联想深腾1800高性能服务器

180021800ƒ学计发ƒ软调术ƒ计术ƒ应软务3ƒ1.论学ƒ2.实验学ƒ3.计学ƒ计学计学计动学计学计学计学计学计学计电学……4ƒ––ƒ––––5ƒ(级计)CRAY,银IƒSMP()IBMp690SUNƒMPP(规)T3E1000HPƒccNUMA(储)SGIQriginƒ统联腾1800,IBMSP267CPU49511061200600MHzMIPS/R16000SGI711107621001050MHzUltraSparcSun782192945002250MHzAthlon2800+AMD1019194525001250MHzAlphaHP1204244456002.8GHzP4XeonIntel1219no36001.8GHzOpteronAMD1285no64003.2GHzP4XeonIntel1576388468001.7GHzPower4IBM2119530360001.5GHzItanium2IntelSpecfp2000PeakTPPn=1000Peak(Mflops)ClockCPUManufacturer8BeijingSocialInsuranceChina/2001245.30384.00Hewlett-PackardSuperDome750MHz/HyperPlex/128486Exploration&DevelopmentRes.Inst.ofDaqingOilFieldChina/2003385.401228.00LegendGroupDeepComp1800-P4Xeon2.4GHz-Giganet/256299SinopecChina/2003402.501228.80IBMxSeriesClusterXeon2.4GHz-Gig-E/256293ShengriLiChina/2003402.501228.80IBMxSeriesClusterXeon2.4GHz-Gig-E/256292YoungwooChina/2003408.20672.00Hewlett-PackardSuperDome875MHz/HyperPlex/192278LASG/InstituteofAtmosphericPhysics,CASChina/2002711.701228.00LegendGroupDeepComp1800-P4Xeon2.4GHz-Myrinet/25698AcademyofMathematicsandSystemScienceChina/20021046.002048.00LegendGroupDeepComp1800-P4Xeon2GHz-Myrinet/51251InstallationSiteCountry/YearRmaxRpeakManufacturerComputer/ProcsRankƒ学计发ƒ软调术ƒ计术ƒ应软务11ƒƒƒƒƒƒƒHotSpotBottleNeck12ƒIOƒ1314SystemLevelSystemLevelApplicationLevelApplicationLevelMicro-ArchitectureLevelMicro-ArchitectureLevel15MicroMicro--ArchitectureLevelArchitectureLevelApplicationLevelApplicationLevelSystemLevelSystemLevelPerfmonPerfmon**ApiMonApiMon**Quantify*Quantify*VTuneVTunePerformancePerformanceAnalyzerAnalyzerNetMonNetMon**LessIntrusiveLessIntrusiveMoreIntrusive16ƒƒƒƒƒƒƒ17Æfor(i=0;iNUM;i++){for(j=0;jNUM;j++){for(k=0;kNUM;k++){c[i][j]=c[i][j]+a[i][k]*b[k][j];}}}for(i=0;iNUM;i++){for(k=0;kNUM;k++){for(j=0;jNUM;j++){c[i][j]=c[i][j]+a[i][k]*b[k][j];}}}10.996sIMSL0.738s0.848s0.762s3.328s5.457s5.445s16.676sMKL60MKL50ATLASCXMLOPT4RELEASESELF181800ƒ学计发ƒ软调术ƒ计术ƒ应软务19ƒPRAMBSPlogPC3ƒSIMDMIMDSPMDMPMD20ƒƒ21CPUMemoryCPUMemoryNetworkCPUMemoryCPUMemoryBusMemoryCPUCPUCPUCPUƒƒƒƒƒOpenMP22ƒ–ƒ–ƒ组–ƒ–23TheProblemInitialtasksCommunicationCombinedTasksFinalProgramƒ–ƒ–ƒ组–ƒ–24AtmosphereModelOceanModelLandSurfaceModelHydrologyModelƒ–ƒ–2526ƒƒƒ27ƒƒƒƒ-ƒ28NNNNNN29N2N4N2N4N4N4N1N2N2N130P2P1P3T1T2T1T1T2T2T3T4T3TN-3TN-1TN-2TN-2TNTN-1TN-1TNTN31-MasterSlaveSlaveSlave32N1N3N233ƒƒ34ƒ数–HPFƒ储–OpenMPPthreadsƒ传递–MPIPVM35ƒƒNavier-Stokesƒƒƒƒ500036NY(i,j)(i,j+1)(i,j-1)(i+1,j)(i-1,j)37(i,j)(i,j+1)(i,j-1)(i+1,j)(i-1,j)38P2P2NYP1P1NYP3P3NYP4P4NY391.9722.5215.162.92816(8节点)7.935.683.9911.124---44.41拟时间/处数401800ƒ学计发ƒ软调术ƒ计术ƒ应软务4118001800¾¾¾¾¾42¾DL_POLY()¾80,000Fortran¾AIX/Power4~Linux/IntelXeon¾DOCK()¾100,000C¾IRIX/SGI~Linux/IntelXeon¾Ccsm()¾170,000Fortran¾AIX/Power4~Linux/IntelXeon43ƒMOLDY调3ƒDL_POLY调2.25ƒ软调调6ƒ庆Geodepth调单节点时络达126MB/sƒ……44ƒIntel/PGI/LF95Compiler……ƒMKLsvmlGoto(MathLibrary)ƒIPP(PerformanceLibrary)ƒTotalView(ParallelDebug)ƒVTune(PerformanceAnalysis)ƒVGV(ClusterPerformanceAnalysis)ƒPerfmon;Apimon;Netmon……4546