Ok, lots of good advice here however there is an important factor thus far not really covered.
I use subs to simplify complex code. I dont mean in the obvious manor where subs can offer logical partitioning but more on the conceptual level. I am a man with a small brain (damn but if pooh does not have a comment for all eventuallities) When dealing with a particularly thorny programming challenge I avoid mapping the problem into the language space as is often done, instead I map the language into the problem space. Consider the nature of the problem you are solving deduce the language of the problem. Code subs to allow you to talk that language; addVector, splitStringByChar. Many of these subs could often contain a single line of code. Then solve your problem using the language of those subs. This frees you from the petty detail of the underlying -programming- language. Once the problem is solved and working the depreciate redundant subs by moving the code from the sub into the main body of the code.
This is an idea most often seen in OO programming and it is very powerful (most good programmers do it without meaning too), it is appropriate for application and algorithmic centric problems.
jplindstrom touched on this idea in his node in this thread, I agree COMPLETELY with what he said it is dangerous to fractionalise code into many small and useless subs. It is important to rationalise the need for each sub at the end and deprecate where required.