| … | |
… | |
| 363 | might affect timers and time-outs. |
363 | might affect timers and time-outs. |
| 364 | |
364 | |
| 365 | When this is the case, you can call this method, which will update the |
365 | When this is the case, you can call this method, which will update the |
| 366 | event loop's idea of "current time". |
366 | event loop's idea of "current time". |
| 367 | |
367 | |
|
|
368 | A typical example would be a script in a web server (e.g. C<mod_perl>) - |
|
|
369 | when mod_perl executes the script, then the event loop will have the wrong |
|
|
370 | idea about the "current time" (being potentially far in the past, when the |
|
|
371 | script ran the last time). In that case you should arrange a call to C<< |
|
|
372 | AnyEvent->now_update >> each time the web server process wakes up again |
|
|
373 | (e.g. at the start of your script, or in a handler). |
|
|
374 | |
| 368 | Note that updating the time I<might> cause some events to be handled. |
375 | Note that updating the time I<might> cause some events to be handled. |
| 369 | |
376 | |
| 370 | =back |
377 | =back |
| 371 | |
378 | |
| 372 | =head2 SIGNAL WATCHERS |
379 | =head2 SIGNAL WATCHERS |
| … | |
… | |
| 395 | correctly. |
402 | correctly. |
| 396 | |
403 | |
| 397 | Example: exit on SIGINT |
404 | Example: exit on SIGINT |
| 398 | |
405 | |
| 399 | my $w = AnyEvent->signal (signal => "INT", cb => sub { exit 1 }); |
406 | my $w = AnyEvent->signal (signal => "INT", cb => sub { exit 1 }); |
|
|
407 | |
|
|
408 | =head3 Restart Behaviour |
|
|
409 | |
|
|
410 | While restart behaviour is up to the event loop implementation, most will |
|
|
411 | not restart syscalls (that includes L<Async::Interrupt> and AnyEvent's |
|
|
412 | pure perl implementation). |
|
|
413 | |
|
|
414 | =head3 Safe/Unsafe Signals |
|
|
415 | |
|
|
416 | Perl signals can be either "safe" (synchronous to opcode handling) or |
|
|
417 | "unsafe" (asynchronous) - the former might get delayed indefinitely, the |
|
|
418 | latter might corrupt your memory. |
|
|
419 | |
|
|
420 | AnyEvent signal handlers are, in addition, synchronous to the event loop, |
|
|
421 | i.e. they will not interrupt your running perl program but will only be |
|
|
422 | called as part of the normal event handling (just like timer, I/O etc. |
|
|
423 | callbacks, too). |
| 400 | |
424 | |
| 401 | =head3 Signal Races, Delays and Workarounds |
425 | =head3 Signal Races, Delays and Workarounds |
| 402 | |
426 | |
| 403 | Many event loops (e.g. Glib, Tk, Qt, IO::Async) do not support attaching |
427 | Many event loops (e.g. Glib, Tk, Qt, IO::Async) do not support attaching |
| 404 | callbacks to signals in a generic way, which is a pity, as you cannot |
428 | callbacks to signals in a generic way, which is a pity, as you cannot |
| … | |
… | |
| 592 | after => 1, |
616 | after => 1, |
| 593 | cb => sub { $result_ready->send }, |
617 | cb => sub { $result_ready->send }, |
| 594 | ); |
618 | ); |
| 595 | |
619 | |
| 596 | # this "blocks" (while handling events) till the callback |
620 | # this "blocks" (while handling events) till the callback |
| 597 | # calls -<send |
621 | # calls ->send |
| 598 | $result_ready->recv; |
622 | $result_ready->recv; |
| 599 | |
623 | |
| 600 | Example: wait for a timer, but take advantage of the fact that condition |
624 | Example: wait for a timer, but take advantage of the fact that condition |
| 601 | variables are also callable directly. |
625 | variables are also callable directly. |
| 602 | |
626 | |
| … | |
… | |
| 666 | one. For example, a function that pings many hosts in parallel might want |
690 | one. For example, a function that pings many hosts in parallel might want |
| 667 | to use a condition variable for the whole process. |
691 | to use a condition variable for the whole process. |
| 668 | |
692 | |
| 669 | Every call to C<< ->begin >> will increment a counter, and every call to |
