The two I'd reach for are Press, Flannery, Teukolsky, & Vetterling, "Numerical Recipes in C, 2nd Ed.", and Acton's "Numerical Methods That Work".
NR is a pretty good book. The code isn't of the same quality, but it's fine for light-duty use if you can stomach the way they use globals.
Long ago I built some shared libraries of the NR routines, and have been using them off and on ever since. My code is mostly prototype stuff--I really wouldn't use NR code in a product, even if it weren't for the licensing issues.
Acton is a super good read about numerical methods generally. It's more of a lore book than a textbook, and I love lore books, as you know. He has zero patience for mathematically neat methods that are useless in practice, e.g. root-squaring methods for solving equations. There are also a bunch of fun war stories.
Cheers
Phil Hobbs