Abstract: In current formal verification practice, information sharing is common across engines attempting to prove multiple formal properties. Recent literature attempts to address multi-property verification via property clustering or ordering based on cone-of-influence (COI), which often becomes ineffective with rising design complexity. In this paper, we introduce PrOFraC, which initially unfolds the design to a certain depth, restructures the goals and employs a modified version of the Property Directed Reachability (PDR) engine that starts proving the goals from the latest frame instead of frame zero. PrOFraC restructures the goals by removing the falsified ones, adding the proven ones as constraints, and reordering the unsolved goals based on statistical data collected during the initial unfolding phase. We demonstrate our results on the hardware model checking competition (HWMCC) benchmarks. We also compare our method with traditional IC3/PDR versions that deploy sequential or concurrent processing of properties. Experimental results demonstrate that PrOFraC solves 2461 more properties across 30 different designs than the traditional sequential approach.
External IDs:dblp:conf/vlsid/DasHDJK25
Loading