期刊网_中国期刊网

Scan Cell Positioning for Boosting the Compression of Fan-Out Networks

简介：Ensuringahighmanufacturingtestqualityofanintegratedelectroniccircuitmandatestheapplicationofalargevolumetestset.Evenifthetestdatacanbefitintothememoryofanexternaltester,theconsequentincreaseintestapplicationtimereflectsintoelevatedproductioncosts.Testdatacompressionsolutionshavebeenproposedtoaddressthetesttimeanddatavolumeproblembystoringanddeliveringthetestdatainacompressedformat,andsubsequentlybyexpandingthedataon-chip.Inthispaper,weproposeascancellpositioningmethodologythataccompaniesacompressiontechniqueinordertoboostthecompressionratio,andsquashthetestdataevenfurther.Whilewepresenttheapplicationoftheproposedapproachinconjunctionwiththefan-outbaseddecompressionarchitecture,thisapproachcanbeextendedforapplicationalongwithothercompressionsolutionsaswell.Theexperimentalresultsalsoconfirmthecompressionenhancementoftheproposedmethodology.
标签：测试数据压缩扫描单元输出网络细胞定位外部存储器测试应用

全文阅读

简介：Themulticoreevolutionhasstimulatedrenewedinterestsinscalingupapplicationsonshared-memorymultiprocessors,significantlyimprovingthescalabilityofmanyapplications.Butthescalabilityislimitedwithinasinglenode;thereforeprogrammersstillhavetoredesignapplicationstoscaleoutovermultiplenodes.Thispaperrevisitsthedesignandimplementationofdistributedsharedmemory(DSM)asawaytoscaleoutapplicationsoptimizedfornon-uniformmemoryaccess(NUMA)architectureoverawell-connectedcluster.ThispaperpresentsMAGI,anefficientDSMsystemthatprovidesatransparentsharedaddressspacewithscalableperformanceonaclusterwithfastnetworkinterfaces.MAGIisuniqueinthatitpresentsaNUMAabstractiontofullyharnessthemulticoreresourcesineachnodethroughhierarchicalsynchronizationandmemorymanagement.MAGIalsoexploitsthememoryaccesspatternsofbig-dataapplicationsandleveragesasetofoptimizationsforremotedirectmemoryaccess(RDMA)toreducethenumberofpagefaultsandthecostofthecoherenceprotocol.MAGIhasbeenimplementedasauser-spacelibrarywithpthread-compatibleinterfacesandcanrunexistingmultithreadedapplicationswithminimizedmodifications.WedeployedMAGIoveran8-nodeRDMA-enabledcluster.ExperimentalevaluationshowsthatMAGIachievesupto9.25:4speedupcomparedwithanunoptimizedimplementation,leadingtoasealableperformanceforlarge-scaledata-intensiveapplications.
标签： distributed shared MEMORY (DSM) SCALABILITY multieore

全文阅读

单看已out，能玩才叫in——TCL炫舞B68接口全接触

全文阅读