Formalized Data Flow Diagrams and Their Relation to Other Computational Models

Symanzik, Jürgen
Baker, Albert
Journal Title
Journal ISSN
Volume Title
Source URI
Research Projects
Organizational Units
Computer Science
Organizational Unit
Journal Issue

Formalized Data Flow Diagrams and Their Relation to Other Computational Models Juergen Symanzik and Albert L. Baker One approach to the formalization of Data Flow Diagrams (DFD's) is presented by Coleman and Leavens, et al. These Formalized Data Flow Diagrams (FDFD's) can be viewed as another model of computation. This paper contains an analysis of the computational power of these FDFD's. We first consider the issue whether certain features of FDFD's affect their computational power. A Reduced Data Flow Diagram (RDFD) is an FDFD with no stores, finite domains for flow values, and no facility for testing for empty flows, but it may contain persistent flows. An RDFD without persistent flows is called a persistent flow--free Reduced Data Flow Diagram (PFF--RDFD). We show that PFF--RDFD's are Turing equivalent. The other features of FDFD's only add to the expressive power of FDFD's. Therefore, any FDFD can be expressed as an PFF--RDFD. Our proof that PFF--RDFD's are Turing equivalent procedes as follows. We first show that any RDFD can be simulated by a FIFO Petri Net. We then show that any Program Machine can be simulated by an PFF--RDFD. It is known that FIFO Petri Nets and Program Machines both are Turing equivalent.

<p>© Copyright 1996 by Jürgen Symanzik and Albert L. Baker. All rights reserved.</p>
Computational Power, Turing Machine, FIFO Petri Net, Program Machine