What about trying to improve and tweak the code which is used to create these benchmarks?
http://shootout.alioth.debian.org
Many Tcl'ers already did a great job, but there are still some scripts missing and some others might have room for optimisations.
The disciplines: