Compact DAG representation and its dynamic scheduling

Scheduling large task graphs is an important issue in parallel computing. I n this paper we tackle the two following problems: (I) how to schedule a ta sk graph, when it is too large to fit into memory? (2) How to build a gener ic program such that parameter values of a task graph can be given at run-t ime? Our answers feature the parameterized task graph (PTG), which is a sym bolic representation of the task graph. We propose a dynamic scheduling alg orithm which takes a PTG as an entry and allows us to generate a generic pr ogram. We present a theoretical study which shows that our algorithm finds good schedules for coarse-grain task graphs, has a low memory cost, and st low computational complexity. When the average number of operations of each task is large enough, we prove that the scheduling overhead is negligible with respect to the makespan. We also provide experimental results that dem onstrate the feasibility of our approach using several compute-intensive ke rnels found in numerical scientific applications. (C) 1999 Academic Press.