This is a live mirror of the Perl 5 development currently hosted at
Add perlthrtut, update pod/* machinery.
[perl5.git] / pod / perlthrtut.pod
1=head1 NAME
3perlthrtut - tutorial on threads in Perl
7One of the most prominent new features of Perl 5.005 is the inclusion
8of threads. Threads make a number of things a lot easier, and are a
9very useful addition to your bag of programming tricks.
11=head1 What Is A Thread Anyway?
13A thread is a flow of control through a program with a single
14execution point.
16Sounds an awful lot like a process, doesn't it? Well, it should.
17Threads are one of the pieces of a process. Every process has at least
18one thread and, up until now, every process running Perl had only one
19thread. With 5.005, though, you can create extra threads. We're going
20to show you how, when, and why.
22=head1 Threaded Program Models
24There are three basic ways that you can structure a threaded
25program. Which model you choose depends on what you need your program
26to do. For many non-trivial threaded programs you'll need to choose
27different models for different pieces of your program.
29=head2 Boss/Worker
31The boss/worker model usually has one `boss' thread and one or more
32`worker' threads. The boss thread gathers or generates tasks that need
33to be done, then parcels those tasks out to the appropriate worker
36This model is common in GUI and server programs, where a main thread
37waits for some event and then passes that event to the appropriate
38worker threads for processing. Once the event has been passed on, the
39boss thread goes back to waiting for another event.
41The boss thread does relatively little work. While tasks aren't
42necessarily performed faster than with any other method, it tends to
43have the best user-response times.
45=head2 Work Crew
47In the work crew model, several threads are created that do
48essentially the same thing to different pieces of data. It closely
49mirrors classical parallel processing and vector processors, where a
50large array of processors do the exact same thing to many pieces of
53This model is particularly useful if the system running the program
54will distribute multiple threads across different processors. It can
55also be useful in ray tracing or rendering engines, where the
56individual threads can pass on interim results to give the user visual
59=head2 Pipeline
61The pipeline model divides up a task into a series of steps, and
62passes the results of one step on to the thread processing the
63next. Each thread does one thing to each piece of data and passes the
64results to the next thread in line.
66This model makes the most sense if you have multiple processors so two
67or more threads will be executing in parallel, though it can often
68make sense in other contexts as well. It tends to keep the individual
69tasks small and simple, as well as allowing some parts of the pipeline
70to block (on I/O or system calls, for example) while other parts keep
71going. If you're running different parts of the pipeline on different
72processors you may also take advantage of the caches on each
75This model is also handy for a form of recursive programming where,
76rather than having a subroutine call itself, it instead creates
77another thread. Prime and Fibonacci generators both map well to this
78form of the pipeline model. (A version of a prime number generator is
79presented later on.)
81=head1 Native threads
83There are several different ways to implement threads on a system. How
84threads are implemented depends both on the vendor and, in some cases,
85the version of the operating system. Often the first implementation
86will be relatively simple, but later versions of the OS will be more
89While the information in this section is useful, it's not necessary,
90so you can skip it if you don't feel up to it.
92There are three basic categories of threads-user-mode threads, kernel
93threads, and multiprocessor kernel threads.
95User-mode threads are threads that live entirely within a program and
96its libraries. In this model, the OS knows nothing about threads. As
97far as it's concerned, your process is just a process.
99This is the easiest way to implement threads, and the way most OSes
100start. The big disadvantage is that, since the OS knows nothing about
101threads, if one thread blocks they all do. Typical blocking activities
102include most system calls, most I/O, and things like sleep().
104Kernel threads are the next step in thread evolution. The OS knows
105about kernel threads, and makes allowances for them. The main
106difference between a kernel thread and a user-mode thread is
107blocking. With kernel threads, things that block a single thread don't
108block other threads. This is not the case with user-mode threads,
109where the kernel blocks at the process level and not the thread level.
111This is a big step forward, and can give a threaded program quite a
112performance boost over non-threaded programs. Threads that block
113performing I/O, for example, won't block threads that are doing other
114things. Each process still has only one thread running at once,
115though, regardless of how many CPUs a system might have.
117Since kernel threading can interrupt a thread at any time, they will
118uncover some of the implicit locking assumptions you may make in your
119program. For example, something as simple as C<$a = $a + 2> can behave
120unpredictably with kernel threads if C<$a> is visible to other
121threads, as another thread may have changed C<$a> between the time it
122was fetched on the right hand side and the time the new value is
125Multiprocessor Kernel Threads are the final step in thread
126support. With multiprocessor kernel threads on a machine with multiple
127CPUs, the OS may schedule two or more threads to run simultaneously on
128different CPUs.
