Algorithm 2. Parallel EXTEND.

1:  function PARALLEL_EXTEND(m, pz, po, pi)

2:   l = len(m)

3:   ml+1.(d, z, o, w) = (pi, pz, po, l = 0 ? 1 : 0)

4:   for $i\leftarrow 2$ to l + 1 in parallel, do

5:    left_w = shuffle(mi.w, i − 1)

6:    mi.w = mi.w ⋅ pz ⋅ (l + 1 − i)/(l + 1)

7:    mi.w = mi.w + po ⋅ left_w ⋅ i/(l + 1)

8:   return m