This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
Add META.yml and Stas' makemeta (plus extensive reworking
[perl5.git] / pod / perlthrtut.pod
CommitLineData
2605996a
JH
1=head1 NAME
2
3perlthrtut - tutorial on threads in Perl
4
5=head1 DESCRIPTION
6
53d7eaa8 7B<NOTE>: this tutorial describes the new Perl threading flavour
9316ed2f
JH
8introduced in Perl 5.6.0 called interpreter threads, or B<ithreads>
9for short. In this model each thread runs in its own Perl interpreter,
10and any data sharing between threads must be explicit.
11
12There is another older Perl threading flavour called the 5.005 model,
13unsurprisingly for 5.005 versions of Perl. The old model is known to
14have problems, deprecated, and will probably be removed around release
155.10. You are strongly encouraged to migrate any existing 5.005
16threads code to the new model as soon as possible.
2a4bf773 17
53d7eaa8 18You can see which (or neither) threading flavour you have by
6eded8f3 19running C<perl -V> and looking at the C<Platform> section.
53d7eaa8
JH
20If you have C<useithreads=define> you have ithreads, if you
21have C<use5005threads=define> you have 5.005 threads.
22If you have neither, you don't have any thread support built in.
23If you have both, you are in trouble.
2605996a 24
bfce6503
DM
25The user-level interface to the 5.005 threads was via the L<Threads>
26class, while ithreads uses the L<threads> class. Note the change in case.
27
28=head1 Status
29
30The ithreads code has been available since Perl 5.6.0, and is considered
31stable. The user-level interface to ithreads (the L<threads> classes)
9e75ef81
IT
32appeared in the 5.8.0 release, and as of this time is considered stable
33although it should be treated with caution as with all new features.
2605996a 34
c975c451
AB
35=head1 What Is A Thread Anyway?
36
37A thread is a flow of control through a program with a single
38execution point.
39
40Sounds an awful lot like a process, doesn't it? Well, it should.
41Threads are one of the pieces of a process. Every process has at least
42one thread and, up until now, every process running Perl had only one
43thread. With 5.8, though, you can create extra threads. We're going
44to show you how, when, and why.
45
46=head1 Threaded Program Models
47
48There are three basic ways that you can structure a threaded
49program. Which model you choose depends on what you need your program
50to do. For many non-trivial threaded programs you'll need to choose
51different models for different pieces of your program.
52
53=head2 Boss/Worker
54
55The boss/worker model usually has one `boss' thread and one or more
56`worker' threads. The boss thread gathers or generates tasks that need
57to be done, then parcels those tasks out to the appropriate worker
58thread.
59
60This model is common in GUI and server programs, where a main thread
61waits for some event and then passes that event to the appropriate
62worker threads for processing. Once the event has been passed on, the
63boss thread goes back to waiting for another event.
64
65The boss thread does relatively little work. While tasks aren't
66necessarily performed faster than with any other method, it tends to
67have the best user-response times.
68
69=head2 Work Crew
70
71In the work crew model, several threads are created that do
72essentially the same thing to different pieces of data. It closely
73mirrors classical parallel processing and vector processors, where a
74large array of processors do the exact same thing to many pieces of
75data.
76
77This model is particularly useful if the system running the program
78will distribute multiple threads across different processors. It can
79also be useful in ray tracing or rendering engines, where the
80individual threads can pass on interim results to give the user visual
81feedback.
82
83=head2 Pipeline
84
85The pipeline model divides up a task into a series of steps, and
86passes the results of one step on to the thread processing the
87next. Each thread does one thing to each piece of data and passes the
88results to the next thread in line.
89
90This model makes the most sense if you have multiple processors so two
91or more threads will be executing in parallel, though it can often
92make sense in other contexts as well. It tends to keep the individual
93tasks small and simple, as well as allowing some parts of the pipeline
94to block (on I/O or system calls, for example) while other parts keep
95going. If you're running different parts of the pipeline on different
96processors you may also take advantage of the caches on each
97processor.
98
99This model is also handy for a form of recursive programming where,
100rather than having a subroutine call itself, it instead creates
101another thread. Prime and Fibonacci generators both map well to this
102form of the pipeline model. (A version of a prime number generator is
103presented later on.)
104
105=head1 Native threads
106
107There are several different ways to implement threads on a system. How
108threads are implemented depends both on the vendor and, in some cases,
109the version of the operating system. Often the first implementation
110will be relatively simple, but later versions of the OS will be more
111sophisticated.
112
113While the information in this section is useful, it's not necessary,
114so you can skip it if you don't feel up to it.
115
6eded8f3 116There are three basic categories of threads: user-mode threads, kernel
c975c451
AB
117threads, and multiprocessor kernel threads.
118
119User-mode threads are threads that live entirely within a program and
120its libraries. In this model, the OS knows nothing about threads. As
121far as it's concerned, your process is just a process.
122
123This is the easiest way to implement threads, and the way most OSes
124start. The big disadvantage is that, since the OS knows nothing about
125threads, if one thread blocks they all do. Typical blocking activities
126include most system calls, most I/O, and things like sleep().
127
128Kernel threads are the next step in thread evolution. The OS knows
129about kernel threads, and makes allowances for them. The main
130difference between a kernel thread and a user-mode thread is
131blocking. With kernel threads, things that block a single thread don't
132block other threads. This is not the case with user-mode threads,
133where the kernel blocks at the process level and not the thread level.
134
135This is a big step forward, and can give a threaded program quite a
136performance boost over non-threaded programs. Threads that block
137performing I/O, for example, won't block threads that are doing other
138things. Each process still has only one thread running at once,
139though, regardless of how many CPUs a system might have.