693 | Every call to C<< ->begin >> will increment a counter, and every call to |
| 670 | C<< ->end >> will decrement it. If the counter reaches C<0> in C<< ->end |
694 | C<< ->end >> will decrement it. If the counter reaches C<0> in C<< ->end |
| 671 | >>, the (last) callback passed to C<begin> will be executed. That callback |
695 | >>, the (last) callback passed to C<begin> will be executed, passing the |
| 672 | is I<supposed> to call C<< ->send >>, but that is not required. If no |
696 | condvar as first argument. That callback is I<supposed> to call C<< ->send |
| 673 | callback was set, C<send> will be called without any arguments. |
697 | >>, but that is not required. If no group callback was set, C<send> will |
|
|
698 | be called without any arguments. |
| 674 | |
699 | |
| 675 | You can think of C<< $cv->send >> giving you an OR condition (one call |
700 | You can think of C<< $cv->send >> giving you an OR condition (one call |
| 676 | sends), while C<< $cv->begin >> and C<< $cv->end >> giving you an AND |
701 | sends), while C<< $cv->begin >> and C<< $cv->end >> giving you an AND |
| 677 | condition (all C<begin> calls must be C<end>'ed before the condvar sends). |
702 | condition (all C<begin> calls must be C<end>'ed before the condvar sends). |
| 678 | |
703 | |
| … | |
… | |
| 705 | begung can potentially be zero: |
730 | begung can potentially be zero: |
| 706 | |
731 | |
| 707 | my $cv = AnyEvent->condvar; |
732 | my $cv = AnyEvent->condvar; |
| 708 | |
733 | |
| 709 | my %result; |
734 | my %result; |
| 710 | $cv->begin (sub { $cv->send (\%result) }); |
735 | $cv->begin (sub { shift->send (\%result) }); |
| 711 | |
736 | |
| 712 | for my $host (@list_of_hosts) { |
737 | for my $host (@list_of_hosts) { |
| 713 | $cv->begin; |
738 | $cv->begin; |
| 714 | ping_host_then_call_callback $host, sub { |
739 | ping_host_then_call_callback $host, sub { |
| 715 | $result{$host} = ...; |
740 | $result{$host} = ...; |
| … | |
… | |
| 941 | You should check C<$AnyEvent::MODEL> before adding to this array, though: |
966 | You should check C<$AnyEvent::MODEL> before adding to this array, though: |
| 942 | if it is defined then the event loop has already been detected, and the |
967 | if it is defined then the event loop has already been detected, and the |
| 943 | array will be ignored. |
968 | array will be ignored. |
| 944 | |
969 | |
| 945 | Best use C<AnyEvent::post_detect { BLOCK }> when your application allows |
970 | Best use C<AnyEvent::post_detect { BLOCK }> when your application allows |
| 946 | it,as it takes care of these details. |
971 | it, as it takes care of these details. |
| 947 | |
972 | |
| 948 | This variable is mainly useful for modules that can do something useful |
973 | This variable is mainly useful for modules that can do something useful |
| 949 | when AnyEvent is used and thus want to know when it is initialised, but do |
974 | when AnyEvent is used and thus want to know when it is initialised, but do |
| 950 | not need to even load it by default. This array provides the means to hook |
975 | not need to even load it by default. This array provides the means to hook |
| 951 | into AnyEvent passively, without loading it. |
976 | into AnyEvent passively, without loading it. |
|
|
977 | |
|
|
978 | Example: To load Coro::AnyEvent whenever Coro and AnyEvent are used |
|
|
979 | together, you could put this into Coro (this is the actual code used by |
|
|
980 | Coro to accomplish this): |
|
|
981 | |
|
|
982 | if (defined $AnyEvent::MODEL) { |
|
|
983 | # AnyEvent already initialised, so load Coro::AnyEvent |
|
|
984 | require Coro::AnyEvent; |
|
|
985 | } else { |
|
|
986 | # AnyEvent not yet initialised, so make sure to load Coro::AnyEvent |
|
|
987 | # as soon as it is |
|
|
988 | push @AnyEvent::post_detect, sub { require Coro::AnyEvent }; |
|
|
989 | } |
| 952 | |
990 | |
| 953 | =back |
991 | =back |
