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 1.84 by root, Sat Oct 28 00:17:30 2006 UTC vs.
Revision 1.89 by root, Sun Oct 29 11:03:18 2006 UTC

… …
50 50
51=head1 DESCRIPTION 51=head1 DESCRIPTION
52 52
53This module implements asynchronous I/O using whatever means your 53This module implements asynchronous I/O using whatever means your
54operating system supports. 54operating system supports.
55
56Asynchronous means that operations that can normally block your program
57(e.g. reading from disk) will be done asynchronously: the operation
58will still block, but you can do something else in the meantime. This
59is extremely useful for programs that need to stay interactive even
60when doing heavy I/O (GUI programs, high performance network servers
61etc.), but can also be used to easily do operations in parallel that are
62normally done sequentially, e.g. stat'ing many files, which is much faster
63on a RAID volume or over NFS when you do a number of stat operations
64concurrently.
65
66While this works on all types of file descriptors (for example sockets),
67using these functions on file descriptors that support nonblocking
68operation (again, sockets, pipes etc.) is very inefficient. Use an event
69loop for that (such as the L<Event|Event> module): IO::AIO will naturally
70fit into such an event loop itself.
55 71
56In this version, a number of threads are started that execute your 72In this version, a number of threads are started that execute your
57requests and signal their completion. You don't need thread support 73requests and signal their completion. You don't need thread support
58in perl, and the threads created by this module will not be visible 74in perl, and the threads created by this module will not be visible
59to perl. In the future, this module might make use of the native aio 75to perl. In the future, this module might make use of the native aio
60functions available on many operating systems. However, they are often 76functions available on many operating systems. However, they are often
61not well-supported or restricted (Linux doesn't allow them on normal 77not well-supported or restricted (GNU/Linux doesn't allow them on normal
62files currently, for example), and they would only support aio_read and 78files currently, for example), and they would only support aio_read and
63aio_write, so the remaining functionality would have to be implemented 79aio_write, so the remaining functionality would have to be implemented
64using threads anyway. 80using threads anyway.
65 81
66Although the module will work with in the presence of other (Perl-) 82Although the module will work with in the presence of other (Perl-)
67threads, it is currently not reentrant in any way, so use appropriate 83threads, it is currently not reentrant in any way, so use appropriate
68locking yourself, always call C<poll_cb> from within the same thread, or 84locking yourself, always call C<poll_cb> from within the same thread, or
69never call C<poll_cb> (or other C<aio_> functions) recursively. 85never call C<poll_cb> (or other C<aio_> functions) recursively.
70 86
87=head2 EXAMPLE
88
89This is a simple example that uses the Event module and loads
90F</etc/passwd> asynchronously:
91
92 use Fcntl;
93 use Event;
94 use IO::AIO;
95
96 # register the IO::AIO callback with Event
97 Event->io (fd => IO::AIO::poll_fileno,
98 poll => 'r',
99 cb => \&IO::AIO::poll_cb);
100
101 # queue the request to open /etc/passwd
102 aio_open "/etc/passwd", O_RDONLY, 0, sub {
103 my $fh = $_[0]
104 or die "error while opening: $!";
105
106 # stat'ing filehandles is generally non-blocking
107 my $size = -s $fh;
108
109 # queue a request to read the file
110 my $contents;
111 aio_read $fh, 0, $size, $contents, 0, sub {
112 $_[0] == $size
113 or die "short read: $!";
114
115 close $fh;
116
117 # file contents now in $contents
118 print $contents;
119
120 # exit event loop and program
121 Event::unloop;
122 };
123 };
124
125 # possibly queue up other requests, or open GUI windows,
126 # check for sockets etc. etc.
127
128 # process events as long as there are some:
129 Event::loop;
130
71=head1 REQUEST ANATOMY AND LIFETIME 131=head1 REQUEST ANATOMY AND LIFETIME
72 132
73Every C<aio_*> function creates a request. which is a C data structure not 133Every C<aio_*> function creates a request. which is a C data structure not
74directly visible to Perl. 134directly visible to Perl.
75 135
… …
116Request has reached the end of its lifetime and holds no resources anymore 176Request 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 177(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 178aio request is severed and calling its methods will either do nothing or
119result in a runtime error). 179result in a runtime error).
