1AIO(3) User Contributed Perl Documentation AIO(3)
2
3
4
6 IO::AIO - Asynchronous Input/Output
7
9 use IO::AIO;
10
11 aio_open "/etc/passwd", O_RDONLY, 0, sub {
12 my $fh = shift
13 or die "/etc/passwd: $!";
14 ...
15 };
16
17 aio_unlink "/tmp/file", sub { };
18
19 aio_read $fh, 30000, 1024, $buffer, 0, sub {
20 $_[0] > 0 or die "read error: $!";
21 };
22
23 # version 2+ has request and group objects
24 use IO::AIO 2;
25
26 aioreq_pri 4; # give next request a very high priority
27 my $req = aio_unlink "/tmp/file", sub { };
28 $req->cancel; # cancel request if still in queue
29
30 my $grp = aio_group sub { print "all stats done\n" };
31 add $grp aio_stat "..." for ...;
32
33 # AnyEvent integration
34 open my $fh, "<&=" . IO::AIO::poll_fileno or die "$!";
35 my $w = AnyEvent->io (fh => $fh, poll => 'r', cb => sub { IO::AIO::poll_cb });
36
37 # Event integration
38 Event->io (fd => IO::AIO::poll_fileno,
39 poll => 'r',
40 cb => \&IO::AIO::poll_cb);
41
42 # Glib/Gtk2 integration
43 add_watch Glib::IO IO::AIO::poll_fileno,
44 in => sub { IO::AIO::poll_cb; 1 };
45
46 # Tk integration
47 Tk::Event::IO->fileevent (IO::AIO::poll_fileno, "",
48 readable => \&IO::AIO::poll_cb);
49
50 # Danga::Socket integration
51 Danga::Socket->AddOtherFds (IO::AIO::poll_fileno =>
52 \&IO::AIO::poll_cb);
53
55 This module implements asynchronous I/O using whatever means your oper‐
56 ating system supports.
57
58 Asynchronous means that operations that can normally block your program
59 (e.g. reading from disk) will be done asynchronously: the operation
60 will still block, but you can do something else in the meantime. This
61 is extremely useful for programs that need to stay interactive even
62 when doing heavy I/O (GUI programs, high performance network servers
63 etc.), but can also be used to easily do operations in parallel that
64 are normally done sequentially, e.g. stat'ing many files, which is much
65 faster on a RAID volume or over NFS when you do a number of stat opera‐
66 tions concurrently.
67
68 While most of this works on all types of file descriptors (for example
69 sockets), using these functions on file descriptors that support non‐
70 blocking operation (again, sockets, pipes etc.) is very inefficient or
71 might not work (aio_read fails on sockets/pipes/fifos). Use an event
72 loop for that (such as the Event module): IO::AIO will naturally fit
73 into such an event loop itself.
74
75 In this version, a number of threads are started that execute your
76 requests and signal their completion. You don't need thread support in
77 perl, and the threads created by this module will not be visible to
78 perl. In the future, this module might make use of the native aio func‐
79 tions available on many operating systems. However, they are often not
80 well-supported or restricted (GNU/Linux doesn't allow them on normal
81 files currently, for example), and they would only support aio_read and
82 aio_write, so the remaining functionality would have to be implemented
83 using threads anyway.
84
85 Although the module will work with in the presence of other (Perl-)
86 threads, it is currently not reentrant in any way, so use appropriate
87 locking yourself, always call "poll_cb" from within the same thread, or
88 never call "poll_cb" (or other "aio_" functions) recursively.
89
90 EXAMPLE
91
92 This is a simple example that uses the Event module and loads
93 /etc/passwd asynchronously:
94
95 use Fcntl;
96 use Event;
97 use IO::AIO;
98
99 # register the IO::AIO callback with Event
100 Event->io (fd => IO::AIO::poll_fileno,
101 poll => 'r',
102 cb => \&IO::AIO::poll_cb);
103
104 # queue the request to open /etc/passwd
105 aio_open "/etc/passwd", O_RDONLY, 0, sub {
106 my $fh = shift
107 or die "error while opening: $!";
108
109 # stat'ing filehandles is generally non-blocking
110 my $size = -s $fh;
111
112 # queue a request to read the file
113 my $contents;
114 aio_read $fh, 0, $size, $contents, 0, sub {
115 $_[0] == $size
116 or die "short read: $!";
117
118 close $fh;
119
120 # file contents now in $contents
121 print $contents;
122
123 # exit event loop and program
124 Event::unloop;
125 };
126 };
127
128 # possibly queue up other requests, or open GUI windows,
129 # check for sockets etc. etc.
