1 | NAME |
1 | NAME |
2 | AnyEvent - the DBI of event loop programming |
2 | AnyEvent - the DBI of event loop programming |
3 | |
3 | |
4 | EV, Event, Glib, Tk, Perl, Event::Lib, Irssi, IO::Async, Qt and POE are |
4 | EV, Event, Glib, Tk, Perl, Event::Lib, Irssi, rxvt-unicode, IO::Async, |
5 | various supported event loops/environments. |
5 | Qt and POE are various supported event loops/environments. |
6 | |
6 | |
7 | SYNOPSIS |
7 | SYNOPSIS |
8 | use AnyEvent; |
8 | use AnyEvent; |
9 | |
9 | |
10 | # file descriptor readable |
10 | # file descriptor readable |
… | |
… | |
173 | Note that "my $w; $w =" combination. This is necessary because in Perl, |
173 | Note that "my $w; $w =" combination. This is necessary because in Perl, |
174 | my variables are only visible after the statement in which they are |
174 | my variables are only visible after the statement in which they are |
175 | declared. |
175 | declared. |
176 | |
176 | |
177 | I/O WATCHERS |
177 | I/O WATCHERS |
|
|
178 | $w = AnyEvent->io ( |
|
|
179 | fh => <filehandle_or_fileno>, |
|
|
180 | poll => <"r" or "w">, |
|
|
181 | cb => <callback>, |
|
|
182 | ); |
|
|
183 | |
178 | You can create an I/O watcher by calling the "AnyEvent->io" method with |
184 | You can create an I/O watcher by calling the "AnyEvent->io" method with |
179 | the following mandatory key-value pairs as arguments: |
185 | the following mandatory key-value pairs as arguments: |
180 | |
186 | |
181 | "fh" is the Perl *file handle* (or a naked file descriptor) to watch for |
187 | "fh" is the Perl *file handle* (or a naked file descriptor) to watch for |
182 | events (AnyEvent might or might not keep a reference to this file |
188 | events (AnyEvent might or might not keep a reference to this file |
… | |
… | |
210 | warn "read: $input\n"; |
216 | warn "read: $input\n"; |
211 | undef $w; |
217 | undef $w; |
212 | }); |
218 | }); |
213 | |
219 | |
214 | TIME WATCHERS |
220 | TIME WATCHERS |
|
|
221 | $w = AnyEvent->timer (after => <seconds>, cb => <callback>); |
|
|
222 | |
|
|
223 | $w = AnyEvent->timer ( |
|
|
224 | after => <fractional_seconds>, |
|
|
225 | interval => <fractional_seconds>, |
|
|
226 | cb => <callback>, |
|
|
227 | ); |
|
|
228 | |
215 | You can create a time watcher by calling the "AnyEvent->timer" method |
229 | You can create a time watcher by calling the "AnyEvent->timer" method |
216 | with the following mandatory arguments: |
230 | with the following mandatory arguments: |
217 | |
231 | |
218 | "after" specifies after how many seconds (fractional values are |
232 | "after" specifies after how many seconds (fractional values are |
219 | supported) the callback should be invoked. "cb" is the callback to |
233 | supported) the callback should be invoked. "cb" is the callback to |
… | |
… | |
340 | time, which might affect timers and time-outs. |
354 | time, which might affect timers and time-outs. |
341 | |
355 | |
342 | When this is the case, you can call this method, which will update |
356 | When this is the case, you can call this method, which will update |
343 | the event loop's idea of "current time". |
357 | the event loop's idea of "current time". |
344 | |
358 | |
|
|
359 | A typical example would be a script in a web server (e.g. |
|
|
360 | "mod_perl") - when mod_perl executes the script, then the event loop |
|
|
361 | will have the wrong idea about the "current time" (being potentially |
|
|
362 | far in the past, when the script ran the last time). In that case |
|
|
363 | you should arrange a call to "AnyEvent->now_update" each time the |
|
|
364 | web server process wakes up again (e.g. at the start of your script, |
|
|
365 | or in a handler). |
|
|
366 | |
345 | Note that updating the time *might* cause some events to be handled. |
367 | Note that updating the time *might* cause some events to be handled. |
346 | |
368 | |
347 | SIGNAL WATCHERS |
369 | SIGNAL WATCHERS |
|
|
370 | $w = AnyEvent->signal (signal => <uppercase_signal_name>, cb => <callback>); |
|
|
371 | |
348 | You can watch for signals using a signal watcher, "signal" is the signal |
372 | You can watch for signals using a signal watcher, "signal" is the signal |
349 | *name* in uppercase and without any "SIG" prefix, "cb" is the Perl |
373 | *name* in uppercase and without any "SIG" prefix, "cb" is the Perl |
350 | callback to be invoked whenever a signal occurs. |
374 | callback to be invoked whenever a signal occurs. |
