简介:现在,内部任务的干扰是在分析multicores的预定行为的主要困难。预定可预言的嵌入的multicore体系结构MERASA,它允许安全最坏的执行时间(WCET)评价,作为一个吸引人的解决方案出现了。在建筑学,WCET能被能被干扰知道的公共汽车仲裁人(IABA)和划分例如columnization或bankization的动态缓存围住的上面的界限延期(UBD)估计。然而,这体系结构面对在减少的UBD和有效分享的缓存利用之间的窘境。获得更紧密的WCET评价,我们建议由优化与IABA和二水平的划分缓存在multicore系统上印射的bank-to-core减少UBD的一条新奇途径。为这,我们首先基于内部任务的干扰延期的分析介绍一个新UBD计算模型,然后提出bank-to-core与最小的UBD印射和优化算法的核心顺序优化方法。试验性的结果证明我们的途径能从4%~37%减少WCET。
简介:包分类被学习十年了;它基于一个给定的规则集合分类包进特定的流动。当定义软件的网络被建议,包分类的一个最近的趋势是放大五元组的模型到多元组。一般来说,多重地上的包分类是一个复杂问题。尽管大多数存在softwarebased算法在实践被证明非凡,他们对经典五元组的模型仅仅合适并且对困难被扩大规模。同时,硬件特定的答案不可弯曲、昂贵,并且他们中的一些是消费的力量。在这份报纸,我们为多核心系统建议一条通用的多维的包分类途径。在我们的途径,新奇数据结构和四个基于分解的算法被设计优化分类并且规则更新。为多地规则,一个规则集合根据领域的数字被切成几部分。每部分独立地工作。这样,这些地在平行被寻找,所有部分结果最后一起被合并。表明我们的途径的可行性,我们实现一个原型并且评估它的产量和潜伏。试验性的结果证明我们的途径比另外的分解底的算法和43%更低的潜伏的完成40%更高的产量平均比另外的算法的统治增长更改。而且,我们的途径平均节省39%记忆消费并且有好可伸缩性。
简介:Thispaperpresentsacoupledmagnetic-circuitmethodforcomputingthemagneticforceofair-corereactorundershort-timecurrent.Thecurrentandthemagneticfluxdensityarecomputedfirstandthenthemagneticforceisobtained.Thus,thedynamicstabilityperformanceofair-corereactorcanbeanalyzedatthedesignstagetoreduceexperimentalcostandshortenthelead-timeofproductdevelopment.
简介:Inthispaper,weinvestigatetheproblemofasize-constrainedk-coregroupquery(SCCGQ)insocialnetworks,takingbothuserclosenessandnetworktopologyintoconsideration.Morespecifically,SCCGQintendstofindagroupofhusersthathasthehighestsocialclosenesswhilebeingak-core.SCCGQcanbewidelyappliedtoeventplanning,taskassignment,socialanalysis,andmanyotherfields.Incontrasttoexistingworkonthek-coredetectionproblem,whichaimstofindak-coreinasocialnetwork,SCCGQnotonlyfocusesonk-coredetectionbutalsotakessizeconstraintsintoconsideration.Althoughtheconventionalk-coredetectionproblemcanbesolvedinlineartime,SCCGQhasahighercomplexity.TosolvetheproblemofSCCGQ,weproposeaBlastScatter(BS)algorithm,whichappointsthequerynodeasthecentertobeginoutwardexpansionsviabreadthsearch.Ineachoutwardexpansion,BSfindsanewcenterthroughagreedystrategyandthenselectsmultipleneighborsofthecenter.TospeeduptheBSalgorithm,weproposeanadvancedsearchalgorithm,calledBoundedExtension(BE).Specifically,BEcombinesaneffectivesocialdistancepruningstrategyandatightupperboundofsocialclosenesstoprunethesearchspaceconsiderably.Inaddition,weproposeanoffiinesocial-awareindextoacceleratethequeryprocessing.Finally,ourexperimentalresultsdemonstratetheefficiencyandeffectivenessofourproposedalgorithmsonlargereal-worldsocialnetworks.
简介:Withtheadventofthebigdataera,theamountsofsamplingdataandthedimensionsofdatafeaturesarerapidlygrowing.Itishighlydesiredtoenablefastandefficientclusteringofunlabeledsamplesbasedonfeaturesimilarities.Asafundamentalprimitivefordataclustering,thek-meansoperationisreceivingincreasinglymoreattentionstoday.Toachievehighperformancek-meanscomputationsonmodernmulti-core/many-coresystems,weproposeamatrix-basedfusedframeworkthatcanachievehighperformancebyconductingcomputationsonadistancematrixandatthesametimecanimprovethememoryreusethroughthefusionofthedistance-matrixcomputationandthenearestcentroidsreduction.Weimplementandoptimizetheparallelk-meansalgorithmontheSW26010many-coreprocessor,whichisthemajorhorsepowerofSunwayTaihuLight.Inparticular,wedesignataskmappingstrategyforload-balancedtaskdistribution,adatasharingschemetoreducethememoryfootprintandaregisterblockingstrategytoincreasethedatalocality.Optimizationtechniquessuchasinstructionreorderinganddoublebufferingarefurtherappliedtoimprovethesustainedperformance.Discussionsonblock-sizetuningandperformancemodelingarealsopresented.Weshowbyexperimentsonbothrandomlygeneratedandreal-worlddatasetsthatourparallelimplementationofk-meansonSW26010cansustainadouble-precisionperformanceofover348.1Gflops,whichis46.9%ofthepeakperformanceand84%ofthetheoreticalperformanceupperboundonasinglecoregroup,andcanachieveanearlyidealscalabilitytothewholeSW26010processoroffourcoregroups.Performancecomparisonswiththepreviousstate-of-the-artonbothCPUandGPUarealsoprovidedtoshowthesuperiorityofouroptimizedk-meanskernel.