When passing from time step i to i +1, half of the processors working on each of the currently processed consecutive chunks j, j −1, ..., j −log n are reassigned to the new chunk to be processed, indexed j +1, while the other half remains with the chunk they started with and pass to a higher layer. More precisely, all the processors with indices ≥n, that is, whose binary representation starts with 1 (those in the bold rectangles in Fig.3), are assigned to the new chunk, while those with indices