ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/AnyEvent/README
(Generate patch)

Comparing AnyEvent/README (file contents):
Revision 1.50 by root, Sat Aug 1 09:14:54 2009 UTC vs.
Revision 1.51 by root, Sun Aug 9 16:05:11 2009 UTC

… …
771SUPPORTED EVENT LOOPS/BACKENDS 771SUPPORTED EVENT LOOPS/BACKENDS
772 The available backend classes are (every class has its own manpage): 772 The available backend classes are (every class has its own manpage):
773 773
774 Backends that are autoprobed when no other event loop can be found. 774 Backends that are autoprobed when no other event loop can be found.
775 EV is the preferred backend when no other event loop seems to be in 775 EV is the preferred backend when no other event loop seems to be in
776 use. If EV is not installed, then AnyEvent will try Event, and, 776 use. If EV is not installed, then AnyEvent will fall back to its own
777 failing that, will fall back to its own pure-perl implementation, 777 pure-perl implementation, which is available everywhere as it comes
778 which is available everywhere as it comes with AnyEvent itself. 778 with AnyEvent itself.
779 779
780 AnyEvent::Impl::EV based on EV (interface to libev, best choice). 780 AnyEvent::Impl::EV based on EV (interface to libev, best choice).
781 AnyEvent::Impl::Event based on Event, very stable, few glitches.
782 AnyEvent::Impl::Perl pure-perl implementation, fast and portable. 781 AnyEvent::Impl::Perl pure-perl implementation, fast and portable.
783 782
784 Backends that are transparently being picked up when they are used. 783 Backends that are transparently being picked up when they are used.
785 These will be used when they are currently loaded when the first 784 These will be used when they are currently loaded when the first
786 watcher is created, in which case it is assumed that the application 785 watcher is created, in which case it is assumed that the application
787 is using them. This means that AnyEvent will automatically pick the 786 is using them. This means that AnyEvent will automatically pick the
788 right backend when the main program loads an event module before 787 right backend when the main program loads an event module before
789 anything starts to create watchers. Nothing special needs to be done 788 anything starts to create watchers. Nothing special needs to be done
790 by the main program. 789 by the main program.
791 790
791 AnyEvent::Impl::Event based on Event, very stable, few glitches.
792 AnyEvent::Impl::Glib based on Glib, slow but very stable. 792 AnyEvent::Impl::Glib based on Glib, slow but very stable.
793 AnyEvent::Impl::Tk based on Tk, very broken. 793 AnyEvent::Impl::Tk based on Tk, very broken.
794 AnyEvent::Impl::EventLib based on Event::Lib, leaks memory and worse. 794 AnyEvent::Impl::EventLib based on Event::Lib, leaks memory and worse.
795 AnyEvent::Impl::POE based on POE, very slow, some limitations. 795 AnyEvent::Impl::POE based on POE, very slow, some limitations.
796 AnyEvent::Impl::Irssi used when running within irssi. 796 AnyEvent::Impl::Irssi used when running within irssi.
… …
1032 Event::ExecFlow 1032 Event::ExecFlow
1033 High level API for event-based execution flow control. 1033 High level API for event-based execution flow control.
1034 1034
1035 Coro 1035 Coro
1036 Has special support for AnyEvent via Coro::AnyEvent. 1036 Has special support for AnyEvent via Coro::AnyEvent.
1037
1038SIMPLIFIED AE API
1039 Starting with version 5.0, AnyEvent officially supports a second, much
1040 simpler, API that is designed to reduce the calling, typing and memory
1041 overhead.
1042
1043 See the AE manpage for details.
1037 1044
1038ERROR AND EXCEPTION HANDLING 1045ERROR AND EXCEPTION HANDLING
1039 In general, AnyEvent does not do any error handling - it relies on the 1046 In general, AnyEvent does not do any error handling - it relies on the
1040 caller to do that if required. The AnyEvent::Strict module (see also the 1047 caller to do that if required. The AnyEvent::Strict module (see also the
1041 "PERL_ANYEVENT_STRICT" environment variable, below) provides strict 1048 "PERL_ANYEVENT_STRICT" environment variable, below) provides strict
… …
1366 through AnyEvent. The benchmark creates a lot of timers (with a zero 1373 through AnyEvent. The benchmark creates a lot of timers (with a zero
1367 timeout) and I/O watchers (watching STDOUT, a pty, to become writable, 1374 timeout) and I/O watchers (watching STDOUT, a pty, to become writable,
1368 which it is), lets them fire exactly once and destroys them again. 1375 which it is), lets them fire exactly once and destroys them again.
