… | |
… | |
40 | points in your program, so locking and parallel access are rarely an |
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 |
41 | issue, making thread programming much safer and easier than using other |
42 | thread models. |
42 | thread models. |
43 | |
43 | |
44 | Unlike the so-called "Perl threads" (which are not actually real threads |
44 | Unlike the so-called "Perl threads" (which are not actually real threads |
45 | but only the windows process emulation ported to unix, and as such act |
45 | but only the windows process emulation (see section of same name for more |
46 | as processes), Coro provides a full shared address space, which makes |
46 | details) ported to unix, and as such act as processes), Coro provides |
47 | communication between threads very easy. And Coro's threads are fast, |
47 | a full shared address space, which makes communication between threads |
48 | too: disabling the Windows process emulation code in your perl and using |
48 | very easy. And Coro's threads are fast, too: disabling the Windows |
49 | Coro can easily result in a two to four times speed increase for your |
49 | process emulation code in your perl and using Coro can easily result in |
50 | programs. A parallel matrix multiplication benchmark runs over 300 times |
50 | a two to four times speed increase for your programs. A parallel matrix |
51 | faster on a single core than perl's pseudo-threads on a quad core using |
51 | multiplication benchmark runs over 300 times faster on a single core than |
52 | all four cores. |
52 | perl's pseudo-threads on a quad core using all four cores. |
53 | |
53 | |
54 | Coro achieves that by supporting multiple running interpreters that share |
54 | Coro achieves that by supporting multiple running interpreters that share |
55 | data, which is especially useful to code pseudo-parallel processes and |
55 | data, which is especially useful to code pseudo-parallel processes and |
56 | for event-based programming, such as multiple HTTP-GET requests running |
56 | for event-based programming, such as multiple HTTP-GET requests running |
57 | concurrently. See L<Coro::AnyEvent> to learn more on how to integrate Coro |
57 | concurrently. See L<Coro::AnyEvent> to learn more on how to integrate Coro |
… | |
… | |
67 | |
67 | |
68 | =cut |
68 | =cut |
69 | |
69 | |
70 | package Coro; |
70 | package Coro; |
71 | |
71 | |
72 | use strict qw(vars subs); |
72 | use common::sense; |
73 | no warnings "uninitialized"; |
73 | |
|
|
74 | use Carp (); |
74 | |
75 | |
75 | use Guard (); |
76 | use Guard (); |
76 | |
77 | |
77 | use Coro::State; |
78 | use Coro::State; |
78 | |
79 | |
… | |
… | |
80 | |
81 | |
81 | our $idle; # idle handler |
82 | our $idle; # idle handler |
82 | our $main; # main coro |
83 | our $main; # main coro |
83 | our $current; # current coro |
84 | our $current; # current coro |
84 | |
85 | |
85 | our $VERSION = 5.14; |
86 | our $VERSION = 5.26; |
86 | |
87 | |
87 | our @EXPORT = qw(async async_pool cede schedule terminate current unblock_sub); |
88 | our @EXPORT = qw(async async_pool cede schedule terminate current unblock_sub rouse_cb rouse_wait); |
88 | our %EXPORT_TAGS = ( |
89 | our %EXPORT_TAGS = ( |
89 | prio => [qw(PRIO_MAX PRIO_HIGH PRIO_NORMAL PRIO_LOW PRIO_IDLE PRIO_MIN)], |
90 | prio => [qw(PRIO_MAX PRIO_HIGH PRIO_NORMAL PRIO_LOW PRIO_IDLE PRIO_MIN)], |
90 | ); |
91 | ); |
91 | our @EXPORT_OK = (@{$EXPORT_TAGS{prio}}, qw(nready)); |
92 | our @EXPORT_OK = (@{$EXPORT_TAGS{prio}}, qw(nready)); |
92 | |
93 | |
… | |
… | |
123 | |
124 | |
124 | This variable is mainly useful to integrate Coro into event loops. It is |
125 | This variable is mainly useful to integrate Coro into event loops. It is |
125 | usually better to rely on L<Coro::AnyEvent> or L<Coro::EV>, as this is |
126 | usually better to rely on L<Coro::AnyEvent> or L<Coro::EV>, as this is |
126 | pretty low-level functionality. |
