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 1.27 by root, Tue Aug 16 22:22:18 2005 UTC vs.
Revision 1.35 by root, Mon Aug 22 23:20:37 2005 UTC

… …
63use base 'Exporter'; 63use base 'Exporter';
64 64
65use Fcntl (); 65use Fcntl ();
66 66
67BEGIN { 67BEGIN {
68 $VERSION = 1.1; 68 $VERSION = 1.5;
69 69
70 @EXPORT = qw(aio_read aio_write aio_open aio_close aio_stat aio_lstat aio_unlink 70 @EXPORT = qw(aio_read aio_write aio_open aio_close aio_stat aio_lstat aio_unlink
71 aio_rmdir aio_symlink aio_fsync aio_fdatasync aio_readahead); 71 aio_rmdir aio_symlink aio_fsync aio_fdatasync aio_readahead);
72 @EXPORT_OK = qw(poll_fileno poll_cb min_parallel max_parallel max_outstanding nreqs); 72 @EXPORT_OK = qw(poll_fileno poll_cb min_parallel max_parallel max_outstanding nreqs);
73 73
… …
88syscall has been executed asynchronously. 88syscall has been executed asynchronously.
89 89
90All functions expecting a filehandle keep a copy of the filehandle 90All functions expecting a filehandle keep a copy of the filehandle
91internally until the request has finished. 91internally until the request has finished.
92 92
93The filenames you pass to these routines I<must> be absolute. The reason 93The pathnames you pass to these routines I<must> be absolute and
94for this is that at the time the request is being executed, the current 94encoded in byte form. The reason for the former is that at the time the
95working directory could have changed. Alternatively, you can make sure 95request is being executed, the current working directory could have
96that you never change the current working directory. 96changed. Alternatively, you can make sure that you never change the
97current working directory.
98
99To encode pathnames to byte form, either make sure you either: a)
100always pass in filenames you got from outside (command line, readdir
101etc.), b) are ASCII or ISO 8859-1, c) use the Encode module and encode
102your pathnames to the locale (or other) encoding in effect in the user
103environment, d) use Glib::filename_from_unicode on unicode filenames or e)
104use something else.
97 105
98=over 4 106=over 4
99 107
100=item aio_open $pathname, $flags, $mode, $callback 108=item aio_open $pathname, $flags, $mode, $callback
101 109
… …
142Reads or writes C<length> bytes from the specified C<fh> and C<offset> 150Reads or writes C<length> bytes from the specified C<fh> and C<offset>
143into the scalar given by C<data> and offset C<dataoffset> and calls the 151into the scalar given by C<data> and offset C<dataoffset> and calls the
144callback without the actual number of bytes read (or -1 on error, just 152callback without the actual number of bytes read (or -1 on error, just
145like the syscall). 153like the syscall).
146 154
155The C<$data> scalar I<MUST NOT> be modified in any way while the request
156is outstanding. Modifying it can result in segfaults or WW3 (if the
157necessary/optional hardware is installed).
158
147Example: Read 15 bytes at offset 7 into scalar C<$buffer>, starting at 159Example: Read 15 bytes at offset 7 into scalar C<$buffer>, starting at
148offset C<0> within the scalar: 160offset C<0> within the scalar:
149 161
150 aio_read $fh, 7, 15, $buffer, 0, sub { 162 aio_read $fh, 7, 15, $buffer, 0, sub {
151 $_[0] > 0 or die "read error: $!"; 163 $_[0] > 0 or die "read error: $!";
152 print "read $_[0] bytes: <$buffer>\n"; 164 print "read $_[0] bytes: <$buffer>\n";
153 }; 165 };
166
167=item aio_sendfile $out_fh, $in_fh, $in_offset, $length, $callback
168
169Tries to copy C<$length> bytes from C<$in_fh> to C<$out_fh>. It starts
170reading at byte offset C<$in_offset>, and starts writing at the current
171file offset of C<$out_fh>. Because of that, it is not safe to issue more
172than one C<aio_sendfile> per C<$out_fh>, as they will interfere with each
173other.
174
175This call tries to make use of a native C<sendfile> syscall to provide
176zero-copy operation. For this to work, C<$out_fh> should refer to a
177socket, and C<$in_fh> should refer to mmap'able file.
178
179If the native sendfile call fails or is not implemented, it will be
180emulated, so you can call C<aio_sendfile> on any filehandles regardless of
181the limitations of the OS.
182
183Please note, however, that C<aio_sendfile> can read more bytes from
184C<$in_fh> than are written, and there is no way to find out how many
185bytes have been read form C<aio_sendfile> alone, as C<aio_sendfile> only
186provides the number of bytes written to C<$out_fh>. Only if the return
187value (the value provided to the callback) equals C<$length> one can
188assume that C<$length> bytes have been read.
154 189
155=item aio_readahead $fh,$offset,$length, $callback 190=item aio_readahead $fh,$offset,$length, $callback
156 191
