On the complexity of scheduling with large communication delays

Abstract Given a directed acyclic graph (dag) with unit execution time tasks and constant communication delays c ⩾ 2, we are interested in deciding if there is a schedule for the dag of length at most L. We prove that the problem is polynomial when L is equal to (c + 1), or (c + 2) for the special case of c = 2, and that it is NP-complete for (c + 3) for any value of c, even in the case of a bipartite dag of depth one.