ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/IO-AIO/AIO.pm
(Generate patch)

Comparing IO-AIO/AIO.pm (file contents):
Revision 1.51 by root, Sat Jun 24 19:14:04 2006 UTC vs.
Revision 1.73 by root, Tue Oct 24 16:35:04 2006 UTC

15 15
16 aio_read $fh, 30000, 1024, $buffer, 0, sub { 16 aio_read $fh, 30000, 1024, $buffer, 0, sub {
17 $_[0] > 0 or die "read error: $!"; 17 $_[0] > 0 or die "read error: $!";
18 }; 18 };
19 19
20 # AnyEvent 20 # version 2+ has request and group objects
21 use IO::AIO 2;
22
23 aioreq_pri 4; # give next request a very high priority
24 my $req = aio_unlink "/tmp/file", sub { };
25 $req->cancel; # cancel request if still in queue
26
27 my $grp = aio_group sub { print "all stats done\n" };
28 add $grp aio_stat "..." for ...;
29
30 # AnyEvent integration
21 open my $fh, "<&=" . IO::AIO::poll_fileno or die "$!"; 31 open my $fh, "<&=" . IO::AIO::poll_fileno or die "$!";
22 my $w = AnyEvent->io (fh => $fh, poll => 'r', cb => sub { IO::AIO::poll_cb }); 32 my $w = AnyEvent->io (fh => $fh, poll => 'r', cb => sub { IO::AIO::poll_cb });
23 33
24 # Event 34 # Event integration
25 Event->io (fd => IO::AIO::poll_fileno, 35 Event->io (fd => IO::AIO::poll_fileno,
26 poll => 'r', 36 poll => 'r',
27 cb => \&IO::AIO::poll_cb); 37 cb => \&IO::AIO::poll_cb);
28 38
29 # Glib/Gtk2 39 # Glib/Gtk2 integration
30 add_watch Glib::IO IO::AIO::poll_fileno, 40 add_watch Glib::IO IO::AIO::poll_fileno,
31 in => sub { IO::AIO::poll_cb; 1 }; 41 in => sub { IO::AIO::poll_cb; 1 };
32 42
33 # Tk 43 # Tk integration
34 Tk::Event::IO->fileevent (IO::AIO::poll_fileno, "", 44 Tk::Event::IO->fileevent (IO::AIO::poll_fileno, "",
35 readable => \&IO::AIO::poll_cb); 45 readable => \&IO::AIO::poll_cb);
36 46
37 # Danga::Socket 47 # Danga::Socket integration
38 Danga::Socket->AddOtherFds (IO::AIO::poll_fileno => 48 Danga::Socket->AddOtherFds (IO::AIO::poll_fileno =>
39 \&IO::AIO::poll_cb); 49 \&IO::AIO::poll_cb);
40 50
41
42=head1 DESCRIPTION 51=head1 DESCRIPTION
43 52
44This module implements asynchronous I/O using whatever means your 53This module implements asynchronous I/O using whatever means your
45operating system supports. 54operating system supports.
46 55
47Currently, a number of threads are started that execute your read/writes 56In this version, a number of threads are started that execute your
48and signal their completion. You don't need thread support in your libc or 57requests and signal their completion. You don't need thread support
49perl, and the threads created by this module will not be visible to the 58in perl, and the threads created by this module will not be visible
50pthreads library. In the future, this module might make use of the native 59to perl. In the future, this module might make use of the native aio
51aio functions available on many operating systems. However, they are often 60functions available on many operating systems. However, they are often
52not well-supported (Linux doesn't allow them on normal files currently, 61not well-supported or restricted (Linux doesn't allow them on normal
53for example), and they would only support aio_read and aio_write, so the 62files currently, for example), and they would only support aio_read and
54remaining functionality would have to be implemented using threads anyway. 63aio_write, so the remaining functionality would have to be implemented
64using threads anyway.
55 65
56Although the module will work with in the presence of other threads, it is 66Although the module will work with in the presence of other (Perl-)
57currently not reentrant, so use appropriate locking yourself, always call 67threads, it is currently not reentrant in any way, so use appropriate
58C<poll_cb> from within the same thread, or never call C<poll_cb> (or other 68locking yourself, always call C<poll_cb> from within the same thread, or
59C<aio_> functions) recursively. 69never call C<poll_cb> (or other C<aio_> functions) recursively.
