Leveraging the Dynamic Program Structure Tree to Detect Data Races in OpenMP Programs

• Published in: 2022 IEEE/ACM Sixth International Workshop on Software Correctness for HPC Applications (Correctness) pp. 54 - 62
• Main Authors: Yu, Lechen, Jin, Feiyang, Protze, Joachim, Sarkar, Vivek
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.11.2022
• Subjects: Benchmark testing; Data models; data race; Detectors; OpenMP; Parallel processing; Prototypes; Runtime; Software
• ISSN: 2831-3925
• DOI: 10.1109/Correctness56720.2022.00012

OpenMP provides a rich set of constructs to support multiple paradigms of parallelization, e.g., single program multiple data (SPMD) and task parallelism. Integrating these disparate paradigms in a single execution model increases the complexity of OpenMP, making OpenMP programs prone to concurrency bugs such as data races. Inspired by the task-oriented execution model in the OpenMP spec, we extended SPD3, a data race detection algorithm designed for async-finish task parallelism to support OpenMP programs. We found that by extending SPD3's key data structure, the Dynamic Program Structure Tree (DPST), SPD3 can support the majority of OpenMP constructs. We have implemented a prototype, TSan-spd3, on top of Google's ThreadSanitizer (TSan). To conduct an apples-to-apples comparison with Archer, the state-of-the-art dynamic race detector for OpenMP programs, we compared TSan-spd3 with an Archer implementation that executes on the same version of TSan. In addition, we evaluated Archer in two modes, the default mode using the original TSan and the accelerated mode enabling the use of SIMD instructions in TSan. The evaluation was conducted on nine benchmarks from the BOTS and SPEC OMP2012 benchmark suites. The evaluation results show that in eight out of nine benchmarks TSan-spd3 achieved similar overhead with Archer, while TSan-spd3 can identify more potential races than Archer.