It looks like a joke to me. The program always just calls exit(0); straight away if the main function is the expected one, which is why it's so fast. All the rest is just noise to hide that. All he seems to be saying is that the competition could be won by implementing fastroot() as just calling exit(0).

Thanks. The choice of ;=) as the operator did make me suspicious.

Andy Wingo linked this on his blog, with this cryptic title and no elaboration. I read it, but I'm not sure I got it. What's the deal?

This isn't really proof that Scheme will be faster than C, and in fact doesn't really mention Scheme at all, except to show that it is a language that supports call-with-current-continuation as a primitive operation.

What I got from the paper is that using continuations with intensional equality (that is to say their contents are the same) can be used to dramatically speed up a program. And so, a primitive way to create intensional equal continuations would be a potentially worthwhile primitive for a language.