Fusing Dataflow with Finite State Machines

by Bilung Lee

Memorandum UCB/ERL M96/20, EECS, University of California, Berkeley, CA, USA 94720.

[PS] [PDF]


The dataflow model of computation has been used extensively in signal processing design, and is particularly convenient for numeric-intensive computation of applications. Finite state machines (FSMs) have been developed to solve a different class of problems, namely sequential control. In this project, we propose to hierarchically nest the dataflow and FSM models of computation. With the two models mixed, concurrency and hierarchy are naturally supported in a manner similar to hierarchical FSMs, like Statecharts. This provides a clean and simple mechanism for describing systems that combine sophisticated signal processing with sophisticated control. We implement the ideas in the Ptolemy software environment, which has been under development at University of California at Berkeley.