A hierarchical data-driven model for multi-grid problem solving

The data-driven principles of execution present an elegant solution to the problem of instruction scheduling in large scale multiprocessor systems. However, much overhead must be expanded in detecting the simplest forms of parallelism such as vector operations. A two-level architecture with powerful processors is presented here. In addition, preemptive execution of certain actors allows better pipelining and tolerance to latencies. The natural application of the machine to numerical computations is demonstrated by the use of a multi-grid Partial Differential Equation solver.