--- IO-AIO/AIO.pm 2005/08/28 11:05:50 1.39 +++ IO-AIO/AIO.pm 2006/10/28 01:40:30 1.85 @@ -17,64 +17,147 @@ $_[0] > 0 or die "read error: $!"; }; - # Event + # version 2+ has request and group objects + use IO::AIO 2; + + aioreq_pri 4; # give next request a very high priority + my $req = aio_unlink "/tmp/file", sub { }; + $req->cancel; # cancel request if still in queue + + my $grp = aio_group sub { print "all stats done\n" }; + add $grp aio_stat "..." for ...; + + # AnyEvent integration + open my $fh, "<&=" . IO::AIO::poll_fileno or die "$!"; + my $w = AnyEvent->io (fh => $fh, poll => 'r', cb => sub { IO::AIO::poll_cb }); + + # Event integration Event->io (fd => IO::AIO::poll_fileno, poll => 'r', cb => \&IO::AIO::poll_cb); - # Glib/Gtk2 + # Glib/Gtk2 integration add_watch Glib::IO IO::AIO::poll_fileno, in => sub { IO::AIO::poll_cb; 1 }; - # Tk + # Tk integration Tk::Event::IO->fileevent (IO::AIO::poll_fileno, "", readable => \&IO::AIO::poll_cb); - # Danga::Socket + # Danga::Socket integration Danga::Socket->AddOtherFds (IO::AIO::poll_fileno => \&IO::AIO::poll_cb); - =head1 DESCRIPTION This module implements asynchronous I/O using whatever means your operating system supports. -Currently, a number of threads are started that execute your read/writes -and signal their completion. You don't need thread support in your libc or -perl, and the threads created by this module will not be visible to the -pthreads library. In the future, this module might make use of the native -aio functions available on many operating systems. However, they are often -not well-supported (Linux doesn't allow them on normal files currently, -for example), and they would only support aio_read and aio_write, so the -remaining functionality would have to be implemented using threads anyway. - -Although the module will work with in the presence of other threads, it is -currently not reentrant, so use appropriate locking yourself, always call -C from within the same thread, or never call C (or other -C functions) recursively. +Asynchronous means that operations that can normally block your program +(e.g. reading from disk) will be done asynchronously: the operation +will still block, but you can do something else in the meantime. This +is extremely useful for programs that need to stay interactive even +when doing heavy I/O (GUI programs, high performance network servers +etc.), but can also be used to easily do operations in parallel that are +normally done sequentially, e.g. stat'ing many files, which is much faster +on a RAID volume or over NFS when you do a number of stat operations +concurrently. + +While this works on all types of file descriptors (for example sockets), +using these functions on file descriptors that support nonblocking +operation (again, sockets, pipes etc.) is very inefficient. Use an event +loop for that (such as the L module): IO::AIO will naturally +fit into such an event loop itself. + +In this version, a number of threads are started that execute your +requests and signal their completion. You don't need thread support +in perl, and the threads created by this module will not be visible +to perl. In the future, this module might make use of the native aio +functions available on many operating systems. However, they are often +not well-supported or restricted (GNU/Linux doesn't allow them on normal +files currently, for example), and they would only support aio_read and +aio_write, so the remaining functionality would have to be implemented +using threads anyway. + +Although the module will work with in the presence of other (Perl-) +threads, it is currently not reentrant in any way, so use appropriate +locking yourself, always call C from within the same thread, or +never call C (or other C functions) recursively. + +=head1 REQUEST ANATOMY AND LIFETIME + +Every C function creates a request. which is a C data structure not +directly visible to Perl. + +If called in non-void context, every request function returns a Perl +object representing the request. In void context, nothing is returned, +which saves a bit of memory. + +The perl object is a fairly standard ref-to-hash object. The hash contents +are not used by IO::AIO so you are free to store anything you like in it. + +During their existance, aio requests travel through the following states, +in order: + +=over 4 + +=item ready + +Immediately after a request is created it is put into the ready state, +waiting for a