Analysis of data flow models using the SARA graph model

A number of investigators have continued to discuss application of asynchronous techniques to rmprove the computational power of computing systems. 2.12,5,9 In fact the need for asynchronous design techniques arose in the earliest machines 26,6.2o which introduced parallel handling of bits in a number and overlapping of independent operations. The concept of distributed autonomous concurrent processors was essential to the visionary architecture proposed by Holland16 to support "operating programs floating in a sea of hardware." The importance of such concurrent and asynchronous systems has increased recently because of the availability of entire processing units on one or a few chips and the potential cost reduction of those units. Moreover, it is expected that work on VLSI technology will permit even more complex systems to be reduced to silicon if their design and market analysis have been validated sufficiently to justify the cost. The need for validation prior to costly physical implementation has increased the value of methods and tools which support modeling and analysis. SARA (System ARchitects Apprentice) is such a supported methodology particulary applicable to multilevel modeling of concurrent systems. Dennis has proposed computation structures for concurrent system design which have aroused considerable interest. This work shows how SARA tools may be used to analyze data flow models utilizing Dennis' computation structure. This work refers to the latter as the Dennis Data Flow (DDF) Model. In the first part of this paper we establish relationships between primitives of two models-the Dennis Data Flow Model and the SARA Graph Model of Behavior (GMB). The second part of the paper presents an example showing how SARA tools can be used to construct and simulaie a data flow model. The discussion and the example should help to increase the reader's understanding of both models. This opens the way for application of supported multilevel design methodologies, like that supported by SARA,8 to higher-level design of systems incorporating devices which implement data flow primitives.