351 | |
375 | |
352 | Although the callback might get passed parameters, their value and |
376 | Although the callback might get passed parameters, their value and |
… | |
… | |
368 | |
392 | |
369 | Example: exit on SIGINT |
393 | Example: exit on SIGINT |
370 | |
394 | |
371 | my $w = AnyEvent->signal (signal => "INT", cb => sub { exit 1 }); |
395 | my $w = AnyEvent->signal (signal => "INT", cb => sub { exit 1 }); |
372 | |
396 | |
|
|
397 | Restart Behaviour |
|
|
398 | While restart behaviour is up to the event loop implementation, most |
|
|
399 | will not restart syscalls (that includes Async::Interrupt and AnyEvent's |
|
|
400 | pure perl implementation). |
|
|
401 | |
|
|
402 | Safe/Unsafe Signals |
|
|
403 | Perl signals can be either "safe" (synchronous to opcode handling) or |
|
|
404 | "unsafe" (asynchronous) - the former might get delayed indefinitely, the |
|
|
405 | latter might corrupt your memory. |
|
|
406 | |
|
|
407 | AnyEvent signal handlers are, in addition, synchronous to the event |
|
|
408 | loop, i.e. they will not interrupt your running perl program but will |
|
|
409 | only be called as part of the normal event handling (just like timer, |
|
|
410 | I/O etc. callbacks, too). |
|
|
411 | |
373 | Signal Races, Delays and Workarounds |
412 | Signal Races, Delays and Workarounds |
374 | Many event loops (e.g. Glib, Tk, Qt, IO::Async) do not support attaching |
413 | Many event loops (e.g. Glib, Tk, Qt, IO::Async) do not support attaching |
375 | callbacks to signals in a generic way, which is a pity, as you cannot do |
414 | callbacks to signals in a generic way, which is a pity, as you cannot do |
376 | race-free signal handling in perl. AnyEvent will try to do it's best, |
415 | race-free signal handling in perl, requiring C libraries for this. |
377 | but in some cases, signals will be delayed. The maximum time a signal |
416 | AnyEvent will try to do it's best, which means in some cases, signals |
378 | might be delayed is specified in $AnyEvent::MAX_SIGNAL_LATENCY (default: |
417 | will be delayed. The maximum time a signal might be delayed is specified |
379 | 10 seconds). This variable can be changed only before the first signal |
418 | in $AnyEvent::MAX_SIGNAL_LATENCY (default: 10 seconds). This variable |
380 | watcher is created, and should be left alone otherwise. Higher values |
419 | can be changed only before the first signal watcher is created, and |
|
|
420 | should be left alone otherwise. This variable determines how often |
|
|
421 | AnyEvent polls for signals (in case a wake-up was missed). Higher values |
381 | will cause fewer spurious wake-ups, which is better for power and CPU |
422 | will cause fewer spurious wake-ups, which is better for power and CPU |
|
|
423 | saving. |
|
|
424 | |
382 | saving. All these problems can be avoided by installing the optional |
425 | All these problems can be avoided by installing the optional |
383 | Async::Interrupt module. This will not work with inherently broken event |
426 | Async::Interrupt module, which works with most event loops. It will not |
384 | loops such as Event or Event::Lib (and not with POE currently, as POE |
427 | work with inherently broken event loops such as Event or Event::Lib (and |
385 | does it's own workaround with one-second latency). With those, you just |
428 | not with POE currently, as POE does it's own workaround with one-second |
386 | have to suffer the delays. |
429 | latency). For those, you just have to suffer the delays. |
387 | |
430 | |
388 | CHILD PROCESS WATCHERS |
431 | CHILD PROCESS WATCHERS |
|
|
432 | $w = AnyEvent->child (pid => <process id>, cb => <callback>); |
|
|
433 | |
389 | You can also watch on a child process exit and catch its exit status. |
434 | You can also watch on a child process exit and catch its exit status. |
390 | |
435 | |
391 | The child process is specified by the "pid" argument (one some backends, |
436 | The child process is specified by the "pid" argument (one some backends, |
392 | using 0 watches for any child process exit, on others this will croak). |
437 | using 0 watches for any child process exit, on others this will croak). |
393 | The watcher will be triggered only when the child process has finished |
438 | The watcher will be triggered only when the child process has finished |
… | |
… | |
439 | |
484 | |
440 | # do something else, then wait for process exit |
