Searching Optimal Compiler Optimization Passes Sequence for Reducing Runtime Memory Profile using Ensemble Reinforcement Learning

This is the source code repository of our work "Searching Optimal Compiler Optimization Passes Sequence for Reducing Runtime Memory Profile using Ensemble Reinforcement Learning", at EMSOFT 2023. This study aims to optimize memory usage by optimizing LLVM optimization flags using reinforcement learning. The repository in consisted of the followings:

1. Feature Extraction

Feature_Extraction/ includes randomly generated source codes and feature extraction LLVM pass. More information is in Feature_Extraction/README.md.

2. Preprocessing

Preprocess/ includes flag and IR selection that are used for the RL model. It also reduces feature dimension by applying PCA and KMeans clustering to the raw features. More information is in Preprocess/README.md.

3. Reinforcement Learning

RL/ includes RL training and testing models. PPO, A3C, PG algorithms are investigated. More information and experiment result is in RL/README.md.

License of the code and data will be changed to BSD-3-Clause license after EMSOFT 2023's notification.