ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/Coro/Coro.pm

Comparing Coro/Coro.pm (file contents):
Revision 1.92 by root, Fri Dec 1 03:47:55 2006 UTC vs.
Revision 1.287 by root, Sat Feb 19 06:51:22 2011 UTC

1	=head1 NAME	1	=head1 NAME
2		2
3	Coro - coroutine process abstraction	3	Coro - the only real threads in perl
4		4
5	=head1 SYNOPSIS	5	=head1 SYNOPSIS
6		6
7	use Coro;	7	use Coro;
8		8
9	async {	9	async {
10	# some asynchronous thread of execution	10	# some asynchronous thread of execution
		11	print "2\n";
		12	cede; # yield back to main
		13	print "4\n";
11	};	14	};
12		15	print "1\n";
13	# alternatively create an async coroutine like this:	16	cede; # yield to coro
14		17	print "3\n";
15	sub some_func : Coro {	18	cede; # and again
16	# some more async code	19
17	}	20	# use locking
18		21	use Coro::Semaphore;
19	cede;	22	my $lock = new Coro::Semaphore;
		23	my $locked;
		24
		25	$lock->down;
		26	$locked = 1;
		27	$lock->up;
20		28
21	=head1 DESCRIPTION	29	=head1 DESCRIPTION
22		30
23	This module collection manages coroutines. Coroutines are similar to	31	For a tutorial-style introduction, please read the L<Coro::Intro>
24	threads but don't run in parallel.	32	manpage. This manpage mainly contains reference information.
25		33
		34	This module collection manages continuations in general, most often in
		35	the form of cooperative threads (also called coros, or simply "coro"
		36	in the documentation). They are similar to kernel threads but don't (in
		37	general) run in parallel at the same time even on SMP machines. The
		38	specific flavor of thread offered by this module also guarantees you that
		39	it will not switch between threads unless necessary, at easily-identified
		40	points in your program, so locking and parallel access are rarely an
		41	issue, making thread programming much safer and easier than using other
		42	thread models.
		43
		44	Unlike the so-called "Perl threads" (which are not actually real threads
		45	but only the windows process emulation (see section of same name for
		46	more details) ported to UNIX, and as such act as processes), Coro
		47	provides a full shared address space, which makes communication between
		48	threads very easy. And coro threads are fast, too: disabling the Windows
		49	process emulation code in your perl and using Coro can easily result in
		50	a two to four times speed increase for your programs. A parallel matrix
		51	multiplication benchmark (very communication-intensive) runs over 300
		52	times faster on a single core than perls pseudo-threads on a quad core
		53	using all four cores.
		54
		55	Coro achieves that by supporting multiple running interpreters that share
		56	data, which is especially useful to code pseudo-parallel processes and
		57	for event-based programming, such as multiple HTTP-GET requests running
		58	concurrently. See L<Coro::AnyEvent> to learn more on how to integrate Coro
		59	into an event-based environment.
		60
26	In this module, coroutines are defined as "callchain + lexical variables	61	In this module, a thread is defined as "callchain + lexical variables +
27	+ @_ + $_ + $@ + $^W + C stack), that is, a coroutine has it's own	62	some package variables + C stack), that is, a thread has its own callchain,
28	callchain, it's own set of lexicals and it's own set of perl's most	63	its own set of lexicals and its own set of perls most important global
29	important global variables.	64	variables (see L<Coro::State> for more configuration and background info).
		65
		66	See also the C<SEE ALSO> section at the end of this document - the Coro
		67	module family is quite large.
		68
		69	=head1 CORO THREAD LIFE CYCLE
		70
		71	During the long and exciting (or not) life of a coro thread, it goes
		72	through a number of states:
		73
		74	=over 4
		75
		76	=item 1. Creation
		77
		78	The first thing in the life of a coro thread is it's creation -
		79	obviously. The typical way to create a thread is to call the C<async
		80	BLOCK> function:
		81
		82	async {
		83	# thread code goes here
		84	};
		85
		86	You can also pass arguments, which are put in C<@_>:
		87
		88	async {
		89	print $_[1]; # prints 2
		90	} 1, 2, 3;
		91
		92	This creates a new coro thread and puts it into the ready queue, meaning
		93	it will run as soon as the CPU is free for it.
		94
		95	C<async> will return a coro object - you can store this for future
		96	reference or ignore it, the thread itself will keep a reference to it's
		97	thread object - threads are alive on their own.
		98
		99	Another way to create a thread is to call the C<new> constructor with a
		100	code-reference:
		101
		102	new Coro sub {
		103	# thread code goes here
		104	}, @optional_arguments;
		105
		106	This is quite similar to calling C<async>, but the important difference is
		107	that the new thread is not put into the ready queue, so the thread will
		108	not run until somebody puts it there. C<async> is, therefore, identical to
		109	this sequence:
		110
		111	my $coro = new Coro sub {
		112	# thread code goes here
		113	};
		114	$coro->ready;
		115	return $coro;
		116
		117	=item 2. Startup
		118
		119	When a new coro thread is created, only a copy of the code reference
		120	and the arguments are stored, no extra memory for stacks and so on is
		121	allocated, keeping the coro thread in a low-memory state.
		122
		123	Only when it actually starts executing will all the resources be finally
		124	allocated.
		125
		126	The optional arguments specified at coro creation are available in C<@_>,
		127	similar to function calls.
		128
		129	=item 3. Running / Blocking
		130
		131	A lot can happen after the coro thread has started running. Quite usually,
		132	it will not run to the end in one go (because you could use a function
		133	instead), but it will give up the CPU regularly because it waits for
		134	external events.
		135
		136	As long as a coro thread runs, it's coro object is available in the global
		137	variable C<$Coro::current>.
		138
		139	The low-level way to give up the CPU is to call the scheduler, which
		140	selects a new coro thread to run:
		141
		142	Coro::schedule;
		143
		144	Since running threads are not in the ready queue, calling the scheduler
		145	without doing anything else will block the coro thread forever - you need
		146	to arrange either for the coro to put woken up (readied) by some other
		147	event or some other thread, or you can put it into the ready queue before
		148	scheduling:
		149
		150	# this is exactly what Coro::cede does
		151	$Coro::current->ready;
		152	Coro::schedule;
		153
		154	All the higher-level synchronisation methods (Coro::Semaphore,
		155	Coro::rouse_*...) are actually implemented via C<< ->ready >> and C<<
		156	Coro::schedule >>.
		157
		158	While the coro thread is running it also might get assigned a C-level
		159	thread, or the C-level thread might be unassigned from it, as the Coro
		160	runtime wishes. A C-level thread needs to be assigned when your perl
		161	thread calls into some C-level function and that function in turn calls
		162	perl and perl then wants to switch coroutines. This happens most often
		163	when you run an event loop and block in the callback, or when perl
		164	itself calls some function such as C<AUTOLOAD> or methods via the C<tie>
		165	mechanism.
