[ViewVC] Diff of: cvs/libev/ev.pod

Comparing libev/ev.pod (file contents):
Revision 1.186 by root, Wed Sep 24 07:56:14 2008 UTC vs.
Revision 1.196 by root, Tue Oct 21 20:04:14 2008 UTC

…		…
685	C<EVUNLOOP_ONE>, which will make the innermost C<ev_loop> call return, or	685	C<EVUNLOOP_ONE>, which will make the innermost C<ev_loop> call return, or
686	C<EVUNLOOP_ALL>, which will make all nested C<ev_loop> calls return.	686	C<EVUNLOOP_ALL>, which will make all nested C<ev_loop> calls return.
687		687
688	This "unloop state" will be cleared when entering C<ev_loop> again.	688	This "unloop state" will be cleared when entering C<ev_loop> again.
689		689
		690	It is safe to call C<ev_unloop> from otuside any C<ev_loop> calls.
		691
690	=item ev_ref (loop)	692	=item ev_ref (loop)
691		693
692	=item ev_unref (loop)	694	=item ev_unref (loop)
693		695
694	Ref/unref can be used to add or remove a reference count on the event	696	Ref/unref can be used to add or remove a reference count on the event
…		…
967		969
968	ev_io_start (EV_DEFAULT_UC, &w);	970	ev_io_start (EV_DEFAULT_UC, &w);
969		971
970	=item C<ev_TYPE_stop> (loop , ev_TYPE watcher)	972	=item C<ev_TYPE_stop> (loop , ev_TYPE watcher)
971		973
972	Stops the given watcher again (if active) and clears the pending	974	Stops the given watcher if active, and clears the pending status (whether
		975	the watcher was active or not).
		976
973	status. It is possible that stopped watchers are pending (for example,	977	It is possible that stopped watchers are pending - for example,
974	non-repeating timers are being stopped when they become pending), but	978	non-repeating timers are being stopped when they become pending - but
975	C<ev_TYPE_stop> ensures that the watcher is neither active nor pending. If	979	calling C<ev_TYPE_stop> ensures that the watcher is neither active nor
976	you want to free or reuse the memory used by the watcher it is therefore a	980	pending. If you want to free or reuse the memory used by the watcher it is
977	good idea to always call its C<ev_TYPE_stop> function.	981	therefore a good idea to always call its C<ev_TYPE_stop> function.
978		982
979	=item bool ev_is_active (ev_TYPE *watcher)	983	=item bool ev_is_active (ev_TYPE *watcher)
980		984
981	Returns a true value iff the watcher is active (i.e. it has been started	985	Returns a true value iff the watcher is active (i.e. it has been started
982	and not yet been stopped). As long as a watcher is active you must not modify	986	and not yet been stopped). As long as a watcher is active you must not modify
…		…
1625		1629
1626	=back	1630	=back
1627		1631
1628	=head3 Examples	1632	=head3 Examples
1629		1633
1630	Example: Try to exit cleanly on SIGINT and SIGTERM.	1634	Example: Try to exit cleanly on SIGINT.
1631		1635
1632	static void	1636	static void
1633	sigint_cb (struct ev_loop loop, struct ev_signal w, int revents)	1637	sigint_cb (struct ev_loop loop, struct ev_signal w, int revents)
1634	{	1638	{
1635	ev_unloop (loop, EVUNLOOP_ALL);	1639	ev_unloop (loop, EVUNLOOP_ALL);
1636	}	1640	}
1637		1641
1638	struct ev_signal signal_watcher;	1642	struct ev_signal signal_watcher;
1639	ev_signal_init (&signal_watcher, sigint_cb, SIGINT);	1643	ev_signal_init (&signal_watcher, sigint_cb, SIGINT);
1640	ev_signal_start (loop, &sigint_cb);	1644	ev_signal_start (loop, &signal_watcher);
1641		1645
1642		1646
1643	=head2 C<ev_child> - watch out for process status changes	1647	=head2 C<ev_child> - watch out for process status changes
1644		1648
1645	Child watchers trigger when your process receives a SIGCHLD in response to	1649	Child watchers trigger when your process receives a SIGCHLD in response to
…		…
1792	to exchange stat structures with application programs compiled using the	1796	to exchange stat structures with application programs compiled using the
1793	default compilation environment.	1797	default compilation environment.
