In Linux ( unsure for Windows, any knowledge regarding is appreciated too ), a technique named lazy binding is used for boosting performance when only a small portion of shared library is required, imagine when you invoke a program with -h –help arg, etc.

I was wondering how is it implemented? is it related to lazy allocation technique of OS virtual memory management? i.e., instead of loading all instructions into memory, a mapping to unmapped memory is conducted first, and the valid memory is loaded only when it is really needed/referred during run time.