[ViewVC] Diff of: cvs/AnyEvent/lib/AnyEvent.pm

Comparing AnyEvent/lib/AnyEvent.pm (file contents):
Revision 1.90 by root, Fri Apr 25 14:24:29 2008 UTC vs.
Revision 1.107 by root, Tue May 6 12:15:50 2008 UTC

…		…
15	my $w = AnyEvent->timer (after => $seconds, cb => sub {	15	my $w = AnyEvent->timer (after => $seconds, cb => sub {
16	...	16	...
17	});	17	});
18		18
19	my $w = AnyEvent->condvar; # stores whether a condition was flagged	19	my $w = AnyEvent->condvar; # stores whether a condition was flagged
20	$w->wait; # enters "main loop" till $condvar gets ->broadcast	20	$w->wait; # enters "main loop" till $condvar gets ->send
21	$w->broadcast; # wake up current and all future wait's	21	$w->send; # wake up current and all future wait's
22		22
23	=head1 WHY YOU SHOULD USE THIS MODULE (OR NOT)	23	=head1 WHY YOU SHOULD USE THIS MODULE (OR NOT)
24		24
25	Glib, POE, IO::Async, Event... CPAN offers event models by the dozen	25	Glib, POE, IO::Async, Event... CPAN offers event models by the dozen
26	nowadays. So what is different about AnyEvent?	26	nowadays. So what is different about AnyEvent?
…		…
65	technically possible.	65	technically possible.
66		66
67	Of course, if you want lots of policy (this can arguably be somewhat	67	Of course, if you want lots of policy (this can arguably be somewhat
68	useful) and you want to force your users to use the one and only event	68	useful) and you want to force your users to use the one and only event
69	model, you should I<not> use this module.	69	model, you should I<not> use this module.
70
71		70
72	=head1 DESCRIPTION	71	=head1 DESCRIPTION
73		72
74	L<AnyEvent> provides an identical interface to multiple event loops. This	73	L<AnyEvent> provides an identical interface to multiple event loops. This
75	allows module authors to utilise an event loop without forcing module	74	allows module authors to utilise an event loop without forcing module
…		…
289	my $w = AnyEvent->child (	288	my $w = AnyEvent->child (
290	pid => $pid,	289	pid => $pid,
291	cb => sub {	290	cb => sub {
292	my ($pid, $status) = @_;	291	my ($pid, $status) = @_;
293	warn "pid $pid exited with status $status";	292	warn "pid $pid exited with status $status";
294	$done->broadcast;	293	$done->send;
295	},	294	},
296	);	295	);
297		296
298	# do something else, then wait for process exit	297	# do something else, then wait for process exit
299	$done->wait;	298	$done->wait;
300		299
301	=head2 CONDITION VARIABLES	300	=head2 CONDITION VARIABLES
302		301
		302	If you are familiar with some event loops you will know that all of them
		303	require you to run some blocking "loop", "run" or similar function that
		304	will actively watch for new events and call your callbacks.
		305
		306	AnyEvent is different, it expects somebody else to run the event loop and
		307	will only block when necessary (usually when told by the user).
		308
		309	The instrument to do that is called a "condition variable", so called
		310	because they represent a condition that must become true.
		311
303	Condition variables can be created by calling the C<< AnyEvent->condvar >>	312	Condition variables can be created by calling the C<< AnyEvent->condvar
304	method without any arguments.	313	>> method, usually without arguments. The only argument pair allowed is
		314	C<cb>, which specifies a callback to be called when the condition variable
		315	becomes true.
305		316
306	A condition variable waits for a condition - precisely that the C<<	317	After creation, the conditon variable is "false" until it becomes "true"
307	->broadcast >> method has been called.	318	by calling the C<send> method.
308		319
309	They are very useful to signal that a condition has been fulfilled, for	320	Condition variables are similar to callbacks, except that you can
		321	optionally wait for them. They can also be called merge points - points
		322	in time where multiple outstandign events have been processed. And yet
		323	another way to call them is transations - each condition variable can be
		324	used to represent a transaction, which finishes at some point and delivers
		325	a result.
		326
		327	Condition variables are very useful to signal that something has finished,
310	example, if you write a module that does asynchronous http requests,	328	for example, if you write a module that does asynchronous http requests,
311	then a condition variable would be the ideal candidate to signal the	329	then a condition variable would be the ideal candidate to signal the
312	availability of results.	330	availability of results. The user can either act when the callback is
		331	called or can synchronously C<< ->wait >> for the results.
