ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/libev/ev.3
(Generate patch)

Comparing libev/ev.3 (file contents):
Revision 1.36 by root, Thu Nov 29 20:05:59 2007 UTC vs.
Revision 1.40 by root, Fri Dec 7 19:23:48 2007 UTC

… …
127.\} 127.\}
128.rm #[ #] #H #V #F C 128.rm #[ #] #H #V #F C
129.\" ======================================================================== 129.\" ========================================================================
130.\" 130.\"
131.IX Title ""<STANDARD INPUT>" 1" 131.IX Title ""<STANDARD INPUT>" 1"
132.TH "<STANDARD INPUT>" 1 "2007-11-29" "perl v5.8.8" "User Contributed Perl Documentation" 132.TH "<STANDARD INPUT>" 1 "2007-12-07" "perl v5.8.8" "User Contributed Perl Documentation"
133.SH "NAME" 133.SH "NAME"
134libev \- a high performance full\-featured event loop written in C 134libev \- a high performance full\-featured event loop written in C
135.SH "SYNOPSIS" 135.SH "SYNOPSIS"
136.IX Header "SYNOPSIS" 136.IX Header "SYNOPSIS"
137.Vb 1 137.Vb 1
… …
196\& return 0; 196\& return 0;
197\& } 197\& }
198.Ve 198.Ve
199.SH "DESCRIPTION" 199.SH "DESCRIPTION"
200.IX Header "DESCRIPTION" 200.IX Header "DESCRIPTION"
201The newest version of this document is also available as a html-formatted
202web page you might find easier to navigate when reading it for the first
203time: <http://cvs.schmorp.de/libev/ev.html>.
204.PP
201Libev is an event loop: you register interest in certain events (such as a 205Libev is an event loop: you register interest in certain events (such as a
202file descriptor being readable or a timeout occuring), and it will manage 206file descriptor being readable or a timeout occuring), and it will manage
203these event sources and provide your program with events. 207these event sources and provide your program with events.
204.PP 208.PP
205To do this, it must take more or less complete control over your process 209To do this, it must take more or less complete control over your process
… …
420a fork, you can also make libev check for a fork in each iteration by 424a fork, you can also make libev check for a fork in each iteration by
421enabling this flag. 425enabling this flag.
422.Sp 426.Sp
423This works by calling \f(CW\*(C`getpid ()\*(C'\fR on every iteration of the loop, 427This works by calling \f(CW\*(C`getpid ()\*(C'\fR on every iteration of the loop,
424and thus this might slow down your event loop if you do a lot of loop 428and thus this might slow down your event loop if you do a lot of loop
425iterations and little real work, but is usually not noticable (on my 429iterations and little real work, but is usually not noticeable (on my
426Linux system for example, \f(CW\*(C`getpid\*(C'\fR is actually a simple 5\-insn sequence 430Linux system for example, \f(CW\*(C`getpid\*(C'\fR is actually a simple 5\-insn sequence
427without a syscall and thus \fIvery\fR fast, but my Linux system also has 431without a syscall and thus \fIvery\fR fast, but my Linux system also has
428\&\f(CW\*(C`pthread_atfork\*(C'\fR which is even faster). 432\&\f(CW\*(C`pthread_atfork\*(C'\fR which is even faster).
429.Sp 433.Sp
430The big advantage of this flag is that you can forget about fork (and 434The big advantage of this flag is that you can forget about fork (and
… …
581.IP "ev_loop_fork (loop)" 4 585.IP "ev_loop_fork (loop)" 4
582.IX Item "ev_loop_fork (loop)" 586.IX Item "ev_loop_fork (loop)"
583Like \f(CW\*(C`ev_default_fork\*(C'\fR, but acts on an event loop created by 587Like \f(CW\*(C`ev_default_fork\*(C'\fR, but acts on an event loop created by
584\&\f(CW\*(C`ev_loop_new\*(C'\fR. Yes, you have to call this on every allocated event loop 588\&\f(CW\*(C`ev_loop_new\*(C'\fR. Yes, you have to call this on every allocated event loop
585after fork, and how you do this is entirely your own problem. 589after fork, and how you do this is entirely your own problem.
