5 perlthrtut - Tutorial on threads in Perl
9 This tutorial describes the use of Perl interpreter threads (sometimes
10 referred to as I<ithreads>) that was first introduced in Perl 5.6.0. In this
11 model, each thread runs in its own Perl interpreter, and any data sharing
12 between threads must be explicit. The user-level interface for I<ithreads>
13 uses the L<threads> class.
15 B<NOTE>: There was another older Perl threading flavor called the 5.005 model
16 that used the L<Threads> class. This old model was known to have problems, is
17 deprecated, and was removed for release 5.10. You are
18 strongly encouraged to migrate any existing 5.005 threads code to the new
19 model as soon as possible.
21 You can see which (or neither) threading flavour you have by
22 running C<perl -V> and looking at the C<Platform> section.
23 If you have C<useithreads=define> you have ithreads, if you
24 have C<use5005threads=define> you have 5.005 threads.
25 If you have neither, you don't have any thread support built in.
26 If you have both, you are in trouble.
28 The L<threads> and L<threads::shared> modules are included in the core Perl
29 distribution. Additionally, they are maintained as a separate modules on
30 CPAN, so you can check there for any updates.
32 =head1 What Is A Thread Anyway?
34 A thread is a flow of control through a program with a single
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.
43 =head1 Threaded Program Models
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.
52 The boss/worker model usually has one I<boss> thread and one or more
53 I<worker> threads. The boss thread gathers or generates tasks that need
54 to be done, then parcels those tasks out to the appropriate worker
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.
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.
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
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
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.
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
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
102 =head1 What kind of threads are Perl threads?
104 If you have experience with other thread implementations, you might
105 find that things aren't quite what you expect. It's very important to
106 remember when dealing with Perl threads that I<Perl Threads Are Not X
107 Threads> for all values of X. They aren't POSIX threads, or
108 DecThreads, or Java's Green threads, or Win32 threads. There are
109 similarities, and the broad concepts are the same, but if you start
110 looking for implementation details you're going to be either
111 disappointed or confused. Possibly both.
113 This is not to say that Perl threads are completely different from
114 everything that's ever come before. They're not. Perl's threading
115 model owes a lot to other thread models, especially POSIX. Just as
116 Perl is not C, though, Perl threads are not POSIX threads. So if you
117 find yourself looking for mutexes, or thread priorities, it's time to
118 step back a bit and think about what you want to do and how Perl can
121 However, it is important to remember that Perl threads cannot magically
122 do things unless your operating system's threads allow it. So if your
123 system blocks the entire process on C<sleep()>, Perl usually will, as well.
125 B<Perl Threads Are Different.>
127 =head1 Thread-Safe Modules
129 The addition of threads has changed Perl's internals
130 substantially. There are implications for people who write
131 modules with XS code or external libraries. However, since Perl data is
132 not shared among threads by default, Perl modules stand a high chance of
133 being thread-safe or can be made thread-safe easily. Modules that are not
134 tagged as thread-safe should be tested or code reviewed before being used
137 Not all modules that you might use are thread-safe, and you should
138 always assume a module is unsafe unless the documentation says
139 otherwise. This includes modules that are distributed as part of the
140 core. Threads are a relatively new feature, and even some of the standard
141 modules aren't thread-safe.
143 Even if a module is thread-safe, it doesn't mean that the module is optimized
144 to work well with threads. A module could possibly be rewritten to utilize
145 the new features in threaded Perl to increase performance in a threaded
148 If you're using a module that's not thread-safe for some reason, you
149 can protect yourself by using it from one, and only one thread at all.
150 If you need multiple threads to access such a module, you can use semaphores and
151 lots of programming discipline to control access to it. Semaphores
152 are covered in L</"Basic semaphores">.
154 See also L</"Thread-Safety of System Libraries">.
158 The L<threads> module provides the basic functions you need to write
159 threaded programs. In the following sections, we'll cover the basics,
160 showing you what you need to do to create a threaded program. After
161 that, we'll go over some of the features of the L<threads> module that
162 make threaded programming easier.
164 =head2 Basic Thread Support
166 Thread support is a Perl compile-time option. It's something that's
167 turned on or off when Perl is built at your site, rather than when
168 your programs are compiled. If your Perl wasn't compiled with thread
169 support enabled, then any attempt to use threads will fail.
171 Your programs can use the Config module to check whether threads are
172 enabled. If your program can't run without them, you can say something
176 $Config{useithreads} or die('Recompile Perl with threads to run this program.');
178 A possibly-threaded program using a possibly-threaded module might
185 if ($Config{useithreads}) {
187 require MyMod_threaded;
188 import MyMod_threaded;
190 require MyMod_unthreaded;
191 import MyMod_unthreaded;
195 Since code that runs both with and without threads is usually pretty
196 messy, it's best to isolate the thread-specific code in its own
197 module. In our example above, that's what C<MyMod_threaded> is, and it's
198 only imported if we're running on a threaded Perl.
