Would it make sense if that were the case? Think about what we want to show with this benchmark: How does the performance behave if we go through an array not with stride 1 but with a larger stride? N is in the code, n is an exponent that ensures that you will get equidistant points if you plot performance vs. M in a log-linear plot.
The same applies to your previous question about the value of s: Choose a value that does not create any problems; apart from that, it does not matter.