590.IP "unsigned int ev_loop_count (loop)" 4
591.IX Item "unsigned int ev_loop_count (loop)"
592Returns the count of loop iterations for the loop, which is identical to
593the number of times libev did poll for new events. It starts at \f(CW0\fR and
594happily wraps around with enough iterations.
595.Sp
596This value can sometimes be useful as a generation counter of sorts (it
597\&\*(L"ticks\*(R" the number of loop iterations), as it roughly corresponds with
598\&\f(CW\*(C`ev_prepare\*(C'\fR and \f(CW\*(C`ev_check\*(C'\fR calls.
586.IP "unsigned int ev_backend (loop)" 4 599.IP "unsigned int ev_backend (loop)" 4
587.IX Item "unsigned int ev_backend (loop)" 600.IX Item "unsigned int ev_backend (loop)"
588Returns one of the \f(CW\*(C`EVBACKEND_*\*(C'\fR flags indicating the event backend in 601Returns one of the \f(CW\*(C`EVBACKEND_*\*(C'\fR flags indicating the event backend in
589use. 602use.
590.IP "ev_tstamp ev_now (loop)" 4 603.IP "ev_tstamp ev_now (loop)" 4
… …
885Returns the callback currently set on the watcher. 898Returns the callback currently set on the watcher.
886.IP "ev_cb_set (ev_TYPE *watcher, callback)" 4 899.IP "ev_cb_set (ev_TYPE *watcher, callback)" 4
887.IX Item "ev_cb_set (ev_TYPE *watcher, callback)" 900.IX Item "ev_cb_set (ev_TYPE *watcher, callback)"
888Change the callback. You can change the callback at virtually any time 901Change the callback. You can change the callback at virtually any time
889(modulo threads). 902(modulo threads).
903.IP "ev_set_priority (ev_TYPE *watcher, priority)" 4
904.IX Item "ev_set_priority (ev_TYPE *watcher, priority)"
905.PD 0
906.IP "int ev_priority (ev_TYPE *watcher)" 4
907.IX Item "int ev_priority (ev_TYPE *watcher)"
908.PD
909Set and query the priority of the watcher. The priority is a small
910integer between \f(CW\*(C`EV_MAXPRI\*(C'\fR (default: \f(CW2\fR) and \f(CW\*(C`EV_MINPRI\*(C'\fR
911(default: \f(CW\*(C`\-2\*(C'\fR). Pending watchers with higher priority will be invoked
912before watchers with lower priority, but priority will not keep watchers
913from being executed (except for \f(CW\*(C`ev_idle\*(C'\fR watchers).
914.Sp
915This means that priorities are \fIonly\fR used for ordering callback
916invocation after new events have been received. This is useful, for
917example, to reduce latency after idling, or more often, to bind two
918watchers on the same event and make sure one is called first.
919.Sp
920If you need to suppress invocation when higher priority events are pending
921you need to look at \f(CW\*(C`ev_idle\*(C'\fR watchers, which provide this functionality.
922.Sp
923The default priority used by watchers when no priority has been set is
924always \f(CW0\fR, which is supposed to not be too high and not be too low :).
925.Sp
926Setting a priority outside the range of \f(CW\*(C`EV_MINPRI\*(C'\fR to \f(CW\*(C`EV_MAXPRI\*(C'\fR is
927fine, as long as you do not mind that the priority value you query might
928or might not have been adjusted to be within valid range.