70
71=head1 REQUEST ANATOMY AND LIFETIME
72
73Every C<aio_*> function creates a request. which is a C data structure not
74directly visible to Perl.
75
76If called in non-void context, every request function returns a Perl
77object representing the request. In void context, nothing is returned,
78which saves a bit of memory.
79
80The perl object is a fairly standard ref-to-hash object. The hash contents
81are not used by IO::AIO so you are free to store anything you like in it.
82
83During their existance, aio requests travel through the following states,
84in order:
85
86=over 4
87
88=item ready
89
90Immediately after a request is created it is put into the ready state,
91waiting for a thread to execute it.
92
93=item execute
94
95A thread has accepted the request for processing and is currently
96executing it (e.g. blocking in read).
97
98=item pending
99
100The request has been executed and is waiting for result processing.
101
102While request submission and execution is fully asynchronous, result
103processing is not and relies on the perl interpreter calling C<poll_cb>
104(or another function with the same effect).
105
106=item result
107
108The request results are processed synchronously by C<poll_cb>.
109
110The C<poll_cb> function will process all outstanding aio requests by
111calling their callbacks, freeing memory associated with them and managing
112any groups they are contained in.
113
114=item done
115
116Request has reached the end of its lifetime and holds no resources anymore
117(except possibly for the Perl object, but its connection to the actual
118aio request is severed and calling its methods will either do nothing or
119result in a runtime error).
60 120
61=cut 121=cut
62 122
63package IO::AIO; 123package IO::AIO;
64 124
66use strict 'vars'; 126use strict 'vars';
67 127
68use base 'Exporter'; 128use base 'Exporter';
69 129
70BEGIN { 130BEGIN {
71 our $VERSION = '1.8'; 131 our $VERSION = '2.0';
72 132
73 our @EXPORT = qw(aio_sendfile aio_read aio_write aio_open aio_close aio_stat 133 our @AIO_REQ = qw(aio_sendfile aio_read aio_write aio_open aio_close aio_stat
74 aio_lstat aio_unlink aio_rmdir aio_readdir aio_scandir aio_symlink 134 aio_lstat aio_unlink aio_rmdir aio_readdir aio_scandir aio_symlink
75 aio_fsync aio_fdatasync aio_readahead aio_rename aio_link aio_move); 135 aio_fsync aio_fdatasync aio_readahead aio_rename aio_link aio_move
76 our @EXPORT_OK = qw(poll_fileno poll_cb min_parallel max_parallel max_outstanding nreqs); 136 aio_group aio_nop);
137 our @EXPORT = (@AIO_REQ, qw(aioreq_pri aioreq_nice));
138 our @EXPORT_OK = qw(poll_fileno poll_cb poll_wait flush
139 min_parallel max_parallel max_outstanding nreqs);
140
141 @IO::AIO::GRP::ISA = 'IO::AIO::REQ';
77 142
78 require XSLoader; 143 require XSLoader;
79 XSLoader::load ("IO::AIO", $VERSION); 144 XSLoader::load ("IO::AIO", $VERSION);
80} 145}
81 146
92syscall has been executed asynchronously. 157syscall has been executed asynchronously.
93 158
94All functions expecting a filehandle keep a copy of the filehandle 159All functions expecting a filehandle keep a copy of the filehandle
95internally until the request has finished. 160internally until the request has finished.
96 161
162All requests return objects of type L<IO::AIO::REQ> that allow further
163manipulation of those requests while they are in-flight.
164
97The pathnames you pass to these routines I<must> be absolute and 165The pathnames you pass to these routines I<must> be absolute and
98encoded in byte form. The reason for the former is that at the time the 166encoded in byte form. The reason for the former is that at the time the
99request is being executed, the current working directory could have 167request is being executed, the current working directory could have
100changed. Alternatively, you can make sure that you never change the 168changed. Alternatively, you can make sure that you never change the
101current working directory. 169current working directory.
106your pathnames to the locale (or other) encoding in effect in the user 174your pathnames to the locale (or other) encoding in effect in the user
107environment, d) use Glib::filename_from_unicode on unicode filenames or e) 175environment, d) use Glib::filename_from_unicode on unicode filenames or e)
108use something else. 176use something else.