1794		1798
1795	=head3 Inotify and Kqueue	1799	=head3 Inotify and Kqueue
1796		1800
1797	When C<inotify (7)> support has been compiled into libev (generally only	1801	When C<inotify (7)> support has been compiled into libev (generally
		1802	only available with Linux 2.6.25 or above due to bugs in earlier
1798	available with Linux) and present at runtime, it will be used to speed up	1803	implementations) and present at runtime, it will be used to speed up
1799	change detection where possible. The inotify descriptor will be created lazily	1804	change detection where possible. The inotify descriptor will be created
1800	when the first C<ev_stat> watcher is being started.	1805	lazily when the first C<ev_stat> watcher is being started.
1801		1806
1802	Inotify presence does not change the semantics of C<ev_stat> watchers	1807	Inotify presence does not change the semantics of C<ev_stat> watchers
1803	except that changes might be detected earlier, and in some cases, to avoid	1808	except that changes might be detected earlier, and in some cases, to avoid
1804	making regular C<stat> calls. Even in the presence of inotify support	1809	making regular C<stat> calls. Even in the presence of inotify support
1805	there are many cases where libev has to resort to regular C<stat> polling,	1810	there are many cases where libev has to resort to regular C<stat> polling,
…		…
2242	So when you want to use this feature you will always have to be prepared	2247	So when you want to use this feature you will always have to be prepared
2243	that you cannot get an embeddable loop. The recommended way to get around	2248	that you cannot get an embeddable loop. The recommended way to get around
2244	this is to have a separate variables for your embeddable loop, try to	2249	this is to have a separate variables for your embeddable loop, try to
2245	create it, and if that fails, use the normal loop for everything.	2250	create it, and if that fails, use the normal loop for everything.
2246		2251
		2252	=head3 C<ev_embed> and fork
		2253
		2254	While the C<ev_embed> watcher is running, forks in the embedding loop will
		2255	automatically be applied to the embedded loop as well, so no special
		2256	fork handling is required in that case. When the watcher is not running,
		2257	however, it is still the task of the libev user to call C<ev_loop_fork ()>
		2258	as applicable.
		2259
2247	=head3 Watcher-Specific Functions and Data Members	2260	=head3 Watcher-Specific Functions and Data Members
2248		2261
2249	=over 4	2262	=over 4
2250		2263
2251	=item ev_embed_init (ev_embed , callback, struct ev_loop embedded_loop)	2264	=item ev_embed_init (ev_embed , callback, struct ev_loop embedded_loop)
…		…
2376	=over 4	2389	=over 4
2377		2390
2378	=item queueing from a signal handler context	2391	=item queueing from a signal handler context
2379		2392
2380	To implement race-free queueing, you simply add to the queue in the signal	2393	To implement race-free queueing, you simply add to the queue in the signal
2381	handler but you block the signal handler in the watcher callback. Here is an example that does that for	2394	handler but you block the signal handler in the watcher callback. Here is
2382	some fictitious SIGUSR1 handler:	2395	an example that does that for some fictitious SIGUSR1 handler:
2383		2396
2384	static ev_async mysig;	2397	static ev_async mysig;
2385		2398
2386	static void	2399	static void
2387	sigusr1_handler (void)	2400	sigusr1_handler (void)
…		…
2494	=over 4	2507	=over 4
2495		2508
2496	=item ev_once (loop, int fd, int events, ev_tstamp timeout, callback)	2509	=item ev_once (loop, int fd, int events, ev_tstamp timeout, callback)
2497		2510
2498	This function combines a simple timer and an I/O watcher, calls your	2511	This function combines a simple timer and an I/O watcher, calls your
2499	callback on whichever event happens first and automatically stop both	2512	callback on whichever event happens first and automatically stops both
2500	watchers. This is useful if you want to wait for a single event on an fd	2513	watchers. This is useful if you want to wait for a single event on an fd
2501	or timeout without having to allocate/configure/start/stop/free one or	2514	or timeout without having to allocate/configure/start/stop/free one or
2502	more watchers yourself.	2515	more watchers yourself.
