1 |
=head1 NAME |
2 |
|
3 |
Coro::State - create and manage simple coroutines |
4 |
|
5 |
=head1 SYNOPSIS |
6 |
|
7 |
use Coro::State; |
8 |
|
9 |
$new = new Coro::State sub { |
10 |
print "in coroutine (called with @_), switching back\n"; |
11 |
$new->transfer($main); |
12 |
print "in coroutine again, switching back\n"; |
13 |
$new->transfer($main); |
14 |
}, 5; |
15 |
|
16 |
$main = new Coro::State; |
17 |
|
18 |
print "in main, switching to coroutine\n"; |
19 |
$main->transfer($new); |
20 |
print "back in main, switch to coroutine again\n"; |
21 |
$main->transfer($new); |
22 |
print "back in main\n"; |
23 |
|
24 |
=head1 DESCRIPTION |
25 |
|
26 |
This module implements coroutines. Coroutines, similar to continuations, |
27 |
allow you to run more than one "thread of execution" in parallel. Unlike |
28 |
threads this, only voluntary switching is used so locking problems are |
29 |
greatly reduced. |
30 |
|
31 |
This module provides only low-level functionality. See L<Coro> and related |
32 |
modules for a more useful process abstraction including scheduling. |
33 |
|
34 |
=head2 MEMORY CONSUMPTION |
35 |
|
36 |
A newly created coroutine that has not been used only allocates a |
37 |
relatively small (a few hundred bytes) structure. Only on the first |
38 |
C<transfer> will perl stacks (a few k) and optionally C stack (4-16k) be |
39 |
allocated. On systems supporting mmap a 128k stack is allocated, on the |
40 |
assumption that the OS has on-demand virtual memory. All this is very |
41 |
system-dependent. On my i686-pc-linux-gnu system this amounts to about 10k |
42 |
per coroutine, 5k when the experimental context sharing is enabled. |
43 |
|
44 |
=over 4 |
45 |
|
46 |
=cut |
47 |
|
48 |
package Coro::State; |
49 |
|
50 |
BEGIN { |
51 |
$VERSION = 0.45; |
52 |
|
53 |
require XSLoader; |
54 |
XSLoader::load Coro::State, $VERSION; |
55 |
} |
56 |
|
57 |
use base 'Exporter'; |
58 |
|
59 |
@EXPORT_OK = qw(SAVE_DEFAV SAVE_DEFSV SAVE_ERRSV SAVE_CCTXT); |
60 |
|
61 |
=item $coro = new [$coderef] [, @args...] |
62 |
|
63 |
Create a new coroutine and return it. The first C<transfer> call to this |
64 |
coroutine will start execution at the given coderef. If the subroutine |
65 |
returns it will be executed again. |
66 |
|
67 |
If the coderef is omitted this function will create a new "empty" |
68 |
coroutine, i.e. a coroutine that cannot be transfered to but can be used |
69 |
to save the current coroutine in. |
70 |
|
71 |
=cut |
72 |
|
73 |
# this is called (or rather: goto'ed) for each and every |
74 |
# new coroutine. IT MUST NEVER RETURN and should not call |
75 |
# anything that changes the stacklevel (like eval). |
76 |
sub initialize { |
77 |
my $proc = shift; |
78 |
eval { |
79 |
&$proc while 1; |
80 |
}; |
81 |
if ($@) { |
82 |
print STDERR "FATAL: uncaught exception\n$@"; |
83 |
} |
84 |
_exit 255; |
85 |
} |
86 |
|
87 |
sub new { |
88 |
my $class = shift; |
89 |
my $proc = shift || sub { die "tried to transfer to an empty coroutine" }; |
90 |
bless _newprocess [$proc, @_], $class; |
91 |
} |
92 |
|
93 |
=item $prev->transfer($next,$flags) |
94 |
|
95 |
Save the state of the current subroutine in C<$prev> and switch to the |
96 |
coroutine saved in C<$next>. |
97 |
|
98 |
The "state" of a subroutine includes the scope, i.e. lexical variables and |
99 |
the current execution state. The C<$flags> value can be used to specify |
100 |
that additional state be saved (and later restored), by C<||>-ing the |
101 |
following constants together: |
102 |
|
103 |
Constant Effect |
104 |
SAVE_DEFAV save/restore @_ |
105 |
SAVE_DEFSV save/restore $_ |
106 |
SAVE_ERRSV save/restore $@ |
107 |
SAVE_CCTXT save/restore C-stack (you usually want this) |
108 |
|
109 |
These constants are not exported by default. |
110 |
|
111 |
If you feel that something important is missing then tell me. Also |
112 |
remember that every function call that might call C<transfer> (such |
113 |
as C<Coro::Channel::put>) might clobber any global and/or special |
114 |
variables. Yes, this is by design ;) You can always create your own |
115 |
process abstraction model that saves these variables. |
116 |
|
117 |
The easiest way to do this is to create your own scheduling primitive like |
118 |
this: |
119 |
|
120 |
sub schedule { |
121 |
local ($_, $@, ...); |
122 |
$old->transfer($new); |
123 |
} |
124 |
|
125 |
IMPLEMENTORS NOTE: all Coro::State functions/methods expect either the |
126 |
usual Coro::State object or a hashref with a key named "_coro_state" that |
127 |
contains the real Coro::State object. That is, you can do: |
128 |
|
129 |
$obj->{_coro_state} = new Coro::State ...; |
130 |
Coro::State::transfer(..., $obj); |
131 |
|
132 |
This exists mainly to ease subclassing (wether through @ISA or not). |
133 |
|
134 |
=cut |
135 |
|
136 |
=item Coro::State::flush |
137 |
|
138 |
To be efficient (actually, to not be abysmaly slow), this module does |
139 |
some fair amount of caching (a possibly complex structure for every |
140 |
subroutine in use). If you don't use coroutines anymore or you want to |
141 |
reclaim some memory then you can call this function which will flush all |
142 |
internal caches. The caches will be rebuilt when needed so this is a safe |
143 |
operation. |
144 |
|
145 |
=cut |
146 |
|
147 |
1; |
148 |
|
149 |
=back |
150 |
|
151 |
=head1 BUGS |
152 |
|
153 |
This module has not yet been extensively tested. Expect segfaults and |
154 |
specially memleaks. |
155 |
|
156 |
This module is not thread-safe. You must only ever use this module from |
157 |
the same thread (this requirenmnt might be loosened in the future). |
158 |
|
159 |
=head1 SEE ALSO |
160 |
|
161 |
L<Coro>. |
162 |
|
163 |
=head1 AUTHOR |
164 |
|
165 |
Marc Lehmann <pcg@goof.com> |
166 |
http://www.goof.com/pcg/marc/ |
167 |
|
168 |
=cut |
169 |
|