1 | =head1 NAME |
1 | =head1 NAME |
2 | |
2 | |
3 | AnyEvent - provide framework for multiple event loops |
3 | AnyEvent - provide framework for multiple event loops |
4 | |
4 | |
5 | EV, Event, Coro::EV, Coro::Event, Glib, Tk, Perl, Event::Lib, Qt - various supported event loops |
5 | EV, Event, Coro::EV, Coro::Event, Glib, Tk, Perl, Event::Lib, Qt, POE - various supported event loops |
6 | |
6 | |
7 | =head1 SYNOPSIS |
7 | =head1 SYNOPSIS |
8 | |
8 | |
9 | use AnyEvent; |
9 | use AnyEvent; |
10 | |
10 | |
… | |
… | |
78 | |
78 | |
79 | The interface itself is vaguely similar, but not identical to the L<Event> |
79 | The interface itself is vaguely similar, but not identical to the L<Event> |
80 | module. |
80 | module. |
81 | |
81 | |
82 | During the first call of any watcher-creation method, the module tries |
82 | During the first call of any watcher-creation method, the module tries |
83 | to detect the currently loaded event loop by probing whether one of |
83 | to detect the currently loaded event loop by probing whether one of the |
84 | the following modules is already loaded: L<Coro::EV>, L<Coro::Event>, |
84 | following modules is already loaded: L<Coro::EV>, L<Coro::Event>, L<EV>, |
85 | L<EV>, L<Event>, L<Glib>, L<Tk>, L<Event::Lib>, L<Qt>. The first one |
85 | L<Event>, L<Glib>, L<AnyEvent::Impl::Perl>, L<Tk>, L<Event::Lib>, L<Qt>, |
86 | found is used. If none are found, the module tries to load these modules |
86 | L<POE>. The first one found is used. If none are found, the module tries |
87 | (excluding Event::Lib and Qt) in the order given. The first one that can |
87 | to load these modules (excluding Tk, Event::Lib, Qt and POE as the pure perl |
|
|
88 | adaptor should always succeed) in the order given. The first one that can |
88 | be successfully loaded will be used. If, after this, still none could be |
89 | be successfully loaded will be used. If, after this, still none could be |
89 | found, AnyEvent will fall back to a pure-perl event loop, which is not |
90 | found, AnyEvent will fall back to a pure-perl event loop, which is not |
90 | very efficient, but should work everywhere. |
91 | very efficient, but should work everywhere. |
91 | |
92 | |
92 | Because AnyEvent first checks for modules that are already loaded, loading |
93 | Because AnyEvent first checks for modules that are already loaded, loading |
… | |
… | |
135 | |
136 | |
136 | Note that C<my $w; $w => combination. This is necessary because in Perl, |
137 | Note that C<my $w; $w => combination. This is necessary because in Perl, |
137 | my variables are only visible after the statement in which they are |
138 | my variables are only visible after the statement in which they are |
138 | declared. |
139 | declared. |
139 | |
140 | |
140 | =head2 IO WATCHERS |
141 | =head2 I/O WATCHERS |
141 | |
142 | |
142 | You can create an I/O watcher by calling the C<< AnyEvent->io >> method |
143 | You can create an I/O watcher by calling the C<< AnyEvent->io >> method |
143 | with the following mandatory key-value pairs as arguments: |
144 | with the following mandatory key-value pairs as arguments: |
144 | |
145 | |
145 | C<fh> the Perl I<file handle> (I<not> file descriptor) to watch for |
146 | C<fh> the Perl I<file handle> (I<not> file descriptor) to watch |
146 | events. C<poll> must be a string that is either C<r> or C<w>, which |
147 | for events. C<poll> must be a string that is either C<r> or C<w>, |
147 | creates a watcher waiting for "r"eadable or "w"ritable events, |
148 | which creates a watcher waiting for "r"eadable or "w"ritable events, |
148 | respectively. C<cb> is the callback to invoke each time the file handle |
149 | respectively. C<cb> is the callback to invoke each time the file handle |
149 | becomes ready. |
150 | becomes ready. |
150 | |
151 | |
151 | As long as the I/O watcher exists it will keep the file descriptor or a |
152 | Although the callback might get passed parameters, their value and |
152 | copy of it alive/open. |
153 | presence is undefined and you cannot rely on them. Portable AnyEvent |
|
|
154 | callbacks cannot use arguments passed to I/O watcher callbacks. |
153 | |
155 | |
|
|