		166
		167	=item 4. Termination
		168
		169	Many threads actually terminate after some time. There are a number of
		170	ways to terminate a coro thread, the simplest is returning from the
		171	top-level code reference:
		172
		173	async {
		174	# after returning from here, the coro thread is terminated
		175	};
		176
		177	async {
		178	return if 0.5 < rand; # terminate a little earlier, maybe
		179	print "got a chance to print this\n";
		180	# or here
		181	};
		182
		183	Any values returned from the coroutine can be recovered using C<< ->join
		184	>>:
		185
		186	my $coro = async {
		187	"hello, world\n" # return a string
		188	};
		189
		190	my $hello_world = $coro->join;
		191
		192	print $hello_world;
		193
		194	Another way to terminate is to call C<< Coro::terminate >>, which at any
		195	subroutine call nesting level:
		196
		197	async {
		198	Coro::terminate "return value 1", "return value 2";
		199	};
		200
		201	And yet another way is to C<< ->cancel >> the coro thread from another
		202	thread:
		203
		204	my $coro = async {
		205	exit 1;
		206	};
		207
		208	$coro->cancel; # an also accept values for ->join to retrieve
		209
		210	Cancellation I<can> be dangerous - it's a bit like calling C<exit> without
		211	actually exiting, and might leave C libraries and XS modules in a weird
		212	state. Unlike other thread implementations, however, Coro is exceptionally
		213	safe with regards to cancellation, as perl will always be in a consistent
		214	state.
		215
		216	So, cancelling a thread that runs in an XS event loop might not be the
		217	best idea, but any other combination that deals with perl only (cancelling
		218	when a thread is in a C<tie> method or an C<AUTOLOAD> for example) is
		219	safe.
		220
		221	=item 5. Cleanup
		222
		223	Threads will allocate various resources. Most but not all will be returned
		224	when a thread terminates, during clean-up.
		225
		226	Cleanup is quite similar to throwing an uncaught exception: perl will
		227	work it's way up through all subroutine calls and blocks. On it's way, it
		228	will release all C<my> variables, undo all C<local>'s and free any other
		229	resources truly local to the thread.
		230
		231	So, a common way to free resources is to keep them referenced only by my
		232	variables:
		233
		234	async {
		235	my $big_cache = new Cache ...;
		236	};
		237
		238	If there are no other references, then the C<$big_cache> object will be
		239	freed when the thread terminates, regardless of how it does so.
		240
		241	What it does C<NOT> do is unlock any Coro::Semaphores or similar
		242	resources, but that's where the C<guard> methods come in handy:
		243
		244	my $sem = new Coro::Semaphore;
		245
		246	async {
		247	my $lock_guard = $sem->guard;
		248	# if we reutrn, or die or get cancelled, here,
		249	# then the semaphore will be "up"ed.
		250	};
		251
		252	The C<Guard::guard> function comes in handy for any custom cleanup you
		253	might want to do:
		254
		255	async {
		256	my $window = new Gtk2::Window "toplevel";
		257	# The window will not be cleaned up automatically, even when $window
		258	# gets freed, so use a guard to ensure it's destruction
		259	# in case of an error:
		260	my $window_guard = Guard::guard { $window->destroy };
		261
		262	# we are safe here
		263	};
		264
		265	Last not least, C<local> can often be handy, too, e.g. when temporarily
		266	replacing the coro thread description:
		267
		268	sub myfunction {
		269	local $Coro::current->{desc} = "inside myfunction(@_)";
		270
		271	# if we return or die here, the description will be restored
		272	}
		273
		274	=item 6. Viva La Zombie Muerte
		275
		276	Even after a thread has terminated and cleaned up it's resources, the coro
		277	object still is there and stores the return values of the thread. Only in
		278	this state will the coro object be "reference counted" in the normal perl
		279	sense: the thread code keeps a reference to it when it is active, but not
		280	after it has terminated.
		281
		282	The means the coro object gets freed automatically when the thread has
		283	terminated and cleaned up and there arenot other references.
		284
		285	If there are, the coro object will stay around, and you can call C<<
		286	->join >> as many times as you wish to retrieve the result values:
		287
		288	async {
		289	print "hi\n";
		290	1
		291	};
		292
		293	# run the async above, and free everything before returning
		294	# from Coro::cede:
		295	Coro::cede;
		296
		297	{
		298	my $coro = async {
		299	print "hi\n";
		300	1
		301	};
		302
		303	# run the async above, and clean up, but do not free the coro
		304	# object:
		305	Coro::cede;
		306
		307	# optionally retrieve the result values
		308	my @results = $coro->join;
		309
		310	# now $coro goes out of scope, and presumably gets freed
		311	};
		312
		313	=back
30		314
31	=cut	315	=cut
32		316
33	package Coro;	317	package Coro;
34		318
35	use strict;	319	use common::sense;
36	no warnings "uninitialized";	320
		321	use Carp ();
		322
		323	use Guard ();
37		324
38	use Coro::State;	325	use Coro::State;
39		326
40	use base qw(Coro::State Exporter);	327	use base qw(Coro::State Exporter);
41		328
42	our $idle; # idle handler	329	our $idle; # idle handler
43	our $main; # main coroutine	330	our $main; # main coro
44	our $current; # current coroutine	331	our $current; # current coro
45		332
46	our $VERSION = '3.0';	333	our $VERSION = 5.37;
47		334
48	our @EXPORT = qw(async cede schedule terminate current unblock_sub);	335	our @EXPORT = qw(async async_pool cede schedule terminate current unblock_sub rouse_cb rouse_wait);
49	our %EXPORT_TAGS = (	336	our %EXPORT_TAGS = (
50	prio => [qw(PRIO_MAX PRIO_HIGH PRIO_NORMAL PRIO_LOW PRIO_IDLE PRIO_MIN)],	337	prio => [qw(PRIO_MAX PRIO_HIGH PRIO_NORMAL PRIO_LOW PRIO_IDLE PRIO_MIN)],
51	);	338	);
52	our @EXPORT_OK = @{$EXPORT_TAGS{prio}};	339	our @EXPORT_OK = (@{$EXPORT_TAGS{prio}}, qw(nready));
53		340
54	{	341	=head1 GLOBAL VARIABLES
55	my @async;
56	my $init;
57
58	# this way of handling attributes simply is NOT scalable ;()
59	sub import {
60	no strict 'refs';
61
62	Coro->export_to_level(1, @_);
63
64	my $old = *{(caller)[0]."::MODIFY_CODE_ATTRIBUTES"}{CODE};
65	*{(caller)[0]."::MODIFY_CODE_ATTRIBUTES"} = sub {
66	my ($package, $ref) = (shift, shift);
67	my @attrs;
68	for (@_) {
69	if ($_ eq "Coro") {
70	push @async, $ref;
71	unless ($init++) {
72	eval q{
73	sub INIT {
74	&async(pop @async) while @async;
75	}
76	};
77	}
78	} else {
79	push @attrs, $_;
80	}
81	}
82	return $old ? $old->($package, $ref, @attrs) : @attrs;
83	};
84	}
85
86	}
87		342
88	=over 4	343	=over 4
89		344
90	=item $main	345	=item $Coro::main
91		346
92	This coroutine represents the main program.	347	This variable stores the Coro object that represents the main
		348	program. While you cna C<ready> it and do most other things you can do to
		349	coro, it is mainly useful to compare again C<$Coro::current>, to see
		350	whether you are running in the main program or not.