120 180
181=back
182
121=cut 183=cut
122 184
123package IO::AIO; 185package IO::AIO;
124 186
125no warnings; 187no warnings;
126use strict 'vars'; 188use strict 'vars';
127 189
128use base 'Exporter'; 190use base 'Exporter';
129 191
130BEGIN { 192BEGIN {
131 our $VERSION = '2.0'; 193 our $VERSION = '2.1';
132 194
133 our @AIO_REQ = qw(aio_sendfile aio_read aio_write aio_open aio_close aio_stat 195 our @AIO_REQ = qw(aio_sendfile aio_read aio_write aio_open aio_close aio_stat
134 aio_lstat aio_unlink aio_rmdir aio_readdir aio_scandir aio_symlink 196 aio_lstat aio_unlink aio_rmdir aio_readdir aio_scandir aio_symlink
135 aio_fsync aio_fdatasync aio_readahead aio_rename aio_link aio_move 197 aio_fsync aio_fdatasync aio_readahead aio_rename aio_link aio_move
136 aio_copy aio_group aio_nop aio_mknod); 198 aio_copy aio_group aio_nop aio_mknod);
137 our @EXPORT = (@AIO_REQ, qw(aioreq_pri aioreq_nice)); 199 our @EXPORT = (@AIO_REQ, qw(aioreq_pri aioreq_nice));
138 our @EXPORT_OK = qw(poll_fileno poll_cb poll_wait flush 200 our @EXPORT_OK = qw(poll_fileno poll_cb poll_wait flush
139 min_parallel max_parallel nreqs nready npending); 201 min_parallel max_parallel max_idle
202 nreqs nready npending nthreads
203 max_poll_time max_poll_reqs);
140 204
141 @IO::AIO::GRP::ISA = 'IO::AIO::REQ'; 205 @IO::AIO::GRP::ISA = 'IO::AIO::REQ';
142 206
143 require XSLoader; 207 require XSLoader;
144 XSLoader::load ("IO::AIO", $VERSION); 208 XSLoader::load ("IO::AIO", $VERSION);
145} 209}
146 210
147=head1 FUNCTIONS 211=head1 FUNCTIONS
148 212
149=head2 AIO FUNCTIONS 213=head2 AIO REQUEST FUNCTIONS
150 214
151All the C<aio_*> calls are more or less thin wrappers around the syscall 215All the C<aio_*> calls are more or less thin wrappers around the syscall
152with the same name (sans C<aio_>). The arguments are similar or identical, 216with the same name (sans C<aio_>). The arguments are similar or identical,
153and they all accept an additional (and optional) C<$callback> argument 217and they all accept an additional (and optional) C<$callback> argument
154which must be a code reference. This code reference will get called with 218which must be a code reference. This code reference will get called with
… …
157syscall has been executed asynchronously. 221syscall has been executed asynchronously.
158 222
159All functions expecting a filehandle keep a copy of the filehandle 223All functions expecting a filehandle keep a copy of the filehandle
160internally until the request has finished. 224internally until the request has finished.
161 225
162All requests return objects of type L<IO::AIO::REQ> that allow further 226All functions return request objects of type L<IO::AIO::REQ> that allow
163manipulation of those requests while they are in-flight. 227further manipulation of those requests while they are in-flight.
164 228
165The pathnames you pass to these routines I<must> be absolute and 229The pathnames you pass to these routines I<must> be absolute and
166encoded in byte form. The reason for the former is that at the time the 230encoded as octets. The reason for the former is that at the time the
167request is being executed, the current working directory could have 231request is being executed, the current working directory could have
168changed. Alternatively, you can make sure that you never change the 232changed. Alternatively, you can make sure that you never change the
169current working directory. 233current working directory anywhere in the program and then use relative
234paths.
