Probably. In addition, there's a lot of literature on optimizing hardware implementations of fundamental arithmetic operations like addition and multiplication

Probably. In addition, there's a lot of literature on optimizing hardware implementations of fundamental arithmetic operations like addition and multiplication. I recall seeing a paper a while ago which talked about reducing the number of gates by allowing some bounded imprecision in the results - unfortunately, I don't remember the title right now, but it sounds like that's what they may be doing.