200 =head2 A Note about the Examples
202 In a real situation, care should be taken that all threads are finished
203 executing before the program exits. That care has B<not> been taken in these
204 examples in the interest of simplicity. Running these examples I<as is> will
205 produce error messages, usually caused by the fact that there are still
206 threads running when the program exits. You should not be alarmed by this.
208 =head2 Creating Threads
210 The L<threads> module provides the tools you need to create new
211 threads. Like any other module, you need to tell Perl that you want to use
212 it; C<use threads;> imports all the pieces you need to create basic
215 The simplest, most straightforward way to create a thread is with C<create()>:
219 my $thr = threads->create(\&sub1);
222 print("In the thread\n");
225 The C<create()> method takes a reference to a subroutine and creates a new
226 thread that starts executing in the referenced subroutine. Control
227 then passes both to the subroutine and the caller.
229 If you need to, your program can pass parameters to the subroutine as
230 part of the thread startup. Just include the list of parameters as
231 part of the C<threads-E<gt>create()> call, like this:
236 my $thr1 = threads->create(\&sub1, 'Param 1', 'Param 2', $Param3);
237 my @ParamList = (42, 'Hello', 3.14);
238 my $thr2 = threads->create(\&sub1, @ParamList);
239 my $thr3 = threads->create(\&sub1, qw(Param1 Param2 Param3));
242 my @InboundParameters = @_;
243 print("In the thread\n");
244 print('Got parameters >', join('<>', @InboundParameters), "<\n");
247 The last example illustrates another feature of threads. You can spawn
248 off several threads using the same subroutine. Each thread executes
249 the same subroutine, but in a separate thread with a separate
250 environment and potentially separate arguments.
252 C<new()> is a synonym for C<create()>.
254 =head2 Waiting For A Thread To Exit
256 Since threads are also subroutines, they can return values. To wait
257 for a thread to exit and extract any values it might return, you can
258 use the C<join()> method:
262 my ($thr) = threads->create(\&sub1);
264 my @ReturnData = $thr->join();
265 print('Thread returned ', join(', ', @ReturnData), "\n");
267 sub sub1 { return ('Fifty-six', 'foo', 2); }
269 In the example above, the C<join()> method returns as soon as the thread
270 ends. In addition to waiting for a thread to finish and gathering up
271 any values that the thread might have returned, C<join()> also performs
272 any OS cleanup necessary for the thread. That cleanup might be
273 important, especially for long-running programs that spawn lots of
274 threads. If you don't want the return values and don't want to wait
275 for the thread to finish, you should call the C<detach()> method
276 instead, as described next.
278 NOTE: In the example above, the thread returns a list, thus necessitating
279 that the thread creation call be made in list context (i.e., C<my ($thr)>).
280 See L<< threads/"$thr->join()" >> and L<threads/"THREAD CONTEXT"> for more
281 details on thread context and return values.
283 =head2 Ignoring A Thread
285 C<join()> does three things: it waits for a thread to exit, cleans up
286 after it, and returns any data the thread may have produced. But what
287 if you're not interested in the thread's return values, and you don't
288 really care when the thread finishes? All you want is for the thread
289 to get cleaned up after when it's done.
291 In this case, you use the C<detach()> method. Once a thread is detached,
292 it'll run until it's finished; then Perl will clean up after it
297 my $thr = threads->create(\&sub1); # Spawn the thread
299 $thr->detach(); # Now we officially don't care any more
301 sleep(15); # Let thread run for awhile
307 print("\$a is $a\n");
312 Once a thread is detached, it may not be joined, and any return data
313 that it might have produced (if it was done and waiting for a join) is
316 C<detach()> can also be called as a class method to allow a thread to
321 my $thr = threads->create(\&sub1);
328 =head2 Process and Thread Termination
330 With threads one must be careful to make sure they all have a chance to
331 run to completion, assuming that is what you want.
333 An action that terminates a process will terminate I<all> running
334 threads. die() and exit() have this property,
335 and perl does an exit when the main thread exits,
336 perhaps implicitly by falling off the end of your code,
337 even if that's not what you want.
339 As an example of this case, this code prints the message
340 "Perl exited with active threads: 2 running and unjoined":
343 my $thr1 = threads->new(\&thrsub, "test1");
344 my $thr2 = threads->new(\&thrsub, "test2");
348 print "thread $message\n";
351 But when the following lines are added at the end:
356 it prints two lines of output, a perhaps more useful outcome.
358 =head1 Threads And Data
360 Now that we've covered the basics of threads, it's time for our next
361 topic: Data. Threading introduces a couple of complications to data
362 access that non-threaded programs never need to worry about.
