This sort of thing is in fact exactly what I've been doing for the last year. (Took a widely used, but no longer maintained, program and sped/scaled up the imp

This sort of thing is in fact exactly what I've been doing for the last year. (Took a widely used, but no longer maintained, program and sped/scaled up the important parts by 2-3+ orders of magnitude; this will soon become the official new version.) I'm looking for other high-impact optimization opportunities, so if you've got any problems which currently require supercomputers but you have a nagging suspicion that they shouldn't, that nagging feeling is probably right and I'd love to hear from you.