… | |
… | |
18 | }; |
18 | }; |
19 | |
19 | |
20 | # version 2+ has request and group objects |
20 | # version 2+ has request and group objects |
21 | use IO::AIO 2; |
21 | use IO::AIO 2; |
22 | |
22 | |
|
|
23 | aioreq_pri 4; # give next request a very high priority |
23 | my $req = aio_unlink "/tmp/file", sub { }; |
24 | my $req = aio_unlink "/tmp/file", sub { }; |
24 | $req->cancel; # cancel request if still in queue |
25 | $req->cancel; # cancel request if still in queue |
25 | |
26 | |
26 | my $grp = aio_group sub { print "all stats done\n" }; |
27 | my $grp = aio_group sub { print "all stats done\n" }; |
27 | add $grp aio_stat "..." for ...; |
28 | add $grp aio_stat "..." for ...; |
… | |
… | |
51 | |
52 | |
52 | This module implements asynchronous I/O using whatever means your |
53 | This module implements asynchronous I/O using whatever means your |
53 | operating system supports. |
54 | operating system supports. |
54 | |
55 | |
55 | Currently, a number of threads are started that execute your read/writes |
56 | Currently, a number of threads are started that execute your read/writes |
56 | and signal their completion. You don't need thread support in your libc or |
57 | and signal their completion. You don't need thread support in perl, and |
57 | perl, and the threads created by this module will not be visible to the |
58 | the threads created by this module will not be visible to perl. In the |
58 | pthreads library. In the future, this module might make use of the native |
59 | future, this module might make use of the native aio functions available |
59 | aio functions available on many operating systems. However, they are often |
60 | on many operating systems. However, they are often not well-supported |
60 | not well-supported (Linux doesn't allow them on normal files currently, |
61 | (Linux doesn't allow them on normal files currently, for example), |
61 | for example), and they would only support aio_read and aio_write, so the |
62 | and they would only support aio_read and aio_write, so the remaining |
62 | remaining functionality would have to be implemented using threads anyway. |
63 | functionality would have to be implemented using threads anyway. |
63 | |
64 | |
64 | Although the module will work with in the presence of other threads, it is |
65 | Although the module will work with in the presence of other threads, |
65 | currently not reentrant, so use appropriate locking yourself, always call |
66 | it is currently not reentrant in any way, so use appropriate locking |
66 | C<poll_cb> from within the same thread, or never call C<poll_cb> (or other |
67 | yourself, always call C<poll_cb> from within the same thread, or never |
67 | C<aio_> functions) recursively. |
68 | call C<poll_cb> (or other C<aio_> functions) recursively. |
68 | |
69 | |
69 | =cut |
70 | =cut |
70 | |
71 | |
71 | package IO::AIO; |
72 | package IO::AIO; |
72 | |
73 | |
… | |
… | |
76 | use base 'Exporter'; |
77 | use base 'Exporter'; |
77 | |
78 | |
78 | BEGIN { |
79 | BEGIN { |
79 | our $VERSION = '2.0'; |
80 | our $VERSION = '2.0'; |
80 | |
81 | |
81 | our @EXPORT = qw(aio_sendfile aio_read aio_write aio_open aio_close aio_stat |
82 | our @AIO_REQ = qw(aio_sendfile aio_read aio_write aio_open aio_close aio_stat |
82 | aio_lstat aio_unlink aio_rmdir aio_readdir aio_scandir aio_symlink |
83 | aio_lstat aio_unlink aio_rmdir aio_readdir aio_scandir aio_symlink |
83 | aio_fsync aio_fdatasync aio_readahead aio_rename aio_link aio_move |
84 | aio_fsync aio_fdatasync aio_readahead aio_rename aio_link aio_move |
84 | aio_group); |
85 | aio_group aio_nop); |
85 | our @EXPORT_OK = qw(poll_fileno poll_cb min_parallel max_parallel max_outstanding nreqs); |
86 | our @EXPORT = (@AIO_REQ, qw(aioreq_pri)); |
|
|
87 | our @EXPORT_OK = qw(poll_fileno poll_cb poll_wait flush |
|
|
