Efficient Mutation-Analysis Coverage for Constrained Random Verification

TL;DR: This paper presents an efficient integration of mutation analysis into CRV flows, not only as a coverage gauge for simulation adequacy but also, a step further, to direct a dynamic adjustment of the test probability distribution.

Abstract: Constrained random simulation based verification (CRV) becomes an important means of verifying the functional correctness of the increasingly complex hardware designs. Effective coverage metric still lacks for assessing the adequacy of these processes. In contrast to other coverage metrics, the syntax-based Mutation Analysis (MA) defines a systematic correlation between the coverage results and the test’s ability to reveal design errors. However, it always suffers from extremely high computation cost. In this paper we present an efficient integration of mutation analysis into CRV flows, not only as a coverage gauge for simulation adequacy but also, a step further, to direct a dynamic adjustment of the test probability distribution. We consider the distinct cost model of this MA-based random simulation flow and try to optimize the coverage process. From the probabilistic analysis of the simulation cost, a heuristics for steering the test generation is derived. The automated flow is implemented by the SystemC Verification Library and by CertitudeTM for mutation analysis. Results from the experiment with an IEEE floating point arithmetic design show the efficiency of our approach.