170 235
171To encode pathnames to byte form, either make sure you either: a) 236To encode pathnames as octets, either make sure you either: a) always pass
172always pass in filenames you got from outside (command line, readdir 237in filenames you got from outside (command line, readdir etc.) without
173etc.), b) are ASCII or ISO 8859-1, c) use the Encode module and encode 238tinkering, b) are ASCII or ISO 8859-1, c) use the Encode module and encode
174your pathnames to the locale (or other) encoding in effect in the user 239your pathnames to the locale (or other) encoding in effect in the user
175environment, d) use Glib::filename_from_unicode on unicode filenames or e) 240environment, d) use Glib::filename_from_unicode on unicode filenames or e)
176use something else. 241use something else to ensure your scalar has the correct contents.
242
243This works, btw. independent of the internal UTF-8 bit, which IO::AIO
244handles correctly wether it is set or not.
177 245
178=over 4 246=over 4
179 247
180=item $prev_pri = aioreq_pri [$pri] 248=item $prev_pri = aioreq_pri [$pri]
181 249
… …
204 }; 272 };
205 273
206=item aioreq_nice $pri_adjust 274=item aioreq_nice $pri_adjust
207 275
208Similar to C<aioreq_pri>, but subtracts the given value from the current 276Similar to C<aioreq_pri>, but subtracts the given value from the current
209priority, so effects are cumulative. 277priority, so the effect is cumulative.
210 278
211=item aio_open $pathname, $flags, $mode, $callback->($fh) 279=item aio_open $pathname, $flags, $mode, $callback->($fh)
212 280
213Asynchronously open or create a file and call the callback with a newly 281Asynchronously open or create a file and call the callback with a newly
214created filehandle for the file. 282created filehandle for the file.
… …
331Asynchronously unlink (delete) a file and call the callback with the 399Asynchronously unlink (delete) a file and call the callback with the
332result code. 400result code.
333 401
334=item aio_mknod $path, $mode, $dev, $callback->($status) 402=item aio_mknod $path, $mode, $dev, $callback->($status)
335 403
404[EXPERIMENTAL]
405
336Asynchronously create a device node (or fifo). See mknod(2). 406Asynchronously create a device node (or fifo). See mknod(2).
337 407
338The only portable (POSIX) way of calling this function is: 408The only (POSIX-) portable way of calling this function is:
339 409
340 aio_mknod $path, IO::AIO::S_IFIFO | $mode, 0, sub { ... 410 aio_mknod $path, IO::AIO::S_IFIFO | $mode, 0, sub { ...
341 411
342=item aio_link $srcpath, $dstpath, $callback->($status) 412=item aio_link $srcpath, $dstpath, $callback->($status)
343 413
… …
829 899
830=back 900=back
831 901
832=head2 SUPPORT FUNCTIONS 902=head2 SUPPORT FUNCTIONS
833 903
904=head3 EVENT PROCESSING AND EVENT LOOP INTEGRATION
905
834=over 4 906=over 4
835 907
836=item $fileno = IO::AIO::poll_fileno 908=item $fileno = IO::AIO::poll_fileno
837 909
838Return the I<request result pipe file descriptor>. This filehandle must be 910Return the I<request result pipe file descriptor>. This filehandle must be
… …
842 914
843See C<poll_cb> for an example. 915See C<poll_cb> for an example.
844 916
845=item IO::AIO::poll_cb 917=item IO::AIO::poll_cb
846 918
847Process all outstanding events on the result pipe. You have to call this 919Process some outstanding events on the result pipe. You have to call this
848regularly. Returns the number of events processed. Returns immediately 920regularly. Returns the number of events processed. Returns immediately
849when no events are outstanding. 921when no events are outstanding. The amount of events processed depends on
922the settings of C<IO::AIO::max_poll_req> and C<IO::AIO::max_poll_time>.
850 923
851If not all requests were processed for whatever reason, the filehandle 924If not all requests were processed for whatever reason, the filehandle
852will still be ready when C<poll_cb> returns. 925will still be ready when C<poll_cb> returns.
853 926
854Example: Install an Event watcher that automatically calls 927Example: Install an Event watcher that automatically calls
… …
856 929
857 Event->io (fd => IO::AIO::poll_fileno, 930 Event->io (fd => IO::AIO::poll_fileno,
858 poll => 'r', async => 1, 931 poll => 'r', async => 1,
859 cb => \&IO::AIO::poll_cb); 932 cb => \&IO::AIO::poll_cb);
860 933
861=item IO::AIO::poll_some $max_requests 934=item IO::AIO::max_poll_reqs $nreqs
862 935
863Similar to C<poll_cb>, but only processes up to C<$max_requests> requests 936=item IO::AIO::max_poll_time $seconds
864at a time.