93		351
94	=cut	352	=cut
95		353
96	$main = new Coro;	354	# $main is now being initialised by Coro::State
97		355
98	=item $current (or as function: current)	356	=item $Coro::current
99		357
100	The current coroutine (the last coroutine switched to). The initial value	358	The Coro object representing the current coro (the last
		359	coro that the Coro scheduler switched to). The initial value is
101	is C<$main> (of course).	360	C<$Coro::main> (of course).
102		361
103	This variable is B<strictly> I<read-only>. It is provided for performance	362	This variable is B<strictly> I<read-only>. You can take copies of the
104	reasons. If performance is not essentiel you are encouraged to use the	363	value stored in it and use it as any other Coro object, but you must
105	C<Coro::current> function instead.	364	not otherwise modify the variable itself.
106		365
107	=cut	366	=cut
108		367
109	# maybe some other module used Coro::Specific before...
110	if ($current) {
111	$main->{specific} = $current->{specific};
112	}
113
114	$current = $main;
115
116	sub current() { $current }	368	sub current() { $current } # [DEPRECATED]
117		369
118	=item $idle	370	=item $Coro::idle
119		371
120	A callback that is called whenever the scheduler finds no ready coroutines	372	This variable is mainly useful to integrate Coro into event loops. It is
		373	usually better to rely on L<Coro::AnyEvent> or L<Coro::EV>, as this is
		374	pretty low-level functionality.
		375
		376	This variable stores a Coro object that is put into the ready queue when
		377	there are no other ready threads (without invoking any ready hooks).
		378
121	to run. The default implementation prints "FATAL: deadlock detected" and	379	The default implementation dies with "FATAL: deadlock detected.", followed
122	exits, because the program has no other way to continue.	380	by a thread listing, because the program has no other way to continue.
123		381
124	This hook is overwritten by modules such as C<Coro::Timer> and	382	This hook is overwritten by modules such as C<Coro::EV> and
125	C<Coro::Event> to wait on an external event that hopefully wake up a	383	C<Coro::AnyEvent> to wait on an external event that hopefully wakes up a
126	coroutine so the scheduler can run it.	384	coro so the scheduler can run it.
127		385
128	Please note that if your callback recursively invokes perl (e.g. for event	386	See L<Coro::EV> or L<Coro::AnyEvent> for examples of using this technique.
129	handlers), then it must be prepared to be called recursively.
130		387
131	=cut	388	=cut
132		389
133	$idle = sub {	390	# ||= because other modules could have provided their own by now
		391	$idle ||= new Coro sub {
		392	require Coro::Debug;
134	print STDERR "FATAL: deadlock detected\n";	393	die "FATAL: deadlock detected.\n"
135	exit (51);	394	. Coro::Debug::ps_listing ();
136	};	395	};
137		396
138	# this coroutine is necessary because a coroutine	397	# this coro is necessary because a coro
139	# cannot destroy itself.	398	# cannot destroy itself.
140	my @destroy;	399	our @destroy;
		400	our $manager;
		401
141	my $manager; $manager = new Coro sub {	402	$manager = new Coro sub {
142	while () {	403	while () {
143	# by overwriting the state object with the manager we destroy it	404	Coro::State::cancel shift @destroy
144	# while still being able to schedule this coroutine (in case it has
145	# been readied multiple times. this is harmless since the manager
146	# can be called as many times as neccessary and will always
147	# remove itself from the runqueue
148	while (@destroy) {	405	while @destroy;
149	my $coro = pop @destroy;
150	$coro->{status} ||= [];
151	$_->ready for @{delete $coro->{join} || []};
152		406
153	# the next line destroys the coro state, but keeps the
154	# coroutine itself intact (we basically make it a zombie
155	# coroutine that always runs the manager thread, so it's possible
156	# to transfer() to this coroutine).
157	$coro->_clone_state_from ($manager);
158	}
159	&schedule;	407	&schedule;
160	}	408	}
161	};	409	};
162		410	$manager->{desc} = "[coro manager]";
163	# static methods. not really.	411	$manager->prio (PRIO_MAX);
164		412
165	=back	413	=back
166		414
167	=head2 STATIC METHODS	415	=head1 SIMPLE CORO CREATION
168
169	Static methods are actually functions that operate on the current coroutine only.
170		416
171	=over 4	417	=over 4
172		418
173	=item async { ... } [@args...]	419	=item async { ... } [@args...]
174		420
175	Create a new asynchronous coroutine and return it's coroutine object	421	Create a new coro and return its Coro object (usually
176	(usually unused). When the sub returns the new coroutine is automatically	422	unused). The coro will be put into the ready queue, so
		423	it will start running automatically on the next scheduler run.
		424
		425	The first argument is a codeblock/closure that should be executed in the
		426	coro. When it returns argument returns the coro is automatically
177	terminated.	427	terminated.
178		428
179	Calling C<exit> in a coroutine will not work correctly, so do not do that.	429	The remaining arguments are passed as arguments to the closure.
180		430
181	When the coroutine dies, the program will exit, just as in the main	431	See the C<Coro::State::new> constructor for info about the coro
182	program.	432	environment in which coro are executed.
183		433
		434	Calling C<exit> in a coro will do the same as calling exit outside
		435	the coro. Likewise, when the coro dies, the program will exit,
		436	just as it would in the main program.
		437
		438	If you do not want that, you can provide a default C<die> handler, or
		439	simply avoid dieing (by use of C<eval>).
		440
184	# create a new coroutine that just prints its arguments	441	Example: Create a new coro that just prints its arguments.
		442
185	async {	443	async {
186	print "@_\n";	444	print "@_\n";
187	} 1,2,3,4;	445	} 1,2,3,4;
188		446
189	=cut	447	=item async_pool { ... } [@args...]
190		448
191	sub async(&@) {	449	Similar to C<async>, but uses a coro pool, so you should not call
192	my $pid = new Coro @_;	450	terminate or join on it (although you are allowed to), and you get a
193	$pid->ready;	451	coro that might have executed other code already (which can be good
194	$pid	452	or bad :).
		453
		454	On the plus side, this function is about twice as fast as creating (and
		455	destroying) a completely new coro, so if you need a lot of generic
		456	coros in quick successsion, use C<async_pool>, not C<async>.
		457
		458	The code block is executed in an C<eval> context and a warning will be
		459	issued in case of an exception instead of terminating the program, as
		460	C<async> does. As the coro is being reused, stuff like C<on_destroy>
		461	will not work in the expected way, unless you call terminate or cancel,
		462	which somehow defeats the purpose of pooling (but is fine in the
		463	exceptional case).
		464
		465	The priority will be reset to C<0> after each run, tracing will be
		466	disabled, the description will be reset and the default output filehandle
		467	gets restored, so you can change all these. Otherwise the coro will
		468	be re-used "as-is": most notably if you change other per-coro global
		469	stuff such as C<$/> you I<must needs> revert that change, which is most
		470	simply done by using local as in: C<< local $/ >>.