1369 1376
1370 Source code for this benchmark is found as eg/bench in the AnyEvent 1377 Source code for this benchmark is found as eg/bench in the AnyEvent
1371 distribution. 1378 distribution. It uses the AE interface, which makes a real difference
1379 for the EV and Perl backends only.
1372 1380
1373 Explanation of the columns 1381 Explanation of the columns
1374 *watcher* is the number of event watchers created/destroyed. Since 1382 *watcher* is the number of event watchers created/destroyed. Since
1375 different event models feature vastly different performances, each event 1383 different event models feature vastly different performances, each event
1376 loop was given a number of watchers so that overall runtime is 1384 loop was given a number of watchers so that overall runtime is
… …
1395 *destroy* is the time, in microseconds, that it takes to destroy a 1403 *destroy* is the time, in microseconds, that it takes to destroy a
1396 single watcher. 1404 single watcher.
1397 1405
1398 Results 1406 Results
1399 name watchers bytes create invoke destroy comment 1407 name watchers bytes create invoke destroy comment
1400 EV/EV 400000 224 0.47 0.35 0.27 EV native interface 1408 EV/EV 100000 223 0.47 0.43 0.27 EV native interface
1401 EV/Any 100000 224 2.88 0.34 0.27 EV + AnyEvent watchers 1409 EV/Any 100000 223 0.48 0.42 0.26 EV + AnyEvent watchers
1402 CoroEV/Any 100000 224 2.85 0.35 0.28 coroutines + Coro::Signal 1410 Coro::EV/Any 100000 223 0.47 0.42 0.26 coroutines + Coro::Signal
1403 Perl/Any 100000 452 4.13 0.73 0.95 pure perl implementation 1411 Perl/Any 100000 431 2.70 0.74 0.92 pure perl implementation
1404 Event/Event 16000 517 32.20 31.80 0.81 Event native interface 1412 Event/Event 16000 516 31.16 31.84 0.82 Event native interface
1405 Event/Any 16000 590 35.85 31.55 1.06 Event + AnyEvent watchers 1413 Event/Any 16000 1203 42.61 34.79 1.80 Event + AnyEvent watchers
1406 IOAsync/Any 16000 989 38.10 32.77 11.13 via IO::Async::Loop::IO_Poll 1414 IOAsync/Any 16000 1911 41.92 27.45 16.81 via IO::Async::Loop::IO_Poll
1407 IOAsync/Any 16000 990 37.59 29.50 10.61 via IO::Async::Loop::Epoll 1415 IOAsync/Any 16000 1726 40.69 26.37 15.25 via IO::Async::Loop::Epoll
1408 Glib/Any 16000 1357 102.33 12.31 51.00 quadratic behaviour 1416 Glib/Any 16000 1118 89.00 12.57 51.17 quadratic behaviour
1409 Tk/Any 2000 1860 27.20 66.31 14.00 SEGV with >> 2000 watchers 1417 Tk/Any 2000 1346 20.96 10.75 8.00 SEGV with >> 2000 watchers
1410 POE/Event 2000 6328 109.99 751.67 14.02 via POE::Loop::Event 1418 POE/Any 2000 6951 108.97 795.32 14.24 via POE::Loop::Event
1411 POE/Select 2000 6027 94.54 809.13 579.80 via POE::Loop::Select 1419 POE/Any 2000 6648 94.79 774.40 575.51 via POE::Loop::Select
1412 1420
1413 Discussion 1421 Discussion
1414 The benchmark does *not* measure scalability of the event loop very 1422 The benchmark does *not* measure scalability of the event loop very
1415 well. For example, a select-based event loop (such as the pure perl one) 1423 well. For example, a select-based event loop (such as the pure perl one)
1416 can never compete with an event loop that uses epoll when the number of 1424 can never compete with an event loop that uses epoll when the number of
… …
1427 benchmark machine, handling an event takes roughly 1600 CPU cycles with 1435 benchmark machine, handling an event takes roughly 1600 CPU cycles with
1428 EV, 3100 CPU cycles with AnyEvent's pure perl loop and almost 3000000 1436 EV, 3100 CPU cycles with AnyEvent's pure perl loop and almost 3000000