865 937
866Useful if you want to ensure some level of interactiveness when perl is 938These set the maximum number of requests (default C<0>, meaning infinity)
867not fast enough to process all requests in time. 939that are being processed by C<IO::AIO::poll_cb> in one call, respectively
940the maximum amount of time (default C<0>, meaning infinity) spent in
941C<IO::AIO::poll_cb> to process requests (more correctly the mininum amount
942of time C<poll_cb> is allowed to use).
943
944Setting C<max_poll_time> to a non-zero value creates an overhead of one
945syscall per request processed, which is not normally a problem unless your
946callbacks are really really fast or your OS is really really slow (I am
947not mentioning Solaris here). Using C<max_poll_reqs> incurs no overhead.
948
949Setting these is useful if you want to ensure some level of
950interactiveness when perl is not fast enough to process all requests in
951time.
952
953For interactive programs, values such as C<0.01> to C<0.1> should be fine.
868 954
869Example: Install an Event watcher that automatically calls 955Example: Install an Event watcher that automatically calls
870IO::AIO::poll_some with low priority, to ensure that other parts of the 956IO::AIO::poll_cb with low priority, to ensure that other parts of the
871program get the CPU sometimes even under high AIO load. 957program get the CPU sometimes even under high AIO load.
872 958
959 # try not to spend much more than 0.1s in poll_cb
960 IO::AIO::max_poll_time 0.1;
961
962 # use a low priority so other tasks have priority
873 Event->io (fd => IO::AIO::poll_fileno, 963 Event->io (fd => IO::AIO::poll_fileno,
874 poll => 'r', nice => 1, 964 poll => 'r', nice => 1,
875 cb => sub { IO::AIO::poll_some 256 }); 965 cb => &IO::AIO::poll_cb);
876 966
877=item IO::AIO::poll_wait 967=item IO::AIO::poll_wait
878 968
879Wait till the result filehandle becomes ready for reading (simply does a 969Wait till the result filehandle becomes ready for reading (simply does a
880C<select> on the filehandle. This is useful if you want to synchronously wait 970C<select> on the filehandle. This is useful if you want to synchronously
881for some requests to finish). 971wait for some requests to finish).
882 972
883See C<nreqs> for an example. 973See C<nreqs> for an example.
884 974
975=item IO::AIO::poll
976
977Waits until some requests have been handled.
978
979Strictly equivalent to:
980
981 IO::AIO::poll_wait, IO::AIO::poll_cb
982 if IO::AIO::nreqs;
983
885=item IO::AIO::nreqs 984=item IO::AIO::flush
886 985
887Returns the number of requests currently in the ready, execute or pending 986Wait till all outstanding AIO requests have been handled.
888states (i.e. for which their callback has not been invoked yet).
889 987
890Example: wait till there are no outstanding requests anymore: 988Strictly equivalent to:
891 989
892 IO::AIO::poll_wait, IO::AIO::poll_cb 990 IO::AIO::poll_wait, IO::AIO::poll_cb
893 while IO::AIO::nreqs; 991 while IO::AIO::nreqs;
894 992
895=item IO::AIO::nready 993=head3 CONTROLLING THE NUMBER OF THREADS
896
897Returns the number of requests currently in the ready state (not yet
898executed).
899
900=item IO::AIO::npending
901
902Returns the number of requests currently in the pending state (executed,
903but not yet processed by poll_cb).
904
905=item IO::AIO::flush
906
907Wait till all outstanding AIO requests have been handled.
908
909Strictly equivalent to:
910
911 IO::AIO::poll_wait, IO::AIO::poll_cb
912 while IO::AIO::nreqs;
913
914=item IO::AIO::poll
915
916Waits until some requests have been handled.
