sulfericacid,
Have you ever read the article I wrote on iterators for perl.com? In there I give the example of needing to break an extremely simple password. You may also be interested in Demonstrate Weakness of "Standard Format" Passwords in which I showed why even long passwords that conform to some standard become very weak. If you are interested in the math involved, please have a look at Re: How do I generate a sequential string? and consider using a module like Algorithm::Loops.

If you seriously wanted to crack passwords, you would write the code in C (or the like) and not perl. Additionally, using the dictionary and variations before going to brute force is certainly something you want in your arsenal. I say variations because people tend to replace zero for the letter O or put a symbol like _ or - between two short dictionary words or add numbers to the end. This human nature flaw is one you should exploit^*. The divide and conquer approach is good but there is only so much improvement that can be done on a single CPU. Writing distributed algorithms is better suited for other languages like Erlang. If you don't know what Rainbow Tables are then look them up, it will make your attempt of caching all possible 3 character passwords seem like child's play.

* Update: Lest anyone assume I feel cracking passwords is ethical, let me be clear - I don't. sulfericacid has indicated that this is for educational purposes and that is all I support. In Re: Character Combinations I point out how silly brute force cracking can be (555 years for the problem presented). I in know way think any of these techniques should be used to gain unauthorized access nor do I feel you should engage in such practices without authorization even if you feel proving the weakness is for everyone's own good. You can ask a well known monk how well that worked out for him.