88 | min_parallel max_parallel max_outstanding nreqs); |
86 | |
89 | |
87 | @IO::AIO::GRP::ISA = 'IO::AIO::REQ'; |
90 | @IO::AIO::GRP::ISA = 'IO::AIO::REQ'; |
88 | |
91 | |
89 | require XSLoader; |
92 | require XSLoader; |
90 | XSLoader::load ("IO::AIO", $VERSION); |
93 | XSLoader::load ("IO::AIO", $VERSION); |
… | |
… | |
120 | your pathnames to the locale (or other) encoding in effect in the user |
123 | your pathnames to the locale (or other) encoding in effect in the user |
121 | environment, d) use Glib::filename_from_unicode on unicode filenames or e) |
124 | environment, d) use Glib::filename_from_unicode on unicode filenames or e) |
122 | use something else. |
125 | use something else. |
123 | |
126 | |
124 | =over 4 |
127 | =over 4 |
|
|
128 | |
|
|
129 | =item aioreq_pri $pri |
|
|
130 | |
|
|
131 | Sets the priority for the next aio request. The default priority |
|
|
132 | is C<0>, the minimum and maximum priorities are C<-4> and C<4>, |
|
|
133 | respectively. Requests with higher priority will be serviced first. |
|
|
134 | |
|
|
135 | The priority will be reset to C<0> after each call to one of the C<aio_> |
|
|
136 | functions. |
|
|
137 | |
|
|
138 | Example: open a file with low priority, then read something from it with |
|
|
139 | higher priority so the read request is serviced before other low priority |
|
|
140 | open requests (potentially spamming the cache): |
|
|
141 | |
|
|
142 | aioreq_pri -3; |
|
|
143 | aio_open ..., sub { |
|
|
144 | return unless $_[0]; |
|
|
145 | |
|
|
146 | aioreq_pri -2; |
|
|
147 | aio_read $_[0], ..., sub { |
|
|
148 | ... |
|
|
149 | }; |
|
|
150 | }; |
|
|
151 | |
|
|
152 | =item aioreq_nice $pri_adjust |
|
|
153 | |
|
|
154 | Similar to C<aioreq_pri>, but subtracts the given value from the current |
|
|
155 | priority, so effects are cumulative. |
125 | |
156 | |
126 | =item aio_open $pathname, $flags, $mode, $callback->($fh) |
157 | =item aio_open $pathname, $flags, $mode, $callback->($fh) |
127 | |
158 | |
128 | Asynchronously open or create a file and call the callback with a newly |
159 | Asynchronously open or create a file and call the callback with a newly |
129 | created filehandle for the file. |
160 | created filehandle for the file. |
… | |
… | |
352 | Scans a directory (similar to C<aio_readdir>) but additionally tries to |
383 | Scans a directory (similar to C<aio_readdir>) but additionally tries to |
353 | separate the entries of directory C<$path> into two sets of names, ones |
384 | separate the entries of directory C<$path> into two sets of names, ones |
354 | you can recurse into (directories or links to them), and ones you cannot |
385 | you can recurse into (directories or links to them), and ones you cannot |
355 | recurse into (everything else). |
386 | recurse into (everything else). |
356 | |
387 | |
357 | C<aio_scandir> is a composite request that consists of many sub |
388 | C<aio_scandir> is a composite request that creates of many sub requests_ |
358 | requests. C<$maxreq> specifies the maximum number of outstanding aio |
389 | C<$maxreq> specifies the maximum number of outstanding aio requests that |
359 | requests that this function generates. If it is C<< <= 0 >>, then a |
390 | this function generates. If it is C<< <= 0 >>, then a suitable default |
360 | suitable default will be chosen (currently 8). |
391 | will be chosen (currently 6). |
361 | |
392 | |
362 | On error, the callback is called without arguments, otherwise it receives |
393 | On error, the callback is called without arguments, otherwise it receives |
363 | two array-refs with path-relative entry names. |
394 | two array-refs with path-relative entry names. |
364 | |
395 | |
365 | Example: |
396 | Example: |
… | |
… | |
404 | sub aio_scandir($$$) { |
435 | sub aio_scandir($$$) { |
405 | my ($path, $maxreq, $cb) = @_; |
436 | my ($path, $maxreq, $cb) = @_; |
