bird - 10:32 pm on Sep 11, 2006 (gmt 0) In all other cases: If it ain't broken, don't fix it.
Ballparks aren't useful information in this context. You can't optimize your code based on what other people measured with their (completely different) code. If you have a program where you think it doesn't run fast enough, profile it, and then you know where your bottlenecks are. Only with this very specific information can you take any useful countermeasures.
In all other cases: If it ain't broken, don't fix it.