The AI-Enabled Continuous Program Optimization Infrastructure for LLVM

Welcome to the ACPO Extension to LLVM!

LLVM is the most popular open-source compiler framework, with massive community involvement. With the addition of ACPO framework, we are providing tools to include ML models for decision making, example models for some passes and a wide range of feature collectors to make performance-oriented compiler development easier to perform.

The changes provided comprise LLVM-side changes to enable the use of ML models in the LLVM compiler, which are meant to be used in tandem with CPLLab-Huawei/ACPO repository that comprises models, ML-framework interfaces and data analysis scripts to make compiler developers more productive.

Getting the Source Code and Building LLVM

Consult the Getting Started with LLVM page for information on building and running LLVM.

For information on how to contribute to the LLVM project, please take a look at the Contributing to LLVM guide.

Adding ACPO into your flow

With the changes provided the LLVM-side of the compiler is already included. To make it work, you will need to include the CPLLab-Huawei/ACPO added to your build to ensure interface with ML frameworks and models can be developed, stored and deployed.

Here are the steps to build and test LLVM with ACPO:

1. Build LLVM with ENABLE_ACPO macro turned on.
2. Include a repo https://github.com/Huawei-CPLLab/ACPO to obtain models and model training infrastructure.
3. Test the flow on existing ACPO-enabled passes, such as inliner to ensure your set up is correct.
4. Begin adding ACPO to passes of your choice.