Random stimulus generation with self-tuning

Constrained random simulation methodology still plays an important role in hardware verification due to the limited scalability of formal verification, especially for the large and complex design in industry. There are two aspects to measure the stimulus generator which are the quality of the stimulus generated and the efficiency of the generator. In this paper, we propose a self-tuning method to guide the generation for constrained random simulation by SAT solvers. We use a greedy search strategy in solving process to get the high-uniform distribution of the stimulus, and improve the efficiency of the generator by affinity grouping. Experimental results show that our methods can generate more uniform random stimulus with good performance.