But being strictly picky (for this size of operation at least), isn't it cheaper (disk/cpu wise) to do my own small recursion?

Uh, File::Find is a particularly well-written piece of code, and has optimizations to ensure that it do unnecessary stat'ing (an optimization that made File::Find actually beat /usr/bin/find on Sunos 4.1.3 on an NFS-mounted directory.

Also, File::Find means your program won't chase symlinks incorrectly (did you think of those?) and your program will be portable to non-Unix platforms trivially.

So, don't reinvent your own directory recursing code. Use File::Find.