157C<aio_readahead> populates the page cache with data from a file so that 192C<aio_readahead> populates the page cache with data from a file so that
158subsequent reads from that file will not block on disk I/O. The C<$offset> 193subsequent reads from that file will not block on disk I/O. The C<$offset>
… …
275 IO::AIO::poll_wait, IO::AIO::poll_cb 310 IO::AIO::poll_wait, IO::AIO::poll_cb
276 if IO::AIO::nreqs; 311 if IO::AIO::nreqs;
277 312
278=item IO::AIO::min_parallel $nthreads 313=item IO::AIO::min_parallel $nthreads
279 314
280Set the minimum number of AIO threads to C<$nthreads>. The default is 315Set the minimum number of AIO threads to C<$nthreads>. The current default
281C<1>, which means a single asynchronous operation can be done at one time 316is C<4>, which means four asynchronous operations can be done at one time
282(the number of outstanding operations, however, is unlimited). 317(the number of outstanding operations, however, is unlimited).
318
319IO::AIO starts threads only on demand, when an AIO request is queued and
320no free thread exists.
283 321
284It is recommended to keep the number of threads low, as some Linux 322It is recommended to keep the number of threads low, as some Linux
285kernel versions will scale negatively with the number of threads (higher 323kernel versions will scale negatively with the number of threads (higher
286parallelity => MUCH higher latency). With current Linux 2.6 versions, 4-32 324parallelity => MUCH higher latency). With current Linux 2.6 versions, 4-32
287threads should be fine. 325threads should be fine.
288 326
289Under normal circumstances you don't need to call this function, as this 327Under most circumstances you don't need to call this function, as the
290module automatically starts some threads (the exact number might change, 328module selects a default that is suitable for low to moderate load.
291and is currently 4).
292 329
293=item IO::AIO::max_parallel $nthreads 330=item IO::AIO::max_parallel $nthreads
294 331
295Sets the maximum number of AIO threads to C<$nthreads>. If more than 332Sets the maximum number of AIO threads to C<$nthreads>. If more than the
296the specified number of threads are currently running, kill them. This 333specified number of threads are currently running, this function kills
297function blocks until the limit is reached. 334them. This function blocks until the limit is reached.
335
336While C<$nthreads> are zero, aio requests get queued but not executed
337until the number of threads has been increased again.
298 338
299This module automatically runs C<max_parallel 0> at program end, to ensure 339This module automatically runs C<max_parallel 0> at program end, to ensure
300that all threads are killed and that there are no outstanding requests. 340that all threads are killed and that there are no outstanding requests.
301 341
302Under normal circumstances you don't need to call this function. 342Under normal circumstances you don't need to call this function.
… …
306Sets the maximum number of outstanding requests to C<$nreqs>. If you 346Sets the maximum number of outstanding requests to C<$nreqs>. If you
307try to queue up more than this number of requests, the caller will block until 347try to queue up more than this number of requests, the caller will block until
308some requests have been handled. 348some requests have been handled.
309 349
310The default is very large, so normally there is no practical limit. If you 350The default is very large, so normally there is no practical limit. If you
311queue up many requests in a loop it it often improves speed if you set 351queue up many requests in a loop it often improves speed if you set
312this to a relatively low number, such as C<100>. 352this to a relatively low number, such as C<100>.
313 353
314Under normal circumstances you don't need to call this function. 354Under normal circumstances you don't need to call this function.
315 355
316=back 356=back
… …
341 381
3421; 3821;
343 383
344=head2 FORK BEHAVIOUR 384=head2 FORK BEHAVIOUR
345 385
346IO::AIO handles all outstanding AIO requests before the fork, destroys all 386Before the fork, IO::AIO enters a quiescent state where no requests
347AIO threads, and recreates them in both the parent and the child after the 387can be added in other threads and no results will be processed. After
348fork. 388the fork the parent simply leaves the quiescent state and continues
349 389request/result processing, while the child clears the request/result
390queue (so the requests started before the fork will only be handled in
391the parent). Threats will be started on demand until the limit ste in the
392parent process has been reached again.
350 393
351=head1 SEE ALSO 394=head1 SEE ALSO
352 395
353L<Coro>, L<Linux::AIO>. 396L<Coro>, L<Linux::AIO>.
354 397

Diff Legend

– Removed lines
+ Added lines
< Changed lines
> Changed lines