364 =head2 Shared And Unshared Data
366 The biggest difference between Perl I<ithreads> and the old 5.005 style
367 threading, or for that matter, to most other threading systems out there,
368 is that by default, no data is shared. When a new Perl thread is created,
369 all the data associated with the current thread is copied to the new
370 thread, and is subsequently private to that new thread!
371 This is similar in feel to what happens when a Unix process forks,
372 except that in this case, the data is just copied to a different part of
373 memory within the same process rather than a real fork taking place.
375 To make use of threading, however, one usually wants the threads to share
376 at least some data between themselves. This is done with the
377 L<threads::shared> module and the C<:shared> attribute:
384 threads->create(sub { $foo++; $bar++; })->join();
386 print("$foo\n"); # Prints 2 since $foo is shared
387 print("$bar\n"); # Prints 1 since $bar is not shared
389 In the case of a shared array, all the array's elements are shared, and for
390 a shared hash, all the keys and values are shared. This places
391 restrictions on what may be assigned to shared array and hash elements: only
392 simple values or references to shared variables are allowed - this is
393 so that a private variable can't accidentally become shared. A bad
394 assignment will cause the thread to die. For example:
400 my $svar :shared = 2;
403 ... create some threads ...
405 $hash{a} = 1; # All threads see exists($hash{a}) and $hash{a} == 1
406 $hash{a} = $var; # okay - copy-by-value: same effect as previous
407 $hash{a} = $svar; # okay - copy-by-value: same effect as previous
408 $hash{a} = \$svar; # okay - a reference to a shared variable
409 $hash{a} = \$var; # This will die
410 delete($hash{a}); # okay - all threads will see !exists($hash{a})
412 Note that a shared variable guarantees that if two or more threads try to
413 modify it at the same time, the internal state of the variable will not
414 become corrupted. However, there are no guarantees beyond this, as
415 explained in the next section.
417 =head2 Thread Pitfalls: Races
419 While threads bring a new set of useful tools, they also bring a
420 number of pitfalls. One pitfall is the race condition:
426 my $thr1 = threads->create(\&sub1);
427 my $thr2 = threads->create(\&sub2);
433 sub sub1 { my $foo = $a; $a = $foo + 1; }
434 sub sub2 { my $bar = $a; $a = $bar + 1; }
436 What do you think C<$a> will be? The answer, unfortunately, is I<it
437 depends>. Both C<sub1()> and C<sub2()> access the global variable C<$a>, once
438 to read and once to write. Depending on factors ranging from your
439 thread implementation's scheduling algorithm to the phase of the moon,
442 Race conditions are caused by unsynchronized access to shared
443 data. Without explicit synchronization, there's no way to be sure that
444 nothing has happened to the shared data between the time you access it
445 and the time you update it. Even this simple code fragment has the
446 possibility of error:
452 my $thr1 = threads->create(sub { $b = $a; $a = $b + 1; });
453 my $thr2 = threads->create(sub { $c = $a; $a = $c + 1; });
457 Two threads both access C<$a>. Each thread can potentially be interrupted
458 at any point, or be executed in any order. At the end, C<$a> could be 3
459 or 4, and both C<$b> and C<$c> could be 2 or 3.
461 Even C<$a += 5> or C<$a++> are not guaranteed to be atomic.
463 Whenever your program accesses data or resources that can be accessed
464 by other threads, you must take steps to coordinate access or risk
465 data inconsistency and race conditions. Note that Perl will protect its
466 internals from your race conditions, but it won't protect you from you.
468 =head1 Synchronization and control
470 Perl provides a number of mechanisms to coordinate the interactions
471 between themselves and their data, to avoid race conditions and the like.
472 Some of these are designed to resemble the common techniques used in thread
473 libraries such as C<pthreads>; others are Perl-specific. Often, the
474 standard techniques are clumsy and difficult to get right (such as
475 condition waits). Where possible, it is usually easier to use Perlish
476 techniques such as queues, which remove some of the hard work involved.
478 =head2 Controlling access: lock()
480 The C<lock()> function takes a shared variable and puts a lock on it.
481 No other thread may lock the variable until the variable is unlocked
482 by the thread holding the lock. Unlocking happens automatically
483 when the locking thread exits the block that contains the call to the
484 C<lock()> function. Using C<lock()> is straightforward: This example has
485 several threads doing some calculations in parallel, and occasionally
486 updating a running total:
491 my $total :shared = 0;
496 # (... do some calculations and set $result ...)
498 lock($total); # Block until we obtain the lock
500 } # Lock implicitly released at end of scope
501 last if $result == 0;
505 my $thr1 = threads->create(\&calc);
506 my $thr2 = threads->create(\&calc);
507 my $thr3 = threads->create(\&calc);
511 print("total=$total\n");
513 C<lock()> blocks the thread until the variable being locked is
514 available. When C<lock()> returns, your thread can be sure that no other
515 thread can lock that variable until the block containing the
518 It's important to note that locks don't prevent access to the variable
519 in question, only lock attempts. This is in keeping with Perl's
520 longstanding tradition of courteous programming, and the advisory file
521 locking that C<flock()> gives you.
