Automatic Constraint Generation for guided random simulation

Abstract

In this paper, we proposed an Automatic Target Constraint Generation (ATCG) technique to automatically generate compact and high-quality constraints for the guided random simulation environment. Our objective is to tackle the biggest bottleneck of the entire constrained random simulation process — the time-consuming and error-prone manual testbench composition process. By taking only the design under verification and simulation coverage as our inputs, our automatic constraint generation technique can successfully generate just a few key constraints while achieving very high simulation coverage. Our experimental results show that the proposed approach can outperform both directed and random simulations in both coverage and simulation runtime for a variety of designs