Summarizing, we first evaluate an array of constants C[i] = C2i−1 = 2d×2i−1 to be used in layer i for i = 1, 2,...,logn.
This is easily done noticing that C[1] = C and C[i + 1] = C[i]2 for i ≥ 1. The parallel procedure for the higher layers is then
given in Fig. 4.