140
141Since kernel threading can interrupt a thread at any time, they will
142uncover some of the implicit locking assumptions you may make in your
143program. For example, something as simple as C<$a = $a + 2> can behave
144unpredictably with kernel threads if $a is visible to other
145threads, as another thread may have changed $a between the time it
146was fetched on the right hand side and the time the new value is
147stored.
148
6eded8f3 149Multiprocessor kernel threads are the final step in thread
c975c451
AB
150support. With multiprocessor kernel threads on a machine with multiple
151CPUs, the OS may schedule two or more threads to run simultaneously on
152different CPUs.
153
154This can give a serious performance boost to your threaded program,
155since more than one thread will be executing at the same time. As a
156tradeoff, though, any of those nagging synchronization issues that
157might not have shown with basic kernel threads will appear with a
158vengeance.
159
160In addition to the different levels of OS involvement in threads,
161different OSes (and different thread implementations for a particular
162OS) allocate CPU cycles to threads in different ways.
163
164Cooperative multitasking systems have running threads give up control
165if one of two things happen. If a thread calls a yield function, it
166gives up control. It also gives up control if the thread does
167something that would cause it to block, such as perform I/O. In a
168cooperative multitasking implementation, one thread can starve all the
169others for CPU time if it so chooses.
170
171Preemptive multitasking systems interrupt threads at regular intervals
172while the system decides which thread should run next. In a preemptive
173multitasking system, one thread usually won't monopolize the CPU.
174
175On some systems, there can be cooperative and preemptive threads
176running simultaneously. (Threads running with realtime priorities
177often behave cooperatively, for example, while threads running at
178normal priorities behave preemptively.)
179
bfce6503 180=head1 What kind of threads are Perl threads?
c975c451
AB
181
182If you have experience with other thread implementations, you might
183find that things aren't quite what you expect. It's very important to
184remember when dealing with Perl threads that Perl Threads Are Not X
185Threads, for all values of X. They aren't POSIX threads, or
186DecThreads, or Java's Green threads, or Win32 threads. There are
187similarities, and the broad concepts are the same, but if you start
188looking for implementation details you're going to be either
189disappointed or confused. Possibly both.
190
191This is not to say that Perl threads are completely different from
192everything that's ever come before--they're not. Perl's threading
193model owes a lot to other thread models, especially POSIX. Just as
194Perl is not C, though, Perl threads are not POSIX threads. So if you
195find yourself looking for mutexes, or thread priorities, it's time to
196step back a bit and think about what you want to do and how Perl can
197do it.
198
6eded8f3 199However it is important to remember that Perl threads cannot magically
c975c451 200do things unless your operating systems threads allows it. So if your
bfce6503 201system blocks the entire process on sleep(), Perl usually will as well.
c975c451 202
9316ed2f
JH
203Perl Threads Are Different.
204
cf5baa48 205=head1 Thread-Safe Modules
c975c451 206
cf5baa48 207The addition of threads has changed Perl's internals
c975c451 208substantially. There are implications for people who write
cf5baa48
JH
209modules with XS code or external libraries. However, since perl data is
210not shared among threads by default, Perl modules stand a high chance of
211being thread-safe or can be made thread-safe easily. Modules that are not
212tagged as thread-safe should be tested or code reviewed before being used
213in production code.
c975c451
AB
214
215Not all modules that you might use are thread-safe, and you should
216always assume a module is unsafe unless the documentation says
217otherwise. This includes modules that are distributed as part of the
218core. Threads are a new feature, and even some of the standard
bfce6503 219modules aren't thread-safe.
c975c451 220
cf5baa48 221Even if a module is thread-safe, it doesn't mean that the module is optimized
6eded8f3
SG
222to work well with threads. A module could possibly be rewritten to utilize
223the new features in threaded Perl to increase performance in a threaded
224environment.
c975c451
AB
225
226If you're using a module that's not thread-safe for some reason, you
cf5baa48
JH
227can protect yourself by using it from one, and only one thread at all.
228If you need multiple threads to access such a module, you can use semaphores and
229lots of programming discipline to control access to it. Semaphores
230are covered in L</"Basic semaphores">.
9316ed2f 231
cf5baa48 232See also L</"Thread-Safety of System Libraries">.
c975c451
AB
233
234=head1 Thread Basics
235
236The core L<threads> module provides the basic functions you need to write
237threaded programs. In the following sections we'll cover the basics,
238showing you what you need to do to create a threaded program. After
239that, we'll go over some of the features of the L<threads> module that
240make threaded programming easier.
241
242=head2 Basic Thread Support
243
6eded8f3 244Thread support is a Perl compile-time option - it's something that's
c975c451
AB
245turned on or off when Perl is built at your site, rather than when
246your programs are compiled. If your Perl wasn't compiled with thread
247support enabled, then any attempt to use threads will fail.
248
c975c451
AB
249Your programs can use the Config module to check whether threads are
250enabled. If your program can't run without them, you can say something
251like:
252
9316ed2f 253 $Config{useithreads} or die "Recompile Perl with threads to run this program.";
c975c451
AB
254
255A possibly-threaded program using a possibly-threaded module might
256have code like this:
257
cf5baa48
JH
258 use Config;
259 use MyMod;
c975c451 260
9316ed2f 261 BEGIN {
cf5baa48
JH
262 if ($Config{useithreads}) {
263 # We have threads
264 require MyMod_threaded;
265 import MyMod_threaded;
266 } else {
267 require MyMod_unthreaded;
268 import MyMod_unthreaded;
9316ed2f 269 }
cf5baa48 270 }
c975c451
AB
271
272Since code that runs both with and without threads is usually pretty
273messy, it's best to isolate the thread-specific code in its own
274module. In our example above, that's what MyMod_threaded is, and it's
275only imported if we're running on a threaded Perl.