127 | pretty low-level functionality. |
127 | |
128 | |
128 | This variable stores either a Coro object or a callback. |
129 | This variable stores a Coro object that is put into the ready queue when |
|
|
130 | there are no other ready threads (without invoking any ready hooks). |
129 | |
131 | |
130 | If it is a callback, the it is called whenever the scheduler finds no |
132 | The default implementation dies with "FATAL: deadlock detected.", followed |
131 | ready coros to run. The default implementation prints "FATAL: |
133 | by a thread listing, because the program has no other way to continue. |
132 | deadlock detected" and exits, because the program has no other way to |
|
|
133 | continue. |
|
|
134 | |
|
|
135 | If it is a coro object, then this object will be readied (without |
|
|
136 | invoking any ready hooks, however) when the scheduler finds no other ready |
|
|
137 | coros to run. |
|
|
138 | |
134 | |
139 | This hook is overwritten by modules such as C<Coro::EV> and |
135 | This hook is overwritten by modules such as C<Coro::EV> and |
140 | C<Coro::AnyEvent> to wait on an external event that hopefully wake up a |
136 | C<Coro::AnyEvent> to wait on an external event that hopefully wake up a |
141 | coro so the scheduler can run it. |
137 | coro so the scheduler can run it. |
142 | |
138 | |
143 | Note that the callback I<must not>, under any circumstances, block |
|
|
144 | the current coro. Normally, this is achieved by having an "idle |
|
|
145 | coro" that calls the event loop and then blocks again, and then |
|
|
146 | readying that coro in the idle handler, or by simply placing the idle |
|
|
147 | coro in this variable. |
|
|
148 | |
|
|
149 | See L<Coro::Event> or L<Coro::AnyEvent> for examples of using this |
139 | See L<Coro::EV> or L<Coro::AnyEvent> for examples of using this technique. |
150 | technique. |
|
|
151 | |
140 | |
152 | Please note that if your callback recursively invokes perl (e.g. for event |
|
|
153 | handlers), then it must be prepared to be called recursively itself. |
|
|
154 | |
|
|
155 | =cut |
141 | =cut |
156 | |
142 | |
157 | $idle = sub { |
143 | # ||= because other modules could have provided their own by now |
158 | require Carp; |
144 | $idle ||= new Coro sub { |
159 | Carp::croak ("FATAL: deadlock detected"); |
145 | require Coro::Debug; |
|
|
146 | die "FATAL: deadlock detected.\n" |
|
|
147 | . Coro::Debug::ps_listing (); |
160 | }; |
148 | }; |
161 | |
149 | |
162 | # this coro is necessary because a coro |
150 | # this coro is necessary because a coro |
163 | # cannot destroy itself. |
151 | # cannot destroy itself. |
164 | our @destroy; |
152 | our @destroy; |
… | |
… | |
206 | Example: Create a new coro that just prints its arguments. |
194 | Example: Create a new coro that just prints its arguments. |
207 | |
195 | |
208 | async { |
196 | async { |
209 | print "@_\n"; |
197 | print "@_\n"; |
210 | } 1,2,3,4; |
198 | } 1,2,3,4; |
211 | |
|
|
212 | =cut |
|
|
213 | |
|
|
214 | sub async(&@) { |
|
|
215 | my $coro = new Coro @_; |
|
|
216 | $coro->ready; |
|
|
217 | $coro |
|
|
218 | } |
|
|
219 | |
199 | |
220 | =item async_pool { ... } [@args...] |
200 | =item async_pool { ... } [@args...] |
221 | |
201 | |
222 | Similar to C<async>, but uses a coro pool, so you should not call |
202 | Similar to C<async>, but uses a coro pool, so you should not call |
223 | terminate or join on it (although you are allowed to), and you get a |
203 | terminate or join on it (although you are allowed to), and you get a |
… | |
… | |
280 | =item schedule |
260 | =item schedule |
281 | |
261 | |
282 | Calls the scheduler. The scheduler will find the next coro that is |
262 | Calls the scheduler. The scheduler will find the next coro that is |
283 | to be run from the ready queue and switches to it. The next coro |
263 | to be run from the ready queue and switches to it. The next coro |