156 | The I/O watcher might use the underlying file descriptor or a copy of it. |
154 | It is not allowed to close a file handle as long as any watcher is active |
157 | You must not close a file handle as long as any watcher is active on the |
155 | on the underlying file descriptor. |
158 | underlying file descriptor. |
156 | |
159 | |
157 | Some event loops issue spurious readyness notifications, so you should |
160 | Some event loops issue spurious readyness notifications, so you should |
158 | always use non-blocking calls when reading/writing from/to your file |
161 | always use non-blocking calls when reading/writing from/to your file |
159 | handles. |
162 | handles. |
160 | |
163 | |
… | |
… | |
171 | |
174 | |
172 | You can create a time watcher by calling the C<< AnyEvent->timer >> |
175 | You can create a time watcher by calling the C<< AnyEvent->timer >> |
173 | method with the following mandatory arguments: |
176 | method with the following mandatory arguments: |
174 | |
177 | |
175 | C<after> specifies after how many seconds (fractional values are |
178 | C<after> specifies after how many seconds (fractional values are |
176 | supported) should the timer activate. C<cb> the callback to invoke in that |
179 | supported) the callback should be invoked. C<cb> is the callback to invoke |
177 | case. |
180 | in that case. |
|
|
181 | |
|
|
182 | Although the callback might get passed parameters, their value and |
|
|
183 | presence is undefined and you cannot rely on them. Portable AnyEvent |
|
|
184 | callbacks cannot use arguments passed to time watcher callbacks. |
178 | |
185 | |
179 | The timer callback will be invoked at most once: if you want a repeating |
186 | The timer callback will be invoked at most once: if you want a repeating |
180 | timer you have to create a new watcher (this is a limitation by both Tk |
187 | timer you have to create a new watcher (this is a limitation by both Tk |
181 | and Glib). |
188 | and Glib). |
182 | |
189 | |
… | |
… | |
227 | |
234 | |
228 | You can watch for signals using a signal watcher, C<signal> is the signal |
235 | You can watch for signals using a signal watcher, C<signal> is the signal |
229 | I<name> without any C<SIG> prefix, C<cb> is the Perl callback to |
236 | I<name> without any C<SIG> prefix, C<cb> is the Perl callback to |
230 | be invoked whenever a signal occurs. |
237 | be invoked whenever a signal occurs. |
231 | |
238 | |
|
|
239 | Although the callback might get passed parameters, their value and |
|
|
240 | presence is undefined and you cannot rely on them. Portable AnyEvent |
|
|
241 | callbacks cannot use arguments passed to signal watcher callbacks. |
|
|
242 | |
232 | Multiple signal occurances can be clumped together into one callback |
243 | Multiple signal occurances can be clumped together into one callback |
233 | invocation, and callback invocation will be synchronous. synchronous means |
244 | invocation, and callback invocation will be synchronous. synchronous means |
234 | that it might take a while until the signal gets handled by the process, |
245 | that it might take a while until the signal gets handled by the process, |
235 | but it is guarenteed not to interrupt any other callbacks. |
246 | but it is guarenteed not to interrupt any other callbacks. |
236 | |
247 | |
… | |
… | |
250 | |
261 | |
251 | The child process is specified by the C<pid> argument (if set to C<0>, it |
262 | The child process is specified by the C<pid> argument (if set to C<0>, it |
252 | watches for any child process exit). The watcher will trigger as often |
263 | watches for any child process exit). The watcher will trigger as often |
253 | as status change for the child are received. This works by installing a |
264 | as status change for the child are received. This works by installing a |
254 | signal handler for C<SIGCHLD>. The callback will be called with the pid |
265 | signal handler for C<SIGCHLD>. The callback will be called with the pid |
255 | and exit status (as returned by waitpid). |
266 | and exit status (as returned by waitpid), so unlike other watcher types, |
|
|
267 | you I<can> rely on child watcher callback arguments. |
256 | |
268 | |
257 | Example: wait for pid 1333 |
