Assertion Ranking using RTL Source Code Analysis

We present a systematic and efficient ranking method to quantify the goodness of an assertion. We model dependencies among design variables as a directed graph called a variable dependency graph . We define assertion importance and assertion complexity metrics and use the dependency graph to algorithmically compute those two metrics. We repurpose an assertion coverage algorithm from the literature to form a statement-coverage-based ranking as our baseline. We compare our assertion ranking both qualitatively and quantitatively to this baseline. We demonstrate that our ranking is computationally more efficient than statement-coverage-based ranking and takes up to $4366\times $ less computation time. We identify the potential design intents that each ranking prioritizes. We also discuss at length the effect of those prioritizations on the rank agreement and the bug detection ability of the top-ranked assertions according to the two rankings. Finally, we provide a comprehensive ranking for a set of assertions by combining our ranking and the statement-coverage-based ranking.