Large-scale simulations on multiple Graphics Processing Units (GPUs) for the direct simulation Monte Carlo method

Su, C. C.; Smith, Matthew R.; Kuo, Fang-An; Wu, Jingwei; Hsieh, Chih‐Wei; Tseng, Kuo-Hung

doi:10.1016/j.jcp.2012.07.038

Cited by 29 publications

(21 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Simulation Case C: supersonic flow over a block A two dimensional supersonic flow over a rectangular block was investigated in this case. It was before simulated by bird, 18 Su et al 19 and Goldsworthy. …”

Section: Simulation Case B: Hypersonic Flow Around a Cylinder Argonmentioning

confidence: 99%

DgSMC-B code: A robust and autonomous direct simulation Monte Carlo code for arbitrary geometries

2016

View full text Add to dashboard Cite

In this paper, we describe the structure of a new Direct Simulation Monte Carlo (DSMC) code that takes advantage of combinatorial geometry (CG) to simulate any rarefied gas flows Medias. The developed code, called DgSMC-B, has been written in FORTRAN90 language with capability of parallel processing using OpenMP framework. The DgSMC-B is capable of handling 3-dimensional (3D) geometries, which is created with first-and second-order surfaces. It performs independent particle tracking for the complex geometry without the intervention of mesh. In addition, it resolves the computational domain boundary and volume computing in border grids using hexahedral mesh. The developed code is robust and selfgoverning code, which does not use any separate code such as mesh generators. The results of six test cases have been presented to indicate its ability to deal with wide range of benchmark problems with sophisticated geometries such as airfoil NACA 0012. The DgSMC-B code demonstrates its performance and accuracy in a variety of problems. The results are found to be in good agreement with references and experimental data. C 2016 Author(s). All article content, except where otherwise noted, is licensed under a Creative Commons Attribution (CC BY) license

show abstract

Section: Simulation Case B: Hypersonic Flow Around a Cylinder Argonmentioning

confidence: 99%

DgSMC-B code: A robust and autonomous direct simulation Monte Carlo code for arbitrary geometries

2016

View full text Add to dashboard Cite

show abstract

“…This flexible method allows a single subroutine to manage communications without concern for (i) the device type, or (ii) if the device is located on the same physical node (see Ref. [52] for more details).…”

Section: Multi-gpu Implementation Of the Dsmc Solvermentioning

confidence: 99%

Adaptive kinetic-fluid solvers for heterogeneous computing architectures

Zabelok

Arslanbekov

Kolobov

2015

Journal of Computational Physics

View full text Add to dashboard Cite

Abstract. We show feasibility and benefits of porting an adaptive multi-scale kinetic-fluid code to CPU-GPU systems. Challenges are due to the irregular data access for adaptive Cartesian mesh, vast difference of computational cost between kinetic and fluid cells, and desire to evenly load all CPUs and GPUs. Our Unified Flow Solver (UFS) combines Adaptive Mesh Refinement (AMR) with automatic cell-by-cell selection of kinetic or fluid solvers based on continuum breakdown criteria. Using GPUs enables hybrid simulations of mixed rarefied-continuum flows with a million of Boltzmann cells with 24x24x24 velocity mesh. We describe the implementation of CUDA kernels for three modules in UFS: the direct Boltzmann solver using discrete velocity method, the Direct Simulation Monte Carlo (DSMC) solver, and a mesoscopic solver based on Lattice Boltzmann Method, all using octree Cartesian mesh. Double digit speedups on single GPU and good scaling for multiGPUs have been demonstrated.

show abstract

“…Of the many domain decomposition approaches, the one used by Ivanov et al 27 and Su et al, 26 is to divide the domain into equal sub-domains, depending on the number of GPUs. Each sub-domain is assigned to one GPU, which constructs a uniform Cartesian mesh 27,26 on its sub-domain and independently proceeds with the DSMC steps.…”

Section: B Domain Decompositionmentioning

confidence: 99%

A Hybrid CPU-GPU Parallel Octree Direct Simulation Monte Carlo Approach

Jambunathan

Levin

2015

22nd AIAA Computational Fluid Dynamics Conference

View full text Add to dashboard Cite

The gridfree octree DSMC approach is parallelized with GPGPUs using CUDA. A linear space filling Morton Z-curve is employed to represent the three dimensional octree structure in a linear array and the advantages of this implementation is discussed. The methodology for a multi-GPU, hybrid MPI-CUDA implementation, involving load balanced domain decomposition and inter-GPU communications is presented. Nearly 70-90 times speedup was achieved by the multi-GPU CHAOS solver using 8 GPUs as compared to the serial octree DSMC implementation. When compared to a single GPU, the simulation performed using 8 GPUs showed 5-6 times speedup, i.e., an efficiency of 55-75% was achieved depending on the type of flow condition being analysed. The single-GPU simulations exhibited 12-30 times speedup compared to the serial gridfree octree DSMC simulation.

show abstract

Large-scale simulations on multiple Graphics Processing Units (GPUs) for the direct simulation Monte Carlo method

Cited by 29 publications

References 9 publications

DgSMC-B code: A robust and autonomous direct simulation Monte Carlo code for arbitrary geometries

DgSMC-B code: A robust and autonomous direct simulation Monte Carlo code for arbitrary geometries

Adaptive kinetic-fluid solvers for heterogeneous computing architectures

A Hybrid CPU-GPU Parallel Octree Direct Simulation Monte Carlo Approach

Contact Info

Product

Resources

About