AINeutralarXiv – CS AI · Feb 274/106
🧠
From Prompts to Performance: Evaluating LLMs for Task-based Parallel Code Generation
Researchers evaluated Large Language Models' ability to generate parallel code across three programming frameworks (OpenMP, C++, HPX) using different input prompts. The study found LLMs show varying performance depending on problem complexity and framework, revealing both capabilities and limitations in high-performance computing applications.