MileStone: A Multi-Objective Compiler Phase Ordering Framework
MileStone is a new framework designed to optimize compiler phase ordering by treating it as a multi-objective optimization problem. It utilizes graph representations of programs and employs a reinforcement-learning agent to explore pass sequences while adhering to user constraints. Experimental results demonstrate that MileStone significantly improves execution time and energy efficiency compared to existing methods.
- ▪MileStone models compiler phase ordering as a multi-objective optimization problem.
- ▪The framework predicts performance metrics using a graph neural network.
- ▪Experiments show that MileStone can reduce execution time by up to 45 percent under the same energy budget.
Opening excerpt (first ~120 words) tap to expand
Computer Science > Programming Languages arXiv:2605.23435 (cs) [Submitted on 22 May 2026] Title:MileStone: A Multi-Objective Compiler Phase Ordering Framework for Graph-based IR-Level Optimization Authors:Amirhosein Sadr, Mehran Alidoost Nia View a PDF of the paper titled MileStone: A Multi-Objective Compiler Phase Ordering Framework for Graph-based IR-Level Optimization, by Amirhosein Sadr and Mehran Alidoost Nia View PDF HTML (experimental) 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.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at arXiv.org.