Shared problems and community action. A working session bringing together users and developers across the whole open-source simulation community to turn shared problems into community-owned next steps.
This BoF builds on our SCA/HPCAsia26 session, where we mapped the shared problems facing open-source computer architecture simulators. This time the aim is to move from naming those problems to acting on them.
The discussion intends to serve the entire open-source architecture-simulation community, that includes both users and developers of many kinds of tools (cycle-level, functional, network/PDES, emulation, ML-driven, and more), rather than any single niche.
Potential discussion topics are framed to lead towards something the community can collectively do, and everything we agree is captured here so the outcomes reach people who couldn't be in the room.
Whilst the session will encourage audience participation to guide the BoF's content, a selection of themes have been curated to serve as possible discussion starting points.
How separate tools can effectively compose to empower a researcher's simulation environment, without losing speed, accuracy or usability along the way.
What keeps an open-source simulator healthy over a decade, and what usually kills it.
Trusting simulation results and comparing them fairly between tools, as hardware moves fast.
How a community governs shared efforts, aligns different priorities, and turns one-off contributions into a lasting ecosystem rather than fragmented, single-use tools.
What practices from the closed world, such as maintenance, validation, interoperability and support, are worth adopting in open source.
How AI is changing the hardware we need to simulate, and what would make AI/ML-driven simulation credible.
An interactive, discussion-led session: short panellist presentations, then most of the hour for the room to talk, prioritise, and identify actionable ways forward.
Organised by Jack Jones (University of Bristol) and Jens Domke (RIKEN Center for Computational Science).
A deliberately balanced group, across the US, Asia and Europe, and across tool developers and users.
HPC network performance analysis and simulation (CODES); multi-fidelity and surrogate-based parallel discrete-event simulation.
Parallel discrete-event simulation (ROSS/CODES); ML models as logical processes in PDES; an open-source industry perspective.
Processor Research Team lead; FPGA and reconfigurable computing and accelerators; works across vendor and open tooling.
MONAKA R&D lead; hybrid HPC-AI simulation; open-source advocate and chair of the UXL AI & Scientific Computing SIG.
gem5 user and extender for architecture research; design-space exploration and the tooling ecosystem behind it.
Developer of open-source performance analysis and prediction tools (OSACA, Kerncraft, WINIC); evaluates their trustworthiness across sites.
This hub is where the session's outputs live, so the discussion and takeaways reach the wider community, not just the people in the room.
This hub carries forward from venue to venue. Earlier sessions and their outcomes are below.
How this community might work together best is itself one of the outcomes the BoF. The mechanisms below are planned to be agreed at the session and generated here.
A shared way for people to add tools, resources and priorities. The exact mechanism is one of the things the session will decide, and we'll format and publish it here as a BoF outcome.
A channel to keep the discussion going beyond SC26. It will be established as part of the session's outcomes and posted here.