406 | |
437 | |
407 | my $grp = aio_group $cb; |
438 | my $grp = aio_group $cb; |
408 | |
439 | |
409 | $maxreq = 8 if $maxreq <= 0; |
440 | $maxreq = 6 if $maxreq <= 0; |
410 | |
441 | |
411 | # stat once |
442 | # stat once |
412 | add $grp aio_stat $path, sub { |
443 | add $grp aio_stat $path, sub { |
413 | return $grp->result () if $_[0]; |
444 | return $grp->result () if $_[0]; |
414 | my $now = time; |
445 | my $now = time; |
… | |
… | |
445 | my (@dirs, @nondirs); |
476 | my (@dirs, @nondirs); |
446 | |
477 | |
447 | my ($statcb, $schedcb); |
478 | my ($statcb, $schedcb); |
448 | my $nreq = 0; |
479 | my $nreq = 0; |
449 | |
480 | |
|
|
481 | my $statgrp = add $grp aio_group; |
|
|
482 | |
450 | $schedcb = sub { |
483 | $schedcb = sub { |
451 | if (@$entries) { |
484 | if (@$entries) { |
452 | if ($nreq < $maxreq) { |
485 | if ($nreq < $maxreq) { |
453 | my $ent = pop @$entries; |
486 | my $ent = pop @$entries; |
454 | $nreq++; |
487 | $nreq++; |
455 | add $grp aio_stat "$path/$ent/.", sub { $statcb->($_[0], $ent) }; |
488 | add $statgrp aio_stat "$path/$ent/.", sub { $statcb->($_[0], $ent) }; |
456 | } |
489 | } |
457 | } elsif (!$nreq) { |
490 | } elsif (!$nreq) { |
458 | # finished |
491 | # finished |
|
|
492 | $statgrp->cancel; |
459 | undef $statcb; |
493 | undef $statcb; |
460 | undef $schedcb; |
494 | undef $schedcb; |
461 | $grp->result (\@dirs, \@nondirs) if $cb; |
495 | $grp->result (\@dirs, \@nondirs); |
462 | undef $cb; |
|
|
463 | } |
496 | } |
464 | }; |
497 | }; |
465 | $statcb = sub { |
498 | $statcb = sub { |
466 | my ($status, $entry) = @_; |
499 | my ($status, $entry) = @_; |
467 | |
500 | |
… | |
… | |
531 | add $grp |
564 | add $grp |
532 | (aio_stat ...), |
565 | (aio_stat ...), |
533 | (aio_stat ...), |
566 | (aio_stat ...), |
534 | ...; |
567 | ...; |
535 | |
568 | |
|
|
569 | =item aio_nop $callback->() |
|
|
570 | |
|
|
571 | This is a special request - it does nothing in itself and is only used for |
|
|
572 | side effects, such as when you want to add a dummy request to a group so |
|
|
573 | that finishing the requests in the group depends on executing the given |
|
|
574 | code. |
|
|
575 | |
|
|
576 | While this request does nothing, it still goes through the execution |
|
|
577 | phase and still requires a worker thread. Thus, the callback will not |
|
|
578 | be executed immediately but only after other requests in the queue have |
|
|
579 | entered their execution phase. This can be used to measure request |
|
|
580 | latency. |
|
|
581 | |
536 | =item IO::AIO::aio_sleep $fractional_seconds, $callback->() *NOT EXPORTED* |
582 | =item IO::AIO::aio_sleep $fractional_seconds, $callback->() *NOT EXPORTED* |
537 | |
583 | |
538 | Mainly used for debugging and benchmarking, this aio request puts one of |
584 | Mainly used for debugging and benchmarking, this aio request puts one of |
539 | the request workers to sleep for the given time. |
585 | the request workers to sleep for the given time. |
540 | |
586 | |
… | |
… | |
558 | callback) and B<done> (request has reached the end of its lifetime and |
604 | callback) and B<done> (request has reached the end of its lifetime and |
559 | holds no resources anymore). |
605 | holds no resources anymore). |
560 | |
606 | |
561 | =over 4 |
607 | =over 4 |
562 | |
608 | |
563 | =item $req->cancel |
609 | =item cancel $req |
564 | |
610 | |
565 | Cancels the request, if possible. Has the effect of skipping execution |
611 | Cancels the request, if possible. Has the effect of skipping execution |
566 | when entering the B<execute> state and skipping calling the callback when |
612 | when entering the B<execute> state and skipping calling the callback when |
567 | entering the the B<result> state, but will leave the request otherwise |
613 | entering the the B<result> state, but will leave the request otherwise |