313		332
314	You can also use condition variables to block your main program until	333	You can also use them to simulate traditional event loops - for example,
315	an event occurs - for example, you could C<< ->wait >> in your main	334	you can block your main program until an event occurs - for example, you
316	program until the user clicks the Quit button in your app, which would C<<	335	could C<< ->wait >> in your main program until the user clicks the Quit
317	->broadcast >> the "quit" event.	336	button of your app, which would C<< ->send >> the "quit" event.
318		337
319	Note that condition variables recurse into the event loop - if you have	338	Note that condition variables recurse into the event loop - if you have
320	two pirces of code that call C<< ->wait >> in a round-robbin fashion, you	339	two pieces of code that call C<< ->wait >> in a round-robbin fashion, you
321	lose. Therefore, condition variables are good to export to your caller, but	340	lose. Therefore, condition variables are good to export to your caller, but
322	you should avoid making a blocking wait yourself, at least in callbacks,	341	you should avoid making a blocking wait yourself, at least in callbacks,
323	as this asks for trouble.	342	as this asks for trouble.
324		343
325	This object has two methods:	344	Condition variables are represented by hash refs in perl, and the keys
		345	used by AnyEvent itself are all named C<_ae_XXX> to make subclassing
		346	easy (it is often useful to build your own transaction class on top of
		347	AnyEvent). To subclass, use C<AnyEvent::CondVar> as base class and call
		348	it's C<new> method in your own C<new> method.
		349
		350	There are two "sides" to a condition variable - the "producer side" which
		351	eventually calls C<< -> send >>, and the "consumer side", which waits
		352	for the send to occur.
		353
		354	Example:
		355
		356	# wait till the result is ready
		357	my $result_ready = AnyEvent->condvar;
		358
		359	# do something such as adding a timer
		360	# or socket watcher the calls $result_ready->send
		361	# when the "result" is ready.
		362	# in this case, we simply use a timer:
		363	my $w = AnyEvent->timer (
		364	after => 1,
		365	cb => sub { $result_ready->send },
		366	);
		367
		368	# this "blocks" (while handling events) till the callback
		369	# calls send
		370	$result_ready->wait;
		371
		372	=head3 METHODS FOR PRODUCERS
		373
		374	These methods should only be used by the producing side, i.e. the
		375	code/module that eventually sends the signal. Note that it is also
		376	the producer side which creates the condvar in most cases, but it isn't
		377	uncommon for the consumer to create it as well.
326		378
327	=over 4	379	=over 4
328		380
		381	=item $cv->send (...)
		382
		383	Flag the condition as ready - a running C<< ->wait >> and all further
		384	calls to C<wait> will (eventually) return after this method has been
		385	called. If nobody is waiting the send will be remembered.
		386
		387	If a callback has been set on the condition variable, it is called
		388	immediately from within send.
		389
		390	Any arguments passed to the C<send> call will be returned by all
		391	future C<< ->wait >> calls.
		392
		393	=item $cv->croak ($error)
		394
		395	Similar to send, but causes all call's wait C<< ->wait >> to invoke
		396	C<Carp::croak> with the given error message/object/scalar.
		397
		398	This can be used to signal any errors to the condition variable
		399	user/consumer.
		400
		401	=item $cv->begin ([group callback])
		402
		403	=item $cv->end
		404
		405	These two methods can be used to combine many transactions/events into
		406	one. For example, a function that pings many hosts in parallel might want
		407	to use a condition variable for the whole process.
		408
		409	Every call to C<< ->begin >> will increment a counter, and every call to
		410	C<< ->end >> will decrement it. If the counter reaches C<0> in C<< ->end
		411	>>, the (last) callback passed to C<begin> will be executed. That callback
		412	is I<supposed> to call C<< ->send >>, but that is not required. If no
		413	callback was set, C<send> will be called without any arguments.
		414
		415	Let's clarify this with the ping example:
		416
		417	my $cv = AnyEvent->condvar;
		418
		419	my %result;
		420	$cv->begin (sub { $cv->send (\%result) });
		421
		422	for my $host (@list_of_hosts) {
		423	$cv->begin;
		424	ping_host_then_call_callback $host, sub {
		425	$result{$host} = ...;
		426	$cv->end;
		427	};
		428	}
		429
		430	$cv->end;
		431
		432	This code fragment supposedly pings a number of hosts and calls
		433	C<send> after results for all then have have been gathered - in any
		434	order. To achieve this, the code issues a call to C<begin> when it starts
		435	each ping request and calls C<end> when it has received some result for
		436	it. Since C<begin> and C<end> only maintain a counter, the order in which
		437	results arrive is not relevant.
		438
		439	There is an additional bracketing call to C<begin> and C<end> outside the
		440	loop, which serves two important purposes: first, it sets the callback
		441	to be called once the counter reaches C<0>, and second, it ensures that
		442	C<send> is called even when C<no> hosts are being pinged (the loop
		443	doesn't execute once).