130
131 # process events as long as there are some:
132 Event::loop;
133
135 Every "aio_*" function creates a request. which is a C data structure
136 not directly visible to Perl.
137
138 If called in non-void context, every request function returns a Perl
139 object representing the request. In void context, nothing is returned,
140 which saves a bit of memory.
141
142 The perl object is a fairly standard ref-to-hash object. The hash con‐
143 tents are not used by IO::AIO so you are free to store anything you
144 like in it.
145
146 During their existance, aio requests travel through the following
147 states, in order:
148
149 ready
150 Immediately after a request is created it is put into the ready
151 state, waiting for a thread to execute it.
152
153 execute
154 A thread has accepted the request for processing and is currently
155 executing it (e.g. blocking in read).
156
157 pending
158 The request has been executed and is waiting for result processing.
159
160 While request submission and execution is fully asynchronous,
161 result processing is not and relies on the perl interpreter calling
162 "poll_cb" (or another function with the same effect).
163
164 result
165 The request results are processed synchronously by "poll_cb".
166
167 The "poll_cb" function will process all outstanding aio requests by
168 calling their callbacks, freeing memory associated with them and
169 managing any groups they are contained in.
170
171 done
172 Request has reached the end of its lifetime and holds no resources
173 anymore (except possibly for the Perl object, but its connection to
174 the actual aio request is severed and calling its methods will
175 either do nothing or result in a runtime error).
176
178 AIO REQUEST FUNCTIONS
179
180 All the "aio_*" calls are more or less thin wrappers around the syscall
181 with the same name (sans "aio_"). The arguments are similar or identi‐
182 cal, and they all accept an additional (and optional) $callback argu‐
183 ment which must be a code reference. This code reference will get
184 called with the syscall return code (e.g. most syscalls return "-1" on
185 error, unlike perl, which usually delivers "false") as it's sole argu‐
186 ment when the given syscall has been executed asynchronously.
187
188 All functions expecting a filehandle keep a copy of the filehandle
189 internally until the request has finished.
190
191 All functions return request objects of type IO::AIO::REQ that allow
192 further manipulation of those requests while they are in-flight.
193
194 The pathnames you pass to these routines must be absolute and encoded
195 as octets. The reason for the former is that at the time the request is
196 being executed, the current working directory could have changed.
197 Alternatively, you can make sure that you never change the current
198 working directory anywhere in the program and then use relative paths.
199
200 To encode pathnames as octets, either make sure you either: a) always
201 pass in filenames you got from outside (command line, readdir etc.)
202 without tinkering, b) are ASCII or ISO 8859-1, c) use the Encode module
203 and encode your pathnames to the locale (or other) encoding in effect
204 in the user environment, d) use Glib::filename_from_unicode on unicode
205 filenames or e) use something else to ensure your scalar has the cor‐
206 rect contents.
207
208 This works, btw. independent of the internal UTF-8 bit, which IO::AIO
209 handles correctly wether it is set or not.
210
211 $prev_pri = aioreq_pri [$pri]
212 Returns the priority value that would be used for the next request
213 and, if $pri is given, sets the priority for the next aio request.
214
215 The default priority is 0, the minimum and maximum priorities are
216 "-4" and 4, respectively. Requests with higher priority will be
217 serviced first.
218
219 The priority will be reset to 0 after each call to one of the
220 "aio_*" functions.
221
222 Example: open a file with low priority, then read something from it
223 with higher priority so the read request is serviced before other
224 low priority open requests (potentially spamming the cache):
225
226 aioreq_pri -3;
227 aio_open ..., sub {
228 return unless $_[0];
229
230 aioreq_pri -2;
231 aio_read $_[0], ..., sub {
232 ...
233 };
234 };
235
236 aioreq_nice $pri_adjust
237 Similar to "aioreq_pri", but subtracts the given value from the
238 current priority, so the effect is cumulative.
239
240 aio_open $pathname, $flags, $mode, $callback->($fh)
241 Asynchronously open or create a file and call the callback with a
242 newly created filehandle for the file.
