Dynamic Partial Order Reduction Based on Interference Solution

Abstract

Abstract In verifying the process of concurrent programs, dynamic partial order reduction (DPOR) exhaustively explores all the interactions of the equivalent execution traces in the concurrent system through dynamic judgment. Existing DPOR algorithms strive for optimal exploration (i.e., exploring only one interleaving per equivalence class), leading to exponential redundancy in the interleaving, state-space explosion, and memory overflow. In this study, the Interference Solution (IntSo) algorithm is proposed to achieve linear memory consumption by exploring only one interleaving per equivalence class through the exchange of causally-happened-before (CHB) conflicting relationships. In the sequential consistency memory model, the IntSo algorithm realizes parallelism while curtailing time and memory consumption in data-independent scenarios, as well as maximum path lengths. Empirical validation using GenMC demonstrates that the IntSo algorithm compares favorably with state-of-the-art techniques in terms of time and memory, fulfills linear time complexity, and reduces memory consumption by more than 30%.