Integrate perlio:
[p5sagit/p5-mst-13.2.git] / pod / perlothrtut.pod
CommitLineData
53d7eaa8 1=head1 NAME
2
3perlothrtut - old tutorial on threads in Perl
4
5=head1 DESCRIPTION
6
7B<WARNING>: Threading is an experimental feature. Both the interface
8and implementation are subject to change drastically. In fact, this
9documentation describes the flavor of threads that was in version
105.005. Perl 5.6.0 and later have the beginnings of support for
11interpreter threads, which (when finished) is expected to be
12significantly different from what is described here. The information
13contained here may therefore soon be obsolete. Use at your own risk!
14
15One of the most prominent new features of Perl 5.005 is the inclusion
16of threads. Threads make a number of things a lot easier, and are a
17very useful addition to your bag of programming tricks.
18
19=head1 What Is A Thread Anyway?
20
21A thread is a flow of control through a program with a single
22execution point.
23
24Sounds an awful lot like a process, doesn't it? Well, it should.
25Threads are one of the pieces of a process. Every process has at least
26one thread and, up until now, every process running Perl had only one
27thread. With 5.005, though, you can create extra threads. We're going
28to show you how, when, and why.
29
30=head1 Threaded Program Models
31
32There are three basic ways that you can structure a threaded
33program. Which model you choose depends on what you need your program
34to do. For many non-trivial threaded programs you'll need to choose
35different models for different pieces of your program.
36
37=head2 Boss/Worker
38
39The boss/worker model usually has one `boss' thread and one or more
40`worker' threads. The boss thread gathers or generates tasks that need
41to be done, then parcels those tasks out to the appropriate worker
42thread.
43
44This model is common in GUI and server programs, where a main thread
45waits for some event and then passes that event to the appropriate
46worker threads for processing. Once the event has been passed on, the
47boss thread goes back to waiting for another event.
48
49The boss thread does relatively little work. While tasks aren't
50necessarily performed faster than with any other method, it tends to
51have the best user-response times.
52
53=head2 Work Crew
54
55In the work crew model, several threads are created that do
56essentially the same thing to different pieces of data. It closely
57mirrors classical parallel processing and vector processors, where a
58large array of processors do the exact same thing to many pieces of
59data.
60
61This model is particularly useful if the system running the program
62will distribute multiple threads across different processors. It can
63also be useful in ray tracing or rendering engines, where the
64individual threads can pass on interim results to give the user visual
65feedback.
66
67=head2 Pipeline
68
69The pipeline model divides up a task into a series of steps, and
70passes the results of one step on to the thread processing the
71next. Each thread does one thing to each piece of data and passes the
72results to the next thread in line.
73
74This model makes the most sense if you have multiple processors so two
75or more threads will be executing in parallel, though it can often
76make sense in other contexts as well. It tends to keep the individual
77tasks small and simple, as well as allowing some parts of the pipeline
78to block (on I/O or system calls, for example) while other parts keep
79going. If you're running different parts of the pipeline on different
80processors you may also take advantage of the caches on each
81processor.
82
83This model is also handy for a form of recursive programming where,
84rather than having a subroutine call itself, it instead creates
85another thread. Prime and Fibonacci generators both map well to this
86form of the pipeline model. (A version of a prime number generator is
87presented later on.)
88
89=head1 Native threads
90
91There are several different ways to implement threads on a system. How
92threads are implemented depends both on the vendor and, in some cases,
93the version of the operating system. Often the first implementation
94will be relatively simple, but later versions of the OS will be more
95sophisticated.
96
97While the information in this section is useful, it's not necessary,
98so you can skip it if you don't feel up to it.
99
100There are three basic categories of threads-user-mode threads, kernel
101threads, and multiprocessor kernel threads.
102
103User-mode threads are threads that live entirely within a program and
104its libraries. In this model, the OS knows nothing about threads. As
105far as it's concerned, your process is just a process.
106
107This is the easiest way to implement threads, and the way most OSes
108start. The big disadvantage is that, since the OS knows nothing about
109threads, if one thread blocks they all do. Typical blocking activities
110include most system calls, most I/O, and things like sleep().
111
112Kernel threads are the next step in thread evolution. The OS knows
113about kernel threads, and makes allowances for them. The main
114difference between a kernel thread and a user-mode thread is
115blocking. With kernel threads, things that block a single thread don't
116block other threads. This is not the case with user-mode threads,
117where the kernel blocks at the process level and not the thread level.
118
119This is a big step forward, and can give a threaded program quite a
120performance boost over non-threaded programs. Threads that block
121performing I/O, for example, won't block threads that are doing other
122things. Each process still has only one thread running at once,
123though, regardless of how many CPUs a system might have.