276
8f95bfb9
EM
277=head2 A Note about the Examples
278
279Although thread support is considered to be stable, there are still a number
280of quirks that may startle you when you try out any of the examples below.
281In a real situation, care should be taken that all threads are finished
282executing before the program exits. That care has B<not> been taken in these
283examples in the interest of simplicity. Running these examples "as is" will
284produce error messages, usually caused by the fact that there are still
285threads running when the program exits. You should not be alarmed by this.
286Future versions of Perl may fix this problem.
287
c975c451
AB
288=head2 Creating Threads
289
290The L<threads> package provides the tools you need to create new
9e75ef81 291threads. Like any other module, you need to tell Perl that you want to use
c975c451
AB
292it; C<use threads> imports all the pieces you need to create basic
293threads.
294
9e75ef81 295The simplest, most straightforward way to create a thread is with new():
c975c451
AB
296
297 use threads;
298
299 $thr = threads->new(\&sub1);
300
301 sub sub1 {
302 print "In the thread\n";
303 }
304
305The new() method takes a reference to a subroutine and creates a new
306thread, which starts executing in the referenced subroutine. Control
307then passes both to the subroutine and the caller.
308
309If you need to, your program can pass parameters to the subroutine as
310part of the thread startup. Just include the list of parameters as
311part of the C<threads::new> call, like this:
312
313 use threads;
bfce6503 314
c975c451
AB
315 $Param3 = "foo";
316 $thr = threads->new(\&sub1, "Param 1", "Param 2", $Param3);
317 $thr = threads->new(\&sub1, @ParamList);
8f95bfb9 318 $thr = threads->new(\&sub1, qw(Param1 Param2 Param3));
c975c451
AB
319
320 sub sub1 {
321 my @InboundParameters = @_;
322 print "In the thread\n";
323 print "got parameters >", join("<>", @InboundParameters), "<\n";
324 }
325
326
327The last example illustrates another feature of threads. You can spawn
328off several threads using the same subroutine. Each thread executes
329the same subroutine, but in a separate thread with a separate
330environment and potentially separate arguments.
331
9316ed2f 332C<create()> is a synonym for C<new()>.
bfce6503 333
c975c451
AB
334=head2 Giving up control
335
336There are times when you may find it useful to have a thread
337explicitly give up the CPU to another thread. Your threading package
338might not support preemptive multitasking for threads, for example, or
9e75ef81 339you may be doing something processor-intensive and want to make sure
c975c451
AB
340that the user-interface thread gets called frequently. Regardless,
341there are times that you might want a thread to give up the processor.
342
343Perl's threading package provides the yield() function that does
344this. yield() is pretty straightforward, and works like this:
345
346 use threads;
818c4caa 347
bfce6503
DM
348 sub loop {
349 my $thread = shift;
350 my $foo = 50;
351 while($foo--) { print "in thread $thread\n" }
8f95bfb9 352 threads->yield;
bfce6503
DM
353 $foo = 50;
354 while($foo--) { print "in thread $thread\n" }
355 }
c975c451 356
bfce6503
DM
357 my $thread1 = threads->new(\&loop, 'first');
358 my $thread2 = threads->new(\&loop, 'second');
359 my $thread3 = threads->new(\&loop, 'third');
818c4caa 360
c975c451
AB
361It is important to remember that yield() is only a hint to give up the CPU,
362it depends on your hardware, OS and threading libraries what actually happens.
363Therefore it is important to note that one should not build the scheduling of
364the threads around yield() calls. It might work on your platform but it won't
365work on another platform.
366
367=head2 Waiting For A Thread To Exit
368
369Since threads are also subroutines, they can return values. To wait
6eded8f3
SG
370for a thread to exit and extract any values it might return, you can
371use the join() method:
c975c451
AB
372
373 use threads;
bfce6503 374
c975c451
AB
375 $thr = threads->new(\&sub1);
376
377 @ReturnData = $thr->join;
378 print "Thread returned @ReturnData";
379
380 sub sub1 { return "Fifty-six", "foo", 2; }
381
382In the example above, the join() method returns as soon as the thread
383ends. In addition to waiting for a thread to finish and gathering up
384any values that the thread might have returned, join() also performs
385any OS cleanup necessary for the thread. That cleanup might be
386important, especially for long-running programs that spawn lots of
387threads. If you don't want the return values and don't want to wait
388for the thread to finish, you should call the detach() method
bfce6503 389instead, as described next.
c975c451
AB
390
391=head2 Ignoring A Thread
392
393join() does three things: it waits for a thread to exit, cleans up
394after it, and returns any data the thread may have produced. But what
395if you're not interested in the thread's return values, and you don't
396really care when the thread finishes? All you want is for the thread
397to get cleaned up after when it's done.
398
399In this case, you use the detach() method. Once a thread is detached,
400it'll run until it's finished, then Perl will clean up after it
401automatically.
402
403 use threads;
bfce6503 404
6eded8f3 405 $thr = threads->new(\&sub1); # Spawn the thread
c975c451
AB
406
407 $thr->detach; # Now we officially don't care any more
408
cf5baa48 409 sub sub1 {
c975c451
AB
410 $a = 0;
411 while (1) {
412 $a++;
413 print "\$a is $a\n";
414 sleep 1;
415 }
416 }
417
bfce6503
DM
418Once a thread is detached, it may not be joined, and any return data
419that it might have produced (if it was done and waiting for a join) is
c975c451
AB
420lost.
421
422=head1 Threads And Data
423
424Now that we've covered the basics of threads, it's time for our next
425topic: data. Threading introduces a couple of complications to data
426access that non-threaded programs never need to worry about.
427
428=head2 Shared And Unshared Data
429
bfce6503
DM
430The biggest difference between Perl ithreads and the old 5.005 style
431threading, or for that matter, to most other threading systems out there,
432is that by default, no data is shared. When a new perl thread is created,
433all the data associated with the current thread is copied to the new
434thread, and is subsequently private to that new thread!
