Computer aided dataflow analysis for the conversion of sequential programs into parallel form

The need for more processing power has spurred the development of new processor architectures. A promising branch of research in this area is devoted to prototypes of a new type of computer organization, based on the dataflow concept. While powerful and attractive through its simplicity, major milestones still have to be passed before such machines reach industrial maturity. However, the principles of dataflow analysis provide a sound basis for the detection and scheduling of parallel operations from ordinary sequential programs. In this paper the basic dataflow architectures are discussed, It is shown that the dataflow principles can be applied to the analysis of sequential programs, with a "conventional" shared memory multiprocessor as target machine. This is realized in the LEM-analyzer /simulator and will be illustrated by several numericalapplications.The need for more processing power has spurred the development of new processor architectures. A promising branch of research in this area is devoted to prototypes of a new type of computer organization, based on the dataflow concept. While powerful and attractive through its simplicity, major milestones still have to be passed before such machines reach industrial maturity. However, the principles of dataflow analysis provide a sound basis for the detection and scheduling of parallel operations from ordinary sequential programs. In this paper the basic dataflow architectures are discussed, It is shown that the dataflow principles can be applied to the analysis of sequential programs, with a "conventional" shared memory multiprocessor as target machine. This is realized in the LEM-analyzer /simulator and will be illustrated by several numericalapplications.B