I'm playing around with the OP's code. I add random code `j = sin(j)` in the loop. Now python takes 21 second., nim takes about 18.6 seconds. For python, the time increases by 5 seconds, and for nim, it increases by only 0.6 seconds. I think we can see if more complicated code is added, nim is likely to win. As many others have mentioned, we need a real problem to do the benchmark.
I use julia for scientific computing. The same thing happens. If we just try simple examples, python with numpy will win, but for real work, julia will be much faster. As Araq said, benchmarking isn't that simple.