435This is similar in feel to what happens when a UNIX process forks,
436except that in this case, the data is just copied to a different part of
437memory within the same process rather than a real fork taking place.
c975c451 438
9e75ef81 439To make use of threading however, one usually wants the threads to share
bfce6503
DM
440at least some data between themselves. This is done with the
441L<threads::shared> module and the C< : shared> attribute:
442
443 use threads;
444 use threads::shared;
445
446 my $foo : shared = 1;
447 my $bar = 1;
448 threads->new(sub { $foo++; $bar++ })->join;
818c4caa 449
bfce6503
DM
450 print "$foo\n"; #prints 2 since $foo is shared
451 print "$bar\n"; #prints 1 since $bar is not shared
452
453In the case of a shared array, all the array's elements are shared, and for
454a shared hash, all the keys and values are shared. This places
455restrictions on what may be assigned to shared array and hash elements: only
456simple values or references to shared variables are allowed - this is
f3278b06 457so that a private variable can't accidentally become shared. A bad
bfce6503
DM
458assignment will cause the thread to die. For example:
459
460 use threads;
461 use threads::shared;
462
463 my $var = 1;
464 my $svar : shared = 2;
465 my %hash : shared;
466
467 ... create some threads ...
468
469 $hash{a} = 1; # all threads see exists($hash{a}) and $hash{a} == 1
56ca1794
IT
470 $hash{a} = $var # okay - copy-by-value: same effect as previous
471 $hash{a} = $svar # okay - copy-by-value: same effect as previous
bfce6503
DM
472 $hash{a} = \$svar # okay - a reference to a shared variable
473 $hash{a} = \$var # This will die
474 delete $hash{a} # okay - all threads will see !exists($hash{a})
475
476Note that a shared variable guarantees that if two or more threads try to
477modify it at the same time, the internal state of the variable will not
478become corrupted. However, there are no guarantees beyond this, as
479explained in the next section.
c975c451 480
6eded8f3 481=head2 Thread Pitfalls: Races
c975c451
AB
482
483While threads bring a new set of useful tools, they also bring a
484number of pitfalls. One pitfall is the race condition:
485
486 use threads;
487 use threads::shared;
bfce6503 488
c975c451
AB
489 my $a : shared = 1;
490 $thr1 = threads->new(\&sub1);
491 $thr2 = threads->new(\&sub2);
492
493 $thr1->join;
494 $thr2->join;
495 print "$a\n";
496
bfce6503
DM
497 sub sub1 { my $foo = $a; $a = $foo + 1; }
498 sub sub2 { my $bar = $a; $a = $bar + 1; }
c975c451
AB
499
500What do you think $a will be? The answer, unfortunately, is "it
501depends." Both sub1() and sub2() access the global variable $a, once
502to read and once to write. Depending on factors ranging from your
503thread implementation's scheduling algorithm to the phase of the moon,
504$a can be 2 or 3.
505
506Race conditions are caused by unsynchronized access to shared
507data. Without explicit synchronization, there's no way to be sure that
508nothing has happened to the shared data between the time you access it
509and the time you update it. Even this simple code fragment has the
510possibility of error:
511
512 use threads;
513 my $a : shared = 2;
514 my $b : shared;
515 my $c : shared;
516 my $thr1 = threads->create(sub { $b = $a; $a = $b + 1; });
517 my $thr2 = threads->create(sub { $c = $a; $a = $c + 1; });
8f95bfb9
EM
518 $thr1->join;
519 $thr2->join;
c975c451
AB
520
521Two threads both access $a. Each thread can potentially be interrupted
522at any point, or be executed in any order. At the end, $a could be 3
523or 4, and both $b and $c could be 2 or 3.
524
bfce6503
DM
525Even C<$a += 5> or C<$a++> are not guaranteed to be atomic.
526
c975c451
AB
527Whenever your program accesses data or resources that can be accessed
528by other threads, you must take steps to coordinate access or risk
bfce6503
DM
529data inconsistency and race conditions. Note that Perl will protect its
530internals from your race conditions, but it won't protect you from you.
531
f3278b06 532=head1 Synchronization and control
bfce6503
DM
533
534Perl provides a number of mechanisms to coordinate the interactions
535between themselves and their data, to avoid race conditions and the like.
536Some of these are designed to resemble the common techniques used in thread
537libraries such as C<pthreads>; others are Perl-specific. Often, the
9e75ef81 538standard techniques are clumsy and difficult to get right (such as
bfce6503
DM
539condition waits). Where possible, it is usually easier to use Perlish
540techniques such as queues, which remove some of the hard work involved.
c975c451
AB
541
542=head2 Controlling access: lock()
543
544The lock() function takes a shared variable and puts a lock on it.
a6d05634 545No other thread may lock the variable until the variable is unlocked
bfce6503 546by the thread holding the lock. Unlocking happens automatically
8f95bfb9 547when the locking thread exits the outermost block that contains
bfce6503 548C<lock()> function. Using lock() is straightforward: this example has
f3278b06 549several threads doing some calculations in parallel, and occasionally
bfce6503
DM
550updating a running total:
551
552 use threads;
553 use threads::shared;
554
555 my $total : shared = 0;
556
557 sub calc {
558 for (;;) {
559 my $result;
560 # (... do some calculations and set $result ...)
561 {
562 lock($total); # block until we obtain the lock
8f95bfb9 563 $total += $result;
f3278b06 564 } # lock implicitly released at end of scope
bfce6503
DM
565 last if $result == 0;
566 }
567 }
568
569 my $thr1 = threads->new(\&calc);
570 my $thr2 = threads->new(\&calc);
571 my $thr3 = threads->new(\&calc);
572 $thr1->join;
573 $thr2->join;
574 $thr3->join;
575 print "total=$total\n";
c975c451 576
c975c451
AB
577
578lock() blocks the thread until the variable being locked is
579available. When lock() returns, your thread can be sure that no other
bfce6503 580thread can lock that variable until the outermost block containing the
c975c451
AB
581lock exits.