243
244 The pathname passed to "aio_open" must be absolute. See API NOTES,
245 above, for an explanation.
246
247 The $flags argument is a bitmask. See the "Fcntl" module for a
248 list. They are the same as used by "sysopen".
249
250 Likewise, $mode specifies the mode of the newly created file, if it
251 didn't exist and "O_CREAT" has been given, just like perl's
252 "sysopen", except that it is mandatory (i.e. use 0 if you don't
253 create new files, and 0666 or 0777 if you do). Note that the $mode
254 will be modified by the umask in effect then the request is being
255 executed, so better never change the umask.
256
257 Example:
258
259 aio_open "/etc/passwd", O_RDONLY, 0, sub {
260 if ($_[0]) {
261 print "open successful, fh is $_[0]\n";
262 ...
263 } else {
264 die "open failed: $!\n";
265 }
266 };
267
268 aio_close $fh, $callback->($status)
269 Asynchronously close a file and call the callback with the result
270 code. WARNING: although accepted, you should not pass in a perl
271 filehandle here, as perl will likely close the file descriptor
272 another time when the filehandle is destroyed. Normally, you can
273 safely call perls "close" or just let filehandles go out of scope.
274
275 This is supposed to be a bug in the API, so that might change. It's
276 therefore best to avoid this function.
277
278 aio_read $fh,$offset,$length, $data,$dataoffset, $callback->($retval)
279 aio_write $fh,$offset,$length, $data,$dataoffset, $callback->($retval)
280 Reads or writes "length" bytes from the specified "fh" and "offset"
281 into the scalar given by "data" and offset "dataoffset" and calls
282 the callback without the actual number of bytes read (or -1 on
283 error, just like the syscall).
284
285 The $data scalar MUST NOT be modified in any way while the request
286 is outstanding. Modifying it can result in segfaults or WW3 (if the
287 necessary/optional hardware is installed).
288
289 Example: Read 15 bytes at offset 7 into scalar $buffer, starting at
290 offset 0 within the scalar:
291
292 aio_read $fh, 7, 15, $buffer, 0, sub {
293 $_[0] > 0 or die "read error: $!";
294 print "read $_[0] bytes: <$buffer>\n";
295 };
296
297 aio_sendfile $out_fh, $in_fh, $in_offset, $length, $callback->($retval)
298 Tries to copy $length bytes from $in_fh to $out_fh. It starts read‐
299 ing at byte offset $in_offset, and starts writing at the current
300 file offset of $out_fh. Because of that, it is not safe to issue
301 more than one "aio_sendfile" per $out_fh, as they will interfere
302 with each other.
303
304 This call tries to make use of a native "sendfile" syscall to pro‐
305 vide zero-copy operation. For this to work, $out_fh should refer to
306 a socket, and $in_fh should refer to mmap'able file.
307
308 If the native sendfile call fails or is not implemented, it will be
309 emulated, so you can call "aio_sendfile" on any type of filehandle
310 regardless of the limitations of the operating system.
311
312 Please note, however, that "aio_sendfile" can read more bytes from
313 $in_fh than are written, and there is no way to find out how many
314 bytes have been read from "aio_sendfile" alone, as "aio_sendfile"
315 only provides the number of bytes written to $out_fh. Only if the
316 result value equals $length one can assume that $length bytes have
317 been read.
318
319 aio_readahead $fh,$offset,$length, $callback->($retval)
320 "aio_readahead" populates the page cache with data from a file so
321 that subsequent reads from that file will not block on disk I/O.
322 The $offset argument specifies the starting point from which data
323 is to be read and $length specifies the number of bytes to be read.
324 I/O is performed in whole pages, so that offset is effectively
325 rounded down to a page boundary and bytes are read up to the next
326 page boundary greater than or equal to (off-set+length).
327 "aio_readahead" does not read beyond the end of the file. The cur‐
328 rent file offset of the file is left unchanged.
329
330 If that syscall doesn't exist (likely if your OS isn't Linux) it
331 will be emulated by simply reading the data, which would have a
332 similar effect.
333
334 aio_stat $fh_or_path, $callback->($status)
335 aio_lstat $fh, $callback->($status)
336 Works like perl's "stat" or "lstat" in void context. The callback
337 will be called after the stat and the results will be available
338 using "stat _" or "-s _" etc...
