A Methodology for Automated Mining of Compact and Accurate Assertion Sets

Assertion - based verification has become a viable solution for functional verification. Manual definition of assertions is costly and needs human expertise. On the other hand, the automatic approaches are still struggling with generating high -quality assertion sets in terms of accuracy, and readability. To overcome the mentioned shortcomings, this paper exceeds the state-of-the-art by introducing a new fully automated approach to generate a set of complete and accurate assertions. Moreover, the mined assertions are readable thereby, easy to understand by the verification engineer. Furthermore, the method is equipped with a feature that communicates with users to select the assertions related to a specific block based on the user's need for further debugging and design analysis. Experimental results present the effectiveness of the proposed approach by showing that it generates significantly more compact assertion sets than the state-of-the-art while achieving 100% detection of the injected mutants.