582
583It's important to note that locks don't prevent access to the variable
584in question, only lock attempts. This is in keeping with Perl's
585longstanding tradition of courteous programming, and the advisory file
586locking that flock() gives you.
587
588You may lock arrays and hashes as well as scalars. Locking an array,
589though, will not block subsequent locks on array elements, just lock
590attempts on the array itself.
591
bfce6503 592Locks are recursive, which means it's okay for a thread to
c975c451 593lock a variable more than once. The lock will last until the outermost
bfce6503
DM
594lock() on the variable goes out of scope. For example:
595
596 my $x : shared;
597 doit();
598
599 sub doit {
600 {
601 {
602 lock($x); # wait for lock
8f95bfb9 603 lock($x); # NOOP - we already have the lock
bfce6503
DM
604 {
605 lock($x); # NOOP
606 {
607 lock($x); # NOOP
608 lockit_some_more();
609 }
610 }
611 } # *** implicit unlock here ***
612 }
613 }
614
615 sub lockit_some_more {
616 lock($x); # NOOP
617 } # nothing happens here
618
619Note that there is no unlock() function - the only way to unlock a
620variable is to allow it to go out of scope.
621
622A lock can either be used to guard the data contained within the variable
623being locked, or it can be used to guard something else, like a section
624of code. In this latter case, the variable in question does not hold any
625useful data, and exists only for the purpose of being locked. In this
626respect, the variable behaves like the mutexes and basic semaphores of
627traditional thread libraries.
c975c451 628
bfce6503 629=head2 A Thread Pitfall: Deadlocks
c975c451 630
bfce6503 631Locks are a handy tool to synchronize access to data, and using them
c975c451 632properly is the key to safe shared data. Unfortunately, locks aren't
f3278b06 633without their dangers, especially when multiple locks are involved.
bfce6503 634Consider the following code:
c975c451
AB
635
636 use threads;
bfce6503 637
c975c451
AB
638 my $a : shared = 4;
639 my $b : shared = "foo";
640 my $thr1 = threads->new(sub {
641 lock($a);
bfce6503 642 threads->yield;
c975c451 643 sleep 20;
bfce6503 644 lock($b);
c975c451
AB
645 });
646 my $thr2 = threads->new(sub {
647 lock($b);
bfce6503 648 threads->yield;
c975c451 649 sleep 20;
bfce6503 650 lock($a);
c975c451
AB
651 });
652
653This program will probably hang until you kill it. The only way it
bfce6503 654won't hang is if one of the two threads acquires both locks
c975c451
AB
655first. A guaranteed-to-hang version is more complicated, but the
656principle is the same.
657
bfce6503
DM
658The first thread will grab a lock on $a, then, after a pause during which
659the second thread has probably had time to do some work, try to grab a
660lock on $b. Meanwhile, the second thread grabs a lock on $b, then later
661tries to grab a lock on $a. The second lock attempt for both threads will
662block, each waiting for the other to release its lock.
c975c451
AB
663
664This condition is called a deadlock, and it occurs whenever two or
665more threads are trying to get locks on resources that the others
666own. Each thread will block, waiting for the other to release a lock
667on a resource. That never happens, though, since the thread with the
668resource is itself waiting for a lock to be released.
669
670There are a number of ways to handle this sort of problem. The best
671way is to always have all threads acquire locks in the exact same
672order. If, for example, you lock variables $a, $b, and $c, always lock
673$a before $b, and $b before $c. It's also best to hold on to locks for
674as short a period of time to minimize the risks of deadlock.
675
48b96218 676The other synchronization primitives described below can suffer from
bfce6503
DM
677similar problems.
678
c975c451
AB
679=head2 Queues: Passing Data Around
680
681A queue is a special thread-safe object that lets you put data in one
682end and take it out the other without having to worry about
683synchronization issues. They're pretty straightforward, and look like
684this:
685
686 use threads;
83272a45 687 use Thread::Queue;
c975c451 688
83272a45 689 my $DataQueue = Thread::Queue->new;
c975c451
AB
690 $thr = threads->new(sub {
691 while ($DataElement = $DataQueue->dequeue) {
692 print "Popped $DataElement off the queue\n";
693 }
694 });
695
696 $DataQueue->enqueue(12);
697 $DataQueue->enqueue("A", "B", "C");
698 $DataQueue->enqueue(\$thr);
699 sleep 10;
700 $DataQueue->enqueue(undef);
8f95bfb9 701 $thr->join;
c975c451 702
83272a45 703You create the queue with C<new Thread::Queue>. Then you can
6eded8f3
SG
704add lists of scalars onto the end with enqueue(), and pop scalars off
705the front of it with dequeue(). A queue has no fixed size, and can grow
706as needed to hold everything pushed on to it.
c975c451
AB
707
708If a queue is empty, dequeue() blocks until another thread enqueues
709something. This makes queues ideal for event loops and other
710communications between threads.
711
c975c451
AB
712=head2 Semaphores: Synchronizing Data Access
713
bfce6503
DM
714Semaphores are a kind of generic locking mechanism. In their most basic
715form, they behave very much like lockable scalars, except that thay
716can't hold data, and that they must be explicitly unlocked. In their
717advanced form, they act like a kind of counter, and can allow multiple
718threads to have the 'lock' at any one time.