523 You may lock arrays and hashes as well as scalars. Locking an array,
524 though, will not block subsequent locks on array elements, just lock
525 attempts on the array itself.
527 Locks are recursive, which means it's okay for a thread to
528 lock a variable more than once. The lock will last until the outermost
529 C<lock()> on the variable goes out of scope. For example:
537 lock($x); # Wait for lock
538 lock($x); # NOOP - we already have the lock
546 } # *** Implicit unlock here ***
550 sub lockit_some_more {
552 } # Nothing happens here
554 Note that there is no C<unlock()> function - the only way to unlock a
555 variable is to allow it to go out of scope.
557 A lock can either be used to guard the data contained within the variable
558 being locked, or it can be used to guard something else, like a section
559 of code. In this latter case, the variable in question does not hold any
560 useful data, and exists only for the purpose of being locked. In this
561 respect, the variable behaves like the mutexes and basic semaphores of
562 traditional thread libraries.
564 =head2 A Thread Pitfall: Deadlocks
566 Locks are a handy tool to synchronize access to data, and using them
567 properly is the key to safe shared data. Unfortunately, locks aren't
568 without their dangers, especially when multiple locks are involved.
569 Consider the following code:
574 my $b :shared = 'foo';
575 my $thr1 = threads->create(sub {
580 my $thr2 = threads->create(sub {
586 This program will probably hang until you kill it. The only way it
587 won't hang is if one of the two threads acquires both locks
588 first. A guaranteed-to-hang version is more complicated, but the
589 principle is the same.
591 The first thread will grab a lock on C<$a>, then, after a pause during which
592 the second thread has probably had time to do some work, try to grab a
593 lock on C<$b>. Meanwhile, the second thread grabs a lock on C<$b>, then later
594 tries to grab a lock on C<$a>. The second lock attempt for both threads will
595 block, each waiting for the other to release its lock.
597 This condition is called a deadlock, and it occurs whenever two or
598 more threads are trying to get locks on resources that the others
599 own. Each thread will block, waiting for the other to release a lock
600 on a resource. That never happens, though, since the thread with the
601 resource is itself waiting for a lock to be released.
603 There are a number of ways to handle this sort of problem. The best
604 way is to always have all threads acquire locks in the exact same
605 order. If, for example, you lock variables C<$a>, C<$b>, and C<$c>, always lock
606 C<$a> before C<$b>, and C<$b> before C<$c>. It's also best to hold on to locks for
607 as short a period of time to minimize the risks of deadlock.
609 The other synchronization primitives described below can suffer from
612 =head2 Queues: Passing Data Around
614 A queue is a special thread-safe object that lets you put data in one
615 end and take it out the other without having to worry about
616 synchronization issues. They're pretty straightforward, and look like
622 my $DataQueue = Thread::Queue->new();
623 my $thr = threads->create(sub {
624 while (my $DataElement = $DataQueue->dequeue()) {
625 print("Popped $DataElement off the queue\n");
629 $DataQueue->enqueue(12);
630 $DataQueue->enqueue("A", "B", "C");
632 $DataQueue->enqueue(undef);
635 You create the queue with C<Thread::Queue-E<gt>new()>. Then you can
636 add lists of scalars onto the end with C<enqueue()>, and pop scalars off
637 the front of it with C<dequeue()>. A queue has no fixed size, and can grow
638 as needed to hold everything pushed on to it.
640 If a queue is empty, C<dequeue()> blocks until another thread enqueues
641 something. This makes queues ideal for event loops and other
642 communications between threads.
644 =head2 Semaphores: Synchronizing Data Access
646 Semaphores are a kind of generic locking mechanism. In their most basic
647 form, they behave very much like lockable scalars, except that they
648 can't hold data, and that they must be explicitly unlocked. In their
649 advanced form, they act like a kind of counter, and can allow multiple
650 threads to have the I<lock> at any one time.
652 =head2 Basic semaphores
654 Semaphores have two methods, C<down()> and C<up()>: C<down()> decrements the resource
655 count, while C<up()> increments it. Calls to C<down()> will block if the
656 semaphore's current count would decrement below zero. This program
657 gives a quick demonstration:
660 use Thread::Semaphore;
662 my $semaphore = Thread::Semaphore->new();
663 my $GlobalVariable :shared = 0;
665 $thr1 = threads->create(\&sample_sub, 1);
666 $thr2 = threads->create(\&sample_sub, 2);
667 $thr3 = threads->create(\&sample_sub, 3);
670 my $SubNumber = shift(@_);
674 while ($TryCount--) {
676 $LocalCopy = $GlobalVariable;
677 print("$TryCount tries left for sub $SubNumber (\$GlobalVariable is $GlobalVariable)\n");
680 $GlobalVariable = $LocalCopy;
689 The three invocations of the subroutine all operate in sync. The
690 semaphore, though, makes sure that only one thread is accessing the
691 global variable at once.