130This can give a serious performance boost to your threaded program,
131since more than one thread will be executing at the same time. As a
132tradeoff, though, any of those nagging synchronization issues that
133might not have shown with basic kernel threads will appear with a
136In addition to the different levels of OS involvement in threads,
137different OSes (and different thread implementations for a particular
138OS) allocate CPU cycles to threads in different ways.
140Cooperative multitasking systems have running threads give up control
141if one of two things happen. If a thread calls a yield function, it
142gives up control. It also gives up control if the thread does
143something that would cause it to block, such as perform I/O. In a
144cooperative multitasking implementation, one thread can starve all the
145others for CPU time if it so chooses.
147Preemptive multitasking systems interrupt threads at regular intervals
148while the system decides which thread should run next. In a preemptive
149multitasking system, one thread usually won't monopolize the CPU.
151On some systems, there can be cooperative and preemptive threads
152running simultaneously. (Threads running with realtime priorities
153often behave cooperatively, for example, while threads running at
154normal priorities behave preemptively.)
156=head1 What kind of threads are perl threads?
158If you have experience with other thread implementations, you might
159find that things aren't quite what you expect. It's very important to
160remember when dealing with Perl threads that Perl Threads Are Not X
161Threads, for all values of X. They aren't POSIX threads, or
162DecThreads, or Java's Green threads, or Win32 threads. There are
163similarities, and the broad concepts are the same, but if you start
164looking for implementation details you're going to be either
165disappointed or confused. Possibly both.
167This is not to say that Perl threads are completely different from
168everything that's ever come before--they're not. Perl's threading
169model owes a lot to other thread models, especially POSIX. Just as
170Perl is not C, though, Perl threads are not POSIX threads. So if you
171find yourself looking for mutexes, or thread priorities, it's time to
172step back a bit and think about what you want to do and how Perl can
173do it.
175=head1 Threadsafe Modules
177The addition of threads has changed Perl's internals
178substantially. There are implications for people who write
179modules--especially modules with XS code or external libraries. While
180most modules won't encounter any problems, modules that aren't
181explicitly tagged as thread-safe should be tested before being used in
182production code.
184Not all modules that you might use are thread-safe, and you should
185always assume a module is unsafe unless the documentation says
186otherwise. This includes modules that are distributed as part of the
187core. Threads are a beta feature, and even some of the standard
188modules aren't thread-safe.
190If you're using a module that's not thread-safe for some reason, you
191can protect yourself by using semaphores and lots of programming
192discipline to control access to the module. Semaphores are covered
193later in the article. Perl Threads Are Different
195=head1 Thread Basics
197The core Thread module provides the basic functions you need to write
198threaded programs. In the following sections we'll cover the basics,
199showing you what you need to do to create a threaded program. After
200that, we'll go over some of the features of the Thread module that
201make threaded programming easier.
203=head2 Basic Thread Support
205Thread support is a Perl compile-time option-it's something that's
206turned on or off when Perl is built at your site, rather than when
207your programs are compiled. If your Perl wasn't compiled with thread
208support enabled, then any attempt to use threads will fail.
210Remember that the threading support in 5.005 is in beta release, and
211should be treated as such. You should expect that it may not function
212entirely properly, and the thread interface may well change some
213before it is a fully supported, production release. The beta version
214shouldn't be used for mission-critical projects. Having said that,
215threaded Perl is pretty nifty, and worth a look.
217Your programs can use the Config module to check whether threads are
218enabled. If your program can't run without them, you can say something
221 $Config{usethreads} or die "Recompile Perl with threads to run this program.";
223A possibly-threaded program using a possibly-threaded module might
224have code like this:
226 use Config;
227 use MyMod;
229 if ($Config{usethreads}) {
230 # We have threads
231 require MyMod_threaded;
232 import MyMod_threaded;
233 } else {
234 require MyMod_unthreaded;
235 import MyMod_unthreaded;
236 }
238Since code that runs both with and without threads is usually pretty
239messy, it's best to isolate the thread-specific code in its own
240module. In our example above, that's what MyMod_threaded is, and it's
241only imported if we're running on a threaded Perl.
243=head2 Creating Threads
245The Thread package provides the tools you need to create new
246threads. Like any other module, you need to tell Perl you want to use
247it; use Thread imports all the pieces you need to create basic
250The simplest, straightforward way to create a thread is with new():
252 use Thread;
254 $thr = new Thread \&sub1;
256 sub sub1 {
257 print "In the thread\n";
258 }
260The new() method takes a reference to a subroutine and creates a new
261thread, which starts executing in the referenced subroutine. Control
262then passes both to the subroutine and the caller.