890.Sh "\s-1ASSOCIATING\s0 \s-1CUSTOM\s0 \s-1DATA\s0 \s-1WITH\s0 A \s-1WATCHER\s0" 929.Sh "\s-1ASSOCIATING\s0 \s-1CUSTOM\s0 \s-1DATA\s0 \s-1WITH\s0 A \s-1WATCHER\s0"
891.IX Subsection "ASSOCIATING CUSTOM DATA WITH A WATCHER" 930.IX Subsection "ASSOCIATING CUSTOM DATA WITH A WATCHER"
892Each watcher has, by default, a member \f(CW\*(C`void *data\*(C'\fR that you can change 931Each watcher has, by default, a member \f(CW\*(C`void *data\*(C'\fR that you can change
893and read at any time, libev will completely ignore it. This can be used 932and read at any time, libev will completely ignore it. This can be used
894to associate arbitrary data with your watcher. If you need more data and 933to associate arbitrary data with your watcher. If you need more data and
… …
1005it is best to always use non-blocking I/O: An extra \f(CW\*(C`read\*(C'\fR(2) returning 1044it is best to always use non-blocking I/O: An extra \f(CW\*(C`read\*(C'\fR(2) returning
1006\&\f(CW\*(C`EAGAIN\*(C'\fR is far preferable to a program hanging until some data arrives. 1045\&\f(CW\*(C`EAGAIN\*(C'\fR is far preferable to a program hanging until some data arrives.
1007.PP 1046.PP
1008If you cannot run the fd in non-blocking mode (for example you should not 1047If you cannot run the fd in non-blocking mode (for example you should not
1009play around with an Xlib connection), then you have to seperately re-test 1048play around with an Xlib connection), then you have to seperately re-test
1010wether a file descriptor is really ready with a known-to-be good interface 1049whether a file descriptor is really ready with a known-to-be good interface
1011such as poll (fortunately in our Xlib example, Xlib already does this on 1050such as poll (fortunately in our Xlib example, Xlib already does this on
1012its own, so its quite safe to use). 1051its own, so its quite safe to use).
1013.IP "ev_io_init (ev_io *, callback, int fd, int events)" 4 1052.IP "ev_io_init (ev_io *, callback, int fd, int events)" 4
1014.IX Item "ev_io_init (ev_io *, callback, int fd, int events)" 1053.IX Item "ev_io_init (ev_io *, callback, int fd, int events)"
1015.PD 0 1054.PD 0
… …
1499\& ev_stat_start (loop, &passwd); 1538\& ev_stat_start (loop, &passwd);
1500.Ve 1539.Ve
1501.ie n .Sh """ev_idle"" \- when you've got nothing better to do..." 1540.ie n .Sh """ev_idle"" \- when you've got nothing better to do..."
1502.el .Sh "\f(CWev_idle\fP \- when you've got nothing better to do..." 1541.el .Sh "\f(CWev_idle\fP \- when you've got nothing better to do..."
1503.IX Subsection "ev_idle - when you've got nothing better to do..." 1542.IX Subsection "ev_idle - when you've got nothing better to do..."
1504Idle watchers trigger events when there are no other events are pending 1543Idle watchers trigger events when no other events of the same or higher
1505(prepare, check and other idle watchers do not count). That is, as long 1544priority are pending (prepare, check and other idle watchers do not
1506as your process is busy handling sockets or timeouts (or even signals, 1545count).
1507imagine) it will not be triggered. But when your process is idle all idle 1546.PP
1508watchers are being called again and again, once per event loop iteration \- 1547That is, as long as your process is busy handling sockets or timeouts
1548(or even signals, imagine) of the same or higher priority it will not be
1549triggered. But when your process is idle (or only lower-priority watchers
1550are pending), the idle watchers are being called once per event loop
1509until stopped, that is, or your process receives more events and becomes 1551iteration \- until stopped, that is, or your process receives more events
1510busy. 1552and becomes busy again with higher priority stuff.
1511.PP 1553.PP
1512The most noteworthy effect is that as long as any idle watchers are 1554The most noteworthy effect is that as long as any idle watchers are
1513active, the process will not block when waiting for new events. 1555active, the process will not block when waiting for new events.
