Exact stimuli minimization for simulation-based verification

Due to the ever increasing complexity of modern circuits and systems, verification represents one of the most time-consuming tasks in the entire design process for embedded systems. For this purpose, simulation-based techniques are widely applied in industrial contexts. Here, stimuli are determined which are used as input for the Design under Verification (DUV) and are supposed to trigger different aspects of the new design. However, usually much more stimuli are generated than actually needed to comprehensively cover all aspects. This obviously increases the run time of the verification significantly. Consequently, verification engineers aim for minimizing the number of stimuli after their generation - without loosing their coverage. Existing solutions, however, usually generate results which are far from being optimal. Besides that, their scalability is severely limited. In this work, we propose a solution for an exact minimization of stimuli. To this end, we utilize the computational power of modern reasoning engines such as MAX-SAT solvers which can efficiently minimize a given set of stimuli. Experimental evaluations confirm that, compared to previous work, up to 63% further reduction can be obtained and scalability significantly increases.