264If you need to, your program can pass parameters to the subroutine as
265part of the thread startup. Just include the list of parameters as
266part of the C<Thread::new> call, like this:
268 use Thread;
269 $Param3 = "foo";
270 $thr = new Thread \&sub1, "Param 1", "Param 2", $Param3;
271 $thr = new Thread \&sub1, @ParamList;
272 $thr = new Thread \&sub1, qw(Param1 Param2 $Param3);
274 sub sub1 {
275 my @InboundParameters = @_;
276 print "In the thread\n";
277 print "got parameters >", join("<>", @InboundParameters), "<\n";
278 }
281The subroutine runs like a normal Perl subroutine, and the call to new
282Thread returns whatever the subroutine returns.
284The last example illustrates another feature of threads. You can spawn
285off several threads using the same subroutine. Each thread executes
286the same subroutine, but in a separate thread with a separate
287environment and potentially separate arguments.
289The other way to spawn a new thread is with async(), which is a way to
290spin off a chunk of code like eval(), but into its own thread:
292 use Thread qw(async);
294 $LineCount = 0;
296 $thr = async {
297 while(<>) {$LineCount++}
298 print "Got $LineCount lines\n";
299 };
301 print "Waiting for the linecount to end\n";
302 $thr->join;
303 print "All done\n";
305You'll notice we did a use Thread qw(async) in that example. async is
306not exported by default, so if you want it, you'll either need to
307import it before you use it or fully qualify it as
308Thread::async. You'll also note that there's a semicolon after the
309closing brace. That's because async() treats the following block as an
310anonymous subroutine, so the semicolon is necessary.
312Like eval(), the code executes in the same context as it would if it
313weren't spun off. Since both the code inside and after the async start
314executing, you need to be careful with any shared resources. Locking
315and other synchronization techniques are covered later.
317=head2 Giving up control
319There are times when you may find it useful to have a thread
320explicitly give up the CPU to another thread. Your threading package
321might not support preemptive multitasking for threads, for example, or
322you may be doing something compute-intensive and want to make sure
323that the user-interface thread gets called frequently. Regardless,
324there are times that you might want a thread to give up the processor.
326Perl's threading package provides the yield() function that does
327this. yield() is pretty straightforward, and works like this:
329 use Thread qw(yield async);
330 async {
331 my $foo = 50;
332 while ($foo--) { print "first async\n" }
333 yield;
334 $foo = 50;
335 while ($foo--) { print "first async\n" }
336 };
337 async {
338 my $foo = 50;
339 while ($foo--) { print "second async\n" }
340 yield;
341 $foo = 50;
342 while ($foo--) { print "second async\n" }
343 };
345=head2 Waiting For A Thread To Exit
347Since threads are also subroutines, they can return values. To wait
348for a thread to exit and extract any scalars it might return, you can
349use the join() method.
351 use Thread;
352 $thr = new Thread \&sub1;
354 @ReturnData = $thr->join;
355 print "Thread returned @ReturnData";
357 sub sub1 { return "Fifty-six", "foo", 2; }
359In the example above, the join() method returns as soon as the thread
360ends. In addition to waiting for a thread to finish and gathering up
361any values that the thread might have returned, join() also performs
362any OS cleanup necessary for the thread. That cleanup might be
363important, especially for long-running programs that spawn lots of
364threads. If you don't want the return values and don't want to wait
365for the thread to finish, you should call the detach() method
366instead. detach() is covered later in the article.
368=head2 Errors In Threads
370So what happens when an error occurs in a thread? Any errors that
371could be caught with eval() are postponed until the thread is
372joined. If your program never joins, the errors appear when your
373program exits.
375Errors deferred until a join() can be caught with eval():
377 use Thread qw(async);
378 $thr = async {$b = 3/0}; # Divide by zero error
379 $foo = eval {$thr->join};
380 if ($@) {
381 print "died with error $@\n";
382 } else {
383 print "Hey, why aren't you dead?\n";
384 }
386eval() passes any results from the joined thread back unmodified, so
387if you want the return value of the thread, this is your only chance
388to get them.
390=head2 Ignoring A Thread
392join() does three things:it waits for a thread to exit, cleans up
393after it, and returns any data the thread may have produced. But what
394if you're not interested in the thread's return values, and you don't
395really care when the thread finishes? All you want is for the thread
396to get cleaned up after when it's done.