124
125Since kernel threading can interrupt a thread at any time, they will
126uncover some of the implicit locking assumptions you may make in your
127program. For example, something as simple as C<$a = $a + 2> can behave
128unpredictably with kernel threads if $a is visible to other
129threads, as another thread may have changed $a between the time it
130was fetched on the right hand side and the time the new value is
131stored.
132
133Multiprocessor Kernel Threads are the final step in thread
134support. With multiprocessor kernel threads on a machine with multiple
135CPUs, the OS may schedule two or more threads to run simultaneously on
136different CPUs.
137
138This can give a serious performance boost to your threaded program,
139since more than one thread will be executing at the same time. As a
140tradeoff, though, any of those nagging synchronization issues that
141might not have shown with basic kernel threads will appear with a
142vengeance.
143
144In addition to the different levels of OS involvement in threads,
145different OSes (and different thread implementations for a particular
146OS) allocate CPU cycles to threads in different ways.
147
148Cooperative multitasking systems have running threads give up control
149if one of two things happen. If a thread calls a yield function, it
150gives up control. It also gives up control if the thread does
151something that would cause it to block, such as perform I/O. In a
152cooperative multitasking implementation, one thread can starve all the
153others for CPU time if it so chooses.
154
155Preemptive multitasking systems interrupt threads at regular intervals
156while the system decides which thread should run next. In a preemptive
157multitasking system, one thread usually won't monopolize the CPU.
158
159On some systems, there can be cooperative and preemptive threads
160running simultaneously. (Threads running with realtime priorities
161often behave cooperatively, for example, while threads running at
162normal priorities behave preemptively.)
163
164=head1 What kind of threads are perl threads?
165
166If you have experience with other thread implementations, you might
167find that things aren't quite what you expect. It's very important to
168remember when dealing with Perl threads that Perl Threads Are Not X
169Threads, for all values of X. They aren't POSIX threads, or
170DecThreads, or Java's Green threads, or Win32 threads. There are
171similarities, and the broad concepts are the same, but if you start
172looking for implementation details you're going to be either
173disappointed or confused. Possibly both.
174
175This is not to say that Perl threads are completely different from
176everything that's ever come before--they're not. Perl's threading
177model owes a lot to other thread models, especially POSIX. Just as
178Perl is not C, though, Perl threads are not POSIX threads. So if you
179find yourself looking for mutexes, or thread priorities, it's time to
180step back a bit and think about what you want to do and how Perl can
181do it.
182
183=head1 Threadsafe Modules
184
185The addition of threads has changed Perl's internals
186substantially. There are implications for people who write
187modules--especially modules with XS code or external libraries. While
188most modules won't encounter any problems, modules that aren't
189explicitly tagged as thread-safe should be tested before being used in
190production code.
191
192Not all modules that you might use are thread-safe, and you should
193always assume a module is unsafe unless the documentation says
194otherwise. This includes modules that are distributed as part of the
195core. Threads are a beta feature, and even some of the standard
196modules aren't thread-safe.
197
198If you're using a module that's not thread-safe for some reason, you
199can protect yourself by using semaphores and lots of programming
200discipline to control access to the module. Semaphores are covered
201later in the article. Perl Threads Are Different
202
203=head1 Thread Basics
204
205The core Thread module provides the basic functions you need to write
206threaded programs. In the following sections we'll cover the basics,
207showing you what you need to do to create a threaded program. After
208that, we'll go over some of the features of the Thread module that
209make threaded programming easier.
210
211=head2 Basic Thread Support
212
213Thread support is a Perl compile-time option-it's something that's
214turned on or off when Perl is built at your site, rather than when
215your programs are compiled. If your Perl wasn't compiled with thread
216support enabled, then any attempt to use threads will fail.
217
218Remember that the threading support in 5.005 is in beta release, and
219should be treated as such. You should expect that it may not function
220entirely properly, and the thread interface may well change some
221before it is a fully supported, production release. The beta version
222shouldn't be used for mission-critical projects. Having said that,
223threaded Perl is pretty nifty, and worth a look.
224
225Your programs can use the Config module to check whether threads are
226enabled. If your program can't run without them, you can say something
227like:
228
229 $Config{usethreads} or die "Recompile Perl with threads to run this program.";
230
231A possibly-threaded program using a possibly-threaded module might
232have code like this:
233
234 use Config;
235 use MyMod;
236
237 if ($Config{usethreads}) {
238 # We have threads
239 require MyMod_threaded;
240 import MyMod_threaded;
241 } else {
242 require MyMod_unthreaded;
243 import MyMod_unthreaded;
244 }
245
246Since code that runs both with and without threads is usually pretty
247messy, it's best to isolate the thread-specific code in its own
248module. In our example above, that's what MyMod_threaded is, and it's
249only imported if we're running on a threaded Perl.
250
251=head2 Creating Threads
252
253The Thread package provides the tools you need to create new
254threads. Like any other module, you need to tell Perl you want to use
255it; use Thread imports all the pieces you need to create basic
256threads.
