Highly efficient lattice Boltzmann multiphase simulations of immiscible fluids at high-density ratios on CPUs and GPUs through code generation

Holzer, Markus; Bauer, Martin; Köstler, Harald; Rüde, Ulrich

doi:10.1177/10943420211016525

Cited by 19 publications

(5 citation statements)

References 43 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Hereafter, the performance model relies on the memory bandwidth and the number of memory accesses, as LB schemes are generally memory-bound on GPUs. 12,13 Applying this model to our family of LB schemes leads to the following theoretical peak performance:…”

Section: Performance Analysismentioning

confidence: 99%

“…For all these reasons, isothermal and weakly compressible LBMs are very well-suited for parallel and high performance computing on both central processing units (CPUs) 11 and graphical processing units (GPUs). 12,13 Unfortunately, the development of fully compressible LBMs is not as straightforward and successful. 14 In fact, it took more than a decade to get the right formulation of the velocity discretization, equilibrium state, and collision model to recover the polyatomic compressible NSF equations with variable Prandtl number.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Exponential distribution functions for positivity-preserving lattice Boltzmann schemes: Application to 2D compressible flow simulations

Thyagarajan,

Coreixas,

Latt

2023

Physics of Fluids

View full text Add to dashboard Cite

A family of positivity-preserving lattice Boltzmann methods (LBMs) is proposed for compressible flow simulations in the continuum regime. It relies on the efficient collide-and-stream algorithm with a collision step based on exponential distribution functions. The latter serves as a generalization of Grad's post-collision distribution functions for which here (1) the linearized non-equilibrium contributions are replaced by their exponential forms and (2) the number of these contributions can be chosen arbitrary. In practice, post-collision moments of our exponential formulation are enforced through an iterative moment-matching approach to recover any macroscopic physics of interest, with or without external forces. This methodology directly flows from the extended framework on numerical equilibria [J. Latt et al., Philos. Trans. R. Soc. A 378, 20190559 (2020)] and goes one step further by allowing for the independent relaxation of hydrodynamic and high-order modes in a given moment space, notably, making the Prandtl number freely adjustable. The model is supplemented by a shock-capturing technique, based on the deviation of non-equilibrium moments from their equilibrium counterparts, to ensure good numerical properties of the model in inviscid and under-resolved conditions. A second exponential distribution accounts for extra degrees of freedom of molecules and allows for the simulation of polyatomic gases. To validate this novel approach and to quantify the accuracy of different lattices and moment closures, several 2D benchmark tests of increasing complexity are considered: double shear layer, linear wave decay, Poiseuille flow, Riemann problem, compressible Blasius flow over a flat plate, and supersonic flow past an airfoil. Corresponding results confirm the accuracy and stability properties of our approach for the simulation of compressible flows with LBMs. Eventually, the performance analysis further highlights its efficiency on general purpose graphical processing units.

show abstract

Section: Performance Analysismentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Exponential distribution functions for positivity-preserving lattice Boltzmann schemes: Application to 2D compressible flow simulations

Thyagarajan,

Coreixas,

Latt

2023

Physics of Fluids

View full text Add to dashboard Cite

show abstract

“…3.1 waLBerla waLBerla [9,10] is a massively parallel multi-physics framework developed at Friedrich-Alexander University Erlangen-Nuremberg (FAU) with focus on simulations using the lattice Boltzmann method. Its specific software design combined with the code generation framework lbmpy [18] allows for highly optimised, performance-portable solutions on block-structured grids for various application areas, like phase-field simulations, and particle-laden flows [19,20]. It supports shared and distributed memory parallelism with OpenMP and MPI, respectively, automated SIMD vectorisation, and the execution on NVIDIA graphics cards.…”

Section: Framework Descriptionmentioning

confidence: 99%

Evaluation of a lattice Boltzmann-based wind-turbine actuator line model against a Navier-Stokes approach