693 =head2 Advanced Semaphores
695 By default, semaphores behave like locks, letting only one thread
696 C<down()> them at a time. However, there are other uses for semaphores.
698 Each semaphore has a counter attached to it. By default, semaphores are
699 created with the counter set to one, C<down()> decrements the counter by
700 one, and C<up()> increments by one. However, we can override any or all
701 of these defaults simply by passing in different values:
704 use Thread::Semaphore;
706 my $semaphore = Thread::Semaphore->new(5);
707 # Creates a semaphore with the counter set to five
709 my $thr1 = threads->create(\&sub1);
710 my $thr2 = threads->create(\&sub1);
713 $semaphore->down(5); # Decrements the counter by five
715 $semaphore->up(5); # Increment the counter by five
721 If C<down()> attempts to decrement the counter below zero, it blocks until
722 the counter is large enough. Note that while a semaphore can be created
723 with a starting count of zero, any C<up()> or C<down()> always changes the
724 counter by at least one, and so C<< $semaphore->down(0) >> is the same as
725 C<< $semaphore->down(1) >>.
727 The question, of course, is why would you do something like this? Why
728 create a semaphore with a starting count that's not one, or why
729 decrement or increment it by more than one? The answer is resource
730 availability. Many resources that you want to manage access for can be
731 safely used by more than one thread at once.
733 For example, let's take a GUI driven program. It has a semaphore that
734 it uses to synchronize access to the display, so only one thread is
735 ever drawing at once. Handy, but of course you don't want any thread
736 to start drawing until things are properly set up. In this case, you
737 can create a semaphore with a counter set to zero, and up it when
738 things are ready for drawing.
740 Semaphores with counters greater than one are also useful for
741 establishing quotas. Say, for example, that you have a number of
742 threads that can do I/O at once. You don't want all the threads
743 reading or writing at once though, since that can potentially swamp
744 your I/O channels, or deplete your process's quota of filehandles. You
745 can use a semaphore initialized to the number of concurrent I/O
746 requests (or open files) that you want at any one time, and have your
747 threads quietly block and unblock themselves.
749 Larger increments or decrements are handy in those cases where a
750 thread needs to check out or return a number of resources at once.
752 =head2 Waiting for a Condition
754 The functions C<cond_wait()> and C<cond_signal()>
755 can be used in conjunction with locks to notify
756 co-operating threads that a resource has become available. They are
757 very similar in use to the functions found in C<pthreads>. However
758 for most purposes, queues are simpler to use and more intuitive. See
759 L<threads::shared> for more details.
761 =head2 Giving up control
763 There are times when you may find it useful to have a thread
764 explicitly give up the CPU to another thread. You may be doing something
765 processor-intensive and want to make sure that the user-interface thread
766 gets called frequently. Regardless, there are times that you might want
767 a thread to give up the processor.
769 Perl's threading package provides the C<yield()> function that does
770 this. C<yield()> is pretty straightforward, and works like this:
777 while($foo--) { print("In thread $thread\n"); }
780 while($foo--) { print("In thread $thread\n"); }
783 my $thr1 = threads->create(\&loop, 'first');
784 my $thr2 = threads->create(\&loop, 'second');
785 my $thr3 = threads->create(\&loop, 'third');
787 It is important to remember that C<yield()> is only a hint to give up the CPU,
788 it depends on your hardware, OS and threading libraries what actually happens.
789 B<On many operating systems, yield() is a no-op.> Therefore it is important
790 to note that one should not build the scheduling of the threads around
791 C<yield()> calls. It might work on your platform but it won't work on another
794 =head1 General Thread Utility Routines
796 We've covered the workhorse parts of Perl's threading package, and
797 with these tools you should be well on your way to writing threaded
798 code and packages. There are a few useful little pieces that didn't
799 really fit in anyplace else.
801 =head2 What Thread Am I In?
803 The C<threads-E<gt>self()> class method provides your program with a way to
804 get an object representing the thread it's currently in. You can use this
805 object in the same way as the ones returned from thread creation.
809 C<tid()> is a thread object method that returns the thread ID of the
810 thread the object represents. Thread IDs are integers, with the main
811 thread in a program being 0. Currently Perl assigns a unique TID to
812 every thread ever created in your program, assigning the first thread
813 to be created a TID of 1, and increasing the TID by 1 for each new
814 thread that's created. When used as a class method, C<threads-E<gt>tid()>
815 can be used by a thread to get its own TID.
817 =head2 Are These Threads The Same?
