\input{headjpdc} \onecolumn \input{defs} \input{hyph} \begin{document} \title{ A Compositional Approach to Graphically Based Parallel Programming% \footnotemark[1] } \author{ J.C. Browne, J.Werth, Ravi Jain, Peter Newton\\ \\ Department of Computer Sciences\\ University of Texas, Austin, Texas 78712\\ browne@cs.utexas.edu 512-471-9579 } \footnotetext[1]{ Paper to appear in {\it Proc. 26th Hawaii Intl. Conf. Sys. Sci.}, Hawaii, Jan. 1993. } \date{} \maketitle \begin{centering} \section*{Abstract} \end{centering} CODE (Computationally Oriented Display Environment) [BRO89] is a graphical parallel programming environment which is based on a formal model of parallel computation. This model of parallel computation defines units of computation and relationships among units of computation. In the graphical environment the units of computation are annotated nodes in a hypergraph and the relationships are specified as annotated arcs or hyperarcs. Basing the programming language on a formal model of computation leads to the possibility of an alternative representation for parallel computation structures. The interface specifications for nodes (units of computation) can be regarded as data structures to which compositional operators can be applied. The compositional operators carry the same information as the annotated arcs and hyperarcs. A program becomes an algebraic expression over the units of computation which evaluates to a parallel program. The precise analogy is that a compositional expression evaluates to a generalized dependency graph in the CODE representation. Several aspects of this representation of parallel programs are of interest: (i) It is complementary to the graphical display representation since it is a compact means of specification of complex structures which are not under direct study or analysis and (ii) it strongly enforces a component oriented approach to programming since the composition operators can act only on the interfaces of the units of composition. The full paper will define the model of parallel computation and the calculus of composition which results and illustrates the use of this alternative mode of representation for parallelism. It will also compare this approach to compositional programming to others which have been proposed, for example [CHA91], and others. {\bf References} [BRO89] Browne, J.C., Sobek, S and Azam, M., CODE, A Graphical Parallel Programming Environment, {\it IEEE Software}, July, 1989. [CHA91] Chandy, K.M. and Taylor, S.J., {\it An Introduction to Parallel Programming}, Jones and Bartlett, Boston, 1991. \end{document}