Comparison of Search-Based Algorithms for Stress-Testing Integrated Circuits

• Published in: Search-Based Software Engineering Vol. 11036; pp. 198 - 212
• Main Authors: Eljuse, Basil, Walkinshaw, Neil
• Format: Book Chapter
• Language: English
• Published: Switzerland Springer International Publishing AG 2018; Springer International Publishing
• Series: Lecture Notes in Computer Science
• Subjects: Automated search-based testing; Cache coherent interconnect; Random-restart hill-climbing; Simulated annealing; Stress testing
• ISBN: 9783319992402; 3319992406
• ISSN: 0302-9743; 1611-3349
• DOI: 10.1007/978-3-319-99241-9_10

This paper is concerned with the task of ‘stress testing’an integrated circuit in its operational environment with the goal of identifying any circumstances under which the circuit might suffer from performance issues. Previous attempts to use simple hill-climbing algorithms to automate the generation of tests have faltered because the behaviour of the circuits can be subject to non-determinism, with a search space that can give rise to local maxima. In this paper we seek to work around these problems by experimenting with different search algorithms which ought to be better at handling such search-space properties (random-restart hill-climbing and simulated annealing). We evaluate these enhancements by applying the approach to test the Arm Cache Coherent Interconnect Unit (CCI) on a new 64-bit development platform, and show that both simulated annealing and random-restart hill-climbing outperforms simple hill-climbing algorithm.