Multi-block adaptive mesh refinement (AMR) for a lattice Boltzmann solver using GPUs

Fu Sheng Hsu, Keh-Chin Chang, Matt-Hew Smith

Research output: Contribution to journalArticle

1 Citation (Scopus)

Abstract

A parallelized adaptive mesh refinement (AMR) approach using lattice Boltzmann solver [1] using GPU acceleration is developed based around a multi-block structured uniform mesh code [2, 3]. AMR is obtained by deploying multiple levels of mesh blocks, with varying resolution employed within each block, for increased resolution of flows in regions requiring higher accuracy. A simple three dimensional benchmark is employed for confirming the accuracy and parallel performance of the AMR implementation. A series of benchmark tests has been performed with the flow fields and parallel performance compared. Comparisons of the computation time between serial and parallel implementations are discussed with a maximum reported speedup of approximately 55 obtained with the AMR code.

Original languageEnglish
Pages (from-to)48-52
Number of pages5
JournalComputers and Fluids
Volume175
DOIs
Publication statusPublished - 2018 Oct 15

Fingerprint

Flow fields
Graphics processing unit

All Science Journal Classification (ASJC) codes

  • Computer Science(all)
  • Engineering(all)

Cite this

@article{40bfaa315f60461c8526bb1605dd6ae2,
title = "Multi-block adaptive mesh refinement (AMR) for a lattice Boltzmann solver using GPUs",
abstract = "A parallelized adaptive mesh refinement (AMR) approach using lattice Boltzmann solver [1] using GPU acceleration is developed based around a multi-block structured uniform mesh code [2, 3]. AMR is obtained by deploying multiple levels of mesh blocks, with varying resolution employed within each block, for increased resolution of flows in regions requiring higher accuracy. A simple three dimensional benchmark is employed for confirming the accuracy and parallel performance of the AMR implementation. A series of benchmark tests has been performed with the flow fields and parallel performance compared. Comparisons of the computation time between serial and parallel implementations are discussed with a maximum reported speedup of approximately 55 obtained with the AMR code.",
author = "Hsu, {Fu Sheng} and Keh-Chin Chang and Matt-Hew Smith",
year = "2018",
month = "10",
day = "15",
doi = "10.1016/j.compfluid.2018.01.033",
language = "English",
volume = "175",
pages = "48--52",
journal = "Computers and Fluids",
issn = "0045-7930",
publisher = "Elsevier Limited",

}

Multi-block adaptive mesh refinement (AMR) for a lattice Boltzmann solver using GPUs. / Hsu, Fu Sheng; Chang, Keh-Chin; Smith, Matt-Hew.

In: Computers and Fluids, Vol. 175, 15.10.2018, p. 48-52.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Multi-block adaptive mesh refinement (AMR) for a lattice Boltzmann solver using GPUs

AU - Hsu, Fu Sheng

AU - Chang, Keh-Chin

AU - Smith, Matt-Hew

PY - 2018/10/15

Y1 - 2018/10/15

N2 - A parallelized adaptive mesh refinement (AMR) approach using lattice Boltzmann solver [1] using GPU acceleration is developed based around a multi-block structured uniform mesh code [2, 3]. AMR is obtained by deploying multiple levels of mesh blocks, with varying resolution employed within each block, for increased resolution of flows in regions requiring higher accuracy. A simple three dimensional benchmark is employed for confirming the accuracy and parallel performance of the AMR implementation. A series of benchmark tests has been performed with the flow fields and parallel performance compared. Comparisons of the computation time between serial and parallel implementations are discussed with a maximum reported speedup of approximately 55 obtained with the AMR code.

AB - A parallelized adaptive mesh refinement (AMR) approach using lattice Boltzmann solver [1] using GPU acceleration is developed based around a multi-block structured uniform mesh code [2, 3]. AMR is obtained by deploying multiple levels of mesh blocks, with varying resolution employed within each block, for increased resolution of flows in regions requiring higher accuracy. A simple three dimensional benchmark is employed for confirming the accuracy and parallel performance of the AMR implementation. A series of benchmark tests has been performed with the flow fields and parallel performance compared. Comparisons of the computation time between serial and parallel implementations are discussed with a maximum reported speedup of approximately 55 obtained with the AMR code.

UR - http://www.scopus.com/inward/record.url?scp=85053138008&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85053138008&partnerID=8YFLogxK

U2 - 10.1016/j.compfluid.2018.01.033

DO - 10.1016/j.compfluid.2018.01.033

M3 - Article

AN - SCOPUS:85053138008

VL - 175

SP - 48

EP - 52

JO - Computers and Fluids

JF - Computers and Fluids

SN - 0045-7930

ER -