Prompting for Performance: Exploring LLMs for Configuring Software
This addresses the problem of high computational costs in configuring software for performance, but it is an incremental step as an exploratory study with preliminary findings.
The study explored whether large language models (LLMs) can assist in performance-oriented software configuration by evaluating them on tasks like identifying relevant options and recommending configurations across systems such as compilers and video encoders, with preliminary results showing both alignment with expert knowledge and limitations like hallucinations.
Software systems usually provide numerous configuration options that can affect performance metrics such as execution time, memory usage, binary size, or bitrate. On the one hand, making informed decisions is challenging and requires domain expertise in options and their combinations. On the other hand, machine learning techniques can search vast configuration spaces, but with a high computational cost, since concrete executions of numerous configurations are required. In this exploratory study, we investigate whether large language models (LLMs) can assist in performance-oriented software configuration through prompts. We evaluate several LLMs on tasks including identifying relevant options, ranking configurations, and recommending performant configurations across various configurable systems, such as compilers, video encoders, and SAT solvers. Our preliminary results reveal both positive abilities and notable limitations: depending on the task and systems, LLMs can well align with expert knowledge, whereas hallucinations or superficial reasoning can emerge in other cases. These findings represent a first step toward systematic evaluations and the design of LLM-based solutions to assist with software configuration.