257
258The simplest, straightforward way to create a thread is with new():
259
260 use Thread;
261
262 $thr = new Thread \&sub1;
263
264 sub sub1 {
265 print "In the thread\n";
266 }
267
268The new() method takes a reference to a subroutine and creates a new
269thread, which starts executing in the referenced subroutine. Control
270then passes both to the subroutine and the caller.
271
272If you need to, your program can pass parameters to the subroutine as
273part of the thread startup. Just include the list of parameters as
274part of the C<Thread::new> call, like this:
275
276 use Thread;
277 $Param3 = "foo";
278 $thr = new Thread \&sub1, "Param 1", "Param 2", $Param3;
279 $thr = new Thread \&sub1, @ParamList;
280 $thr = new Thread \&sub1, qw(Param1 Param2 $Param3);
281
282 sub sub1 {
283 my @InboundParameters = @_;
284 print "In the thread\n";
285 print "got parameters >", join("<>", @InboundParameters), "<\n";
286 }
287
288
289The subroutine runs like a normal Perl subroutine, and the call to new
290Thread returns whatever the subroutine returns.
291
292The last example illustrates another feature of threads. You can spawn
293off several threads using the same subroutine. Each thread executes
294the same subroutine, but in a separate thread with a separate
295environment and potentially separate arguments.
296
297The other way to spawn a new thread is with async(), which is a way to
298spin off a chunk of code like eval(), but into its own thread:
299
300 use Thread qw(async);
301
302 $LineCount = 0;
303
304 $thr = async {
305 while(<>) {$LineCount++}
306 print "Got $LineCount lines\n";
307 };
308
309 print "Waiting for the linecount to end\n";
310 $thr->join;
311 print "All done\n";
312
313You'll notice we did a use Thread qw(async) in that example. async is
314not exported by default, so if you want it, you'll either need to
315import it before you use it or fully qualify it as
316Thread::async. You'll also note that there's a semicolon after the
317closing brace. That's because async() treats the following block as an
318anonymous subroutine, so the semicolon is necessary.
319
320Like eval(), the code executes in the same context as it would if it
321weren't spun off. Since both the code inside and after the async start
322executing, you need to be careful with any shared resources. Locking
323and other synchronization techniques are covered later.
324
325=head2 Giving up control
326
327There are times when you may find it useful to have a thread
328explicitly give up the CPU to another thread. Your threading package
329might not support preemptive multitasking for threads, for example, or
330you may be doing something compute-intensive and want to make sure
331that the user-interface thread gets called frequently. Regardless,
332there are times that you might want a thread to give up the processor.
333
334Perl's threading package provides the yield() function that does
335this. yield() is pretty straightforward, and works like this:
336
337 use Thread qw(yield async);
338 async {
339 my $foo = 50;
340 while ($foo--) { print "first async\n" }
341 yield;
342 $foo = 50;
343 while ($foo--) { print "first async\n" }
344 };
345 async {
346 my $foo = 50;
347 while ($foo--) { print "second async\n" }
348 yield;
349 $foo = 50;
350 while ($foo--) { print "second async\n" }
351 };
352
353=head2 Waiting For A Thread To Exit
354
355Since threads are also subroutines, they can return values. To wait
356for a thread to exit and extract any scalars it might return, you can
357use the join() method.
358
359 use Thread;
360 $thr = new Thread \&sub1;
361
362 @ReturnData = $thr->join;
363 print "Thread returned @ReturnData";
364
365 sub sub1 { return "Fifty-six", "foo", 2; }
366
367In the example above, the join() method returns as soon as the thread
368ends. In addition to waiting for a thread to finish and gathering up
369any values that the thread might have returned, join() also performs
370any OS cleanup necessary for the thread. That cleanup might be
371important, especially for long-running programs that spawn lots of
372threads. If you don't want the return values and don't want to wait
373for the thread to finish, you should call the detach() method
374instead. detach() is covered later in the article.
375
376=head2 Errors In Threads
377
378So what happens when an error occurs in a thread? Any errors that
379could be caught with eval() are postponed until the thread is
380joined. If your program never joins, the errors appear when your
381program exits.
382
383Errors deferred until a join() can be caught with eval():
384
385 use Thread qw(async);
386 $thr = async {$b = 3/0}; # Divide by zero error
387 $foo = eval {$thr->join};
388 if ($@) {
389 print "died with error $@\n";
390 } else {
391 print "Hey, why aren't you dead?\n";
392 }
393
394eval() passes any results from the joined thread back unmodified, so
395if you want the return value of the thread, this is your only chance
396to get them.
397
398=head2 Ignoring A Thread
399
400join() does three things: it waits for a thread to exit, cleans up
401after it, and returns any data the thread may have produced. But what
402if you're not interested in the thread's return values, and you don't
403really care when the thread finishes? All you want is for the thread
404to get cleaned up after when it's done.