1514.PP 1556.PP
1515Apart from keeping your process non-blocking (which is a useful 1557Apart from keeping your process non-blocking (which is a useful
… …
1610\& if (revents & EV_READ ) fd->revents |= fd->events & POLLIN; 1652\& if (revents & EV_READ ) fd->revents |= fd->events & POLLIN;
1611\& if (revents & EV_WRITE) fd->revents |= fd->events & POLLOUT; 1653\& if (revents & EV_WRITE) fd->revents |= fd->events & POLLOUT;
1612\& } 1654\& }
1613.Ve 1655.Ve
1614.PP 1656.PP
1615.Vb 7 1657.Vb 8
1616\& // create io watchers for each fd and a timer before blocking 1658\& // create io watchers for each fd and a timer before blocking
1617\& static void 1659\& static void
1618\& adns_prepare_cb (ev_loop *loop, ev_prepare *w, int revents) 1660\& adns_prepare_cb (ev_loop *loop, ev_prepare *w, int revents)
1619\& { 1661\& {
1620\& int timeout = 3600000;truct pollfd fds [nfd]; 1662\& int timeout = 3600000;
1663\& struct pollfd fds [nfd];
1621\& // actual code will need to loop here and realloc etc. 1664\& // actual code will need to loop here and realloc etc.
1622\& adns_beforepoll (ads, fds, &nfd, &timeout, timeval_from (ev_time ())); 1665\& adns_beforepoll (ads, fds, &nfd, &timeout, timeval_from (ev_time ()));
1623.Ve 1666.Ve
1624.PP 1667.PP
1625.Vb 3 1668.Vb 3
… …
1952\& } 1995\& }
1953.Ve 1996.Ve
1954.SH "MACRO MAGIC" 1997.SH "MACRO MAGIC"
1955.IX Header "MACRO MAGIC" 1998.IX Header "MACRO MAGIC"
1956Libev can be compiled with a variety of options, the most fundemantal is 1999Libev can be compiled with a variety of options, the most fundemantal is
1957\&\f(CW\*(C`EV_MULTIPLICITY\*(C'\fR. This option determines wether (most) functions and 2000\&\f(CW\*(C`EV_MULTIPLICITY\*(C'\fR. This option determines whether (most) functions and
1958callbacks have an initial \f(CW\*(C`struct ev_loop *\*(C'\fR argument. 2001callbacks have an initial \f(CW\*(C`struct ev_loop *\*(C'\fR argument.
1959.PP 2002.PP
1960To make it easier to write programs that cope with either variant, the 2003To make it easier to write programs that cope with either variant, the
1961following macros are defined: 2004following macros are defined:
1962.ie n .IP """EV_A""\fR, \f(CW""EV_A_""" 4 2005.ie n .IP """EV_A""\fR, \f(CW""EV_A_""" 4
… …
1998.IX Item "EV_DEFAULT, EV_DEFAULT_" 2041.IX Item "EV_DEFAULT, EV_DEFAULT_"
1999Similar to the other two macros, this gives you the value of the default 2042Similar to the other two macros, this gives you the value of the default
2000loop, if multiple loops are supported (\*(L"ev loop default\*(R"). 2043loop, if multiple loops are supported (\*(L"ev loop default\*(R").
2001.PP 2044.PP
2002Example: Declare and initialise a check watcher, utilising the above 2045Example: Declare and initialise a check watcher, utilising the above
2003macros so it will work regardless of wether multiple loops are supported 2046macros so it will work regardless of whether multiple loops are supported
2004or not. 2047or not.
2005.PP 2048.PP
2006.Vb 5 2049.Vb 5
2007\& static void 2050\& static void
2008\& check_cb (EV_P_ ev_timer *w, int revents) 2051\& check_cb (EV_P_ ev_timer *w, int revents)
… …
2235If undefined or defined to \f(CW1\fR, then all event-loop-specific functions 2278If undefined or defined to \f(CW1\fR, then all event-loop-specific functions
2236will have the \f(CW\*(C`struct ev_loop *\*(C'\fR as first argument, and you can create 2279will have the \f(CW\*(C`struct ev_loop *\*(C'\fR as first argument, and you can create
2237additional independent event loops. Otherwise there will be no support 2280additional independent event loops. Otherwise there will be no support
2238for multiple event loops and there is no first event loop pointer 2281for multiple event loops and there is no first event loop pointer
2239argument. Instead, all functions act on the single default loop. 2282argument. Instead, all functions act on the single default loop.
