Performance improvements of differential operators code for MPS method on GPU

Murotani, Kenta; Masaie, Issei; Matsunaga, Takuya; Koshizuka, Seiichi; Shioya, Ryuji; Ogino, Masao; Fujisawa, Toshimitsu

doi:10.1007/s40571-015-0059-2

Cited by 17 publications

(6 citation statements)

References 41 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…To ensure computational stability and efficiency, it is recommended that the influence radius be 2-4 times the initial particle distance [16]. For particle i, the particle number density can be defined as…”

Section: Explicit Mps Methodsmentioning

confidence: 99%

“…The main limitation of MPS is that it can be computationally expensive and requires a large number of particles to accurately capture the flow behavior. To overcome this limitation, the parallel computing technique can be used to improve its efficiency and make it feasible for practical applications [16][17][18]. As a massively parallel processor, the GPU can achieve high accelerations and efficiency using CUDA or OpenCL programming frameworks [19,20].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

An Efficient Explicit Moving Particle Simulation Solver for Simulating Free Surface Flow on Multicore CPU/GPUs

Zhao,

Jiang,

Mochizuki

2024

Modelling

View full text Add to dashboard Cite

The moving particle simulation (MPS) method is a simulation technique capable of calculating free surface and incompressible flows. As a particle-based method, MPS requires significant computational resources when simulating flow in a large-scale domain with a huge number of particles. Therefore, improving computational speed is a crucial aspect of current research in particle methods. In recent decades, many-core CPUs and GPUs have been widely utilized in scientific simulations to significantly enhance computational efficiency. However, the implementation of MPS on different types of hardware is not a trivial task. In this study, we present an implementation method for the explicit MPS that utilizes the Taichi parallel programming language. When it comes to CPU computing, Taichi’s computational efficiency is comparable to that of OpenMP. Nevertheless, when GPU computing is utilized, the acceleration of Taichi in parallel computing is not as fast as the CUDA implementation. Our developed explicit MPS solver demonstrates significant performance improvements in simulating dam-break flow dynamics.

show abstract

Section: Explicit Mps Methodsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

An Efficient Explicit Moving Particle Simulation Solver for Simulating Free Surface Flow on Multicore CPU/GPUs

Zhao,

Jiang,

Mochizuki

2024

Modelling

View full text Add to dashboard Cite

show abstract

“…Schematic diagram of the CD‐WC‐MPS coupled with geometrically nonlinear shell numerical algorithm (The reader interested on the bucket‐based domain decomposition is referred to Reference 92) [Colour figure can be viewed at wileyonlinelibrary.com]…”

Section: Methodsmentioning

confidence: 99%

Three‐dimensional weakly compressible moving particle simulation coupled with geometrically nonlinear shell for hydro‐elastic free‐surface flows

Junior

Neto

Cheng

2022

Numerical Methods in Fluids

View full text Add to dashboard Cite

A three-dimensional fluid-structure interaction solver based on an improved weakly compressible moving particle simulation (WC-MPS) method and a geometrically nonlinear shell structural model is developed and applied to hydro-elastic free-surface flows. The fluid-structure coupling is performed by a polygon wall boundary model that can handle particles and finite elements of distinct sizes. In WC-MPS, a tuning-free diffusive term is introduced to the continuity equation to mitigate nonphysical pressure oscillations. Discrete divergence operators are derived and applied to the polygon wall boundary, of which the numerical stability is enhanced by a repulsive Lennard-Jones force.Additionally, an efficient technique to deal with the interaction between fluid particles placed at opposite sides of zero-thickness walls is proposed. The geometrically nonlinear shell is modeled by an unstructured mesh of six-node triangular elements. Finite rotations are considered with Rodrigues parameters and a hyperelastic constitutive model is adopted. Benchmark examples involving free-surface flows and thin-walled structures demonstrate that the proposed model is robust, numerically stable and offers more efficient computation by allowing mesh size larger than that of fluid particles.

show abstract

“…During recent years, increased focus has shifted to improving the computational efficiency of particle methods through a multigrid scheme (Södersten et al, 2019), parallelization (Guo et al, 2018), and running particle methods on graphics processing units (GPUs) (Murotani et al, 2015;Chow et al, 2018). Moreover, several multi-resolution (Tang Axel SÖDERSTEN*, Takuya MATSUNAGA*, Seiichi KOSHIZUKA*, Tomoyuki HOSAKA** and Eiji ISHII* et al, 2016a;Tang et al, 2016b;Chen et al, 2016;Tanaka et al, 2018;Khayyer et al, 2019;Wang et al, 2019) schemes have been proposed, where high spatial resolution is only considered over critical regions of the flow, such as around clearances (Tanaka et al, 2018) and at impact with walls (Chen et al, 2016;Tanaka et al, 2018), rigid structures (Wang et al, 2019) and elastic beams (Khayyer et al, 2019;Khayyer et al, 2021).…”

Section: Introductionmentioning

confidence: 99%

Adaptive resizing-based multi-resolution particle method

Södersten

Matsunaga

Koshizuka

et al. 2022

Mechanical Engineering Journal

Self Cite

View full text Add to dashboard Cite

Multi-resolution techniques are essential for the performance of large-scale simulations of computational fluid dynamics with particle methods. In this study, a novel adaptive multi-resolution scheme has been developed for the least squares moving particle semi-implicit (LSMPS) method. With the proposed technique, particles are dynamically resized based on a local error estimate. The error estimate is defined by how well the velocity of a particle can be approximated by the Taylor-series expansions of its fluid neighbours, and vice versa for wall neighbours. Due to the adaptiveness of the multi-resolution technique, time-consuming optimization of predefined particle size targets is avoided. The adaptiveness also enables particle resizing which tracks transient resolution changes of the flow. Therefore, the adaptiveness should improve the computational efficiency of the multi-resolution method. In this study, the multi-resolution technique was tested for a two-dimensional eccentric rotating cylinder problem with a small clearance and a known steady-state solution. As expected, initially uniform particle sizes quickly decreased around the clearance. The particle size distribution evolution was smooth in both time and space throughout the simulations. Consequently, the multi-resolution method gave significantly more accurate results than a single resolution method with the same number of particles and time-step length. A drawback with the multi-resolution scheme is that the restrictions on time-step lengths become tighter. This issue is considered by an ongoing development of a multi-time stepping scheme.

show abstract

Performance improvements of differential operators code for MPS method on GPU

Cited by 17 publications

References 41 publications

An Efficient Explicit Moving Particle Simulation Solver for Simulating Free Surface Flow on Multicore CPU/GPUs

An Efficient Explicit Moving Particle Simulation Solver for Simulating Free Surface Flow on Multicore CPU/GPUs

Three‐dimensional weakly compressible moving particle simulation coupled with geometrically nonlinear shell for hydro‐elastic free‐surface flows

Adaptive resizing-based multi-resolution particle method

Contact Info

Product

Resources

About