Coarse grained parallel maximum matching in convex bipartite graphs

We present a coarse grained parallel algorithm for computing a maximum matching in a convex bipartite graph G=(A,B,E). For p processors with N/p memory per processor, N=|A|+|B|,N/p/spl ges/p, the algorithm requires O(log p) communication rounds and O(T/sub sequ/(n/p,m/p)+n/p log p) local computation, where n=|A|,m=|B| and T/sub sequ/(n,m) is the sequential time complexity for the problem. For the BSP model, this implies O(log p) supersteps with O(gN+gn/p log p) communication cost and O(T/sub sequ/(n/p,m/p)+n/p log p) local computation.