Designing systems for highly parallel and distributed computing

The parallel computational model with data flow sequencing is introduced. Description of the basic principles and features of mpDF, a massively parallel architecture, based on the dataflow operational model and RISC organizational principles is presented. The analytical performance model is developed to evaluate proposed architectural solution for distributed/parallel computing based on data flow sequencing of instructions. Algorithmic performance model is extended to include characterization of the parallel programs in terms of the average parallelism. Model is solved for number of different workloads. The values of the basic architectural characteristics are analyzed showing that architecture is well balanced providing consistent performance made for wide rage of parallel applications.<
>