Function multi-versioning in GCC 6

Ah, yes, I see the first example did in fact do this, and I had forgotten about it because I had the second example firmly wedged in my mind by the end of the article.

But I still think there could be a problem. These inline functions provide basic functionality that is used in a number of different higher level functions. The cpu type annotation needs to propagate up from the static inline functions to the functions calling them, and it's the callers that should end up with multiple versions rather than the annotated static inline functions. I wonder if gcc-6 will do that. It's time to do some testing...