| 954 | |
992 | |
| 955 | =head1 WHAT TO DO IN A MODULE |
993 | =head1 WHAT TO DO IN A MODULE |
| 956 | |
994 | |
| … | |
… | |
| 1105 | |
1143 | |
| 1106 | package AnyEvent; |
1144 | package AnyEvent; |
| 1107 | |
1145 | |
| 1108 | # basically a tuned-down version of common::sense |
1146 | # basically a tuned-down version of common::sense |
| 1109 | sub common_sense { |
1147 | sub common_sense { |
| 1110 | # no warnings |
1148 | # from common:.sense 1.0 |
| 1111 | ${^WARNING_BITS} ^= ${^WARNING_BITS}; |
1149 | ${^WARNING_BITS} = "\xfc\x3f\x33\x00\x0f\xf3\xcf\xc0\xf3\xfc\x33\x03"; |
| 1112 | # use strict vars subs |
1150 | # use strict vars subs |
| 1113 | $^H |= 0x00000600; |
1151 | $^H |= 0x00000600; |
| 1114 | } |
1152 | } |
| 1115 | |
1153 | |
| 1116 | BEGIN { AnyEvent::common_sense } |
1154 | BEGIN { AnyEvent::common_sense } |
| 1117 | |
1155 | |
| 1118 | use Carp (); |
1156 | use Carp (); |
| 1119 | |
1157 | |
| 1120 | our $VERSION = 4.92; |
1158 | our $VERSION = '5.22'; |
| 1121 | our $MODEL; |
1159 | our $MODEL; |
| 1122 | |
1160 | |
| 1123 | our $AUTOLOAD; |
1161 | our $AUTOLOAD; |
| 1124 | our @ISA; |
1162 | our @ISA; |
| 1125 | |
1163 | |
| 1126 | our @REGISTRY; |
1164 | our @REGISTRY; |
| 1127 | |
|
|
| 1128 | our $WIN32; |
|
|
| 1129 | |
1165 | |
| 1130 | our $VERBOSE; |
1166 | our $VERBOSE; |
| 1131 | |
1167 | |
| 1132 | BEGIN { |
1168 | BEGIN { |
| 1133 | eval "sub WIN32(){ " . (($^O =~ /mswin32/i)*1) ." }"; |
1169 | eval "sub WIN32(){ " . (($^O =~ /mswin32/i)*1) ." }"; |
| … | |
… | |
| 1168 | [Prima:: => AnyEvent::Impl::POE::], |
1204 | [Prima:: => AnyEvent::Impl::POE::], |
| 1169 | # IO::Async is just too broken - we would need workarounds for its |
1205 | # IO::Async is just too broken - we would need workarounds for its |
| 1170 | # byzantine signal and broken child handling, among others. |
1206 | # byzantine signal and broken child handling, among others. |
| 1171 | # IO::Async is rather hard to detect, as it doesn't have any |
1207 | # IO::Async is rather hard to detect, as it doesn't have any |
| 1172 | # obvious default class. |
1208 | # obvious default class. |
| 1173 | # [0, IO::Async:: => AnyEvent::Impl::IOAsync::], # requires special main program |
1209 | [IO::Async:: => AnyEvent::Impl::IOAsync::], # requires special main program |
| 1174 | # [0, IO::Async::Loop:: => AnyEvent::Impl::IOAsync::], # requires special main program |
1210 | [IO::Async::Loop:: => AnyEvent::Impl::IOAsync::], # requires special main program |
| 1175 | # [0, IO::Async::Notifier:: => AnyEvent::Impl::IOAsync::], # requires special main program |
1211 | [IO::Async::Notifier:: => AnyEvent::Impl::IOAsync::], # requires special main program |
|
|
1212 | [AnyEvent::Impl::IOAsync:: => AnyEvent::Impl::IOAsync::], # requires special main program |
| 1176 | ); |
1213 | ); |
| 1177 | |
1214 | |
| 1178 | our %method = map +($_ => 1), |
1215 | our %method = map +($_ => 1), |
| 1179 | qw(io timer time now now_update signal child idle condvar one_event DESTROY); |
1216 | qw(io timer time now now_update signal child idle condvar one_event DESTROY); |
| 1180 | |
1217 | |
| … | |
… | |
| 1287 | # we assume CLOEXEC is already set by perl in all important cases |
1324 | # we assume CLOEXEC is already set by perl in all important cases |
| 1288 | |
1325 | |
| 1289 | ($fh2, $rw) |
1326 | ($fh2, $rw) |
| 1290 | } |
1327 | } |
| 1291 | |
1328 | |
| 1292 | ############################################################################# |
1329 | =head1 SIMPLIFIED AE API |
| 1293 | # "new" API, currently only emulation of it |
1330 | |
| 1294 | ############################################################################# |
1331 | Starting with version 5.0, AnyEvent officially supports a second, much |
|
|
1332 | simpler, API that is designed to reduce the calling, typing and memory |
|
|
1333 | overhead. |
|
|
1334 | |
|
|
1335 | See the L<AE> manpage for details. |
