out the other without having to worry about |
662 | synchronization issues. They're pretty straightforward, and look like |
663 | this: |
664 | |
665 | use threads; |
666 | use threads::shared::queue; |
667 | |
bfce6503 |
668 | my $DataQueue = threads::shared::queue->new(); |
c975c451 |
669 | $thr = threads->new(sub { |
670 | while ($DataElement = $DataQueue->dequeue) { |
671 | print "Popped $DataElement off the queue\n"; |
672 | } |
673 | }); |
674 | |
675 | $DataQueue->enqueue(12); |
676 | $DataQueue->enqueue("A", "B", "C"); |
677 | $DataQueue->enqueue(\$thr); |
678 | sleep 10; |
679 | $DataQueue->enqueue(undef); |
680 | $thr->join(); |
681 | |
6eded8f3 |
682 | You create the queue with C<new threads::shared::queue>. Then you can |
683 | add lists of scalars onto the end with enqueue(), and pop scalars off |
684 | the front of it with dequeue(). A queue has no fixed size, and can grow |
685 | as needed to hold everything pushed on to it. |
c975c451 |
686 | |
687 | If a queue is empty, dequeue() blocks until another thread enqueues |
688 | something. This makes queues ideal for event loops and other |
689 | communications between threads. |
690 | |
c975c451 |
691 | =head2 Semaphores: Synchronizing Data Access |
692 | |
bfce6503 |
693 | Semaphores are a kind of generic locking mechanism. In their most basic |
694 | form, they behave very much like lockable scalars, except that thay |
695 | can't hold data, and that they must be explicitly unlocked. In their |
696 | advanced form, they act like a kind of counter, and can allow multiple |
697 | threads to have the 'lock' at any one time. |
2605996a |
698 | |
bfce6503 |
699 | =head2 Basic semaphores |
2605996a |
700 | |
bfce6503 |
701 | Semaphores have two methods, down() and up(): down() decrements the resource |
702 | count, while up increments it. Calls to down() will block if the |
c975c451 |
703 | semaphore's current count would decrement below zero. This program |
704 | gives a quick demonstration: |
705 | |
706 | use threads qw(yield); |
707 | use threads::shared::semaphore; |
bfce6503 |
708 | |
c975c451 |
709 | my $semaphore = new threads::shared::semaphore; |
bfce6503 |
710 | my $GlobalVariable : shared = 0; |
2605996a |
711 | |
c975c451 |
712 | $thr1 = new threads \&sample_sub, 1; |
713 | $thr2 = new threads \&sample_sub, 2; |
714 | $thr3 = new threads \&sample_sub, 3; |
2605996a |
715 | |
c975c451 |
716 | sub sample_sub { |
717 | my $SubNumber = shift @_; |
718 | my $TryCount = 10; |
719 | my $LocalCopy; |
720 | sleep 1; |
721 | while ($TryCount--) { |
722 | $semaphore->down; |
723 | $LocalCopy = $GlobalVariable; |
724 | print "$TryCount tries left for sub $SubNumber (\$GlobalVariable is $GlobalVariable)\n"; |
725 | yield; |
726 | sleep 2; |
727 | $LocalCopy++; |
728 | $GlobalVariable = $LocalCopy; |
729 | $semaphore->up; |
730 | } |
731 | } |
6eded8f3 |
732 | |
c975c451 |
733 | $thr1->join(); |
734 | $thr2->join(); |
735 | $thr3->join(); |
2605996a |
736 | |
c975c451 |
737 | The three invocations of the subroutine all operate in sync. The |
738 | semaphore, though, makes sure that only one thread is accessing the |
739 | global variable at once. |
2605996a |
740 | |
bfce6503 |
741 | =head2 Advanced Semaphores |
2605996a |
742 | |
c975c451 |
743 | By default, semaphores behave like locks, letting only one thread |
744 | down() them at a time. However, there are other uses for semaphores. |
2605996a |
745 | |
6eded8f3 |
746 | Each semaphore has a counter attached to it. By default, semaphores are |
747 | created with the counter set to one, down() decrements the counter by |
748 | one, and up() increments by one. However, we can override any or all |
749 | of these defaults simply by passing in different values: |
750 | |
751 | use threads; |
752 | use threads::shared::semaphore; |
753 | my $semaphore = threads::shared::semaphore->new(5); |
754 | # Creates a semaphore with the counter set to five |
755 | |
756 | $thr1 = threads->new(\&sub1); |
757 | $thr2 = threads->new(\&sub1); |
758 | |
759 | sub sub1 { |
760 | $semaphore->down(5); # Decrements the counter by five |
761 | # Do stuff here |
762 | $semaphore->up(5); # Increment the counter by five |
763 | } |
764 | |
765 | $thr1->detach(); |
766 | $thr2->detach(); |
767 | |
768 | If down() attempts to decrement the counter below zero, it blocks until |
769 | the counter is large enough. Note that while a semaphore can be created |