2503		2516
2504	If C<fd> is less than 0, then no I/O watcher will be started and events	2517	If C<fd> is less than 0, then no I/O watcher will be started and the
2505	is being ignored. Otherwise, an C<ev_io> watcher for the given C<fd> and	2518	C<events> argument is being ignored. Otherwise, an C<ev_io> watcher for
2506	C<events> set will be created and started.	2519	the given C<fd> and C<events> set will be created and started.
2507		2520
2508	If C<timeout> is less than 0, then no timeout watcher will be	2521	If C<timeout> is less than 0, then no timeout watcher will be
2509	started. Otherwise an C<ev_timer> watcher with after = C<timeout> (and	2522	started. Otherwise an C<ev_timer> watcher with after = C<timeout> (and
2510	repeat = 0) will be started. While C<0> is a valid timeout, it is of	2523	repeat = 0) will be started. C<0> is a valid timeout.
2511	dubious value.
2512		2524
2513	The callback has the type C<void (cb)(int revents, void arg)> and gets	2525	The callback has the type C<void (cb)(int revents, void arg)> and gets
2514	passed an C<revents> set like normal event callbacks (a combination of	2526	passed an C<revents> set like normal event callbacks (a combination of
2515	C<EV_ERROR>, C<EV_READ>, C<EV_WRITE> or C<EV_TIMEOUT>) and the C<arg>	2527	C<EV_ERROR>, C<EV_READ>, C<EV_WRITE> or C<EV_TIMEOUT>) and the C<arg>
2516	value passed to C<ev_once>:	2528	value passed to C<ev_once>. Note that it is possible to receive I<both>
		2529	a timeout and an io event at the same time - you probably should give io
		2530	events precedence.
		2531
		2532	Example: wait up to ten seconds for data to appear on STDIN_FILENO.
2517		2533
2518	static void stdin_ready (int revents, void *arg)	2534	static void stdin_ready (int revents, void *arg)
2519	{	2535	{
		2536	if (revents & EV_READ)
		2537	/* stdin might have data for us, joy! */;
2520	if (revents & EV_TIMEOUT)	2538	else if (revents & EV_TIMEOUT)
2521	/* doh, nothing entered */;	2539	/* doh, nothing entered */;
2522	else if (revents & EV_READ)
2523	/* stdin might have data for us, joy! */;
2524	}	2540	}
2525		2541
2526	ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0);	2542	ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0);
2527		2543
2528	=item ev_feed_event (ev_loop , watcher , int revents)	2544	=item ev_feed_event (ev_loop , watcher , int revents)
…		…
3298	And a F<ev_cpp.C> implementation file that contains libev proper and is compiled:	3314	And a F<ev_cpp.C> implementation file that contains libev proper and is compiled:
3299		3315
3300	#include "ev_cpp.h"	3316	#include "ev_cpp.h"
3301	#include "ev.c"	3317	#include "ev.c"
3302		3318
		3319	=head1 INTERACTION WITH OTHER PROGRAMS OR LIBRARIES
3303		3320
3304	=head1 THREADS AND COROUTINES	3321	=head2 THREADS AND COROUTINES
3305		3322
3306	=head2 THREADS	3323	=head3 THREADS
3307		3324
3308	All libev functions are reentrant and thread-safe unless explicitly	3325	All libev functions are reentrant and thread-safe unless explicitly
3309	documented otherwise, but it uses no locking itself. This means that you	3326	documented otherwise, but libev implements no locking itself. This means
3310	can use as many loops as you want in parallel, as long as there are no	3327	that you can use as many loops as you want in parallel, as long as there
3311	concurrent calls into any libev function with the same loop parameter	3328	are no concurrent calls into any libev function with the same loop
3312	(C<ev_default_*> calls have an implicit default loop parameter, of	3329	parameter (C<ev_default_*> calls have an implicit default loop parameter,
3313	course): libev guarantees that different event loops share no data	3330	of course): libev guarantees that different event loops share no data
3314	structures that need any locking.	3331	structures that need any locking.
