No, they should be constant. RANDSIZL is always 8 (unless one were using non-ub8 types in the mm/randrsl arrays), RANDSIZ is the length of seed/output array, generally 256, but it should work with any multiple of 8.
I presume using m[i], r[i], i+=1 rather than *(m++), *(r++) is slower, but it should not be this kind of magnitude. I will play around with it a bit more.