thread to execute it. + +=item execute + +A thread has accepted the request for processing and is currently +executing it (e.g. blocking in read). + +=item pending + +The request has been executed and is waiting for result processing. + +While request submission and execution is fully asynchronous, result +processing is not and relies on the perl interpreter calling C +(or another function with the same effect). + +=item result + +The request results are processed synchronously by C. + +The C function will process all outstanding aio requests by +calling their callbacks, freeing memory associated with them and managing +any groups they are contained in. + +=item done + +Request has reached the end of its lifetime and holds no resources anymore +(except possibly for the Perl object, but its connection to the actual +aio request is severed and calling its methods will either do nothing or +result in a runtime error). =cut package IO::AIO; no warnings; +use strict 'vars'; use base 'Exporter'; -use Fcntl (); - BEGIN { - $VERSION = 1.6; + our $VERSION = '2.0'; + + our @AIO_REQ = qw(aio_sendfile aio_read aio_write aio_open aio_close aio_stat + aio_lstat aio_unlink aio_rmdir aio_readdir aio_scandir aio_symlink + aio_fsync aio_fdatasync aio_readahead aio_rename aio_link aio_move + aio_copy aio_group aio_nop aio_mknod); + our @EXPORT = (@AIO_REQ, qw(aioreq_pri aioreq_nice)); + our @EXPORT_OK = qw(poll_fileno poll_cb poll_wait flush + min_parallel max_parallel nreqs nready npending); - @EXPORT = qw(aio_sendfile aio_read aio_write aio_open aio_close aio_stat - aio_lstat aio_unlink aio_rmdir aio_readdir aio_symlink - aio_fsync aio_fdatasync aio_readahead); - @EXPORT_OK = qw(poll_fileno poll_cb min_parallel max_parallel - max_outstanding nreqs); + @IO::AIO::GRP::ISA = 'IO::AIO::REQ'; require XSLoader; - XSLoader::load IO::AIO, $VERSION; + XSLoader::load ("IO::AIO", $VERSION); } =head1 FUNCTIONS @@ -92,6 +175,9 @@ All functions expecting a filehandle keep a copy of the filehandle internally until the request has finished. +All requests return objects of type L that allow further +manipulation of those requests while they are in-flight. + The pathnames you pass to these routines I be absolute and encoded in byte form. The reason for the former is that at the time the request is being executed, the current working directory could have @@ -107,7 +193,38 @@ =over 4 -=item aio_open $pathname, $flags, $mode, $callback +=item $prev_pri = aioreq_pri [$pri] + +Returns the priority value that would be used for the next request and, if +C<$pri> is given, sets the priority for the next aio request. + +The default priority is C<0>, the minimum and maximum priorities are C<-4> +and C<4>, respectively. Requests with higher priority will be serviced +first. + +The priority will be reset to C<0> after each call to one of the C +functions. + +Example: open a file with low priority, then read something from it with +higher priority so the read request is serviced before other low priority +open requests (potentially spamming the cache): + + aioreq_pri -3; + aio_open ..., sub { + return unless $_[0]; + + aioreq_pri -2; + aio_read $_[0], ..., sub { + ... + }; + }; + +=item aioreq_nice $pri_adjust + +Similar to C, but subtracts the given value from the current +priority, so effects are cumulative. + +=item aio_open $pathname, $flags, $mode, $callback->($fh) Asynchronously open or create a file and call the callback with a newly created filehandle for the file. @@ -134,7 +251,7 @@ } }; -=item aio_close $fh, $callback +=item aio_close $fh, $callback->($status) Asynchronously close a file and call the callback with the result code. I although accepted, you should not pass in a perl @@ -145,9 +262,9 @@ This is supposed to be a bug in the API, so that might change. It's therefore best to avoid this function. -=item aio_read $fh,$offset,$length, $data,$dataoffset,$callback +=item aio_read $fh,$offset,$length, $data,$dataoffset, $callback->($retval) -=item aio_write $fh,$offset,$length, $data,$dataoffset,$callback +=item aio_write $fh,$offset,$length, $data,$dataoffset, $callback->($retval) Reads or writes C bytes from the specified C and C into the scalar given by C and offset C and calls the @@ -166,7 +283,7 @@ print "read $_[0] bytes: <$buffer>\n"; }; -=item aio_sendfile $out_fh, $in_fh, $in_offset, $length, $callback +=item aio_sendfile $out_fh, $in_fh, $in_offset, $length, $callback->($retval) Tries to copy