284 | to be run is simply the one with the highest priority that is longest |
264 | to be run is simply the one with the highest priority that is longest |
285 | in its ready queue. If there is no coro ready, it will clal the |
265 | in its ready queue. If there is no coro ready, it will call the |
286 | C<$Coro::idle> hook. |
266 | C<$Coro::idle> hook. |
287 | |
267 | |
288 | Please note that the current coro will I<not> be put into the ready |
268 | Please note that the current coro will I<not> be put into the ready |
289 | queue, so calling this function usually means you will never be called |
269 | queue, so calling this function usually means you will never be called |
290 | again unless something else (e.g. an event handler) calls C<< ->ready >>, |
270 | again unless something else (e.g. an event handler) calls C<< ->ready >>, |
… | |
… | |
587 | wantarray ? @{$self->{_status}} : $self->{_status}[0]; |
567 | wantarray ? @{$self->{_status}} : $self->{_status}[0]; |
588 | } |
568 | } |
589 | |
569 | |
590 | =item $coro->on_destroy (\&cb) |
570 | =item $coro->on_destroy (\&cb) |
591 | |
571 | |
592 | Registers a callback that is called when this coro gets destroyed, |
572 | Registers a callback that is called when this coro thread gets destroyed, |
593 | but before it is joined. The callback gets passed the terminate arguments, |
573 | but before it is joined. The callback gets passed the terminate arguments, |
594 | if any, and I<must not> die, under any circumstances. |
574 | if any, and I<must not> die, under any circumstances. |
595 | |
575 | |
|
|
576 | There can be any number of C<on_destroy> callbacks per coro. |
|
|
577 | |
596 | =cut |
578 | =cut |
597 | |
579 | |
598 | sub on_destroy { |
580 | sub on_destroy { |
599 | my ($self, $cb) = @_; |
581 | my ($self, $cb) = @_; |
600 | |
582 | |
… | |
… | |
602 | } |
584 | } |
603 | |
585 | |
604 | =item $oldprio = $coro->prio ($newprio) |
586 | =item $oldprio = $coro->prio ($newprio) |
605 | |
587 | |
606 | Sets (or gets, if the argument is missing) the priority of the |
588 | Sets (or gets, if the argument is missing) the priority of the |
607 | coro. Higher priority coro get run before lower priority |
589 | coro thread. Higher priority coro get run before lower priority |
608 | coro. Priorities are small signed integers (currently -4 .. +3), |
590 | coros. Priorities are small signed integers (currently -4 .. +3), |
609 | that you can refer to using PRIO_xxx constants (use the import tag :prio |
591 | that you can refer to using PRIO_xxx constants (use the import tag :prio |
610 | to get then): |
592 | to get then): |
611 | |
593 | |
612 | PRIO_MAX > PRIO_HIGH > PRIO_NORMAL > PRIO_LOW > PRIO_IDLE > PRIO_MIN |
594 | PRIO_MAX > PRIO_HIGH > PRIO_NORMAL > PRIO_LOW > PRIO_IDLE > PRIO_MIN |
613 | 3 > 1 > 0 > -1 > -3 > -4 |
595 | 3 > 1 > 0 > -1 > -3 > -4 |
614 | |
596 | |
615 | # set priority to HIGH |
597 | # set priority to HIGH |
616 | current->prio (PRIO_HIGH); |
598 | current->prio (PRIO_HIGH); |
617 | |
599 | |
618 | The idle coro ($Coro::idle) always has a lower priority than any |
600 | The idle coro thread ($Coro::idle) always has a lower priority than any |
619 | existing coro. |
601 | existing coro. |
620 | |
602 | |
621 | Changing the priority of the current coro will take effect immediately, |
603 | Changing the priority of the current coro will take effect immediately, |
622 | but changing the priority of coro in the ready queue (but not |
604 | but changing the priority of a coro in the ready queue (but not running) |
623 | running) will only take effect after the next schedule (of that |
605 | will only take effect after the next schedule (of that coro). This is a |
624 | coro). This is a bug that will be fixed in some future version. |
606 | bug that will be fixed in some future version. |
625 | |
607 | |
626 | =item $newprio = $coro->nice ($change) |