405
406In this case, you use the detach() method. Once a thread is detached,
407it'll run until it's finished, then Perl will clean up after it
408automatically.
409
410 use Thread;
411 $thr = new Thread \&sub1; # Spawn the thread
412
413 $thr->detach; # Now we officially don't care any more
414
415 sub sub1 {
416 $a = 0;
417 while (1) {
418 $a++;
419 print "\$a is $a\n";
420 sleep 1;
421 }
422 }
423
424
425Once a thread is detached, it may not be joined, and any output that
426it might have produced (if it was done and waiting for a join) is
427lost.
428
429=head1 Threads And Data
430
431Now that we've covered the basics of threads, it's time for our next
432topic: data. Threading introduces a couple of complications to data
433access that non-threaded programs never need to worry about.
434
435=head2 Shared And Unshared Data
436
437The single most important thing to remember when using threads is that
438all threads potentially have access to all the data anywhere in your
439program. While this is true with a nonthreaded Perl program as well,
440it's especially important to remember with a threaded program, since
441more than one thread can be accessing this data at once.
442
443Perl's scoping rules don't change because you're using threads. If a
444subroutine (or block, in the case of async()) could see a variable if
445you weren't running with threads, it can see it if you are. This is
446especially important for the subroutines that create, and makes C<my>
447variables even more important. Remember--if your variables aren't
448lexically scoped (declared with C<my>) you're probably sharing them
449between threads.
450
451=head2 Thread Pitfall: Races
452
453While threads bring a new set of useful tools, they also bring a
454number of pitfalls. One pitfall is the race condition:
455
456 use Thread;
457 $a = 1;
458 $thr1 = Thread->new(\&sub1);
459 $thr2 = Thread->new(\&sub2);
460
461 sleep 10;
462 print "$a\n";
463
464 sub sub1 { $foo = $a; $a = $foo + 1; }
465 sub sub2 { $bar = $a; $a = $bar + 1; }
466
467What do you think $a will be? The answer, unfortunately, is "it
468depends." Both sub1() and sub2() access the global variable $a, once
469to read and once to write. Depending on factors ranging from your
470thread implementation's scheduling algorithm to the phase of the moon,
471$a can be 2 or 3.
472
473Race conditions are caused by unsynchronized access to shared
474data. Without explicit synchronization, there's no way to be sure that
475nothing has happened to the shared data between the time you access it
476and the time you update it. Even this simple code fragment has the
477possibility of error:
478
479 use Thread qw(async);
480 $a = 2;
481 async{ $b = $a; $a = $b + 1; };
482 async{ $c = $a; $a = $c + 1; };
483
484Two threads both access $a. Each thread can potentially be interrupted
485at any point, or be executed in any order. At the end, $a could be 3
486or 4, and both $b and $c could be 2 or 3.
487
488Whenever your program accesses data or resources that can be accessed
489by other threads, you must take steps to coordinate access or risk
490data corruption and race conditions.
491
492=head2 Controlling access: lock()
493
494The lock() function takes a variable (or subroutine, but we'll get to
495that later) and puts a lock on it. No other thread may lock the
496variable until the locking thread exits the innermost block containing
497the lock. Using lock() is straightforward:
498
499 use Thread qw(async);
500 $a = 4;
501 $thr1 = async {
502 $foo = 12;
503 {
504 lock ($a); # Block until we get access to $a
505 $b = $a;
506 $a = $b * $foo;
507 }
508 print "\$foo was $foo\n";
509 };
510 $thr2 = async {
511 $bar = 7;
512 {
513 lock ($a); # Block until we can get access to $a
514 $c = $a;
515 $a = $c * $bar;
516 }
517 print "\$bar was $bar\n";
518 };
519 $thr1->join;
520 $thr2->join;
521 print "\$a is $a\n";
522
523lock() blocks the thread until the variable being locked is
524available. When lock() returns, your thread can be sure that no other
525thread can lock that variable until the innermost block containing the
526lock exits.
527
528It's important to note that locks don't prevent access to the variable
529in question, only lock attempts. This is in keeping with Perl's
530longstanding tradition of courteous programming, and the advisory file
531locking that flock() gives you. Locked subroutines behave differently,
532however. We'll cover that later in the article.
533
534You may lock arrays and hashes as well as scalars. Locking an array,
535though, will not block subsequent locks on array elements, just lock
536attempts on the array itself.
537
538Finally, locks are recursive, which means it's okay for a thread to
539lock a variable more than once. The lock will last until the outermost
540lock() on the variable goes out of scope.
