![[2602.18511] Beyond Pass-by-Pass Optimization: Intent-Driven IR Optimization with Large Language Models](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.18511] Beyond Pass-by-Pass Optimization: Intent-Driven IR Optimization with Large Language Models
Summary
The paper presents IntOpt, an intent-driven IR optimizer that enhances program optimization by separating high-level intent from low-level transformations, achieving significant performance improvements over traditional methods.
Why It Matters
This research addresses the limitations of current compiler optimization techniques by introducing a new paradigm that enhances the efficiency and correctness of program transformations. As software complexity grows, effective optimization strategies become crucial for performance, making this work highly relevant for developers and researchers in the field of programming languages and AI.
Key Takeaways
- IntOpt separates high-level optimization intent from low-level transformations.
- Achieves 90.5% verified correctness and 2.660x average speedup on a 200-program test set.
- Outperforms state-of-the-art LLM-based optimizers and modern compilers.
- Introduces a three-stage process: intent formulation, refinement, and realization.
- Addresses the pass coordination problem inherent in traditional optimization methods.
Computer Science > Programming Languages arXiv:2602.18511 (cs) [Submitted on 19 Feb 2026] Title:Beyond Pass-by-Pass Optimization: Intent-Driven IR Optimization with Large Language Models Authors:Lei Qiu, Zi Yang, Fang Lyu, Ming Zhong, Huimin Cui, Xiaobing Feng View a PDF of the paper titled Beyond Pass-by-Pass Optimization: Intent-Driven IR Optimization with Large Language Models, by Lei Qiu and 5 other authors View PDF HTML (experimental) Abstract:Modern compilers optimize programs through a sequence of modular passes over intermediate representations (IR). While this pass-by-pass paradigm offers engineering benefits, it suffers from a pass coordination problem: locally beneficial transformations may block more profitable optimizations in later stages. This limitation stems from the lack of an explicit notion of optimization intent, defined as a holistic strategy for coordinating multiple transformations toward a global performance objective. Recent LLM-based approaches formulate IR optimization as an end-to-end generation task, thereby avoiding the traditional pass-by-pass structure. However, optimization intent remains implicit in these methods, forcing models to jointly infer optimization strategy and generate low-level transformations, which limits both correctness and performance. We propose IntOpt, the first intent-driven IR optimizer that explicitly separates high-level optimization intent from low-level analysis and transformation. IntOpt organizes IR optimization...
