Automatic Identification of Parallelizable Loops Using Transformer-Based Source Code Representations

This paper presents a method for the automatic identification of parallelizable loops using transformer-based source code representations. This research has significant real-world potential in optimizing software performance and development, particularly in areas requiring high-performance computing and efficient resource utilization, by automating a complex task in code analysis.