I'm sure the people who made those optimisations took very good benchmarks before they made them. And, most likely, they do work under the right circumstances.

However, to make a generalisation/optimisation of my own, in 99% of the cases, trying to write overly clever code like that is a waste of time and does not result in faster code - though it does obscure your code and make it harder to optimise later when you actually need to. It would probably be better for 90+% of the people who will stumble on this article to never have read it.