Unformatted text preview:

Computer Science and EngineeringAICCSA-06Hesham El-Rewini, March 9, 2006Parallelism Challenges: Parallelism Challenges: Rewind Rewind and and Fast ForwardFast ForwardHesham ElHesham El--Rewini, Ph.D., P.E.Rewini, Ph.D., P.E.Department of Computer Science Department of Computer Science and Engineeringand [email protected]/~rewiniComputer Science and EngineeringAICCSA-06Hesham El-Rewini, March 9, 2006Talk Outline1.1.Why Parallel Computing? Major HighlightsWhy Parallel Computing? Major Highlights2.2.Ups and Downs of Parallel ComputingUps and Downs of Parallel Computing3.3.Recent Trends and Challenges for the futureRecent Trends and Challenges for the futureComputer Science and EngineeringAICCSA-06Hesham El-Rewini, March 9, 2006Let’s Start with a Couple of Interesting Pictures!!Computer Science and EngineeringAICCSA-06Hesham El-Rewini, March 9, 2006Picture 1 --- Home Computer as imagined 50 years ago.In 1954, scientists from the RAND Corporation have created this model (shown in the next slide) to illustrate how a home computer could look like in the year 2004. However, the needed technology will not be economically feasible for the average home. Also the scientists readily admit that the computer will require not yet invented technology to actually work, but 50 years from now scientific progress is expected to solve these problems. With the teletype interface and the Fortran language, the computer will be easy to use.From 1954 popular mechanics magazineComputer Science and EngineeringAICCSA-06Hesham El-Rewini, March 9, 2006Computer Science and EngineeringAICCSA-06Hesham El-Rewini, March 9, 2006Picture 2 --- Mobile Computer in the Microsoft EraTech Support CTR in the middle of nowhere!Tech Support CTR in the middle of nowhere!Computer Science and EngineeringAICCSA-06Hesham El-Rewini, March 9, 2006Computer Science and EngineeringAICCSA-06Hesham El-Rewini, March 9, 2006Picture 3 --- Future Personal ComputerWonder Computer TechnologyWonder Computer TechnologyComputer Science and EngineeringAICCSA-06Hesham El-Rewini, March 9, 2006Computer Science and EngineeringAICCSA-06Hesham El-Rewini, March 9, 2006Computer Science and EngineeringAICCSA-06Hesham El-Rewini, March 9, 2006Computer Science and EngineeringAICCSA-06Hesham El-Rewini, March 9, 2006Computer Science and EngineeringAICCSA-06Hesham El-Rewini, March 9, 2006Computer Science and EngineeringAICCSA-06Hesham El-Rewini, March 9, 2006Computer Science and EngineeringAICCSA-06Hesham El-Rewini, March 9, 20061- Why Parallel Computing?Major HighlightsComputer Science and EngineeringAICCSA-06Hesham El-Rewini, March 9, 2006EveryoneDepartments/individualsDivisional IS shopsCorporate computer centersOwnersInternet/WirelessLANPeripheral cableNoneConnectivityOrchestrateLayoutEditProcessOperationAsk and TellSee and pointKeyboard, CRTPunched cardsInterfaceCommunicatePresentAccessCalculateObjectiveMultimediaFont, GraphsText, numbersAlphanumericDataGroupsIndividualsSpecialistsExpertsUsersMobileDesktopTerminal roomComputer roomLocation1990s/2000s1980s1970s1960sTime PeriodShared NetworkDesktopTime SharingBatchFeatureFour Eras of ComputingComputer Science and EngineeringAICCSA-06Hesham El-Rewini, March 9, 2006Moore’s Law1965 prediction by Intel founder Gordon Moore:The number of transistors that can be built on the same size piece of silicon will double every 18 monthsComputer Science and EngineeringAICCSA-06Hesham El-Rewini, March 9, 2006Computer Science and EngineeringAICCSA-06Hesham El-Rewini, March 9, 2006Processor Evolution• Gate delay reduces by 1/ (frequency up by )• Number of transistors in a constant area goes up by 2• Additional transistors enable an additional increase in performance• Result: 2x performance at roughly equal costGeneration NGeneration N +1GenerationN2 22Computer Science and EngineeringAICCSA-06Hesham El-Rewini, March 9, 2006Can this Growth be sustained forever?