Empirical Studies of Control Dependence Graph Size forC Programs

  • Authors:

  • Mary Jean Harrold;James A. Jones;Gregg Rothermel

  • Affiliations:

  • Computer and Information Science, Ohio State University, 395 Dreese Lab, Columbus, OH 43210;Computer and Information Science, Ohio State University, 395 Dreese Lab, Columbus, OH 43210;Computer Science, Oregon State University, 307A Dearborn Hall, Corvallis, OR 97331

  • Venue:
  • Empirical Software Engineering
  • Year:
  • 1998

Quantified Score

Hi-index 0.00

Visualization

Abstract

Many tools and techniques for performing software engineeringtasks require control-dependence information, represented inthe form of control-dependence graphs. Worst-case analysis ofthese graphs has shown that their size may be quadratic in thenumber of statements in the procedure that they represent. Despitethis result, two empirical studies suggest that in practice,the relationship between control-dependence graph size and programsize is linear. These studies, however, were performed on a relativelysmall number of Fortran procedures, all of which were derivedfrom numerical methods programs. To further investigate control-dependencesize, we implemented tools for constructing the two most populartypes of control-dependence graphs, and ran our tools on over3000 C functions extracted from a wide range of source programs.Our results support the earlier conclusions about control-dependencegraph size.