Yep, it is very fast, but why is that better than this:

    open(F, "<", $file) or die "$file: $!";
    binmode(F);
    undef $/; # switch off end-of-line separating

    # read file in large chunks
    while (<F>) {
        while ( m/$re/oigsm ) {
            print "$1\n";
        }
    }

    $/ = '\n'; # switch back to line mode
    close(F);
[download]

?

Thanks,
Tamas