269 | There is a slight catch to child watchers, however: you usually start them |
|
|
270 | I<after> the child process was created, and this means the process could |
|
|
271 | have exited already (and no SIGCHLD will be sent anymore). |
|
|
272 | |
|
|
273 | Not all event models handle this correctly (POE doesn't), but even for |
|
|
274 | event models that I<do> handle this correctly, they usually need to be |
|
|
275 | loaded before the process exits (i.e. before you fork in the first place). |
|
|
276 | |
|
|
277 | This means you cannot create a child watcher as the very first thing in an |
|
|
278 | AnyEvent program, you I<have> to create at least one watcher before you |
|
|
279 | C<fork> the child (alternatively, you can call C<AnyEvent::detect>). |
|
|
280 | |
|
|
281 | Example: fork a process and wait for it |
|
|
282 | |
|
|
283 | my $done = AnyEvent->condvar; |
|
|
284 | |
|
|
285 | AnyEvent::detect; # force event module to be initialised |
|
|
286 | |
|
|
287 | my $pid = fork or exit 5; |
258 | |
288 | |
259 | my $w = AnyEvent->child ( |
289 | my $w = AnyEvent->child ( |
260 | pid => 1333, |
290 | pid => $pid, |
261 | cb => sub { |
291 | cb => sub { |
262 | my ($pid, $status) = @_; |
292 | my ($pid, $status) = @_; |
263 | warn "pid $pid exited with status $status"; |
293 | warn "pid $pid exited with status $status"; |
|
|
294 | $done->broadcast; |
264 | }, |
295 | }, |
265 | ); |
296 | ); |
|
|
297 | |
|
|
298 | # do something else, then wait for process exit |
|
|
299 | $done->wait; |
266 | |
300 | |
267 | =head2 CONDITION VARIABLES |
301 | =head2 CONDITION VARIABLES |
268 | |
302 | |
269 | Condition variables can be created by calling the C<< AnyEvent->condvar >> |
303 | Condition variables can be created by calling the C<< AnyEvent->condvar >> |
270 | method without any arguments. |
304 | method without any arguments. |
… | |
… | |
358 | AnyEvent::Impl::CoroEV based on Coro::EV, best choice. |
392 | AnyEvent::Impl::CoroEV based on Coro::EV, best choice. |
359 | AnyEvent::Impl::CoroEvent based on Coro::Event, second best choice. |
393 | AnyEvent::Impl::CoroEvent based on Coro::Event, second best choice. |
360 | AnyEvent::Impl::EV based on EV (an interface to libev, best choice). |
394 | AnyEvent::Impl::EV based on EV (an interface to libev, best choice). |
361 | AnyEvent::Impl::Event based on Event, second best choice. |
395 | AnyEvent::Impl::Event based on Event, second best choice. |
362 | AnyEvent::Impl::Glib based on Glib, third-best choice. |
396 | AnyEvent::Impl::Glib based on Glib, third-best choice. |
|
|
397 | AnyEvent::Impl::Perl pure-perl implementation, inefficient but portable. |
363 | AnyEvent::Impl::Tk based on Tk, very bad choice. |
398 | AnyEvent::Impl::Tk based on Tk, very bad choice. |
364 | AnyEvent::Impl::Perl pure-perl implementation, inefficient but portable. |
|
|
365 | AnyEvent::Impl::Qt based on Qt, cannot be autoprobed (see its docs). |
399 | AnyEvent::Impl::Qt based on Qt, cannot be autoprobed (see its docs). |
366 | AnyEvent::Impl::EventLib based on Event::Lib, leaks memory and worse. |
400 | AnyEvent::Impl::EventLib based on Event::Lib, leaks memory and worse. |
|
|
401 | AnyEvent::Impl::POE based on POE, not generic enough for full support. |
|
|
402 | |
|
|
403 | There is no support for WxWidgets, as WxWidgets has no support for |
|
|
404 | watching file handles. However, you can use WxWidgets through the |
|
|
405 | POE Adaptor, as POE has a Wx backend that simply polls 20 times per |
|
|
406 | second, which was considered to be too horrible to even consider for |
|
|
407 | AnyEvent. Likewise, other POE backends can be used by AnyEvent by using |
|
|
408 | it's adaptor. |
|
|
409 | |
|
|
410 | AnyEvent knows about L<Prima> and L<Wx> and will try to use L<POE> when |
|
|
411 | autodetecting them. |
367 | |
412 | |
368 | =item AnyEvent::detect |
413 | =item AnyEvent::detect |
369 | |
414 | |
370 | Returns C<$AnyEvent::MODEL>, forcing autodetection of the event model |
415 | Returns C<$AnyEvent::MODEL>, forcing autodetection of the event model |
371 | if necessary. You should only call this function right before you would |
416 | if necessary. You should only call this function right before you would |
