Exploration of design space for parallel programs

A key point in rapid system prototyping is to be able to incorporate into the design and development process specification of hardware and software systems in a more-or-less uniform framework. This paper presents a representation basis for parallel computation structures which covers both hardware and software systems, which is declarative and in which specification of components and structures is well separated. The top level and most abstract specification of the representation basis is: (a) All parallel computation structures can be represented as an extended form of directed graphs called data/control flow graphs (DCGs) with appropriate semantics chosen for the nodes and the arcs. The nodes are transformational components and the arcs are communication components. (b) All executions of parallel computation structures can be described as traversals of these graphs by appropriately defined transactions. (c) There exist straightforward algorithms for translations between this and other representations of parallel computation systems such as equational or production rule programs or finite state machines