485 | # do something else, then wait for process exit |
441 | $done->recv; |
486 | $done->recv; |
442 | |
487 | |
443 | IDLE WATCHERS |
488 | IDLE WATCHERS |
|
|
489 | $w = AnyEvent->idle (cb => <callback>); |
|
|
490 | |
444 | Sometimes there is a need to do something, but it is not so important to |
491 | Sometimes there is a need to do something, but it is not so important to |
445 | do it instantly, but only when there is nothing better to do. This |
492 | do it instantly, but only when there is nothing better to do. This |
446 | "nothing better to do" is usually defined to be "no other events need |
493 | "nothing better to do" is usually defined to be "no other events need |
447 | attention by the event loop". |
494 | attention by the event loop". |
448 | |
495 | |
… | |
… | |
473 | } |
520 | } |
474 | }); |
521 | }); |
475 | }); |
522 | }); |
476 | |
523 | |
477 | CONDITION VARIABLES |
524 | CONDITION VARIABLES |
|
|
525 | $cv = AnyEvent->condvar; |
|
|
526 | |
|
|
527 | $cv->send (<list>); |
|
|
528 | my @res = $cv->recv; |
|
|
529 | |
478 | If you are familiar with some event loops you will know that all of them |
530 | If you are familiar with some event loops you will know that all of them |
479 | require you to run some blocking "loop", "run" or similar function that |
531 | require you to run some blocking "loop", "run" or similar function that |
480 | will actively watch for new events and call your callbacks. |
532 | will actively watch for new events and call your callbacks. |
481 | |
533 | |
482 | AnyEvent is slightly different: it expects somebody else to run the |
534 | AnyEvent is slightly different: it expects somebody else to run the |
… | |
… | |
547 | after => 1, |
599 | after => 1, |
548 | cb => sub { $result_ready->send }, |
600 | cb => sub { $result_ready->send }, |
549 | ); |
601 | ); |
550 | |
602 | |
551 | # this "blocks" (while handling events) till the callback |
603 | # this "blocks" (while handling events) till the callback |
552 | # calls -<send |
604 | # calls ->send |
553 | $result_ready->recv; |
605 | $result_ready->recv; |
554 | |
606 | |
555 | Example: wait for a timer, but take advantage of the fact that condition |
607 | Example: wait for a timer, but take advantage of the fact that condition |
556 | variables are also callable directly. |
608 | variables are also callable directly. |
557 | |
609 | |
… | |
… | |
614 | into one. For example, a function that pings many hosts in parallel |
666 | into one. For example, a function that pings many hosts in parallel |
615 | might want to use a condition variable for the whole process. |
667 | might want to use a condition variable for the whole process. |
616 | |
668 | |
617 | Every call to "->begin" will increment a counter, and every call to |
669 | Every call to "->begin" will increment a counter, and every call to |
618 | "->end" will decrement it. If the counter reaches 0 in "->end", the |
670 | "->end" will decrement it. If the counter reaches 0 in "->end", the |
619 | (last) callback passed to "begin" will be executed. That callback is |
671 | (last) callback passed to "begin" will be executed, passing the |
620 | *supposed* to call "->send", but that is not required. If no |
672 | condvar as first argument. That callback is *supposed* to call |
|
|
673 | "->send", but that is not required. If no group callback was set, |
621 | callback was set, "send" will be called without any arguments. |
674 | "send" will be called without any arguments. |
622 | |
675 | |
623 | You can think of "$cv->send" giving you an OR condition (one call |
676 | You can think of "$cv->send" giving you an OR condition (one call |
624 | sends), while "$cv->begin" and "$cv->end" giving you an AND |
677 | sends), while "$cv->begin" and "$cv->end" giving you an AND |
625 | condition (all "begin" calls must be "end"'ed before the condvar |
678 | condition (all "begin" calls must be "end"'ed before the condvar |
626 | sends). |
679 | sends). |
… | |
… | |
654 | that are begung can potentially be zero: |
707 | that are begung can potentially be zero: |
655 | |
708 | |
656 | my $cv = AnyEvent->condvar; |
709 | my $cv = AnyEvent->condvar; |
657 | |
710 | |
658 | my %result; |
711 | my %result; |
659 | $cv->begin (sub { $cv->send (\%result) }); |
712 | $cv->begin (sub { shift->send (\%result) }); |
660 | |
713 | |
661 | for my $host (@list_of_hosts) { |
714 | for my $host (@list_of_hosts) { |
662 | $cv->begin; |
