use Sort::Key::Top 'natpos';

use Benchmark qw(cmpthese);

my @f = map rand, 0..1e7;

my ($r1, $r2);

cmpthese -1, { sort => sub { $r1 = (sort { $a <=> $b } @f)[@f/2] },
               skt  => sub { $r2 = natpos @f/2, @f } };

print "median $r1, $r2\n";


##</code><code>##

$ perl /tmp/median.pl 
     s/iter  sort   skt
sort   20.5    --  -94%
skt    1.25 1542%    --
mediam 0.499785519518444, 0.499785519518444