		471
		472	The idle pool size is limited to C<8> idle coros (this can be
		473	adjusted by changing $Coro::POOL_SIZE), but there can be as many non-idle
		474	coros as required.
		475
		476	If you are concerned about pooled coros growing a lot because a
		477	single C<async_pool> used a lot of stackspace you can e.g. C<async_pool
		478	{ terminate }> once per second or so to slowly replenish the pool. In
		479	addition to that, when the stacks used by a handler grows larger than 32kb
		480	(adjustable via $Coro::POOL_RSS) it will also be destroyed.
		481
		482	=cut
		483
		484	our $POOL_SIZE = 8;
		485	our $POOL_RSS = 32 * 1024;
		486	our @async_pool;
		487
		488	sub pool_handler {
		489	while () {
		490	eval {
		491	&{&_pool_handler} while 1;
		492	};
		493
		494	warn $@ if $@;
		495	}
195	}	496	}
196		497
		498	=back
		499
		500	=head1 STATIC METHODS
		501
		502	Static methods are actually functions that implicitly operate on the
		503	current coro.
		504
		505	=over 4
		506
197	=item schedule	507	=item schedule
198		508
199	Calls the scheduler. Please note that the current coroutine will not be put	509	Calls the scheduler. The scheduler will find the next coro that is
		510	to be run from the ready queue and switches to it. The next coro
		511	to be run is simply the one with the highest priority that is longest
		512	in its ready queue. If there is no coro ready, it will call the
		513	C<$Coro::idle> hook.
		514
		515	Please note that the current coro will I<not> be put into the ready
200	into the ready queue, so calling this function usually means you will	516	queue, so calling this function usually means you will never be called
201	never be called again unless something else (e.g. an event handler) calls	517	again unless something else (e.g. an event handler) calls C<< ->ready >>,
202	ready.	518	thus waking you up.
203		519
204	The canonical way to wait on external events is this:	520	This makes C<schedule> I<the> generic method to use to block the current
		521	coro and wait for events: first you remember the current coro in
		522	a variable, then arrange for some callback of yours to call C<< ->ready
		523	>> on that once some event happens, and last you call C<schedule> to put
		524	yourself to sleep. Note that a lot of things can wake your coro up,
		525	so you need to check whether the event indeed happened, e.g. by storing the
		526	status in a variable.
205		527
		528	See B<HOW TO WAIT FOR A CALLBACK>, below, for some ways to wait for callbacks.
		529
		530	=item cede
		531
		532	"Cede" to other coros. This function puts the current coro into
		533	the ready queue and calls C<schedule>, which has the effect of giving
		534	up the current "timeslice" to other coros of the same or higher
		535	priority. Once your coro gets its turn again it will automatically be
		536	resumed.
		537
		538	This function is often called C<yield> in other languages.
		539
		540	=item Coro::cede_notself
		541
		542	Works like cede, but is not exported by default and will cede to I<any>
		543	coro, regardless of priority. This is useful sometimes to ensure
		544	progress is made.
		545
		546	=item terminate [arg...]
		547
		548	Terminates the current coro with the given status values (see L<cancel>).
		549
		550	=item Coro::on_enter BLOCK, Coro::on_leave BLOCK
		551
		552	These function install enter and leave winders in the current scope. The
		553	enter block will be executed when on_enter is called and whenever the
		554	current coro is re-entered by the scheduler, while the leave block is
		555	executed whenever the current coro is blocked by the scheduler, and
		556	also when the containing scope is exited (by whatever means, be it exit,
		557	die, last etc.).
		558
		559	I<Neither invoking the scheduler, nor exceptions, are allowed within those
		560	BLOCKs>. That means: do not even think about calling C<die> without an
		561	eval, and do not even think of entering the scheduler in any way.
		562
		563	Since both BLOCKs are tied to the current scope, they will automatically
		564	be removed when the current scope exits.
		565
		566	These functions implement the same concept as C<dynamic-wind> in scheme
		567	does, and are useful when you want to localise some resource to a specific
		568	coro.
		569
		570	They slow down thread switching considerably for coros that use them
		571	(about 40% for a BLOCK with a single assignment, so thread switching is
		572	still reasonably fast if the handlers are fast).
		573
		574	These functions are best understood by an example: The following function
		575	will change the current timezone to "Antarctica/South_Pole", which
		576	requires a call to C<tzset>, but by using C<on_enter> and C<on_leave>,
		577	which remember/change the current timezone and restore the previous
		578	value, respectively, the timezone is only changed for the coro that
		579	installed those handlers.
		580
		581	use POSIX qw(tzset);
		582
		583	async {
		584	my $old_tz; # store outside TZ value here
		585
		586	Coro::on_enter {
		587	$old_tz = $ENV{TZ}; # remember the old value
		588
		589	$ENV{TZ} = "Antarctica/South_Pole";
		590	tzset; # enable new value
		591	};
		592
		593	Coro::on_leave {
		594	$ENV{TZ} = $old_tz;
		595	tzset; # restore old value
		596	};
		597
		598	# at this place, the timezone is Antarctica/South_Pole,
		599	# without disturbing the TZ of any other coro.
		600	};
		601
		602	This can be used to localise about any resource (locale, uid, current
		603	working directory etc.) to a block, despite the existance of other
		604	coros.
		605
		606	Another interesting example implements time-sliced multitasking using
		607	interval timers (this could obviously be optimised, but does the job):
		608
		609	# "timeslice" the given block
		610	sub timeslice(&) {
		611	use Time::HiRes ();
		612
		613	Coro::on_enter {
		614	# on entering the thread, we set an VTALRM handler to cede
		615	$SIG{VTALRM} = sub { cede };
		616	# and then start the interval timer
		617	Time::HiRes::setitimer &Time::HiRes::ITIMER_VIRTUAL, 0.01, 0.01;
		618	};
		619	Coro::on_leave {
		620	# on leaving the thread, we stop the interval timer again
		621	Time::HiRes::setitimer &Time::HiRes::ITIMER_VIRTUAL, 0, 0;
		622	};
		623
		624	&{+shift};
		625	}
		626
		627	# use like this:
		628	timeslice {
		629	# The following is an endless loop that would normally
		630	# monopolise the process. Since it runs in a timesliced
		631	# environment, it will regularly cede to other threads.
		632	while () { }
		633	};
		634
		635
		636	=item killall
		637
		638	Kills/terminates/cancels all coros except the currently running one.
		639
		640	Note that while this will try to free some of the main interpreter
		641	resources if the calling coro isn't the main coro, but one
		642	cannot free all of them, so if a coro that is not the main coro
		643	calls this function, there will be some one-time resource leak.
		644
		645	=cut
		646
		647	sub killall {
		648	for (Coro::State::list) {
		649	$_->cancel
		650	if $_ != $current && UNIVERSAL::isa $_, "Coro";
206	{	651	}
207	# remember current coroutine	652	}
		653
		654	=back
		655
		656	=head1 CORO OBJECT METHODS
		657
		658	These are the methods you can call on coro objects (or to create
		659	them).