C<$length> bytes from C<$in_fh> to C<$out_fh>. It starts reading at byte offset C<$in_offset>, and starts writing at the current @@ -189,7 +306,7 @@ value equals C<$length> one can assume that C<$length> bytes have been read. -=item aio_readahead $fh,$offset,$length, $callback +=item aio_readahead $fh,$offset,$length, $callback->($retval) C populates the page cache with data from a file so that subsequent reads from that file will not block on disk I/O. The C<$offset> @@ -203,9 +320,9 @@ If that syscall doesn't exist (likely if your OS isn't Linux) it will be emulated by simply reading the data, which would have a similar effect. -=item aio_stat $fh_or_path, $callback +=item aio_stat $fh_or_path, $callback->($status) -=item aio_lstat $fh, $callback +=item aio_lstat $fh, $callback->($status) Works like perl's C or C in void context. The callback will be called after the stat and the results will be available using C @@ -225,17 +342,40 @@ print "size is ", -s _, "\n"; }; -=item aio_unlink $pathname, $callback +=item aio_unlink $pathname, $callback->($status) Asynchronously unlink (delete) a file and call the callback with the result code. -=item aio_rmdir $pathname, $callback +=item aio_mknod $path, $mode, $dev, $callback->($status) + +Asynchronously create a device node (or fifo). See mknod(2). + +The only portable (POSIX) way of calling this function is: + + aio_mknod $path, IO::AIO::S_IFIFO | $mode, 0, sub { ... + +=item aio_link $srcpath, $dstpath, $callback->($status) + +Asynchronously create a new link to the existing object at C<$srcpath> at +the path C<$dstpath> and call the callback with the result code. + +=item aio_symlink $srcpath, $dstpath, $callback->($status) + +Asynchronously create a new symbolic link to the existing object at C<$srcpath> at +the path C<$dstpath> and call the callback with the result code. + +=item aio_rename $srcpath, $dstpath, $callback->($status) + +Asynchronously rename the object at C<$srcpath> to C<$dstpath>, just as +rename(2) and call the callback with the result code. + +=item aio_rmdir $pathname, $callback->($status) Asynchronously rmdir (delete) a directory and call the callback with the result code. -=item aio_readdir $pathname $callback +=item aio_readdir $pathname, $callback->($entries) Unlike the POSIX call of the same name, C reads an entire directory (i.e. opendir + readdir + closedir). The entries will not be @@ -244,12 +384,254 @@ The callback a single argument which is either C or an array-ref with the filenames. -=item aio_fsync $fh, $callback +=item aio_copy $srcpath, $dstpath, $callback->($status) + +Try to copy the I (directories not supported as either source or +destination) from C<$srcpath> to C<$dstpath> and call the callback with +the C<0> (error) or C<-1> ok. + +This is a composite request that it creates the destination file with +mode 0200 and copies the contents of the source file into it using +C, followed by restoring atime, mtime, access mode and +uid/gid, in that order. + +If an error occurs, the partial destination file will be unlinked, if +possible, except when setting atime, mtime, access mode and uid/gid, where +errors are being ignored. + +=cut + +sub aio_copy($$;$) { + my ($src, $dst, $cb) = @_; + + my $pri = aioreq_pri; + my $grp = aio_group $cb; + + aioreq_pri $pri; + add $grp aio_open $src, O_RDONLY, 0, sub { + if (my $src_fh = $_[0]) { + my @stat = stat $src_fh; + + aioreq_pri $pri; + add $grp aio_open $dst, O_CREAT | O_WRONLY | O_TRUNC, 0200, sub { + if (my $dst_fh = $_[0]) { + aioreq_pri $pri; + add $grp aio_sendfile $dst_fh, $src_fh, 0, $stat[7], sub { + if ($_[0] == $stat[7]) { + $grp->result (0); + close $src_fh; + + # those should not normally block. should. should. + utime $stat[8], $stat[9], $dst; + chmod $stat[2] & 07777, $dst_fh; + chown $stat[4], $stat[5], $dst_fh; + close $dst_fh; + } else { + $grp->result (-1); + close $src_fh; + close $dst_fh; + + aioreq $pri; + add $grp aio_unlink $dst; + } + }; + } else { + $grp->result (-1); + } + }, + + } else { + $grp->result (-1); + } + }; + + $grp +} + +=item aio_move $srcpath, $dstpath, $callback->($status) + +Try to move the I (directories not supported as either source or +destination) from C<$srcpath> to C<$dstpath> and call the callback with +the C<0> (error) or C<-1> ok. + +This is a composite request that tries to rename(2) the file first. If +rename files with C, it copies the file with C and, if +that is successful, unlinking the C<$srcpath>. + +=cut + +sub aio_move($$;$) { + my ($src, $dst, $cb) = @_; + + my $pri = aioreq_pri; + my $grp = aio_group $cb; + + aioreq_pri $pri; + add $grp aio_rename $src, $dst, sub { + if ($_[0] && $! == EXDEV) { + aioreq_pri $pri; + add $grp aio_copy $src, $dst, sub { + $grp->result ($_[0]); + + if (!