- remove the memoized versions (obviously it's faster)
- show the executed instructions for each language of the main loop (which would be a nice exercise with all the vms and dynamic languages)