Control-flow and data-flow analysis for the translation of control graphs into data graphs
The author developed and analyzed two notions: control-flow analysis and conditional-reaching definition. Then, these two notions are used to detect the implicit parallelism available in an ordinary program (that is a serial program, written in a higher-level programming language, for a serial computer). The two theories, control flow analysis and the conditional-reaching definition, are developed for structured control graphs, CG, which are language-independent representation of programs written in a higher-level programming language. The definition of CG is similar to what is known as a flow graph. The theory of control flow answers the following fundamental question: as early as possible, when the control reaches a given point in the program and whose execution is conditional, what is the set of instruction which will be executed eventually. The solution of the control-flow equation of some node, n, is the set of nodes which will execute eventually after the execution of n. The theory of conditional-reaching definition answers the following question: as early as possible, and when the control reaches a given point, c, in the program, which single definition, d, of the variable x will reach a given instruction in the program. Finally, the author shows how to use the two theories to map a structured control graph into an equivalent data graph, DG, with its permit execution rule.
- Research Organization:
- Wayne State Univ., Detroit, MI (USA)
- OSTI ID:
- 5254747
- Resource Relation:
- Other Information: Thesis (Ph. D.)
- Country of Publication:
- United States
- Language:
- English
Similar Records
Translating data flow graphs to architectures
Piecewise data flow architecture: architectural concepts