Regularized lattice Boltzmann method parallel model on heterogeneous platforms

SummaryThe execution of agent‐based simulations (ABSs) on hardware accelerator devices such as graphics processing units (GPUs) has been shown to offer great performance potentials. However, in heterogeneous hardware environments, it can become increasingly difficult to find viable partitions of the simulation and provide implementations for different hardware devices. To automate this process, we present OpenABLext, an extension to OpenABL, a model specification language for ABSs. By providing a device‐aware OpenCL backend, OpenABLext enables the co‐execution of ABS on heterogeneous hardware platforms consisting of central processing units, GPUs, and field programmable gate arrays (FPGAs). We present a novel online dispatching method that efficiently profiles partitions of the simulation during run‐time to optimize the hardware assignment while using the profiling results to advance the simulation itself. In addition, OpenABLext features automated conflict resolution based on user‐specified rules, supports graph‐based simulation spaces, and utilizes an efficient neighbor search algorithm. We show the improved performance of OpenABLext and demonstrate the potential of FPGAs in the context of ABS. We illustrate how co‐execution can be used to further lower execution times. OpenABLext can be seen as an enabler to tap the computing power of heterogeneous hardware platforms for ABS.

On the coupling of a direct-forcing immersed boundary method and the regularized lattice Boltzmann method for fluid-structure interaction

The generalized lattice Boltzmann equation (GLBE), with the addition of the standard Smagorinsky subgrid-stress (SGS) model, has been proved that it is more suitable for simulating high Reynolds number turbulent flows when compared with the lattice BGK Boltzmann equation (LBGK). However, the computing efficiency of lattice Boltzmann method (LBM) is too low to make it for practical applications, unless using a massive parallel computing clusters facility. In this study, the massive parallel computing power from an inexpensive graphic processor unit (GPU) and a typical personal computer has been developed for improving the computing efficiency, more than 100 times. This developed three-dimensional (3-D) GLBE-SGS model, with the D3Q19 scheme for simplifying collision and streaming courses, has been successfully used to study 3-D rectangular cavity flows with Reynolds number up to 10000. generalized lattice Boltzmann equation (GLBE), subgrid-stress (SGS), D3Q19, 3-D rectangular cavity flows, GPU

Three-dimensional simulations of buoyancy-driven flow of two immiscible liquids are performed using lattice Boltzmann method (LBM) implemented on a graphics processing unit (GPU). Graphics processing unit is a new paradigm for computing fluid flows and has become more popular in the recent years. It is a powerful and convenient to use. LBM, which is an excellent alternative technique for fluid flow simulation, when implemented on GPUs gives a very high computational speed-up. Our present GPU based LBM solver gives a speed-up 25 times corresponding CPU based code.

I present two novel thread-safe in-place streaming schemes for the lattice Boltzmann method (LBM) on graphics processing units (GPUs), termed Esoteric Pull and Esoteric Push, that result in the LBM only requiring one copy of the density distribution functions (DDFs) instead of two, greatly reducing memory demand. These build upon the idea of the existing Esoteric Twist scheme, to stream half of the DDFs at the end of one stream-collide kernel and the remaining half at the beginning of the next, and offer the same beneficial properties over the AA-Pattern scheme—reduced memory bandwidth due to implicit bounce-back boundaries and the possibility of swapping pointers between even and odd time steps. However, the streaming directions are chosen in a way that allows the algorithm to be implemented in about one tenth the amount of code, as two simple loops, and is compatible with all velocity sets and suitable for automatic code-generation. The performance of the new streaming schemes is slightly increased over Esoteric Twist due to better memory coalescence. Benchmarks across a large variety of GPUs and CPUs show that for most dedicated GPUs, performance differs only insignificantly from the One-Step Pull scheme; however, for integrated GPUs and CPUs, performance is significantly improved. The two proposed algorithms greatly facilitate modifying existing code to in-place streaming, even with extensions already in place, such as demonstrated here for the Free Surface LBM implementation FluidX3D. Their simplicity, together with their ideal performance characteristics, may enable more widespread adoption of in-place streaming across LBM GPU codes.

GPU accelerated numerical study of PCM melting process in an enclosure with internal fins using lattice Boltzmann method

A Lattice-Boltzmann solver for 3D fluid simulation on GPU

Suspension flows are ubiquitous in many fields, and reducing computational costs for numerical simulations of suspension flows remains a great challenge. Numerical simulations of particle-laden flows were performed here using the regularized lattice Boltzmann method (RLBM). RLBM, coupled with the virtual flux method (VFM) or the immersed boundary method (IBM) with the multi-direct forcing approach, conducted a series of three simulations for stationary and moving boundaries. The VFM and IBM results were compared for validation, and the computational efficiency was evaluated. For the two simulations using stationary and moving boundaries (i.e., a steady flow past a stationary cylinder and a plane Poiseuille flow including a single moving circular particle), the results based on the VFM and IBM were qualitatively comparable. The computational time for IBM was longer than that for VFM due to the iterations in its scheme. The pressure-driven suspension flow simulations showed that a dilute suspension's (<5%) relative viscosity was almost equivalent. The difference between the results obtained by the VFM and IBM in the semidilute regime (>10%) was discussed in terms of particle rotational motion. Even though the scalability for parallel computing was comparable between the VFM and IBM, the computational time could be effectively reduced by using the VFM instead of the IBM because of the increased suspension concentration.

