represents. If so, the index of the start node of the link and the link’s weight are stored in a RAM within the PE referred to as the adjacency RAM. Thus, the input graph representation is distributed among the PEs such that each PE stores its incoming links in its adjacency RAM