		660
		661	=over 4
		662
		663	=item new Coro \&sub [, @args...]
		664
		665	Create a new coro and return it. When the sub returns, the coro
		666	automatically terminates as if C<terminate> with the returned values were
		667	called. To make the coro run you must first put it into the ready
		668	queue by calling the ready method.
		669
		670	See C<async> and C<Coro::State::new> for additional info about the
		671	coro environment.
		672
		673	=cut
		674
		675	sub _coro_run {
		676	terminate &{+shift};
		677	}
		678
		679	=item $success = $coro->ready
		680
		681	Put the given coro into the end of its ready queue (there is one
		682	queue for each priority) and return true. If the coro is already in
		683	the ready queue, do nothing and return false.
		684
		685	This ensures that the scheduler will resume this coro automatically
		686	once all the coro of higher priority and all coro of the same
		687	priority that were put into the ready queue earlier have been resumed.
		688
		689	=item $coro->suspend
		690
		691	Suspends the specified coro. A suspended coro works just like any other
		692	coro, except that the scheduler will not select a suspended coro for
		693	execution.
		694
		695	Suspending a coro can be useful when you want to keep the coro from
		696	running, but you don't want to destroy it, or when you want to temporarily
		697	freeze a coro (e.g. for debugging) to resume it later.
		698
		699	A scenario for the former would be to suspend all (other) coros after a
		700	fork and keep them alive, so their destructors aren't called, but new
		701	coros can be created.
		702
		703	=item $coro->resume
		704
		705	If the specified coro was suspended, it will be resumed. Note that when
		706	the coro was in the ready queue when it was suspended, it might have been
		707	unreadied by the scheduler, so an activation might have been lost.
		708
		709	To avoid this, it is best to put a suspended coro into the ready queue
		710	unconditionally, as every synchronisation mechanism must protect itself
		711	against spurious wakeups, and the one in the Coro family certainly do
		712	that.
		713
		714	=item $is_ready = $coro->is_ready
		715
		716	Returns true iff the Coro object is in the ready queue. Unless the Coro
		717	object gets destroyed, it will eventually be scheduled by the scheduler.
		718
		719	=item $is_running = $coro->is_running
		720
		721	Returns true iff the Coro object is currently running. Only one Coro object
		722	can ever be in the running state (but it currently is possible to have
		723	multiple running Coro::States).
		724
		725	=item $is_suspended = $coro->is_suspended
		726
		727	Returns true iff this Coro object has been suspended. Suspended Coros will
		728	not ever be scheduled.
		729
		730	=item $coro->cancel (arg...)
		731
		732	Terminates the given Coro and makes it return the given arguments as
		733	status (default: the empty list). Never returns if the Coro is the
		734	current Coro.
		735
		736	=cut
		737
		738	sub cancel {
		739	my $self = shift;
		740
		741	if ($current == $self) {
		742	terminate @_;
		743	} else {
		744	$self->{_status} = [@_];
		745	Coro::State::cancel $self;
		746	}
		747	}
		748
		749	=item $coro->schedule_to
		750
		751	Puts the current coro to sleep (like C<Coro::schedule>), but instead
		752	of continuing with the next coro from the ready queue, always switch to
		753	the given coro object (regardless of priority etc.). The readyness
		754	state of that coro isn't changed.
		755
		756	This is an advanced method for special cases - I'd love to hear about any
		757	uses for this one.
		758
		759	=item $coro->cede_to
		760
		761	Like C<schedule_to>, but puts the current coro into the ready
		762	queue. This has the effect of temporarily switching to the given
		763	coro, and continuing some time later.
		764
		765	This is an advanced method for special cases - I'd love to hear about any
		766	uses for this one.
		767
		768	=item $coro->throw ([$scalar])
		769
		770	If C<$throw> is specified and defined, it will be thrown as an exception
		771	inside the coro at the next convenient point in time. Otherwise
		772	clears the exception object.
		773
		774	Coro will check for the exception each time a schedule-like-function
		775	returns, i.e. after each C<schedule>, C<cede>, C<< Coro::Semaphore->down
		776	>>, C<< Coro::Handle->readable >> and so on. Most of these functions
		777	detect this case and return early in case an exception is pending.
		778
		779	The exception object will be thrown "as is" with the specified scalar in
		780	C<$@>, i.e. if it is a string, no line number or newline will be appended
		781	(unlike with C<die>).
		782
		783	This can be used as a softer means than C<cancel> to ask a coro to
		784	end itself, although there is no guarantee that the exception will lead to
		785	termination, and if the exception isn't caught it might well end the whole
		786	program.
		787
		788	You might also think of C<throw> as being the moral equivalent of
		789	C<kill>ing a coro with a signal (in this case, a scalar).
		790
		791	=item $coro->join
		792
		793	Wait until the coro terminates and return any values given to the
		794	C<terminate> or C<cancel> functions. C<join> can be called concurrently
		795	from multiple coro, and all will be resumed and given the status
		796	return once the C<$coro> terminates.
		797
		798	=cut
		799
		800	sub join {
		801	my $self = shift;
		802
		803	unless ($self->{_status}) {
208	my $current = $Coro::current;	804	my $current = $current;
209		805
210	# register a hypothetical event handler	806	push @{$self->{_on_destroy}}, sub {
211	on_event_invoke sub {
212	# wake up sleeping coroutine
213	$current->ready;	807	$current->ready;
214	undef $current;	808	undef $current;
215	};	809	};
216		810
217	# call schedule until event occured.
218	# in case we are woken up for other reasons
219	# (current still defined), loop.
220	Coro::schedule while $current;	811	&schedule while $current;
221	}	812	}
222		813
223	=item cede	814	wantarray ? @{$self->{_status}} : $self->{_status}[0];
224
225	"Cede" to other coroutines. This function puts the current coroutine into the
226	ready queue and calls C<schedule>, which has the effect of giving up the
227	current "timeslice" to other coroutines of the same or higher priority.
228
229	=item terminate [arg...]
230
231	Terminates the current coroutine with the given status values (see L<cancel>).
232
233	=cut
234
235	sub terminate {
236	$current->cancel (@_);
237	}	815	}
238		816
239	=back	817	=item $coro->on_destroy (\&cb)
240		818
241	# dynamic methods	819	Registers a callback that is called when this coro thread gets destroyed,
		820	but before it is joined. The callback gets passed the terminate arguments,
		821	if any, and I<must not> die, under any circumstances.
242		822
243	=head2 COROUTINE METHODS	823	There can be any number of C<on_destroy> callbacks per coro.
244		824
245	These are the methods you can call on coroutine objects.
246
247	=over 4
248
249	=item new Coro \&sub [, @args...]
250
251	Create a new coroutine and return it. When the sub returns the coroutine
252	automatically terminates as if C<terminate> with the returned values were
253	called. To make the coroutine run you must first put it into the ready queue
254	by calling the ready method.