109 177
110=over 4 178=over 4
179
180=item aioreq_pri $pri
181
182Sets the priority for the next aio request. The default priority
183is C<0>, the minimum and maximum priorities are C<-4> and C<4>,
184respectively. Requests with higher priority will be serviced first.
185
186The priority will be reset to C<0> after each call to one of the C<aio_>
187functions.
188
189Example: open a file with low priority, then read something from it with
190higher priority so the read request is serviced before other low priority
191open requests (potentially spamming the cache):
192
193 aioreq_pri -3;
194 aio_open ..., sub {
195 return unless $_[0];
196
197 aioreq_pri -2;
198 aio_read $_[0], ..., sub {
199 ...
200 };
201 };
202
203=item aioreq_nice $pri_adjust
204
205Similar to C<aioreq_pri>, but subtracts the given value from the current
206priority, so effects are cumulative.
111 207
112=item aio_open $pathname, $flags, $mode, $callback->($fh) 208=item aio_open $pathname, $flags, $mode, $callback->($fh)
113 209
114Asynchronously open or create a file and call the callback with a newly 210Asynchronously open or create a file and call the callback with a newly
115created filehandle for the file. 211created filehandle for the file.
168 print "read $_[0] bytes: <$buffer>\n"; 264 print "read $_[0] bytes: <$buffer>\n";
169 }; 265 };
170 266
171=item aio_move $srcpath, $dstpath, $callback->($status) 267=item aio_move $srcpath, $dstpath, $callback->($status)
172 268
173[EXPERIMENTAL]
174
175Try to move the I<file> (directories not supported as either source or destination) 269Try to move the I<file> (directories not supported as either source or
176from C<$srcpath> to C<$dstpath> and call the callback with the C<0> (error) or C<-1> ok. 270destination) from C<$srcpath> to C<$dstpath> and call the callback with
271the C<0> (error) or C<-1> ok.
177 272
178This is a composite request that tries to rename(2) the file first. If 273This is a composite request that tries to rename(2) the file first. If
179rename files with C<EXDEV>, it creates the destination file with mode 0200 274rename files with C<EXDEV>, it creates the destination file with mode 0200
180and copies the contents of the source file into it using C<aio_sendfile>, 275and copies the contents of the source file into it using C<aio_sendfile>,
181followed by restoring atime, mtime, access mode and uid/gid, in that 276followed by restoring atime, mtime, access mode and uid/gid, in that
188=cut 283=cut
189 284
190sub aio_move($$$) { 285sub aio_move($$$) {
191 my ($src, $dst, $cb) = @_; 286 my ($src, $dst, $cb) = @_;
192 287
288 my $grp = aio_group $cb;
289
193 aio_rename $src, $dst, sub { 290 add $grp aio_rename $src, $dst, sub {
194 if ($_[0] && $! == EXDEV) { 291 if ($_[0] && $! == EXDEV) {
195 aio_open $src, O_RDONLY, 0, sub { 292 add $grp aio_open $src, O_RDONLY, 0, sub {
196 if (my $src_fh = $_[0]) { 293 if (my $src_fh = $_[0]) {
197 my @stat = stat $src_fh; 294 my @stat = stat $src_fh;
198 295
199 aio_open $dst, O_WRONLY, 0200, sub { 296 add $grp aio_open $dst, O_WRONLY, 0200, sub {
200 if (my $dst_fh = $_[0]) { 297 if (my $dst_fh = $_[0]) {
201 aio_sendfile $dst_fh, $src_fh, 0, $stat[7], sub { 298 add $grp aio_sendfile $dst_fh, $src_fh, 0, $stat[7], sub {
202 close $src_fh; 299 close $src_fh;
203 300
204 if ($_[0] == $stat[7]) { 301 if ($_[0] == $stat[7]) {
205 utime $stat[8], $stat[9], $dst; 302 utime $stat[8], $stat[9], $dst;
206 chmod $stat[2] & 07777, $dst_fh; 303 chmod $stat[2] & 07777, $dst_fh;
207 chown $stat[4], $stat[5], $dst_fh; 304 chown $stat[4], $stat[5], $dst_fh;
208 close $dst_fh; 305 close $dst_fh;
209 306
210 aio_unlink $src, sub { 307 add $grp aio_unlink $src, sub {
211 $cb->($_[0]); 308 $grp->result ($_[0]);
212 }; 309 };
213 } else { 310 } else {
214 my $errno = $!; 311 my $errno = $!;
215 aio_unlink $dst, sub { 312 add $grp aio_unlink $dst, sub {
216 $! = $errno; 313 $! = $errno;
217 $cb->(-1); 314 $grp->result (-1);
218 }; 315 };
219 } 316 }
220 }; 317 };
221 } else { 318 } else {
222 $cb->(-1); 319 $grp->result (-1);
