MCE::Examples - A list of examples demonstrating Many-core Engine
This document describes MCE::Examples version 1.509
MCE comes with various examples showing real-world scenarios on parallelizing something as small as cat (try with -n) to searching for patterns and word count aggregation.
bin/mce_grep A wrapper script with support for the following C binaries. agrep, grep, egrep, fgrep, and tre-agrep Chunking may be applied either at the [file] level, for large file(s), or at the [list] level when parsing many files recursively. The gain in performance is noticeable for expensive patterns, especially with agrep and tre-agrep. barrier_sync.pl A barrier sync demonstration. cat.pl Concatenation script, similar to the cat binary. egrep.pl Egrep script, similar to the egrep binary. wc.pl Word count script, similar to the wc binary. findnull.pl A parallel script for reporting lines containing null fields. It is many times faster than the egrep binary. Try this against a large file containing very long lines. flow_model.pl Demonstrates MCE::Flow, MCE::Queue, and MCE->gather. foreach.pl, forseq.pl, forchunk.pl These take the same sqrt example from Parallel::Loops and measures the overhead of the engine. The number indicates the size of @input which can be submitted and displayed in under 1 second. Parallel::Loops is based on Parallel::ForkManager. MCE utilizes a pool of workers which persist while running. Parallel::Loops: 600 Forking each @input is expensive MCE foreach....: 34,000 Sends result after each @input MCE forseq.....: 70,000 Loops through sequence of numbers MCE forchunk...: 480,000 Chunking reduces overhead interval.pl Demonstration of the interval option appearing in MCE 1.5. iterator.pl Similar to forseq.pl. Specifies an iterator for input_data. A factory function is called which returns a closure (the iterator itself). matmult/matmult_base.pl, matmult_mce.pl, strassen_mce.pl Various matrix multiplication demonstrations benchmarking PDL, PDL + MCE, as well as parallelizing Strassen's divide-and-conquer algorithm. Also included are 2 plain Perl examples. scaling_pings.pl Perform ping test and report back failed IPs to standard output. seq_demo.pl A demonstration of the new sequence option appearing in MCE 1.3. Run with seq_demo.pl | sort tbray/wf_mce1.pl, wf_mce2.pl, wf_mce3.pl An implementation of wide finder utilizing MCE. As fast as MMAP IO when file resides in OS FS cache. 2x ~ 3x faster when reading directly from disk.
Imagine a long running process and wanting to parallelize an array against a pool of workers. Note: The sequence option may be used if simply wanting to loop through a sequence of numbers instead.
Below, a callback function is used for displaying results. The logic shows how one can output results immediately while still preserving order as if processing serially. The %tmp hash is a temporary cache for out-of-order results.
my @input_data = (0 .. 18000 - 1); ## Make an output iterator for gather. Output order is preserved. sub output_iterator { my %tmp; my $order_id = 1; return sub { my ($result, $chunk_id) = @_; $tmp{$chunk_id} = $result; while (1) { last unless (exists $tmp{$order_id}); printf "i: %d sqrt(i): %f\n", $input_data[$order_id - 1], $tmp{$order_id}; delete $tmp{$order_id++}; } }; } ## Use $chunk_ref->[0] or $_ to retrieve the element. my $mce = MCE->new( input_data => \@input_data, gather => output_iterator(), chunk_size => 1, max_workers => 3, user_func => sub { my ($mce, $chunk_ref, $chunk_id) = @_; my $result = sqrt($chunk_ref->[0]); MCE->gather($result, $chunk_id); } ); MCE->run();
This does the same thing using the foreach "sugar" method.
my $mce = MCE->new( chunk_size => 1, max_workers => 3, gather => output_iterator() ); MCE->foreach( \@input_data, sub { my ($mce, $chunk_ref, $chunk_id) = @_; my $result = sqrt($chunk_ref->[0]); MCE->gather($result, $chunk_id); });
The two examples described above were done using the core API. MCE 1.5 comes with various models. The MCE::Loop model is used for the next demonstration.
use MCE::Loop; MCE::Loop::init { chunk_size => 1, max_workers => 3, gather => output_iterator() }; mce_loop { my ($mce, $chunk_ref, $chunk_id) = @_; my $result = sqrt($chunk_ref->[0]); MCE->gather($result, $chunk_id); } @input_data;
Chunking has the effect of reducing the IPC overhead by many folds. A chunk containing up to $chunk_size items is sent to the next available worker.