541
542=head2 Thread Pitfall: Deadlocks
543
544Locks are a handy tool to synchronize access to data. Using them
545properly is the key to safe shared data. Unfortunately, locks aren't
546without their dangers. Consider the following code:
547
548 use Thread qw(async yield);
549 $a = 4;
550 $b = "foo";
551 async {
552 lock($a);
553 yield;
554 sleep 20;
555 lock ($b);
556 };
557 async {
558 lock($b);
559 yield;
560 sleep 20;
561 lock ($a);
562 };
563
564This program will probably hang until you kill it. The only way it
565won't hang is if one of the two async() routines acquires both locks
566first. A guaranteed-to-hang version is more complicated, but the
567principle is the same.
568
569The first thread spawned by async() will grab a lock on $a then, a
570second or two later, try to grab a lock on $b. Meanwhile, the second
571thread grabs a lock on $b, then later tries to grab a lock on $a. The
572second lock attempt for both threads will block, each waiting for the
573other to release its lock.
574
575This condition is called a deadlock, and it occurs whenever two or
576more threads are trying to get locks on resources that the others
577own. Each thread will block, waiting for the other to release a lock
578on a resource. That never happens, though, since the thread with the
579resource is itself waiting for a lock to be released.
580
581There are a number of ways to handle this sort of problem. The best
582way is to always have all threads acquire locks in the exact same
583order. If, for example, you lock variables $a, $b, and $c, always lock
584$a before $b, and $b before $c. It's also best to hold on to locks for
585as short a period of time to minimize the risks of deadlock.
586
587=head2 Queues: Passing Data Around
588
589A queue is a special thread-safe object that lets you put data in one
590end and take it out the other without having to worry about
591synchronization issues. They're pretty straightforward, and look like
592this:
593
594 use Thread qw(async);
595 use Thread::Queue;
596
597 my $DataQueue = new Thread::Queue;
598 $thr = async {
599 while ($DataElement = $DataQueue->dequeue) {
600 print "Popped $DataElement off the queue\n";
601 }
602 };
603
604 $DataQueue->enqueue(12);
605 $DataQueue->enqueue("A", "B", "C");
606 $DataQueue->enqueue(\$thr);
607 sleep 10;
608 $DataQueue->enqueue(undef);
609
610You create the queue with new Thread::Queue. Then you can add lists of
611scalars onto the end with enqueue(), and pop scalars off the front of
612it with dequeue(). A queue has no fixed size, and can grow as needed
613to hold everything pushed on to it.
614
615If a queue is empty, dequeue() blocks until another thread enqueues
616something. This makes queues ideal for event loops and other
617communications between threads.
618
619=head1 Threads And Code
620
621In addition to providing thread-safe access to data via locks and
622queues, threaded Perl also provides general-purpose semaphores for
623coarser synchronization than locks provide and thread-safe access to
624entire subroutines.
625
626=head2 Semaphores: Synchronizing Data Access
627
628Semaphores are a kind of generic locking mechanism. Unlike lock, which
629gets a lock on a particular scalar, Perl doesn't associate any
630particular thing with a semaphore so you can use them to control
631access to anything you like. In addition, semaphores can allow more
632than one thread to access a resource at once, though by default
633semaphores only allow one thread access at a time.
634
635=over 4
636
637=item Basic semaphores
638
639Semaphores have two methods, down and up. down decrements the resource
640count, while up increments it. down calls will block if the
641semaphore's current count would decrement below zero. This program
642gives a quick demonstration:
643
644 use Thread qw(yield);
645 use Thread::Semaphore;
646 my $semaphore = new Thread::Semaphore;
647 $GlobalVariable = 0;
648
649 $thr1 = new Thread \&sample_sub, 1;
650 $thr2 = new Thread \&sample_sub, 2;
651 $thr3 = new Thread \&sample_sub, 3;
652
653 sub sample_sub {
654 my $SubNumber = shift @_;
655 my $TryCount = 10;
656 my $LocalCopy;
657 sleep 1;
658 while ($TryCount--) {
659 $semaphore->down;
660 $LocalCopy = $GlobalVariable;
661 print "$TryCount tries left for sub $SubNumber (\$GlobalVariable is $GlobalVariable)\n";
662 yield;
663 sleep 2;
664 $LocalCopy++;
665 $GlobalVariable = $LocalCopy;
666 $semaphore->up;
667 }
668 }
669
670The three invocations of the subroutine all operate in sync. The
671semaphore, though, makes sure that only one thread is accessing the
672global variable at once.
673
674=item Advanced Semaphores
675
676By default, semaphores behave like locks, letting only one thread
677down() them at a time. However, there are other uses for semaphores.
678
679Each semaphore has a counter attached to it. down() decrements the
680counter and up() increments the counter. By default, semaphores are
681created with the counter set to one, down() decrements by one, and
682up() increments by one. If down() attempts to decrement the counter
683below zero, it blocks until the counter is large enough. Note that
684while a semaphore can be created with a starting count of zero, any
685up() or down() always changes the counter by at least
686one. $semaphore->down(0) is the same as $semaphore->down(1).