223 } 320 }
224 }, 321 },
225 322
226 } else { 323 } else {
227 $cb->(-1); 324 $grp->result (-1);
228 } 325 }
229 }; 326 };
230 } else { 327 } else {
231 $cb->($_[0]); 328 $grp->result ($_[0]);
232 } 329 }
233 }; 330 };
331
332 $grp
234} 333}
235 334
236=item aio_sendfile $out_fh, $in_fh, $in_offset, $length, $callback->($retval) 335=item aio_sendfile $out_fh, $in_fh, $in_offset, $length, $callback->($retval)
237 336
238Tries to copy C<$length> bytes from C<$in_fh> to C<$out_fh>. It starts 337Tries to copy C<$length> bytes from C<$in_fh> to C<$out_fh>. It starts
326The callback a single argument which is either C<undef> or an array-ref 425The callback a single argument which is either C<undef> or an array-ref
327with the filenames. 426with the filenames.
328 427
329=item aio_scandir $path, $maxreq, $callback->($dirs, $nondirs) 428=item aio_scandir $path, $maxreq, $callback->($dirs, $nondirs)
330 429
331Scans a directory (similar to C<aio_readdir>) and tries to separate the 430Scans a directory (similar to C<aio_readdir>) but additionally tries to
332entries of directory C<$path> into two sets of names, ones you can recurse 431separate the entries of directory C<$path> into two sets of names, ones
333into (directories), and ones you cannot recurse into (everything else). 432you can recurse into (directories or links to them), and ones you cannot
433recurse into (everything else).
334 434
335C<aio_scandir> is a composite request that consists of many 435C<aio_scandir> is a composite request that creates of many sub requests_
336aio-primitives. C<$maxreq> specifies the maximum number of outstanding 436C<$maxreq> specifies the maximum number of outstanding aio requests that
337aio requests that this function generates. If it is C<< <= 0 >>, then a 437this function generates. If it is C<< <= 0 >>, then a suitable default
338suitable default will be chosen (currently 8). 438will be chosen (currently 6).
339 439
340On error, the callback is called without arguments, otherwise it receives 440On error, the callback is called without arguments, otherwise it receives
341two array-refs with path-relative entry names. 441two array-refs with path-relative entry names.
342 442
343Example: 443Example:
351Implementation notes. 451Implementation notes.
352 452
353The C<aio_readdir> cannot be avoided, but C<stat()>'ing every entry can. 453The C<aio_readdir> cannot be avoided, but C<stat()>'ing every entry can.
354 454
355After reading the directory, the modification time, size etc. of the 455After reading the directory, the modification time, size etc. of the
356directory before and after the readdir is checked, and if they match, the 456directory before and after the readdir is checked, and if they match (and
357link count will be used to decide how many entries are directories (if 457isn't the current time), the link count will be used to decide how many
358>= 2). Otherwise, no knowledge of the number of subdirectories will be 458entries are directories (if >= 2). Otherwise, no knowledge of the number
359assumed. 459of subdirectories will be assumed.
360 460
361Then entires will be sorted into likely directories (everything without a 461Then entries will be sorted into likely directories (everything without
362non-initial dot) and likely non-directories (everything else). Then every 462a non-initial dot currently) and likely non-directories (everything
363entry + C</.> will be C<stat>'ed, likely directories first. This is often 463else). Then every entry plus an appended C</.> will be C<stat>'ed,
464likely directories first. If that succeeds, it assumes that the entry
465is a directory or a symlink to directory (which will be checked
466seperately). This is often faster than stat'ing the entry itself because
364faster because filesystems might detect the type of the entry without 467filesystems might detect the type of the entry without reading the inode
365reading the inode data (e.g. ext2fs filetype feature). If that succeeds, 468data (e.g. ext2fs filetype feature).