|
|
1336 | |
|
|
1337 | =cut |
| 1295 | |
1338 | |
| 1296 | package AE; |
1339 | package AE; |
| 1297 | |
1340 | |
| 1298 | our $VERSION = $AnyEvent::VERSION; |
1341 | our $VERSION = $AnyEvent::VERSION; |
| 1299 | |
1342 | |
| 1300 | sub io($$$) { |
1343 | sub io($$$) { |
| 1301 | AnyEvent->io (fh => $_[0], poll => $_[1] ? "w" : "r", cb => $_[2]) |
1344 | AnyEvent->io (fh => $_[0], poll => $_[1] ? "w" : "r", cb => $_[2]) |
| 1302 | } |
1345 | } |
| 1303 | |
1346 | |
| 1304 | sub timer($$$) { |
1347 | sub timer($$$) { |
| 1305 | AnyEvent->timer (after => $_[0], interval => $_[1], cb => $_[2]); |
1348 | AnyEvent->timer (after => $_[0], interval => $_[1], cb => $_[2]) |
| 1306 | } |
1349 | } |
| 1307 | |
1350 | |
| 1308 | sub signal($$) { |
1351 | sub signal($$) { |
| 1309 | AnyEvent->signal (signal => $_[0], cb => $_[1]); |
1352 | AnyEvent->signal (signal => $_[0], cb => $_[1]) |
| 1310 | } |
1353 | } |
| 1311 | |
1354 | |
| 1312 | sub child($$) { |
1355 | sub child($$) { |
| 1313 | AnyEvent->child (pid => $_[0], cb => $_[1]); |
1356 | AnyEvent->child (pid => $_[0], cb => $_[1]) |
| 1314 | } |
1357 | } |
| 1315 | |
1358 | |
| 1316 | sub idle($) { |
1359 | sub idle($) { |
| 1317 | AnyEvent->idle (cb => $_[0]); |
1360 | AnyEvent->idle (cb => $_[0]) |
| 1318 | } |
1361 | } |
| 1319 | |
1362 | |
| 1320 | sub cv(;&) { |
1363 | sub cv(;&) { |
| 1321 | AnyEvent->condvar (@_ ? (cb => $_[0]) : ()) |
1364 | AnyEvent->condvar (@_ ? (cb => $_[0]) : ()) |
| 1322 | } |
1365 | } |
| … | |
… | |
| 1335 | |
1378 | |
| 1336 | package AnyEvent::Base; |
1379 | package AnyEvent::Base; |
| 1337 | |
1380 | |
| 1338 | # default implementations for many methods |
1381 | # default implementations for many methods |
| 1339 | |
1382 | |
| 1340 | sub _time { |
1383 | sub _time() { |
| 1341 | # probe for availability of Time::HiRes |
1384 | # probe for availability of Time::HiRes |
| 1342 | if (eval "use Time::HiRes (); Time::HiRes::time (); 1") { |
1385 | if (eval "use Time::HiRes (); Time::HiRes::time (); 1") { |
| 1343 | warn "AnyEvent: using Time::HiRes for sub-second timing accuracy.\n" if $VERBOSE >= 8; |
1386 | warn "AnyEvent: using Time::HiRes for sub-second timing accuracy.\n" if $VERBOSE >= 8; |
| 1344 | *_time = \&Time::HiRes::time; |
1387 | *_time = \&Time::HiRes::time; |
| 1345 | # if (eval "use POSIX (); (POSIX::times())... |
1388 | # if (eval "use POSIX (); (POSIX::times())... |
| … | |
… | |
| 1365 | |
1408 | |
| 1366 | our $HAVE_ASYNC_INTERRUPT; |
1409 | our $HAVE_ASYNC_INTERRUPT; |
| 1367 | |
1410 | |
| 1368 | sub _have_async_interrupt() { |
1411 | sub _have_async_interrupt() { |
| 1369 | $HAVE_ASYNC_INTERRUPT = 1*(!$ENV{PERL_ANYEVENT_AVOID_ASYNC_INTERRUPT} |
1412 | $HAVE_ASYNC_INTERRUPT = 1*(!$ENV{PERL_ANYEVENT_AVOID_ASYNC_INTERRUPT} |
| 1370 | && eval "use Async::Interrupt 1.0 (); 1") |
1413 | && eval "use Async::Interrupt 1.02 (); 1") |
| 1371 | unless defined $HAVE_ASYNC_INTERRUPT; |
1414 | unless defined $HAVE_ASYNC_INTERRUPT; |
| 1372 | |
1415 | |
| 1373 | $HAVE_ASYNC_INTERRUPT |
1416 | $HAVE_ASYNC_INTERRUPT |
| 1374 | } |
1417 | } |
| 1375 | |
1418 | |
| … | |
… | |
| 1378 | our ($SIG_COUNT, $SIG_TW); |
1421 | our ($SIG_COUNT, $SIG_TW); |
| 1379 | |
1422 | |
| 1380 | sub _signal_exec { |
1423 | sub _signal_exec { |
| 1381 | $HAVE_ASYNC_INTERRUPT |
1424 | $HAVE_ASYNC_INTERRUPT |
| 1382 | ? $SIGPIPE_R->drain |
1425 | ? $SIGPIPE_R->drain |
| 1383 | : sysread $SIGPIPE_R, my $dummy, 9; |
1426 | : sysread $SIGPIPE_R, (my $dummy), 9; |
| 1384 | |
1427 | |
| 1385 | while (%SIG_EV) { |
1428 | while (%SIG_EV) { |
| 1386 | for (keys %SIG_EV) { |
1429 | for (keys %SIG_EV) { |
| 1387 | delete $SIG_EV{$_}; |
1430 | delete $SIG_EV{$_}; |
| 1388 | $_->() for values %{ $SIG_CB{$_} || {} }; |
1431 | $_->() for values %{ $SIG_CB{$_} || {} }; |
| … | |
… | |
| 1904 | warn "read: $input\n"; # output what has been read |
1947 | warn "read: $input\n"; # output what has been read |
