Last week, members of the EU’s CoE RAISE project participated in an “all hands” meeting at CERN. This innovative project is developing an artificial intelligence (AI) approach for next-generation ‘exascale’ supercomputers for use in both science and industry. Use cases explored throughout this project include wind farm layout optimization, efficient aircraft design, acoustic engineering improvement, and seismic imaging with remote sensing.
CoE RAISE – European Center of Excellence in Exascale Computing “Research on AI and Simulation-Based Engineering at Exascale” – is funded under the EU’s Horizon 2020 research and innovation programme. The project will start in 2021 and run for three years.
Fifty-four project members attended the four-day conference held in CERN’s Council Chamber. Participants discussed progress in developing AI technologies in Europe for complex applications running on future “exascale” high-performance computing (HPC) systems. Exascale refers to next-generation high-performance computers capable of doing 10 or more operations.18 Floating point operations per second (FLOPS). Currently, only the Frontier supercomputer at Oak Ridge National Laboratory in the United States has reached this level. However, with more exascale HPC systems coming soon, it is important to ensure that AI approaches used in science and industry can fully exploit their enormous potential. In June, the European High Performance Computing Joint Undertaking (EuroHPC JU) announced that his Forschungszentrum Jülich GmbH in Germany was selected to host and operate Europe’s first exascale supercomputer. Undertake innovative and transformative exascale research pioneers).
CoE RAISE develops innovative AI methodologies in heterogeneous HPC architectures containing multiple types of processors. Such architectures can provide higher performance and energy efficiency, but code must be adapted to efficiently use different types of processors. The AI methods under development focus on nine key use cases and are designed to scale well to run on exascale HPC systems.
CoE RAISE supports technology transfer to industry, especially small and medium enterprises, and carries out education and training initiatives. In addition to this, CoE RAISE will provide consultancy and work with other European initiatives to maximize synergies, capitalize on co-design opportunities and share knowledge. All aspects of the project’s work were discussed over four days at his CERN.
CERN is also a partner and brings one of its use cases to the project. This work focuses on improving methods for reconstructing particle collision events at the upgraded High Luminous Large Hadron Accelerator (HL-LHC). HL-LHC is scheduled to come online in 2029. Exabytes of data are generated each year, creating unprecedented computing challenges. Hundreds of different algorithms run simultaneously to reconstruct today’s particle impact events (data sets on the order of terabytes or petabytes). Some are traditional algorithms optimized for specific hardware configurations, while others already include AI-driven methods such as deep neural networks. (DNN). Members of the project team at CERN are working to make the system more modular and code optimized to take full advantage of heterogeneous architectures. We are also increasing our use of machine learning and other AI techniques for collision reconstruction and particle classification.
“Supercomputers are reaching exascale and capable of providing unprecedented scales of processing resources for HPC and AI workflows,” says Maria Girone, CERN OpenLabs CTO, who leads CERN’s contributions to the project. said. “The research done at CoE RAISE will drive the collaborative design of HPC computing resources for future AI and HPC applications for both science and industry. We have been able to develop new perspectives and bring new perspectives, as well as give researchers in other fields unique insights into the environment and challenges facing CERN, fostering cross-fertilization and understanding.”