770 | with a starting count of zero, any up() or down() always changes the |
771 | counter by at least one, and so $semaphore->down(0) is the same as |
772 | $semaphore->down(1). |
2605996a |
773 | |
c975c451 |
774 | The question, of course, is why would you do something like this? Why |
775 | create a semaphore with a starting count that's not one, or why |
776 | decrement/increment it by more than one? The answer is resource |
777 | availability. Many resources that you want to manage access for can be |
778 | safely used by more than one thread at once. |
2605996a |
779 | |
c975c451 |
780 | For example, let's take a GUI driven program. It has a semaphore that |
781 | it uses to synchronize access to the display, so only one thread is |
782 | ever drawing at once. Handy, but of course you don't want any thread |
783 | to start drawing until things are properly set up. In this case, you |
784 | can create a semaphore with a counter set to zero, and up it when |
785 | things are ready for drawing. |
2605996a |
786 | |
c975c451 |
787 | Semaphores with counters greater than one are also useful for |
788 | establishing quotas. Say, for example, that you have a number of |
789 | threads that can do I/O at once. You don't want all the threads |
790 | reading or writing at once though, since that can potentially swamp |
791 | your I/O channels, or deplete your process' quota of filehandles. You |
792 | can use a semaphore initialized to the number of concurrent I/O |
793 | requests (or open files) that you want at any one time, and have your |
794 | threads quietly block and unblock themselves. |
2605996a |
795 | |
c975c451 |
796 | Larger increments or decrements are handy in those cases where a |
797 | thread needs to check out or return a number of resources at once. |
2605996a |
798 | |
bfce6503 |
799 | =head2 cond_wait() and cond_signal() |
800 | |
801 | These two functions can be used in conjunction with locks to notify |
802 | co-operating threads that a resource has become available. They are |
803 | very similar in use to the functions found in C<pthreads>. However |
804 | for most purposes, queues are simpler to use and more intuitive. See |
805 | L<threads::shared> for more details. |
2605996a |
806 | |
c975c451 |
807 | =head1 General Thread Utility Routines |
808 | |
809 | We've covered the workhorse parts of Perl's threading package, and |
810 | with these tools you should be well on your way to writing threaded |
811 | code and packages. There are a few useful little pieces that didn't |
812 | really fit in anyplace else. |
813 | |
814 | =head2 What Thread Am I In? |
815 | |
bfce6503 |
816 | The C<< threads->self >> class method provides your program with a way to |
817 | get an object representing the thread it's currently in. You can use this |
6eded8f3 |
818 | object in the same way as the ones returned from thread creation. |
c975c451 |
819 | |
820 | =head2 Thread IDs |
821 | |
822 | tid() is a thread object method that returns the thread ID of the |
823 | thread the object represents. Thread IDs are integers, with the main |
824 | thread in a program being 0. Currently Perl assigns a unique tid to |
825 | every thread ever created in your program, assigning the first thread |
826 | to be created a tid of 1, and increasing the tid by 1 for each new |
827 | thread that's created. |
828 | |
829 | =head2 Are These Threads The Same? |
830 | |
831 | The equal() method takes two thread objects and returns true |
832 | if the objects represent the same thread, and false if they don't. |
833 | |
834 | Thread objects also have an overloaded == comparison so that you can do |
835 | comparison on them as you would with normal objects. |
836 | |
837 | =head2 What Threads Are Running? |
838 | |
bfce6503 |
839 | C<< threads->list >> returns a list of thread objects, one for each thread |
c975c451 |
840 | that's currently running and not detached. Handy for a number of things, |
841 | including cleaning up at the end of your program: |
842 | |
843 | # Loop through all the threads |
844 | foreach $thr (threads->list) { |
845 | # Don't join the main thread or ourselves |
846 | if ($thr->tid && !threads::equal($thr, threads->self)) { |
847 | $thr->join; |
848 | } |
849 | } |
850 | |
bfce6503 |
851 | If some threads have not finished running when the main Perl thread |
