Meh, at least it's sublinear latency with the number of bits. (I'm assuming, based on the description. Unless there's something I'm missing, you've got log2(d)

Meh, at least it's sublinear latency with the number of bits. (I'm assuming, based on the description. Unless there's something I'm missing, you've got log2(d) stages of O(1) latency.) The nasty operations are the ones that are linear (or worse!).