| 1905 | $cv->send if $input =~ /^q/i; # quit program if /^q/i |
1948 | $cv->send if $input =~ /^q/i; # quit program if /^q/i |
| 1906 | }, |
1949 | }, |
| 1907 | ); |
1950 | ); |
| 1908 | |
1951 | |
| 1909 | my $time_watcher; # can only be used once |
|
|
| 1910 | |
|
|
| 1911 | sub new_timer { |
|
|
| 1912 | $timer = AnyEvent->timer (after => 1, cb => sub { |
1952 | my $time_watcher = AnyEvent->timer (after => 1, interval => 1, cb => sub { |
| 1913 | warn "timeout\n"; # print 'timeout' about every second |
1953 | warn "timeout\n"; # print 'timeout' at most every second |
| 1914 | &new_timer; # and restart the time |
|
|
| 1915 | }); |
1954 | }); |
| 1916 | } |
|
|
| 1917 | |
|
|
| 1918 | new_timer; # create first timer |
|
|
| 1919 | |
1955 | |
| 1920 | $cv->recv; # wait until user enters /^q/i |
1956 | $cv->recv; # wait until user enters /^q/i |
| 1921 | |
1957 | |
| 1922 | =head1 REAL-WORLD EXAMPLE |
1958 | =head1 REAL-WORLD EXAMPLE |
| 1923 | |
1959 | |
| … | |
… | |
| 2054 | through AnyEvent. The benchmark creates a lot of timers (with a zero |
2090 | through AnyEvent. The benchmark creates a lot of timers (with a zero |
| 2055 | timeout) and I/O watchers (watching STDOUT, a pty, to become writable, |
2091 | timeout) and I/O watchers (watching STDOUT, a pty, to become writable, |
| 2056 | which it is), lets them fire exactly once and destroys them again. |
2092 | which it is), lets them fire exactly once and destroys them again. |
| 2057 | |
2093 | |
| 2058 | Source code for this benchmark is found as F<eg/bench> in the AnyEvent |
2094 | Source code for this benchmark is found as F<eg/bench> in the AnyEvent |
| 2059 | distribution. |
2095 | distribution. It uses the L<AE> interface, which makes a real difference |
|
|
2096 | for the EV and Perl backends only. |
| 2060 | |
2097 | |
| 2061 | =head3 Explanation of the columns |
2098 | =head3 Explanation of the columns |
| 2062 | |
2099 | |
| 2063 | I<watcher> is the number of event watchers created/destroyed. Since |
2100 | I<watcher> is the number of event watchers created/destroyed. Since |
| 2064 | different event models feature vastly different performances, each event |
2101 | different event models feature vastly different performances, each event |
| … | |
… | |
| 2085 | watcher. |
2122 | watcher. |
| 2086 | |
2123 | |
| 2087 | =head3 Results |
2124 | =head3 Results |
| 2088 | |
2125 | |
| 2089 | name watchers bytes create invoke destroy comment |
2126 | name watchers bytes create invoke destroy comment |
| 2090 | EV/EV 400000 224 0.47 0.35 0.27 EV native interface |
2127 | EV/EV 100000 223 0.47 0.43 0.27 EV native interface |
| 2091 | EV/Any 100000 224 2.88 0.34 0.27 EV + AnyEvent watchers |
2128 | EV/Any 100000 223 0.48 0.42 0.26 EV + AnyEvent watchers |
| 2092 | CoroEV/Any 100000 224 2.85 0.35 0.28 coroutines + Coro::Signal |
2129 | Coro::EV/Any 100000 223 0.47 0.42 0.26 coroutines + Coro::Signal |
| 2093 | Perl/Any 100000 452 4.13 0.73 0.95 pure perl implementation |
2130 | Perl/Any 100000 431 2.70 0.74 0.92 pure perl implementation |
| 2094 | Event/Event 16000 517 32.20 31.80 0.81 Event native interface |
2131 | Event/Event 16000 516 31.16 31.84 0.82 Event native interface |
| 2095 | Event/Any 16000 590 35.85 31.55 1.06 Event + AnyEvent watchers |
2132 | Event/Any 16000 1203 42.61 34.79 1.80 Event + AnyEvent watchers |
| 2096 | IOAsync/Any 16000 989 38.10 32.77 11.13 via IO::Async::Loop::IO_Poll |
2133 | IOAsync/Any 16000 1911 41.92 27.45 16.81 via IO::Async::Loop::IO_Poll |
| 2097 | IOAsync/Any 16000 990 37.59 29.50 10.61 via IO::Async::Loop::Epoll |
2134 | IOAsync/Any 16000 1726 40.69 26.37 15.25 via IO::Async::Loop::Epoll |
| 2098 | Glib/Any 16000 1357 102.33 12.31 51.00 quadratic behaviour |
2135 | Glib/Any 16000 1118 89.00 12.57 51.17 quadratic behaviour |
| 2099 | Tk/Any 2000 1860 27.20 66.31 14.00 SEGV with >> 2000 watchers |