		444
		445	This is the general pattern when you "fan out" into multiple subrequests:
		446	use an outer C<begin>/C<end> pair to set the callback and ensure C<end>
		447	is called at least once, and then, for each subrequest you start, call
		448	C<begin> and for eahc subrequest you finish, call C<end>.
		449
		450	=back
		451
		452	=head3 METHODS FOR CONSUMERS
		453
		454	These methods should only be used by the consuming side, i.e. the
		455	code awaits the condition.
		456
		457	=over 4
		458
329	=item $cv->wait	459	=item $cv->wait
330		460
331	Wait (blocking if necessary) until the C<< ->broadcast >> method has been	461	Wait (blocking if necessary) until the C<< ->send >> or C<< ->croak
332	called on c<$cv>, while servicing other watchers normally.	462	>> methods have been called on c<$cv>, while servicing other watchers
		463	normally.
333		464
334	You can only wait once on a condition - additional calls will return	465	You can only wait once on a condition - additional calls are valid but
335	immediately.	466	will return immediately.
		467
		468	If an error condition has been set by calling C<< ->croak >>, then this
		469	function will call C<croak>.
		470
		471	In list context, all parameters passed to C<send> will be returned,
		472	in scalar context only the first one will be returned.
336		473
337	Not all event models support a blocking wait - some die in that case	474	Not all event models support a blocking wait - some die in that case
338	(programs might want to do that to stay interactive), so I<if you are	475	(programs might want to do that to stay interactive), so I<if you are
339	using this from a module, never require a blocking wait>, but let the	476	using this from a module, never require a blocking wait>, but let the
340	caller decide whether the call will block or not (for example, by coupling	477	caller decide whether the call will block or not (for example, by coupling
…		…
347	multiple interpreters or coroutines/threads, none of which C<AnyEvent>	484	multiple interpreters or coroutines/threads, none of which C<AnyEvent>
348	can supply (the coroutine-aware backends L<AnyEvent::Impl::CoroEV> and	485	can supply (the coroutine-aware backends L<AnyEvent::Impl::CoroEV> and
349	L<AnyEvent::Impl::CoroEvent> explicitly support concurrent C<< ->wait >>'s	486	L<AnyEvent::Impl::CoroEvent> explicitly support concurrent C<< ->wait >>'s
350	from different coroutines, however).	487	from different coroutines, however).
351		488
352	=item $cv->broadcast	489	You can ensure that C<< -wait >> never blocks by setting a callback and
		490	only calling C<< ->wait >> from within that callback (or at a later
		491	time). This will work even when the event loop does not support blocking
		492	waits otherwise.
353		493
354	Flag the condition as ready - a running C<< ->wait >> and all further	494	=item $bool = $cv->ready
355	calls to C<wait> will (eventually) return after this method has been	495
356	called. If nobody is waiting the broadcast will be remembered..	496	Returns true when the condition is "true", i.e. whether C<send> or
		497	C<croak> have been called.
		498
		499	=item $cb = $cv->cb ([new callback])
		500
		501	This is a mutator function that returns the callback set and optionally
		502	replaces it before doing so.
		503
		504	The callback will be called when the condition becomes "true", i.e. when
		505	C<send> or C<croak> are called. Calling C<wait> inside the callback
		506	or at any later time is guaranteed not to block.
357		507
358	=back	508	=back
359
360	Example:
361
362	# wait till the result is ready
363	my $result_ready = AnyEvent->condvar;
364
365	# do something such as adding a timer
366	# or socket watcher the calls $result_ready->broadcast
367	# when the "result" is ready.
368	# in this case, we simply use a timer:
369	my $w = AnyEvent->timer (
370	after => 1,
371	cb => sub { $result_ready->broadcast },
372	);
373
374	# this "blocks" (while handling events) till the watcher
375	# calls broadcast
376	$result_ready->wait;
377		509
378	=head1 GLOBAL VARIABLES AND FUNCTIONS	510	=head1 GLOBAL VARIABLES AND FUNCTIONS
379		511
380	=over 4	512	=over 4
381		513
…		…
391		523
392	AnyEvent::Impl::CoroEV based on Coro::EV, best choice.	524	AnyEvent::Impl::CoroEV based on Coro::EV, best choice.
393	AnyEvent::Impl::CoroEvent based on Coro::Event, second best choice.	525	AnyEvent::Impl::CoroEvent based on Coro::Event, second best choice.
394	AnyEvent::Impl::EV based on EV (an interface to libev, best choice).	526	AnyEvent::Impl::EV based on EV (an interface to libev, best choice).
395	AnyEvent::Impl::Event based on Event, second best choice.	527	AnyEvent::Impl::Event based on Event, second best choice.