366it assumes that the entry is a directory or a symlink to directory (which
367will be checked seperately).
368 469
369If the known number of directories has been reached, the rest of the 470If the known number of directories (link count - 2) has been reached, the
370entries is assumed to be non-directories. 471rest of the entries is assumed to be non-directories.
472
473This only works with certainty on POSIX (= UNIX) filesystems, which
474fortunately are the vast majority of filesystems around.
475
476It will also likely work on non-POSIX filesystems with reduced efficiency
477as those tend to return 0 or 1 as link counts, which disables the
478directory counting heuristic.
371 479
372=cut 480=cut
373 481
374sub aio_scandir($$$) { 482sub aio_scandir($$$) {
375 my ($path, $maxreq, $cb) = @_; 483 my ($path, $maxreq, $cb) = @_;
376 484
485 my $grp = aio_group $cb;
486
377 $maxreq = 8 if $maxreq <= 0; 487 $maxreq = 6 if $maxreq <= 0;
378 488
379 # stat once 489 # stat once
380 aio_stat $path, sub { 490 add $grp aio_stat $path, sub {
381 return $cb->() if $_[0]; 491 return $grp->result () if $_[0];
492 my $now = time;
382 my $hash1 = join ":", (stat _)[0,1,3,7,9]; 493 my $hash1 = join ":", (stat _)[0,1,3,7,9];
383 494
384 # read the directory entries 495 # read the directory entries
385 aio_readdir $path, sub { 496 add $grp aio_readdir $path, sub {
386 my $entries = shift 497 my $entries = shift
387 or return $cb->(); 498 or return $grp->result ();
388 499
389 # stat the dir another time 500 # stat the dir another time
390 aio_stat $path, sub { 501 add $grp aio_stat $path, sub {
391 my $hash2 = join ":", (stat _)[0,1,3,7,9]; 502 my $hash2 = join ":", (stat _)[0,1,3,7,9];
392 503
393 my $ndirs; 504 my $ndirs;
394 505
395 # take the slow route if anything looks fishy 506 # take the slow route if anything looks fishy
396 if ($hash1 ne $hash2) { 507 if ($hash1 ne $hash2 or (stat _)[9] == $now) {
397 $ndirs = -1; 508 $ndirs = -1;
398 } else { 509 } else {
399 # if nlink == 2, we are finished 510 # if nlink == 2, we are finished
400 # on non-posix-fs's, we rely on nlink < 2 511 # on non-posix-fs's, we rely on nlink < 2
401 $ndirs = (stat _)[3] - 2 512 $ndirs = (stat _)[3] - 2
402 or return $cb->([], $entries); 513 or return $grp->result ([], $entries);
403 } 514 }
404 515
405 # sort into likely dirs and likely nondirs 516 # sort into likely dirs and likely nondirs
406 # dirs == files without ".", short entries first 517 # dirs == files without ".", short entries first
407 $entries = [map $_->[0], 518 $entries = [map $_->[0],
412 my (@dirs, @nondirs); 523 my (@dirs, @nondirs);
413 524
414 my ($statcb, $schedcb); 525 my ($statcb, $schedcb);
415 my $nreq = 0; 526 my $nreq = 0;
416 527
528 my $statgrp = add $grp aio_group;
529
417 $schedcb = sub { 530 $schedcb = sub {
418 if (@$entries) { 531 if (@$entries) {
419 if ($nreq < $maxreq) { 532 if ($nreq < $maxreq) {
420 my $ent = pop @$entries; 533 my $ent = pop @$entries;
421 $nreq++; 534 $nreq++;
422 aio_stat "$path/$ent/.", sub { $statcb->($_[0], $ent) }; 535 add $statgrp aio_stat "$path/$ent/.", sub { $statcb->($_[0], $ent) };
423 } 536 }
424 } elsif (!$nreq) { 537 } elsif (!$nreq) {
425 # finished 538 # finished
539 $statgrp->cancel;
426 undef $statcb; 540 undef $statcb;
427 undef $schedcb; 541 undef $schedcb;
428 $cb->(\@dirs, \@nondirs) if $cb; 542 $grp->result (\@dirs, \@nondirs);
429 undef $cb;
430 } 543 }
431 }; 544 };
432 $statcb = sub { 545 $statcb = sub {
433 my ($status, $entry) = @_; 546 my ($status, $entry) = @_;
434 547
436 $nreq--; 549 $nreq--;
437 push @nondirs, $entry; 550 push @nondirs, $entry;
438 &$schedcb; 551 &$schedcb;
439 } else { 552 } else {
440 # need to check for real directory 553 # need to check for real directory
441 aio_lstat "$path/$entry", sub { 554 add $grp aio_lstat "$path/$entry", sub {
442 $nreq--; 555 $nreq--;
443 556
444 if (-d _) { 557 if (-d _) {
445 push @dirs, $entry; 558 push @dirs, $entry;
446 559
459 572
460 &$schedcb while @$entries && $nreq < $maxreq; 573 &$schedcb while @$entries && $nreq < $maxreq;
461 }; 574 };
462 }; 575 };
463 }; 576 };
577
578 $grp
464} 579}
465 580
466=item aio_fsync $fh, $callback->($status) 581=item aio_fsync $fh, $callback->($status)
467 582
468Asynchronously call fsync on the given filehandle and call the callback 583Asynchronously call fsync on the given filehandle and call the callback
473Asynchronously call fdatasync on the given filehandle and call the 588Asynchronously call fdatasync on the given filehandle and call the
474callback with the fdatasync result code. 589callback with the fdatasync result code.