2136 | Tk/Any 2000 1346 20.96 10.75 8.00 SEGV with >> 2000 watchers |
| 2100 | POE/Event 2000 6328 109.99 751.67 14.02 via POE::Loop::Event |
2137 | POE/Any 2000 6951 108.97 795.32 14.24 via POE::Loop::Event |
| 2101 | POE/Select 2000 6027 94.54 809.13 579.80 via POE::Loop::Select |
2138 | POE/Any 2000 6648 94.79 774.40 575.51 via POE::Loop::Select |
| 2102 | |
2139 | |
| 2103 | =head3 Discussion |
2140 | =head3 Discussion |
| 2104 | |
2141 | |
| 2105 | The benchmark does I<not> measure scalability of the event loop very |
2142 | The benchmark does I<not> measure scalability of the event loop very |
| 2106 | well. For example, a select-based event loop (such as the pure perl one) |
2143 | well. For example, a select-based event loop (such as the pure perl one) |
| … | |
… | |
| 2118 | benchmark machine, handling an event takes roughly 1600 CPU cycles with |
2155 | benchmark machine, handling an event takes roughly 1600 CPU cycles with |
| 2119 | EV, 3100 CPU cycles with AnyEvent's pure perl loop and almost 3000000 CPU |
2156 | EV, 3100 CPU cycles with AnyEvent's pure perl loop and almost 3000000 CPU |
| 2120 | cycles with POE. |
2157 | cycles with POE. |
| 2121 | |
2158 | |
| 2122 | C<EV> is the sole leader regarding speed and memory use, which are both |
2159 | C<EV> is the sole leader regarding speed and memory use, which are both |
| 2123 | maximal/minimal, respectively. Even when going through AnyEvent, it uses |
2160 | maximal/minimal, respectively. When using the L<AE> API there is zero |
|
|
2161 | overhead (when going through the AnyEvent API create is about 5-6 times |
|
|
2162 | slower, with other times being equal, so still uses far less memory than |
| 2124 | far less memory than any other event loop and is still faster than Event |
2163 | any other event loop and is still faster than Event natively). |
| 2125 | natively. |
|
|
| 2126 | |
2164 | |
| 2127 | The pure perl implementation is hit in a few sweet spots (both the |
2165 | The pure perl implementation is hit in a few sweet spots (both the |
| 2128 | constant timeout and the use of a single fd hit optimisations in the perl |
2166 | constant timeout and the use of a single fd hit optimisations in the perl |
| 2129 | interpreter and the backend itself). Nevertheless this shows that it |
2167 | interpreter and the backend itself). Nevertheless this shows that it |
| 2130 | adds very little overhead in itself. Like any select-based backend its |
2168 | adds very little overhead in itself. Like any select-based backend its |
| … | |
… | |
| 2204 | In this benchmark, we use 10000 socket pairs (20000 sockets), of which 100 |
2242 | In this benchmark, we use 10000 socket pairs (20000 sockets), of which 100 |
| 2205 | (1%) are active. This mirrors the activity of large servers with many |
2243 | (1%) are active. This mirrors the activity of large servers with many |
| 2206 | connections, most of which are idle at any one point in time. |
2244 | connections, most of which are idle at any one point in time. |
| 2207 | |
2245 | |
| 2208 | Source code for this benchmark is found as F<eg/bench2> in the AnyEvent |
2246 | Source code for this benchmark is found as F<eg/bench2> in the AnyEvent |
| 2209 | distribution. |
2247 | distribution. It uses the L<AE> interface, which makes a real difference |
|
|
2248 | for the EV and Perl backends only. |
| 2210 | |
2249 | |
| 2211 | =head3 Explanation of the columns |
2250 | =head3 Explanation of the columns |
| 2212 | |
2251 | |
| 2213 | I<sockets> is the number of sockets, and twice the number of "servers" (as |
2252 | I<sockets> is the number of sockets, and twice the number of "servers" (as |
| 2214 | each server has a read and write socket end). |
2253 | each server has a read and write socket end). |
| … | |
… | |
| 2222 | a new one that moves the timeout into the future. |
2261 | a new one that moves the timeout into the future. |
| 2223 | |
