Open Access

We evaluate masked block-structured grids for ocean domains which allow to represent small-scale geometric features without resorting to very small blocks or excessive mesh resolution. The considered approach aims to combine the geometric flexibility of unstructured meshes with the computational efficiency of stencil-based discretizations and is implemented and tested in a quadrature-free discontinuous Galerkin shallow water solver. We investigate the accuracy and the computational performance of the scheme on a range of realistic ocean domains meshed with blocks of different size and provide some comparisons to results obtained on unmasked block-structured grids and unstructured meshes.

We employ the Partially Saturated Cells Method to model the interaction between the fluid flow and solid moving objects as an extension to the conventional lattice Boltzmann method. We introduce an efficient and accurate method for mapping complex moving geometries onto uniform Cartesian grids suitable for massively parallel processing. A validation of the physical accuracy of the solid–fluid coupling and the proposed mapping of complex geometries is presented. The implementation is integrated into the code generation pipeline of the waLBerla framework so that highly optimized kernels for Central Processing Unit (CPU) and Graphical Processing Unit (GPU) architectures become available. We study the node-level performance of the automatically generated solver routines. 71% of the theoretical peak performance can be achieved on CPU nodes and 86% on GPU accelerated nodes. Only a moderate overhead is observed for the processing of the solid–fluid coupling when compared to the fluids simulations without moving objects. Finally, a counter-rotating open rotor is presented as a prototype industrial scenario, resulting in a mesh size involving up to 4.3 × 109 fluid grid cells. For this scenario, excellent parallel efficiency is reported in a strong scaling study on up to 32 768 CPU cores on the LUMI-C supercomputer and on up to 1024 NVIDIA A100 GPUs on the JUWELS Booster system.

Current supercomputers often have a heterogeneous architecture using both conventional Central Processing Units (CPUs) and Graphics Processing Units (GPUs). At the same time, numerical simulation tasks frequently involve multiphysics scenarios whose components run on different hardware due to multiple reasons, e.g., architectural requirements, pragmatism, etc. This leads naturally to a software design where different simulation modules are mapped to different subsystems of the heterogeneous architecture. We present a detailed performance analysis for such a hybrid four-way coupled simulation of a fully resolved particle-laden flow. The Eulerian representation of the flow utilizes GPUs, while the Lagrangian model for the particles runs on conventional CPUs. Two characteristic model situations involving dense and dilute particle systems are used as benchmark scenarios. First, a roofline model is employed to predict the node level performance and to show that the lattice-Boltzmann-based Eulerian fluid simulation reaches very good performance on a single GPU. Furthermore, the GPU-GPU communication for a large-scale Eulerian flow simulation results in only moderate slowdowns. This is due to the efficiency of the CUDA-aware MPI communication, combined with the use of communication hiding techniques. On 1024 A100 GPUs, an overall parallel efficiency of up to 71% is achieved. While the flow simulation has good performance characteristics, the integration of the stiff Lagrangian particle system requires frequent CPU-CPU communications that can become a bottleneck, especially when simulating the dense particle system. Additionally, special attention is paid to the CPU-GPU communication overhead since this is essential for coupling the particles to the flow simulation. However, thanks to our problem-aware co-partitioning, the CPU-GPU communication overhead is found to be negligible. As a lesson learned from this development, four criteria are postulated that a hybrid implementation must meet for the efficient use of heterogeneous supercomputers.

This poster provides an overview of the recent advances in simulating piping erosion during the installation process of a suction bucket foundation. The lattice Boltzmann method is coupled with the discrete element method. Numerical results are presented for three different suction conditions.

We present our approach and methodology for simulating piping erosion, which occurs during the installation process of suction bucket foundations for offshore wind turbines. We show several simulations and analyze the hydrodynamic and contact forces acting on the granular fabric and the differential pressure of the fluid phase. We demonstrate weak scaling performance on the LUMI supercomputer with a parallel efficiency of up to 90% for 4096 Graphics Compute Dies.