back
1 comments
Author could do worse than to read http://jvns.ca/blog/2016/07/23/rigorous-benchmarking-in-reas...

I think another good benchmarking resource was on HN recently but I didn't find it right away.

fwiw, in my own benchmark, pypy, python2, python3, and c++ were all within a factor of 3 (6-18us per fib(300) on an i5-3320M) using 100,000 repetitions (timeit's autotuned repetition count). Since most of the time in fib.cr was probably startup time, it's not directly comparable to any of their figures (691us to invoke fib.cr and calculat fib(42) once)

.. and I see now that the "big" fib was 300_000 and not 300