2262 | |
| 2224 | =head3 Results |
2263 | =head3 Results |
| 2225 | |
2264 | |
| 2226 | name sockets create request |
2265 | name sockets create request |
| 2227 | EV 20000 69.01 11.16 |
2266 | EV 20000 62.66 7.99 |
| 2228 | Perl 20000 73.32 35.87 |
2267 | Perl 20000 68.32 32.64 |
| 2229 | IOAsync 20000 157.00 98.14 epoll |
2268 | IOAsync 20000 174.06 101.15 epoll |
| 2230 | IOAsync 20000 159.31 616.06 poll |
2269 | IOAsync 20000 174.67 610.84 poll |
| 2231 | Event 20000 212.62 257.32 |
2270 | Event 20000 202.69 242.91 |
| 2232 | Glib 20000 651.16 1896.30 |
2271 | Glib 20000 557.01 1689.52 |
| 2233 | POE 20000 349.67 12317.24 uses POE::Loop::Event |
2272 | POE 20000 341.54 12086.32 uses POE::Loop::Event |
| 2234 | |
2273 | |
| 2235 | =head3 Discussion |
2274 | =head3 Discussion |
| 2236 | |
2275 | |
| 2237 | This benchmark I<does> measure scalability and overall performance of the |
2276 | This benchmark I<does> measure scalability and overall performance of the |
| 2238 | particular event loop. |
2277 | particular event loop. |
| … | |
… | |
| 2364 | As you can see, the AnyEvent + EV combination even beats the |
2403 | As you can see, the AnyEvent + EV combination even beats the |
| 2365 | hand-optimised "raw sockets benchmark", while AnyEvent + its pure perl |
2404 | hand-optimised "raw sockets benchmark", while AnyEvent + its pure perl |
| 2366 | backend easily beats IO::Lambda and POE. |
2405 | backend easily beats IO::Lambda and POE. |
| 2367 | |
2406 | |
| 2368 | And even the 100% non-blocking version written using the high-level (and |
2407 | And even the 100% non-blocking version written using the high-level (and |
| 2369 | slow :) L<AnyEvent::Handle> abstraction beats both POE and IO::Lambda by a |
2408 | slow :) L<AnyEvent::Handle> abstraction beats both POE and IO::Lambda |
| 2370 | large margin, even though it does all of DNS, tcp-connect and socket I/O |
2409 | higher level ("unoptimised") abstractions by a large margin, even though |
| 2371 | in a non-blocking way. |
2410 | it does all of DNS, tcp-connect and socket I/O in a non-blocking way. |
| 2372 | |
2411 | |
| 2373 | The two AnyEvent benchmarks programs can be found as F<eg/ae0.pl> and |
2412 | The two AnyEvent benchmarks programs can be found as F<eg/ae0.pl> and |
| 2374 | F<eg/ae2.pl> in the AnyEvent distribution, the remaining benchmarks are |
2413 | F<eg/ae2.pl> in the AnyEvent distribution, the remaining benchmarks are |
| 2375 | part of the IO::lambda distribution and were used without any changes. |
2414 | part of the IO::Lambda distribution and were used without any changes. |
| 2376 | |
2415 | |
| 2377 | |
2416 | |
| 2378 | =head1 SIGNALS |
2417 | =head1 SIGNALS |
| 2379 | |
2418 | |
| 2380 | AnyEvent currently installs handlers for these signals: |
2419 | AnyEvent currently installs handlers for these signals: |
| … | |
… | |
| 2422 | it's built-in modules) are required to use it. |
2461 | it's built-in modules) are required to use it. |
| 2423 | |
2462 | |
| 2424 | That does not mean that AnyEvent won't take advantage of some additional |
2463 | That does not mean that AnyEvent won't take advantage of some additional |
| 2425 | modules if they are installed. |
2464 | modules if they are installed. |
| 2426 | |
2465 | |
| 2427 | This section epxlains which additional modules will be used, and how they |
2466 | This section explains which additional modules will be used, and how they |
| 2428 | affect AnyEvent's operetion. |
2467 | affect AnyEvent's operation. |
| 2429 | |
2468 | |
| 2430 | =over 4 |
2469 | =over 4 |
| 2431 | |
2470 | |
| 2432 | =item L<Async::Interrupt> |
2471 | =item L<Async::Interrupt> |
| 2433 | |
2472 | |
| … | |
… | |
| 2438 | catch the signals) with some delay (default is 10 seconds, look for |
2477 | catch the signals) with some delay (default is 10 seconds, look for |
| 2439 | C<$AnyEvent::MAX_SIGNAL_LATENCY>). |
