A framework for hybrid parallel flow simulations with a trillion cells in complex geometries

Godenschwager, Christian; Schornbaum, Florian; Bauer, Martin; Köstler, Harald; Rüde, Ulrich

doi:10.1145/2503210.2503273

Cited by 78 publications

(82 citation statements)

References 28 publications

(37 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Starting as a framework for massively-parallel simulations using the Lattice Boltzmann method, it evolved over the years into a general framework for implementing parallel algorithms on blockstructured grids that scale on various kinds of large HPC clusters [21][22][23]. Therefore, it offers datastructures and algorithms for load balancing, efficient input and output of simulation data as well as helper functions to implement stencil based schemes.…”

Section: Walberla Frameworkmentioning

confidence: 99%

Large scale phase-field simulations of directional ternary eutectic solidification

et al. 2015

Self Cite

View full text Add to dashboard Cite

Section: Walberla Frameworkmentioning

confidence: 99%

Large scale phase-field simulations of directional ternary eutectic solidification

et al. 2015

Self Cite

View full text Add to dashboard Cite

“…Due to the LBM's local nature, this method is well-suited for large parallel computations. Its integration into the WALBERLA framework [6], developed at the University of Erlangen-Nürnberg, provides the necessary performance for more complex, large-scale simulations with several thousand particles [7].…”

Section: Discussionmentioning

confidence: 99%

Simulations of Particle‐laden Flows with the Lattice Boltzmann Method

Rettinger

Rüde

2016

Proc Appl Math and Mech

Self Cite

View full text Add to dashboard Cite

A lattice Boltzmann method is presented which enables the direct numerical simulation of particle-laden flows. It applies a fluid-solid coupling technique based on the momentum exchange method to represent geometrically fully resolved particles of arbitrary shape. The test case of a single moving sphere allows an in-depth evaluation of the algorithm. The method is fully parallelizable and yields results with an accuracy comparable to classical CFD simulations. This forms the basis for simulations of complex systems involving several thousand particles like the sediment transport in river beds. MotivationNumerical simulations have become an important tool to investigate particulate flows, like sediment transport in river beds or fluidization of particle beds. They are used to gain a better understanding of the behavior of such complex systems. To avoid empirical modeling assumptions, one has to resort to direct numerical simulations (DNS) with appropriate fluid-solid coupling algorithms. In computational fluid dynamics (CFD), the lattice Boltzmann method (LBM) has developed into a viable alternative to solve the Navier-Stokes equations numerically. Over the years, various algorithms have been developed for the DNS of particulate flows such as the immersed boundary method, Noble and Torczynski's immersed moving boundary method, and the momentum exchange method. The lattice Boltzmann method presented in this article is suitable for the accurate simulation of particle-laden flows on massively parallel supercomputers. Numerical AlgorithmLattice Boltzmann method: The LBM originates from statistical mechanics and models the evolution of distribution functions f q , corresponding to the lattice velocities c q , on a rectangular grid. Macroscopic quantities can be calculated via moments such as density ρ = q f q and momentum ρ u = q f q c q . By applying the BGK collision model, the evolution equation of the distribution functions readswhich models the linear relaxation of the f q towards their equilibrium values f eq q . The relaxation parameter τ determines the physical viscosity ν. For improved stability and accuracy, more enhanced models like TRT or MRT should be used which then feature several relaxation parameters, see e.g. [1]. Momentum exchange method: This fluid-solid coupling method (see [2]) maps the body explicitly into the domain by marking cells as solid rather than fluid. In order to fulfill the no-slip condition along the body's surface with a local velocity u p , the boundary conditionis applied at the boundary cells. This simple bounce-back scheme can be improved to account for the exact boundary location by schemes like Central Linear Interpolation (CLI) or Multireflection (MR) [1]. On the other hand, the fluid exerts a hydrodynamic force onto the body. For each boundary cell, the local contribution to this force acting at position x b is computed as in [3], taking special care of Galilean invariance, byThe total force and torque on the body is then obtained by summing up all local contributions. Toge...

show abstract

“…In Ref. 3, the flow through a coronary artery tree is computed on a massively parallel scale. The difficulty of this setup arises due to the complex, sparse structure of the artery tree, covering only 0.3% of the volume of the enclosing axis-aligned bounding box.…”

Section: Example Applicationsmentioning

confidence: 99%

“…Here, performance and efficiency are crucial since simulation-based science often requires very fine spatial and temporal resolution to resolve all relevant physical effects. waLBerla is carefully designed as a framework for massively parallel HPC systems, employing only fully distributed data structures [3,4]. Every process holds information only about local and adjacent data.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

waLBerla: A block-structured high-performance framework for multiphysics simulations

Bauer

Eibl

Godenschwager

et al. 2021

Computers & Mathematics with Applications

Self Cite

View full text Add to dashboard Cite

Programming current supercomputers efficiently is a challenging task. Multiple levels of parallelism on the core, on the compute node, and between nodes need to be exploited to make full use of the system. Heterogeneous hardware architectures with accelerators further complicate the development process. waLBerla addresses these challenges by providing the user with highly efficient building blocks for developing simulations on block-structured grids. The block-structured domain partitioning is flexible enough to handle complex geometries, while the structured grid within each block allows for highly efficient implementations of stencil-based algorithms. We present several example applications realized with waLBerla, ranging from lattice Boltzmann methods to rigid particle simulations. Most importantly, these methods can be coupled together, enabling multiphysics simulations. The framework uses meta-programming techniques to generate highly efficient code for CPUs and GPUs from a symbolic method formulation. To ensure software quality and performance portability, a continuous integration toolchain automatically runs an extensive test suite encompassing multiple compilers, hardware architectures, and software configurations.

show abstract

A framework for hybrid parallel flow simulations with a trillion cells in complex geometries

Cited by 78 publications

References 28 publications

Large scale phase-field simulations of directional ternary eutectic solidification

Large scale phase-field simulations of directional ternary eutectic solidification

Simulations of Particle‐laden Flows with the Lattice Boltzmann Method

waLBerla: A block-structured high-performance framework for multiphysics simulations

Contact Info

Product

Resources

About