for my $m (m/$regex/g) { # Update: reworked 
            $matched{$m} = 1;
            print "$file $_";
        }
[download]

Full code:

#!/usr/bin/env perl

use strict;
use warnings;

my @files = <c:/perl64/myfiles/*>;

# record matching regexes
our %matched;

my @nums = ('1203', '1204', '1207');
my $regex = '\b(' . join('|', @nums) . ')\b'; # update: fixed + to .

for my $file ( @files ) {

    open my $file_h, '<', $file or die "Can't open $file: $!";

    while ( <$file_h> ) {
        for my $m (m/$regex/g) { # Update: reworked 
            $matched{$m} = 1;
            print "$file $_";
        }
    }
}

# Check all nums have been seen
for my $num (@nums) {
    if (not exists($matched{$num})) {
        print "$num not found\n";
    }
}
[download]

What Anonymous Monk said:

while (my ($match) = m/$regex/g) {
    $matched{$match} = 1;
    print "$file $_";
}
[download]

gives you an infinite loop.

c:\@Work\Perl\monks>perl -wMstrict -le
"my $regex = qr{ \b f[eio]e \b }xms;
 ;;
 $_ = 'xx fee fie foe yy';
 while (m/($regex)/g) {
   my $match = $1;
   print qq{captured '$match'};
   }
"
captured 'fee'
captured 'fie'
captured 'foe'
[download]

works better (at least it stops). If you want to get a bit fancy,

while (m/($regex)/g and my $match = $1) {
  print qq{captured '$match'};
  }
[download]

also works. But the whole "extract multiple matches per line" thing may not even be a real issue! The OPer says nothing about it, and I only raised it as a cautionary point of interest.

---

Give a man a fish:  <%-{-{-{-<

Fixed original.

-QM
--
Quantum Mechanics: The dreams stuff is made of

this matches, but loops infinitly on the same line print, the script never finishes