255
256	Calling C<exit> in a coroutine will not work correctly, so do not do that.
257
258	=cut	825	=cut
259		826
260	sub _new_coro {	827	sub on_destroy {
261	terminate &{+shift};	828	my ($self, $cb) = @_;
		829
		830	push @{ $self->{_on_destroy} }, $cb;
262	}	831	}
263		832
264	sub new {
265	my $class = shift;
266
267	$class->SUPER::new (\&_new_coro, @_)
268	}
269
270	=item $success = $coroutine->ready
271
272	Put the given coroutine into the ready queue (according to it's priority)
273	and return true. If the coroutine is already in the ready queue, do nothing
274	and return false.
275
276	=item $is_ready = $coroutine->is_ready
277
278	Return wether the coroutine is currently the ready queue or not,
279
280	=item $coroutine->cancel (arg...)
281
282	Terminates the given coroutine and makes it return the given arguments as
283	status (default: the empty list).
284
285	=cut
286
287	sub cancel {
288	my $self = shift;
289	$self->{status} = [@_];
290	push @destroy, $self;
291	$manager->ready;
292	&schedule if $current == $self;
293	}
294
295	=item $coroutine->join
296
297	Wait until the coroutine terminates and return any values given to the
298	C<terminate> or C<cancel> functions. C<join> can be called multiple times
299	from multiple coroutine.
300
301	=cut
302
303	sub join {
304	my $self = shift;
305	unless ($self->{status}) {
306	push @{$self->{join}}, $current;
307	&schedule;
308	}
309	wantarray ? @{$self->{status}} : $self->{status}[0];
310	}
311
312	=item $oldprio = $coroutine->prio ($newprio)	833	=item $oldprio = $coro->prio ($newprio)
313		834
314	Sets (or gets, if the argument is missing) the priority of the	835	Sets (or gets, if the argument is missing) the priority of the
315	coroutine. Higher priority coroutines get run before lower priority	836	coro thread. Higher priority coro get run before lower priority
316	coroutines. Priorities are small signed integers (currently -4 .. +3),	837	coros. Priorities are small signed integers (currently -4 .. +3),
317	that you can refer to using PRIO_xxx constants (use the import tag :prio	838	that you can refer to using PRIO_xxx constants (use the import tag :prio
318	to get then):	839	to get then):
319		840
320	PRIO_MAX > PRIO_HIGH > PRIO_NORMAL > PRIO_LOW > PRIO_IDLE > PRIO_MIN	841	PRIO_MAX > PRIO_HIGH > PRIO_NORMAL > PRIO_LOW > PRIO_IDLE > PRIO_MIN
321	3 > 1 > 0 > -1 > -3 > -4	842	3 > 1 > 0 > -1 > -3 > -4
322		843
323	# set priority to HIGH	844	# set priority to HIGH
324	current->prio(PRIO_HIGH);	845	current->prio (PRIO_HIGH);
325		846
326	The idle coroutine ($Coro::idle) always has a lower priority than any	847	The idle coro thread ($Coro::idle) always has a lower priority than any
327	existing coroutine.	848	existing coro.
328		849
329	Changing the priority of the current coroutine will take effect immediately,	850	Changing the priority of the current coro will take effect immediately,
330	but changing the priority of coroutines in the ready queue (but not	851	but changing the priority of a coro in the ready queue (but not running)
331	running) will only take effect after the next schedule (of that	852	will only take effect after the next schedule (of that coro). This is a
332	coroutine). This is a bug that will be fixed in some future version.	853	bug that will be fixed in some future version.
333		854
334	=item $newprio = $coroutine->nice ($change)	855	=item $newprio = $coro->nice ($change)
335		856
336	Similar to C<prio>, but subtract the given value from the priority (i.e.	857	Similar to C<prio>, but subtract the given value from the priority (i.e.
337	higher values mean lower priority, just as in unix).	858	higher values mean lower priority, just as in UNIX's nice command).
338		859
339	=item $olddesc = $coroutine->desc ($newdesc)	860	=item $olddesc = $coro->desc ($newdesc)
340		861
341	Sets (or gets in case the argument is missing) the description for this	862	Sets (or gets in case the argument is missing) the description for this
342	coroutine. This is just a free-form string you can associate with a coroutine.	863	coro thread. This is just a free-form string you can associate with a
		864	coro.
		865
		866	This method simply sets the C<< $coro->{desc} >> member to the given
		867	string. You can modify this member directly if you wish, and in fact, this
		868	is often preferred to indicate major processing states that cna then be
		869	seen for example in a L<Coro::Debug> session:
		870
		871	sub my_long_function {
		872	local $Coro::current->{desc} = "now in my_long_function";
		873	...
		874	$Coro::current->{desc} = "my_long_function: phase 1";
		875	...
		876	$Coro::current->{desc} = "my_long_function: phase 2";
		877	...
		878	}
343		879
344	=cut	880	=cut
345		881
346	sub desc {	882	sub desc {
347	my $old = $_[0]{desc};	883	my $old = $_[0]{desc};
348	$_[0]{desc} = $_[1] if @_ > 1;	884	$_[0]{desc} = $_[1] if @_ > 1;
349	$old;	885	$old;
350	}	886	}
351		887
		888	sub transfer {
		889	require Carp;
		890	Carp::croak ("You must not call ->transfer on Coro objects. Use Coro::State objects or the ->schedule_to method. Caught");
		891	}
		892
352	=back	893	=back
353		894
354	=head2 UTILITY FUNCTIONS	895	=head1 GLOBAL FUNCTIONS
355		896
356	=over 4	897	=over 4
357		898
		899	=item Coro::nready
		900
		901	Returns the number of coro that are currently in the ready state,
		902	i.e. that can be switched to by calling C<schedule> directory or
		903	indirectly. The value C<0> means that the only runnable coro is the
		904	currently running one, so C<cede> would have no effect, and C<schedule>
		905	would cause a deadlock unless there is an idle handler that wakes up some
		906	coro.
		907
		908	=item my $guard = Coro::guard { ... }
		909
		910	This function still exists, but is deprecated. Please use the
		911	C<Guard::guard> function instead.
		912
		913	=cut
		914
		915	BEGIN { *guard = \&Guard::guard }
		916
358	=item unblock_sub { ... }	917	=item unblock_sub { ... }
359		918
360	This utility function takes a BLOCK or code reference and "unblocks" it,	919	This utility function takes a BLOCK or code reference and "unblocks" it,
361	returning the new coderef. This means that the new coderef will return	920	returning a new coderef. Unblocking means that calling the new coderef
362	immediately without blocking, returning nothing, while the original code	921	will return immediately without blocking, returning nothing, while the
363	ref will be called (with parameters) from within its own coroutine.	922	original code ref will be called (with parameters) from within another
		923	coro.
364		924
365	The reason this fucntion exists is that many event libraries (such as the	925	The reason this function exists is that many event libraries (such as
366	venerable L<Event|Event> module) are not coroutine-safe (a weaker form	926	the venerable L<Event|Event> module) are not thread-safe (a weaker form
367	of thread-safety). This means you must not block within event callbacks,	927	of reentrancy). This means you must not block within event callbacks,
368	otherwise you might suffer from crashes or worse.	928	otherwise you might suffer from crashes or worse. The only event library
		929	currently known that is safe to use without C<unblock_sub> is L<EV> (but
		930	you might still run into deadlocks if all event loops are blocked).
