Note that the primes subroutine is quite inefficient and returns 1 as well, which is usually not considered prime.

Here's a faster one:

#! /usr/bin/perl
use warnings;
use strict;
use feature qw{ say };

sub primes {
    my $n = shift;
    return if $n < 2;

    my @primes = (2);
    for my $i (3 .. $n) {
        my $sqrt = sqrt $i;
        my $notprime;
        for my $p (@primes) {
            last if $p > $sqrt;

            $notprime = 1, last if 0 == $i % $p;
        }
        push @primes, $i unless $notprime;
    }
    return @primes
}


use List::Util qw{ sum };
sub primes_la {
    # Copy your code here.
}

use Test::More tests => 1;
is_deeply([1, primes(10000)], [primes_la(10000)], 'same');

use Benchmark qw{ cmpthese };

cmpthese(-10,
        { ch => 'primes(10000)',
          la => 'primes_la(10000)',
        });
__END__
1..1
ok 1 - same
     s/iter     la     ch
la     1.35     --   -99%
ch 1.06e-02 12662%     --
[download]