资源描述：
《A Lattice-Boltzmann solver for 3D fluid simulation on GPU.pdf》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。
1、SimulationModellingPracticeandTheory25(2012)163–171ContentslistsavailableatSciVerseScienceDirectSimulationModellingPracticeandTheoryjournalhomepage:www.elsevier.com/locate/simpatALattice-Boltzmannsolverfor3DfluidsimulationonGPUa,⇑baaP.R.Rinaldi,E.A.Dari,M.J.Vénere,A.C
2、lausseaCNEA-CONICETandUniversidadNacionaldelCentro,7000Tandil,ArgentinabCONICET-CNEAandInstitutoBalseiro,8400Bariloche,ArgentinaarticleinfoabstractArticlehistory:Athree-dimensionalLattice-BoltzmannfluidmodelwithnineteendiscretevelocitieswasReceived24October2011impleme
3、ntedusingNVIDIAGraphicProcessingUnit(GPU)programinglanguage‘‘ComputeReceivedinrevisedform26February2012UnifiedDeviceArchitecture’’(CUDA).PreviousLBMGPUimplementationsrequiredtwoAccepted13March2012stepstomaximizememorybandwidthduetomemoryaccessrestrictionsofearlierAvai
4、lableonline9April2012versionsofCUDAtoolkitandhardwarecapabilities.Inthiswork,anewapproachbasedonsingle-stepalgorithmwithareversedcollision–propagationschemeisdevelopedtoKeywords:maximizeGPUmemorybandwidth,takingadvantageofthenewerversionsofCUDApro-GPGPUgrammingmodela
5、ndnewerNVIDIAGraphicCards.Thecodewastestedonthenumerical3DLattice-BoltzmannMethodsCUDAcalculationofliddrivencubiccavityflowatReynoldsnumber100and1000showinggreatprecisionandstability.SimulationsrunningonlowcostGPUcardscancalculate400cellupdatespersecondwithmorethan65%
6、hardwarebandwidth.Ó2012ElsevierB.V.Allrightsreserved.1.IntroductionTheLattice-BoltzmannMethod(LBM)isaclassofcellularautomata(CA)thatapproximatestheNavier–Stokesequationstosecondorderwithanexplicitcollision–advectionscheme[4].DerivedfromtheLatticeGasAutomata[2],discre
7、teveloc-itiesactasarcsbetweenlatticecellswhosepopulationsarestatevariables.LikeanyexplicitCAschemethesamesimplecodeisrunovertheentiregridineverytimestep,makingLBMespeciallysuitableforparallelimplementations[5,15,18].AninterestingeconomicalternativeforparallelizingLBM
8、aretheGraphicProcessingUnits(GPUs)[6,16,20].AGPUisthechipusedbygraphiccardstorenderpixelsonthescreen.ModernGPUsareoptimizedforexecu