687
688The question, of course, is why would you do something like this? Why
689create a semaphore with a starting count that's not one, or why
690decrement/increment it by more than one? The answer is resource
691availability. Many resources that you want to manage access for can be
692safely used by more than one thread at once.
693
694For example, let's take a GUI driven program. It has a semaphore that
695it uses to synchronize access to the display, so only one thread is
696ever drawing at once. Handy, but of course you don't want any thread
697to start drawing until things are properly set up. In this case, you
698can create a semaphore with a counter set to zero, and up it when
699things are ready for drawing.
700
701Semaphores with counters greater than one are also useful for
702establishing quotas. Say, for example, that you have a number of
703threads that can do I/O at once. You don't want all the threads
704reading or writing at once though, since that can potentially swamp
705your I/O channels, or deplete your process' quota of filehandles. You
706can use a semaphore initialized to the number of concurrent I/O
707requests (or open files) that you want at any one time, and have your
708threads quietly block and unblock themselves.
709
710Larger increments or decrements are handy in those cases where a
711thread needs to check out or return a number of resources at once.
712
713=back
714
715=head2 Attributes: Restricting Access To Subroutines
716
717In addition to synchronizing access to data or resources, you might
718find it useful to synchronize access to subroutines. You may be
719accessing a singular machine resource (perhaps a vector processor), or
720find it easier to serialize calls to a particular subroutine than to
721have a set of locks and semaphores.
722
723One of the additions to Perl 5.005 is subroutine attributes. The
724Thread package uses these to provide several flavors of
725serialization. It's important to remember that these attributes are
726used in the compilation phase of your program so you can't change a
727subroutine's behavior while your program is actually running.
728
729=head2 Subroutine Locks
730
731The basic subroutine lock looks like this:
732
733 sub test_sub :locked {
734 }
735
736This ensures that only one thread will be executing this subroutine at
737any one time. Once a thread calls this subroutine, any other thread
738that calls it will block until the thread in the subroutine exits
739it. A more elaborate example looks like this:
740
741 use Thread qw(yield);
742
743 new Thread \&thread_sub, 1;
744 new Thread \&thread_sub, 2;
745 new Thread \&thread_sub, 3;
746 new Thread \&thread_sub, 4;
747
748 sub sync_sub :locked {
749 my $CallingThread = shift @_;
750 print "In sync_sub for thread $CallingThread\n";
751 yield;
752 sleep 3;
753 print "Leaving sync_sub for thread $CallingThread\n";
754 }
755
756 sub thread_sub {
757 my $ThreadID = shift @_;
758 print "Thread $ThreadID calling sync_sub\n";
759 sync_sub($ThreadID);
760 print "$ThreadID is done with sync_sub\n";
761 }
762
763The C<locked> attribute tells perl to lock sync_sub(), and if you run
764this, you can see that only one thread is in it at any one time.
765
766=head2 Methods
767
768Locking an entire subroutine can sometimes be overkill, especially
769when dealing with Perl objects. When calling a method for an object,
770for example, you want to serialize calls to a method, so that only one
771thread will be in the subroutine for a particular object, but threads
772calling that subroutine for a different object aren't blocked. The
773method attribute indicates whether the subroutine is really a method.
774
775 use Thread;
776
777 sub tester {
778 my $thrnum = shift @_;
779 my $bar = new Foo;
780 foreach (1..10) {
781 print "$thrnum calling per_object\n";
782 $bar->per_object($thrnum);
783 print "$thrnum out of per_object\n";
784 yield;
785 print "$thrnum calling one_at_a_time\n";
786 $bar->one_at_a_time($thrnum);
787 print "$thrnum out of one_at_a_time\n";
788 yield;
789 }
790 }
791
792 foreach my $thrnum (1..10) {
793 new Thread \&tester, $thrnum;
794 }
795
796 package Foo;
797 sub new {
798 my $class = shift @_;
799 return bless [@_], $class;
800 }
801
802 sub per_object :locked :method {
803 my ($class, $thrnum) = @_;
804 print "In per_object for thread $thrnum\n";
805 yield;
806 sleep 2;
807 print "Exiting per_object for thread $thrnum\n";
808 }
809
810 sub one_at_a_time :locked {
811 my ($class, $thrnum) = @_;
812 print "In one_at_a_time for thread $thrnum\n";
813 yield;
814 sleep 2;
815 print "Exiting one_at_a_time for thread $thrnum\n";
816 }
817
818As you can see from the output (omitted for brevity; it's 800 lines)
819all the threads can be in per_object() simultaneously, but only one
820thread is ever in one_at_a_time() at once.