398In this case, you use the detach() method. Once a thread is detached,
399it'll run until it's finished, then Perl will clean up after it
402 use Thread;
403 $thr = new Thread \&sub1; # Spawn the thread
405 $thr->detach; # Now we officially don't care any more
407 sub sub1 {
408 $a = 0;
409 while (1) {
410 $a++;
411 print "\$a is $a\n";
412 sleep 1;
413 }
414 }
417Once a thread is detached, it may not be joined, and any output that
418it might have produced (if it was done and waiting for a join) is
421=head1 Threads And Data
423Now that we've covered the basics of threads, it's time for our next
424topic: data. Threading introduces a couple of complications to data
425access that non-threaded programs never need to worry about.
427=head2 Shared And Unshared Data
429The single most important thing to remember when using threads is that
430all threads potentially have access to all the data anywhere in your
431program. While this is true with a nonthreaded Perl program as well,
432it's especially important to remember with a threaded program, since
433more than one thread can be accessing this data at once.
435Perl's scoping rules don't change because you're using threads. If a
436subroutine (or block, in the case of async()) could see a variable if
437you weren't running with threads, it can see it if you are. This is
438especially important for the subroutines that create, and makes my
439variables even more important. Remember--if your variables aren't
440lexically scoped (declared with C<my>) you're probably sharing it between
443=head2 Thread Pitfall: Races
445While threads bring a new set of useful tools, they also bring a
446number of pitfalls. One pitfall is the race condition:
448 use Thread;
449 $a = 1;
450 $thr1 = Thread->new(\&sub1);
451 $thr2 = Thread->new(\&sub2);
453 sleep 10;
454 print "$a\n";
456 sub sub1 { $foo = $a; $a = $foo + 1; }
457 sub sub2 { $bar = $a; $a = $bar + 1; }
459What do you think $a will be? The answer, unfortunately, is "it
460depends." Both sub1() and sub2() access the global variable $a, once
461to read and once to write. Depending on factors ranging from your
462thread implementation's scheduling algorithm to the phase of the moon,
463$a can be 2 or 3.
465Race conditions are caused by unsynchronized access to shared
466data. Without explicit synchronization, there's no way to be sure that
467nothing has happened to the shared data between the time you access it
468and the time you update it. Even this simple code fragment has the
469possibility of error:
471 use Thread qw(async);
472 $a = 2;
473 async{ $b = $a; $a = $b + 1; };
474 async{ $c = $a; $a = $c + 1; };
476Two threads both access $a. Each thread can potentially be interrupted
477at any point, or be executed in any order. At the end, $a could be 3
478or 4, and both $b and $c could be 2 or 3.
480Whenever your program accesses data or resources that can be accessed
481by other threads, you must take steps to coordinate access or risk
482data corruption and race conditions.
484=head2 Controlling access: lock()
486The lock() function takes a variable (or subroutine, but we'll get to
487that later) and puts a lock on it. No other thread may lock the
488variable until the locking thread exits the innermost block containing
489the lock. Using lock() is straightforward:
491 use Thread qw(async);
492 $a = 4;
493 $thr1 = async {
494 $foo = 12;
495 {
496 lock ($a); # Block until we get access to $a
497 $b = $a;
498 $a = $b * $foo;
499 }
500 print "\$foo was $foo\n";
501 };
502 $thr2 = async {
503 $bar = 7;
504 {
505 lock ($a); # Block until we can get access to $a
506 $c = $a;
507 $a = $c * $bar;
508 }
509 print "\$bar was $bar\n";
510 };
511 $thr1->join;
512 $thr2->join;
513 print "\$a is $a\n";
515lock() blocks the thread until the variable being locked is
516available. When lock() returns, your thread can be sure that no other
517thread can lock that variable until the innermost block containing the
518lock exits.
520It's important to note that locks don't prevent access to the variable
521in question, only lock attempts. This is in keeping with Perl's
522longstanding tradition of courteous programming, and the advisory file
523locking that flock() gives you. Locked subroutines behave differently,
524however. We'll cover that later in the article.
526You may lock arrays and hashes as well as scalars. Locking an array,
527though, will not block subsequent locks on array elements, just lock
528attempts on the array itself.
530Finally, locks are recursive, which means it's okay for a thread to
531lock a variable more than once. The lock will last until the outermost
532lock() on the variable goes out of scope.
534=head2 Thread Pitfall: Deadlocks
536Locks are a handy tool to synchronize access to data. Using them
537properly is the key to safe shared data. Unfortunately, locks aren't
538without their dangers. Consider the following code:
540 use Thread qw(async yield);
541 $a = 4;
542 $b = "foo";
543 async {
544 lock($a);
545 yield;
546 sleep 20;
547 lock ($b);
548 };
549 async {
550 lock($b);
551 yield;
552 sleep 20;
553 lock ($a);
554 };
556This program will probably hang until you kill it. The only way it
557won't hang is if one of the two async() routines acquires both locks
558first. A guaranteed-to-hang version is more complicated, but the
559principle is the same.
