Motivation

Python is commonly used in the fields of scientific computing, data analysis and machine learning. However, the Python packages used in these disciplines aren’t wholly written in Python - they also contain a lot of code written in C/C++ or other languages which needs to be built as part of the module. Such projects typically create binary wheels for each of their releases and upload these wheels to pypi, the Python Package Index. This allows users to easily and quickly install tested, prebuilt versions of their favourite projects using the pip (or uv) tool.

Until recently, Python packaging infrastructure like auditwheel, cibuildwheel, and manylinux did not support riscv64, and no native runners for GitHub Actions were available. This made supporting the architecture difficult for open-source projects without complicated build processes and emulated systems. However, the aforementioned infrastructure now supports riscv64, and with the RISE RISC-V Runners project, maintainers have the option of building binary wheels on native riscv64 platforms. The RISE Python Wheels project’s goal is to accelerate this adoption and ensure that the riscv64 architecture is fully-supported for data science and machine learning applications.