Scheduling task graphs containing branches on parallel systems

The problem of scheduling task graphs with conditional branching is considered one of the most difficult problems in scheduling parallel programs on multiprocessor computers. The major problem in having branches in task graphs is the non-determinism, since the direction of a branch may be unknown until the program is midway in execution. The authors overcome the problem of non-determinism by proposing a new probabilistic model that distinguishes between branch and precedence relations in parallel programs. They present two different approaches for solving this problem. In the first approach, a schedule is obtained by merging the schedules generated for several possible task graphs. The second technique generates a schedule for a single deterministic task graph that approximates all possible task graphs.