561The first thread spawned by async() will grab a lock on $a then, a
562second or two later, try to grab a lock on $b. Meanwhile, the second
563thread grabs a lock on $b, then later tries to grab a lock on $a. The
564second lock attempt for both threads will block, each waiting for the
565other to release its lock.
567This condition is called a deadlock, and it occurs whenever two or
568more threads are trying to get locks on resources that the others
569own. Each thread will block, waiting for the other to release a lock
570on a resource. That never happens, though, since the thread with the
571resource is itself waiting for a lock to be released.
573There are a number of ways to handle this sort of problem. The best
574way is to always have all threads acquire locks in the exact same
575order. If, for example, you lock variables $a, $b, and $c, always lock
576$a before $b, and $b before $c. It's also best to hold on to locks for
577as short a period of time to minimize the risks of deadlock.
579=head2 Queues: Passing Data Around
581A queue is a special thread-safe object that lets you put data in one
582end and take it out the other without having to worry about
583synchronization issues. They're pretty straightforward, and look like
586 use Thread qw(async);
587 use Thread::Queue;
589 my $DataQueue = new Thread::Queue;
590 $thr = async {
591 while ($DataElement = $DataQueue->dequeue) {
592 print "Popped $DataElement off the queue\n";
593 }
594 };
596 $DataQueue->enqueue(12);
597 $DataQueue->enqueue("A", "B", "C");
598 $DataQueue->enqueue(\$thr);
599 sleep 10;
600 $DataQueue->enqueue(undef);
602You create the queue with new Thread::Queue. Then you can add lists of
603scalars onto the end with enqueue(), and pop scalars off the front of
604it with dequeue(). A queue has no fixed size, and can grow as needed
605to hold everything pushed on to it.
607If a queue is empty, dequeue() blocks until another thread enqueues
608something. This makes queues ideal for event loops and other
609communications between threads.
611=head1 Threads And Code
613In addition to providing thread-safe access to data via locks and
614queues, threaded Perl also provides general-purpose semaphores for
615coarser synchronization than locks provide and thread-safe access to
616entire subroutines.
618=head2 Semaphores: Synchronizing Data Access
620Semaphores are a kind of generic locking mechanism. Unlike lock, which
621gets a lock on a particular scalar, Perl doesn't associate any
622particular thing with a semaphore so you can use them to control
623access to anything you like. In addition, semaphores can allow more
624than one thread to access a resource at once, though by default
625semaphores only allow one thread access at a time.
627=over 4
629=item Basic semaphores
631Semaphores have two methods, down and up. down decrements the resource
632count, while up increments it. down calls will block if the
633semaphore's current count would decrement below zero. This program
634gives a quick demonstration:
636 use Thread qw(yield);
637 use Thread::Semaphore;
638 my $semaphore = new Thread::Semaphore;
639 $GlobalVariable = 0;
641 $thr1 = new Thread \&sample_sub, 1;
642 $thr2 = new Thread \&sample_sub, 2;
643 $thr3 = new Thread \&sample_sub, 3;
645 sub sample_sub {
646 my $SubNumber = shift @_;
647 my $TryCount = 10;
648 my $LocalCopy;
649 sleep 1;
650 while ($TryCount--) {
651 $semaphore->down;
652 $LocalCopy = $GlobalVariable;
653 print "$TryCount tries left for sub $SubNumber (\$GlobalVariable is $GlobalVariable)\n";
654 yield;
655 sleep 2;
656 $LocalCopy++;
657 $GlobalVariable = $LocalCopy;
658 $semaphore->up;
659 }
660 }
662The three invocations of the subroutine all operate in sync. The
663semaphore, though, makes sure that only one thread is accessing the
664global variable at once.
666=item Advanced Semaphores
668By default, semaphores behave like locks, letting only one thread
669down() them at a time. However, there are other uses for semaphores.
671Each semaphore has a counter attached to it. down() decrements the
672counter and up() increments the counter. By default, semaphores are
673created with the counter set to one, down() decrements by one, and
674up() increments by one. If down() attempts to decrement the counter
675below zero, it blocks until the counter is large enough. Note that
676while a semaphore can be created with a starting count of zero, any
677up() or down() always changes the counter by at least
678one. $semaphore->down(0) is the same as $semaphore->down(1).
