kinda? this is sort of a direct result of the 'vectors are just sliced registers' model. if i do a pairwise operation and divide my domain by 2 at each step, is the resulting vector sparse or dense? if its dense then I only really top out when i'm in the last log2slice steps.