715 | $cv->begin; |
663 | ping_host_then_call_callback $host, sub { |
716 | ping_host_then_call_callback $host, sub { |
664 | $result{$host} = ...; |
717 | $result{$host} = ...; |
… | |
… | |
731 | |
784 | |
732 | $cb = $cv->cb ($cb->($cv)) |
785 | $cb = $cv->cb ($cb->($cv)) |
733 | This is a mutator function that returns the callback set and |
786 | This is a mutator function that returns the callback set and |
734 | optionally replaces it before doing so. |
787 | optionally replaces it before doing so. |
735 | |
788 | |
736 | The callback will be called when the condition becomes "true", i.e. |
789 | The callback will be called when the condition becomes (or already |
737 | when "send" or "croak" are called, with the only argument being the |
790 | was) "true", i.e. when "send" or "croak" are called (or were |
738 | condition variable itself. Calling "recv" inside the callback or at |
791 | called), with the only argument being the condition variable itself. |
|
|
792 | Calling "recv" inside the callback or at any later time is |
739 | any later time is guaranteed not to block. |
793 | guaranteed not to block. |
740 | |
794 | |
741 | SUPPORTED EVENT LOOPS/BACKENDS |
795 | SUPPORTED EVENT LOOPS/BACKENDS |
742 | The available backend classes are (every class has its own manpage): |
796 | The available backend classes are (every class has its own manpage): |
743 | |
797 | |
744 | Backends that are autoprobed when no other event loop can be found. |
798 | Backends that are autoprobed when no other event loop can be found. |
745 | EV is the preferred backend when no other event loop seems to be in |
799 | EV is the preferred backend when no other event loop seems to be in |
746 | use. If EV is not installed, then AnyEvent will try Event, and, |
800 | use. If EV is not installed, then AnyEvent will fall back to its own |
747 | failing that, will fall back to its own pure-perl implementation, |
801 | pure-perl implementation, which is available everywhere as it comes |
748 | which is available everywhere as it comes with AnyEvent itself. |
802 | with AnyEvent itself. |
749 | |
803 | |
750 | AnyEvent::Impl::EV based on EV (interface to libev, best choice). |
804 | AnyEvent::Impl::EV based on EV (interface to libev, best choice). |
751 | AnyEvent::Impl::Event based on Event, very stable, few glitches. |
|
|
752 | AnyEvent::Impl::Perl pure-perl implementation, fast and portable. |
805 | AnyEvent::Impl::Perl pure-perl implementation, fast and portable. |
753 | |
806 | |
754 | Backends that are transparently being picked up when they are used. |
807 | Backends that are transparently being picked up when they are used. |
755 | These will be used when they are currently loaded when the first |
808 | These will be used when they are currently loaded when the first |
756 | watcher is created, in which case it is assumed that the application |
809 | watcher is created, in which case it is assumed that the application |
757 | is using them. This means that AnyEvent will automatically pick the |
810 | is using them. This means that AnyEvent will automatically pick the |
758 | right backend when the main program loads an event module before |
811 | right backend when the main program loads an event module before |
759 | anything starts to create watchers. Nothing special needs to be done |
812 | anything starts to create watchers. Nothing special needs to be done |
760 | by the main program. |
813 | by the main program. |
761 | |
814 | |
|
|
815 | AnyEvent::Impl::Event based on Event, very stable, few glitches. |
762 | AnyEvent::Impl::Glib based on Glib, slow but very stable. |
816 | AnyEvent::Impl::Glib based on Glib, slow but very stable. |
763 | AnyEvent::Impl::Tk based on Tk, very broken. |
817 | AnyEvent::Impl::Tk based on Tk, very broken. |
764 | AnyEvent::Impl::EventLib based on Event::Lib, leaks memory and worse. |
818 | AnyEvent::Impl::EventLib based on Event::Lib, leaks memory and worse. |
765 | AnyEvent::Impl::POE based on POE, very slow, some limitations. |
819 | AnyEvent::Impl::POE based on POE, very slow, some limitations. |
766 | AnyEvent::Impl::Irssi used when running within irssi. |
820 | AnyEvent::Impl::Irssi used when running within irssi. |
… | |
… | |
1002 | Event::ExecFlow |
1056 | Event::ExecFlow |
1003 | High level API for event-based execution flow control. |
1057 | High level API for event-based execution flow control. |
1004 | |
1058 | |