680The question, of course, is why would you do something like this? Why
681create a semaphore with a starting count that's not one, or why
682decrement/increment it by more than one? The answer is resource
683availability. Many resources that you want to manage access for can be
684safely used by more than one thread at once.
686For example, let's take a GUI driven program. It has a semaphore that
687it uses to synchronize access to the display, so only one thread is
688ever drawing at once. Handy, but of course you don't want any thread
689to start drawing until things are properly set up. In this case, you
690can create a semaphore with a counter set to zero, and up it when
691things are ready for drawing.
693Semaphores with counters greater than one are also useful for
694establishing quotas. Say, for example, that you have a number of
695threads that can do I/O at once. You don't want all the threads
696reading or writing at once though, since that can potentially swamp
697your I/O channels, or deplete your process' quota of filehandles. You
698can use a semaphore initialized to the number of concurrent I/O
699requests (or open files) that you want at any one time, and have your
700threads quietly block and unblock themselves.
702Larger increments or decrements are handy in those cases where a
703thread needs to check out or return a number of resources at once.
707=head2 Attributes: Restricting Access To Subroutines
709In addition to synchronizing access to data or resources, you might
710find it useful to synchronize access to subroutines. You may be
711accessing a singular machine resource (perhaps a vector processor), or
712find it easier to serialize calls to a particular subroutine than to
713have a set of locks and sempahores.
715One of the additions to Perl 5.005 is subroutine attributes. The
716Thread package uses these to provide several flavors of
717serialization. It's important to remember that these attributes are
718used in the compilation phase of your program so you can't change a
719subroutine's behavior while your program is actually running.
721=head2 Subroutine Locks
723The basic subroutine lock looks like this:
725 sub test_sub {
726 use attrs qw(locked);
727 }
729This ensures that only one thread will be executing this subroutine at
730any one time. Once a thread calls this subroutine, any other thread
731that calls it will block until the thread in the subroutine exits
732it. A more elaborate example looks like this:
734 use Thread qw(yield);
736 new Thread \&thread_sub, 1;
737 new Thread \&thread_sub, 2;
738 new Thread \&thread_sub, 3;
739 new Thread \&thread_sub, 4;
741 sub sync_sub {
742 use attrs qw(locked);
743 my $CallingThread = shift @_;
744 print "In sync_sub for thread $CallingThread\n";
745 yield;
746 sleep 3;
747 print "Leaving sync_sub for thread $CallingThread\n";
748 }
750 sub thread_sub {
751 my $ThreadID = shift @_;
752 print "Thread $ThreadID calling sync_sub\n";
753 sync_sub($ThreadID);
754 print "$ThreadID is done with sync_sub\n";
755 }
757The use attrs qw(locked) locks sync_sub(), and if you run this, you
758can see that only one thread is in it at any one time.
760=head2 Methods
762Locking an entire subroutine can sometimes be overkill, especially
763when dealing with Perl objects. When calling a method for an object,
764for example, you want to serialize calls to a method, so that only one
765thread will be in the subroutine for a particular object, but threads
766calling that subroutine for a different object aren't blocked. The
767method attribute indicates whether the subroutine is really a method.
769 use Thread;
771 sub tester {
772 my $thrnum = shift @_;
773 my $bar = new Foo;
774 foreach (1..10) {
775 print "$thrnum calling per_object\n";
776 $bar->per_object($thrnum);
777 print "$thrnum out of per_object\n";
778 yield;
779 print "$thrnum calling one_at_a_time\n";
780 $bar->one_at_a_time($thrnum);
781 print "$thrnum out of one_at_a_time\n";
782 yield;
783 }
784 }
786 foreach my $thrnum (1..10) {
787 new Thread \&tester, $thrnum;
788 }
790 package Foo;
791 sub new {
792 my $class = shift @_;
793 return bless [@_], $class;
794 }
796 sub per_object {
797 use attrs qw(locked method);
798 my ($class, $thrnum) = @_;
799 print "In per_object for thread $thrnum\n";
800 yield;
801 sleep 2;
802 print "Exiting per_object for thread $thrnum\n";
803 }
805 sub one_at_a_time {
806 use attrs qw(locked);
807 my ($class, $thrnum) = @_;
808 print "In one_at_a_time for thread $thrnum\n";
809 yield;
810 sleep 2;
811 print "Exiting one_at_a_time for thread $thrnum\n";
812 }
814As you can see from the output (omitted for brevity; it's 800 lines)
815all the threads can be in per_object() simultaneously, but only one
816thread is ever in one_at_a_time() at once.