821
822=head2 Locking A Subroutine
823
824You can lock a subroutine as you would lock a variable. Subroutine locks
825work the same as specifying a C<locked> attribute for the subroutine,
826and block all access to the subroutine for other threads until the
827lock goes out of scope. When the subroutine isn't locked, any number
828of threads can be in it at once, and getting a lock on a subroutine
829doesn't affect threads already in the subroutine. Getting a lock on a
830subroutine looks like this:
831
832 lock(\&sub_to_lock);
833
834Simple enough. Unlike the C<locked> attribute, which is a compile time
835option, locking and unlocking a subroutine can be done at runtime at your
836discretion. There is some runtime penalty to using lock(\&sub) instead
837of the C<locked> attribute, so make sure you're choosing the proper
838method to do the locking.
839
840You'd choose lock(\&sub) when writing modules and code to run on both
841threaded and unthreaded Perl, especially for code that will run on
8425.004 or earlier Perls. In that case, it's useful to have subroutines
843that should be serialized lock themselves if they're running threaded,
844like so:
845
846 package Foo;
847 use Config;
848 $Running_Threaded = 0;
849
850 BEGIN { $Running_Threaded = $Config{'usethreads'} }
851
852 sub sub1 { lock(\&sub1) if $Running_Threaded }
853
854
855This way you can ensure single-threadedness regardless of which
856version of Perl you're running.
857
858=head1 General Thread Utility Routines
859
860We've covered the workhorse parts of Perl's threading package, and
861with these tools you should be well on your way to writing threaded
862code and packages. There are a few useful little pieces that didn't
863really fit in anyplace else.
864
865=head2 What Thread Am I In?
866
867The Thread->self method provides your program with a way to get an
868object representing the thread it's currently in. You can use this
869object in the same way as the ones returned from the thread creation.
870
871=head2 Thread IDs
872
873tid() is a thread object method that returns the thread ID of the
874thread the object represents. Thread IDs are integers, with the main
875thread in a program being 0. Currently Perl assigns a unique tid to
876every thread ever created in your program, assigning the first thread
877to be created a tid of 1, and increasing the tid by 1 for each new
878thread that's created.
879
880=head2 Are These Threads The Same?
881
882The equal() method takes two thread objects and returns true
883if the objects represent the same thread, and false if they don't.
884
885=head2 What Threads Are Running?
886
887Thread->list returns a list of thread objects, one for each thread
888that's currently running. Handy for a number of things, including
889cleaning up at the end of your program:
890
891 # Loop through all the threads
892 foreach $thr (Thread->list) {
893 # Don't join the main thread or ourselves
894 if ($thr->tid && !Thread::equal($thr, Thread->self)) {
895 $thr->join;
896 }
897 }
898
899The example above is just for illustration. It isn't strictly
900necessary to join all the threads you create, since Perl detaches all
901the threads before it exits.
902
903=head1 A Complete Example
904
905Confused yet? It's time for an example program to show some of the
906things we've covered. This program finds prime numbers using threads.
907
908 1 #!/usr/bin/perl -w
909 2 # prime-pthread, courtesy of Tom Christiansen
910 3
911 4 use strict;
912 5
913 6 use Thread;
914 7 use Thread::Queue;
915 8
916 9 my $stream = new Thread::Queue;
917 10 my $kid = new Thread(\&check_num, $stream, 2);
918 11
919 12 for my $i ( 3 .. 1000 ) {
920 13 $stream->enqueue($i);
921 14 }
922 15
923 16 $stream->enqueue(undef);
924 17 $kid->join();
925 18
926 19 sub check_num {
927 20 my ($upstream, $cur_prime) = @_;
928 21 my $kid;
929 22 my $downstream = new Thread::Queue;
930 23 while (my $num = $upstream->dequeue) {
931 24 next unless $num % $cur_prime;
932 25 if ($kid) {
933 26 $downstream->enqueue($num);
934 27 } else {
935 28 print "Found prime $num\n";
936 29 $kid = new Thread(\&check_num, $downstream, $num);
937 30 }
938 31 }
939 32 $downstream->enqueue(undef) if $kid;
940 33 $kid->join() if $kid;
941 34 }
942
943This program uses the pipeline model to generate prime numbers. Each
944thread in the pipeline has an input queue that feeds numbers to be
945checked, a prime number that it's responsible for, and an output queue
946that it funnels numbers that have failed the check into. If the thread
947has a number that's failed its check and there's no child thread, then
948the thread must have found a new prime number. In that case, a new
949child thread is created for that prime and stuck on the end of the
950pipeline.
951
952This probably sounds a bit more confusing than it really is, so lets
953go through this program piece by piece and see what it does. (For
954those of you who might be trying to remember exactly what a prime
955number is, it's a number that's only evenly divisible by itself and 1)
956
957The bulk of the work is done by the check_num() subroutine, which
958takes a reference to its input queue and a prime number that it's
959responsible for. After pulling in the input queue and the prime that
960the subroutine's checking (line 20), we create a new queue (line 22)
961and reserve a scalar for the thread that we're likely to create later
962(line 21).
