CONTRIBUTING.md

Contributing to TPU-MLIR

Thanks for taking the time to contribute! 🎉

Reporting issues

Before opening an issue, please:

1. Search existing issues to avoid duplicates.
2. If you find a related but unresolved issue, comment there instead of opening a new one.
3. When opening a new issue, use the appropriate template and include:
   • TPU-MLIR version (pip show tpu_mlir or commit SHA)
   • Target chip (e.g. bm1684x)
   • Source framework + model link (if reproducible)
   • Exact command line and full error output
   • Minimal reproducer when possible

Submitting pull requests

1. Discuss first — for non-trivial changes, open an issue describing the design before sending a PR.
2. One topic per PR — keep changes focused and reviewable.
3. Match the existing style:
   • C/C++ — formatted with clang-format (config: .clang-format).
   • Python — formatted with yapf (config: .style.yapf, 100-column limit).
   • The repo also ships an .editorconfig that most editors pick up automatically.
4. Run the regression tests under regression/ for any change touching the compiler or runtime.
5. Write good commit messages — short imperative summary, optional body explaining why.
6. Sign your commits with an email registered on your GitHub account so the commit is properly attributed (see GitHub → Settings → Emails).

Building from source

See README.md → Installation and README_cn.md → 安装.

Quick reference inside the official Docker image:

pip install -r requirements.txt
source ./envsetup.sh
./build.sh           # RELEASE build
./build.sh DEBUG     # debug build with symbols

Code review

• A maintainer will review your PR and may request changes.
• CI must pass before merge.
• Once approved, a maintainer will squash-merge or rebase-merge the change.

License

By contributing, you agree that your contributions will be licensed under the same terms as the project (see LICENSE).