ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/IO-AIO/README
Revision: 1.22
Committed: Sat Jan 6 02:47:11 2007 UTC (17 years, 10 months ago) by root
Branch: MAIN
CVS Tags: rel-2_31
Changes since 1.21: +4 -0 lines
Log Message:
*** empty log message ***

File Contents

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