1005 | Coro |
1059 | Coro |
1006 | Has special support for AnyEvent via Coro::AnyEvent. |
1060 | Has special support for AnyEvent via Coro::AnyEvent. |
|
|
1061 | |
|
|
1062 | SIMPLIFIED AE API |
|
|
1063 | Starting with version 5.0, AnyEvent officially supports a second, much |
|
|
1064 | simpler, API that is designed to reduce the calling, typing and memory |
|
|
1065 | overhead. |
|
|
1066 | |
|
|
1067 | See the AE manpage for details. |
1007 | |
1068 | |
1008 | ERROR AND EXCEPTION HANDLING |
1069 | ERROR AND EXCEPTION HANDLING |
1009 | In general, AnyEvent does not do any error handling - it relies on the |
1070 | In general, AnyEvent does not do any error handling - it relies on the |
1010 | caller to do that if required. The AnyEvent::Strict module (see also the |
1071 | caller to do that if required. The AnyEvent::Strict module (see also the |
1011 | "PERL_ANYEVENT_STRICT" environment variable, below) provides strict |
1072 | "PERL_ANYEVENT_STRICT" environment variable, below) provides strict |
… | |
… | |
1190 | warn "read: $input\n"; # output what has been read |
1251 | warn "read: $input\n"; # output what has been read |
1191 | $cv->send if $input =~ /^q/i; # quit program if /^q/i |
1252 | $cv->send if $input =~ /^q/i; # quit program if /^q/i |
1192 | }, |
1253 | }, |
1193 | ); |
1254 | ); |
1194 | |
1255 | |
1195 | my $time_watcher; # can only be used once |
|
|
1196 | |
|
|
1197 | sub new_timer { |
|
|
1198 | $timer = AnyEvent->timer (after => 1, cb => sub { |
1256 | my $time_watcher = AnyEvent->timer (after => 1, interval => 1, cb => sub { |
1199 | warn "timeout\n"; # print 'timeout' about every second |
1257 | warn "timeout\n"; # print 'timeout' at most every second |
1200 | &new_timer; # and restart the time |
|
|
1201 | }); |
|
|
1202 | } |
1258 | }); |
1203 | |
|
|
1204 | new_timer; # create first timer |
|
|
1205 | |
1259 | |
1206 | $cv->recv; # wait until user enters /^q/i |
1260 | $cv->recv; # wait until user enters /^q/i |
1207 | |
1261 | |
1208 | REAL-WORLD EXAMPLE |
1262 | REAL-WORLD EXAMPLE |
1209 | Consider the Net::FCP module. It features (among others) the following |
1263 | Consider the Net::FCP module. It features (among others) the following |
… | |
… | |
1336 | through AnyEvent. The benchmark creates a lot of timers (with a zero |
1390 | through AnyEvent. The benchmark creates a lot of timers (with a zero |
1337 | timeout) and I/O watchers (watching STDOUT, a pty, to become writable, |
1391 | timeout) and I/O watchers (watching STDOUT, a pty, to become writable, |
1338 | which it is), lets them fire exactly once and destroys them again. |
1392 | which it is), lets them fire exactly once and destroys them again. |
1339 | |
1393 | |
1340 | Source code for this benchmark is found as eg/bench in the AnyEvent |
1394 | Source code for this benchmark is found as eg/bench in the AnyEvent |
1341 | distribution. |
1395 | distribution. It uses the AE interface, which makes a real difference |
|
|
1396 | for the EV and Perl backends only. |
1342 | |
1397 | |
1343 | Explanation of the columns |
1398 | Explanation of the columns |
1344 | *watcher* is the number of event watchers created/destroyed. Since |
1399 | *watcher* is the number of event watchers created/destroyed. Since |
1345 | different event models feature vastly different performances, each event |
1400 | different event models feature vastly different performances, each event |
1346 | loop was given a number of watchers so that overall runtime is |
1401 | loop was given a number of watchers so that overall runtime is |
… | |
… | |
1365 | *destroy* is the time, in microseconds, that it takes to destroy a |
1420 | *destroy* is the time, in microseconds, that it takes to destroy a |
1366 | single watcher. |
1421 | single watcher. |
1367 | |
1422 | |
1368 | Results |
1423 | Results |
1369 | name watchers bytes create invoke destroy comment |
1424 | name watchers bytes create invoke destroy comment |
1370 | EV/EV 400000 224 0.47 0.35 0.27 EV native interface |
1425 | EV/EV 100000 223 0.47 0.43 0.27 EV native interface |
1371 | EV/Any 100000 224 2.88 0.34 0.27 EV + AnyEvent watchers |
1426 | EV/Any 100000 223 0.48 0.42 0.26 EV + AnyEvent watchers |
1372 | CoroEV/Any 100000 224 2.85 0.35 0.28 coroutines + Coro::Signal |
1427 | Coro::EV/Any 100000 223 0.47 0.42 0.26 coroutines + Coro::Signal |
1373 | Perl/Any 100000 452 4.13 0.73 0.95 pure perl implementation |
