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

Comparing libev/ev.pod (file contents):
Revision 1.121 by root, Mon Jan 28 12:13:54 2008 UTC vs.
Revision 1.130 by root, Wed Feb 6 18:34:24 2008 UTC

… …
774=item C<EV_FORK> 774=item C<EV_FORK>
775 775
776The event loop has been resumed in the child process after fork (see 776The event loop has been resumed in the child process after fork (see
777C<ev_fork>). 777C<ev_fork>).
778 778
779=item C<EV_ASYNC>
780
781The given async watcher has been asynchronously notified (see C<ev_async>).
782
779=item C<EV_ERROR> 783=item C<EV_ERROR>
780 784
781An unspecified error has occured, the watcher has been stopped. This might 785An unspecified error has occured, the watcher has been stopped. This might
782happen because the watcher could not be properly started because libev 786happen because the watcher could not be properly started because libev
783ran out of memory, a file descriptor was found to be closed or any other 787ran out of memory, a file descriptor was found to be closed or any other
… …
2046believe me. 2050believe me.
2047 2051
2048=back 2052=back
2049 2053
2050 2054
2055=head2 C<ev_async> - how to wake up another event loop
2056
2057In general, you cannot use an C<ev_loop> from multiple threads or other
2058asynchronous sources such as signal handlers (as opposed to multiple event
2059loops - those are of course safe to use in different threads).
2060
2061Sometimes, however, you need to wake up another event loop you do not
2062control, for example because it belongs to another thread. This is what
2063C<ev_async> watchers do: as long as the C<ev_async> watcher is active, you
2064can signal it by calling C<ev_async_send>, which is thread- and signal
2065safe.
2066
2067This functionality is very similar to C<ev_signal> watchers, as signals,
2068too, are asynchronous in nature, and signals, too, will be compressed
2069(i.e. the number of callback invocations may be less than the number of
2070C<ev_async_sent> calls).
2071
2072Unlike C<ev_signal> watchers, C<ev_async> works with any event loop, not
2073just the default loop.
2074
2075=head3 Queueing
2076
2077C<ev_async> does not support queueing of data in any way. The reason
2078is that the author does not know of a simple (or any) algorithm for a
2079multiple-writer-single-reader queue that works in all cases and doesn't
2080need elaborate support such as pthreads.
2081
2082That means that if you want to queue data, you have to provide your own
2083queue. But at least I can tell you would implement locking around your
2084queue:
2085
2086=over 4
2087
2088=item queueing from a signal handler context
2089
2090To implement race-free queueing, you simply add to the queue in the signal
2091handler but you block the signal handler in the watcher callback. Here is an example that does that for
2092some fictitiuous SIGUSR1 handler:
2093
2094 static ev_async mysig;
2095
2096 static void
2097 sigusr1_handler (void)
2098 {
2099 sometype data;
2100
2101 // no locking etc.
2102 queue_put (data);
2103 ev_async_send (DEFAULT_ &mysig);
2104 }
2105
2106 static void
2107 mysig_cb (EV_P_ ev_async *w, int revents)
2108 {
2109 sometype data;
2110 sigset_t block, prev;
2111
2112 sigemptyset (&block);
2113 sigaddset (&block, SIGUSR1);
2114 sigprocmask (SIG_BLOCK, &block, &prev);
2115
2116 while (queue_get (&data))
2117 process (data);
2118
2119 if (sigismember (&prev, SIGUSR1)
2120 sigprocmask (SIG_UNBLOCK, &block, 0);
2121 }
2122
2123(Note: pthreads in theory requires you to use C<pthread_setmask>
2124instead of C<sigprocmask> when you use threads, but libev doesn't do it
2125either...).
2126
2127=item queueing from a thread context
2128
2129The strategy for threads is different, as you cannot (easily) block
2130threads but you can easily preempt them, so to queue safely you need to
2131employ a traditional mutex lock, such as in this pthread example:
2132
2133 static ev_async mysig;
2134 static pthread_mutex_t mymutex = PTHREAD_MUTEX_INITIALIZER;
2135
2136 static void
2137 otherthread (void)
2138 {
2139 // only need to lock the actual queueing operation
2140 pthread_mutex_lock (&mymutex);
2141 queue_put (data);
2142 pthread_mutex_unlock (&mymutex);
2143
2144 ev_async_send (DEFAULT_ &mysig);
2145 }
2146
2147 static void
2148 mysig_cb (EV_P_ ev_async *w, int revents)
2149 {
2150 pthread_mutex_lock (&mymutex);
2151
2152 while (queue_get (&data))
2153 process (data);
2154
2155 pthread_mutex_unlock (&mymutex);
2156 }
2157
2158=back
2159
2160
2161=head3 Watcher-Specific Functions and Data Members
2162
2163=over 4
2164
2165=item ev_async_init (ev_async *, callback)
2166
2167Initialises and configures the async watcher - it has no parameters of any
2168kind. There is a C<ev_asynd_set> macro, but using it is utterly pointless,
2169believe me.
2170
2171=item ev_async_send (loop, ev_async *)
2172
2173Sends/signals/activates the given C<ev_async> watcher, that is, feeds
2174an C<EV_ASYNC> event on the watcher into the event loop. Unlike
2175C<ev_feed_event>, this call is safe to do in other threads, signal or
2176similar contexts (see the dicusssion of C<EV_ATOMIC_T> in the embedding
2177section below on what exactly this means).
2178
2179This call incurs the overhead of a syscall only once per loop iteration,
2180so while the overhead might be noticable, it doesn't apply to repeated
