This research evaluates how well Large Language Models understand and implement code optimization for high-performance computing applications.

• Introduces a benchmark suite of critical HPC computational motifs
• Assesses state-of-the-art LLMs on their ability to generate efficient and correct code
• Examines performance in complex computational contexts requiring specialized optimization knowledge
• Provides insights into where LLMs succeed and fail in engineering optimization tasks

For engineering teams, this research offers valuable understanding of when LLMs can effectively assist with performance optimization, and when human expertise remains essential for complex computational challenges.

Do Large Language Models Understand Performance Optimization?