3315		3332
3316	Or to put it differently: calls with different loop parameters can be done	3333	Or to put it differently: calls with different loop parameters can be done
3317	concurrently from multiple threads, calls with the same loop parameter	3334	concurrently from multiple threads, calls with the same loop parameter
3318	must be done serially (but can be done from different threads, as long as	3335	must be done serially (but can be done from different threads, as long as
…		…
3358	default loop and triggering an C<ev_async> watcher from the default loop	3375	default loop and triggering an C<ev_async> watcher from the default loop
3359	watcher callback into the event loop interested in the signal.	3376	watcher callback into the event loop interested in the signal.
3360		3377
3361	=back	3378	=back
3362		3379
3363	=head2 COROUTINES	3380	=head3 COROUTINES
3364		3381
3365	Libev is much more accommodating to coroutines ("cooperative threads"):	3382	Libev is very accommodating to coroutines ("cooperative threads"):
3366	libev fully supports nesting calls to it's functions from different	3383	libev fully supports nesting calls to its functions from different
3367	coroutines (e.g. you can call C<ev_loop> on the same loop from two	3384	coroutines (e.g. you can call C<ev_loop> on the same loop from two
3368	different coroutines and switch freely between both coroutines running the	3385	different coroutines, and switch freely between both coroutines running the
3369	loop, as long as you don't confuse yourself). The only exception is that	3386	loop, as long as you don't confuse yourself). The only exception is that
3370	you must not do this from C<ev_periodic> reschedule callbacks.	3387	you must not do this from C<ev_periodic> reschedule callbacks.
3371		3388
3372	Care has been taken to ensure that libev does not keep local state inside	3389	Care has been taken to ensure that libev does not keep local state inside
3373	C<ev_loop>, and other calls do not usually allow coroutine switches.	3390	C<ev_loop>, and other calls do not usually allow for coroutine switches as
		3391	they do not clal any callbacks.
3374		3392
		3393	=head2 COMPILER WARNINGS
3375		3394
3376	=head1 COMPLEXITIES	3395	Depending on your compiler and compiler settings, you might get no or a
		3396	lot of warnings when compiling libev code. Some people are apparently
		3397	scared by this.
3377		3398
3378	In this section the complexities of (many of) the algorithms used inside	3399	However, these are unavoidable for many reasons. For one, each compiler
3379	libev will be explained. For complexity discussions about backends see the	3400	has different warnings, and each user has different tastes regarding
3380	documentation for C<ev_default_init>.	3401	warning options. "Warn-free" code therefore cannot be a goal except when
		3402	targeting a specific compiler and compiler-version.
3381		3403
3382	All of the following are about amortised time: If an array needs to be	3404	Another reason is that some compiler warnings require elaborate
3383	extended, libev needs to realloc and move the whole array, but this	3405	workarounds, or other changes to the code that make it less clear and less
3384	happens asymptotically never with higher number of elements, so O(1) might	3406	maintainable.
3385	mean it might do a lengthy realloc operation in rare cases, but on average
3386	it is much faster and asymptotically approaches constant time.
3387		3407
3388	=over 4	3408	And of course, some compiler warnings are just plain stupid, or simply
		3409	wrong (because they don't actually warn about the condition their message
		3410	seems to warn about). For example, certain older gcc versions had some
		3411	warnings that resulted an extreme number of false positives. These have
		3412	been fixed, but some people still insist on making code warn-free with
		3413	such buggy versions.
3389		3414
3390	=item Starting and stopping timer/periodic watchers: O(log skipped_other_timers)	3415	While libev is written to generate as few warnings as possible,
		3416	"warn-free" code is not a goal, and it is recommended not to build libev
		3417	with any compiler warnings enabled unless you are prepared to cope with
		3418	them (e.g. by ignoring them). Remember that warnings are just that:
		3419	warnings, not errors, or proof of bugs.
3391		3420
3392	This means that, when you have a watcher that triggers in one hour and
3393	there are 100 watchers that would trigger before that then inserting will
3394	have to skip roughly seven (C<ld 100>) of these watchers.
3395		3421
3396	=item Changing timer/periodic watchers (by autorepeat or calling again): O(log skipped_other_timers)	3422	=head2 VALGRIND
3397		3423
3398	That means that changing a timer costs less than removing/adding them	3424	Valgrind has a special section here because it is a popular tool that is
3399	as only the relative motion in the event queue has to be paid for.	3425	highly useful. Unfortunately, valgrind reports are very hard to interpret.