2283.IP "\s-1EV_MINPRI\s0" 4
2284.IX Item "EV_MINPRI"
2285.PD 0
2286.IP "\s-1EV_MAXPRI\s0" 4
2287.IX Item "EV_MAXPRI"
2288.PD
2289The range of allowed priorities. \f(CW\*(C`EV_MINPRI\*(C'\fR must be smaller or equal to
2290\&\f(CW\*(C`EV_MAXPRI\*(C'\fR, but otherwise there are no non-obvious limitations. You can
2291provide for more priorities by overriding those symbols (usually defined
2292to be \f(CW\*(C`\-2\*(C'\fR and \f(CW2\fR, respectively).
2293.Sp
2294When doing priority-based operations, libev usually has to linearly search
2295all the priorities, so having many of them (hundreds) uses a lot of space
2296and time, so using the defaults of five priorities (\-2 .. +2) is usually
2297fine.
2298.Sp
2299If your embedding app does not need any priorities, defining these both to
2300\&\f(CW0\fR will save some memory and cpu.
2240.IP "\s-1EV_PERIODIC_ENABLE\s0" 4 2301.IP "\s-1EV_PERIODIC_ENABLE\s0" 4
2241.IX Item "EV_PERIODIC_ENABLE" 2302.IX Item "EV_PERIODIC_ENABLE"
2242If undefined or defined to be \f(CW1\fR, then periodic timers are supported. If 2303If undefined or defined to be \f(CW1\fR, then periodic timers are supported. If
2304defined to be \f(CW0\fR, then they are not. Disabling them saves a few kB of
2305code.
2306.IP "\s-1EV_IDLE_ENABLE\s0" 4
2307.IX Item "EV_IDLE_ENABLE"
2308If undefined or defined to be \f(CW1\fR, then idle watchers are supported. If
2243defined to be \f(CW0\fR, then they are not. Disabling them saves a few kB of 2309defined to be \f(CW0\fR, then they are not. Disabling them saves a few kB of
2244code. 2310code.
2245.IP "\s-1EV_EMBED_ENABLE\s0" 4 2311.IP "\s-1EV_EMBED_ENABLE\s0" 4
2246.IX Item "EV_EMBED_ENABLE" 2312.IX Item "EV_EMBED_ENABLE"
2247If undefined or defined to be \f(CW1\fR, then embed watchers are supported. If 2313If undefined or defined to be \f(CW1\fR, then embed watchers are supported. If
… …
2338.SH "COMPLEXITIES" 2404.SH "COMPLEXITIES"
2339.IX Header "COMPLEXITIES" 2405.IX Header "COMPLEXITIES"
2340In this section the complexities of (many of) the algorithms used inside 2406In this section the complexities of (many of) the algorithms used inside
2341libev will be explained. For complexity discussions about backends see the 2407libev will be explained. For complexity discussions about backends see the
2342documentation for \f(CW\*(C`ev_default_init\*(C'\fR. 2408documentation for \f(CW\*(C`ev_default_init\*(C'\fR.
2409.Sp
2410All of the following are about amortised time: If an array needs to be
2411extended, libev needs to realloc and move the whole array, but this
2412happens asymptotically never with higher number of elements, so O(1) might
2413mean it might do a lengthy realloc operation in rare cases, but on average
2414it is much faster and asymptotically approaches constant time.
2343.RS 4 2415.RS 4
2344.IP "Starting and stopping timer/periodic watchers: O(log skipped_other_timers)" 4 2416.IP "Starting and stopping timer/periodic watchers: O(log skipped_other_timers)" 4
2345.IX Item "Starting and stopping timer/periodic watchers: O(log skipped_other_timers)" 2417.IX Item "Starting and stopping timer/periodic watchers: O(log skipped_other_timers)"
2346.PD 0 2418This means that, when you have a watcher that triggers in one hour and
2419there are 100 watchers that would trigger before that then inserting will
2420have to skip those 100 watchers.