my @input_data = (0 .. 6000 - 1); my $chunk_size = 500; my $max_workers = 3; ## Make an output iterator for gather. Output order is preserved. sub output_iterator { my %tmp; my $order_id = 1; return sub { my ($result, $chunk_id) = @_; $tmp{$chunk_id} = $result; while (1) { last unless (exists $tmp{$order_id}); my $i = ($order_id - 1) * $chunk_size; foreach ( @{ $tmp{$order_id} } ) { printf "i: %d sqrt(i): %f\n", $input_data[$i++], $_; } delete $tmp{$order_id++}; } }; } ## Use $chunk_ref or $_ to access the array reference. ## Chunking requires one to loop inside the code block. my $mce = MCE->new( input_data => \@input_data, gather => output_iterator(), chunk_size => $chunk_size, max_workers => $max_workers, user_func => sub { my ($mce, $chunk_ref, $chunk_id) = @_; my @result; foreach ( @{ $chunk_ref } ) { push @result, sqrt($_); } MCE->gather(\@result, $chunk_id); } ); MCE->run();
This does the same thing using the forchunk "sugar" method.
my $mce = MCE->new( chunk_size => $chunk_size, max_workers => $max_workers, gather => output_iterator() ); MCE->forchunk( \@input_data, sub { my ($mce, $chunk_ref, $chunk_id) = @_; my @result; foreach ( @{ $chunk_ref } ) { push @result, sqrt($_); } MCE->gather(\@result, $chunk_id); });
The MCE::Loop model is shown next. Looping is required inside the code block, just like above.
use MCE::Loop; MCE::Loop::init { chunk_size => $chunk_size, max_workers => $max_workers, gather => output_iterator() }; mce_loop { my ($mce, $chunk_ref, $chunk_id) = @_; my @result; foreach ( @{ $chunk_ref } ) { push @result, sqrt($_); } MCE->gather(\@result, $chunk_id); } @input_data;
The following is an extract from the seq_demo.pl example included with MCE. Think of having several MCEs running simultaneously in parallel. The sequence option including chunk_size may be specified individually under each task.
The input scalar $_ (not shown below) contains the same value as $seq_n.
use MCE; ## Run with seq_demo.pl | sort sub user_func { my ($mce, $seq_n, $chunk_id) = @_; my $wid = MCE->wid(); my $task_id = MCE->task_id(); my $task_wid = MCE->task_wid(); if (ref $seq_n eq 'ARRAY') { ## Received the next "chunked" sequence of numbers ## e.g. when chunk_size > 1, $seq_n will be an array ref above foreach (@{ $seq_n }) { printf( "task_id %d: seq_n %s: chunk_id %d: wid %d: task_wid %d\n", $task_id, $_, $chunk_id, $wid, $task_wid ); } } else { printf( "task_id %d: seq_n %s: chunk_id %d: wid %d: task_wid %d\n", $task_id, $seq_n, $chunk_id, $wid, $task_wid ); } } ## Each task is configured independently. my $mce = MCE->new( user_tasks => [{ max_workers => 2, chunk_size => 1, sequence => { begin => 11, end => 19, step => 1 }, user_func => \&user_func },{ max_workers => 2, chunk_size => 5, sequence => { begin => 21, end => 29, step => 1 }, user_func => \&user_func },{ max_workers => 2, chunk_size => 3, sequence => { begin => 31, end => 39, step => 1 }, user_func => \&user_func }] ); MCE->run(); -- Output task_id 0: seq_n 11: chunk_id 1: wid 1: task_wid 1 task_id 0: seq_n 12: chunk_id 2: wid 2: task_wid 2 task_id 0: seq_n 13: chunk_id 3: wid 1: task_wid 1 task_id 0: seq_n 14: chunk_id 4: wid 2: task_wid 2 task_id 0: seq_n 15: chunk_id 5: wid 1: task_wid 1 task_id 0: seq_n 16: chunk_id 6: wid 2: task_wid 2 task_id 0: seq_n 17: chunk_id 7: wid 1: task_wid 1 task_id 0: seq_n 18: chunk_id 8: wid 2: task_wid 2 task_id 0: seq_n 19: chunk_id 9: wid 1: task_wid 1 task_id 1: seq_n 21: chunk_id 1: wid 3: task_wid 1 task_id 1: seq_n 22: chunk_id 1: wid 3: task_wid 1 task_id 1: seq_n 23: chunk_id 1: wid 3: task_wid 1 task_id 1: seq_n 24: chunk_id 1: wid 3: task_wid 1 task_id 1: seq_n 25: chunk_id 1: wid 3: task_wid 1 task_id 1: seq_n 26: chunk_id 2: wid 4: task_wid 2 task_id 1: seq_n 27: chunk_id 2: wid 4: task_wid 2 task_id 1: seq_n 28: chunk_id 2: wid 4: task_wid 2 task_id 1: seq_n 29: chunk_id 2: wid 4: task_wid 2 task_id 2: seq_n 31: chunk_id 1: wid 5: task_wid 1 task_id 2: seq_n 32: chunk_id 1: wid 5: task_wid 1 task_id 2: seq_n 33: chunk_id 1: wid 5: task_wid 1 task_id 2: seq_n 34: chunk_id 2: wid 6: task_wid 2 task_id 2: seq_n 35: chunk_id 2: wid 6: task_wid 2 task_id 2: seq_n 36: chunk_id 2: wid 6: task_wid 2 task_id 2: seq_n 37: chunk_id 3: wid 5: task_wid 1 task_id 2: seq_n 38: chunk_id 3: wid 5: task_wid 1 task_id 2: seq_n 39: chunk_id 3: wid 5: task_wid 1
It's possible that Perl may create a new code ref on subsequent runs causing MCE models to re-spawn. One solution to this is to declare global variables, referenced by workers, with "our" instead of "my".