… | |
… | |
422 | no warnings; |
467 | no warnings; |
423 | use strict; |
468 | use strict; |
424 | |
469 | |
425 | use Carp; |
470 | use Carp; |
426 | |
471 | |
427 | our $VERSION = '3.2'; |
472 | our $VERSION = '3.3'; |
428 | our $MODEL; |
473 | our $MODEL; |
429 | |
474 | |
430 | our $AUTOLOAD; |
475 | our $AUTOLOAD; |
431 | our @ISA; |
476 | our @ISA; |
432 | |
477 | |
… | |
… | |
439 | [Coro::Event:: => AnyEvent::Impl::CoroEvent::], |
484 | [Coro::Event:: => AnyEvent::Impl::CoroEvent::], |
440 | [EV:: => AnyEvent::Impl::EV::], |
485 | [EV:: => AnyEvent::Impl::EV::], |
441 | [Event:: => AnyEvent::Impl::Event::], |
486 | [Event:: => AnyEvent::Impl::Event::], |
442 | [Glib:: => AnyEvent::Impl::Glib::], |
487 | [Glib:: => AnyEvent::Impl::Glib::], |
443 | [Tk:: => AnyEvent::Impl::Tk::], |
488 | [Tk:: => AnyEvent::Impl::Tk::], |
|
|
489 | [Wx:: => AnyEvent::Impl::POE::], |
|
|
490 | [Prima:: => AnyEvent::Impl::POE::], |
444 | [AnyEvent::Impl::Perl:: => AnyEvent::Impl::Perl::], |
491 | [AnyEvent::Impl::Perl:: => AnyEvent::Impl::Perl::], |
445 | ); |
492 | # everything below here will not be autoprobed as the pureperl backend should work everywhere |
446 | my @models_detect = ( |
493 | [Event::Lib:: => AnyEvent::Impl::EventLib::], # too buggy |
447 | [Qt:: => AnyEvent::Impl::Qt::], # requires special main program |
494 | [Qt:: => AnyEvent::Impl::Qt::], # requires special main program |
448 | [Event::Lib:: => AnyEvent::Impl::EventLib::], # too buggy |
495 | [POE::Kernel:: => AnyEvent::Impl::POE::], # lasciate ogni speranza |
449 | ); |
496 | ); |
450 | |
497 | |
451 | our %method = map +($_ => 1), qw(io timer signal child condvar broadcast wait one_event DESTROY); |
498 | our %method = map +($_ => 1), qw(io timer signal child condvar broadcast wait one_event DESTROY); |
452 | |
499 | |
453 | sub detect() { |
500 | sub detect() { |
… | |
… | |
457 | if ($ENV{PERL_ANYEVENT_MODEL} =~ /^([a-zA-Z]+)$/) { |
504 | if ($ENV{PERL_ANYEVENT_MODEL} =~ /^([a-zA-Z]+)$/) { |
458 | my $model = "AnyEvent::Impl::$1"; |
505 | my $model = "AnyEvent::Impl::$1"; |
459 | if (eval "require $model") { |
506 | if (eval "require $model") { |
460 | $MODEL = $model; |
507 | $MODEL = $model; |
461 | warn "AnyEvent: loaded model '$model' (forced by \$PERL_ANYEVENT_MODEL), using it.\n" if $verbose > 1; |
508 | warn "AnyEvent: loaded model '$model' (forced by \$PERL_ANYEVENT_MODEL), using it.\n" if $verbose > 1; |
|
|
509 | } else { |
|
|
510 | warn "AnyEvent: unable to load model '$model' (from \$PERL_ANYEVENT_MODEL):\n$@" if $verbose; |
462 | } |
511 | } |
463 | } |
512 | } |
464 | |
513 | |
465 | # check for already loaded models |
514 | # check for already loaded models |
466 | unless ($MODEL) { |
515 | unless ($MODEL) { |
467 | for (@REGISTRY, @models, @models_detect) { |
516 | for (@REGISTRY, @models) { |
468 | my ($package, $model) = @$_; |
517 | my ($package, $model) = @$_; |
469 | if (${"$package\::VERSION"} > 0) { |
518 | if (${"$package\::VERSION"} > 0) { |
470 | if (eval "require $model") { |
519 | if (eval "require $model") { |
471 | $MODEL = $model; |
520 | $MODEL = $model; |
472 | warn "AnyEvent: autodetected model '$model', using it.\n" if $verbose > 1; |
521 | warn "AnyEvent: autodetected model '$model', using it.\n" if $verbose > 1; |
… | |
… | |
659 | |
708 | |
660 | =over 4 |
709 | =over 4 |
661 | |
710 | |
662 | =item C<PERL_ANYEVENT_VERBOSE> |
711 | =item C<PERL_ANYEVENT_VERBOSE> |
663 | |
712 | |
|
|
713 | By default, AnyEvent will be completely silent except in fatal |
|
|
714 | conditions. You can set this environment variable to make AnyEvent more |
|
|
715 | talkative. |
|
|
716 | |
|
|
717 | When set to C<1> or higher, causes AnyEvent to warn about unexpected |
|
|
718 | conditions, such as not being able to load the event model specified by |
|
|
719 | C<PERL_ANYEVENT_MODEL>. |
|
|
720 | |
664 | When set to C<2> or higher, cause AnyEvent to report to STDERR which event |
721 | When set to C<2> or higher, cause AnyEvent to report to STDERR which event |
665 | model it chooses. |
722 | model it chooses. |
666 | |
723 | |
667 | =item C<PERL_ANYEVENT_MODEL> |
724 | =item C<PERL_ANYEVENT_MODEL> |
668 | |
725 | |
… | |
… | |