568 | untouched. That means that requests that currently execute will not be |
614 | untouched. That means that requests that currently execute will not be |
569 | stopped and resources held by the request will not be freed prematurely. |
615 | stopped and resources held by the request will not be freed prematurely. |
570 | |
616 | |
|
|
617 | =item cb $req $callback->(...) |
|
|
618 | |
|
|
619 | Replace (or simply set) the callback registered to the request. |
|
|
620 | |
571 | =back |
621 | =back |
572 | |
622 | |
573 | =head2 IO::AIO::GRP CLASS |
623 | =head2 IO::AIO::GRP CLASS |
574 | |
624 | |
575 | This class is a subclass of L<IO::AIO::REQ>, so all its methods apply to |
625 | This class is a subclass of L<IO::AIO::REQ>, so all its methods apply to |
… | |
… | |
601 | }; |
651 | }; |
602 | |
652 | |
603 | This makes it very easy to create composite requests (see the source of |
653 | This makes it very easy to create composite requests (see the source of |
604 | C<aio_move> for an application) that work and feel like simple requests. |
654 | C<aio_move> for an application) that work and feel like simple requests. |
605 | |
655 | |
|
|
656 | =over 4 |
|
|
657 | |
606 | The IO::AIO::GRP objects will be cleaned up during calls to |
658 | =item * The IO::AIO::GRP objects will be cleaned up during calls to |
607 | C<IO::AIO::poll_cb>, just like any other request. |
659 | C<IO::AIO::poll_cb>, just like any other request. |
608 | |
660 | |
609 | They can be canceled like any other request. Canceling will cancel not |
661 | =item * They can be canceled like any other request. Canceling will cancel not |
610 | only the request itself, but also all requests it contains. |
662 | only the request itself, but also all requests it contains. |
611 | |
663 | |
612 | They can also can also be added to other IO::AIO::GRP objects. |
664 | =item * They can also can also be added to other IO::AIO::GRP objects. |
|
|
665 | |
|
|
666 | =item * You must not add requests to a group from within the group callback (or |
|
|
667 | any later time). |
|
|
668 | |
|
|
669 | =item * This does not harmonise well with C<max_outstanding>, so best do |
|
|
670 | not combine C<aio_group> with it. Groups and feeders are recommended for |
|
|
671 | this kind of concurrency-limiting. |
|
|
672 | |
|
|
673 | =back |
613 | |
674 | |
614 | Their lifetime, simplified, looks like this: when they are empty, they |
675 | Their lifetime, simplified, looks like this: when they are empty, they |
615 | will finish very quickly. If they contain only requests that are in the |
676 | will finish very quickly. If they contain only requests that are in the |
616 | C<done> state, they will also finish. Otherwise they will continue to |
677 | C<done> state, they will also finish. Otherwise they will continue to |
617 | exist. |
678 | exist. |
… | |
… | |
621 | group. And only when all those requests have finished will the the group |
682 | group. And only when all those requests have finished will the the group |
622 | itself finish. |
683 | itself finish. |
623 | |
684 | |
624 | =over 4 |
685 | =over 4 |
625 | |
686 | |
|
|
687 | =item add $grp ... |
|
|
688 | |
626 | =item $grp->add (...) |
689 | =item $grp->add (...) |
627 | |
|
|
628 | =item add $grp ... |
|
|
629 | |
690 | |
630 | Add one or more requests to the group. Any type of L<IO::AIO::REQ> can |
691 | Add one or more requests to the group. Any type of L<IO::AIO::REQ> can |
631 | be added, including other groups, as long as you do not create circular |
692 | be added, including other groups, as long as you do not create circular |
632 | dependencies. |
693 | dependencies. |
633 | |
694 | |
… | |
… | |
635 | |
696 | |
636 | =item $grp->result (...) |
697 | =item $grp->result (...) |
637 | |
698 | |
638 | Set the result value(s) that will be passed to the group callback when all |