		528	AnyEvent::Impl::Perl pure-perl implementation, fast and portable.
396	AnyEvent::Impl::Glib based on Glib, third-best choice.	529	AnyEvent::Impl::Glib based on Glib, third-best choice.
397	AnyEvent::Impl::Perl pure-perl implementation, inefficient but portable.
398	AnyEvent::Impl::Tk based on Tk, very bad choice.	530	AnyEvent::Impl::Tk based on Tk, very bad choice.
399	AnyEvent::Impl::Qt based on Qt, cannot be autoprobed (see its docs).	531	AnyEvent::Impl::Qt based on Qt, cannot be autoprobed (see its docs).
400	AnyEvent::Impl::EventLib based on Event::Lib, leaks memory and worse.	532	AnyEvent::Impl::EventLib based on Event::Lib, leaks memory and worse.
401	AnyEvent::Impl::POE based on POE, not generic enough for full support.	533	AnyEvent::Impl::POE based on POE, not generic enough for full support.
402		534
…		…
428	decide which event module to use as soon as the first method is called, so	560	decide which event module to use as soon as the first method is called, so
429	by calling AnyEvent in your module body you force the user of your module	561	by calling AnyEvent in your module body you force the user of your module
430	to load the event module first.	562	to load the event module first.
431		563
432	Never call C<< ->wait >> on a condition variable unless you I<know> that	564	Never call C<< ->wait >> on a condition variable unless you I<know> that
433	the C<< ->broadcast >> method has been called on it already. This is	565	the C<< ->send >> method has been called on it already. This is
434	because it will stall the whole program, and the whole point of using	566	because it will stall the whole program, and the whole point of using
435	events is to stay interactive.	567	events is to stay interactive.
436		568
437	It is fine, however, to call C<< ->wait >> when the user of your module	569	It is fine, however, to call C<< ->wait >> when the user of your module
438	requests it (i.e. if you create a http request object ad have a method	570	requests it (i.e. if you create a http request object ad have a method
…		…
458		590
459	You can chose to use a rather inefficient pure-perl implementation by	591	You can chose to use a rather inefficient pure-perl implementation by
460	loading the C<AnyEvent::Impl::Perl> module, which gives you similar	592	loading the C<AnyEvent::Impl::Perl> module, which gives you similar
461	behaviour everywhere, but letting AnyEvent chose is generally better.	593	behaviour everywhere, but letting AnyEvent chose is generally better.
462		594
		595	=head1 OTHER MODULES
		596
		597	The following is a non-exhaustive list of additional modules that use
		598	AnyEvent and can therefore be mixed easily with other AnyEvent modules
		599	in the same program. Some of the modules come with AnyEvent, some are
		600	available via CPAN.
		601
		602	=over 4
		603
		604	=item L<AnyEvent::Util>
		605
		606	Contains various utility functions that replace often-used but blocking
		607	functions such as C<inet_aton> by event-/callback-based versions.
		608
		609	=item L<AnyEvent::Handle>
		610
		611	Provide read and write buffers and manages watchers for reads and writes.
		612
		613	=item L<AnyEvent::Socket>
		614
		615	Provides a means to do non-blocking connects, accepts etc.
		616
		617	=item L<AnyEvent::HTTPD>
		618
		619	Provides a simple web application server framework.
		620
		621	=item L<AnyEvent::DNS>
		622
		623	Provides asynchronous DNS resolver capabilities, beyond what
		624	L<AnyEvent::Util> offers.
		625
		626	=item L<AnyEvent::FastPing>
		627
		628	The fastest ping in the west.
		629
		630	=item L<Net::IRC3>
		631
		632	AnyEvent based IRC client module family.
		633
		634	=item L<Net::XMPP2>
		635
		636	AnyEvent based XMPP (Jabber protocol) module family.
		637
		638	=item L<Net::FCP>
		639
		640	AnyEvent-based implementation of the Freenet Client Protocol, birthplace
		641	of AnyEvent.
		642
		643	=item L<Event::ExecFlow>
		644
		645	High level API for event-based execution flow control.
		646
		647	=item L<Coro>
		648
		649	Has special support for AnyEvent.
		650
		651	=item L<IO::Lambda>
		652
		653	The lambda approach to I/O - don't ask, look there. Can use AnyEvent.
		654
		655	=item L<IO::AIO>
		656
		657	Truly asynchronous I/O, should be in the toolbox of every event
		658	programmer. Can be trivially made to use AnyEvent.