2181calls to C<ev_async_send>.
2182
2183=back
2184
2185
2051=head1 OTHER FUNCTIONS 2186=head1 OTHER FUNCTIONS
2052 2187
2053There are some other functions of possible interest. Described. Here. Now. 2188There are some other functions of possible interest. Described. Here. Now.
2054 2189
2055=over 4 2190=over 4
… …
2558 2693
2559If defined to be C<1>, libev will compile in support for the Linux inotify 2694If defined to be C<1>, libev will compile in support for the Linux inotify
2560interface to speed up C<ev_stat> watchers. Its actual availability will 2695interface to speed up C<ev_stat> watchers. Its actual availability will
2561be detected at runtime. 2696be detected at runtime.
2562 2697
2698=item EV_ATOMIC_T
2699
2700Libev requires an integer type (suitable for storing C<0> or C<1>) whose
2701access is atomic with respect to other threads or signal contexts. No such
2702type is easily found in the C language, so you can provide your own type
2703that you know is safe for your purposes. It is used both for signal handler "locking"
2704as well as for signal and thread safety in C<ev_async> watchers.
2705
2706In the absense of this define, libev will use C<sig_atomic_t volatile>
2707(from F<signal.h>), which is usually good enough on most platforms.
2708
2563=item EV_H 2709=item EV_H
2564 2710
2565The name of the F<ev.h> header file used to include it. The default if 2711The name of the F<ev.h> header file used to include it. The default if
2566undefined is C<"ev.h"> in F<event.h>, F<ev.c> and F<ev++.h>. This can be 2712undefined is C<"ev.h"> in F<event.h>, F<ev.c> and F<ev++.h>. This can be
2567used to virtually rename the F<ev.h> header file in case of conflicts. 2713used to virtually rename the F<ev.h> header file in case of conflicts.
… …
2632defined to be C<0>, then they are not. 2778defined to be C<0>, then they are not.
2633 2779
2634=item EV_FORK_ENABLE 2780=item EV_FORK_ENABLE
2635 2781
2636If undefined or defined to be C<1>, then fork watchers are supported. If 2782If undefined or defined to be C<1>, then fork watchers are supported. If
2783defined to be C<0>, then they are not.
2784
2785=item EV_ASYNC_ENABLE
2786
2787If undefined or defined to be C<1>, then async watchers are supported. If
2637defined to be C<0>, then they are not. 2788defined to be C<0>, then they are not.
2638 2789
2639=item EV_MINIMAL 2790=item EV_MINIMAL
2640 2791
2641If you need to shave off some kilobytes of code at the expense of some 2792If you need to shave off some kilobytes of code at the expense of some
… …
2762=item Changing timer/periodic watchers (by autorepeat or calling again): O(log skipped_other_timers) 2913=item Changing timer/periodic watchers (by autorepeat or calling again): O(log skipped_other_timers)
2763 2914
2764That means that changing a timer costs less than removing/adding them 2915That means that changing a timer costs less than removing/adding them
2765as only the relative motion in the event queue has to be paid for. 2916as only the relative motion in the event queue has to be paid for.
2766 2917
2767=item Starting io/check/prepare/idle/signal/child watchers: O(1) 2918=item Starting io/check/prepare/idle/signal/child/fork/async watchers: O(1)
2768 2919
2769These just add the watcher into an array or at the head of a list. 2920These just add the watcher into an array or at the head of a list.
2770 2921
2771=item Stopping check/prepare/idle watchers: O(1) 2922=item Stopping check/prepare/idle/fork/async watchers: O(1)
2772 2923
2773=item Stopping an io/signal/child watcher: O(number_of_watchers_for_this_(fd/signal/pid % EV_PID_HASHSIZE)) 2924=item Stopping an io/signal/child watcher: O(number_of_watchers_for_this_(fd/signal/pid % EV_PID_HASHSIZE))
2774 2925
2775These watchers are stored in lists then need to be walked to find the 2926These watchers are stored in lists then need to be walked to find the
2776correct watcher to remove. The lists are usually short (you don't usually 2927correct watcher to remove. The lists are usually short (you don't usually
… …
2792=item Priority handling: O(number_of_priorities) 2943=item Priority handling: O(number_of_priorities)
2793 2944
2794Priorities are implemented by allocating some space for each 2945Priorities are implemented by allocating some space for each
2795priority. When doing priority-based operations, libev usually has to 2946priority. When doing priority-based operations, libev usually has to
2796linearly search all the priorities, but starting/stopping and activating 2947linearly search all the priorities, but starting/stopping and activating
2797watchers becomes O(1) w.r.t. prioritiy handling. 2948watchers becomes O(1) w.r.t. priority handling.
2949
2950=item Sending an ev_async: O(1)
2951
2952=item Processing ev_async_send: O(number_of_async_watchers)
2953
2954=item Processing signals: O(max_signal_number)
2955
2956Sending involves a syscall I<iff> there were no other C<ev_async_send>
2957calls in the current loop iteration. Checking for async and signal events
2958involves iterating over all running async watchers or all signal numbers.
2798 2959
2799=back 2960=back
2800 2961
2801 2962
2802=head1 Win32 platform limitations and workarounds 2963=head1 Win32 platform limitations and workarounds

Diff Legend

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