MMFuzz: Towards Enhancing RTL Fuzz Testing using Metric Feedbacks based on Markov Chain

Coverage guided dynamic verification is a widely used verification technique for RTL designs described using domain-specific languages for hardware and representing in some intermediate representations. Although the embedding of fuzz testing promote the abilities of coverage guided dynamic verification, there are lack of efficiently metric feedbacks utilization. In this paper, we proposed MMFuzz, a novel fuzzing tool enhanced by metric feedbacks. The proposed method utilze metric feedbacks efficiently in two aspects: seeds selection and mutators selection. The experimental results on several practical designs show that our method is able to achieve up to 1.0x improvements over the state-of-the-art RTL fuzzing tool with the same times of mutation.