819 The C<equal()> method takes two thread objects and returns true
820 if the objects represent the same thread, and false if they don't.
822 Thread objects also have an overloaded C<==> comparison so that you can do
823 comparison on them as you would with normal objects.
825 =head2 What Threads Are Running?
827 C<threads-E<gt>list()> returns a list of thread objects, one for each thread
828 that's currently running and not detached. Handy for a number of things,
829 including cleaning up at the end of your program (from the main Perl thread,
832 # Loop through all the threads
833 foreach my $thr (threads->list()) {
837 If some threads have not finished running when the main Perl thread
838 ends, Perl will warn you about it and die, since it is impossible for Perl
839 to clean up itself while other threads are running.
841 NOTE: The main Perl thread (thread 0) is in a I<detached> state, and so
842 does not appear in the list returned by C<threads-E<gt>list()>.
844 =head1 A Complete Example
846 Confused yet? It's time for an example program to show some of the
847 things we've covered. This program finds prime numbers using threads.
850 2 # prime-pthread, courtesy of Tom Christiansen
859 11 my ($upstream, $cur_prime) = @_;
861 13 my $downstream = Thread::Queue->new();
862 14 while (my $num = $upstream->dequeue()) {
863 15 next unless ($num % $cur_prime);
865 17 $downstream->enqueue($num);
867 19 print("Found prime: $num\n");
868 20 $kid = threads->create(\&check_num, $downstream, $num);
870 22 warn("Sorry. Ran out of threads.\n");
876 28 $downstream->enqueue(undef);
881 33 my $stream = Thread::Queue->new(3..1000, undef);
882 34 check_num($stream, 2);
884 This program uses the pipeline model to generate prime numbers. Each
885 thread in the pipeline has an input queue that feeds numbers to be
886 checked, a prime number that it's responsible for, and an output queue
887 into which it funnels numbers that have failed the check. If the thread
888 has a number that's failed its check and there's no child thread, then
889 the thread must have found a new prime number. In that case, a new
890 child thread is created for that prime and stuck on the end of the
893 This probably sounds a bit more confusing than it really is, so let's
894 go through this program piece by piece and see what it does. (For
895 those of you who might be trying to remember exactly what a prime
896 number is, it's a number that's only evenly divisible by itself and 1.)
898 The bulk of the work is done by the C<check_num()> subroutine, which
899 takes a reference to its input queue and a prime number that it's
900 responsible for. After pulling in the input queue and the prime that
901 the subroutine is checking (line 11), we create a new queue (line 13)
902 and reserve a scalar for the thread that we're likely to create later
905 The while loop from line 14 to line 26 grabs a scalar off the input
906 queue and checks against the prime this thread is responsible
907 for. Line 15 checks to see if there's a remainder when we divide the
908 number to be checked by our prime. If there is one, the number
909 must not be evenly divisible by our prime, so we need to either pass
910 it on to the next thread if we've created one (line 17) or create a
911 new thread if we haven't.
913 The new thread creation is line 20. We pass on to it a reference to
914 the queue we've created, and the prime number we've found. In lines 21
915 through 24, we check to make sure that our new thread got created, and
916 if not, we stop checking any remaining numbers in the queue.
918 Finally, once the loop terminates (because we got a 0 or C<undef> in the
919 queue, which serves as a note to terminate), we pass on the notice to our
920 child, and wait for it to exit if we've created a child (lines 27 and
923 Meanwhile, back in the main thread, we first create a queue (line 33) and
924 queue up all the numbers from 3 to 1000 for checking, plus a termination
925 notice. Then all we have to do to get the ball rolling is pass the queue
926 and the first prime to the C<check_num()> subroutine (line 34).
928 That's how it works. It's pretty simple; as with many Perl programs,
929 the explanation is much longer than the program.
931 =head1 Different implementations of threads
933 Some background on thread implementations from the operating system
934 viewpoint. There are three basic categories of threads: user-mode threads,
935 kernel threads, and multiprocessor kernel threads.
937 User-mode threads are threads that live entirely within a program and
938 its libraries. In this model, the OS knows nothing about threads. As
939 far as it's concerned, your process is just a process.
941 This is the easiest way to implement threads, and the way most OSes
942 start. The big disadvantage is that, since the OS knows nothing about
943 threads, if one thread blocks they all do. Typical blocking activities
944 include most system calls, most I/O, and things like C<sleep()>.
946 Kernel threads are the next step in thread evolution. The OS knows
947 about kernel threads, and makes allowances for them. The main
948 difference between a kernel thread and a user-mode thread is
949 blocking. With kernel threads, things that block a single thread don't
950 block other threads. This is not the case with user-mode threads,
951 where the kernel blocks at the process level and not the thread level.
953 This is a big step forward, and can give a threaded program quite a
954 performance boost over non-threaded programs. Threads that block
955 performing I/O, for example, won't block threads that are doing other
956 things. Each process still has only one thread running at once,
957 though, regardless of how many CPUs a system might have.
