What are you referring to when you say it's "fundamentally computationally inefficient"? It's pretty efficient because it's content-addressed, plus optimizations to reduce storage and data transfer with packfiles.
I suspect they were referring to some of the things git allows for non centralized version control. There are simplifications if you just wanted a centralized system like cvs had.