		659
		660	=item L<BDB>
		661
		662	Truly asynchronous Berkeley DB access. Can be trivially made to use
		663	AnyEvent.
		664
		665	=back
		666
463	=cut	667	=cut
464		668
465	package AnyEvent;	669	package AnyEvent;
466		670
467	no warnings;	671	no warnings;
…		…
482	my @models = (	686	my @models = (
483	[Coro::EV:: => AnyEvent::Impl::CoroEV::],	687	[Coro::EV:: => AnyEvent::Impl::CoroEV::],
484	[Coro::Event:: => AnyEvent::Impl::CoroEvent::],	688	[Coro::Event:: => AnyEvent::Impl::CoroEvent::],
485	[EV:: => AnyEvent::Impl::EV::],	689	[EV:: => AnyEvent::Impl::EV::],
486	[Event:: => AnyEvent::Impl::Event::],	690	[Event:: => AnyEvent::Impl::Event::],
487	[Glib:: => AnyEvent::Impl::Glib::],
488	[Tk:: => AnyEvent::Impl::Tk::],	691	[Tk:: => AnyEvent::Impl::Tk::],
489	[Wx:: => AnyEvent::Impl::POE::],	692	[Wx:: => AnyEvent::Impl::POE::],
490	[Prima:: => AnyEvent::Impl::POE::],	693	[Prima:: => AnyEvent::Impl::POE::],
491	[AnyEvent::Impl::Perl:: => AnyEvent::Impl::Perl::],	694	[AnyEvent::Impl::Perl:: => AnyEvent::Impl::Perl::],
492	# everything below here will not be autoprobed as the pureperl backend should work everywhere	695	# everything below here will not be autoprobed as the pureperl backend should work everywhere
		696	[Glib:: => AnyEvent::Impl::Glib::],
493	[Event::Lib:: => AnyEvent::Impl::EventLib::], # too buggy	697	[Event::Lib:: => AnyEvent::Impl::EventLib::], # too buggy
494	[Qt:: => AnyEvent::Impl::Qt::], # requires special main program	698	[Qt:: => AnyEvent::Impl::Qt::], # requires special main program
495	[POE::Kernel:: => AnyEvent::Impl::POE::], # lasciate ogni speranza	699	[POE::Kernel:: => AnyEvent::Impl::POE::], # lasciate ogni speranza
496	);	700	);
497		701
498	our %method = map +($_ => 1), qw(io timer signal child condvar broadcast wait one_event DESTROY);	702	our %method = map +($_ => 1), qw(io timer signal child condvar one_event DESTROY);
499		703
500	sub detect() {	704	sub detect() {
501	unless ($MODEL) {	705	unless ($MODEL) {
502	no strict 'refs';	706	no strict 'refs';
503		707
…		…
894	});	1098	});
895		1099
896	$quit->wait;	1100	$quit->wait;
897		1101
898		1102
899	=head1 BENCHMARK	1103	=head1 BENCHMARKS
900		1104
901	To give you an idea of the performance and overheads that AnyEvent adds	1105	To give you an idea of the performance and overheads that AnyEvent adds
902	over the event loops themselves (and to give you an impression of the	1106	over the event loops themselves and to give you an impression of the speed
903	speed of various event loops), here is a benchmark of various supported	1107	of various event loops I prepared some benchmarks.
904	event models natively and with anyevent. The benchmark creates a lot of	1108
905	timers (with a zero timeout) and I/O watchers (watching STDOUT, a pty, to	1109	=head2 BENCHMARKING ANYEVENT OVERHEAD
		1110
		1111	Here is a benchmark of various supported event models used natively and
		1112	through anyevent. The benchmark creates a lot of timers (with a zero
		1113	timeout) and I/O watchers (watching STDOUT, a pty, to become writable,
906	become writable, which it is), lets them fire exactly once and destroys	1114	which it is), lets them fire exactly once and destroys them again.
907	them again.
908		1115
909	Rewriting the benchmark to use many different sockets instead of using	1116	Source code for this benchmark is found as F<eg/bench> in the AnyEvent
910	the same filehandle for all I/O watchers results in a much longer runtime	1117	distribution.
911	(socket creation is expensive), but qualitatively the same figures, so it
912	was not used.
913		1118
914	=head2 Explanation of the columns	1119	=head3 Explanation of the columns
915		1120
916	I<watcher> is the number of event watchers created/destroyed. Since	1121	I<watcher> is the number of event watchers created/destroyed. Since
917	different event models feature vastly different performances, each event	1122	different event models feature vastly different performances, each event
918	loop was given a number of watchers so that overall runtime is acceptable	1123	loop was given a number of watchers so that overall runtime is acceptable
919	and similar between tested event loop (and keep them from crashing): Glib	1124	and similar between tested event loop (and keep them from crashing): Glib
…		…
935	signal the end of this phase.	1140	signal the end of this phase.