917
918Strictly equivalent to:
919
920 IO::AIO::poll_wait, IO::AIO::poll_cb
921 if IO::AIO::nreqs;
922 994
923=item IO::AIO::min_parallel $nthreads 995=item IO::AIO::min_parallel $nthreads
924 996
925Set the minimum number of AIO threads to C<$nthreads>. The current 997Set the minimum number of AIO threads to C<$nthreads>. The current
926default is C<8>, which means eight asynchronous operations can execute 998default is C<8>, which means eight asynchronous operations can execute
927concurrently at any one time (the number of outstanding requests, 999concurrently at any one time (the number of outstanding requests,
928however, is unlimited). 1000however, is unlimited).
929 1001
930IO::AIO starts threads only on demand, when an AIO request is queued and 1002IO::AIO starts threads only on demand, when an AIO request is queued and
931no free thread exists. 1003no free thread exists. Please note that queueing up a hundred requests can
1004create demand for a hundred threads, even if it turns out that everything
1005is in the cache and could have been processed faster by a single thread.
932 1006
933It is recommended to keep the number of threads relatively low, as some 1007It is recommended to keep the number of threads relatively low, as some
934Linux kernel versions will scale negatively with the number of threads 1008Linux kernel versions will scale negatively with the number of threads
935(higher parallelity => MUCH higher latency). With current Linux 2.6 1009(higher parallelity => MUCH higher latency). With current Linux 2.6
936versions, 4-32 threads should be fine. 1010versions, 4-32 threads should be fine.
… …
950This module automatically runs C<max_parallel 0> at program end, to ensure 1024This module automatically runs C<max_parallel 0> at program end, to ensure
951that all threads are killed and that there are no outstanding requests. 1025that all threads are killed and that there are no outstanding requests.
952 1026
953Under normal circumstances you don't need to call this function. 1027Under normal circumstances you don't need to call this function.
954 1028
1029=item IO::AIO::max_idle $nthreads
1030
1031Limit the number of threads (default: 4) that are allowed to idle (i.e.,
1032threads that did not get a request to process within 10 seconds). That
1033means if a thread becomes idle while C<$nthreads> other threads are also
1034idle, it will free its resources and exit.
1035
1036This is useful when you allow a large number of threads (e.g. 100 or 1000)
1037to allow for extremely high load situations, but want to free resources
1038under normal circumstances (1000 threads can easily consume 30MB of RAM).
1039
1040The default is probably ok in most situations, especially if thread
1041creation is fast. If thread creation is very slow on your system you might
1042want to use larger values.
1043
955=item $oldmaxreqs = IO::AIO::max_outstanding $maxreqs 1044=item $oldmaxreqs = IO::AIO::max_outstanding $maxreqs
956 1045
957This is a very bad function to use in interactive programs because it 1046This is a very bad function to use in interactive programs because it
958blocks, and a bad way to reduce concurrency because it is inexact: Better 1047blocks, and a bad way to reduce concurrency because it is inexact: Better
959use an C<aio_group> together with a feed callback. 1048use an C<aio_group> together with a feed callback.
… …
967number of outstanding requests. 1056number of outstanding requests.
968 1057
969You can still queue as many requests as you want. Therefore, 1058You can still queue as many requests as you want. Therefore,
970C<max_oustsanding> is mainly useful in simple scripts (with low values) or 1059C<max_oustsanding> is mainly useful in simple scripts (with low values) or
971as a stop gap to shield against fatal memory overflow (with large values). 1060as a stop gap to shield against fatal memory overflow (with large values).
1061
1062=head3 STATISTICAL INFORMATION
1063
1064=item IO::AIO::nreqs
1065
1066Returns the number of requests currently in the ready, execute or pending
1067states (i.e. for which their callback has not been invoked yet).
1068
1069Example: wait till there are no outstanding requests anymore:
1070
1071 IO::AIO::poll_wait, IO::AIO::poll_cb
1072 while IO::AIO::nreqs;
1073
1074=item IO::AIO::nready
1075
1076Returns the number of requests currently in the ready state (not yet
1077executed).
1078
1079=item IO::AIO::npending
1080
1081Returns the number of requests currently in the pending state (executed,
1082but not yet processed by poll_cb).
972 1083
973=back 1084=back
974 1085
975=cut 1086=cut
976 1087

Diff Legend

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