Notice: Function _load_textdomain_just_in_time was called incorrectly. Translation loading for the complianz-gdpr domain was triggered too early. This is usually an indicator for some code in the plugin or theme running too early. Translations should be loaded at the init action or later. Please see Debugging in WordPress for more information. (This message was added in version 6.7.0.) in /srv/wpfarm/ceec_coe_eu/wordpress/wp-includes/functions.php on line 6114
Our Codes – CEEC CoE

The Codes and Frameworks Used in CEEC

Interested in learning more about the codes and frameworks we’re working with to run our lighthouse cases?

Click on each code below for a technical description, important links, and answers to common questions about the codes and our work with them.

Graph depicting speedup versus the number of GPUs. The x-axis is labeled "GPUs" with values ranging from 8 to 256, and the y-axis is labeled "Speedup" with values from 10^1 to 10^2. Three lines represent different systems: MareNostrum 5 (H100) in blue, Leonardo (A100) in orange, and Karolina (A100) in green. The lines show an upward trend, indicating increased speedup with more GPUs.

Alya/SOD2D

Alya and SOD2D are an HPC simulation framework designed for multiphysics problems. They explicitly address structural mechanics and compressible fluid flows, with a special focus on fluid-structure interaction (FSI). The framework integrates two finite element codes which are optimized to harness the capabilities of modern HPC infrastructures, supporting both CPU-only and GPU-accelerated computing nodes. Alya has shown an excellent speed-up on MareNostrum5-GPP, and SOD2D demonstrates outstanding performance on various GPU-accelerated EuroHPC machines, including MareNostrum5-ACC (H100), Leonardo Booster (A100), and Karolina (A100) using CUDA-aware MPI and OpenACC.

Graph showing the relationship between the number of CPUs/GPUs and speedup. The x-axis is labeled "# CPUs/GPUs" with values ranging from 10^1 to 10^5, and the y-axis is labeled "Speedup" with values from 10^1 to 10^5. Multiple colored lines represent different systems: a dashed line for Ideal, blue for LUMI-C, green for Leonardo-B, orange for HAWK, magenta for MN5-GPP, yellow for JEDI, and orange for Juwels-B. All lines show an upward trend, indicating increased speedup with more CPUs/GPUs.

FLEXI/GALÆXI

FLEXI/GALÆXI is a high-order accurate HPC framework based on the discontinuous Galerkin (DG) spectral element method for hyperbolic-parabolic conservation laws with special focus on compressible flows (at higher Mack numbers). FLEXI has demonstrated excellent scaling on up to five hundred thousand CPUs, and GALÆXI scales well up to thousands of GPUs utilizing MPI-aware CUDA and HIP. The strong scaling results presented here were conducted on various EuroHPC machines, including LUMI-C  (AMD EPYC 7763), MareNostrum5 GPP (Intel Sapphire Rapids 8480+), JEDI (GH200), Leonardo Booster (A100), as well as HAWK (AMD EPYC 7742) and Juwels Booster (A100).

Graph depicting the average time per timestep in seconds on the y-axis versus the number of logical GPUs on the x-axis. The x-axis values are 4,096, 8,192, and 16,384. Three lines are shown: a teal solid line with circles representing "LUMI", an orange solid line with triangles representing "Leonardo", and a black dashed line representing "Ideal". All lines show a downward trend, indicating that as the number of logical GPUs increases, the average time per timestep decreases. The legend is located in the upper right corner of the graph.

Neko

Neko is a portable framework for high-order spectral element based simulations, focusing on the incompressible regime. The framework adopts an object-oriented approach, allowing for multi-tier abstractions of the solver stack and facilitating various hardware backends, ranging from general-purpose processors, accelerators, and vector processors, to limited support of FPGA. Neko demonstrates excellent scaling on the EuroHPC machines LUMI-G (AMD MI250X) and Leonardo Booster (A100).

Graph titled "NekRS strong scaling" with a logarithmic scale on both axes. The vertical axis represents time per step in seconds, ranging from 10^-2 to 10^1, and the horizontal axis represents the number of MPI Ranks (P), ranging from 10^1 to 10^4. Three sets of data are plotted: circles for n = 2048^3, squares for n = 1024^3, and triangles for n = 512^3. Lines are color-coded: dashed cyan for "ideal," solid black for "Juwels Booster (NVIDIA A100)," and solid red for "Summit (NVIDIA V100)." Data points show a downward trend as the number of MPI Ranks increases, indicating improved performance with more ranks.

NekRS/Nek5000

Nek5000/NekRS is a high-order spectral element framework for thermal-fluids simulations, solving the incompressible Navier-Stokes and thermal transport equations. The GPU-accelerated version  NekRS is designed for high-performance on exascale platforms, leveraging performance-portable frameworks like OCCA and optimized kernels from libParanumal. NekRS demonstrates efficient scaling on Summit (V100) and JUWELS Booster (A100) up to thousands of GPUs.

Graph showing the performance of four datasets: LUMI-C (blue), LUMI-G (orange), MN5-GPP (green), and MN5-ACC (red). The x-axis represents "Nodes" ranging from 2^0 to 2^6, and the y-axis represents "Time to solution (ms)" ranging from 2^2 to 2^8. Each dataset shows a downward trend, indicating reduced time to solution as the number of nodes increases.

waLBerla

waLBerla is a modern highly scalable open source multi-physics simulation software framework with a focus on CFD applications such as three-dimensional fully-resolved coupled fluid-particle simulations. The framework employs the lattice Boltzmann method for the fluid phase, and the discrete element method for the granular soil. waLBerla has shown perfect scalability on conventional CPU architectures, and on up to thousands of GPUs utilizing MPI-aware CUDA and HIP. The strong scaling results presented below were obtained on a variety of EuroHPC machines, including LUMI-C (AMD EPYC 7763), LUMI-G (AMD MI250X), MareNostrum5-GPP (Intel Sapphire Rapids 8480+), and MareNostrum5-ACC (H100).