936		1141
937	I<destroy> is the time, in microseconds, that it takes to destroy a single	1142	I<destroy> is the time, in microseconds, that it takes to destroy a single
938	watcher.	1143	watcher.
939		1144
940	=head2 Results	1145	=head3 Results
941		1146
942	name watchers bytes create invoke destroy comment	1147	name watchers bytes create invoke destroy comment
943	EV/EV 400000 244 0.56 0.46 0.31 EV native interface	1148	EV/EV 400000 244 0.56 0.46 0.31 EV native interface
944	EV/Any 100000 244 2.50 0.46 0.29 EV + AnyEvent watchers	1149	EV/Any 100000 244 2.50 0.46 0.29 EV + AnyEvent watchers
945	CoroEV/Any 100000 244 2.49 0.44 0.29 coroutines + Coro::Signal	1150	CoroEV/Any 100000 244 2.49 0.44 0.29 coroutines + Coro::Signal
946	Perl/Any 100000 513 4.92 0.87 1.12 pure perl implementation	1151	Perl/Any 100000 513 4.92 0.87 1.12 pure perl implementation
947	Event/Event 16000 516 31.88 31.30 0.85 Event native interface	1152	Event/Event 16000 516 31.88 31.30 0.85 Event native interface
948	Event/Any 16000 936 39.17 33.63 1.43 Event + AnyEvent watchers	1153	Event/Any 16000 590 35.75 31.42 1.08 Event + AnyEvent watchers
949	Glib/Any 16000 1357 98.22 12.41 54.00 quadratic behaviour	1154	Glib/Any 16000 1357 98.22 12.41 54.00 quadratic behaviour
950	Tk/Any 2000 1860 26.97 67.98 14.00 SEGV with >> 2000 watchers	1155	Tk/Any 2000 1860 26.97 67.98 14.00 SEGV with >> 2000 watchers
951	POE/Event 2000 6644 108.64 736.02 14.73 via POE::Loop::Event	1156	POE/Event 2000 6644 108.64 736.02 14.73 via POE::Loop::Event
952	POE/Select 2000 6343 94.13 809.12 565.96 via POE::Loop::Select	1157	POE/Select 2000 6343 94.13 809.12 565.96 via POE::Loop::Select
953		1158
954	=head2 Discussion	1159	=head3 Discussion
955		1160
956	The benchmark does I<not> measure scalability of the event loop very	1161	The benchmark does I<not> measure scalability of the event loop very
957	well. For example, a select-based event loop (such as the pure perl one)	1162	well. For example, a select-based event loop (such as the pure perl one)
958	can never compete with an event loop that uses epoll when the number of	1163	can never compete with an event loop that uses epoll when the number of
959	file descriptors grows high. In this benchmark, all events become ready at	1164	file descriptors grows high. In this benchmark, all events become ready at
960	the same time, so select/poll-based implementations get an unnatural speed	1165	the same time, so select/poll-based implementations get an unnatural speed
961	boost.	1166	boost.
		1167
		1168	Also, note that the number of watchers usually has a nonlinear effect on
		1169	overall speed, that is, creating twice as many watchers doesn't take twice
		1170	the time - usually it takes longer. This puts event loops tested with a
		1171	higher number of watchers at a disadvantage.
		1172
		1173	To put the range of results into perspective, consider that on the
		1174	benchmark machine, handling an event takes roughly 1600 CPU cycles with
		1175	EV, 3100 CPU cycles with AnyEvent's pure perl loop and almost 3000000 CPU
		1176	cycles with POE.
962		1177
963	C<EV> is the sole leader regarding speed and memory use, which are both	1178	C<EV> is the sole leader regarding speed and memory use, which are both
964	maximal/minimal, respectively. Even when going through AnyEvent, it uses	1179	maximal/minimal, respectively. Even when going through AnyEvent, it uses
965	far less memory than any other event loop and is still faster than Event	1180	far less memory than any other event loop and is still faster than Event
966	natively.	1181	natively.
…		…
989	file descriptor is dup()ed for each watcher. This shows that the dup()	1204	file descriptor is dup()ed for each watcher. This shows that the dup()
990	employed by some adaptors is not a big performance issue (it does incur a	1205	employed by some adaptors is not a big performance issue (it does incur a
991	hidden memory cost inside the kernel which is not reflected in the figures	1206	hidden memory cost inside the kernel which is not reflected in the figures
992	above).	1207	above).
