NAME

MCE - Many-Core Engine for Perl providing parallel processing capabilities

VERSION

This document describes MCE version 1.799_02

Many-Core Engine (MCE) for Perl helps enable a new level of performance by maximizing all available cores.

MCE

DESCRIPTION

MCE spawns a pool of workers and therefore does not fork a new process per each element of data. Instead, MCE follows a bank queuing model. Imagine the line being the data and bank-tellers the parallel workers. MCE enhances that model by adding the ability to chunk the next n elements from the input stream to the next available worker.

Bank Queuing Model

SYNOPSIS

This is a simplistic use case of MCE running with 5 workers.

# Construction using the Core API

use MCE;

my $mce = MCE->new(
   max_workers => 5,
   user_func => sub {
      my ($mce) = @_;
      $mce->say("Hello from " . $mce->wid);
   }
);

$mce->run;

# Construction using a MCE model

use MCE::Flow max_workers => 5;

mce_flow sub {
   my ($mce) = @_;
   MCE->say("Hello from " . MCE->wid);
};

The following is a demonstration for parsing a huge log file in parallel.

use MCE::Loop;

MCE::Loop::init { max_workers => 8, use_slurpio => 1 };

my $pattern  = 'something';
my $hugefile = 'very_huge.file';

my @result = mce_loop_f {
   my ($mce, $slurp_ref, $chunk_id) = @_;

   # Quickly determine if a match is found.
   # Process the slurped chunk only if true.

   if ($$slurp_ref =~ /$pattern/m) {
      my @matches;

      # The following is fast on Unix, but performance degrades
      # drastically on Windows beyond 4 workers.

      open my $MEM_FH, '<', $slurp_ref;
      binmode $MEM_FH, ':raw';
      while (<$MEM_FH>) { push @matches, $_ if (/$pattern/); }
      close   $MEM_FH;

      # Therefore, use the following construction on Windows.

      while ( $$slurp_ref =~ /([^\n]+\n)/mg ) {
         my $line = $1; # save $1 to not lose the value
         push @matches, $line if ($line =~ /$pattern/);
      }

      # Gather matched lines.

      MCE->gather(@matches);
   }

} $hugefile;

print join('', @result);

The next demonstration loops through a sequence of numbers with MCE::Flow.

use MCE::Flow;

my $N = shift || 4_000_000;

sub compute_pi {
   my ( $beg_seq, $end_seq ) = @_;
   my ( $pi, $t ) = ( 0.0 );

   foreach my $i ( $beg_seq .. $end_seq ) {
      $t = ( $i + 0.5 ) / $N;
      $pi += 4.0 / ( 1.0 + $t * $t );
   }

   MCE->gather( $pi );
}

# Compute bounds only, workers receive [ begin, end ] values

MCE::Flow::init(
   chunk_size  => 200_000,
   max_workers => 8,
   bounds_only => 1
);

my @ret = mce_flow_s sub {
   compute_pi( $_->[0], $_->[1] );
}, 0, $N - 1;

my $pi = 0.0;  $pi += $_ for @ret;

printf "pi = %0.13f\n", $pi / $N;  # 3.1415926535898

CORE MODULES

Three modules make up the core engine for MCE.

MCE::Core

Provides the Core API for Many-Core Engine. The various MCE options are described here.

MCE::Signal

Temporary directory creation, cleanup, and signal handling.

MCE::Util

Utility functions for Many-Core Engine.

MCE EXTRAS

There are 4 add-on modules for use with MCE.

MCE::Candy

Provides a collection of sugar methods and output iterators for preserving output order.

MCE::Mutex

Provides a simple semaphore implementation supporting threads and processes.

MCE::Queue

Provides a hybrid queuing implementation for MCE supporting normal queues and priority queues from a single module. MCE::Queue exchanges data via the core engine to enable queuing to work for both children (spawned from fork) and threads.

MCE::Relay

Enables workers to receive and pass on information orderly with zero involvement by the manager process while running.

MCE MODELS

The models take Many-Core Engine to a new level for ease of use. Two options (chunk_size and max_workers) are configured automatically as well as spawning and shutdown.

MCE::Loop

Provides a parallel loop utilizing MCE for building creative loops.

MCE::Flow

A parallel flow model for building creative applications. This makes use of user_tasks in MCE. The author has full control when utilizing this model. MCE::Flow is similar to MCE::Loop, but allows for multiple code blocks to run in parallel with a slight change to syntax.

MCE::Grep

Provides a parallel grep implementation similar to the native grep function.

MCE::Map

Provides a parallel map model similar to the native map function.

MCE::Step

Provides a parallel step implementation utilizing MCE::Queue between user tasks. MCE::Step is a spin off from MCE::Flow with a touch of MCE::Stream. This model, introduced in 1.506, allows one to pass data from one sub-task into the next transparently.

MCE::Stream

Provides an efficient parallel implementation for chaining multiple maps and greps together through user_tasks and MCE::Queue. Like with MCE::Flow, MCE::Stream can run multiple code blocks in parallel with a slight change to syntax from MCE::Map and MCE::Grep.

MISCELLANEOUS

Miscellaneous additions included with the distribution.

MCE::Examples

Describes various demonstrations for MCE including a Monte Carlo simulation.

MCE::Subs

Exports functions mapped directly to MCE methods; e.g. mce_wid. The module allows 3 options; :manager, :worker, and :getter.

REQUIREMENTS

Perl 5.8.0 or later. PDL::IO::Storable is required in scripts running PDL.

SOURCE AND FURTHER READING

The source, cookbook, and examples are hosted at GitHub.

SEE ALSO

MCE::Shared provides data sharing capabilities for MCE. It includes MCE::Hobo for running code asynchronously.

AUTHOR

Mario E. Roy, <marioeroy AT gmail DOT com>

COPYRIGHT AND LICENSE

Copyright (C) 2012-2016 by Mario E. Roy

MCE is released under the same license as Perl.

See http://dev.perl.org/licenses/ for more information.