Genetic algorithms: the philosopher's stone or an effective solution for high-level TPG?

Abstract

The paper examines the potentialities of genetic algorithms (GAs) with respect to the development of high-level TPGs. It summarizes at first the most relevant test pattern generation techniques based on genetic algorithms (GAs). This analysis distinguishes the considered techniques with respect to the abstraction level of the design under test. In particular, the effectiveness of gate-level GA-based TPGs is compared with the effectiveness of high-level GA-based TPGs. Differences are deeply investigated. They mainly concern the way genetic operators exploit specific simulation information to heuristically guide the genetic evolution. Moreover, a functional testing framework is described and used to actually measure on high-level descriptions the effectiveness of sophisticated GA-based TPGs in comparison to random approaches. Results are reported on a variety of benchmarks.