2605996a 719
bfce6503 720=head2 Basic semaphores
2605996a 721
bfce6503
DM
722Semaphores have two methods, down() and up(): down() decrements the resource
723count, while up increments it. Calls to down() will block if the
c975c451
AB
724semaphore's current count would decrement below zero. This program
725gives a quick demonstration:
726
727 use threads qw(yield);
83272a45 728 use Thread::Semaphore;
bfce6503 729
83272a45 730 my $semaphore = new Thread::Semaphore;
bfce6503 731 my $GlobalVariable : shared = 0;
2605996a 732
c975c451
AB
733 $thr1 = new threads \&sample_sub, 1;
734 $thr2 = new threads \&sample_sub, 2;
735 $thr3 = new threads \&sample_sub, 3;
2605996a 736
c975c451
AB
737 sub sample_sub {
738 my $SubNumber = shift @_;
739 my $TryCount = 10;
740 my $LocalCopy;
741 sleep 1;
742 while ($TryCount--) {
743 $semaphore->down;
744 $LocalCopy = $GlobalVariable;
745 print "$TryCount tries left for sub $SubNumber (\$GlobalVariable is $GlobalVariable)\n";
746 yield;
747 sleep 2;
748 $LocalCopy++;
749 $GlobalVariable = $LocalCopy;
750 $semaphore->up;
751 }
752 }
6eded8f3 753
8f95bfb9
EM
754 $thr1->join;
755 $thr2->join;
756 $thr3->join;
2605996a 757
c975c451
AB
758The three invocations of the subroutine all operate in sync. The
759semaphore, though, makes sure that only one thread is accessing the
760global variable at once.
2605996a 761
bfce6503 762=head2 Advanced Semaphores
2605996a 763
c975c451
AB
764By default, semaphores behave like locks, letting only one thread
765down() them at a time. However, there are other uses for semaphores.
2605996a 766
6eded8f3
SG
767Each semaphore has a counter attached to it. By default, semaphores are
768created with the counter set to one, down() decrements the counter by
769one, and up() increments by one. However, we can override any or all
770of these defaults simply by passing in different values:
771
772 use threads;
83272a45
JH
773 use Thread::Semaphore;
774 my $semaphore = Thread::Semaphore->new(5);
6eded8f3
SG
775 # Creates a semaphore with the counter set to five
776
777 $thr1 = threads->new(\&sub1);
778 $thr2 = threads->new(\&sub1);
779
780 sub sub1 {
781 $semaphore->down(5); # Decrements the counter by five
782 # Do stuff here
783 $semaphore->up(5); # Increment the counter by five
784 }
785
8f95bfb9
EM
786 $thr1->detach;
787 $thr2->detach;
6eded8f3
SG
788
789If down() attempts to decrement the counter below zero, it blocks until
790the counter is large enough. Note that while a semaphore can be created
791with a starting count of zero, any up() or down() always changes the
792counter by at least one, and so $semaphore->down(0) is the same as
793$semaphore->down(1).
2605996a 794
c975c451
AB
795The question, of course, is why would you do something like this? Why
796create a semaphore with a starting count that's not one, or why
797decrement/increment it by more than one? The answer is resource
798availability. Many resources that you want to manage access for can be
799safely used by more than one thread at once.
2605996a 800
c975c451
AB
801For example, let's take a GUI driven program. It has a semaphore that
802it uses to synchronize access to the display, so only one thread is
803ever drawing at once. Handy, but of course you don't want any thread
804to start drawing until things are properly set up. In this case, you
805can create a semaphore with a counter set to zero, and up it when
806things are ready for drawing.
2605996a 807
c975c451
AB
808Semaphores with counters greater than one are also useful for
809establishing quotas. Say, for example, that you have a number of
810threads that can do I/O at once. You don't want all the threads
811reading or writing at once though, since that can potentially swamp
812your I/O channels, or deplete your process' quota of filehandles. You
813can use a semaphore initialized to the number of concurrent I/O
814requests (or open files) that you want at any one time, and have your
815threads quietly block and unblock themselves.
2605996a 816
c975c451
AB
817Larger increments or decrements are handy in those cases where a
818thread needs to check out or return a number of resources at once.
2605996a 819
bfce6503
DM
820=head2 cond_wait() and cond_signal()
821
822These two functions can be used in conjunction with locks to notify
823co-operating threads that a resource has become available. They are
824very similar in use to the functions found in C<pthreads>. However
825for most purposes, queues are simpler to use and more intuitive. See
826L<threads::shared> for more details.
2605996a 827
c975c451
AB
828=head1 General Thread Utility Routines
829
830We've covered the workhorse parts of Perl's threading package, and
831with these tools you should be well on your way to writing threaded
832code and packages. There are a few useful little pieces that didn't
833really fit in anyplace else.
834
835=head2 What Thread Am I In?
836
bfce6503
DM
837The C<< threads->self >> class method provides your program with a way to
838get an object representing the thread it's currently in. You can use this
6eded8f3 839object in the same way as the ones returned from thread creation.
c975c451
AB
840
841=head2 Thread IDs
842
843tid() is a thread object method that returns the thread ID of the
844thread the object represents. Thread IDs are integers, with the main
845thread in a program being 0. Currently Perl assigns a unique tid to
846every thread ever created in your program, assigning the first thread
847to be created a tid of 1, and increasing the tid by 1 for each new
848thread that's created.
849
850=head2 Are These Threads The Same?
851
852The equal() method takes two thread objects and returns true
853if the objects represent the same thread, and false if they don't.
854
855Thread objects also have an overloaded == comparison so that you can do
856comparison on them as you would with normal objects.
857
858=head2 What Threads Are Running?
