Scheduling of multiprocessor tasks for numerical applications

The authors investigate the efficient implementation of algorithms with a two-level parallelism on distributed memory machines. They consider parallel specifications consisting of an upper level of multiprocessor tasks each of which having an internal structure of uni-processor tasks. To achieve an optimal parallel execution time, the parallel execution of such a program requires an optimal scheduling of the multiprocessor tasks and an appropriate treatment of uni-processor tasks. In particular they consider an important class of parallel programs that are generated within a specific parallel programming model designing group-SPMD programs for scientific computing. They show how the costs of data redistributions between M-tasks can be taken into consideration and how the special structure of the resulting program can be exploited by using a simple approximation algorithm with a provable good performance.