699 | Set the result value(s) that will be passed to the group callback when all |
639 | subrequests have finished. By default, no argument will be passed. |
700 | subrequests have finished. By default, no argument will be passed. |
|
|
701 | |
|
|
702 | =item feed $grp $callback->($grp) |
|
|
703 | |
|
|
704 | [VERY EXPERIMENTAL] |
|
|
705 | |
|
|
706 | Sets a feeder/generator on this group: every group can have an attached |
|
|
707 | generator that generates requests if idle. The idea behind this is that, |
|
|
708 | although you could just queue as many requests as you want in a group, |
|
|
709 | this might starve other requests for a potentially long time. For |
|
|
710 | example, C<aio_scandir> might generate hundreds of thousands C<aio_stat> |
|
|
711 | requests, delaying any later requests for a long time. |
|
|
712 | |
|
|
713 | To avoid this, and allow incremental generation of requests, you can |
|
|
714 | instead a group and set a feeder on it that generates those requests. The |
|
|
715 | feed callback will be called whenever there are few enough (see C<limit>, |
|
|
716 | below) requests active in the group itself and is expected to queue more |
|
|
717 | requests. |
|
|
718 | |
|
|
719 | The feed callback can queue as many requests as it likes (i.e. C<add> does |
|
|
720 | not impose any limits). |
|
|
721 | |
|
|
722 | If the feed does not queue more requests when called, it will be |
|
|
723 | automatically removed from the group. |
|
|
724 | |
|
|
725 | If the feed limit is C<0>, it will be set to C<2> automatically. |
|
|
726 | |
|
|
727 | Example: |
|
|
728 | |
|
|
729 | # stat all files in @files, but only ever use four aio requests concurrently: |
|
|
730 | |
|
|
731 | my $grp = aio_group sub { print "finished\n" }; |
|
|
732 | limit $grp 4; |
|
|
733 | feed $grp sub { |
|
|
734 | my $file = pop @files |
|
|
735 | or return; |
|
|
736 | |
|
|
737 | add $grp aio_stat $file, sub { ... }; |
|
|
738 | }; |
|
|
739 | |
|
|
740 | =item limit $grp $num |
|
|
741 | |
|
|
742 | Sets the feeder limit for the group: The feeder will be called whenever |
|
|
743 | the group contains less than this many requests. |
|
|
744 | |
|
|
745 | Setting the limit to C<0> will pause the feeding process. |
640 | |
746 | |
641 | =back |
747 | =back |
642 | |
748 | |
643 | =head2 SUPPORT FUNCTIONS |
749 | =head2 SUPPORT FUNCTIONS |
644 | |
750 | |
… | |
… | |
702 | IO::AIO::poll_wait, IO::AIO::poll_cb |
808 | IO::AIO::poll_wait, IO::AIO::poll_cb |
703 | if IO::AIO::nreqs; |
809 | if IO::AIO::nreqs; |
704 | |
810 | |
705 | =item IO::AIO::min_parallel $nthreads |
811 | =item IO::AIO::min_parallel $nthreads |
706 | |
812 | |
707 | Set the minimum number of AIO threads to C<$nthreads>. The current default |
813 | Set the minimum number of AIO threads to C<$nthreads>. The current |
708 | is C<4>, which means four asynchronous operations can be done at one time |
814 | default is C<8>, which means eight asynchronous operations can execute |
709 | (the number of outstanding operations, however, is unlimited). |
815 | concurrently at any one time (the number of outstanding requests, |
|
|
816 | however, is unlimited). |
710 | |
817 | |
711 | IO::AIO starts threads only on demand, when an AIO request is queued and |
818 | IO::AIO starts threads only on demand, when an AIO request is queued and |
712 | no free thread exists. |
819 | no free thread exists. |
713 | |
820 | |
714 | It is recommended to keep the number of threads low, as some Linux |
821 | It is recommended to keep the number of threads relatively low, as some |
715 | kernel versions will scale negatively with the number of threads (higher |
822 | Linux kernel versions will scale negatively with the number of threads |
716 | parallelity => MUCH higher latency). With current Linux 2.6 versions, 4-32 |
