Microprocessor Hazard Analysis via Formal Verification of Parameterized Systems

Implementation of pipeline-based execution of instructions in purpose-specific microprocessors is an error prone task, which implies a need of proper verification of the resulting designs. Our long-term goal is to develop a set of verification techniques with formal roots, each of them specialised in checking absence of a certain kind of errors in purpose-specific microprocessors. The main idea is that, this way, a high degree of automation and scalability can be achieved since only parts of a design related to a specific error are to be investigated. In our previous works, we proposed, with the above goal in mind, fully automated approaches for checking correctness of the implementation of individual instructions and for verifying absence of read-after-write (RAW) hazards. In this paper, we extend our approach by aiming at write-after-write (WAW) and write-after-read (WAR) in microprocessors with a single pipeline.