Graphs of Data Flow Dependencies

Abstract

Data flow dependencies of a program are important for code-optimisation of compilers and for detecting errors in programs. The control flow of a program can be described by its control flow graph. The mutual dependencies of variables in a program can be represented by graphs, too. For each variable of a program a tree structure can be found that shows its dependency on other variables. Different kinds of tree structures depict the data flow dependencies more or less detailed, according to the more or less exhaustive use of the source code’s information. The most accurate representation of data flow by tree structures can be achieved for individual paths of a program only. A less accurate representation results from the set of unordered program statements. In this case it is not possible to distinguish whether a statement precedes another or not. By use of the static control flow graph, trees can be established without regard to branching conditions. With these trees distinctions can be made between different actual values of the same variable. The methods of building tree structures that show the data flow of programs are suitable for data flow analysis by hand as well as for automatic analysis.