Although glob performs well for small word sizes, for larger sizes a hand-rolled loop is much faster:

#! perl
use strict;
use warnings;
use Benchmark qw(cmpthese);

cmpthese(1, {
    'loop' => sub { my @p = permute_loop(10); print 'loop: ', scalar @
+p, "\n"; },
    'glob' => sub { my @q = permute_glob(10); print 'glob: ', scalar @
+q, "\n"; },
});

sub permute_glob
{
    my ($size) = @_;
    my  $r     = 'A,T,G,C';

    return glob "{$r}" x $size;
}

sub permute_loop
{
    my ($size) = @_;
    my  @a     = qw(A T G C);

    while (--$size)
    {
        @a = map { $_ . 'A', $_ . 'T', $_ . 'G', $_ . 'C' } @a;
    }

    return @a;
}
[download]

Output:

glob: 1048576
            (warning: too few iterations for a reliable count)
loop: 1048576
            (warning: too few iterations for a reliable count)
     s/iter  glob  loop
glob    281    --  -98%
loop   4.60 6011%    --
[download]

Hope that helps,