339
340 The pathname passed to "aio_stat" must be absolute. See API NOTES,
341 above, for an explanation.
342
343 Currently, the stats are always 64-bit-stats, i.e. instead of
344 returning an error when stat'ing a large file, the results will be
345 silently truncated unless perl itself is compiled with large file
346 support.
347
348 Example: Print the length of /etc/passwd:
349
350 aio_stat "/etc/passwd", sub {
351 $_[0] and die "stat failed: $!";
352 print "size is ", -s _, "\n";
353 };
354
355 aio_unlink $pathname, $callback->($status)
356 Asynchronously unlink (delete) a file and call the callback with
357 the result code.
358
359 aio_mknod $path, $mode, $dev, $callback->($status)
360 [EXPERIMENTAL]
361
362 Asynchronously create a device node (or fifo). See mknod(2).
363
364 The only (POSIX-) portable way of calling this function is:
365
366 aio_mknod $path, IO::AIO::S_IFIFO ⎪ $mode, 0, sub { ...
367
368 aio_link $srcpath, $dstpath, $callback->($status)
369 Asynchronously create a new link to the existing object at $srcpath
370 at the path $dstpath and call the callback with the result code.
371
372 aio_symlink $srcpath, $dstpath, $callback->($status)
373 Asynchronously create a new symbolic link to the existing object at
374 $srcpath at the path $dstpath and call the callback with the result
375 code.
376
377 aio_readlink $path, $callback->($link)
378 Asynchronously read the symlink specified by $path and pass it to
379 the callback. If an error occurs, nothing or undef gets passed to
380 the callback.
381
382 aio_rename $srcpath, $dstpath, $callback->($status)
383 Asynchronously rename the object at $srcpath to $dstpath, just as
384 rename(2) and call the callback with the result code.
385
386 aio_mkdir $pathname, $mode, $callback->($status)
387 Asynchronously mkdir (create) a directory and call the callback
388 with the result code. $mode will be modified by the umask at the
389 time the request is executed, so do not change your umask.
390
391 aio_rmdir $pathname, $callback->($status)
392 Asynchronously rmdir (delete) a directory and call the callback
393 with the result code.
394
395 aio_readdir $pathname, $callback->($entries)
396 Unlike the POSIX call of the same name, "aio_readdir" reads an
397 entire directory (i.e. opendir + readdir + closedir). The entries
398 will not be sorted, and will NOT include the "." and ".." entries.
399
400 The callback a single argument which is either "undef" or an array-
401 ref with the filenames.
402
403 aio_load $path, $data, $callback->($status)
404 This is a composite request that tries to fully load the given file
405 into memory. Status is the same as with aio_read.
406
407 aio_copy $srcpath, $dstpath, $callback->($status)
408 Try to copy the file (directories not supported as either source or
409 destination) from $srcpath to $dstpath and call the callback with
410 the 0 (error) or "-1" ok.
411
412 This is a composite request that it creates the destination file
413 with mode 0200 and copies the contents of the source file into it
414 using "aio_sendfile", followed by restoring atime, mtime, access
415 mode and uid/gid, in that order.
416
417 If an error occurs, the partial destination file will be unlinked,
418 if possible, except when setting atime, mtime, access mode and
419 uid/gid, where errors are being ignored.
420
421 aio_move $srcpath, $dstpath, $callback->($status)
422 Try to move the file (directories not supported as either source or
423 destination) from $srcpath to $dstpath and call the callback with
424 the 0 (error) or "-1" ok.
425
426 This is a composite request that tries to rename(2) the file first.
427 If rename files with "EXDEV", it copies the file with "aio_copy"
428 and, if that is successful, unlinking the $srcpath.
429
430 aio_scandir $path, $maxreq, $callback->($dirs, $nondirs)
431 Scans a directory (similar to "aio_readdir") but additionally tries
432 to efficiently separate the entries of directory $path into two
433 sets of names, directories you can recurse into (directories), and
434 ones you cannot recurse into (everything else, including symlinks
435 to directories).
436
437 "aio_scandir" is a composite request that creates of many sub
438 requests_ $maxreq specifies the maximum number of outstanding aio
439 requests that this function generates. If it is "<= 0", then a
440 suitable default will be chosen (currently 4).