608 | =item $newprio = $coro->nice ($change) |
627 | |
609 | |
628 | Similar to C<prio>, but subtract the given value from the priority (i.e. |
610 | Similar to C<prio>, but subtract the given value from the priority (i.e. |
629 | higher values mean lower priority, just as in unix). |
611 | higher values mean lower priority, just as in UNIX's nice command). |
630 | |
612 | |
631 | =item $olddesc = $coro->desc ($newdesc) |
613 | =item $olddesc = $coro->desc ($newdesc) |
632 | |
614 | |
633 | Sets (or gets in case the argument is missing) the description for this |
615 | Sets (or gets in case the argument is missing) the description for this |
634 | coro. This is just a free-form string you can associate with a |
616 | coro thread. This is just a free-form string you can associate with a |
635 | coro. |
617 | coro. |
636 | |
618 | |
637 | This method simply sets the C<< $coro->{desc} >> member to the given |
619 | This method simply sets the C<< $coro->{desc} >> member to the given |
638 | string. You can modify this member directly if you wish. |
620 | string. You can modify this member directly if you wish, and in fact, this |
|
|
621 | is often preferred to indicate major processing states that cna then be |
|
|
622 | seen for example in a L<Coro::Debug> session: |
|
|
623 | |
|
|
624 | sub my_long_function { |
|
|
625 | local $Coro::current->{desc} = "now in my_long_function"; |
|
|
626 | ... |
|
|
627 | $Coro::current->{desc} = "my_long_function: phase 1"; |
|
|
628 | ... |
|
|
629 | $Coro::current->{desc} = "my_long_function: phase 2"; |
|
|
630 | ... |
|
|
631 | } |
639 | |
632 | |
640 | =cut |
633 | =cut |
641 | |
634 | |
642 | sub desc { |
635 | sub desc { |
643 | my $old = $_[0]{desc}; |
636 | my $old = $_[0]{desc}; |
… | |
… | |
680 | returning a new coderef. Unblocking means that calling the new coderef |
673 | returning a new coderef. Unblocking means that calling the new coderef |
681 | will return immediately without blocking, returning nothing, while the |
674 | will return immediately without blocking, returning nothing, while the |
682 | original code ref will be called (with parameters) from within another |
675 | original code ref will be called (with parameters) from within another |
683 | coro. |
676 | coro. |
684 | |
677 | |
685 | The reason this function exists is that many event libraries (such as the |
678 | The reason this function exists is that many event libraries (such as |
686 | venerable L<Event|Event> module) are not thread-safe (a weaker form |
679 | the venerable L<Event|Event> module) are not thread-safe (a weaker form |
687 | of reentrancy). This means you must not block within event callbacks, |
680 | of reentrancy). This means you must not block within event callbacks, |
688 | otherwise you might suffer from crashes or worse. The only event library |
681 | otherwise you might suffer from crashes or worse. The only event library |
689 | currently known that is safe to use without C<unblock_sub> is L<EV>. |
682 | currently known that is safe to use without C<unblock_sub> is L<EV> (but |
|
|
683 | you might still run into deadlocks if all event loops are blocked). |
|
|
684 | |
|
|
685 | Coro will try to catch you when you block in the event loop |
|
|
686 | ("FATAL:$Coro::IDLE blocked itself"), but this is just best effort and |
|
|
687 | only works when you do not run your own event loop. |
690 | |
688 | |
691 | This function allows your callbacks to block by executing them in another |
689 | This function allows your callbacks to block by executing them in another |
692 | coro where it is safe to block. One example where blocking is handy |
690 | coro where it is safe to block. One example where blocking is handy |
693 | is when you use the L<Coro::AIO|Coro::AIO> functions to save results to |
691 | is when you use the L<Coro::AIO|Coro::AIO> functions to save results to |
694 | disk, for example. |
692 | disk, for example. |
