Comments by "LoneTech" (@0LoneTech) on "Numberphile2" channel.

Python has big integers, and Pypy can compile it. C or assembler don't, so you'll be calling an arbitrary precision library like gmp, and the difference becomes insignificant after you hit big numbers. Close to metal isn't magic. That said, try Julia or Haskell. And for a task like this, maybe Decimal instead of Integer.
7
@00O3O1B If you're doing that level of assembly work (I have), the answer is typically get a bigger microcontroller or stop toying with the CPU and use the GPU already. Which can be eased greatly with e.g. Clyther or OpenAcc. Searches like these are embarassingly parallel so distributing it easy, and offers gains in thousands compared to cpu fiddling often 2-5x. Algorithmic work is often more significant still.
2