475 590
476If this call isn't available because your OS lacks it or it couldn't be 591If this call isn't available because your OS lacks it or it couldn't be
477detected, it will be emulated by calling C<fsync> instead. 592detected, it will be emulated by calling C<fsync> instead.
593
594=item aio_group $callback->(...)
595
596[EXPERIMENTAL]
597
598This is a very special aio request: Instead of doing something, it is a
599container for other aio requests, which is useful if you want to bundle
600many requests into a single, composite, request with a definite callback
601and the ability to cancel the whole request with its subrequests.
602
603Returns an object of class L<IO::AIO::GRP>. See its documentation below
604for more info.
605
606Example:
607
608 my $grp = aio_group sub {
609 print "all stats done\n";
610 };
611
612 add $grp
613 (aio_stat ...),
614 (aio_stat ...),
615 ...;
616
617=item aio_nop $callback->()
618
619This is a special request - it does nothing in itself and is only used for
620side effects, such as when you want to add a dummy request to a group so
621that finishing the requests in the group depends on executing the given
622code.
623
624While this request does nothing, it still goes through the execution
625phase and still requires a worker thread. Thus, the callback will not
626be executed immediately but only after other requests in the queue have
627entered their execution phase. This can be used to measure request
628latency.
629
630=item IO::AIO::aio_busy $fractional_seconds, $callback->() *NOT EXPORTED*
631
632Mainly used for debugging and benchmarking, this aio request puts one of
633the request workers to sleep for the given time.
634
635While it is theoretically handy to have simple I/O scheduling requests
636like sleep and file handle readable/writable, the overhead this creates is
637immense (it blocks a thread for a long time) so do not use this function
638except to put your application under artificial I/O pressure.
639
640=back
641
642=head2 IO::AIO::REQ CLASS
643
644All non-aggregate C<aio_*> functions return an object of this class when
645called in non-void context.
646
647=over 4
648
649=item cancel $req
650
651Cancels the request, if possible. Has the effect of skipping execution
652when entering the B<execute> state and skipping calling the callback when
653entering the the B<result> state, but will leave the request otherwise
654untouched. That means that requests that currently execute will not be
655stopped and resources held by the request will not be freed prematurely.
656
657=item cb $req $callback->(...)
658
659Replace (or simply set) the callback registered to the request.
660
661=back
662
663=head2 IO::AIO::GRP CLASS
664
665This class is a subclass of L<IO::AIO::REQ>, so all its methods apply to
666objects of this class, too.
667
668A IO::AIO::GRP object is a special request that can contain multiple other
669aio requests.
670
671You create one by calling the C<aio_group> constructing function with a
672callback that will be called when all contained requests have entered the
673C<done> state:
674
675 my $grp = aio_group sub {
676 print "all requests are done\n";
677 };
678
679You add requests by calling the C<add> method with one or more
680C<IO::AIO::REQ> objects:
681
682 $grp->add (aio_unlink "...");
683
684 add $grp aio_stat "...", sub {
685 $_[0] or return $grp->result ("error");
686
687 # add another request dynamically, if first succeeded
688 add $grp aio_open "...", sub {
689 $grp->result ("ok");
690 };
691 };
692
693This makes it very easy to create composite requests (see the source of
694C<aio_move> for an application) that work and feel like simple requests.