Helen

Anciaux–Sedrakian

Blondel

et al. 2022

J. Phys.: Conf. Ser.

Self Cite

View full text Add to dashboard Cite

Due to the cost and difficulty to precisely measure aerodynamic quantities in onshore and offshore wind farms, researchers often rely on high-fidelity large eddy simulation, based on Navier-Stokes flow solvers. However, the cost of such simulation is very high and does not allow, in practice, extensive parametric studies for large wind farms. Among others, the lattice Boltzmann method is a good candidate for much faster, ExaScale wind farm flow simulations. The present paper aims to assess the validity of a lattice Boltzmann-based actuator line model and highlights its strengths and potential weaknesses. With this intent, comparisons against a Navier-Stokes approach commonly used in the wind energy community are performed. We assess the potential of the lattice Boltzmann method to reduce the computational cost of such simulations by analyzing the performance of the different solvers and their scalability. The lattice Boltzmann-based waLBerla solver reduces the computational costs significantly compared to SOWFA while maintaining the same accuracy as the Navier-Stokes-based method. Furthermore, we show that a multi-GPU implementation leads to an even more drastic reduction of the computational time, achieving faster-than-real-time simulations. This performance will allow extensive parametric studies over large wind farms in future studies.

show abstract

“…Thus, complex mathematical models like multi-phase solidification models [1] can be described in a high-level latex-like representation close to usual descriptions in the literature. Furthermore, other packages like the Lattice-Boltzmann-Method (LBM) code generation framework lbmpy [2], [3] build on top of pystencils to derive highly optimized numerical schemes for solving flow problems. From the numerical schemes, pystencils then generates low-level C/OpenMP code to target CPUs or CUDA/OpenCL for GPUs.…”

Section: B Pystencilsmentioning

confidence: 99%

“…The curvature of the phase field is computed with a finite difference discretization, which adds a 3D7pt stencil to the conventional, very memory intensive D3Q15 LBM stencil. This makes it particularly complicated to achieve good performance results [3], but at the same time interesting for automatic performance estimations, which can give significant insight into the complex structure of the compute kernels.…”

Section: Multi Phase Lbmmentioning

confidence: 99%

Opening the Black Box: Performance Estimation during Code Generation for GPUs

Ernst¹,

Hager²,

Holzer³

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

Automatic code generation is frequently used to create implementations of algorithms specifically tuned to particular hardware and application parameters. The code generation process involves the selection of adequate code transformations, tuning parameters, and parallelization strategies. To cover the huge search space, code generation frameworks may apply time-intensive autotuning, exploit scenario-specific performance models, or treat performance as an intangible black box that must be described via machine learning.This paper addresses the selection problem by identifying the relevant performance-defining mechanisms through a performance model coupled with an analytic hardware metric estimator. This enables a quick exploration of large configuration spaces to identify highly efficient candidates with high accuracy.Our current approach targets memory-intensive GPGPU applications and focuses on the correct modeling of data transfer volumes to all levels of the memory hierarchy. We show how our method can be coupled to the "pystencils" stencil code generator, which is used to generate kernels for a range four 3D25pt stencil and a complex two phase fluid solver based on the Lattice Boltzmann Method. For both, it delivers a ranking that can be used to select the best performing candidate.The method is not limited to stencil kernels, but can be integrated into any code generator that can generate the required address expressions.

show abstract

Highly efficient lattice Boltzmann multiphase simulations of immiscible fluids at high-density ratios on CPUs and GPUs through code generation

Cited by 19 publications

References 43 publications

Exponential distribution functions for positivity-preserving lattice Boltzmann schemes: Application to 2D compressible flow simulations

Exponential distribution functions for positivity-preserving lattice Boltzmann schemes: Application to 2D compressible flow simulations

Evaluation of a lattice Boltzmann-based wind-turbine actuator line model against a Navier-Stokes approach

Opening the Black Box: Performance Estimation during Code Generation for GPUs

Contact Info

Product

Resources

About