441
442 On error, the callback is called without arguments, otherwise it
443 receives two array-refs with path-relative entry names.
444
445 Example:
446
447 aio_scandir $dir, 0, sub {
448 my ($dirs, $nondirs) = @_;
449 print "real directories: @$dirs\n";
450 print "everything else: @$nondirs\n";
451 };
452
453 Implementation notes.
454
455 The "aio_readdir" cannot be avoided, but "stat()"'ing every entry
456 can.
457
458 After reading the directory, the modification time, size etc. of
459 the directory before and after the readdir is checked, and if they
460 match (and isn't the current time), the link count will be used to
461 decide how many entries are directories (if >= 2). Otherwise, no
462 knowledge of the number of subdirectories will be assumed.
463
464 Then entries will be sorted into likely directories (everything
465 without a non-initial dot currently) and likely non-directories
466 (everything else). Then every entry plus an appended "/." will be
467 "stat"'ed, likely directories first. If that succeeds, it assumes
468 that the entry is a directory or a symlink to directory (which will
469 be checked seperately). This is often faster than stat'ing the
470 entry itself because filesystems might detect the type of the entry
471 without reading the inode data (e.g. ext2fs filetype feature).
472
473 If the known number of directories (link count - 2) has been
474 reached, the rest of the entries is assumed to be non-directories.
475
476 This only works with certainty on POSIX (= UNIX) filesystems, which
477 fortunately are the vast majority of filesystems around.
478
479 It will also likely work on non-POSIX filesystems with reduced
480 efficiency as those tend to return 0 or 1 as link counts, which
481 disables the directory counting heuristic.
482
483 aio_rmtree $path, $callback->($status)
484 Delete a directory tree starting (and including) $path, return the
485 status of the final "rmdir" only. This is a composite request that
486 uses "aio_scandir" to recurse into and rmdir directories, and
487 unlink everything else.
488
489 aio_fsync $fh, $callback->($status)
490 Asynchronously call fsync on the given filehandle and call the
491 callback with the fsync result code.
492
493 aio_fdatasync $fh, $callback->($status)
494 Asynchronously call fdatasync on the given filehandle and call the
495 callback with the fdatasync result code.
496
497 If this call isn't available because your OS lacks it or it
498 couldn't be detected, it will be emulated by calling "fsync"
499 instead.
500
501 aio_group $callback->(...)
502 This is a very special aio request: Instead of doing something, it
503 is a container for other aio requests, which is useful if you want
504 to bundle many requests into a single, composite, request with a
505 definite callback and the ability to cancel the whole request with
506 its subrequests.
507
508 Returns an object of class IO::AIO::GRP. See its documentation
509 below for more info.
510
511 Example:
512
513 my $grp = aio_group sub {
514 print "all stats done\n";
515 };
516
517 add $grp
518 (aio_stat ...),
519 (aio_stat ...),
520 ...;
521
522 aio_nop $callback->()
523 This is a special request - it does nothing in itself and is only
524 used for side effects, such as when you want to add a dummy request
525 to a group so that finishing the requests in the group depends on
526 executing the given code.
527
528 While this request does nothing, it still goes through the execu‐
529 tion phase and still requires a worker thread. Thus, the callback
530 will not be executed immediately but only after other requests in
531 the queue have entered their execution phase. This can be used to
532 measure request latency.
533
534 IO::AIO::aio_busy $fractional_seconds, $callback->() *NOT EXPORTED*
535 Mainly used for debugging and benchmarking, this aio request puts
536 one of the request workers to sleep for the given time.
537
538 While it is theoretically handy to have simple I/O scheduling
539 requests like sleep and file handle readable/writable, the overhead
540 this creates is immense (it blocks a thread for a long time) so do
541 not use this function except to put your application under artifi‐
542 cial I/O pressure.
543
544 IO::AIO::REQ CLASS
545
546 All non-aggregate "aio_*" functions return an object of this class when
547 called in non-void context.
548
549 cancel $req
550 Cancels the request, if possible. Has the effect of skipping execu‐
551 tion when entering the execute state and skipping calling the call‐
552 back when entering the the result state, but will leave the request
553 otherwise untouched. That means that requests that currently exe‐
554 cute will not be stopped and resources held by the request will not
555 be freed prematurely.
