I'm not sure if it helps, but for 130 bits, we only need to divide the point by 2, 101 times. The issue is figuring out where to subtract 1 to avoid floating-point errors.