993		1208
994	C<POE>, regardless of underlying event loop (whether using its pure	1209	C<POE>, regardless of underlying event loop (whether using its pure perl
995	perl select-based backend or the Event module, the POE-EV backend	1210	select-based backend or the Event module, the POE-EV backend couldn't
996	couldn't be tested because it wasn't working) shows abysmal performance	1211	be tested because it wasn't working) shows abysmal performance and
997	and memory usage: Watchers use almost 30 times as much memory as	1212	memory usage with AnyEvent: Watchers use almost 30 times as much memory
998	EV watchers, and 10 times as much memory as Event (the high memory	1213	as EV watchers, and 10 times as much memory as Event (the high memory
999	requirements are caused by requiring a session for each watcher). Watcher	1214	requirements are caused by requiring a session for each watcher). Watcher
1000	invocation speed is almost 900 times slower than with AnyEvent's pure perl	1215	invocation speed is almost 900 times slower than with AnyEvent's pure perl
		1216	implementation.
		1217
1001	implementation. The design of the POE adaptor class in AnyEvent can not	1218	The design of the POE adaptor class in AnyEvent can not really account
1002	really account for this, as session creation overhead is small compared	1219	for the performance issues, though, as session creation overhead is
1003	to execution of the state machine, which is coded pretty optimally within	1220	small compared to execution of the state machine, which is coded pretty
1004	L<AnyEvent::Impl::POE>. POE simply seems to be abysmally slow.	1221	optimally within L<AnyEvent::Impl::POE> (and while everybody agrees that
		1222	using multiple sessions is not a good approach, especially regarding
		1223	memory usage, even the author of POE could not come up with a faster
		1224	design).
1005		1225
1006	=head2 Summary	1226	=head3 Summary
1007		1227
1008	=over 4	1228	=over 4
1009		1229
1010	=item * Using EV through AnyEvent is faster than any other event loop	1230	=item * Using EV through AnyEvent is faster than any other event loop
1011	(even when used without AnyEvent), but most event loops have acceptable	1231	(even when used without AnyEvent), but most event loops have acceptable
…		…
1018	=item * You should avoid POE like the plague if you want performance or	1238	=item * You should avoid POE like the plague if you want performance or
1019	reasonable memory usage.	1239	reasonable memory usage.
1020		1240
1021	=back	1241	=back
1022		1242
		1243	=head2 BENCHMARKING THE LARGE SERVER CASE
		1244
		1245	This benchmark atcually benchmarks the event loop itself. It works by
		1246	creating a number of "servers": each server consists of a socketpair, a
		1247	timeout watcher that gets reset on activity (but never fires), and an I/O
		1248	watcher waiting for input on one side of the socket. Each time the socket
		1249	watcher reads a byte it will write that byte to a random other "server".
		1250
		1251	The effect is that there will be a lot of I/O watchers, only part of which
		1252	are active at any one point (so there is a constant number of active
		1253	fds for each loop iterstaion, but which fds these are is random). The
		1254	timeout is reset each time something is read because that reflects how
		1255	most timeouts work (and puts extra pressure on the event loops).
		1256
		1257	In this benchmark, we use 10000 socketpairs (20000 sockets), of which 100
		1258	(1%) are active. This mirrors the activity of large servers with many
		1259	connections, most of which are idle at any one point in time.
		1260
		1261	Source code for this benchmark is found as F<eg/bench2> in the AnyEvent
		1262	distribution.
		1263
		1264	=head3 Explanation of the columns
		1265
		1266	I<sockets> is the number of sockets, and twice the number of "servers" (as
		1267	each server has a read and write socket end).
		1268
		1269	I<create> is the time it takes to create a socketpair (which is
		1270	nontrivial) and two watchers: an I/O watcher and a timeout watcher.
		1271
		1272	I<request>, the most important value, is the time it takes to handle a
		1273	single "request", that is, reading the token from the pipe and forwarding
		1274	it to another server. This includes deleting the old timeout and creating
		1275	a new one that moves the timeout into the future.
		1276
		1277	=head3 Results
		1278
		1279	name sockets create request
		1280	EV 20000 69.01 11.16
		1281	Perl 20000 73.32 35.87
		1282	Event 20000 212.62 257.32
		1283	Glib 20000 651.16 1896.30
		1284	POE 20000 349.67 12317.24 uses POE::Loop::Event
		1285
		1286	=head3 Discussion
		1287
		1288	This benchmark I<does> measure scalability and overall performance of the
		1289	particular event loop.
		1290
		1291	EV is again fastest. Since it is using epoll on my system, the setup time
		1292	is relatively high, though.
		1293
		1294	Perl surprisingly comes second. It is much faster than the C-based event
		1295	loops Event and Glib.
		1296
		1297	Event suffers from high setup time as well (look at its code and you will
		1298	understand why). Callback invocation also has a high overhead compared to
		1299	the C<< $_->() for .. >>-style loop that the Perl event loop uses. Event
		1300	uses select or poll in basically all documented configurations.
		1301
		1302	Glib is hit hard by its quadratic behaviour w.r.t. many watchers. It
		1303	clearly fails to perform with many filehandles or in busy servers.