859
bfce6503 860C<< threads->list >> returns a list of thread objects, one for each thread
c975c451
AB
861that's currently running and not detached. Handy for a number of things,
862including cleaning up at the end of your program:
863
864 # Loop through all the threads
865 foreach $thr (threads->list) {
866 # Don't join the main thread or ourselves
867 if ($thr->tid && !threads::equal($thr, threads->self)) {
868 $thr->join;
869 }
870 }
871
bfce6503
DM
872If some threads have not finished running when the main Perl thread
873ends, Perl will warn you about it and die, since it is impossible for Perl
6eded8f3 874to clean up itself while other threads are running
c975c451
AB
875
876=head1 A Complete Example
877
878Confused yet? It's time for an example program to show some of the
879things we've covered. This program finds prime numbers using threads.
880
881 1 #!/usr/bin/perl -w
882 2 # prime-pthread, courtesy of Tom Christiansen
883 3
884 4 use strict;
885 5
886 6 use threads;
83272a45 887 7 use Thread::Queue;
c975c451 888 8
83272a45 889 9 my $stream = new Thread::Queue;
c975c451
AB
890 10 my $kid = new threads(\&check_num, $stream, 2);
891 11
892 12 for my $i ( 3 .. 1000 ) {
893 13 $stream->enqueue($i);
894 14 }
895 15
896 16 $stream->enqueue(undef);
8f95bfb9 897 17 $kid->join;
c975c451
AB
898 18
899 19 sub check_num {
900 20 my ($upstream, $cur_prime) = @_;
901 21 my $kid;
83272a45 902 22 my $downstream = new Thread::Queue;
c975c451
AB
903 23 while (my $num = $upstream->dequeue) {
904 24 next unless $num % $cur_prime;
905 25 if ($kid) {
906 26 $downstream->enqueue($num);
907 27 } else {
908 28 print "Found prime $num\n";
909 29 $kid = new threads(\&check_num, $downstream, $num);
910 30 }
911 31 }
912 32 $downstream->enqueue(undef) if $kid;
8f95bfb9 913 33 $kid->join if $kid;
c975c451
AB
914 34 }
915
916This program uses the pipeline model to generate prime numbers. Each
917thread in the pipeline has an input queue that feeds numbers to be
918checked, a prime number that it's responsible for, and an output queue
9e75ef81 919into which it funnels numbers that have failed the check. If the thread
c975c451
AB
920has a number that's failed its check and there's no child thread, then
921the thread must have found a new prime number. In that case, a new
922child thread is created for that prime and stuck on the end of the
923pipeline.
924
6eded8f3 925This probably sounds a bit more confusing than it really is, so let's
c975c451
AB
926go through this program piece by piece and see what it does. (For
927those of you who might be trying to remember exactly what a prime
928number is, it's a number that's only evenly divisible by itself and 1)
929
930The bulk of the work is done by the check_num() subroutine, which
931takes a reference to its input queue and a prime number that it's
932responsible for. After pulling in the input queue and the prime that
933the subroutine's checking (line 20), we create a new queue (line 22)
934and reserve a scalar for the thread that we're likely to create later
935(line 21).
936
937The while loop from lines 23 to line 31 grabs a scalar off the input
938queue and checks against the prime this thread is responsible
939for. Line 24 checks to see if there's a remainder when we modulo the
940number to be checked against our prime. If there is one, the number
941must not be evenly divisible by our prime, so we need to either pass
942it on to the next thread if we've created one (line 26) or create a
943new thread if we haven't.
944
945The new thread creation is line 29. We pass on to it a reference to
946the queue we've created, and the prime number we've found.
947
948Finally, once the loop terminates (because we got a 0 or undef in the
949queue, which serves as a note to die), we pass on the notice to our
6eded8f3 950child and wait for it to exit if we've created a child (lines 32 and
c975c451
AB
95137).
952
953Meanwhile, back in the main thread, we create a queue (line 9) and the
954initial child thread (line 10), and pre-seed it with the first prime:
9552. Then we queue all the numbers from 3 to 1000 for checking (lines
95612-14), then queue a die notice (line 16) and wait for the first child
957thread to terminate (line 17). Because a child won't die until its
958child has died, we know that we're done once we return from the join.
959
960That's how it works. It's pretty simple; as with many Perl programs,
961the explanation is much longer than the program.
962
bfce6503
DM
963=head1 Performance considerations
964
965The main thing to bear in mind when comparing ithreads to other threading
966models is the fact that for each new thread created, a complete copy of
967all the variables and data of the parent thread has to be taken. Thus
968thread creation can be quite expensive, both in terms of memory usage and
969time spent in creation. The ideal way to reduce these costs is to have a
970relatively short number of long-lived threads, all created fairly early
971on - before the base thread has accumulated too much data. Of course, this
972may not always be possible, so compromises have to be made. However, after
973a thread has been created, its performance and extra memory usage should
974be little different than ordinary code.
975
976Also note that under the current implementation, shared variables
977use a little more memory and are a little slower than ordinary variables.
978
cf5baa48
JH
979=head1 Process-scope Changes
980
981Note that while threads themselves are separate execution threads and
982Perl data is thread-private unless explicitly shared, the threads can
983affect process-scope state, affecting all the threads.
984
985The most common example of this is changing the current working
986directory using chdir(). One thread calls chdir(), and the working
987directory of all the threads changes.
bdcfa4c7 988
cf5baa48
JH
989Even more drastic example of a process-scope change is chroot():
990the root directory of all the threads changes, and no thread can
991undo it (as opposed to chdir()).
992
993Further examples of process-scope changes include umask() and
994changing uids/gids.
995
996Thinking of mixing fork() and threads? Please lie down and wait
997until the feeling passes-- but in case you really want to know,
998the semantics is that fork() duplicates all the threads.
999(In UNIX, at least, other platforms will do something different.)
1000
b03ad8f6
JH
1001Similarly, mixing signals and threads should not be attempted.