852 | ends, Perl will warn you about it and die, since it is impossible for Perl |
6eded8f3 |
853 | to clean up itself while other threads are running |
c975c451 |
854 | |
855 | =head1 A Complete Example |
856 | |
857 | Confused yet? It's time for an example program to show some of the |
858 | things we've covered. This program finds prime numbers using threads. |
859 | |
860 | 1 #!/usr/bin/perl -w |
861 | 2 # prime-pthread, courtesy of Tom Christiansen |
862 | 3 |
863 | 4 use strict; |
864 | 5 |
865 | 6 use threads; |
866 | 7 use threads::shared::queue; |
867 | 8 |
868 | 9 my $stream = new threads::shared::queue; |
869 | 10 my $kid = new threads(\&check_num, $stream, 2); |
870 | 11 |
871 | 12 for my $i ( 3 .. 1000 ) { |
872 | 13 $stream->enqueue($i); |
873 | 14 } |
874 | 15 |
875 | 16 $stream->enqueue(undef); |
876 | 17 $kid->join(); |
877 | 18 |
878 | 19 sub check_num { |
879 | 20 my ($upstream, $cur_prime) = @_; |
880 | 21 my $kid; |
881 | 22 my $downstream = new threads::shared::queue; |
882 | 23 while (my $num = $upstream->dequeue) { |
883 | 24 next unless $num % $cur_prime; |
884 | 25 if ($kid) { |
885 | 26 $downstream->enqueue($num); |
886 | 27 } else { |
887 | 28 print "Found prime $num\n"; |
888 | 29 $kid = new threads(\&check_num, $downstream, $num); |
889 | 30 } |
890 | 31 } |
891 | 32 $downstream->enqueue(undef) if $kid; |
892 | 33 $kid->join() if $kid; |
893 | 34 } |
894 | |
895 | This program uses the pipeline model to generate prime numbers. Each |
896 | thread in the pipeline has an input queue that feeds numbers to be |
897 | checked, a prime number that it's responsible for, and an output queue |
6eded8f3 |
898 | that into which it funnels numbers that have failed the check. If the thread |
c975c451 |
899 | has a number that's failed its check and there's no child thread, then |
900 | the thread must have found a new prime number. In that case, a new |
901 | child thread is created for that prime and stuck on the end of the |
902 | pipeline. |
903 | |
6eded8f3 |
904 | This probably sounds a bit more confusing than it really is, so let's |
c975c451 |
905 | go through this program piece by piece and see what it does. (For |
906 | those of you who might be trying to remember exactly what a prime |
907 | number is, it's a number that's only evenly divisible by itself and 1) |
908 | |
909 | The bulk of the work is done by the check_num() subroutine, which |
910 | takes a reference to its input queue and a prime number that it's |
911 | responsible for. After pulling in the input queue and the prime that |
912 | the subroutine's checking (line 20), we create a new queue (line 22) |
913 | and reserve a scalar for the thread that we're likely to create later |
914 | (line 21). |
915 | |
916 | The while loop from lines 23 to line 31 grabs a scalar off the input |
917 | queue and checks against the prime this thread is responsible |
918 | for. Line 24 checks to see if there's a remainder when we modulo the |
919 | number to be checked against our prime. If there is one, the number |
920 | must not be evenly divisible by our prime, so we need to either pass |
921 | it on to the next thread if we've created one (line 26) or create a |
922 | new thread if we haven't. |
923 | |
924 | The new thread creation is line 29. We pass on to it a reference to |
925 | the queue we've created, and the prime number we've found. |
926 | |
927 | Finally, once the loop terminates (because we got a 0 or undef in the |
928 | queue, which serves as a note to die), we pass on the notice to our |
6eded8f3 |
929 | child and wait for it to exit if we've created a child (lines 32 and |
c975c451 |
930 | 37). |
931 | |
932 | Meanwhile, back in the main thread, we create a queue (line 9) and the |
933 | initial child thread (line 10), and pre-seed it with the first prime: |
934 | 2. Then we queue all the numbers from 3 to 1000 for checking (lines |
935 | 12-14), then queue a die notice (line 16) and wait for the first child |
936 | thread to terminate (line 17). Because a child won't die until its |
937 | child has died, we know that we're done once we return from the join. |
938 | |
939 | That's how it works. It's pretty simple; as with many Perl programs, |
940 | the explanation is much longer than the program. |
941 | |
bfce6503 |
942 | =head1 Performance considerations |
943 | |
944 | The main thing to bear in mind when comparing ithreads to other threading |