818=head2 Locking A Subroutine
820You can lock a subroutine as you would lock a variable. Subroutine
821locks work the same as a C<use attrs qw(locked)> in the subroutine,
822and block all access to the subroutine for other threads until the
823lock goes out of scope. When the subroutine isn't locked, any number
824of threads can be in it at once, and getting a lock on a subroutine
825doesn't affect threads already in the subroutine. Getting a lock on a
826subroutine looks like this:
828 lock(\&sub_to_lock);
830Simple enough. Unlike use attrs, which is a compile time option,
831locking and unlocking a subroutine can be done at runtime at your
832discretion. There is some runtime penalty to using lock(\&sub) instead
833of use attrs qw(locked), so make sure you're choosing the proper
834method to do the locking.
836You'd choose lock(\&sub) when writing modules and code to run on both
837threaded and unthreaded Perl, especially for code that will run on
8385.004 or earlier Perls. In that case, it's useful to have subroutines
839that should be serialized lock themselves if they're running threaded,
840like so:
842 package Foo;
843 use Config;
844 $Running_Threaded = 0;
846 BEGIN { $Running_Threaded = $Config{'usethreaded'} }
848 sub sub1 { lock(\&sub1) if $Running_Threaded }
851This way you can ensure single-threadedness regardless of which
852version of Perl you're running.
854=head1 General Thread Utility Routines
856We've covered the workhorse parts of Perl's threading package, and
857with these tools you should be well on your way to writing threaded
858code and packages. There are a few useful little pieces that didn't
859really fit in anyplace else.
861=head2 What Thread Am I In?
863The Thread->self method provides your program with a way to get an
864object representing the thread it's currently in. You can use this
865object in the same way as the ones returned from the thread creation.
867=head2 Thread IDs
869tid() is a thread object method that returns the thread ID of the
870thread the object represents. Thread IDs are integers, with the main
871thread in a program being 0. Currently Perl assigns a unique tid to
872every thread ever created in your program, assigning the first thread
873to be created a tid of 1, and increasing the tid by 1 for each new
874thread that's created.
876=head2 Are These Threads The Same?
878The equal() method takes two thread objects and returns true
879if the objects represent the same thread, and false if they don't.
881=head2 What Threads Are Running?
883Thread->list returns a list of thread objects, one for each thread
884that's currently running. Handy for a number of things, including
885cleaning up at the end of your program:
887 # Loop through all the threads
888 foreach $thr (Thread->list) {
889 # Don't join the main thread or ourselves
890 if ($thr->tid && !Thread::equal($thr, Thread->self)) {
891 $thr->join;
892 }
893 }
895The example above is just for illustration. It isn't strictly
896necessary to join all the threads you create, since Perl detaches all
897the threads before it exits.
899=head1 A Complete Example
901Confused yet? It's time for an example program to show some of the
902things we've covered. This program finds prime numbers using threads.
904 1 #!/usr/bin/perl -w
905 2 # prime-pthread, courtesy of Tom Christiansen
906 3
907 4 use strict;
908 5
909 6 use Thread;
910 7 use Thread::Queue;
911 8
912 9 my $stream = new Thread::Queue;
913 10 my $kid = new Thread(\&check_num, $stream, 2);
914 11
915 12 for my $i ( 3 .. 1000 ) {
916 13 $stream->enqueue($i);
917 14 }
918 15
919 16 $stream->enqueue(undef);
920 17 $kid->join();
921 18
922 19 sub check_num {
923 20 my ($upstream, $cur_prime) = @_;
924 21 my $kid;
925 22 my $downstream = new Thread::Queue;
926 23 while (my $num = $upstream->dequeue) {
927 24 next unless $num % $cur_prime;
928 25 if ($kid) {
929 26 $downstream->enqueue($num);
930 27 } else {
931 28 print "Found prime $num\n";
932 29 $kid = new Thread(\&check_num, $downstream, $num);
933 30 }
934 31 }
935 32 $downstream->enqueue(undef) if $kid;
936 33 $kid->join() if $kid;
937 34 }
939This program uses the pipeline model to generate prime numbers. Each
940thread in the pipeline has an input queue that feeds numbers to be
941checked, a prime number that it's responsible for, and an output queue
942that it funnels numbers that have failed the check into. If the thread
943has a number that's failed its check and there's no child thread, then
944the thread must have found a new prime number. In that case, a new
945child thread is created for that prime and stuck on the end of the
948This probably sounds a bit more confusing than it really is, so lets
949go through this program piece by piece and see what it does. (For
950those of you who might be trying to remember exactly what a prime
951number is, it's a number that's only evenly divisible by itself and 1)
953The bulk of the work is done by the check_num() subroutine, which
954takes a reference to its input queue and a prime number that it's
955responsible for. After pulling in the input queue and the prime that
956the subroutine's checking (line 20), we create a new queue (line 22)
957and reserve a scalar for the thread that we're likely to create later
958(line 21).
