How Good is Your Property? A New Metric for Formal Property Coverage

Formal property verification has been widely used in function verification, where user-specified properties are checked to ensure function correctness. However, as human verification engineers can also make mistakes when writing the properties, it is a general question on how to ensure the quality of formal properties. There are several existing coverage metrics, e.g., cone-of-influence (COI) and Proof Core coverage that assess the quality of properties in the setting of formal property verification. However, they are not sufficient to identify all coverage holes as this paper will show. Therefore, we propose FRCoverage, a new functional coverage metric for formal property verification, which is based on function reduction. Our case study shows that this new coverage metric successfully identifies three coverage holes in the commonly used RISCV- Formal verification framework when it is applied on PicoRV32 processor verification.