3400		3426
3401	=item Starting io/check/prepare/idle/signal/child/fork/async watchers: O(1)	3427	If you think you found a bug (memory leak, uninitialised data access etc.)
		3428	in libev, then check twice: If valgrind reports something like:
3402		3429
3403	These just add the watcher into an array or at the head of a list.	3430	==2274== definitely lost: 0 bytes in 0 blocks.
		3431	==2274== possibly lost: 0 bytes in 0 blocks.
		3432	==2274== still reachable: 256 bytes in 1 blocks.
3404		3433
3405	=item Stopping check/prepare/idle/fork/async watchers: O(1)	3434	Then there is no memory leak, just as memory accounted to global variables
		3435	is not a memleak - the memory is still being refernced, and didn't leak.
3406		3436
3407	=item Stopping an io/signal/child watcher: O(number_of_watchers_for_this_(fd/signal/pid % EV_PID_HASHSIZE))	3437	Similarly, under some circumstances, valgrind might report kernel bugs
		3438	as if it were a bug in libev (e.g. in realloc or in the poll backend,
		3439	although an acceptable workaround has been found here), or it might be
		3440	confused.
3408		3441
3409	These watchers are stored in lists then need to be walked to find the	3442	Keep in mind that valgrind is a very good tool, but only a tool. Don't
3410	correct watcher to remove. The lists are usually short (you don't usually	3443	make it into some kind of religion.
3411	have many watchers waiting for the same fd or signal).
3412		3444
3413	=item Finding the next timer in each loop iteration: O(1)	3445	If you are unsure about something, feel free to contact the mailing list
		3446	with the full valgrind report and an explanation on why you think this
		3447	is a bug in libev (best check the archives, too :). However, don't be
		3448	annoyed when you get a brisk "this is no bug" answer and take the chance
		3449	of learning how to interpret valgrind properly.
3414		3450
3415	By virtue of using a binary or 4-heap, the next timer is always found at a	3451	If you need, for some reason, empty reports from valgrind for your project
3416	fixed position in the storage array.	3452	I suggest using suppression lists.
3417		3453
3418	=item Each change on a file descriptor per loop iteration: O(number_of_watchers_for_this_fd)
3419		3454
3420	A change means an I/O watcher gets started or stopped, which requires	3455	=head1 PORTABILITY NOTES
3421	libev to recalculate its status (and possibly tell the kernel, depending
3422	on backend and whether C<ev_io_set> was used).
3423		3456
3424	=item Activating one watcher (putting it into the pending state): O(1)
3425
3426	=item Priority handling: O(number_of_priorities)
3427
3428	Priorities are implemented by allocating some space for each
3429	priority. When doing priority-based operations, libev usually has to
3430	linearly search all the priorities, but starting/stopping and activating
3431	watchers becomes O(1) with respect to priority handling.
3432
3433	=item Sending an ev_async: O(1)
3434
3435	=item Processing ev_async_send: O(number_of_async_watchers)
3436
3437	=item Processing signals: O(max_signal_number)
3438
3439	Sending involves a system call I<iff> there were no other C<ev_async_send>
3440	calls in the current loop iteration. Checking for async and signal events
3441	involves iterating over all running async watchers or all signal numbers.
3442
3443	=back
3444
3445
3446	=head1 WIN32 PLATFORM LIMITATIONS AND WORKAROUNDS	3457	=head2 WIN32 PLATFORM LIMITATIONS AND WORKAROUNDS
3447		3458
3448	Win32 doesn't support any of the standards (e.g. POSIX) that libev	3459	Win32 doesn't support any of the standards (e.g. POSIX) that libev
3449	requires, and its I/O model is fundamentally incompatible with the POSIX	3460	requires, and its I/O model is fundamentally incompatible with the POSIX
3450	model. Libev still offers limited functionality on this platform in	3461	model. Libev still offers limited functionality on this platform in
3451	the form of the C<EVBACKEND_SELECT> backend, and only supports socket	3462	the form of the C<EVBACKEND_SELECT> backend, and only supports socket
…		…
3538	wrap all I/O functions and provide your own fd management, but the cost of	3549	wrap all I/O functions and provide your own fd management, but the cost of
3539	calling select (O(n²)) will likely make this unworkable.	3550	calling select (O(n²)) will likely make this unworkable.