2347.IP "Changing timer/periodic watchers (by autorepeat, again): O(log skipped_other_timers)" 4 2421.IP "Changing timer/periodic watchers (by autorepeat, again): O(log skipped_other_timers)" 4
2348.IX Item "Changing timer/periodic watchers (by autorepeat, again): O(log skipped_other_timers)" 2422.IX Item "Changing timer/periodic watchers (by autorepeat, again): O(log skipped_other_timers)"
2423That means that for changing a timer costs less than removing/adding them
2424as only the relative motion in the event queue has to be paid for.
2349.IP "Starting io/check/prepare/idle/signal/child watchers: O(1)" 4 2425.IP "Starting io/check/prepare/idle/signal/child watchers: O(1)" 4
2350.IX Item "Starting io/check/prepare/idle/signal/child watchers: O(1)" 2426.IX Item "Starting io/check/prepare/idle/signal/child watchers: O(1)"
2351.IP "Stopping check/prepare/idle watchers: O(1)" 4 2427These just add the watcher into an array or at the head of a list.
2352.IX Item "Stopping check/prepare/idle watchers: O(1)" 2428=item Stopping check/prepare/idle watchers: O(1)
2353.IP "Stopping an io/signal/child watcher: O(number_of_watchers_for_this_(fd/signal/pid % \s-1EV_PID_HASHSIZE\s0))" 4 2429.IP "Stopping an io/signal/child watcher: O(number_of_watchers_for_this_(fd/signal/pid % \s-1EV_PID_HASHSIZE\s0))" 4
2354.IX Item "Stopping an io/signal/child watcher: O(number_of_watchers_for_this_(fd/signal/pid % EV_PID_HASHSIZE))" 2430.IX Item "Stopping an io/signal/child watcher: O(number_of_watchers_for_this_(fd/signal/pid % EV_PID_HASHSIZE))"
2431These watchers are stored in lists then need to be walked to find the
2432correct watcher to remove. The lists are usually short (you don't usually
2433have many watchers waiting for the same fd or signal).
2355.IP "Finding the next timer per loop iteration: O(1)" 4 2434.IP "Finding the next timer per loop iteration: O(1)" 4
2356.IX Item "Finding the next timer per loop iteration: O(1)" 2435.IX Item "Finding the next timer per loop iteration: O(1)"
2436.PD 0
2357.IP "Each change on a file descriptor per loop iteration: O(number_of_watchers_for_this_fd)" 4 2437.IP "Each change on a file descriptor per loop iteration: O(number_of_watchers_for_this_fd)" 4
2358.IX Item "Each change on a file descriptor per loop iteration: O(number_of_watchers_for_this_fd)" 2438.IX Item "Each change on a file descriptor per loop iteration: O(number_of_watchers_for_this_fd)"
2439.PD
2440A change means an I/O watcher gets started or stopped, which requires
2441libev to recalculate its status (and possibly tell the kernel).
2359.IP "Activating one watcher: O(1)" 4 2442.IP "Activating one watcher: O(1)" 4
2360.IX Item "Activating one watcher: O(1)" 2443.IX Item "Activating one watcher: O(1)"
2444.PD 0
2445.IP "Priority handling: O(number_of_priorities)" 4
2446.IX Item "Priority handling: O(number_of_priorities)"
2447.PD
2448Priorities are implemented by allocating some space for each
2449priority. When doing priority-based operations, libev usually has to
2450linearly search all the priorities.
2361.RE 2451.RE
2362.RS 4 2452.RS 4
2363.PD
2364.SH "AUTHOR" 2453.SH "AUTHOR"
2365.IX Header "AUTHOR" 2454.IX Header "AUTHOR"
2366Marc Lehmann <libev@schmorp.de>. 2455Marc Lehmann <libev@schmorp.de>.

Diff Legend

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