682 | |
739 | |
683 | =back |
740 | =back |
684 | |
741 | |
685 | =head1 EXAMPLE PROGRAM |
742 | =head1 EXAMPLE PROGRAM |
686 | |
743 | |
687 | The following program uses an IO watcher to read data from STDIN, a timer |
744 | The following program uses an I/O watcher to read data from STDIN, a timer |
688 | to display a message once per second, and a condition variable to quit the |
745 | to display a message once per second, and a condition variable to quit the |
689 | program when the user enters quit: |
746 | program when the user enters quit: |
690 | |
747 | |
691 | use AnyEvent; |
748 | use AnyEvent; |
692 | |
749 | |
… | |
… | |
836 | $quit->broadcast; |
893 | $quit->broadcast; |
837 | }); |
894 | }); |
838 | |
895 | |
839 | $quit->wait; |
896 | $quit->wait; |
840 | |
897 | |
|
|
898 | |
|
|
899 | =head1 BENCHMARKS |
|
|
900 | |
|
|
901 | 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 speed |
|
|
903 | of various event loops I prepared some benchmarks. |
|
|
904 | |
|
|
905 | =head2 BENCHMARKING ANYEVENT OVERHEAD |
|
|
906 | |
|
|
907 | Here is a benchmark of various supported event models used natively and |
|
|
908 | through anyevent. The benchmark creates a lot of timers (with a zero |
|
|
909 | timeout) and I/O watchers (watching STDOUT, a pty, to become writable, |
|
|
910 | which it is), lets them fire exactly once and destroys them again. |
|
|
911 | |
|
|
912 | Source code for this benchmark is found as F<eg/bench> in the AnyEvent |
|
|
913 | distribution. |
|
|
914 | |
|
|
915 | =head3 Explanation of the columns |
|
|
916 | |
|
|
917 | I<watcher> is the number of event watchers created/destroyed. Since |
|
|
918 | different event models feature vastly different performances, each event |
|
|
919 | loop was given a number of watchers so that overall runtime is acceptable |
|
|
920 | and similar between tested event loop (and keep them from crashing): Glib |
|
|
921 | would probably take thousands of years if asked to process the same number |
|
|
922 | of watchers as EV in this benchmark. |
|
|
923 | |
|
|
924 | I<bytes> is the number of bytes (as measured by the resident set size, |
|
|
925 | RSS) consumed by each watcher. This method of measuring captures both C |
|
|
926 | and Perl-based overheads. |
|
|
927 | |
|
|
928 | I<create> is the time, in microseconds (millionths of seconds), that it |
|
|
929 | takes to create a single watcher. The callback is a closure shared between |
|
|
930 | all watchers, to avoid adding memory overhead. That means closure creation |
|
|
931 | and memory usage is not included in the figures. |
|
|
932 | |
|
|
933 | I<invoke> is the time, in microseconds, used to invoke a simple |
|
|
934 | callback. The callback simply counts down a Perl variable and after it was |
|
|
935 | invoked "watcher" times, it would C<< ->broadcast >> a condvar once to |
|
|
936 | signal the end of this phase. |
|
|
937 | |
|
|
938 | I<destroy> is the time, in microseconds, that it takes to destroy a single |
|
|
939 | watcher. |
|
|
940 | |
|
|
941 | =head3 Results |
|
|
942 | |
|
|
943 | name watchers bytes create invoke destroy comment |
|
|
944 | EV/EV 400000 244 0.56 0.46 0.31 EV native interface |
|
|
945 | EV/Any 100000 244 2.50 0.46 0.29 EV + AnyEvent watchers |
|
|
946 | CoroEV/Any 100000 244 2.49 0.44 0.29 coroutines + Coro::Signal |
|
|
947 | Perl/Any 100000 513 4.92 0.87 1.12 pure perl implementation |
|
|
948 | Event/Event 16000 516 31.88 31.30 0.85 Event native interface |
|
|
949 | Event/Any 16000 936 39.17 33.63 1.43 Event + AnyEvent watchers |
|
|
950 | Glib/Any 16000 1357 98.22 12.41 54.00 quadratic behaviour |
|
|
951 | Tk/Any 2000 1860 26.97 67.98 14.00 SEGV with >> 2000 watchers |
|
|
952 | POE/Event 2000 6644 108.64 736.02 14.73 via POE::Loop::Event |
|
|
953 | POE/Select 2000 6343 94.13 809.12 565.96 via POE::Loop::Select |
|
|
954 | |
|
|
955 | =head3 Discussion |
|
|
956 | |
|
|
957 | The benchmark does I<not> measure scalability of the event loop very |
|
|
958 | well. For example, a select-based event loop (such as the pure perl one) |
|
|
959 | can never compete with an event loop that uses epoll when the number of |
|
|
960 | file descriptors grows high. In this benchmark, all events become ready at |