Let's take a look. The $i variable is declared with my and being reference in both user_begin and mce_loop blocks. This will cause Perl to create a new code ref for mce_loop on subsequent runs.
use MCE::Loop; my $i = 0; ## <-- this is the reason, try our instead MCE::Loop::init { user_begin => sub { print "process_id: $$\n" if MCE->wid() == 1; $i++; }, chunk_size => 1, max_workers => 'auto', }; for (1..2) { ## Perl creates another code block ref causing workers ## to re-spawn on subsequent runs. print "\n"; mce_loop { print "$i: $_\n" } 1..4; } MCE::Loop::finish; -- Output process_id: 51380 1: 1 1: 2 1: 3 1: 4 process_id: 51388 1: 1 1: 2 1: 3 1: 4
By making the one line change, we see that workers persist for the duration of the script.
use MCE::Loop; our $i = 0; ## <-- changed my to our MCE::Loop::init { user_begin => sub { print "process_id: $$\n" if MCE->wid() == 1; $i++; }, chunk_size => 1, max_workers => 'auto', }; for (1..2) { ## Workers persist between runs. No re-spawning. print "\n"; mce_loop { print "$i: $_\n" } 1..4; } -- Output process_id: 51457 1: 1 1: 2 1: 4 1: 3 process_id: 51457 2: 1 2: 2 2: 3 2: 4
One may alternatively specify a code reference to existing routines for user_begin and mce_loop. Take notice of the comma after \&_func though.
use MCE::Loop; my $i = 0; ## my (ok) sub _begin { print "process_id: $$\n" if MCE->wid() == 1; $i++; } sub _func { print "$i: $_\n"; } MCE::Loop::init { user_begin => \&_begin, chunk_size => 1, max_workers => 'auto', }; for (1..2) { print "\n"; mce_loop \&_func, 1..4; } MCE::Loop::finish; -- Output process_id: 51626 1: 1 1: 2 1: 3 1: 4 process_id: 51626 2: 1 2: 2 2: 3 2: 4
There is an article on the web (search for comp.lang.perl.misc MCE) suggesting that MCE::Examples does not cover a simple simulation scenario. This section covers just that.
The serial code is based off the one by "gamo". A sleep is added to imitate extra CPU time. The while loop is wrapped within a for loop to run 10 times. The random number generator is seeded as well.
use Time::HiRes qw(time sleep); srand(5906); my ($var, $foo, $bar) = (1, 2, 3); my ($r, $a, $b); my $start = time(); for (1..10) { while (1) { $r = rand(); $a = $r * ($var + $foo + $bar); $b = sqrt($var + $foo + $bar); last if ($a < $b + 0.001 && $a > $b - 0.001); sleep 0.002; } print "$r -> $a\n"; } my $end = time(); printf STDERR "\n## compute time: %0.03f secs\n\n", $end - $start; -- Output 0.408246276657106 -> 2.44947765994264 0.408099657137821 -> 2.44859794282693 0.408285842931324 -> 2.44971505758794 0.408342292008765 -> 2.45005375205259 0.408333076522673 -> 2.44999845913604 0.408344266898869 -> 2.45006560139321 0.408084104120526 -> 2.44850462472316 0.408197400014714 -> 2.44918440008828 0.408344783704855 -> 2.45006870222913 0.408248062985479 -> 2.44948837791287 ## compute time: 82.355 secs
Next, we'd do the same with MCE. This demonstration requires at least MCE 1.509 to run properly. Folks on prior releases (1.505 - 1.508) will not see output for the 2nd run and beyond.