$_[0]) { + aioreq_pri $pri; + add $grp aio_unlink $src; + } + }; + } else { + $grp->result ($_[0]); + } + }; + + $grp +} + +=item aio_scandir $path, $maxreq, $callback->($dirs, $nondirs) + +Scans a directory (similar to C) but additionally tries to +efficiently separate the entries of directory C<$path> into two sets of +names, directories you can recurse into (directories), and ones you cannot +recurse into (everything else, including symlinks to directories). + +C is a composite request that creates of many sub requests_ +C<$maxreq> specifies the maximum number of outstanding aio requests that +this function generates. If it is C<< <= 0 >>, then a suitable default +will be chosen (currently 4). + +On error, the callback is called without arguments, otherwise it receives +two array-refs with path-relative entry names. + +Example: + + aio_scandir $dir, 0, sub { + my ($dirs, $nondirs) = @_; + print "real directories: @$dirs\n"; + print "everything else: @$nondirs\n"; + }; + +Implementation notes. + +The C cannot be avoided, but C'ing every entry can. + +After reading the directory, the modification time, size etc. of the +directory before and after the readdir is checked, and if they match (and +isn't the current time), the link count will be used to decide how many +entries are directories (if >= 2). Otherwise, no knowledge of the number +of subdirectories will be assumed. + +Then entries will be sorted into likely directories (everything without +a non-initial dot currently) and likely non-directories (everything +else). Then every entry plus an appended C will be C'ed, +likely directories first. If that succeeds, it assumes that the entry +is a directory or a symlink to directory (which will be checked +seperately). This is often faster than stat'ing the entry itself because +filesystems might detect the type of the entry without reading the inode +data (e.g. ext2fs filetype feature). + +If the known number of directories (link count - 2) has been reached, the +rest of the entries is assumed to be non-directories. + +This only works with certainty on POSIX (= UNIX) filesystems, which +fortunately are the vast majority of filesystems around. + +It will also likely work on non-POSIX filesystems with reduced efficiency +as those tend to return 0 or 1 as link counts, which disables the +directory counting heuristic. + +=cut + +sub aio_scandir($$$) { + my ($path, $maxreq, $cb) = @_; + + my $pri = aioreq_pri; + + my $grp = aio_group $cb; + + $maxreq = 4 if $maxreq <= 0; + + # stat once + aioreq_pri $pri; + add $grp aio_stat $path, sub { + return $grp->result () if $_[0]; + my $now = time; + my $hash1 = join ":", (stat _)[0,1,3,7,9]; + + # read the directory entries + aioreq_pri $pri; + add $grp aio_readdir $path, sub { + my $entries = shift + or return $grp->result (); + + # stat the dir another time + aioreq_pri $pri; + add $grp aio_stat $path, sub { + my $hash2 = join ":", (stat _)[0,1,3,7,9]; + + my $ndirs; + + # take the slow route if anything looks fishy + if ($hash1 ne $hash2 or (stat _)[9] == $now) { + $ndirs = -1; + } else { + # if nlink == 2, we are finished + # on non-posix-fs's, we rely on nlink < 2 + $ndirs = (stat _)[3] - 2 + or return $grp->result ([], $entries); + } + + # sort into likely dirs and likely nondirs + # dirs == files without ".", short entries first + $entries = [map $_->[0], + sort { $b->[1] cmp $a->[1] } + map [$_, sprintf "%s%04d", (/.\./ ? "1" : "0"), length], + @$entries]; + + my (@dirs, @nondirs); + + my $statgrp = add $grp aio_group sub { + $grp->result (\@dirs, \@nondirs); + }; + + limit $statgrp $maxreq; + feed $statgrp sub { + return unless @$entries; + my $entry = pop @$entries; + + aioreq_pri $pri; + add $statgrp aio_stat "$path/$entry/.", sub { + if ($_[0] < 0) { + push @nondirs, $entry; + } else { + # need to check for real directory + aioreq_pri $pri; + add $statgrp aio_lstat "$path/$entry", sub { + if (-d _) { + push @dirs, $entry; + + unless (--$ndirs) { + push @nondirs, @$entries; + feed $statgrp; + } + } else { + push @nondirs, $entry; + } + } + } + }; + }; + }; + }; + }; + + $grp +} + +=item aio_fsync $fh, $callback->($status) Asynchronously call fsync on the given filehandle and call the callback with the fsync result code. -=item aio_fdatasync $fh, $callback +=item aio_fdatasync $fh, $callback->($status) Asynchronously call fdatasync on the given filehandle and call the callback with the fdatasync result code. @@ -257,6 +639,210 @@ If this call isn't available because your OS lacks it or it couldn't be detected, it will be emulated by calling C instead. +=item aio_group $callback->(...) + +This is a very special aio request: Instead of doing something, it is a +container for other aio requests, which is useful if you want to bundle +many requests into a single, composite, request with a definite callback +and the ability to cancel the whole request with its subrequests. + +Returns an object of class L. See its documentation below +for more info. + +Example: + + my $grp = aio_group sub { + print "all stats done\n"; + }; + + add $grp + (aio_stat ...), + (aio_stat ...), + ...; + +=item aio_nop $callback->() + +This is a special request - it does nothing in itself and is only used for +side effects, such as when you want to add a dummy request to a group so +that finishing the requests in the group depends on executing the given +code. + +While this request does nothing, it still goes through the execution +phase and still requires a worker thread. Thus, the callback will not +be executed immediately but only after other requests in the queue have +entered their execution phase. This can be used to measure request +latency. + +=item IO::AIO::aio_busy $fractional_seconds, $callback->() *NOT EXPORTED* + +Mainly used for debugging and benchmarking, this aio request puts one of +the request workers to sleep for the given time. + +While it is theoretically handy to have simple I/O scheduling requests +like sleep and file handle readable/writable, the overhead this creates is +immense (it blocks a thread for a long time) so do not use this function +except to put your application under artificial I/O pressure. + +=back + +=head2 IO::AIO::REQ CLASS + +All non-aggregate C functions return an object of this class when +called in non-void context. + +=over 4 + +=item cancel $req + +Cancels the request, if possible. Has the effect of skipping execution +when entering the B state and skipping calling the callback when +entering the the B state, but will leave the request otherwise +untouched. That means that requests that currently execute will not be +stopped and resources held by the request will not be freed prematurely. + +=item cb $req $callback->(...) + +Replace (or simply set) the callback registered to the request. + +=back + +=head2 IO::AIO::GRP CLASS + +This class is a subclass of L, so all its methods apply to +objects of this class, too. + +A IO::AIO::GRP object is a special request that can contain multiple other +aio requests. + +You create one by calling the C constructing function with a +callback that will be called when all contained requests have entered the +C state: + + my $grp = aio_group sub { + print "all requests are done\n"; + }; + +You add requests by calling the C method with one or more +C objects: + + $grp->add (aio_unlink "..."); + + add $grp aio_stat "...", sub { + $_[0] or return $grp->result ("error"); + + # add another request dynamically, if first succeeded + add $grp aio_open "...", sub { + $grp->result ("ok"); + }; + }; + +This makes it very easy to create composite requests (see the source of +C for an application) that work and feel like simple requests. + +=over 4 + +=item * The IO::AIO::GRP objects will be cleaned up during calls to +C, just like any other request. + +=item * They can be canceled like any other request. Canceling will cancel not +only the request itself, but also all requests it contains. + +=item * They can also can also be added to other IO::AIO::GRP objects. + +=item * You must not add requests to a group from within the group callback (or +any later time). + +=back + +Their lifetime, simplified, looks like this: when they are empty, they +will finish very quickly. If they contain only requests that are in the +C state, they will also finish. Otherwise they will continue to +exist. + +That means after creating a group you have some time to add requests. And +in the callbacks of those requests, you can add further requests to the +group. And only when all those requests have finished will the the group +itself finish. + +=over 4 + +=item