		931
		932	Coro will try to catch you when you block in the event loop
		933	("FATAL:$Coro::IDLE blocked itself"), but this is just best effort and
		934	only works when you do not run your own event loop.
369		935
370	This function allows your callbacks to block by executing them in another	936	This function allows your callbacks to block by executing them in another
371	coroutine where it is safe to block. One example where blocking is handy	937	coro where it is safe to block. One example where blocking is handy
372	is when you use the L<Coro::AIO|Coro::AIO> functions to save results to	938	is when you use the L<Coro::AIO|Coro::AIO> functions to save results to
373	disk.	939	disk, for example.
374		940
375	In short: simply use C<unblock_sub { ... }> instead of C<sub { ... }> when	941	In short: simply use C<unblock_sub { ... }> instead of C<sub { ... }> when
376	creating event callbacks that want to block.	942	creating event callbacks that want to block.
377		943
378	=cut	944	If your handler does not plan to block (e.g. simply sends a message to
		945	another coro, or puts some other coro into the ready queue), there is
		946	no reason to use C<unblock_sub>.
379		947
380	our @unblock_pool;	948	Note that you also need to use C<unblock_sub> for any other callbacks that
		949	are indirectly executed by any C-based event loop. For example, when you
		950	use a module that uses L<AnyEvent> (and you use L<Coro::AnyEvent>) and it
		951	provides callbacks that are the result of some event callback, then you
		952	must not block either, or use C<unblock_sub>.
		953
		954	=cut
		955
381	our @unblock_queue;	956	our @unblock_queue;
382	our $UNBLOCK_POOL_SIZE = 2;
383		957
384	sub unblock_handler_ {	958	# we create a special coro because we want to cede,
385	while () {	959	# to reduce pressure on the coro pool (because most callbacks
386	my ($cb, @arg) = @{ delete $Coro::current->{arg} };	960	# return immediately and can be reused) and because we cannot cede
387	$cb->(@arg);	961	# inside an event callback.
388
389	last if @unblock_pool >= $UNBLOCK_POOL_SIZE;
390	push @unblock_pool, $Coro::current;
391	schedule;
392	}
393	}
394
395	our $unblock_scheduler = async {	962	our $unblock_scheduler = new Coro sub {
396	while () {	963	while () {
397	while (my $cb = pop @unblock_queue) {	964	while (my $cb = pop @unblock_queue) {
398	my $handler = (pop @unblock_pool or new Coro \&unblock_handler_);	965	&async_pool (@$cb);
399	$handler->{arg} = $cb;	966
400	$handler->ready;	967	# for short-lived callbacks, this reduces pressure on the coro pool
		968	# as the chance is very high that the async_poll coro will be back
		969	# in the idle state when cede returns
401	cede;	970	cede;
402	}	971	}
403		972	schedule; # sleep well
404	schedule;
405	}	973	}
406	};	974	};
		975	$unblock_scheduler->{desc} = "[unblock_sub scheduler]";
407		976
408	sub unblock_sub(&) {	977	sub unblock_sub(&) {
409	my $cb = shift;	978	my $cb = shift;
410		979
411	sub {	980	sub {
412	push @unblock_queue, [$cb, @_];	981	unshift @unblock_queue, [$cb, @_];
413	$unblock_scheduler->ready;	982	$unblock_scheduler->ready;
414	}	983	}
415	}	984	}
416		985
		986	=item $cb = rouse_cb
		987
		988	Create and return a "rouse callback". That's a code reference that,
		989	when called, will remember a copy of its arguments and notify the owner
		990	coro of the callback.
		991
		992	See the next function.
		993
		994	=item @args = rouse_wait [$cb]
		995
		996	Wait for the specified rouse callback (or the last one that was created in
		997	this coro).
		998
		999	As soon as the callback is invoked (or when the callback was invoked
		1000	before C<rouse_wait>), it will return the arguments originally passed to
		1001	the rouse callback. In scalar context, that means you get the I<last>
		1002	argument, just as if C<rouse_wait> had a C<return ($a1, $a2, $a3...)>
		1003	statement at the end.
		1004
		1005	See the section B<HOW TO WAIT FOR A CALLBACK> for an actual usage example.
		1006
417	=back	1007	=back
418		1008
419	=cut	1009	=cut
420		1010
		1011	for my $module (qw(Channel RWLock Semaphore SemaphoreSet Signal Specific)) {
		1012	my $old = defined &{"Coro::$module\::new"} && \&{"Coro::$module\::new"};
		1013
		1014	*{"Coro::$module\::new"} = sub {
		1015	require "Coro/$module.pm";
		1016
		1017	# some modules have their new predefined in State.xs, some don't
		1018	*{"Coro::$module\::new"} = $old
		1019	if $old;
		1020
		1021	goto &{"Coro::$module\::new"};
		1022	};
		1023	}
		1024
421	1;	1025	1;
422		1026
		1027	=head1 HOW TO WAIT FOR A CALLBACK
		1028
		1029	It is very common for a coro to wait for some callback to be
		1030	called. This occurs naturally when you use coro in an otherwise
		1031	event-based program, or when you use event-based libraries.
		1032
		1033	These typically register a callback for some event, and call that callback
		1034	when the event occured. In a coro, however, you typically want to
		1035	just wait for the event, simplyifying things.
		1036
		1037	For example C<< AnyEvent->child >> registers a callback to be called when
		1038	a specific child has exited:
		1039
		1040	my $child_watcher = AnyEvent->child (pid => $pid, cb => sub { ... });
		1041
		1042	But from within a coro, you often just want to write this:
		1043
		1044	my $status = wait_for_child $pid;
		1045
		1046	Coro offers two functions specifically designed to make this easy,
		1047	C<Coro::rouse_cb> and C<Coro::rouse_wait>.
		1048
		1049	The first function, C<rouse_cb>, generates and returns a callback that,
		1050	when invoked, will save its arguments and notify the coro that
		1051	created the callback.
		1052
		1053	The second function, C<rouse_wait>, waits for the callback to be called
		1054	(by calling C<schedule> to go to sleep) and returns the arguments
		1055	originally passed to the callback.
		1056
		1057	Using these functions, it becomes easy to write the C<wait_for_child>
		1058	function mentioned above:
		1059
		1060	sub wait_for_child($) {
		1061	my ($pid) = @_;
		1062
		1063	my $watcher = AnyEvent->child (pid => $pid, cb => Coro::rouse_cb);
		1064
		1065	my ($rpid, $rstatus) = Coro::rouse_wait;
		1066	$rstatus
		1067	}
		1068
		1069	In the case where C<rouse_cb> and C<rouse_wait> are not flexible enough,
		1070	you can roll your own, using C<schedule>:
		1071
		1072	sub wait_for_child($) {
		1073	my ($pid) = @_;
		1074
		1075	# store the current coro in $current,
		1076	# and provide result variables for the closure passed to ->child
		1077	my $current = $Coro::current;
		1078	my ($done, $rstatus);
		1079
		1080	# pass a closure to ->child
		1081	my $watcher = AnyEvent->child (pid => $pid, cb => sub {
		1082	$rstatus = $_[1]; # remember rstatus
		1083	$done = 1; # mark $rstatus as valud
		1084	});
		1085
		1086	# wait until the closure has been called
		1087	schedule while !$done;
		1088
		1089	$rstatus
		1090	}
		1091
		1092
423	=head1 BUGS/LIMITATIONS	1093	=head1 BUGS/LIMITATIONS
424		1094
425	- you must make very sure that no coro is still active on global	1095	=over 4
426	destruction. very bad things might happen otherwise (usually segfaults).