use Time::HiRes qw(time sleep); use MCE::Loop; ## Configure MCE. Move common variables inside the user_begin ## block when not needed by the manager process. MCE::Loop::init { user_begin => sub { use vars qw($var $foo $bar); our ($var, $foo, $bar) = (1, 2, 3); }, chunk_size => 1, max_workers => 'auto', input_data => \&_input, gather => \&_gather }; ## Callback functions. my ($done, $r, $a); sub _input { return if $done; return rand(); } sub _gather { my ($_r, $_a, $_b) = @_; return if $done; if ($_a < $_b + 0.001 && $_a > $_b - 0.001) { ($done, $r, $a) = (1, $_r, $_a); } return; } ## Compute in parallel. my $start = time(); srand(5906); for (1..10) { $done = 0; ## Reset $done before running mce_loop { # my ($mce, $chunk_ref, $chunk_id) = @_; # my $r = $chunk_ref->[0]; my $r = $_; ## Valid due to chunk_size => 1 my $a = $r * ($var + $foo + $bar); my $b = sqrt($var + $foo + $bar); MCE->gather($r, $a, $b); sleep 0.002; }; print "$r -> $a\n"; } printf "\n## compute time: %0.03f secs\n\n", time() - $start; -- Output 0.408246276657106 -> 2.44947765994264 0.408099657137821 -> 2.44859794282693 0.408285842931324 -> 2.44971505758794 0.408342292008765 -> 2.45005375205259 0.408333076522673 -> 2.44999845913604 0.408344266898869 -> 2.45006560139321 0.408084104120526 -> 2.44850462472316 0.408197400014714 -> 2.44918440008828 0.408344783704855 -> 2.45006870222913 0.408248062985479 -> 2.44948837791287 ## compute time: 11.290 secs
Well, there you have it. MCE is able to complete the same simulation many times faster.
There are occasions when one wants several workers to run in parallel without having to specify input_data or seqeunce. These two options are optional in MCE. The "do" and "sendto" methods, for sending data to the manager process, are also demonstrated below. Both are processed serially by the manager process on a first come, first serve basis.
use MCE; sub report_stats { my ($wid, $msg, $h_ref) = @_; print "Worker $wid says $msg: ", $h_ref->{"counter"}, "\n"; } my $mce = MCE->new( max_workers => 4, user_func => sub { my ($mce) = @_; my $wid = MCE->wid(); if ($wid == 1) { my %h = ("counter" => 0); while (1) { $h{"counter"} += 1; MCE->do("report_stats", $wid, "Hey there", \%h); last if ($h{"counter"} == 4); sleep 2; } } else { my %h = ("counter" => 0); while (1) { $h{"counter"} += 1; MCE->do("report_stats", $wid, "Welcome..", \%h); last if ($h{"counter"} == 2); sleep 4; } } MCE->sendto("STDOUT", "Worker $wid is exiting\n"); } ); MCE->run; -- Output Note how worker 2 comes first in the 2nd run below. $ ./demo.pl Worker 1 says Hey there: 1 Worker 2 says Welcome..: 1 Worker 3 says Welcome..: 1 Worker 4 says Welcome..: 1 Worker 1 says Hey there: 2 Worker 2 says Welcome..: 2 Worker 3 says Welcome..: 2 Worker 1 says Hey there: 3 Worker 2 is exiting Worker 3 is exiting Worker 4 says Welcome..: 2 Worker 4 is exiting Worker 1 says Hey there: 4 Worker 1 is exiting $ ./demo.pl Worker 2 says Welcome..: 1 Worker 1 says Hey there: 1 Worker 4 says Welcome..: 1 Worker 3 says Welcome..: 1 Worker 1 says Hey there: 2 Worker 2 says Welcome..: 2 Worker 4 says Welcome..: 2 Worker 3 says Welcome..: 2 Worker 2 is exiting Worker 4 is exiting Worker 1 says Hey there: 3 Worker 3 is exiting Worker 1 says Hey there: 4 Worker 1 is exiting
MCE
Mario E. Roy, <marioeroy AT gmail DOT com>
This program is free software; you can redistribute it and/or modify it under the terms of either: the GNU General Public License as published by the Free Software Foundation; or the Artistic License.
See http://dev.perl.org/licenses/ for more information.
To install MCE, copy and paste the appropriate command in to your terminal.
cpanm
cpanm MCE
CPAN shell
perl -MCPAN -e shell install MCE
For more information on module installation, please visit the detailed CPAN module installation guide.