When I'm working on performance, I measure twice and cut once. I know measuring is 'hard' as there's a lot of code optimised for x86 in VS, but saying 'It would be slower built for 64 bit' is cutting before measuring even once.

I have the greatest of respect for the author, but everyone needs to be exposed to 'trust, but verify' at least occasionally, just in case they happen to be wrong this once.