1428 | Perl/Any 100000 431 2.70 0.74 0.92 pure perl implementation |
1374 | Event/Event 16000 517 32.20 31.80 0.81 Event native interface |
1429 | Event/Event 16000 516 31.16 31.84 0.82 Event native interface |
1375 | Event/Any 16000 590 35.85 31.55 1.06 Event + AnyEvent watchers |
1430 | Event/Any 16000 1203 42.61 34.79 1.80 Event + AnyEvent watchers |
1376 | IOAsync/Any 16000 989 38.10 32.77 11.13 via IO::Async::Loop::IO_Poll |
1431 | IOAsync/Any 16000 1911 41.92 27.45 16.81 via IO::Async::Loop::IO_Poll |
1377 | IOAsync/Any 16000 990 37.59 29.50 10.61 via IO::Async::Loop::Epoll |
1432 | IOAsync/Any 16000 1726 40.69 26.37 15.25 via IO::Async::Loop::Epoll |
1378 | Glib/Any 16000 1357 102.33 12.31 51.00 quadratic behaviour |
1433 | Glib/Any 16000 1118 89.00 12.57 51.17 quadratic behaviour |
1379 | Tk/Any 2000 1860 27.20 66.31 14.00 SEGV with >> 2000 watchers |
1434 | Tk/Any 2000 1346 20.96 10.75 8.00 SEGV with >> 2000 watchers |
1380 | POE/Event 2000 6328 109.99 751.67 14.02 via POE::Loop::Event |
1435 | POE/Any 2000 6951 108.97 795.32 14.24 via POE::Loop::Event |
1381 | POE/Select 2000 6027 94.54 809.13 579.80 via POE::Loop::Select |
1436 | POE/Any 2000 6648 94.79 774.40 575.51 via POE::Loop::Select |
1382 | |
1437 | |
1383 | Discussion |
1438 | Discussion |
1384 | The benchmark does *not* measure scalability of the event loop very |
1439 | The benchmark does *not* measure scalability of the event loop very |
1385 | well. For example, a select-based event loop (such as the pure perl one) |
1440 | well. For example, a select-based event loop (such as the pure perl one) |
1386 | can never compete with an event loop that uses epoll when the number of |
1441 | can never compete with an event loop that uses epoll when the number of |
… | |
… | |
1397 | benchmark machine, handling an event takes roughly 1600 CPU cycles with |
1452 | benchmark machine, handling an event takes roughly 1600 CPU cycles with |
1398 | EV, 3100 CPU cycles with AnyEvent's pure perl loop and almost 3000000 |
1453 | EV, 3100 CPU cycles with AnyEvent's pure perl loop and almost 3000000 |
1399 | CPU cycles with POE. |
1454 | CPU cycles with POE. |
1400 | |
1455 | |
1401 | "EV" is the sole leader regarding speed and memory use, which are both |
1456 | "EV" is the sole leader regarding speed and memory use, which are both |
1402 | maximal/minimal, respectively. Even when going through AnyEvent, it uses |
1457 | maximal/minimal, respectively. When using the AE API there is zero |
|
|
1458 | overhead (when going through the AnyEvent API create is about 5-6 times |
|
|
1459 | slower, with other times being equal, so still uses far less memory than |
1403 | far less memory than any other event loop and is still faster than Event |
1460 | any other event loop and is still faster than Event natively). |
1404 | natively. |
|
|
1405 | |
1461 | |
1406 | The pure perl implementation is hit in a few sweet spots (both the |
1462 | The pure perl implementation is hit in a few sweet spots (both the |
1407 | constant timeout and the use of a single fd hit optimisations in the |
1463 | constant timeout and the use of a single fd hit optimisations in the |
1408 | perl interpreter and the backend itself). Nevertheless this shows that |
1464 | perl interpreter and the backend itself). Nevertheless this shows that |
1409 | it adds very little overhead in itself. Like any select-based backend |
1465 | it adds very little overhead in itself. Like any select-based backend |
… | |
… | |
1479 | In this benchmark, we use 10000 socket pairs (20000 sockets), of which |
1535 | In this benchmark, we use 10000 socket pairs (20000 sockets), of which |
1480 | 100 (1%) are active. This mirrors the activity of large servers with |
1536 | 100 (1%) are active. This mirrors the activity of large servers with |
1481 | many connections, most of which are idle at any one point in time. |
1537 | many connections, most of which are idle at any one point in time. |
1482 | |
1538 | |
1483 | Source code for this benchmark is found as eg/bench2 in the AnyEvent |
1539 | Source code for this benchmark is found as eg/bench2 in the AnyEvent |
1484 | distribution. |
1540 | distribution. It uses the AE interface, which makes a real difference |
|
|
1541 | for the EV and Perl backends only. |
1485 | |
1542 | |
1486 | Explanation of the columns |
1543 | Explanation of the columns |
1487 | *sockets* is the number of sockets, and twice the number of "servers" |
