Large Language Models (LLMs) have demonstrated remarkable capabilities across a wide spectrum of software engineering tasks, from code generation and summarization to program repair and vulnerability detection. This surge has given rise to the Large Language Models for Software Engineering (LLM4SE) area. However, limited attention has been paid to address two critical issues of LLM4SE techniques: (1) \emph{inefficiency}: they require prohibitive computational cost, time, and memory resources, making them inaccessible to many practitioners, and (2) \emph{unsustainability}: they consume vast amounts of energy, water, and carbon emissions, raising serious environmental concerns. As a result, the current trajectory of LLM4SE risks becoming both economically and ecologically unsustainable, calling for urgent attention to efficiency and greenness.

This paper aims to redirect the focus of the research community towards the efficiency and greenness of LLM4SE, while also sharing potential research directions to achieve this goal. It commences with a brief overview of the significance of LLM4SE and highlights the need for efficient and green LLM4SE solutions, followed by a classification of existing techniques along four complementary dimensions: data-centric, modelcentric, system-centric, and program-centric. Building on this analysis, the paper presents a vision for a future where efficient and green LLM4SE revolutionizes the LLM-based software engineering tool landscape, benefiting various stakeholders, including industry, individual practitioners, and society. The paper then delineates a roadmap for future research, outlining several promising research directions to realize the above vision by 2030, such as constructing new benchmarks, developing efficient training and fine-tuning methods, exploring alternative approaches like retrieval-augmented generation, and advancing model compression, inference acceleration, and program-level optimization. While not intended to be a definitive guide, the paper aims to inspire further progress, with the ultimate goal of establishing efficient and green LLM4SE as a central element in the future of software engineering.

This paper has been accepted for publication in ACM Transactions on Software Engineering and Methodology (TOSEM) on November 07, 2024, and is available at https://doi.org/10.1145/3708525. Rather than being exclusively a secondary study, this paper combines a literature review with a forward-looking vision and detailed research roadmap. Jieke Shi will present the paper at ICSE 2026, if invited.