Debugging Tools for Correctness Analysis of MPI and OpenMP Applications
The correctness analysis of parallel programms is part of the research focus of the Chair I12. From a practical point of view, debugging parallel code often is a hard task, as bugs might depend on the scheduling and scaling of the execution. We have seen applications to only fail at large scale, while succeeding at lower scale.
With the intent to broaden the awareness of correctness analysis of parallel programs, Joachim Protze recorded a 30-minute webinar and presents the two tools MUST and ARCHER. They support runtime correctness analysis of MPI and OpenMP applications and are in part developed at the Chair. These tools, freely available under open-source licenses, allow the analysis even during large-scale execution. Combining both tools for hybrid MPI + OpenMP applications enables unprecedented runtime analysis of data races in multi-threaded MPI communication.