427		1096
		1097	=item fork with pthread backend
		1098
		1099	When Coro is compiled using the pthread backend (which isn't recommended
		1100	but required on many BSDs as their libcs are completely broken), then
		1101	coro will not survive a fork. There is no known workaround except to
		1102	fix your libc and use a saner backend.
		1103
		1104	=item perl process emulation ("threads")
		1105
428	- this module is not thread-safe. You should only ever use this module	1106	This module is not perl-pseudo-thread-safe. You should only ever use this
429	from the same thread (this requirement might be losened in the future	1107	module from the first thread (this requirement might be removed in the
430	to allow per-thread schedulers, but Coro::State does not yet allow	1108	future to allow per-thread schedulers, but Coro::State does not yet allow
431	this).	1109	this). I recommend disabling thread support and using processes, as having
		1110	the windows process emulation enabled under unix roughly halves perl
		1111	performance, even when not used.
		1112
		1113	=item coro switching is not signal safe
		1114
		1115	You must not switch to another coro from within a signal handler (only
		1116	relevant with %SIG - most event libraries provide safe signals), I<unless>
		1117	you are sure you are not interrupting a Coro function.
		1118
		1119	That means you I<MUST NOT> call any function that might "block" the
		1120	current coro - C<cede>, C<schedule> C<< Coro::Semaphore->down >> or
		1121	anything that calls those. Everything else, including calling C<ready>,
		1122	works.
		1123
		1124	=back
		1125
		1126
		1127	=head1 WINDOWS PROCESS EMULATION
		1128
		1129	A great many people seem to be confused about ithreads (for example, Chip
		1130	Salzenberg called me unintelligent, incapable, stupid and gullible,
		1131	while in the same mail making rather confused statements about perl
		1132	ithreads (for example, that memory or files would be shared), showing his
		1133	lack of understanding of this area - if it is hard to understand for Chip,
		1134	it is probably not obvious to everybody).
		1135
		1136	What follows is an ultra-condensed version of my talk about threads in
		1137	scripting languages given on the perl workshop 2009:
		1138
		1139	The so-called "ithreads" were originally implemented for two reasons:
		1140	first, to (badly) emulate unix processes on native win32 perls, and
		1141	secondly, to replace the older, real thread model ("5.005-threads").
		1142
		1143	It does that by using threads instead of OS processes. The difference
		1144	between processes and threads is that threads share memory (and other
		1145	state, such as files) between threads within a single process, while
		1146	processes do not share anything (at least not semantically). That
		1147	means that modifications done by one thread are seen by others, while
		1148	modifications by one process are not seen by other processes.
		1149
		1150	The "ithreads" work exactly like that: when creating a new ithreads
		1151	process, all state is copied (memory is copied physically, files and code
		1152	is copied logically). Afterwards, it isolates all modifications. On UNIX,
		1153	the same behaviour can be achieved by using operating system processes,
		1154	except that UNIX typically uses hardware built into the system to do this
		1155	efficiently, while the windows process emulation emulates this hardware in
		1156	software (rather efficiently, but of course it is still much slower than
		1157	dedicated hardware).
		1158
		1159	As mentioned before, loading code, modifying code, modifying data
		1160	structures and so on is only visible in the ithreads process doing the
		1161	modification, not in other ithread processes within the same OS process.
		1162
		1163	This is why "ithreads" do not implement threads for perl at all, only
		1164	processes. What makes it so bad is that on non-windows platforms, you can
		1165	actually take advantage of custom hardware for this purpose (as evidenced
		1166	by the forks module, which gives you the (i-) threads API, just much
		1167	faster).
		1168
		1169	Sharing data is in the i-threads model is done by transfering data
		1170	structures between threads using copying semantics, which is very slow -
		1171	shared data simply does not exist. Benchmarks using i-threads which are
		1172	communication-intensive show extremely bad behaviour with i-threads (in
		1173	fact, so bad that Coro, which cannot take direct advantage of multiple
		1174	CPUs, is often orders of magnitude faster because it shares data using
		1175	real threads, refer to my talk for details).
		1176
		1177	As summary, i-threads *use* threads to implement processes, while
		1178	the compatible forks module *uses* processes to emulate, uhm,
		1179	processes. I-threads slow down every perl program when enabled, and
		1180	outside of windows, serve no (or little) practical purpose, but
		1181	disadvantages every single-threaded Perl program.
		1182
		1183	This is the reason that I try to avoid the name "ithreads", as it is
		1184	misleading as it implies that it implements some kind of thread model for
		1185	perl, and prefer the name "windows process emulation", which describes the
		1186	actual use and behaviour of it much better.
432		1187
433	=head1 SEE ALSO	1188	=head1 SEE ALSO
434		1189
		1190	Event-Loop integration: L<Coro::AnyEvent>, L<Coro::EV>, L<Coro::Event>.
		1191
		1192	Debugging: L<Coro::Debug>.
		1193
435	Support/Utility: L<Coro::Cont>, L<Coro::Specific>, L<Coro::State>, L<Coro::Util>.	1194	Support/Utility: L<Coro::Specific>, L<Coro::Util>.
436		1195
437	Locking/IPC: L<Coro::Signal>, L<Coro::Channel>, L<Coro::Semaphore>, L<Coro::SemaphoreSet>, L<Coro::RWLock>.	1196	Locking and IPC: L<Coro::Signal>, L<Coro::Channel>, L<Coro::Semaphore>,
		1197	L<Coro::SemaphoreSet>, L<Coro::RWLock>.
438		1198
439	Event/IO: L<Coro::Timer>, L<Coro::Event>, L<Coro::Handle>, L<Coro::Socket>, L<Coro::Select>.	1199	I/O and Timers: L<Coro::Timer>, L<Coro::Handle>, L<Coro::Socket>, L<Coro::AIO>.
440		1200
441	Embedding: L<Coro:MakeMaker>	1201	Compatibility with other modules: L<Coro::LWP> (but see also L<AnyEvent::HTTP> for
		1202	a better-working alternative), L<Coro::BDB>, L<Coro::Storable>,
		1203	L<Coro::Select>.
		1204
		1205	XS API: L<Coro::MakeMaker>.
		1206
		1207	Low level Configuration, Thread Environment, Continuations: L<Coro::State>.
442		1208
443	=head1 AUTHOR	1209	=head1 AUTHOR
444		1210
445	Marc Lehmann <schmorp@schmorp.de>	1211	Marc Lehmann <schmorp@schmorp.de>
446	http://home.schmorp.de/	1212	http://home.schmorp.de/

Diff Legend

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