I would suggest actually looking at a sieve, for example the ones on RosettaCode. They are much faster than the trial division method you link to, albeit generating primes starting at two.
For doing trial division, the skip-2-3 primality test from RosettaCode is twice as fast as the code you link to. Just wrap in something like: my @primes = grep { isprime($_) } 1..1000000;
While the regex solution mentioned in another reply is super clever, it is also super not useful for the task once past toy sizes.