… | |
… | |
736 | unshift @unblock_queue, [$cb, @_]; |
734 | unshift @unblock_queue, [$cb, @_]; |
737 | $unblock_scheduler->ready; |
735 | $unblock_scheduler->ready; |
738 | } |
736 | } |
739 | } |
737 | } |
740 | |
738 | |
741 | =item $cb = Coro::rouse_cb |
739 | =item $cb = rouse_cb |
742 | |
740 | |
743 | Create and return a "rouse callback". That's a code reference that, |
741 | Create and return a "rouse callback". That's a code reference that, |
744 | when called, will remember a copy of its arguments and notify the owner |
742 | when called, will remember a copy of its arguments and notify the owner |
745 | coro of the callback. |
743 | coro of the callback. |
746 | |
744 | |
747 | See the next function. |
745 | See the next function. |
748 | |
746 | |
749 | =item @args = Coro::rouse_wait [$cb] |
747 | =item @args = rouse_wait [$cb] |
750 | |
748 | |
751 | Wait for the specified rouse callback (or the last one that was created in |
749 | Wait for the specified rouse callback (or the last one that was created in |
752 | this coro). |
750 | this coro). |
753 | |
751 | |
754 | As soon as the callback is invoked (or when the callback was invoked |
752 | As soon as the callback is invoked (or when the callback was invoked |
… | |
… | |
760 | See the section B<HOW TO WAIT FOR A CALLBACK> for an actual usage example. |
758 | See the section B<HOW TO WAIT FOR A CALLBACK> for an actual usage example. |
761 | |
759 | |
762 | =back |
760 | =back |
763 | |
761 | |
764 | =cut |
762 | =cut |
|
|
763 | |
|
|
764 | for my $module (qw(Channel RWLock Semaphore SemaphoreSet Signal Specific)) { |
|
|
765 | my $old = defined &{"Coro::$module\::new"} && \&{"Coro::$module\::new"}; |
|
|
766 | |
|
|
767 | *{"Coro::$module\::new"} = sub { |
|
|
768 | require "Coro/$module.pm"; |
|
|
769 | |
|
|
770 | # some modules have their new predefined in State.xs, some don't |
|
|
771 | *{"Coro::$module\::new"} = $old |
|
|
772 | if $old; |
|
|
773 | |
|
|
774 | goto &{"Coro::$module\::new"}; |
|
|
775 | }; |
|
|
776 | } |
765 | |
777 | |
766 | 1; |
778 | 1; |
767 | |
779 | |
768 | =head1 HOW TO WAIT FOR A CALLBACK |
780 | =head1 HOW TO WAIT FOR A CALLBACK |
769 | |
781 | |
… | |
… | |
851 | the windows process emulation enabled under unix roughly halves perl |
863 | the windows process emulation enabled under unix roughly halves perl |
852 | performance, even when not used. |
864 | performance, even when not used. |
853 | |
865 | |
854 | =item coro switching is not signal safe |
866 | =item coro switching is not signal safe |
855 | |
867 | |
856 | You must not switch to another coro from within a signal handler |
868 | You must not switch to another coro from within a signal handler (only |
857 | (only relevant with %SIG - most event libraries provide safe signals). |
869 | relevant with %SIG - most event libraries provide safe signals), I<unless> |
|
|
870 | you are sure you are not interrupting a Coro function. |
858 | |
871 | |
859 | That means you I<MUST NOT> call any function that might "block" the |
872 | That means you I<MUST NOT> call any function that might "block" the |
860 | current coro - C<cede>, C<schedule> C<< Coro::Semaphore->down >> or |
873 | current coro - C<cede>, C<schedule> C<< Coro::Semaphore->down >> or |
861 | anything that calls those. Everything else, including calling C<ready>, |
874 | anything that calls those. Everything else, including calling C<ready>, |
862 | works. |
875 | works. |
863 | |
876 | |
864 | =back |
877 | =back |
865 | |
878 | |
866 | |
879 | |
|
|
880 | =head1 WINDOWS PROCESS EMULATION |
|
|
881 | |
|
|
882 | A great many people seem to be confused about ithreads (for example, Chip |
|
|
883 | Salzenberg called me unintelligent, incapable, stupid and gullible, |
|
|
884 | while in the same mail making rather confused statements about perl |
|
|
885 | ithreads (for example, that memory or files would be shared), showing his |