556
557 cb $req $callback->(...)
558 Replace (or simply set) the callback registered to the request.
559
560 IO::AIO::GRP CLASS
561
562 This class is a subclass of IO::AIO::REQ, so all its methods apply to
563 objects of this class, too.
564
565 A IO::AIO::GRP object is a special request that can contain multiple
566 other aio requests.
567
568 You create one by calling the "aio_group" constructing function with a
569 callback that will be called when all contained requests have entered
570 the "done" state:
571
572 my $grp = aio_group sub {
573 print "all requests are done\n";
574 };
575
576 You add requests by calling the "add" method with one or more
577 "IO::AIO::REQ" objects:
578
579 $grp->add (aio_unlink "...");
580
581 add $grp aio_stat "...", sub {
582 $_[0] or return $grp->result ("error");
583
584 # add another request dynamically, if first succeeded
585 add $grp aio_open "...", sub {
586 $grp->result ("ok");
587 };
588 };
589
590 This makes it very easy to create composite requests (see the source of
591 "aio_move" for an application) that work and feel like simple requests.
592
593 * The IO::AIO::GRP objects will be cleaned up during calls to
594 "IO::AIO::poll_cb", just like any other request.
595 * They can be canceled like any other request. Canceling will cancel
596 not only the request itself, but also all requests it contains.
597 * They can also can also be added to other IO::AIO::GRP objects.
598 * You must not add requests to a group from within the group callback
599 (or any later time).
600
601 Their lifetime, simplified, looks like this: when they are empty, they
602 will finish very quickly. If they contain only requests that are in the
603 "done" state, they will also finish. Otherwise they will continue to
604 exist.
605
606 That means after creating a group you have some time to add requests.
607 And in the callbacks of those requests, you can add further requests to
608 the group. And only when all those requests have finished will the the
609 group itself finish.
610
611 add $grp ...
612 $grp->add (...)
613 Add one or more requests to the group. Any type of IO::AIO::REQ can
614 be added, including other groups, as long as you do not create cir‐
615 cular dependencies.
616
617 Returns all its arguments.
618
619 $grp->cancel_subs
620 Cancel all subrequests and clears any feeder, but not the group
621 request itself. Useful when you queued a lot of events but got a
622 result early.
623
624 $grp->result (...)
625 Set the result value(s) that will be passed to the group callback
626 when all subrequests have finished and set thre groups errno to the
627 current value of errno (just like calling "errno" without an error
628 number). By default, no argument will be passed and errno is zero.
629
630 $grp->errno ([$errno])
631 Sets the group errno value to $errno, or the current value of errno
632 when the argument is missing.
633
634 Every aio request has an associated errno value that is restored
635 when the callback is invoked. This method lets you change this
636 value from its default (0).
637
638 Calling "result" will also set errno, so make sure you either set
639 $! before the call to "result", or call c<errno> after it.
640
641 feed $grp $callback->($grp)
642 Sets a feeder/generator on this group: every group can have an
643 attached generator that generates requests if idle. The idea behind
644 this is that, although you could just queue as many requests as you
645 want in a group, this might starve other requests for a potentially
646 long time. For example, "aio_scandir" might generate hundreds of
647 thousands "aio_stat" requests, delaying any later requests for a
648 long time.
649
650 To avoid this, and allow incremental generation of requests, you
651 can instead a group and set a feeder on it that generates those
652 requests. The feed callback will be called whenever there are few
653 enough (see "limit", below) requests active in the group itself and
654 is expected to queue more requests.
655
656 The feed callback can queue as many requests as it likes (i.e.
657 "add" does not impose any limits).
658
659 If the feed does not queue more requests when called, it will be
660 automatically removed from the group.
661
662 If the feed limit is 0, it will be set to 2 automatically.
663
664 Example:
665
666 # stat all files in @files, but only ever use four aio requests concurrently:
667
668 my $grp = aio_group sub { print "finished\n" };
669 limit $grp 4;
670 feed $grp sub {
671 my $file = pop @files
672 or return;
673
674 add $grp aio_stat $file, sub { ... };
675 };
676
677 limit $grp $num
678 Sets the feeder limit for the group: The feeder will be called
679 whenever the group contains less than this many requests.