3540		3551
3541	=back	3552	=back
3542		3553
3543
3544	=head1 PORTABILITY REQUIREMENTS	3554	=head2 PORTABILITY REQUIREMENTS
3545		3555
3546	In addition to a working ISO-C implementation, libev relies on a few	3556	In addition to a working ISO-C implementation and of course the
3547	additional extensions:	3557	backend-specific APIs, libev relies on a few additional extensions:
3548		3558
3549	=over 4	3559	=over 4
3550		3560
3551	=item C<void ()(ev_watcher_type , int revents)> must have compatible	3561	=item C<void ()(ev_watcher_type , int revents)> must have compatible
3552	calling conventions regardless of C<ev_watcher_type *>.	3562	calling conventions regardless of C<ev_watcher_type *>.
…		…
3577	except the initial one, and run the default loop in the initial thread as	3587	except the initial one, and run the default loop in the initial thread as
3578	well.	3588	well.
3579		3589
3580	=item C<long> must be large enough for common memory allocation sizes	3590	=item C<long> must be large enough for common memory allocation sizes
3581		3591
3582	To improve portability and simplify using libev, libev uses C<long>	3592	To improve portability and simplify its API, libev uses C<long> internally
3583	internally instead of C<size_t> when allocating its data structures. On	3593	instead of C<size_t> when allocating its data structures. On non-POSIX
3584	non-POSIX systems (Microsoft...) this might be unexpectedly low, but	3594	systems (Microsoft...) this might be unexpectedly low, but is still at
3585	is still at least 31 bits everywhere, which is enough for hundreds of	3595	least 31 bits everywhere, which is enough for hundreds of millions of
3586	millions of watchers.	3596	watchers.
3587		3597
3588	=item C<double> must hold a time value in seconds with enough accuracy	3598	=item C<double> must hold a time value in seconds with enough accuracy
3589		3599
3590	The type C<double> is used to represent timestamps. It is required to	3600	The type C<double> is used to represent timestamps. It is required to
3591	have at least 51 bits of mantissa (and 9 bits of exponent), which is good	3601	have at least 51 bits of mantissa (and 9 bits of exponent), which is good
…		…
3595	=back	3605	=back
3596		3606
3597	If you know of other additional requirements drop me a note.	3607	If you know of other additional requirements drop me a note.
3598		3608
3599		3609
3600	=head1 COMPILER WARNINGS	3610	=head1 ALGORITHMIC COMPLEXITIES
3601		3611
3602	Depending on your compiler and compiler settings, you might get no or a	3612	In this section the complexities of (many of) the algorithms used inside
3603	lot of warnings when compiling libev code. Some people are apparently	3613	libev will be documented. For complexity discussions about backends see
3604	scared by this.	3614	the documentation for C<ev_default_init>.
3605		3615
3606	However, these are unavoidable for many reasons. For one, each compiler	3616	All of the following are about amortised time: If an array needs to be
3607	has different warnings, and each user has different tastes regarding	3617	extended, libev needs to realloc and move the whole array, but this
3608	warning options. "Warn-free" code therefore cannot be a goal except when	3618	happens asymptotically rarer with higher number of elements, so O(1) might
3609	targeting a specific compiler and compiler-version.	3619	mean that libev does a lengthy realloc operation in rare cases, but on
		3620	average it is much faster and asymptotically approaches constant time.
3610		3621
3611	Another reason is that some compiler warnings require elaborate	3622	=over 4
3612	workarounds, or other changes to the code that make it less clear and less
3613	maintainable.
3614		3623
3615	And of course, some compiler warnings are just plain stupid, or simply	3624	=item Starting and stopping timer/periodic watchers: O(log skipped_other_timers)
3616	wrong (because they don't actually warn about the condition their message
3617	seems to warn about).
3618		3625
3619	While libev is written to generate as few warnings as possible,	3626	This means that, when you have a watcher that triggers in one hour and
3620	"warn-free" code is not a goal, and it is recommended not to build libev	3627	there are 100 watchers that would trigger before that, then inserting will
3621	with any compiler warnings enabled unless you are prepared to cope with	3628	have to skip roughly seven (C<ld 100>) of these watchers.