Pore scale characteristics of gas flow in shale matrix determined by the regularized lattice Boltzmann method

The TheLMA project: A thermal lattice Boltzmann solver for the GPU

Nanoscale fluid-ion transport is investigated in biophysical chemistry, drug delivery, protein sequencing, etc. Currently, fast three-dimensional models for fluid-ion transport through biological nanopores are unavailable. This study, therefore, focuses on the simulation and parallelization of nanoscale fluid-ion transport on multiple graphics processing units (GPUs). Nanoscale fluid-ion transport is described through the fourth-order Poisson-Nernst-Planck-Bikerman model coupled with the Navier-Stokes equations. The model incorporates the effect of ionic and non-ionic interactions, ion solvation, nonlocal electrostatics, and the finite size of particles. Governing equations are discretized using the lattice Boltzmann method (LBM). For complex geometries, the immersed boundary method has been incorporated with the LBM. To demonstrate the applicability of the developed model, electro-osmotic flow through a relatively large biological nanopore has been simulated. Parallelization on multiple GPUs has enabled us to perform simulations for a full three-dimensional channel geometry. Results showed a good match with the published data. We captured local variations in concentration and fluid flow in both axial and radial directions. Furthermore, flow circulation around the channel was also observed. The impact of external potential difference and the finite size of the particles on the flow has also been assessed. Using our in-house code, we achieved a performance of 1982 × 106 lattice updates per second on 8 A100 80 gigabyte GPUs. The methodology presented here provides an accurate and fast simulation methodology for characterizing symmetric/asymmetric nanopore properties. This model can help in enhancing the fundamental understanding of nanofluidics and characterization of the nanopores.

Computing speed is a significant issue of large-scale flood simulations for real-time response to disaster prevention and mitigation. Even today, most of the large-scale flood simulations are generally run on supercomputers due to the massive amounts of data and computations necessary. In this work, a two-dimensional shallow water model based on an unstructured Godunov-type finite volume scheme was proposed for flood simulation. To realize a fast simulation of large-scale floods on a personal computer, a Graphics Processing Unit (GPU)-based, high-performance computing method using the OpenACC application was adopted to parallelize the shallow water model. An unstructured data management method was presented to control the data transportation between the GPU and CPU (Central Processing Unit) with minimum overhead, and then both computation and data were offloaded from the CPU to the GPU, which exploited the computational capability of the GPU as much as possible. The parallel model was validated using various benchmarks and real-world case studies. The results demonstrate that speed-ups of up to one order of magnitude can be achieved in comparison with the serial model. The proposed parallel model provides a fast and reliable tool with which to quickly assess flood hazards in large-scale areas and, thus, has a bright application prospect for dynamic inundation risk identification and disaster assessment.

A modified power-law (MPL) viscosity model of non-Newtonian fluid flow has been used for the multiple-relaxation-time (MRT) lattice Boltzmann methods (LBM) and then validated with the benchmark problems using the graphics process unit (GPU) parallel computing via Compute Unified Device Architecture (CUDA) C platform. The MPL model for characterizing the non-Newtonian behavior is an empirical correlation that considers the Newtonian behavior of a non-Newtonian fluid at a very low and high shear rate. A new time unit parameter (λ) governing the flow has been identified, and this parameter is the consequence of the induced length scale introduced by the power law. The MPL model is free from any singularities due to the very low or even zero shear-rate. The proposed MPL model was first validated for the benchmark study of the lid-driven cavity and channel flows. The model was then applied for shear-thinning and shear-thickening fluid flows through a backward-facing step with relatively low Reynolds numbers, Re = 100–400. In the case of shear-thinning fluids (n=0.5), laminar to transitional flow arises while Re≥300, and the large vortex breaks into several small vortices. The numerical results are presented regarding the velocity distribution, streamlines, and the lengths of the reattachment points.

Computational Fluid Dynamics (CFD) uses numerical methods to solve problems involving fluid flows, e.g. in automotive engineering, energy, civil engineering, and aerospace. One of these methods is the lattice Boltzmann method (LBM). Beside giving accurate results for a wide range of complex flows, the LBM lends itself to efficient implementations on massively parallel systems such as the general-purpose computing on graphics processing units (GPGPU). This work considers the cumulant LBM, which overcomes some problems of classical LBMs, such as the violation of the Galilean invariance, the spurious coupling of the degrees of freedoms, and the hyper-viscosity. In order to enable the cumulant LBM for complex CFD engineering problems, this work focuses on the discretization of the computational domain and the analysis of turbulent flows in the near-wall region.&#13;\nThe grid generation for LBM deals with several issues such as the second order boundary definition, the free shape grid refinement, and the level wise load balancing for parallel meshes. While discretizing complex geometries, these issues have to be handled simultaneously.&#13;\nTurbulent flows close to walls are characterized by high velocity gradients in a thin region called the boundary layer. In order to resolve numerically the boundary layer, a grid with a high resolution is required resulting in a high computational cost for the simulation. Empirical functions can be used for modelling the boundary layer, allowing to estimate the quantities at the wall even with coarse grids, and thus increasing the efficiency of the numerical simulation.&#13;\nIn this work, a new grid generator that addresses the LBM discretization issues is implemented. It creates free shape multi-level three-dimensional meshes with second order accurate boundary definition for very complex geometries. It can be applied for generating grids of complex bodies, such as cars, porous media, and urban areas.&#13;\nRegarding the near-wall region treatment, a new wall function is introduced. It uses local information at the boundary nodes for recovering the quantities at the wall. For generating a proper turbulence for wall bounded flows, a new set of relaxation parameters is introduced, which eliminates the spurious dependence of the error on the bulk viscosity.&#13;\nIn conclusion, this work addresses two important aspects for enabling the cumulant LBM for complex fluid flow problems: grid generation and boundary treatment for turbulent flows.

Hybrid MPI and CUDA paralleled finite volume unstructured CFD simulations on a multi-GPU system