680
681 Setting the limit to 0 will pause the feeding process.
682
683 SUPPORT FUNCTIONS
684
685 EVENT PROCESSING AND EVENT LOOP INTEGRATION
686
687 $fileno = IO::AIO::poll_fileno
688 Return the request result pipe file descriptor. This filehandle
689 must be polled for reading by some mechanism outside this module
690 (e.g. Event or select, see below or the SYNOPSIS). If the pipe
691 becomes readable you have to call "poll_cb" to check the results.
692
693 See "poll_cb" for an example.
694
695 IO::AIO::poll_cb
696 Process some outstanding events on the result pipe. You have to
697 call this regularly. Returns the number of events processed.
698 Returns immediately when no events are outstanding. The amount of
699 events processed depends on the settings of "IO::AIO::max_poll_req"
700 and "IO::AIO::max_poll_time".
701
702 If not all requests were processed for whatever reason, the file‐
703 handle will still be ready when "poll_cb" returns.
704
705 Example: Install an Event watcher that automatically calls
706 IO::AIO::poll_cb with high priority:
707
708 Event->io (fd => IO::AIO::poll_fileno,
709 poll => 'r', async => 1,
710 cb => \&IO::AIO::poll_cb);
711
712 IO::AIO::max_poll_reqs $nreqs
713 IO::AIO::max_poll_time $seconds
714 These set the maximum number of requests (default 0, meaning infin‐
715 ity) that are being processed by "IO::AIO::poll_cb" in one call,
716 respectively the maximum amount of time (default 0, meaning infin‐
717 ity) spent in "IO::AIO::poll_cb" to process requests (more cor‐
718 rectly the mininum amount of time "poll_cb" is allowed to use).
719
720 Setting "max_poll_time" to a non-zero value creates an overhead of
721 one syscall per request processed, which is not normally a problem
722 unless your callbacks are really really fast or your OS is really
723 really slow (I am not mentioning Solaris here). Using
724 "max_poll_reqs" incurs no overhead.
725
726 Setting these is useful if you want to ensure some level of inter‐
727 activeness when perl is not fast enough to process all requests in
728 time.
729
730 For interactive programs, values such as 0.01 to 0.1 should be
731 fine.
732
733 Example: Install an Event watcher that automatically calls
734 IO::AIO::poll_cb with low priority, to ensure that other parts of
735 the program get the CPU sometimes even under high AIO load.
736
737 # try not to spend much more than 0.1s in poll_cb
738 IO::AIO::max_poll_time 0.1;
739
740 # use a low priority so other tasks have priority
741 Event->io (fd => IO::AIO::poll_fileno,
742 poll => 'r', nice => 1,
743 cb => &IO::AIO::poll_cb);
744
745 IO::AIO::poll_wait
746 If there are any outstanding requests and none of them in the
747 result phase, wait till the result filehandle becomes ready for
748 reading (simply does a "select" on the filehandle. This is useful
749 if you want to synchronously wait for some requests to finish).
750
751 See "nreqs" for an example.
752
753 IO::AIO::poll
754 Waits until some requests have been handled.
755
756 Returns the number of requests processed, but is otherwise strictly
757 equivalent to:
758
759 IO::AIO::poll_wait, IO::AIO::poll_cb
760
761 IO::AIO::flush
762 Wait till all outstanding AIO requests have been handled.
763
764 Strictly equivalent to:
765
766 IO::AIO::poll_wait, IO::AIO::poll_cb
767 while IO::AIO::nreqs;
768
769 CONTROLLING THE NUMBER OF THREADS
770
771 IO::AIO::min_parallel $nthreads
772 Set the minimum number of AIO threads to $nthreads. The current
773 default is 8, which means eight asynchronous operations can execute
774 concurrently at any one time (the number of outstanding requests,
775 however, is unlimited).
776
777 IO::AIO starts threads only on demand, when an AIO request is
778 queued and no free thread exists. Please note that queueing up a
779 hundred requests can create demand for a hundred threads, even if
780 it turns out that everything is in the cache and could have been
781 processed faster by a single thread.
782
783 It is recommended to keep the number of threads relatively low, as
784 some Linux kernel versions will scale negatively with the number of
785 threads (higher parallelity => MUCH higher latency). With current
786 Linux 2.6 versions, 4-32 threads should be fine.
787
788 Under most circumstances you don't need to call this function, as
789 the module selects a default that is suitable for low to moderate
790 load.