1544 | *sockets* is the number of sockets, and twice the number of "servers" |
1488 | (as each server has a read and write socket end). |
1545 | (as each server has a read and write socket end). |
1489 | |
1546 | |
… | |
… | |
1495 | forwarding it to another server. This includes deleting the old timeout |
1552 | forwarding it to another server. This includes deleting the old timeout |
1496 | and creating a new one that moves the timeout into the future. |
1553 | and creating a new one that moves the timeout into the future. |
1497 | |
1554 | |
1498 | Results |
1555 | Results |
1499 | name sockets create request |
1556 | name sockets create request |
1500 | EV 20000 69.01 11.16 |
1557 | EV 20000 62.66 7.99 |
1501 | Perl 20000 73.32 35.87 |
1558 | Perl 20000 68.32 32.64 |
1502 | IOAsync 20000 157.00 98.14 epoll |
1559 | IOAsync 20000 174.06 101.15 epoll |
1503 | IOAsync 20000 159.31 616.06 poll |
1560 | IOAsync 20000 174.67 610.84 poll |
1504 | Event 20000 212.62 257.32 |
1561 | Event 20000 202.69 242.91 |
1505 | Glib 20000 651.16 1896.30 |
1562 | Glib 20000 557.01 1689.52 |
1506 | POE 20000 349.67 12317.24 uses POE::Loop::Event |
1563 | POE 20000 341.54 12086.32 uses POE::Loop::Event |
1507 | |
1564 | |
1508 | Discussion |
1565 | Discussion |
1509 | This benchmark *does* measure scalability and overall performance of the |
1566 | This benchmark *does* measure scalability and overall performance of the |
1510 | particular event loop. |
1567 | particular event loop. |
1511 | |
1568 | |
… | |
… | |
1624 | As you can see, the AnyEvent + EV combination even beats the |
1681 | As you can see, the AnyEvent + EV combination even beats the |
1625 | hand-optimised "raw sockets benchmark", while AnyEvent + its pure perl |
1682 | hand-optimised "raw sockets benchmark", while AnyEvent + its pure perl |
1626 | backend easily beats IO::Lambda and POE. |
1683 | backend easily beats IO::Lambda and POE. |
1627 | |
1684 | |
1628 | And even the 100% non-blocking version written using the high-level (and |
1685 | And even the 100% non-blocking version written using the high-level (and |
1629 | slow :) AnyEvent::Handle abstraction beats both POE and IO::Lambda by a |
1686 | slow :) AnyEvent::Handle abstraction beats both POE and IO::Lambda |
1630 | large margin, even though it does all of DNS, tcp-connect and socket I/O |
1687 | higher level ("unoptimised") abstractions by a large margin, even though |
1631 | in a non-blocking way. |
1688 | it does all of DNS, tcp-connect and socket I/O in a non-blocking way. |
1632 | |
1689 | |
1633 | The two AnyEvent benchmarks programs can be found as eg/ae0.pl and |
1690 | The two AnyEvent benchmarks programs can be found as eg/ae0.pl and |
1634 | eg/ae2.pl in the AnyEvent distribution, the remaining benchmarks are |
1691 | eg/ae2.pl in the AnyEvent distribution, the remaining benchmarks are |
1635 | part of the IO::lambda distribution and were used without any changes. |
1692 | part of the IO::Lambda distribution and were used without any changes. |
1636 | |
1693 | |
1637 | SIGNALS |
1694 | SIGNALS |
1638 | AnyEvent currently installs handlers for these signals: |
1695 | AnyEvent currently installs handlers for these signals: |
1639 | |
1696 | |
1640 | SIGCHLD |
1697 | SIGCHLD |
… | |
… | |
1667 | it's built-in modules) are required to use it. |
1724 | it's built-in modules) are required to use it. |
1668 | |
1725 | |
1669 | That does not mean that AnyEvent won't take advantage of some additional |
1726 | That does not mean that AnyEvent won't take advantage of some additional |
1670 | modules if they are installed. |
1727 | modules if they are installed. |
1671 | |
1728 | |
1672 | This section epxlains which additional modules will be used, and how |
1729 | This section explains which additional modules will be used, and how |
1673 | they affect AnyEvent's operetion. |
1730 | they affect AnyEvent's operation. |
1674 | |
1731 | |
1675 | Async::Interrupt |
1732 | Async::Interrupt |
1676 | This slightly arcane module is used to implement fast signal |
1733 | This slightly arcane module is used to implement fast signal |
1677 | handling: To my knowledge, there is no way to do completely |
1734 | handling: To my knowledge, there is no way to do completely |
1678 | race-free and quick signal handling in pure perl. To ensure that |
1735 | race-free and quick signal handling in pure perl. To ensure that |
… | |
… | |
