Function-Parallel Computation in a Data-Parallel Environment

Asynchromus problems are those which may be decomposed into a set of independenr sub-tasks which are suitable for concurrent execution. Th function paraIIeIism of these problems cannot normally be direcrly expressed using the data-parallel programming model. In this paper, data distribution strategies have been explored which allow an asynchronous problem to be implemented which function-parallelism in a data-parallel environment. When a problem can be implemented using both function-parallelism and data-parallelism, there are tradeoffs in using either approach. We have investigated the optimal balance between function-parallelism and data-parallelism for an asynchronous problem.