SC19 Proceedings

The International Conference for High Performance Computing, Networking, Storage, and Analysis

Can Arm Take the Lead in HPC?

Authors: John Linford (ARM Ltd), Andrew Younge (Sandia National Laboratories), Oscar Hernandez (Oak Ridge National Laboratory), Mitsuhisa Sato (RIKEN Center for Computational Science (R-CCS)), Simon McIntosh-Smith (University of Bristol, GW4), Keith Obenschain (US Naval Research Laboratory)

Abstract: Arm-based processors have gained substantial traction in the HPC community. Sandia’s “Astra” is #156 on the Top500, and projects like the Japanese Post-K “Fugaku”, European Mont-Blanc, U.S. DOE Vanguard, and UK GW4/EPSRC are strong proof points. HPC system integrators like Atos, Cray, Gigabyte, and HPE have Arm commercial offerings, but will Arm become an HPC leader? This BoF brings together experts and luminaries to share their experiences with Arm, discuss the remaining technical and ecosystem challenges, consider the role of codesign in HPC, discuss progress, and lay out a vision for the future state of Arm in the HPC community.

Long Description: Arm availability and usability has grown substantially in the HPC community in just the past few years. This is demonstrated by several Arm-based projects around the world, including the Japanese Post-K, European Mont-Blanc, U.S. DOE/NNSA Astra, the UK’s GW4/EPSRC efforts. Furthermore, the commercial availability of Arm nodes from HPC vendors such as Cray, Gigabyte, HPE, Fujitsu, Atos, and Ingrasys also demonstrate the viability of Arm architectures. The Arm HPC community is growing substantially as the hardware is being tailored for HPC workloads via additions, such as Armv8-A Scalable Vector Extension (SVE) technology, and its software stack is growing to meet HPC applications needs.

This will be the third SC Arm BoF that will bring together experts from the HPC community including national laboratories, vendors, academia, and international research centers to give short lightning-talk presentations discussing how the current state-of-the-art of the Arm HPC ecosystem may be advanced to meet the requirements of the future. These short talks will be offered by ecosystem experts, researchers, and users in various fields and cover a wide array of topics surrounding the Arm ecosystem including hardware architectures, compilers, operating systems, runtime environments, fabric enablement, performance tools, scientific libraries, system software, benchmarks and HPC applications. With recent Arm HPC deployments, presenters will share performance results which will be of direct interest to expected attendees. This will provide an opportunity for the HPC community to learn about the latest research and development, share the state-of-the-art, understand the rapid progress made within the Arm HPC ecosystem, and discuss the future of HPC and Arm.

Following initial presentations, we will hold a panel session to enable active discussion between the BoF participants and audience. This panel will give attendees the opportunity to engage and interact with early adopters and contributors, learn about current development efforts and engage the discussion of future trends of Arm-based technologies as the industry advances. This panel will open the floor to discuss the latest advances in the Arm ecosystem with a multi-pronged approach for panel discussion, seeded by a moderator.

Initial topics will likely include relevant challenges in system architecture, networking, memory system designs, reliability and resilience, and accelerators, all of which are key drivers for the future of the HPC community. In particular, we would like to discuss SVE vector technology and its potential impact on users and compiler toolchains. The moderator will engage the panel and audience to discuss advances in OS and compilers, runtime components, system libraries, scientific tools, programming models, and application porting experiences. Finally, the panel will consider the open question of whether Arm-based supercomputing architectures are poised to take a leadership role in the HPC community.

The result of this BOF will be for attendees to help create a holistic picture of the current advancements as well as future trends in both hardware and software of Arm-based supercomputing. After the BOF, we will produce a report of the topics discussed which relate to the Arm HPC ecosystem and send updates to the Arm user mailing list.


Back to Birds of a Feather Archive Listing