Multi-Armed Bandits for Autonomous Test Application in RISC-V Processor Verification

Abstract

Multi-armed bandit problems have recently received a great deal of attention, because they adequately formalize so called exploration-exploitation trade-offs arising in several relevant applications of recommendation systems. In this article, the multi-armed bandit concept is applied to an equivalent problem within the context of hardware verification. Specifically, a verification recommendation system is designed, based on the multi-armed bandit decision-making model, to automatically suggest which test sequences – constrained-random or direct – should be applied during processor verification. Automating the process of test application not only improves functional coverage, as shown by the experimental results for a two-way superscalar out-of-order RISC-V processor, but it also minimizes the verification effort and, in effect, shortens the time to market.