add $grp ... + +=item $grp->add (...) + +Add one or more requests to the group. Any type of L can +be added, including other groups, as long as you do not create circular +dependencies. + +Returns all its arguments. + +=item $grp->cancel_subs + +Cancel all subrequests and clears any feeder, but not the group request +itself. Useful when you queued a lot of events but got a result early. + +=item $grp->result (...) + +Set the result value(s) that will be passed to the group callback when all +subrequests have finished and set thre groups errno to the current value +of errno (just like calling C without an error number). By default, +no argument will be passed and errno is zero. + +=item $grp->errno ([$errno]) + +Sets the group errno value to C<$errno>, or the current value of errno +when the argument is missing. + +Every aio request has an associated errno value that is restored when +the callback is invoked. This method lets you change this value from its +default (0). + +Calling C will also set errno, so make sure you either set C<$!> +before the call to C, or call c after it. + +=item feed $grp $callback->($grp) + +Sets a feeder/generator on this group: every group can have an attached +generator that generates requests if idle. The idea behind this is that, +although you could just queue as many requests as you want in a group, +this might starve other requests for a potentially long time. For +example, C might generate hundreds of thousands C +requests, delaying any later requests for a long time. + +To avoid this, and allow incremental generation of requests, you can +instead a group and set a feeder on it that generates those requests. The +feed callback will be called whenever there are few enough (see C, +below) requests active in the group itself and is expected to queue more +requests. + +The feed callback can queue as many requests as it likes (i.e. C does +not impose any limits). + +If the feed does not queue more requests when called, it will be +automatically removed from the group. + +If the feed limit is C<0>, it will be set to C<2> automatically. + +Example: + + # stat all files in @files, but only ever use four aio requests concurrently: + + my $grp = aio_group sub { print "finished\n" }; + limit $grp 4; + feed $grp sub { + my $file = pop @files + or return; + + add $grp aio_stat $file, sub { ... }; + }; + +=item limit $grp $num + +Sets the feeder limit for the group: The feeder will be called whenever +the group contains less than this many requests. + +Setting the limit to C<0> will pause the feeding process. + =back =head2 SUPPORT FUNCTIONS @@ -278,6 +864,9 @@ regularly. Returns the number of events processed. Returns immediately when no events are outstanding. +If not all requests were processed for whatever reason, the filehandle +will still be ready when C returns. + Example: Install an Event watcher that automatically calls IO::AIO::poll_cb with high priority: @@ -285,6 +874,22 @@ poll => 'r', async => 1, cb => \&IO::AIO::poll_cb); +=item IO::AIO::poll_some $max_requests + +Similar to C, but only processes up to C<$max_requests> requests +at a time. + +Useful if you want to ensure some level of interactiveness when perl is +not fast enough to process all requests in time. + +Example: Install an Event watcher that automatically calls +IO::AIO::poll_some with low priority, to ensure that other parts of the +program get the CPU sometimes even under high AIO load. + + Event->io (fd => IO::AIO::poll_fileno, + poll => 'r', nice => 1, + cb => sub { IO::AIO::poll_some 256 }); + =item IO::AIO::poll_wait Wait till the result filehandle becomes ready for reading (simply does a @@ -295,14 +900,24 @@ =item IO::AIO::nreqs -Returns the number of requests currently outstanding (i.e. for which their -callback has not been invoked yet). +Returns the number of requests currently in the ready, execute or pending +states (i.e. for which their callback has not been invoked yet). Example: wait till there are no outstanding requests anymore: IO::AIO::poll_wait, IO::AIO::poll_cb while IO::AIO::nreqs; +=item IO::AIO::nready + +Returns the number of requests currently in the ready state (not yet +executed). + +=item IO::AIO::npending + +Returns the number of requests currently in the pending state (executed, +but not yet processed by poll_cb). + =item IO::AIO::flush Wait till all outstanding AIO requests have been handled. @@ -323,17 +938,18 @@ =item IO::AIO::min_parallel $nthreads -Set the minimum number of AIO threads to C<$nthreads>. The current default -is C<4>, which means four asynchronous operations can be done at one time -(the number of outstanding operations, however, is unlimited). +Set the minimum number of AIO threads to C<$nthreads>. The current +default is C<8>, which means eight asynchronous operations can execute +concurrently at any one time (the number of outstanding requests, +however, is unlimited). IO::AIO starts threads only on demand, when an AIO request is queued and no free thread exists. -It is recommended to keep the number of threads low, as some Linux -kernel versions will scale negatively with the number of threads (higher -parallelity => MUCH higher latency). With current Linux 2.6 versions, 4-32 -threads should be fine. +It is recommended to keep the number of threads relatively low, as some +Linux kernel versions will scale negatively with the number of threads +(higher parallelity => MUCH higher latency). With current Linux 2.6 +versions, 4-32 threads should be fine. Under most circumstances you don't need to call this function, as the module selects a default that is suitable for low to moderate load. @@ -352,17 +968,23 @@ Under normal circumstances you don't need to call this function. -=item $oldnreqs = IO::AIO::max_outstanding $nreqs - -Sets the maximum number of outstanding requests to C<$nreqs>. If you -try to queue up more than this number of requests, the caller will block until -some requests have been handled. +=item $oldmaxreqs = IO::AIO::max_outstanding $maxreqs -The default is very large, so normally there is no practical limit. If you -queue up many requests in a loop it often improves speed if you set -this to a relatively low number, such as C<100>. +This is a very bad function to use in interactive programs because it +blocks, and a bad way to reduce concurrency because it is inexact: Better +use an C together with a feed callback. -Under normal circumstances you don't need to call this function. +Sets the maximum number of outstanding requests to C<$nreqs>. If you +to queue up more than this number of requests, the next call to the +C (and C and other functions calling C) +function will block until the limit is no longer exceeded. + +The default value is very large, so there is no practical limit on the +number of outstanding requests. + +You can still queue as many requests as you want. Therefore, +C is mainly useful in simple scripts (with low values) or +as a stop gap to shield against fatal memory overflow (with large values). =back @@ -384,27 +1006,57 @@ *$sym } -min_parallel 4; +min_parallel 8; END { - max_parallel 0; -} + min_parallel 1; + flush; +}; 1; =head2 FORK BEHAVIOUR +This module should do "the right thing" when the process using it forks: + Before the fork, IO::AIO enters a quiescent state where no requests can be added in other threads and no results will be processed. After the fork the parent simply leaves the quiescent state and continues -request/result processing, while the child clears the request/result -queue (so the requests started before the fork will only be handled in -the parent). Threats will be started on demand until the limit ste in the +request/result processing, while the child frees the request/result queue +(so that the requests started before the fork will only be handled in the +parent). Threads will be started on demand until the limit set in the parent process has been reached again. +In short: the parent will, after a short pause, continue as if fork had +not been called, while the child will act as if IO::AIO has not been used +yet. + +=head2 MEMORY USAGE + +Per-request usage: + +Each aio request uses - depending on your architecture - around 100-200 +bytes of memory. In addition, stat requests need a stat buffer (possibly +a few hundred bytes), readdir requires a result buffer and so on. Perl +scalars and other data passed into aio requests will also be locked and +will consume memory till the request has entered the done state. + +This is now awfully much, so queuing lots of requests is not usually a +problem. + +Per-thread usage: + +In the execution phase, some aio requests require more memory for +temporary buffers, and each thread requires a stack and other data +structures (usually around 16k-128k, depending on the OS). + +=head1 KNOWN BUGS + +Known bugs will be fixed in the next release. + =head1 SEE ALSO -L, L. +L. =head1 AUTHOR