as a slave co-processor, the host issues instructions to the SPCU through a 2-bit-wide instruction bus. Data transfer between the host and the SPCU occur through a bidirectional data bus, whose width depends on the number of bits required to represent a node (which in turn depends on the maximum number of nodes) and the number of bits required to represent a link’s weight. In the present implementation, we use 8 bits to represent the weight of a link.