A performance evaluation of CP list scheduling heuristics for communication intensive task graphs

List-based priority schedulers have long been one of the dominant classes of static scheduling algorithms. Such heuristics have been predominantly based around the "critical path, most immediate successors first" (CP/MISF) priority. The ability of this type of scheduler to handle increased levels of communication overhead is examined in this paper. Three of the more popular list scheduling heuristics, HLFET and ISH and DSH, plus the mapping heuristic are subjected to a performance based comparison, with results demonstrating their inadequacies in communication-intensive cases. Performance degradation in these instances is partly due to the level alteration problem, but more significantly to conservative estimation of communication costs due to the assumption of zero link contention. The significance of this component of communication is also examined.