WaLBerla is designed for performance across scales. Written in modern C++20 and utilizing vectorization, OpenMP, CUDA and HIP, the framework makes the most of modern parallel processors of all sizes. Its block-structured computational grids allow for efficient distributed parallelization, effective meshing and grid refinement, as well as dynamic load balancing. Near-perfect scaling behavior has been demonstrated on state-of-the-art supercomputers, such as SuperMUC-NG, JUWELS Booster and LUMI-G.

Powered by the lattice Boltzmann method, waLBerla offers a platform for computational fluid dynamics across a wide range of applications. Using the lbmpy library, applications can choose from a variety of collision models, from the foundational BGK to the highly accurate K17 cumulant method. Complement these with link-wise boundary conditions, use memory-efficient in-place streaming, or include mass-conserving volumetric mesh refinement to construct a flow solver taylored to your needs.

With its versatile Python-based metaprogramming ecosystem, waLBerla tackles the difficulties of modern-day hardware diversity through code generation. Describe numerical methods using a mathematical toolbox and let the code generator take care of the implementation details. Automated generation of hardware-specific parallel kernels makes it easy to target and switch between different platforms without sacrificing performance.

Simulate realistic particle interactions with mesa-pd, waLBerla’s high-performance DEM module. Fully customizable through code generation, mesa-pd supports simulations of rigid body dynamics, contact forces, and complex material models — all with full parallel scalability.

The waLBerla project aims for sustainable software development using latest best practices. A modular code base, clean interfaces, and adherence to modern coding standards are the basis for extensibility and long-term software stability. Continuous integration and testing ensure reproducible builds and code robustness. Being fully open-source, waLBerla welcomes community contributions through its GitLab development platform. The framework is published as free software under the GNU GPLv3.

Tackle complex, real-world problems by coupling LBM, DEM, and other solvers in waLBerla. Construct multiphase and temperature-dependent problems without ever leaving the LBM framework; simulate particulate flows with fully resolved particles; or include complex and moving geometries using the partially saturated cells method.