What GA/GP lib do you use and why?
None, because I haven't used genetic algorithms in all the projects I worked with since college.
In college, I rolled my own of course :) - did some weird variation, with variable length genomes.
I have completed two graduate level research projects using genetic algorithms to automatically design electric motors.
For the first project, I used the EO Evolutionary Computation Framework which is based almost entirely on C++ template programming. It is very powerful but requires (IMO) a lot of concentration to extend it because of the meta nature of the templates and the corresponding nonsense error messages from the compiler.
For the second project, I ported everything over to the Open BEAGLE EC Framework which is based on C++ inheritance instead of templates. It also has a very complete XML-based system for controlling parameters and storing milestones and results. I found it much easier to extend and adapt to my specific needs (mainly mixed integer and float design parameters with manufacturing tolerances).
I also briefly used Open BEAGLE in a simple Genetic Programming experiment and it proved easy to use there as well.
If you are using Python, pyevolve is easy to use. I used it with success. http://pyevolve.sourceforge.net/