695
696=over 4
697
698=item * The IO::AIO::GRP objects will be cleaned up during calls to
699C<IO::AIO::poll_cb>, just like any other request.
700
701=item * They can be canceled like any other request. Canceling will cancel not
702only the request itself, but also all requests it contains.
703
704=item * They can also can also be added to other IO::AIO::GRP objects.
705
706=item * You must not add requests to a group from within the group callback (or
707any later time).
708
709=item * This does not harmonise well with C<max_outstanding>, so best do
710not combine C<aio_group> with it. Groups and feeders are recommended for
711this kind of concurrency-limiting.
712
713=back
714
715Their lifetime, simplified, looks like this: when they are empty, they
716will finish very quickly. If they contain only requests that are in the
717C<done> state, they will also finish. Otherwise they will continue to
718exist.
719
720That means after creating a group you have some time to add requests. And
721in the callbacks of those requests, you can add further requests to the
722group. And only when all those requests have finished will the the group
723itself finish.
724
725=over 4
726
727=item add $grp ...
728
729=item $grp->add (...)
730
731Add one or more requests to the group. Any type of L<IO::AIO::REQ> can
732be added, including other groups, as long as you do not create circular
733dependencies.
734
735Returns all its arguments.
736
737=item $grp->result (...)
738
739Set the result value(s) that will be passed to the group callback when all
740subrequests have finished. By default, no argument will be passed.
741
742=item feed $grp $callback->($grp)
743
744[VERY EXPERIMENTAL]
745
746Sets a feeder/generator on this group: every group can have an attached
747generator that generates requests if idle. The idea behind this is that,
748although you could just queue as many requests as you want in a group,
749this might starve other requests for a potentially long time. For
750example, C<aio_scandir> might generate hundreds of thousands C<aio_stat>
751requests, delaying any later requests for a long time.
752
753To avoid this, and allow incremental generation of requests, you can
754instead a group and set a feeder on it that generates those requests. The
755feed callback will be called whenever there are few enough (see C<limit>,
756below) requests active in the group itself and is expected to queue more
757requests.
758
759The feed callback can queue as many requests as it likes (i.e. C<add> does
760not impose any limits).
761
762If the feed does not queue more requests when called, it will be
763automatically removed from the group.
764
765If the feed limit is C<0>, it will be set to C<2> automatically.
766
767Example:
768
769 # stat all files in @files, but only ever use four aio requests concurrently:
770
771 my $grp = aio_group sub { print "finished\n" };
772 limit $grp 4;
773 feed $grp sub {
774 my $file = pop @files
775 or return;
776
777 add $grp aio_stat $file, sub { ... };
778 };
779
780=item limit $grp $num
781
782Sets the feeder limit for the group: The feeder will be called whenever
783the group contains less than this many requests.
784
785Setting the limit to C<0> will pause the feeding process.
478 786
479=back 787=back
480 788
481=head2 SUPPORT FUNCTIONS 789=head2 SUPPORT FUNCTIONS
482 790
540 IO::AIO::poll_wait, IO::AIO::poll_cb 848 IO::AIO::poll_wait, IO::AIO::poll_cb
541 if IO::AIO::nreqs; 849 if IO::AIO::nreqs;
542 850
543=item IO::AIO::min_parallel $nthreads 851=item IO::AIO::min_parallel $nthreads
544 852
545Set the minimum number of AIO threads to C<$nthreads>. The current default 853Set the minimum number of AIO threads to C<$nthreads>. The current
546is C<4>, which means four asynchronous operations can be done at one time 854default is C<8>, which means eight asynchronous operations can execute
547(the number of outstanding operations, however, is unlimited). 855concurrently at any one time (the number of outstanding requests,
856however, is unlimited).
548 857
549IO::AIO starts threads only on demand, when an AIO request is queued and 858IO::AIO starts threads only on demand, when an AIO request is queued and
550no free thread exists. 859no free thread exists.
