Symbolic implication in test generation

Abstract

All test generation algorithms make use of symbolic algebra. The symbolic value that most test generators use is 'X', to denote the unknown/do not care logic value. The other end of the spectrum is to shade each X differently to fully exploit the information contained in them. This is impractical due to combinatorial explosion that results from such coloring. In this paper, the authors explore use of limited symbolic evaluation in test generation. This symbolic evaluation greatly improves test generation compared with the usual five-valued simulation. Also, and in contrast with other established techniques in test pattern generation such as static learning and dynamic learning, it requires no preprocessing and almost no additional memory. >