2478 | C<$AnyEvent::MAX_SIGNAL_LATENCY>). |
| 2440 | |
2479 | |
| 2441 | If this module is available, then it will be used to implement signal |
2480 | If this module is available, then it will be used to implement signal |
| 2442 | catching, which means that signals will not be delayed, and the event loop |
2481 | catching, which means that signals will not be delayed, and the event loop |
| 2443 | will not be interrupted regularly, which is more efficient (And good for |
2482 | will not be interrupted regularly, which is more efficient (and good for |
| 2444 | battery life on laptops). |
2483 | battery life on laptops). |
| 2445 | |
2484 | |
| 2446 | This affects not just the pure-perl event loop, but also other event loops |
2485 | This affects not just the pure-perl event loop, but also other event loops |
| 2447 | that have no signal handling on their own (e.g. Glib, Tk, Qt). |
2486 | that have no signal handling on their own (e.g. Glib, Tk, Qt). |
| 2448 | |
2487 | |
| … | |
… | |
| 2469 | lot less memory), but otherwise doesn't affect guard operation much. It is |
2508 | lot less memory), but otherwise doesn't affect guard operation much. It is |
| 2470 | purely used for performance. |
2509 | purely used for performance. |
| 2471 | |
2510 | |
| 2472 | =item L<JSON> and L<JSON::XS> |
2511 | =item L<JSON> and L<JSON::XS> |
| 2473 | |
2512 | |
| 2474 | This module is required when you want to read or write JSON data via |
2513 | One of these modules is required when you want to read or write JSON data |
| 2475 | L<AnyEvent::Handle>. It is also written in pure-perl, but can take |
2514 | via L<AnyEvent::Handle>. It is also written in pure-perl, but can take |
| 2476 | advantage of the ultra-high-speed L<JSON::XS> module when it is installed. |
2515 | advantage of the ultra-high-speed L<JSON::XS> module when it is installed. |
| 2477 | |
2516 | |
| 2478 | In fact, L<AnyEvent::Handle> will use L<JSON::XS> by default if it is |
2517 | In fact, L<AnyEvent::Handle> will use L<JSON::XS> by default if it is |
| 2479 | installed. |
2518 | installed. |
| 2480 | |
2519 | |
| … | |
… | |
| 2498 | |
2537 | |
| 2499 | Most event libraries are not fork-safe. The ones who are usually are |
2538 | Most event libraries are not fork-safe. The ones who are usually are |
| 2500 | because they rely on inefficient but fork-safe C<select> or C<poll> |
2539 | because they rely on inefficient but fork-safe C<select> or C<poll> |
| 2501 | calls. Only L<EV> is fully fork-aware. |
2540 | calls. Only L<EV> is fully fork-aware. |
| 2502 | |
2541 | |
|
|
2542 | This means that, in general, you cannot fork and do event processing |
|
|
2543 | in the child if a watcher was created before the fork (which in turn |
|
|
2544 | initialises the event library). |
|
|
2545 | |
| 2503 | If you have to fork, you must either do so I<before> creating your first |
2546 | If you have to fork, you must either do so I<before> creating your first |
| 2504 | watcher OR you must not use AnyEvent at all in the child OR you must do |
2547 | watcher OR you must not use AnyEvent at all in the child OR you must do |
| 2505 | something completely out of the scope of AnyEvent. |
2548 | something completely out of the scope of AnyEvent. |
|
|
2549 | |
|
|
2550 | The problem of doing event processing in the parent I<and> the child |
|
|
2551 | is much more complicated: even for backends that I<are> fork-aware or |
|
|
2552 | fork-safe, their behaviour is not usually what you want: fork clones all |
|
|
2553 | watchers, that means all timers, I/O watchers etc. are active in both |
|
|
2554 | parent and child, which is almost never what you want. |
| 2506 | |
2555 | |
| 2507 | |
2556 | |
| 2508 | =head1 SECURITY CONSIDERATIONS |
2557 | =head1 SECURITY CONSIDERATIONS |
| 2509 | |
2558 | |
| 2510 | AnyEvent can be forced to load any event model via |
2559 | AnyEvent can be forced to load any event model via |