|
|
961 | the same time, so select/poll-based implementations get an unnatural speed |
|
|
962 | boost. |
|
|
963 | |
|
|
964 | C<EV> is the sole leader regarding speed and memory use, which are both |
|
|
965 | maximal/minimal, respectively. Even when going through AnyEvent, it uses |
|
|
966 | far less memory than any other event loop and is still faster than Event |
|
|
967 | natively. |
|
|
968 | |
|
|
969 | The pure perl implementation is hit in a few sweet spots (both the |
|
|
970 | constant timeout and the use of a single fd hit optimisations in the perl |
|
|
971 | interpreter and the backend itself). Nevertheless this shows that it |
|
|
972 | adds very little overhead in itself. Like any select-based backend its |
|
|
973 | performance becomes really bad with lots of file descriptors (and few of |
|
|
974 | them active), of course, but this was not subject of this benchmark. |
|
|
975 | |
|
|
976 | The C<Event> module has a relatively high setup and callback invocation |
|
|
977 | cost, but overall scores in on the third place. |
|
|
978 | |
|
|
979 | C<Glib>'s memory usage is quite a bit higher, but it features a |
|
|
980 | faster callback invocation and overall ends up in the same class as |
|
|
981 | C<Event>. However, Glib scales extremely badly, doubling the number of |
|
|
982 | watchers increases the processing time by more than a factor of four, |
|
|
983 | making it completely unusable when using larger numbers of watchers |
|
|
984 | (note that only a single file descriptor was used in the benchmark, so |
|
|
985 | inefficiencies of C<poll> do not account for this). |
|
|
986 | |
|
|
987 | The C<Tk> adaptor works relatively well. The fact that it crashes with |
|
|
988 | more than 2000 watchers is a big setback, however, as correctness takes |
|
|
989 | precedence over speed. Nevertheless, its performance is surprising, as the |
|
|
990 | file descriptor is dup()ed for each watcher. This shows that the dup() |
|
|
991 | employed by some adaptors is not a big performance issue (it does incur a |
|
|
992 | hidden memory cost inside the kernel which is not reflected in the figures |
|
|
993 | above). |
|
|
994 | |
|
|
995 | C<POE>, regardless of underlying event loop (whether using its pure |
|
|
996 | perl select-based backend or the Event module, the POE-EV backend |
|
|
997 | couldn't be tested because it wasn't working) shows abysmal performance |
|
|
998 | and memory usage: Watchers use almost 30 times as much memory as |
|
|
999 | EV watchers, and 10 times as much memory as Event (the high memory |
|
|
1000 | requirements are caused by requiring a session for each watcher). Watcher |
|
|
1001 | invocation speed is almost 900 times slower than with AnyEvent's pure perl |
|
|
1002 | implementation. The design of the POE adaptor class in AnyEvent can not |
|
|
1003 | really account for this, as session creation overhead is small compared |
|
|
1004 | to execution of the state machine, which is coded pretty optimally within |
|
|
1005 | L<AnyEvent::Impl::POE>. POE simply seems to be abysmally slow. |
|
|
1006 | |
|
|
1007 | =head3 Summary |
|
|
1008 | |
|
|
1009 | =over 4 |
|
|
1010 | |
|
|
1011 | =item * Using EV through AnyEvent is faster than any other event loop |
|
|
1012 | (even when used without AnyEvent), but most event loops have acceptable |
|
|
1013 | performance with or without AnyEvent. |
|
|
1014 | |
|
|
1015 | =item * The overhead AnyEvent adds is usually much smaller than the overhead of |
|
|
1016 | the actual event loop, only with extremely fast event loops such as EV |
|
|
1017 | adds AnyEvent significant overhead. |
|
|
1018 | |
|
|
1019 | =item * You should avoid POE like the plague if you want performance or |
|
|
1020 | reasonable memory usage. |
|
|
1021 | |
|
|
1022 | =back |
|
|
1023 | |
|
|
1024 | =head2 BENCHMARKING THE LARGE SERVER CASE |
|
|
1025 | |
|
|
1026 | This benchmark atcually benchmarks the event loop itself. It works by |
|
|
1027 | creating a number of "servers": each server consists of a socketpair, a |
|
|
1028 | timeout watcher that gets reset on activity (but never fires), and an I/O |
|
|
1029 | watcher waiting for input on one side of the socket. Each time the socket |
