Research Proposal: Efficient and Effective Symbolic Model Checking

In this proposal, we briefly outline some of the major challenges facing the adoption of Formal Verification techniques. We focus on the model checking approach, which is completely automated in principle and quite automated in practice. These challenges include handling the state explosion problem associated with large industrial designs, which manifests itself as large representation sizes and being able to reconcile verification with debugging. Large State Spaces: As is well known, the main challenge in model checking for design verification is what is termed as the “state explosion problem” – given a design, the state space that it encompasses is often exponential in the size of the design description. Traditionally, the state explosion problem has been handled close to the design level, using for example abstraction, symmetry reduction, compositional reasoning, etc. These approaches have been shown to produce significant gains in many cases. Their main drawback is that the user needs to discover the applicability of these techniques on almost a case by case basis; hence, they cannot be easily automated. It is therefore vital to handle large state spaces automatically in a manner that is transparent to the designer. The earliest approach to model checking [6] emphasized an enumerative implementation, while a symbolic technique using BDDs [3] was suggested by McMillan [15]. Symbolic model checking can exhaustively cover the state space when handling small designs but cannot handle industrial sized designs, which can often be orders of magnitude larger. The main issue is that current symbolic data structures quickly grow quite large, thereby often not fitting in main memory and being cumbersome to perform operations upon at such large sizes. One solution is functional partitioning as proposed by Jain, et.al.[13], and extended by Narayan, et.al.[17]. More recently satisfiability-based model checkinghas