CEEC

As supercomputers become more powerful and numerical algorithms more efficient, scientists are able to simulate complex systems with increasing realism. Computational fluid dynamics, a computationally demanding set of methods that are essential for many scientific and engineering fields, is one important field that is driving the continued evolution of high-performance computing toward the exascale, the next order of magnitude in supercomputing power. Because of CFD’s focus on scaling to simulate larger systems at greater precision, and because many highly efficient CFD simulation codes already exist, it is one of the few application areas with the potential to reach exascale performance.

Currently, however, most CFD codes have not been adapted to run reliably and efficiently on extreme-scale systems. They can use huge amounts of energy, run slowly, and be prone to crashing due to errors, making them impractical. Improving codes to run exascale simulations efficiently is thus a major challenge currently facing the field of CFD.

Building on the work of previous, complementary projects like EXCELLERAT, the EuroHPC Joint Undertaking’s Centre of Excellence in Exascale CFD (CEEC) will improve the performance of five European CFD codes, which address a large spectrum of CFD applications, including compressible, incompressible, and multiphase flows. Improvements in these codes will be tested in lighthouse cases, focusing on applications with high impact in industrial practice, including designing more fuel-efficient aircraft, improving static mixers for blending fluids, managing soil erosion in offshore wind farms, modelling atmospheric boundary layer flows, and reducing turbulence in ship hull design. In contrast with other projects, CEEC will focus on running large-scale simulations that require the use of an entire exascale supercomputer.

CEEC will implement exascale-ready workflows for addressing relevant challenges for new and upcoming pre-exascale and exascale systems, including those being procured by the EuroHPC Joint Undertaking. To improve scalability, energy efficiency, and usability, CEEC will implement a variety of techniques and technologies, including:

• efficient exploitation of accelerated hardware architectures (GPUs)
• adaptive mixed-precision algorithms
• efficient uncertainty quantification at scale
• data compression
• turbulence closure models
• hybrid remote visualization

HLRS is contributing its expertise in visualization and data management to CEEC and is leading the project’s communication activities.