资源描述：
《Volume render on GPU CPU and other》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、IEEETRANSACTIONSONVISUALIZATIONANDCOMPUTERGRAPHICS,VOL.15,NO.6,NOVEMBER/DECEMBER20091563MappingHigh-FidelityVolumeRenderingforMedicalImagingtoCPU,GPUandMany-CoreArchitecturesMikhailSmelyanskiy,DavidHolmes,JatinChhugani,AlanLarson,DouglasM.Carmean,DennisHanson,
2、PradeepDubey,KurtAugustine,DaehyunKim,AlanKyker,VictorW.Lee,AnthonyD.Nguyen,LarrySeiler,andRichardRobbAbstract—Medicalvolumetricimagingrequireshighfidelity,highperformancerenderingalgorithms.Wemotivateandanalyzenewvolumetricrenderingalgorithmsthataresuitedtomod
3、ernparallelprocessingarchitectures.First,wedescribethethreemajorcategoriesofvolumerenderingalgorithmsandconfirmthroughanimagingscientist-guidedevaluationthatray-castingisthemostacceptable.Wedescribeathread-anddata-parallelimplementationofray-castingthatmakesita
4、menabletokeyarchitecturaltrendsRofthreemoderncommodityparallelarchitectures:multi-core,GPU,andanupcomingmany-coreIntelarchitecturecode-namedLarrabee.Weachievemorethananorderofmagnitudeperformanceimprovementonanumberoflarge3Dmedicaldatasets.Wefurtherdescribead
5、atacompressionschemethatsignificantlyreducesdata-transferoverhead.ThisallowsourapproachtoscalewelltolargenumbersofLarrabeecores.IndexTerms—VolumeCompositing,ParallelProcessing,Many-coreComputing,MedicalImaging,GraphicsArchitecture,GPGPU.1INTRODUCTIONThepasttwod
6、ecadeshaveseenunprecedentedgrowthintheamountray-castingtobethemostacceptableofthreevolumerenderingandcomplexityofdigitalmedicalimagedatacollectedonpatientsintechniquesforhigh-fidelityrequirementsofmedicalimaging.standardmedicalpractice.Theclinicalneedtoaccurate
7、lydiagnose•Wemap,evaluateandcompareperformanceoftworay-castingdiseaseanddeveloptreatmentstrategiesinaminimally-invasiveman-implementationsonthreemodernparallelarchitectures.Weop-nerhasrequireddevelopingnewimageacquisitionmethods,highres-timizeourimplementation
8、totakefulladvantageofeacharchi-olutionacquisitionhardware,andnovelimagingmodalities.Alloftecture’ssalienthardwarefeatures.theseplacecomputationalburdensontheabilitytosynergisticall
