Dynamic Scheduling on Distributed-Memory Multiprocessors

The problem of scheduling a set of applications on a multiprocessor system has been investigated in a number of diierent points of view. This paper describes our work on the scheduling problem at the user level, where we have to distribute evenly the parallel tasks that compose a program among a set of processors. We investigated dynamic scheduling heuristics applied to loops on distributed-memory multiprocessors. Our approaches try to combine the workload balancing goal with the data locality exploitation, as many scientiic applications exhibit some processor aanity when executing on a distributed-memory machine. Some experimental results on a CM-5 comparing our heuristics with some other approaches are presented.