|
|
1030 | watcher reads a byte it will write that byte to a random other "server". |
|
|
1031 | |
|
|
1032 | The effect is that there will be a lot of I/O watchers, only part of which |
|
|
1033 | are active at any one point (so there is a constant number of active |
|
|
1034 | fds for each loop iterstaion, but which fds these are is random). The |
|
|
1035 | timeout is reset each time something is read because that reflects how |
|
|
1036 | most timeouts work (and puts extra pressure on the event loops). |
|
|
1037 | |
|
|
1038 | In this benchmark, we use 10000 socketpairs (20000 sockets), of which 100 |
|
|
1039 | (1%) are active. This mirrors the activity of large servers with many |
|
|
1040 | connections, most of which are idle at any one point in time. |
|
|
1041 | |
|
|
1042 | Source code for this benchmark is found as F<eg/bench2> in the AnyEvent |
|
|
1043 | distribution. |
|
|
1044 | |
|
|
1045 | =head3 Explanation of the columns |
|
|
1046 | |
|
|
1047 | I<sockets> is the number of sockets, and twice the number of "servers" (as |
|
|
1048 | eahc server has a read and write socket end). |
|
|
1049 | |
|
|
1050 | I<create> is the time it takes to create a socketpair (which is |
|
|
1051 | nontrivial) and two watchers: an I/O watcher and a timeout watcher. |
|
|
1052 | |
|
|
1053 | I<request>, the most important value, is the time it takes to handle a |
|
|
1054 | single "request", that is, reading the token from the pipe and forwarding |
|
|
1055 | it to another server. This includes deleting the old timeout and creating |
|
|
1056 | a new one that moves the timeout into the future. |
|
|
1057 | |
|
|
1058 | =head3 Results |
|
|
1059 | |
|
|
1060 | name sockets create request |
|
|
1061 | EV 20000 69.01 11.16 |
|
|
1062 | Perl 20000 75.28 112.76 |
|
|
1063 | Event 20000 212.62 257.32 |
|
|
1064 | Glib 20000 651.16 1896.30 |
|
|
1065 | POE 20000 349.67 12317.24 uses POE::Loop::Event |
|
|
1066 | |
|
|
1067 | =head3 Discussion |
|
|
1068 | |
|
|
1069 | This benchmark I<does> measure scalability and overall performance of the |
|
|
1070 | particular event loop. |
|
|
1071 | |
|
|
1072 | EV is again fastest. Since it is using epoll on my system, the setup time |
|
|
1073 | is relatively high, though. |
|
|
1074 | |
|
|
1075 | Perl surprisingly comes second. It is much faster than the C-based event |
|
|
1076 | loops Event and Glib. |
|
|
1077 | |
|
|
1078 | Event suffers from high setup time as well (look at its code and you will |
|
|
1079 | understand why). Callback invocation also has a high overhead compared to |
|
|
1080 | the C<< $_->() for .. >>-style loop that the Perl event loop uses. Event |
|
|
1081 | uses select or poll in basically all documented configurations. |
|
|
1082 | |
|
|
1083 | Glib is hit hard by its quadratic behaviour w.r.t. many watchers. It |
|
|
1084 | clearly fails to perform with many filehandles or in busy servers. |
|
|
1085 | |
|
|
1086 | POE is still completely out of the picture, taking over 1000 times as long |
|
|
1087 | as EV, and over 100 times as long as the Perl implementation, even though |
|
|
1088 | it uses a C-based event loop in this case. |
|
|
1089 | |
|
|
1090 | =head3 Summary |
|
|
1091 | |
|
|
1092 | =over 4 |
|
|
1093 | |
|
|
1094 | =item * The pure perl implementation performs extremely well, considering |
|
|
1095 | that it uses select. |
|
|
1096 | |
|
|
1097 | =item * Avoid Glib or POE in large projects where performance matters. |
|
|
1098 | |
|
|
1099 | =back |
|
|
1100 | |
|
|
1101 | =head2 BENCHMARKING SMALL SERVERS |
|
|
1102 | |
|
|
1103 | While event loops should scale (and select-based ones do not...) even to |
|
|
1104 | large servers, most programs we (or I :) actually write have only a few |
|
|
1105 | I/O watchers. |
|
|
1106 | |
|
|
1107 | In this benchmark, I use the same benchmark program as in the large server |
|
|
1108 | case, but it uses only eight "servers", of which three are active at any |
|
|
1109 | one time. This should reflect performance for a small server relatively |
|
|
1110 | well. |
|
|
1111 | |
|
|
1112 | The columns are identical to the previous table. |
|
|
1113 | |
|
|
1114 | =head3 Results |
|
|