963
964The while loop from lines 23 to line 31 grabs a scalar off the input
965queue and checks against the prime this thread is responsible
966for. Line 24 checks to see if there's a remainder when we modulo the
967number to be checked against our prime. If there is one, the number
968must not be evenly divisible by our prime, so we need to either pass
969it on to the next thread if we've created one (line 26) or create a
970new thread if we haven't.
971
972The new thread creation is line 29. We pass on to it a reference to
973the queue we've created, and the prime number we've found.
974
975Finally, once the loop terminates (because we got a 0 or undef in the
976queue, which serves as a note to die), we pass on the notice to our
977child and wait for it to exit if we've created a child (Lines 32 and
97837).
979
980Meanwhile, back in the main thread, we create a queue (line 9) and the
981initial child thread (line 10), and pre-seed it with the first prime:
9822. Then we queue all the numbers from 3 to 1000 for checking (lines
98312-14), then queue a die notice (line 16) and wait for the first child
984thread to terminate (line 17). Because a child won't die until its
985child has died, we know that we're done once we return from the join.
986
987That's how it works. It's pretty simple; as with many Perl programs,
988the explanation is much longer than the program.
989
990=head1 Conclusion
991
992A complete thread tutorial could fill a book (and has, many times),
993but this should get you well on your way. The final authority on how
994Perl's threads behave is the documentation bundled with the Perl
995distribution, but with what we've covered in this article, you should
996be well on your way to becoming a threaded Perl expert.
997
998=head1 Bibliography
999
1000Here's a short bibliography courtesy of Jürgen Christoffel:
1001
1002=head2 Introductory Texts
1003
1004Birrell, Andrew D. An Introduction to Programming with
1005Threads. Digital Equipment Corporation, 1989, DEC-SRC Research Report
1006#35 online as
1007http://www.research.digital.com/SRC/staff/birrell/bib.html (highly
1008recommended)
1009
1010Robbins, Kay. A., and Steven Robbins. Practical Unix Programming: A
1011Guide to Concurrency, Communication, and
1012Multithreading. Prentice-Hall, 1996.
1013
1014Lewis, Bill, and Daniel J. Berg. Multithreaded Programming with
1015Pthreads. Prentice Hall, 1997, ISBN 0-13-443698-9 (a well-written
1016introduction to threads).
1017
1018Nelson, Greg (editor). Systems Programming with Modula-3. Prentice
1019Hall, 1991, ISBN 0-13-590464-1.
1020
1021Nichols, Bradford, Dick Buttlar, and Jacqueline Proulx Farrell.
1022Pthreads Programming. O'Reilly & Associates, 1996, ISBN 156592-115-1
1023(covers POSIX threads).
1024
1025=head2 OS-Related References
1026
1027Boykin, Joseph, David Kirschen, Alan Langerman, and Susan
1028LoVerso. Programming under Mach. Addison-Wesley, 1994, ISBN
10290-201-52739-1.
1030
1031Tanenbaum, Andrew S. Distributed Operating Systems. Prentice Hall,
10321995, ISBN 0-13-219908-4 (great textbook).
1033
1034Silberschatz, Abraham, and Peter B. Galvin. Operating System Concepts,
10354th ed. Addison-Wesley, 1995, ISBN 0-201-59292-4
1036
1037=head2 Other References
1038
1039Arnold, Ken and James Gosling. The Java Programming Language, 2nd
1040ed. Addison-Wesley, 1998, ISBN 0-201-31006-6.
1041
1042Le Sergent, T. and B. Berthomieu. "Incremental MultiThreaded Garbage
1043Collection on Virtually Shared Memory Architectures" in Memory
1044Management: Proc. of the International Workshop IWMM 92, St. Malo,
1045France, September 1992, Yves Bekkers and Jacques Cohen, eds. Springer,
10461992, ISBN 3540-55940-X (real-life thread applications).
1047
1048=head1 Acknowledgements
1049
1050Thanks (in no particular order) to Chaim Frenkel, Steve Fink, Gurusamy
1051Sarathy, Ilya Zakharevich, Benjamin Sugars, Jürgen Christoffel, Joshua
1052Pritikin, and Alan Burlison, for their help in reality-checking and
1053polishing this article. Big thanks to Tom Christiansen for his rewrite
1054of the prime number generator.
1055
1056=head1 AUTHOR
1057
1058Dan Sugalski E<lt>sugalskd@ous.eduE<gt>
1059
1060=head1 Copyrights
1061
1062This article originally appeared in The Perl Journal #10, and is
1063copyright 1998 The Perl Journal. It appears courtesy of Jon Orwant and
1064The Perl Journal. This document may be distributed under the same terms
1065as Perl itself.
1066
1067