Update: Added chunk_size option to the example.

Update: Added note on one worker reading at any given time.

The example by BrowserUk is awesome. Perhaps you're having to work with Perl not compiled with threads; e.g. running Perl on Solaris. A pattern counter can be done with MCE using the following code. The usage is the same between the two scripts.

This is safe from random I/O thrashing. The read pattern is sequential, not random between workers due to one worker reading at any given time.

#! perl -slw

use strict;
use warnings;

use MCE::Loop;

our $T //= 'auto';

my ( $file_name, $target ) = @ARGV;

MCE::Loop::init {
   use_slurpio => 1, max_workers => $T,
   chunk_size => 1024 * 1024 * 16,
};

my @result = mce_loop_f {
   my ( $mce, $slurped_chunk, $chunk_id ) = @_;
   my $count = 0;

   $count++ while ( $$slurped_chunk =~ /$target/g );
   MCE->gather($count);

} $file_name;

my $total = 0;
$total += shift @result while @result;

print "Found $total '$target' lines";
[download]

Usage:

mce_script.pl -T=n theFile.txt "the string"
[download]