SC24 Intel Panel Discussion – Open Source and Open Standards in HPC

Intel kicked off SC24 with a panel of industry experts who discussed how the open ecosystem works together to accelerate science and discovery. The panelists delved into the latest trends and opportunities for the HPC community. They discussed how open standards and software advancements are empowering organizations to stay competitive and agile in our rapidly changing technological environment.

Hosted by Rod Burns, VP of Ecosystem at Codeplay, an Intel company, and Unified Acceleration (UXL) Foundation Steering Committee Chair, the panel featured leaders across the open source and open standards community, including:

• Dr. Hartwig Anzt, Chair of Computational Mathematics, TU Munich and Professor, University of Tennessee
• Dr. Tom Deakin, Senior Lecturer in Advanced Computer Systems, University of Bristol, and Chair of the Khronos SYCL* Working Group
• Tony James, Chief Architect, Science and Space, Red Hat
• Penporn Koanantakool, UXL Foundation Steering Committee Member and Senior Software Engineer, Google Cloud
• Dr. Priyanka Sharma, UXL Foundation Steering Committee Member and Director of Software Engineering at Fujitsu Research of India and Head of the MONAKA SW R&D Unit

Open Source and Open Standards Working Together

The panelists discussed open communities and shared views about how open standards and open source together provide common code bases that different vendors can leverage to develop their applications across different hardware while maintaining adaptability to future technologies.

As Bristol’s Tom Deakin noted, “Open standards…allow application developers to have the kind of guarantees that the applications that they write will stay supported for the long term.”

Taking the Ginkgo high-performance math library as an example, TU Munich’s Anzt explained that “you can access it as a oneAPI product, and as we enable platform portability, then any application that relies on Ginkgo can automatically run on different hardware.”

Tony James commented on how open standards fit into Red Hat’s open-source projects. “Open standards are key for interoperability to ensure open governance so that you’re not locked into any one solution.” He called out OPEA, a Linux Foundation project that stands for Open Platform for Enterprise AI: “One of the things that they’re working on as a community is an effort to build building blocks where you can get started easily with your generative AI use cases. That kind of stuff couldn’t happen without open standards and open source communities.”

Encouraging Adoption within Large Organizations

While open source and open standards have an established history, large organizations may still hesitate to adopt open technologies. However, panelists argued that collaboration with open communities helps drive technology forward while providing opportunities for enterprises to add unique value.

In developing Fujitsu’s MONAKA, a 2nm Arm CPU, Priyanka Sharma stated, “The goal for MONAKA is very clear to us: it’s going to support all open-source software stack…so that’s why one of our prime focuses was to collaborate with the open source community, and we have done that very actively with collaboration from the Unified Acceleration Foundation (and) the Intel oneAPI stack.”

James noted, “One of the things that’s beneficial for everyone here is that all of us up here come from different organizations, and yet we all collaborate together in the open source realm whether it’s through projects under the Linux Foundation, partnerships with organizations like Intel, in some cases customer partnership. But I think all of you benefit from the collaboration that happens between all of our organizations and the open source community.”

Deakin agreed, “The nice thing about open standards like SYCL and the Khronos Group is it allows that common baseline that everybody can implement but then allows vendors to extend it in their own way, bring their own vendor extension, so they can expand and provide some value add.”

Answering security concerns, Google’s Koanantakool explained, “With open source software, you can see the source code exactly what it does, but for the closed source, you actually don’t know what it does, and you have to do a lot more testing to make sure that it’s actually just doing what it claims to do and nothing else. So, in a way, I think open source software is safer to use.”

Supporting AI and HPC Convergence

One common observation was how open standards in AI and HPC are starting to influence each other. Fujitsu’s Sharma shared that many engineers focused on developing AI models may not be aware that there is “something that you can do in PyTorch to enable support to train AI models.”

James added: “One of the things that I think a lot of us can agree on is that there is a convergence between traditional HPC and AI use cases. One of the challenges there is because a lot of open source software is involved in this, and it moves at such a rapid pace that it’s forcing some of these traditional high performance computing centers and academic institutions to reconsider how they’ve been deploying infrastructure up to this point. …As a community and keeping with open standards and open source, we should all try to collaborate to push towards one direction for the betterment of all of us.”

Visit the Intel Business YouTube channel to view the complete panel – and stay tuned for more videos of Intel booth theater talks from SC24.

* Other names and brands may be claimed as the property of others. SYCL is a trademark of the Khronos Group Inc. Quotations have been edited for brevity and clarity.