I take it it's the inlining of "errinvoke" that makes all the
difference ;-)
Meh. I question the validity of a programming language shootout where
everyone's just optimising their code to the hilt, though. The
ability of a language to support *highly manually optimised code* is,
I think, a questionable measurement of its worth, although it would
be interesting to conduct a benchmark where lots of programmers
submit their own implementations of an algorithm, *along with how
many hours it took them to do and how many months of experience they
had with that language*, and the gradients of the resulting best-fit
lines be compared instead of any absolute value :-)