LLOR: Automated Repair of OpenMP Programs

This conference talk presents LLOR, a novel technique for automatically repairing data race errors in parallel programs written in C/C++ and Fortran using the OpenMP API. Learn about this innovative tool developed by researchers from Indian Institute of Technology Hyderabad, University of Cambridge, Supra Research, and independent contributors that not only fixes race conditions but also removes unnecessary barriers to improve performance. Discover how LLOR takes a language-independent approach to provide appropriate placements of synchronization constructs, making it the first tool of its kind capable of repairing OpenMP programs. The presentation showcases extensive experimental results from testing on 415 parallel programs, demonstrating LLOR's capabilities and effectiveness. This 28-minute talk was delivered at the VMCAI conference (January 20-21, 2025), sponsored by ACM SIGPLAN.