1002Implementations are platform-dependent, and even the POSIX
1003semantics may not be what you expect (and Perl doesn't even
1004give you the full POSIX API).
1005
cf5baa48
JH
1006=head1 Thread-Safety of System Libraries
1007
1008Whether various library calls are thread-safe is outside the control
1009of Perl. Calls often suffering from not being thread-safe include:
bdcfa4c7 1010localtime(), gmtime(), get{gr,host,net,proto,serv,pw}*(), readdir(),
cf5baa48
JH
1011rand(), and srand() -- in general, calls that depend on some global
1012external state.
80bbcbc4 1013
cf5baa48 1014If the system Perl is compiled in has thread-safe variants of such
80bbcbc4 1015calls, they will be used. Beyond that, Perl is at the mercy of
cf5baa48 1016the thread-safety or -unsafety of the calls. Please consult your
80bbcbc4
JH
1017C library call documentation.
1018
af685957
JH
1019On some platforms the thread-safe library interfaces may fail if the
1020result buffer is too small (for example the user group databases may
1021be rather large, and the reentrant interfaces may have to carry around
1022a full snapshot of those databases). Perl will start with a small
1023buffer, but keep retrying and growing the result buffer
1024until the result fits. If this limitless growing sounds bad for
1025security or memory consumption reasons you can recompile Perl with
1026PERL_REENTRANT_MAXSIZE defined to the maximum number of bytes you will
1027allow.
bdcfa4c7 1028
c975c451
AB
1029=head1 Conclusion
1030
1031A complete thread tutorial could fill a book (and has, many times),
6eded8f3
SG
1032but with what we've covered in this introduction, you should be well
1033on your way to becoming a threaded Perl expert.
c975c451
AB
1034
1035=head1 Bibliography
1036
1037Here's a short bibliography courtesy of Jürgen Christoffel:
1038
1039=head2 Introductory Texts
1040
1041Birrell, Andrew D. An Introduction to Programming with
1042Threads. Digital Equipment Corporation, 1989, DEC-SRC Research Report
1043#35 online as
6eded8f3
SG
1044http://gatekeeper.dec.com/pub/DEC/SRC/research-reports/abstracts/src-rr-035.html
1045(highly recommended)
c975c451
AB
1046
1047Robbins, Kay. A., and Steven Robbins. Practical Unix Programming: A
1048Guide to Concurrency, Communication, and
1049Multithreading. Prentice-Hall, 1996.
1050
1051Lewis, Bill, and Daniel J. Berg. Multithreaded Programming with
1052Pthreads. Prentice Hall, 1997, ISBN 0-13-443698-9 (a well-written
1053introduction to threads).
1054
1055Nelson, Greg (editor). Systems Programming with Modula-3. Prentice
1056Hall, 1991, ISBN 0-13-590464-1.
1057
1058Nichols, Bradford, Dick Buttlar, and Jacqueline Proulx Farrell.
1059Pthreads Programming. O'Reilly & Associates, 1996, ISBN 156592-115-1
1060(covers POSIX threads).
1061
1062=head2 OS-Related References
1063
1064Boykin, Joseph, David Kirschen, Alan Langerman, and Susan
1065LoVerso. Programming under Mach. Addison-Wesley, 1994, ISBN
10660-201-52739-1.
1067
1068Tanenbaum, Andrew S. Distributed Operating Systems. Prentice Hall,
10691995, ISBN 0-13-219908-4 (great textbook).
1070
1071Silberschatz, Abraham, and Peter B. Galvin. Operating System Concepts,
10724th ed. Addison-Wesley, 1995, ISBN 0-201-59292-4
1073
1074=head2 Other References
1075
1076Arnold, Ken and James Gosling. The Java Programming Language, 2nd
1077ed. Addison-Wesley, 1998, ISBN 0-201-31006-6.
1078
b03ad8f6
JH
1079comp.programming.threads FAQ,
1080L<http://www.serpentine.com/~bos/threads-faq/>
1081
c975c451
AB
1082Le Sergent, T. and B. Berthomieu. "Incremental MultiThreaded Garbage
1083Collection on Virtually Shared Memory Architectures" in Memory
1084Management: Proc. of the International Workshop IWMM 92, St. Malo,
1085France, September 1992, Yves Bekkers and Jacques Cohen, eds. Springer,
10861992, ISBN 3540-55940-X (real-life thread applications).
1087
5e549d84
JH
1088Artur Bergman, "Where Wizards Fear To Tread", June 11, 2002,
1089L<http://www.perl.com/pub/a/2002/06/11/threads.html>
1090
c975c451
AB
1091=head1 Acknowledgements
1092
1093Thanks (in no particular order) to Chaim Frenkel, Steve Fink, Gurusamy
1094Sarathy, Ilya Zakharevich, Benjamin Sugars, Jürgen Christoffel, Joshua
1095Pritikin, and Alan Burlison, for their help in reality-checking and
1096polishing this article. Big thanks to Tom Christiansen for his rewrite
1097of the prime number generator.
1098
1099=head1 AUTHOR
1100
9316ed2f 1101Dan Sugalski E<lt>dan@sidhe.org<gt>
c975c451
AB
1102
1103Slightly modified by Arthur Bergman to fit the new thread model/module.
1104
cf5baa48
JH
1105Reworked slightly by Jörg Walter E<lt>jwalt@cpan.org<gt> to be more concise
1106about thread-safety of perl code.
1107
c975c451
AB
1108=head1 Copyrights
1109
bfce6503
DM
1110The original version of this article originally appeared in The Perl
1111Journal #10, and is copyright 1998 The Perl Journal. It appears courtesy
1112of Jon Orwant and The Perl Journal. This document may be distributed
1113under the same terms as Perl itself.
2605996a 1114
53d7eaa8 1115For more information please see L<threads> and L<threads::shared>.
2605996a 1116