Partners: Hartmut Kaiser, Rod Tohid, Andrew Lumsdaine 

SSEC Engineers: Don Setiawan, Ayush Nag 

Research Goals and Domain 

HPyX bridges the gap between Python and high-performance computing by providing Python bindings to the HPX C++ Parallelism Library. HPX implements modern C++ concurrency and parallelism features, enabling scalable execution across multi-core and distributed systems. HPyX leverages Python 3.13’s free-threading capabilities and Nanobind to expose these features to Python developers, empowering scientific workflows that demand speed and scalability. 

Software Problem 

Python dominates scientific computing but struggles with true multi-threading due to the Global Interpreter Lock (GIL). This limitation reduces performance for large-scale simulations, AI model training, and data-intensive tasks. Existing solutions often require rewriting code in C++ or using complex frameworks, creating barriers for researchers and developers who prefer Python. 

Software Solution 

HPyX removes these barriers by integrating HPX’s asynchronous and parallel execution primitives directly into Python. The project introduces free-threaded Python support for true concurrency, Nanobind-based bindings for efficient C++ interoperability, and high-level APIs for parallel algorithms such as for_loop and async. It also includes robust packaging and documentation, with conda and PyPI distribution, automated builds, and tutorials. It further stabilizes cross-platform behavior with early Windows support, reorganizes configuration files for clarity, and adds a fully automated build‑and‑test workflow using pixi. Coupled with a new MkDocs-based documentation system, expanded examples, improved inline docstrings, and updated packaging metadata, HPyX now provides a cleaner installation path, a more reliable developer experience, and a stronger foundation for distributed and multi-threaded Python execution. This design ensures reproducibility, scalability, and ease of adoption for Python developers in HPC and AI domains. 

Impact 

HPyX democratizes access to advanced parallel computing by combining Python’s usability with HPX’s performance. Researchers can now scale workloads across cores and clusters without leaving Python, accelerating innovation in fields like climate modeling, genomics, and large-scale AI. By fostering open-source collaboration and integrating best practices, HPyX strengthens the scientific software ecosystem and supports reproducible, high-impact research.