1681 | 10 seconds, look for $AnyEvent::MAX_SIGNAL_LATENCY). |
1738 | 10 seconds, look for $AnyEvent::MAX_SIGNAL_LATENCY). |
1682 | |
1739 | |
1683 | If this module is available, then it will be used to implement |
1740 | If this module is available, then it will be used to implement |
1684 | signal catching, which means that signals will not be delayed, and |
1741 | signal catching, which means that signals will not be delayed, and |
1685 | the event loop will not be interrupted regularly, which is more |
1742 | the event loop will not be interrupted regularly, which is more |
1686 | efficient (And good for battery life on laptops). |
1743 | efficient (and good for battery life on laptops). |
1687 | |
1744 | |
1688 | This affects not just the pure-perl event loop, but also other event |
1745 | This affects not just the pure-perl event loop, but also other event |
1689 | loops that have no signal handling on their own (e.g. Glib, Tk, Qt). |
1746 | loops that have no signal handling on their own (e.g. Glib, Tk, Qt). |
1690 | |
1747 | |
1691 | Some event loops (POE, Event, Event::Lib) offer signal watchers |
1748 | Some event loops (POE, Event, Event::Lib) offer signal watchers |
… | |
… | |
1708 | "AnyEvent::Util::guard". This speeds up guards considerably (and |
1765 | "AnyEvent::Util::guard". This speeds up guards considerably (and |
1709 | uses a lot less memory), but otherwise doesn't affect guard |
1766 | uses a lot less memory), but otherwise doesn't affect guard |
1710 | operation much. It is purely used for performance. |
1767 | operation much. It is purely used for performance. |
1711 | |
1768 | |
1712 | JSON and JSON::XS |
1769 | JSON and JSON::XS |
1713 | This module is required when you want to read or write JSON data via |
1770 | One of these modules is required when you want to read or write JSON |
1714 | AnyEvent::Handle. It is also written in pure-perl, but can take |
1771 | data via AnyEvent::Handle. It is also written in pure-perl, but can |
1715 | advantage of the ultra-high-speed JSON::XS module when it is |
1772 | take advantage of the ultra-high-speed JSON::XS module when it is |
1716 | installed. |
1773 | installed. |
1717 | |
1774 | |
1718 | In fact, AnyEvent::Handle will use JSON::XS by default if it is |
1775 | In fact, AnyEvent::Handle will use JSON::XS by default if it is |
1719 | installed. |
1776 | installed. |
1720 | |
1777 | |
… | |
… | |
1733 | FORK |
1790 | FORK |
1734 | Most event libraries are not fork-safe. The ones who are usually are |
1791 | Most event libraries are not fork-safe. The ones who are usually are |
1735 | because they rely on inefficient but fork-safe "select" or "poll" calls. |
1792 | because they rely on inefficient but fork-safe "select" or "poll" calls. |
1736 | Only EV is fully fork-aware. |
1793 | Only EV is fully fork-aware. |
1737 | |
1794 | |
|
|
1795 | This means that, in general, you cannot fork and do event processing in |
|
|
1796 | the child if a watcher was created before the fork (which in turn |
|
|
1797 | initialises the event library). |
|
|
1798 | |
1738 | If you have to fork, you must either do so *before* creating your first |
1799 | If you have to fork, you must either do so *before* creating your first |
1739 | watcher OR you must not use AnyEvent at all in the child OR you must do |
1800 | watcher OR you must not use AnyEvent at all in the child OR you must do |
1740 | something completely out of the scope of AnyEvent. |
1801 | something completely out of the scope of AnyEvent. |
|
|
1802 | |
|
|
1803 | The problem of doing event processing in the parent *and* the child is |
|
|
1804 | much more complicated: even for backends that *are* fork-aware or |
|
|
1805 | fork-safe, their behaviour is not usually what you want: fork clones all |
|
|
1806 | watchers, that means all timers, I/O watchers etc. are active in both |
|
|
1807 | parent and child, which is almost never what you want. |
1741 | |
1808 | |
1742 | SECURITY CONSIDERATIONS |
1809 | SECURITY CONSIDERATIONS |
1743 | AnyEvent can be forced to load any event model via |
1810 | AnyEvent can be forced to load any event model via |
1744 | $ENV{PERL_ANYEVENT_MODEL}. While this cannot (to my knowledge) be used |
1811 | $ENV{PERL_ANYEVENT_MODEL}. While this cannot (to my knowledge) be used |
1745 | to execute arbitrary code or directly gain access, it can easily be used |
1812 | to execute arbitrary code or directly gain access, it can easily be used |