¾ Speed of Light Argument(Most people)¾ The Vanishing Electrons Argument(Joel Birnbaum, HP Labs, ACM 97)¾ The FM Radio Analogy(Erik P. DeBenedictis, Sandia National Labs, 2005)Computer Science and EngineeringAICCSA-06Hesham El-Rewini, March 9, 2006Speed of Light LimitLight travels 1 cm in301nanosecondWhat is the speed if a signal must travel 1 cmduring the execution of an instruction?Computer Science and EngineeringAICCSA-06Hesham El-Rewini, March 9, 2006Joel Birnbaum, HP Labs, 1997Computer Science and EngineeringAICCSA-06Hesham El-Rewini, March 9, 2006FM Radio and End of Moore’s Law(Erik P. DeBenedictis, Sandia National Lab, 5/16/05)Driving away from FM transmitterÆless signalNoise from electrons Æ no changeIncreasing numbers of gatesÆless signal powerNoise from electrons Æ no changeShrinkDistanceComputer Science and EngineeringAICCSA-06Hesham El-Rewini, March 9, 2006Grand Challenges need Computing PowerVision, Human GenomeVision, Human GenomeClimate Model, Ocean CirculationClimate Model, Ocean CirculationFluid Turbulence, Viscous FlowFluid Turbulence, Viscous FlowQuantum ChromodynamicsQuantum ChromodynamicsSuperconductor ModelSuperconductor ModelVehicle DynamicsVehicle DynamicsWhether PredictionWhether PredictionChemical DynamicsChemical Dynamics3D plasma3D plasmaOil Reservoir ModelOil Reservoir ModelParallelism is an obvious answer!!Computer Science and EngineeringAICCSA-06Hesham El-Rewini, March 9, 2006ParallelismMultiple processors cooperate to jointly execute a single computational task in order to speed up its execution.¾ Solve Problems Faster (Speedup)¾ Solve More Problems (Higher Throughput)¾ Solve Larger Problems (Computational Power)¾ Enhance Solutions’ Quality (Quality Up)Computer Science and EngineeringAICCSA-06Hesham El-Rewini, March 9, 2006Types of ParallelismMIMDMultiprocessorsMulticomputersMISDMultipleInstruction StreamSIMDArray ProcessorsVectorSISDUniprocessorsSingleInstructionStreamMultiple Data StreamSingle Data StreamFlynn’s TaxonomyComputer Science and EngineeringAICCSA-06Hesham El-Rewini, March 9, 2006MIMD CategoriesDMMPDistributed MemoryMulticomputersDMSVDistributed Shared MemoryDistributedMemoryGMMPGMSVShared Memory MultiprocessorsGlobalMemoryMessage PassingShared VariablesJohnson’s ExpansionComputer Science and EngineeringAICCSA-06Hesham El-Rewini, March 9, 2006Main Parallel ArchitectureComputer Science and EngineeringAICCSA-06Hesham El-Rewini, March 9, 2006SIMD SystemsProcessorMemoryPMPMPMPMPMPMPMPMPMPMPMPMPMPMPMPMvon Neumann ComputerSome Interconnection


View Full Document

SMU CSE 8383 - Parallelism Challenges

Download Parallelism Challenges
Our administrator received your request to download this document. We will send you the file to your email shortly.
Loading Unlocking...
Login

Join to view Parallelism Challenges and access 3M+ class-specific study document.

or
We will never post anything without your permission.
Don't have an account?
Sign Up

Join to view Parallelism Challenges 2 2 and access 3M+ class-specific study document.

or

By creating an account you agree to our Privacy Policy and Terms Of Use

Already a member?