960The while loop from lines 23 to line 31 grabs a scalar off the input
961queue and checks against the prime this thread is responsible
962for. Line 24 checks to see if there's a remainder when we modulo the
963number to be checked against our prime. If there is one, the number
964must not be evenly divisible by our prime, so we need to either pass
965it on to the next thread if we've created one (line 26) or create a
966new thread if we haven't.
968The new thread creation is line 29. We pass on to it a reference to
969the queue we've created, and the prime number we've found.
971Finally, once the loop terminates (because we got a 0 or undef in the
972queue, which serves as a note to die), we pass on the notice to our
973child and wait for it to exit if we've created a child (Lines 32 and
976Meanwhile, back in the main thread, we create a queue (line 9) and the
977initial child thread (line 10), and pre-seed it with the first prime:
9782. Then we queue all the numbers from 3 to 1000 for checking (lines
97912-14), then queue a die notice (line 16) and wait for the first child
980thread to terminate (line 17). Because a child won't die until its
981child has died, we know that we're done once we return from the join.
983That's how it works. It's pretty simple; as with many Perl programs,
984the explanation is much longer than the program.
986=head1 Conclusion
988A complete thread tutorial could fill a book (and has, many times),
989but this should get you well on your way. The final authority on how
990Perl's threads behave is the documention bundled with the Perl
991distribution, but with what we've covered in this article, you should
992be well on your way to becoming a threaded Perl expert.
994=head1 Bibliography
996Here's a short bibliography courtesy of Jürgen Christoffel:
998=head2 Introductory Texts
1000Birrell, Andrew D. An Introduction to Programming with
1001Threads. Digital Equipment Corporation, 1989, DEC-SRC Research Report
1002#35 online as
1003 (highly
1006Robbins, Kay. A., and Steven Robbins. Practical Unix Programming: A
1007Guide to Concurrency, Communication, and
1008Multithreading. Prentice-Hall, 1996.
1010Lewis, Bill, and Daniel J. Berg. Multithreaded Programming with
1011Pthreads. Prentice Hall, 1997, ISBN 0-13-443698-9 (a well-written
1012introduction to threads).
1014Nelson, Greg (editor). Systems Programming with Modula-3. Prentice
1015Hall, 1991, ISBN 0-13-590464-1.
1017Nichols, Bradford, Dick Buttlar, and Jacqueline Proulx Farrell.
1018Pthreads Programming. O'Reilly & Associates, 1996, ISBN 156592-115-1
1019(covers POSIX threads).
1021=head2 OS-Related References
1023Boykin, Joseph, David Kirschen, Alan Langerman, and Susan
1024LoVerso. Programming under Mach. Addison-Wesley, 1994, ISBN
1027Tanenbaum, Andrew S. Distributed Operating Systems. Prentice Hall,
10281995, ISBN 0-13-143934-0 (great textbook).
1030Silberschatz, Abraham, and Peter B. Galvin. Operating System Concepts,
10314th ed. Addison-Wesley, 1995, ISBN 0-201-59292-4
1033=head2 Other References
1035Arnold, Ken and James Gosling. The Java Programming Language, 2nd
1036ed. Addison-Wesley, 1998, ISBN 0-201-31006-6.
1038Le Sergent, T. and B. Berthomieu. "Incremental MultiThreaded Garbage
1039Collection on Virtually Shared Memory Architectures" in Memory
1040Management: Proc. of the International Workshop IWMM 92, St. Malo,
1041France, September 1992, Yves Bekkers and Jacques Cohen, eds. Springer,
10421992, ISBN 3540-55940-X (real-life thread applications).
1044=head1 Acknowledgements
1046Thanks (in no particular order) to Chaim Frenkel, Steve Fink, Gurusamy
1047Sarathy, Ilya Zakharevich, Benjamin Sugars, Jürgen Christoffel, Joshua
1048Pritikin, and Alan Burlison, for their help in reality-checking and
1049polishing this article. Big thanks to Tom Christiansen for his rewrite
1050of the prime number generator.
1052=head1 AUTHOR
1054Dan Sugalski E<lt>sugalskd@ous.eduE<gt>
1056=head1 Copyrights
1058This article originally appeared in The Perl Journal #10, and is
1059copyright 1998 The Perl Journal. It appears courtesy of Jon Orwant and
1060The Perl Journal. This document may be distributed under the same terms
1061as Perl itself.