Inside HPSF 2026: The Future of HPC Software
As high-performance computing, AI, and accelerated systems continue to reach new industries and research applications, the importance of scalable and sustainable software infrastructure is becoming increasingly clear. Software remains the language of science, and this was a central theme at the High Performance Software Foundation (HPSF) Conference, where the community came together to advance open source collaboration and strengthen the foundations of modern computing.
Kitware presented numerous talks this year, sharing insights from our experience in build systems, package management, and reproducible workflows. You can watch the recordings or continue reading to get the latest updates in high performance software.
The session recordings are live!
What You Should Know About High Performance Software in 2026
Software Infrastructure as a Foundation for HPC Systems
Build systems, package management, and testing frameworks are central to how HPC software is developed and deployed. Across the conference, this aligned closely with Kitware’s long-standing work on core infrastructure like CMake, Spack, and CDash, where these systems are treated as foundational components that determine how software is composed, integrated, and sustained across environments.
Dependency Management at Scale
As software stacks grow, dependency graphs become deeper and more difficult to manage. Version conflicts, transitive dependencies, and platform-specific variations introduce instability. Work surrounding Spack environments and shared stacks reflects a broader effort across the HPC software community, including work involving Kitware, to define and maintain consistent configurations.
Reproducibility Through Environment Definition
Reproducibility depends on controlling the full software environment, including compilers, dependencies, and configuration. This remains a central focus in both the conference discussions and Kitware’s work supporting reproducible workflows, where explicitly defined environments are required to ensure consistent builds across systems.
Build and Test Performance as Critical Factors
Build and test workflows can become bottlenecks in large-scale projects. CMake’s instrumentation framework provides visibility into compilation timing, build execution, system resource usage, and test behavior throughout the software workflow. These capabilities are increasingly important for improving iteration speed and are an area where Kitware continues to invest through ongoing development of build and testing infrastructure.
Standardizing and Scaling Software Stacks
Defining reusable and consistent software stacks helps reduce duplication and improve collaboration across teams. Shared configurations and common stacks, particularly within Spack, reflect a broader push toward standardization that aligns with Kitware’s work in supporting scalable, multi-platform software environments.
Sustaining HPC Infrastructure Over Time
Maintaining widely used HPC tools requires long-term investment in infrastructure, governance, and careful evolution. Kitware’s role in sustaining projects like CMake, VTK, and ParaView highlights the importance of evolving APIs and supporting production systems without fragmenting ecosystems, a challenge that was echoed throughout the conference.
CI Beyond Commodity Cloud Environments
Testing HPC software requires infrastructure that reflects real execution environments. Standard CI pipelines often fall short when dealing with distributed systems and specialized hardware. Discussions at HPSF emphasized extending CI beyond commodity cloud environments, an area where Kitware is actively involved through working group efforts and infrastructure development.
Final Takeaways
These discussions highlight a broader shift in how HPC software is being developed and sustained. As systems continue to scale and diversify, success depends on coordinated approaches to building, testing, packaging, and deploying software across complex environments.
At Kitware, our focus remains on developing scalable, reproducible, and maintainable solutions for real-world HPC and AI applications. If you are interested in assessing and modernizing your software systems, contact our team. Our experts can review your existing structure and processes and identify opportunities to improve performance.