|
|
886 | lack of understanding of this area - if it is hard to understand for Chip, |
|
|
887 | it is probably not obvious to everybody). |
|
|
888 | |
|
|
889 | What follows is an ultra-condensed version of my talk about threads in |
|
|
890 | scripting languages given on the perl workshop 2009: |
|
|
891 | |
|
|
892 | The so-called "ithreads" were originally implemented for two reasons: |
|
|
893 | first, to (badly) emulate unix processes on native win32 perls, and |
|
|
894 | secondly, to replace the older, real thread model ("5.005-threads"). |
|
|
895 | |
|
|
896 | It does that by using threads instead of OS processes. The difference |
|
|
897 | between processes and threads is that threads share memory (and other |
|
|
898 | state, such as files) between threads within a single process, while |
|
|
899 | processes do not share anything (at least not semantically). That |
|
|
900 | means that modifications done by one thread are seen by others, while |
|
|
901 | modifications by one process are not seen by other processes. |
|
|
902 | |
|
|
903 | The "ithreads" work exactly like that: when creating a new ithreads |
|
|
904 | process, all state is copied (memory is copied physically, files and code |
|
|
905 | is copied logically). Afterwards, it isolates all modifications. On UNIX, |
|
|
906 | the same behaviour can be achieved by using operating system processes, |
|
|
907 | except that UNIX typically uses hardware built into the system to do this |
|
|
908 | efficiently, while the windows process emulation emulates this hardware in |
|
|
909 | software (rather efficiently, but of course it is still much slower than |
|
|
910 | dedicated hardware). |
|
|
911 | |
|
|
912 | As mentioned before, loading code, modifying code, modifying data |
|
|
913 | structures and so on is only visible in the ithreads process doing the |
|
|
914 | modification, not in other ithread processes within the same OS process. |
|
|
915 | |
|
|
916 | This is why "ithreads" do not implement threads for perl at all, only |
|
|
917 | processes. What makes it so bad is that on non-windows platforms, you can |
|
|
918 | actually take advantage of custom hardware for this purpose (as evidenced |
|
|
919 | by the forks module, which gives you the (i-) threads API, just much |
|
|
920 | faster). |
|
|
921 | |
|
|
922 | Sharing data is in the i-threads model is done by transfering data |
|
|
923 | structures between threads using copying semantics, which is very slow - |
|
|
924 | shared data simply does not exist. Benchmarks using i-threads which are |
|
|
925 | communication-intensive show extremely bad behaviour with i-threads (in |
|
|
926 | fact, so bad that Coro, which cannot take direct advantage of multiple |
|
|
927 | CPUs, is often orders of magnitude faster because it shares data using |
|
|
928 | real threads, refer to my talk for details). |
|
|
929 | |
|
|
930 | As summary, i-threads *use* threads to implement processes, while |
|
|
931 | the compatible forks module *uses* processes to emulate, uhm, |
|
|
932 | processes. I-threads slow down every perl program when enabled, and |
|
|
933 | outside of windows, serve no (or little) practical purpose, but |
|
|
934 | disadvantages every single-threaded Perl program. |
|
|
935 | |
|
|
936 | This is the reason that I try to avoid the name "ithreads", as it is |
|
|
937 | misleading as it implies that it implements some kind of thread model for |
|
|
938 | perl, and prefer the name "windows process emulation", which describes the |
|
|
939 | actual use and behaviour of it much better. |
|
|
940 | |
867 | =head1 SEE ALSO |
941 | =head1 SEE ALSO |
868 | |
942 | |
869 | Event-Loop integration: L<Coro::AnyEvent>, L<Coro::EV>, L<Coro::Event>. |
943 | Event-Loop integration: L<Coro::AnyEvent>, L<Coro::EV>, L<Coro::Event>. |
870 | |
944 | |
871 | Debugging: L<Coro::Debug>. |
945 | Debugging: L<Coro::Debug>. |