945 | models is the fact that for each new thread created, a complete copy of |
946 | all the variables and data of the parent thread has to be taken. Thus |
947 | thread creation can be quite expensive, both in terms of memory usage and |
948 | time spent in creation. The ideal way to reduce these costs is to have a |
949 | relatively short number of long-lived threads, all created fairly early |
950 | on - before the base thread has accumulated too much data. Of course, this |
951 | may not always be possible, so compromises have to be made. However, after |
952 | a thread has been created, its performance and extra memory usage should |
953 | be little different than ordinary code. |
954 | |
955 | Also note that under the current implementation, shared variables |
956 | use a little more memory and are a little slower than ordinary variables. |
957 | |
c975c451 |
958 | =head1 Conclusion |
959 | |
960 | A complete thread tutorial could fill a book (and has, many times), |
6eded8f3 |
961 | but with what we've covered in this introduction, you should be well |
962 | on your way to becoming a threaded Perl expert. |
c975c451 |
963 | |
964 | =head1 Bibliography |
965 | |
966 | Here's a short bibliography courtesy of Jürgen Christoffel: |
967 | |
968 | =head2 Introductory Texts |
969 | |
970 | Birrell, Andrew D. An Introduction to Programming with |
971 | Threads. Digital Equipment Corporation, 1989, DEC-SRC Research Report |
972 | #35 online as |
6eded8f3 |
973 | http://gatekeeper.dec.com/pub/DEC/SRC/research-reports/abstracts/src-rr-035.html |
974 | (highly recommended) |
c975c451 |
975 | |
976 | Robbins, Kay. A., and Steven Robbins. Practical Unix Programming: A |
977 | Guide to Concurrency, Communication, and |
978 | Multithreading. Prentice-Hall, 1996. |
979 | |
980 | Lewis, Bill, and Daniel J. Berg. Multithreaded Programming with |
981 | Pthreads. Prentice Hall, 1997, ISBN 0-13-443698-9 (a well-written |
982 | introduction to threads). |
983 | |
984 | Nelson, Greg (editor). Systems Programming with Modula-3. Prentice |
985 | Hall, 1991, ISBN 0-13-590464-1. |
986 | |
987 | Nichols, Bradford, Dick Buttlar, and Jacqueline Proulx Farrell. |
988 | Pthreads Programming. O'Reilly & Associates, 1996, ISBN 156592-115-1 |
989 | (covers POSIX threads). |
990 | |
991 | =head2 OS-Related References |
992 | |
993 | Boykin, Joseph, David Kirschen, Alan Langerman, and Susan |
994 | LoVerso. Programming under Mach. Addison-Wesley, 1994, ISBN |
995 | 0-201-52739-1. |
996 | |
997 | Tanenbaum, Andrew S. Distributed Operating Systems. Prentice Hall, |
998 | 1995, ISBN 0-13-219908-4 (great textbook). |
999 | |
1000 | Silberschatz, Abraham, and Peter B. Galvin. Operating System Concepts, |
1001 | 4th ed. Addison-Wesley, 1995, ISBN 0-201-59292-4 |
1002 | |
1003 | =head2 Other References |
1004 | |
1005 | Arnold, Ken and James Gosling. The Java Programming Language, 2nd |
1006 | ed. Addison-Wesley, 1998, ISBN 0-201-31006-6. |
1007 | |
1008 | Le Sergent, T. and B. Berthomieu. "Incremental MultiThreaded Garbage |
1009 | Collection on Virtually Shared Memory Architectures" in Memory |
1010 | Management: Proc. of the International Workshop IWMM 92, St. Malo, |
1011 | France, September 1992, Yves Bekkers and Jacques Cohen, eds. Springer, |
1012 | 1992, ISBN 3540-55940-X (real-life thread applications). |
1013 | |
1014 | =head1 Acknowledgements |
1015 | |
1016 | Thanks (in no particular order) to Chaim Frenkel, Steve Fink, Gurusamy |
1017 | Sarathy, Ilya Zakharevich, Benjamin Sugars, Jürgen Christoffel, Joshua |
1018 | Pritikin, and Alan Burlison, for their help in reality-checking and |
1019 | polishing this article. Big thanks to Tom Christiansen for his rewrite |
1020 | of the prime number generator. |
1021 | |
1022 | =head1 AUTHOR |
1023 | |
1024 | Dan Sugalski E<lt>sugalskd@ous.eduE<gt> |
1025 | |
1026 | Slightly modified by Arthur Bergman to fit the new thread model/module. |
1027 | |
1028 | =head1 Copyrights |
1029 | |
bfce6503 |
1030 | The original version of this article originally appeared in The Perl |
1031 | Journal #10, and is copyright 1998 The Perl Journal. It appears courtesy |
1032 | of Jon Orwant and The Perl Journal. This document may be distributed |
1033 | under the same terms as Perl itself. |
2605996a |
1034 | |
53d7eaa8 |
1035 | For more information please see L<threads> and L<threads::shared>. |
2605996a |
1036 | |