A View of Large Language Models in HPC: Challenges and Opportunities
Abstract: There is a growing interest in using machine learning techniques to automate and improve the process of generating code. With the rapid development of large language models (LLMs), various models have been created to help write and optimize code. However, they do not yet meet the stringent requirements of high-performance computing (HPC), where highly optimized and efficient code is essential. This paper explores the research direction of adapting and using LLMs for HPC code generation. We present the reasoning behind our position and suggest how existing ideas can be adapted and enhanced to meet the demands of HPC applications.
Paper Type: Short
Research Area: NLP Applications
Research Area Keywords: code generation and understanding
Contribution Types: Position papers
Languages Studied: English
Submission Number: 5430
Loading