551 860
552It is recommended to keep the number of threads low, as some Linux 861It is recommended to keep the number of threads relatively low, as some
553kernel versions will scale negatively with the number of threads (higher 862Linux kernel versions will scale negatively with the number of threads
554parallelity => MUCH higher latency). With current Linux 2.6 versions, 4-32 863(higher parallelity => MUCH higher latency). With current Linux 2.6
555threads should be fine. 864versions, 4-32 threads should be fine.
556 865
557Under most circumstances you don't need to call this function, as the 866Under most circumstances you don't need to call this function, as the
558module selects a default that is suitable for low to moderate load. 867module selects a default that is suitable for low to moderate load.
559 868
560=item IO::AIO::max_parallel $nthreads 869=item IO::AIO::max_parallel $nthreads
571 880
572Under normal circumstances you don't need to call this function. 881Under normal circumstances you don't need to call this function.
573 882
574=item $oldnreqs = IO::AIO::max_outstanding $nreqs 883=item $oldnreqs = IO::AIO::max_outstanding $nreqs
575 884
885[DEPRECATED]
886
576Sets the maximum number of outstanding requests to C<$nreqs>. If you 887Sets the maximum number of outstanding requests to C<$nreqs>. If you
577try to queue up more than this number of requests, the caller will block until 888try to queue up more than this number of requests, the caller will block until
578some requests have been handled. 889some requests have been handled.
579 890
580The default is very large, so normally there is no practical limit. If you 891The default is very large, so normally there is no practical limit. If you
581queue up many requests in a loop it often improves speed if you set 892queue up many requests in a loop it often improves speed if you set
582this to a relatively low number, such as C<100>. 893this to a relatively low number, such as C<100>.
894
895This function does not work well together with C<aio_group>'s, and their
896feeder interface is better suited to limiting concurrency, so do not use
897this function.
583 898
584Under normal circumstances you don't need to call this function. 899Under normal circumstances you don't need to call this function.
585 900
586=back 901=back
587 902
601 or return undef; 916 or return undef;
602 917
603 *$sym 918 *$sym
604} 919}
605 920
606min_parallel 4; 921min_parallel 8;
607 922
608END { 923END {
609 max_parallel 0; 924 max_parallel 0;
610} 925}
611 926
6121; 9271;
613 928
614=head2 FORK BEHAVIOUR 929=head2 FORK BEHAVIOUR
615 930
931This module should do "the right thing" when the process using it forks:
932
616Before the fork, IO::AIO enters a quiescent state where no requests 933Before the fork, IO::AIO enters a quiescent state where no requests
617can be added in other threads and no results will be processed. After 934can be added in other threads and no results will be processed. After
618the fork the parent simply leaves the quiescent state and continues 935the fork the parent simply leaves the quiescent state and continues
619request/result processing, while the child clears the request/result 936request/result processing, while the child frees the request/result queue
620queue (so the requests started before the fork will only be handled in 937(so that the requests started before the fork will only be handled in the
621the parent). Threats will be started on demand until the limit ste in the 938parent). Threads will be started on demand until the limit set in the
622parent process has been reached again. 939parent process has been reached again.
623 940
941In short: the parent will, after a short pause, continue as if fork had
942not been called, while the child will act as if IO::AIO has not been used
943yet.
944
945=head2 MEMORY USAGE
946
947Per-request usage:
948
949Each aio request uses - depending on your architecture - around 100-200
950bytes of memory. In addition, stat requests need a stat buffer (possibly
951a few hundred bytes), readdir requires a result buffer and so on. Perl
952scalars and other data passed into aio requests will also be locked and
953will consume memory till the request has entered the done state.
954
955This is now awfully much, so queuing lots of requests is not usually a
956problem.
957
958Per-thread usage:
959
960In the execution phase, some aio requests require more memory for
961temporary buffers, and each thread requires a stack and other data
962structures (usually around 16k-128k, depending on the OS).
963
964=head1 KNOWN BUGS
965
966Known bugs will be fixed in the next release.
967
624=head1 SEE ALSO 968=head1 SEE ALSO
625 969
626L<Coro>, L<Linux::AIO>. 970L<Coro::AIO>.
627 971
628=head1 AUTHOR 972=head1 AUTHOR
629 973
630 Marc Lehmann <schmorp@schmorp.de> 974 Marc Lehmann <schmorp@schmorp.de>
631 http://home.schmorp.de/ 975 http://home.schmorp.de/

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines