We are excited to announce that Anaconda has joined the OpenMP Architecture Review Board (ARB). As more and more researchers and developers adopt Python for their work, parallel computing standards like OpenMP must evolve to include Python as a first-class language. OpenMP enables multi-threaded applications to efficiently use all CPU cores, making it essential for high performance computing (HPC). Python has become indispensable in this space, particularly as machine learning workloads become integral to HPC environments. Anaconda has been developing Python compilers for more than a decade, and ships OpenMP-enabled packages to millions of users every year. That’s why we’ve joined the OpenMP ARB: to drive the official extension of OpenMP to the Python language and represent the unique needs of the Python community in this critical standard.

What is OpenMP?

OpenMP makes parallel programming significantly easier for developers. Instead of manually managing how work gets divided across multiple CPU cores, developers simply mark which parts of their code should run in parallel. The OpenMP compiler and runtime handle all the complex coordination behind the scenes. You may not have ever used OpenMP directly, but many of your favorite Python data science and numerical computing libraries depend on OpenMP to make efficient use of all the CPU cores in your computer.

Because OpenMP plays such a critical role across the computing industry, the standard is managed by an Architecture Review Board. This non-profit consortium of technology companies and research institutions oversees the standard’s evolution, publishes periodic updates, and forms working groups to explore new capabilities.

Why should OpenMP support Python?

Python is heavily used in fields where OpenMP has been very successful: numerical computing, high performance computing, data analytics, and machine learning. Traditionally, using OpenMP has required developers to leave the Python interpreter and write code in a supported language, like C++ or FORTRAN. While this workaround functions, there’s strong demand for native Python support.

One of Numba’s highest rated features in user surveys is its parallel loop capability (contributed to Numba many years ago by researchers at Intel), which covers only a small fraction of what OpenMP can do. Expanding Numba’s parallel programming capabilities to cover more of the OpenMP standard would unlock significant new possibilities for Python developers.

This led several researchers to build a more complete implementation of OpenMP in Python using Numba as a compiler framework. The effort was open sourced as the PyOMP project, with collaboration from Anaconda’s compiler team. This work resulted in the paper Programming GPUs with OpenMP and Python, which won Best Paper Award at the International Workshop on OpenMP in 2025. The research demonstrates that it’s now possible to take a Python function and compile on-the-fly for GPU acceleration using an OpenMP runtime. This breakthrough makes GPU computing even more accessible to Python developers.

Making a standard is hard work

While PyOMP demonstrates what’s possible, it’s an unofficial adaptation. OpenMP doesn’t have an official standard for Python yet, but is now working on one. Python presents unique challenges for OpenMP, as it is designed primarily to be interpreted rather than compiled. Python compilers, like Numba and SPy, have a long history of navigating the complexities of compiling a language built for interpretation. Anaconda’s compiler team has spent the last 13 years working through these challenges, so we are excited to join the OpenMP Python Subcommittee. We’ll provide our technical perspective while representing the Python community’s unique needs to ensure the new standard fits into the broader Python ecosystem.

The future is bright

We’re excited about what a Python-native OpenMP specification will enable. The PyOMP project has already shown the potential for GPU acceleration directly from Python code. As the standard evolves, we expect to see even more innovative implementations that make parallel computing accessible to the entire Python community.

Anaconda is committed to helping define this new standard and ensuring it serves Python developers’ needs. Our compiler team brings over a decade of experience navigating the unique challenges of compiling Python, and we’re ready to put that expertise to work.

Want to stay in the loop? Check out the PyOMP installation guide to get a preview of how OpenMP could work for Python users.