959 Since kernel threading can interrupt a thread at any time, they will
960 uncover some of the implicit locking assumptions you may make in your
961 program. For example, something as simple as C<$a = $a + 2> can behave
962 unpredictably with kernel threads if C<$a> is visible to other
963 threads, as another thread may have changed C<$a> between the time it
964 was fetched on the right hand side and the time the new value is
967 Multiprocessor kernel threads are the final step in thread
968 support. With multiprocessor kernel threads on a machine with multiple
969 CPUs, the OS may schedule two or more threads to run simultaneously on
972 This can give a serious performance boost to your threaded program,
973 since more than one thread will be executing at the same time. As a
974 tradeoff, though, any of those nagging synchronization issues that
975 might not have shown with basic kernel threads will appear with a
978 In addition to the different levels of OS involvement in threads,
979 different OSes (and different thread implementations for a particular
980 OS) allocate CPU cycles to threads in different ways.
982 Cooperative multitasking systems have running threads give up control
983 if one of two things happen. If a thread calls a yield function, it
984 gives up control. It also gives up control if the thread does
985 something that would cause it to block, such as perform I/O. In a
986 cooperative multitasking implementation, one thread can starve all the
987 others for CPU time if it so chooses.
989 Preemptive multitasking systems interrupt threads at regular intervals
990 while the system decides which thread should run next. In a preemptive
991 multitasking system, one thread usually won't monopolize the CPU.
993 On some systems, there can be cooperative and preemptive threads
994 running simultaneously. (Threads running with realtime priorities
995 often behave cooperatively, for example, while threads running at
996 normal priorities behave preemptively.)
998 Most modern operating systems support preemptive multitasking nowadays.
1000 =head1 Performance considerations
1002 The main thing to bear in mind when comparing Perl's I<ithreads> to other threading
1003 models is the fact that for each new thread created, a complete copy of
1004 all the variables and data of the parent thread has to be taken. Thus,
1005 thread creation can be quite expensive, both in terms of memory usage and
1006 time spent in creation. The ideal way to reduce these costs is to have a
1007 relatively short number of long-lived threads, all created fairly early
1008 on (before the base thread has accumulated too much data). Of course, this
1009 may not always be possible, so compromises have to be made. However, after
1010 a thread has been created, its performance and extra memory usage should
1011 be little different than ordinary code.
1013 Also note that under the current implementation, shared variables
1014 use a little more memory and are a little slower than ordinary variables.
1016 =head1 Process-scope Changes
1018 Note that while threads themselves are separate execution threads and
1019 Perl data is thread-private unless explicitly shared, the threads can
1020 affect process-scope state, affecting all the threads.
1022 The most common example of this is changing the current working
1023 directory using C<chdir()>. One thread calls C<chdir()>, and the working
1024 directory of all the threads changes.
1026 Even more drastic example of a process-scope change is C<chroot()>:
1027 the root directory of all the threads changes, and no thread can
1028 undo it (as opposed to C<chdir()>).
1030 Further examples of process-scope changes include C<umask()> and
1031 changing uids and gids.
1033 Thinking of mixing C<fork()> and threads? Please lie down and wait
1034 until the feeling passes. Be aware that the semantics of C<fork()> vary
1035 between platforms. For example, some Unix systems copy all the current
1036 threads into the child process, while others only copy the thread that
1037 called C<fork()>. You have been warned!
1039 Similarly, mixing signals and threads may be problematic.
1040 Implementations are platform-dependent, and even the POSIX
1041 semantics may not be what you expect (and Perl doesn't even
1042 give you the full POSIX API). For example, there is no way to
1043 guarantee that a signal sent to a multi-threaded Perl application
1044 will get intercepted by any particular thread. (However, a recently
1045 added feature does provide the capability to send signals between
1046 threads. See L<threads/"THREAD SIGNALLING> for more details.)
1048 =head1 Thread-Safety of System Libraries
1050 Whether various library calls are thread-safe is outside the control
1051 of Perl. Calls often suffering from not being thread-safe include:
1052 C<localtime()>, C<gmtime()>, functions fetching user, group and
1053 network information (such as C<getgrent()>, C<gethostent()>,
1054 C<getnetent()> and so on), C<readdir()>, C<rand()>, and C<srand()>. In
1055 general, calls that depend on some global external state.
