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