The article mentions mawk which is one of the faster awks I know of. Then there is awka that compiles awk script to C. Do you know of other speed demons apart from these. I think cleaning up the language by removing some of the one-off constructs and making it more regular (npi) could make faster implementations possible. An embeddable/extendable awk that exploits simd and multicore could potentially lift a lot of weight.

I wish awk had more mind share than it does these days.