Observability enhanced coverage analysis of C programs for functional validation

Abstract

Software programs written in some programming languages like C, C++, Java, etc, are mostly verified by functional simulation. Since exhaustive functional simulation is impossible for even a small sized C program, it is important to quantitatively measure the extent of design verification during simulation by a set of test vectors. Various coverage metrics have been proposed for measuring the degree of design verification. Most of them compute the extent of design excitation (controllability) but are unable to say whether the excitation responses have propagated to observable points in the program (observability). In this paper we propose a metric for code coverage analysis of C programs that addresses not only controllability but tackles observability as well. Thus, this metric is able to tell what percentage of the simulation responses have been propagated to observable points in the program like primary outputs or printed variables. We improve upon a recently proposed observability enhanced software coverage metric by increasing the accuracy of the analysis as well as decreasing the simulation runtime overhead. We report some preliminary results on example C programs.