		1304
		1305	POE is still completely out of the picture, taking over 1000 times as long
		1306	as EV, and over 100 times as long as the Perl implementation, even though
		1307	it uses a C-based event loop in this case.
		1308
		1309	=head3 Summary
		1310
		1311	=over 4
		1312
		1313	=item * The pure perl implementation performs extremely well.
		1314
		1315	=item * Avoid Glib or POE in large projects where performance matters.
		1316
		1317	=back
		1318
		1319	=head2 BENCHMARKING SMALL SERVERS
		1320
		1321	While event loops should scale (and select-based ones do not...) even to
		1322	large servers, most programs we (or I :) actually write have only a few
		1323	I/O watchers.
		1324
		1325	In this benchmark, I use the same benchmark program as in the large server
		1326	case, but it uses only eight "servers", of which three are active at any
		1327	one time. This should reflect performance for a small server relatively
		1328	well.
		1329
		1330	The columns are identical to the previous table.
		1331
		1332	=head3 Results
		1333
		1334	name sockets create request
		1335	EV 16 20.00 6.54
		1336	Perl 16 25.75 12.62
		1337	Event 16 81.27 35.86
		1338	Glib 16 32.63 15.48
		1339	POE 16 261.87 276.28 uses POE::Loop::Event
		1340
		1341	=head3 Discussion
		1342
		1343	The benchmark tries to test the performance of a typical small
		1344	server. While knowing how various event loops perform is interesting, keep
		1345	in mind that their overhead in this case is usually not as important, due
		1346	to the small absolute number of watchers (that is, you need efficiency and
		1347	speed most when you have lots of watchers, not when you only have a few of
		1348	them).
		1349
		1350	EV is again fastest.
		1351
		1352	Perl again comes second. It is noticably faster than the C-based event
		1353	loops Event and Glib, although the difference is too small to really
		1354	matter.
		1355
		1356	POE also performs much better in this case, but is is still far behind the
		1357	others.
		1358
		1359	=head3 Summary
		1360
		1361	=over 4
		1362
		1363	=item * C-based event loops perform very well with small number of
		1364	watchers, as the management overhead dominates.
		1365
		1366	=back
		1367
1023		1368
1024	=head1 FORK	1369	=head1 FORK
1025		1370
1026	Most event libraries are not fork-safe. The ones who are usually are	1371	Most event libraries are not fork-safe. The ones who are usually are
1027	because they are so inefficient. Only L<EV> is fully fork-aware.	1372	because they rely on inefficient but fork-safe C<select> or C<poll>
		1373	calls. Only L<EV> is fully fork-aware.
1028		1374
1029	If you have to fork, you must either do so I<before> creating your first	1375	If you have to fork, you must either do so I<before> creating your first
1030	watcher OR you must not use AnyEvent at all in the child.	1376	watcher OR you must not use AnyEvent at all in the child.
1031		1377
1032		1378
…		…
1044		1390
1045	BEGIN { delete $ENV{PERL_ANYEVENT_MODEL} }	1391	BEGIN { delete $ENV{PERL_ANYEVENT_MODEL} }
1046		1392
1047	use AnyEvent;	1393	use AnyEvent;
1048		1394
		1395	Similar considerations apply to $ENV{PERL_ANYEVENT_VERBOSE}, as that can
		1396	be used to probe what backend is used and gain other information (which is
		1397	probably even less useful to an attacker than PERL_ANYEVENT_MODEL).
		1398
1049		1399
1050	=head1 SEE ALSO	1400	=head1 SEE ALSO
1051		1401
1052	Event modules: L<Coro::EV>, L<EV>, L<EV::Glib>, L<Glib::EV>,	1402	Event modules: L<Coro::EV>, L<EV>, L<EV::Glib>, L<Glib::EV>,
1053	L<Coro::Event>, L<Event>, L<Glib::Event>, L<Glib>, L<Coro>, L<Tk>,	1403	L<Coro::Event>, L<Event>, L<Glib::Event>, L<Glib>, L<Coro>, L<Tk>,

Diff Legend

-–
+Removed lines
-+
+Added lines
-<
+Changed lines
->
+Changed lines

Comparing AnyEvent/lib/AnyEvent.pm (file contents): Revision 1.90 by root, Fri Apr 25 14:24:29 2008 UTC vs. Revision 1.107 by root, Tue May 6 12:15:50 2008 UTC

Diff Legend

Comparing AnyEvent/lib/AnyEvent.pm (file contents):
Revision 1.90 by root, Fri Apr 25 14:24:29 2008 UTC vs.
Revision 1.107 by root, Tue May 6 12:15:50 2008 UTC