Open Edge HPC Initiative

Open Edge HPC Initiative

OEHI ISC’2022 Status Report

3,007 attendees and 137 exhibitors from 59 countries crowded ISC 2022. A very welcome comeback after three years of not having had the possibility to meet in person. Alas, we all had the opportunity to come together as a community, catch up on technology and on the development, build new relationships and maintain old ones.

Several members of the OEHI attended the events and presentations about the OEHI were shown at private meetings and on the monitors at the booth’s rolling presentations.

Prof. Dirk Pleiter from KTH Royal Institute of Technology in Stockholm presented the current status and projects of OEHI at the booth of E4.

Here is the presentation of Prof. Dirk Pleiter in a nutshell:

  • Mission of OEHI
    • Foster the development of an open and feature-rich ecosystem for Arm- and RISC-V-based technologies
    • Make the Arm and RISC-V based technology well suited for different IT segments assuming a future computing continuum ranging from HPC to Edge
    • Support the evolving needs of the various industries undergoing digitization and of all their respective stakeholders
  • Goals of OEHI
    • Facilitate shorter time to market
    • Create an environment where members can integrate solutions coming from different suppliers
    • Create an open environment for sharing information
    • Raise awareness for capabilities of Arm®- and RISC-V-based solutions
  • Why Arm or RISC-V for HPC?
  • Arm’s Scalable Vector Extension(SVE)
    • Extension of Arm-v8 that includes NEON SIMD instructions
    • Unique feature of SVE: Vector Length Agnostic (VLA)
  • Integration of high-bandwidth memory technologies like HBM
    • Promising to overcome the “memory wall”
    • Arm-based A64FX first standard CPU using HBM
  • Ongoing Activities
    • Software ecosystem for Arm
      • EasyBuild
        • Software stack for HAICGU cluster similar to JSC’s production environment
        • Consistent support of multiple development environments
      • BeeGFS
        • White paper “Performance Evaluation of the BeeGFS File System on the Arm AArch64 Architecture”
    • System software status
      • Operating system
        • Several enterprise-class Linux distributions are supported
      • Cluster management solution
        • Tested solutions available, few supported solutions
      • Resource managers
        • Tested solutions available (e.g. Slurm), limited support
      • Parallel file systems
        • Tested solutions available (e.g. Lustre, BeeGFS), limited support
      • Application porting and development
        • Preparing for Arm’s Scalable Vector Extension
        • Evaluation of different vectorisation strategies
        • Evaluation of auto-vectorisation using synthetic benchmarks like PolyBench/C and ETSVC
    • Training
      • Organisation of several hackathons for Arm
        • Focus on porting of applications to Arm-based platforms
      • Upcoming: DeepHack
        • https://www.open-edge-hpc-initiative.org/2022/03/ai-high-performance-computing-deephack/
        • Organised jointly with EIT Digital
        • Focus on deep learning using Atlas 500 Edge station with Arm-based processors

The full slide deck can be found here.

Top