791
792 IO::AIO::max_parallel $nthreads
793 Sets the maximum number of AIO threads to $nthreads. If more than
794 the specified number of threads are currently running, this func‐
795 tion kills them. This function blocks until the limit is reached.
796
797 While $nthreads are zero, aio requests get queued but not executed
798 until the number of threads has been increased again.
799
800 This module automatically runs "max_parallel 0" at program end, to
801 ensure that all threads are killed and that there are no outstand‐
802 ing requests.
803
804 Under normal circumstances you don't need to call this function.
805
806 IO::AIO::max_idle $nthreads
807 Limit the number of threads (default: 4) that are allowed to idle
808 (i.e., threads that did not get a request to process within 10 sec‐
809 onds). That means if a thread becomes idle while $nthreads other
810 threads are also idle, it will free its resources and exit.
811
812 This is useful when you allow a large number of threads (e.g. 100
813 or 1000) to allow for extremely high load situations, but want to
814 free resources under normal circumstances (1000 threads can easily
815 consume 30MB of RAM).
816
817 The default is probably ok in most situations, especially if thread
818 creation is fast. If thread creation is very slow on your system
819 you might want to use larger values.
820
821 $oldmaxreqs = IO::AIO::max_outstanding $maxreqs
822 This is a very bad function to use in interactive programs because
823 it blocks, and a bad way to reduce concurrency because it is inex‐
824 act: Better use an "aio_group" together with a feed callback.
825
826 Sets the maximum number of outstanding requests to $nreqs. If you
827 to queue up more than this number of requests, the next call to the
828 "poll_cb" (and "poll_some" and other functions calling "poll_cb")
829 function will block until the limit is no longer exceeded.
830
831 The default value is very large, so there is no practical limit on
832 the number of outstanding requests.
833
834 You can still queue as many requests as you want. Therefore,
835 "max_oustsanding" is mainly useful in simple scripts (with low val‐
836 ues) or as a stop gap to shield against fatal memory overflow (with
837 large values).
838
839 STATISTICAL INFORMATION
840
841 IO::AIO::nreqs
842 Returns the number of requests currently in the ready, execute or
843 pending states (i.e. for which their callback has not been invoked
844 yet).
845
846 Example: wait till there are no outstanding requests anymore:
847
848 IO::AIO::poll_wait, IO::AIO::poll_cb
849 while IO::AIO::nreqs;
850
851 IO::AIO::nready
852 Returns the number of requests currently in the ready state (not
853 yet executed).
854
855 IO::AIO::npending
856 Returns the number of requests currently in the pending state (exe‐
857 cuted, but not yet processed by poll_cb).
858
859 FORK BEHAVIOUR
860
861 This module should do "the right thing" when the process using it
862 forks:
863
864 Before the fork, IO::AIO enters a quiescent state where no requests can
865 be added in other threads and no results will be processed. After the
866 fork the parent simply leaves the quiescent state and continues
867 request/result processing, while the child frees the request/result
868 queue (so that the requests started before the fork will only be han‐
869 dled in the parent). Threads will be started on demand until the limit
870 set in the parent process has been reached again.
871
872 In short: the parent will, after a short pause, continue as if fork had
873 not been called, while the child will act as if IO::AIO has not been
874 used yet.
875
876 MEMORY USAGE
877
878 Per-request usage:
879
880 Each aio request uses - depending on your architecture - around 100-200
881 bytes of memory. In addition, stat requests need a stat buffer (possi‐
882 bly a few hundred bytes), readdir requires a result buffer and so on.
883 Perl scalars and other data passed into aio requests will also be
884 locked and will consume memory till the request has entered the done
885 state.
886
887 This is now awfully much, so queuing lots of requests is not usually a
888 problem.
889
890 Per-thread usage:
891
892 In the execution phase, some aio requests require more memory for tem‐
893 porary buffers, and each thread requires a stack and other data struc‐
894 tures (usually around 16k-128k, depending on the OS).
895
897 Known bugs will be fixed in the next release.
898
900 Coro::AIO.
901
903 Marc Lehmann <schmorp@schmorp.de>
904 http://home.schmorp.de/
905
906
907
908perl v5.8.8 2007-01-23 AIO(3)