1115 | |
|
|
1116 | name sockets create request |
|
|
1117 | EV 16 20.00 6.54 |
|
|
1118 | Event 16 81.27 35.86 |
|
|
1119 | Glib 16 32.63 15.48 |
|
|
1120 | Perl 16 24.62 162.37 |
|
|
1121 | POE 16 261.87 276.28 uses POE::Loop::Event |
|
|
1122 | |
|
|
1123 | =head3 Discussion |
|
|
1124 | |
|
|
1125 | The benchmark tries to test the performance of a typical small |
|
|
1126 | server. While knowing how various event loops perform is interesting, keep |
|
|
1127 | in mind that their overhead in this case is usually not as important, due |
|
|
1128 | to the small absolute number of watchers. |
|
|
1129 | |
|
|
1130 | EV is again fastest. |
|
|
1131 | |
|
|
1132 | The C-based event loops Event and Glib come in second this time, as the |
|
|
1133 | overhead of running an iteration is much smaller in C than in Perl (little |
|
|
1134 | code to execute in the inner loop, and perl's function calling overhead is |
|
|
1135 | high, and updating all the data structures is costly). |
|
|
1136 | |
|
|
1137 | The pure perl event loop is much slower, but still competitive. |
|
|
1138 | |
|
|
1139 | POE also performs much better in this case, but is is stillf ar behind the |
|
|
1140 | others. |
|
|
1141 | |
|
|
1142 | =head3 Summary |
|
|
1143 | |
|
|
1144 | =over 4 |
|
|
1145 | |
|
|
1146 | =item * C-based event loops perform very well with small number of |
|
|
1147 | watchers, as the management overhead dominates. |
|
|
1148 | |
|
|
1149 | =back |
|
|
1150 | |
|
|
1151 | |
841 | =head1 FORK |
1152 | =head1 FORK |
842 | |
1153 | |
843 | Most event libraries are not fork-safe. The ones who are usually are |
1154 | Most event libraries are not fork-safe. The ones who are usually are |
844 | because they are so inefficient. Only L<EV> is fully fork-aware. |
1155 | because they are so inefficient. Only L<EV> is fully fork-aware. |
845 | |
1156 | |
846 | If you have to fork, you must either do so I<before> creating your first |
1157 | If you have to fork, you must either do so I<before> creating your first |
847 | watcher OR you must not use AnyEvent at all in the child. |
1158 | watcher OR you must not use AnyEvent at all in the child. |
|
|
1159 | |
848 | |
1160 | |
849 | =head1 SECURITY CONSIDERATIONS |
1161 | =head1 SECURITY CONSIDERATIONS |
850 | |
1162 | |
851 | AnyEvent can be forced to load any event model via |
1163 | AnyEvent can be forced to load any event model via |
852 | $ENV{PERL_ANYEVENT_MODEL}. While this cannot (to my knowledge) be used to |
1164 | $ENV{PERL_ANYEVENT_MODEL}. While this cannot (to my knowledge) be used to |
… | |
… | |
860 | |
1172 | |
861 | BEGIN { delete $ENV{PERL_ANYEVENT_MODEL} } |
1173 | BEGIN { delete $ENV{PERL_ANYEVENT_MODEL} } |
862 | |
1174 | |
863 | use AnyEvent; |
1175 | use AnyEvent; |
864 | |
1176 | |
|
|
1177 | |
865 | =head1 SEE ALSO |
1178 | =head1 SEE ALSO |
866 | |
1179 | |
867 | Event modules: L<Coro::EV>, L<EV>, L<EV::Glib>, L<Glib::EV>, |
1180 | Event modules: L<Coro::EV>, L<EV>, L<EV::Glib>, L<Glib::EV>, |
868 | L<Coro::Event>, L<Event>, L<Glib::Event>, L<Glib>, L<Coro>, L<Tk>, |
1181 | L<Coro::Event>, L<Event>, L<Glib::Event>, L<Glib>, L<Coro>, L<Tk>, |
869 | L<Event::Lib>, L<Qt>. |
1182 | L<Event::Lib>, L<Qt>, L<POE>. |
870 | |
1183 | |
871 | Implementations: L<AnyEvent::Impl::CoroEV>, L<AnyEvent::Impl::EV>, |
1184 | Implementations: L<AnyEvent::Impl::CoroEV>, L<AnyEvent::Impl::EV>, |
872 | L<AnyEvent::Impl::CoroEvent>, L<AnyEvent::Impl::Event>, L<AnyEvent::Impl::Glib>, |
1185 | L<AnyEvent::Impl::CoroEvent>, L<AnyEvent::Impl::Event>, L<AnyEvent::Impl::Glib>, |
873 | L<AnyEvent::Impl::Tk>, L<AnyEvent::Impl::Perl>, L<AnyEvent::Impl::EventLib>, |
1186 | L<AnyEvent::Impl::Tk>, L<AnyEvent::Impl::Perl>, L<AnyEvent::Impl::EventLib>, |
874 | L<AnyEvent::Impl::Qt>. |
1187 | L<AnyEvent::Impl::Qt>, L<AnyEvent::Impl::POE>. |
875 | |
1188 | |
876 | Nontrivial usage examples: L<Net::FCP>, L<Net::XMPP2>. |
1189 | Nontrivial usage examples: L<Net::FCP>, L<Net::XMPP2>. |
|
|
1190 | |
877 | |
1191 | |
878 | =head1 AUTHOR |
1192 | =head1 AUTHOR |
879 | |
1193 | |
880 | Marc Lehmann <schmorp@schmorp.de> |
1194 | Marc Lehmann <schmorp@schmorp.de> |
881 | http://home.schmorp.de/ |
1195 | http://home.schmorp.de/ |