MileStone: A Multi-Objective Compiler Phase Ordering Framework for Graph-based IR-Level Optimization

2026-05-22

AI Summary

This work addresses the challenge of compiler optimization phase ordering, where trade-offs among execution time, code size, and energy consumption often lead to conflicting objectives that existing approaches struggle to balance. The paper proposes a novel framework that integrates graph neural networks, reinforcement learning, and a self-evolving database to formulate the problem as a constrained multi-objective optimization. Programs are represented as IR graphs, and a graph neural network predicts their performance under different optimization sequences. A reinforcement learning agent then explores Pareto-optimal sequences within user-specified constraints, while the self-evolving database continuously refines prediction accuracy through iterative learning. Experimental results on standard benchmarks demonstrate that the proposed method significantly outperforms LLVM’s default optimization, reducing execution time by up to 45% under the same energy budget and achieving more precise adherence to multidimensional constraints.

Abstract

Compiler phase ordering has a strong effect on program performance. Finding an effective sequence of passes is still a difficult task because the search space is large and execution time, code size and energy consumption often conflict. Existing methods usually depend on fixed optimization levels or limited heuristics and they rarely handle multiple objectives at the same time. This paper presents MileStone, a modular framework that models compiler phase ordering as a multi-objective optimization problem. MileStone represents programs as graphs, predicts performance metrics with a graph neural network and explores pass sequences with a reinforcement-learning agent that follows user constraints. The framework also builds a self-evolving database that collects compiler transformations and improves prediction quality. Experiments on standard benchmarks show that MileStone finds strong Pareto-optimal solutions, meets energy limits more accurately than LLVM optimization levels and other related techniques. MileStone reduces execution time by up to 45 percent under the same energy budget using a multi-objective approach. The results show that MileStone provides an effective and scalable solution for multi-objective compiler phase ordering.

Problem

• compiler phase ordering
• multi-objective optimization
• program performance
• energy consumption
• code size

Innovation

• multi-objective optimization
• graph neural network
• reinforcement learning
• compiler phase ordering
• graph-based IR