1057 If the system Perl is compiled in has thread-safe variants of such
1058 calls, they will be used. Beyond that, Perl is at the mercy of
1059 the thread-safety or -unsafety of the calls. Please consult your
1060 C library call documentation.
1062 On some platforms the thread-safe library interfaces may fail if the
1063 result buffer is too small (for example the user group databases may
1064 be rather large, and the reentrant interfaces may have to carry around
1065 a full snapshot of those databases). Perl will start with a small
1066 buffer, but keep retrying and growing the result buffer
1067 until the result fits. If this limitless growing sounds bad for
1068 security or memory consumption reasons you can recompile Perl with
1069 C<PERL_REENTRANT_MAXSIZE> defined to the maximum number of bytes you will
1074 A complete thread tutorial could fill a book (and has, many times),
1075 but with what we've covered in this introduction, you should be well
1076 on your way to becoming a threaded Perl expert.
1080 Annotated POD for L<threads>:
1081 L<http://annocpan.org/?mode=search&field=Module&name=threads>
1083 Lastest version of L<threads> on CPAN:
1084 L<http://search.cpan.org/search?module=threads>
1086 Annotated POD for L<threads::shared>:
1087 L<http://annocpan.org/?mode=search&field=Module&name=threads%3A%3Ashared>
1089 Lastest version of L<threads::shared> on CPAN:
1090 L<http://search.cpan.org/search?module=threads%3A%3Ashared>
1092 Perl threads mailing list:
1093 L<http://lists.cpan.org/showlist.cgi?name=iThreads>
1097 Here's a short bibliography courtesy of Jürgen Christoffel:
1099 =head2 Introductory Texts
1101 Birrell, Andrew D. An Introduction to Programming with
1102 Threads. Digital Equipment Corporation, 1989, DEC-SRC Research Report
1104 ftp://ftp.dec.com/pub/DEC/SRC/research-reports/SRC-035.pdf
1105 (highly recommended)
1107 Robbins, Kay. A., and Steven Robbins. Practical Unix Programming: A
1108 Guide to Concurrency, Communication, and
1109 Multithreading. Prentice-Hall, 1996.
1111 Lewis, Bill, and Daniel J. Berg. Multithreaded Programming with
1112 Pthreads. Prentice Hall, 1997, ISBN 0-13-443698-9 (a well-written
1113 introduction to threads).
1115 Nelson, Greg (editor). Systems Programming with Modula-3. Prentice
1116 Hall, 1991, ISBN 0-13-590464-1.
1118 Nichols, Bradford, Dick Buttlar, and Jacqueline Proulx Farrell.
1119 Pthreads Programming. O'Reilly & Associates, 1996, ISBN 156592-115-1
1120 (covers POSIX threads).
1122 =head2 OS-Related References
1124 Boykin, Joseph, David Kirschen, Alan Langerman, and Susan
1125 LoVerso. Programming under Mach. Addison-Wesley, 1994, ISBN
1128 Tanenbaum, Andrew S. Distributed Operating Systems. Prentice Hall,
1129 1995, ISBN 0-13-219908-4 (great textbook).
1131 Silberschatz, Abraham, and Peter B. Galvin. Operating System Concepts,
1132 4th ed. Addison-Wesley, 1995, ISBN 0-201-59292-4
1134 =head2 Other References
1136 Arnold, Ken and James Gosling. The Java Programming Language, 2nd
1137 ed. Addison-Wesley, 1998, ISBN 0-201-31006-6.
1139 comp.programming.threads FAQ,
1140 L<http://www.serpentine.com/~bos/threads-faq/>
1142 Le Sergent, T. and B. Berthomieu. "Incremental MultiThreaded Garbage
1143 Collection on Virtually Shared Memory Architectures" in Memory
1144 Management: Proc. of the International Workshop IWMM 92, St. Malo,
1145 France, September 1992, Yves Bekkers and Jacques Cohen, eds. Springer,
1146 1992, ISBN 3540-55940-X (real-life thread applications).
1148 Artur Bergman, "Where Wizards Fear To Tread", June 11, 2002,
1149 L<http://www.perl.com/pub/a/2002/06/11/threads.html>
1151 =head1 Acknowledgements
1153 Thanks (in no particular order) to Chaim Frenkel, Steve Fink, Gurusamy
1154 Sarathy, Ilya Zakharevich, Benjamin Sugars, Jürgen Christoffel, Joshua
1155 Pritikin, and Alan Burlison, for their help in reality-checking and
1156 polishing this article. Big thanks to Tom Christiansen for his rewrite
1157 of the prime number generator.
1161 Dan Sugalski E<lt>dan@sidhe.org<gt>
1163 Slightly modified by Arthur Bergman to fit the new thread model/module.
1165 Reworked slightly by Jörg Walter E<lt>jwalt@cpan.org<gt> to be more concise
1166 about thread-safety of Perl code.
1168 Rearranged slightly by Elizabeth Mattijsen E<lt>liz@dijkmat.nl<gt> to put
1169 less emphasis on yield().
1173 The original version of this article originally appeared in The Perl
1174 Journal #10, and is copyright 1998 The Perl Journal. It appears courtesy
1175 of Jon Orwant and The Perl Journal. This document may be distributed
1176 under the same terms as Perl itself.