1429 CPU cycles with POE. 1437 CPU cycles with POE.
1430 1438
1431 "EV" is the sole leader regarding speed and memory use, which are both 1439 "EV" is the sole leader regarding speed and memory use, which are both
1432 maximal/minimal, respectively. Even when going through AnyEvent, it uses 1440 maximal/minimal, respectively. When using the AE API there is zero
1441 overhead (when going through the AnyEvent API create is about 5-6 times
1442 slower, with other times being equal, so still uses far less memory than
1433 far less memory than any other event loop and is still faster than Event 1443 any other event loop and is still faster than Event natively).
1434 natively.
1435 1444
1436 The pure perl implementation is hit in a few sweet spots (both the 1445 The pure perl implementation is hit in a few sweet spots (both the
1437 constant timeout and the use of a single fd hit optimisations in the 1446 constant timeout and the use of a single fd hit optimisations in the
1438 perl interpreter and the backend itself). Nevertheless this shows that 1447 perl interpreter and the backend itself). Nevertheless this shows that
1439 it adds very little overhead in itself. Like any select-based backend 1448 it adds very little overhead in itself. Like any select-based backend
… …
1509 In this benchmark, we use 10000 socket pairs (20000 sockets), of which 1518 In this benchmark, we use 10000 socket pairs (20000 sockets), of which
1510 100 (1%) are active. This mirrors the activity of large servers with 1519 100 (1%) are active. This mirrors the activity of large servers with
1511 many connections, most of which are idle at any one point in time. 1520 many connections, most of which are idle at any one point in time.
1512 1521
1513 Source code for this benchmark is found as eg/bench2 in the AnyEvent 1522 Source code for this benchmark is found as eg/bench2 in the AnyEvent
1514 distribution. 1523 distribution. It uses the AE interface, which makes a real difference
1524 for the EV and Perl backends only.
1515 1525
1516 Explanation of the columns 1526 Explanation of the columns
1517 *sockets* is the number of sockets, and twice the number of "servers" 1527 *sockets* is the number of sockets, and twice the number of "servers"
1518 (as each server has a read and write socket end). 1528 (as each server has a read and write socket end).
1519 1529
… …
1525 forwarding it to another server. This includes deleting the old timeout 1535 forwarding it to another server. This includes deleting the old timeout
1526 and creating a new one that moves the timeout into the future. 1536 and creating a new one that moves the timeout into the future.
1527 1537
1528 Results 1538 Results
1529 name sockets create request 1539 name sockets create request
1530 EV 20000 69.01 11.16 1540 EV 20000 62.66 7.99
1531 Perl 20000 73.32 35.87 1541 Perl 20000 68.32 32.64
1532 IOAsync 20000 157.00 98.14 epoll 1542 IOAsync 20000 174.06 101.15 epoll
1533 IOAsync 20000 159.31 616.06 poll 1543 IOAsync 20000 174.67 610.84 poll
1534 Event 20000 212.62 257.32 1544 Event 20000 202.69 242.91
1535 Glib 20000 651.16 1896.30 1545 Glib 20000 557.01 1689.52
1536 POE 20000 349.67 12317.24 uses POE::Loop::Event 1546 POE 20000 341.54 12086.32 uses POE::Loop::Event
1537 1547
1538 Discussion 1548 Discussion
1539 This benchmark *does* measure scalability and overall performance of the 1549 This benchmark *does* measure scalability and overall performance of the
1540 particular event loop. 1550 particular event loop.
1541 1551

Diff Legend

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