3622	them (e.g. by ignoring them). Remember that warnings are just that:
3623	warnings, not errors, or proof of bugs.
3624		3629
		3630	=item Changing timer/periodic watchers (by autorepeat or calling again): O(log skipped_other_timers)
3625		3631
3626	=head1 VALGRIND	3632	That means that changing a timer costs less than removing/adding them,
		3633	as only the relative motion in the event queue has to be paid for.
3627		3634
3628	Valgrind has a special section here because it is a popular tool that is	3635	=item Starting io/check/prepare/idle/signal/child/fork/async watchers: O(1)
3629	highly useful, but valgrind reports are very hard to interpret.
3630		3636
3631	If you think you found a bug (memory leak, uninitialised data access etc.)	3637	These just add the watcher into an array or at the head of a list.
3632	in libev, then check twice: If valgrind reports something like:
3633		3638
3634	==2274== definitely lost: 0 bytes in 0 blocks.	3639	=item Stopping check/prepare/idle/fork/async watchers: O(1)
3635	==2274== possibly lost: 0 bytes in 0 blocks.
3636	==2274== still reachable: 256 bytes in 1 blocks.
3637		3640
3638	Then there is no memory leak. Similarly, under some circumstances,	3641	=item Stopping an io/signal/child watcher: O(number_of_watchers_for_this_(fd/signal/pid % EV_PID_HASHSIZE))
3639	valgrind might report kernel bugs as if it were a bug in libev, or it
3640	might be confused (it is a very good tool, but only a tool).
3641		3642
3642	If you are unsure about something, feel free to contact the mailing list	3643	These watchers are stored in lists, so they need to be walked to find the
3643	with the full valgrind report and an explanation on why you think this is	3644	correct watcher to remove. The lists are usually short (you don't usually
3644	a bug in libev. However, don't be annoyed when you get a brisk "this is	3645	have many watchers waiting for the same fd or signal: one is typical, two
3645	no bug" answer and take the chance of learning how to interpret valgrind	3646	is rare).
3646	properly.
3647		3647
3648	If you need, for some reason, empty reports from valgrind for your project	3648	=item Finding the next timer in each loop iteration: O(1)
3649	I suggest using suppression lists.	3649
		3650	By virtue of using a binary or 4-heap, the next timer is always found at a
		3651	fixed position in the storage array.
		3652
		3653	=item Each change on a file descriptor per loop iteration: O(number_of_watchers_for_this_fd)
		3654
		3655	A change means an I/O watcher gets started or stopped, which requires
		3656	libev to recalculate its status (and possibly tell the kernel, depending
		3657	on backend and whether C<ev_io_set> was used).
		3658
		3659	=item Activating one watcher (putting it into the pending state): O(1)
		3660
		3661	=item Priority handling: O(number_of_priorities)
		3662
		3663	Priorities are implemented by allocating some space for each
		3664	priority. When doing priority-based operations, libev usually has to
		3665	linearly search all the priorities, but starting/stopping and activating
		3666	watchers becomes O(1) with respect to priority handling.
		3667
		3668	=item Sending an ev_async: O(1)
		3669
		3670	=item Processing ev_async_send: O(number_of_async_watchers)
		3671
		3672	=item Processing signals: O(max_signal_number)
		3673
		3674	Sending involves a system call I<iff> there were no other C<ev_async_send>
		3675	calls in the current loop iteration. Checking for async and signal events
		3676	involves iterating over all running async watchers or all signal numbers.
		3677
		3678	=back
3650		3679
3651		3680
3652	=head1 AUTHOR	3681	=head1 AUTHOR
3653		3682
3654	Marc Lehmann <libev@schmorp.de>.	3683	Marc Lehmann <libev@schmorp.de>.

Diff Legend

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

Comparing libev/ev.pod (file contents): Revision 1.186 by root, Wed Sep 24 07:56:14 2008 UTC vs. Revision 1.196 by root, Tue Oct 21 20:04:14 2008 UTC

Diff Legend

Comparing libev/ev.pod (file contents):
Revision 1.186 by root, Wed Sep 24 07:56:14 2008 UTC vs.
Revision 1.196 by root, Tue Oct 21 20:04:14 2008 UTC