823 | (higher parallelity => MUCH higher latency). With current Linux 2.6 |
717 | threads should be fine. |
824 | versions, 4-32 threads should be fine. |
718 | |
825 | |
719 | Under most circumstances you don't need to call this function, as the |
826 | Under most circumstances you don't need to call this function, as the |
720 | module selects a default that is suitable for low to moderate load. |
827 | module selects a default that is suitable for low to moderate load. |
721 | |
828 | |
722 | =item IO::AIO::max_parallel $nthreads |
829 | =item IO::AIO::max_parallel $nthreads |
… | |
… | |
733 | |
840 | |
734 | Under normal circumstances you don't need to call this function. |
841 | Under normal circumstances you don't need to call this function. |
735 | |
842 | |
736 | =item $oldnreqs = IO::AIO::max_outstanding $nreqs |
843 | =item $oldnreqs = IO::AIO::max_outstanding $nreqs |
737 | |
844 | |
|
|
845 | [DEPRECATED] |
|
|
846 | |
738 | Sets the maximum number of outstanding requests to C<$nreqs>. If you |
847 | Sets the maximum number of outstanding requests to C<$nreqs>. If you |
739 | try to queue up more than this number of requests, the caller will block until |
848 | try to queue up more than this number of requests, the caller will block until |
740 | some requests have been handled. |
849 | some requests have been handled. |
741 | |
850 | |
742 | The default is very large, so normally there is no practical limit. If you |
851 | The default is very large, so normally there is no practical limit. If you |
743 | queue up many requests in a loop it often improves speed if you set |
852 | queue up many requests in a loop it often improves speed if you set |
744 | this to a relatively low number, such as C<100>. |
853 | this to a relatively low number, such as C<100>. |
|
|
854 | |
|
|
855 | This function does not work well together with C<aio_group>'s, and their |
|
|
856 | feeder interface is better suited to limiting concurrency, so do not use |
|
|
857 | this function. |
745 | |
858 | |
746 | Under normal circumstances you don't need to call this function. |
859 | Under normal circumstances you don't need to call this function. |
747 | |
860 | |
748 | =back |
861 | =back |
749 | |
862 | |
… | |
… | |
763 | or return undef; |
876 | or return undef; |
764 | |
877 | |
765 | *$sym |
878 | *$sym |
766 | } |
879 | } |
767 | |
880 | |
768 | min_parallel 4; |
881 | min_parallel 8; |
769 | |
882 | |
770 | END { |
883 | END { |
771 | max_parallel 0; |
884 | max_parallel 0; |
772 | } |
885 | } |
773 | |
886 | |
… | |
… | |
787 | |
900 | |
788 | In short: the parent will, after a short pause, continue as if fork had |
901 | In short: the parent will, after a short pause, continue as if fork had |
789 | not been called, while the child will act as if IO::AIO has not been used |
902 | not been called, while the child will act as if IO::AIO has not been used |
790 | yet. |
903 | yet. |
791 | |
904 | |
|
|
905 | =head2 MEMORY USAGE |
|
|
906 | |
|
|
907 | Each aio request uses - depending on your architecture - around 128 bytes |
|
|
908 | of memory. In addition, stat requests need a stat buffer (possibly a few |
|
|
909 | hundred bytes). Perl scalars and other data passed into aio requests will |
|
|
910 | also be locked. |
|
|
911 | |
|
|
912 | This is now awfully much, so queuing lots of requests is not usually a |
|
|
913 | problem. |
|
|
914 | |
|
|
915 | Each thread needs a stack area which is usually around 16k, sometimes much |
|
|
916 | larger, depending on the OS. |
|
|
917 | |
792 | =head1 SEE ALSO |
918 | =head1 SEE ALSO |
793 | |
919 | |
794 | L<Coro>, L<Linux::AIO> (obsolete). |
920 | L<Coro::AIO>. |
795 | |
921 | |
796 | =head1 AUTHOR |
922 | =head1 AUTHOR |
797 | |
923 | |
798 | Marc Lehmann <schmorp@schmorp.de> |
924 | Marc Lehmann <schmorp@schmorp.de> |
799 | http://home.schmorp.de/ |
925 | http://home.schmorp.de/ |