Python for HPC
Wednesday, June 24, 2026 5:15 PM to 6:15 PM · 1 hr. (Europe/Berlin)
Hall E - 2nd Floor
Birds of a Feather
Compiler and Tools for Parallel ProgrammingParallel Programming LanguagesParallel Numerical Algorithms
Information
Python is the world’s most widely used programming language, powering everything from machine learning and data analysis to web development and automation. However, using Python effectively on high-performance computing (HPC) systems presents unique and persistent challenges. These issues often prevent new users from realizing the full power of supercomputers, and they frustrate seasoned practitioners who rely on Python for scientific computing.
The Python HPC ecosystem is highly fragmented. Language implementers, library developers, and HPC users often operate in silos with little coordination. As a result, critical challenges faced by HPC Python users go unaddressed at the language and standards level, and efforts to improve performance and usability on HPC platforms remain scattered and under-resourced.
This Birds of a Feather (BoF) session aims to build stronger connections between the global scientific Python community, the Python language maintainers, and the HPC community. Our goal is to create an ongoing dialogue that ensures the needs of HPC users are represented as Python evolves.
Common runtime issues faced by HPC users include:
- Python’s module system, when run on shared filesystems, introduces significant latency during startup.
- Python’s default packaging tools struggle to support specialized libraries such as optimized MPI stacks or system-tuned numerical libraries.
- Parallelism tools in Python - ranging from threading to multiprocessing to MPI bindings - are often incompatible or difficult to integrate.
- Despite the size and impact of the HPC Python user base, it remains underrepresented in broader Python discussions, including at core language and packaging summits.
At the same time, the Python language itself is evolving rapidly. A number of active Python Enhancement Proposals (PEPs) will have far-reaching effects on the HPC ecosystem, but currently, there is minimal overlap between HPC stakeholders and the Python Steering Council or other decision-making bodies.
Relevant ongoing work includes:
- Standardizing array APIs and memory models to improve interoperability between libraries and accelerators.
- Evolving packaging standards to better accommodate system-level and cross-platform deployments.
- Developing just-in-time (JIT) compilation frameworks at the bytecode level, with implications for performance-critical workloads.
- Introducing new asynchronous and parallel programming primitives at the language level.
These changes present both opportunities and risks for the HPC community. Without input from HPC practitioners, new language features may fall short of supercomputing needs or introduce new bottlenecks. With the right engagement, however, these changes could significantly improve Python’s performance, portability, and productivity in HPC environments.
To that end, we are inviting leaders from the core Python development team to participate in this BoF and hear directly from HPC practitioners. Our goals are to:
- Equip attendees with an understanding of upcoming Python language standards and how they may impact scientific and HPC workloads.
- Connect practitioners with one another to share solutions and align on priorities.
- Create shared documentation and communication channels to support sustained collaboration.
- Lay the groundwork for an official Python HPC Working Group that can represent HPC interests to the Python Software Foundation and language governance bodies.
Join us to help shape the future of Python on supercomputers - and ensure your voice is heard.
Organizers:
The Python HPC ecosystem is highly fragmented. Language implementers, library developers, and HPC users often operate in silos with little coordination. As a result, critical challenges faced by HPC Python users go unaddressed at the language and standards level, and efforts to improve performance and usability on HPC platforms remain scattered and under-resourced.
This Birds of a Feather (BoF) session aims to build stronger connections between the global scientific Python community, the Python language maintainers, and the HPC community. Our goal is to create an ongoing dialogue that ensures the needs of HPC users are represented as Python evolves.
Common runtime issues faced by HPC users include:
- Python’s module system, when run on shared filesystems, introduces significant latency during startup.
- Python’s default packaging tools struggle to support specialized libraries such as optimized MPI stacks or system-tuned numerical libraries.
- Parallelism tools in Python - ranging from threading to multiprocessing to MPI bindings - are often incompatible or difficult to integrate.
- Despite the size and impact of the HPC Python user base, it remains underrepresented in broader Python discussions, including at core language and packaging summits.
At the same time, the Python language itself is evolving rapidly. A number of active Python Enhancement Proposals (PEPs) will have far-reaching effects on the HPC ecosystem, but currently, there is minimal overlap between HPC stakeholders and the Python Steering Council or other decision-making bodies.
Relevant ongoing work includes:
- Standardizing array APIs and memory models to improve interoperability between libraries and accelerators.
- Evolving packaging standards to better accommodate system-level and cross-platform deployments.
- Developing just-in-time (JIT) compilation frameworks at the bytecode level, with implications for performance-critical workloads.
- Introducing new asynchronous and parallel programming primitives at the language level.
These changes present both opportunities and risks for the HPC community. Without input from HPC practitioners, new language features may fall short of supercomputing needs or introduce new bottlenecks. With the right engagement, however, these changes could significantly improve Python’s performance, portability, and productivity in HPC environments.
To that end, we are inviting leaders from the core Python development team to participate in this BoF and hear directly from HPC practitioners. Our goals are to:
- Equip attendees with an understanding of upcoming Python language standards and how they may impact scientific and HPC workloads.
- Connect practitioners with one another to share solutions and align on priorities.
- Create shared documentation and communication channels to support sustained collaboration.
- Lay the groundwork for an official Python HPC Working Group that can represent HPC interests to the Python Software Foundation and language governance bodies.
Join us to help shape the future of Python on supercomputers - and ensure your voice is heard.
Organizers:
Format
on-site
Targeted Audience
- Current Python-on-HPC users: HPC practitioners already using Python on HPC systems.
- Potential Python-on-HPC users: HPC practitioners on HPC systems not yet using Python.
- Potential HPC users from Python: Python users who want to adopt HPC.
- Library/framework developers: Builders of HPC-used software (e.g., C++, Fortran, Rust).
BoF Format
Birds of a Feather Presentation
