The original original may be Christopher Seiwald's "seven pillars of pretty code", which among other things suggested to do your ifs first and get rid of the rare cases, then handle only the common case with more complicated logic in the bottom part of each function.

An excellent article that you have to search for.

Just because someone published an article on micro-optimization, does that mean we all have to follow it blindly?