高级计算机体系结构设计及其在数据中心和云计算的应用Schedule·LecturetimeMon.(Dec.9/16)14:00-15:30,15:40-17:10Tue.(Dec.10/17)8:30-10:00,10:10-11:40,14:00-15:30- Wed.(Dec.11/18)16:00-17:30 Thu.(Dec.12/19) 8:30-10:00, 10:10-11:40— Fri.(Dec. 13/20) 8:30-10:00, 10:10-11:40·Duedate (TBA)- Lab & report (within two weeks from class ending date)-Homeworkassignment (withintwoweeksfromclassendingdate)
高级计算机体系结构设计及其在数据中心和云计算的应 用 Schedule •Lecture time – Mon. (Dec. 9/16) 14:00-15:30, 15:40-17:10 – Tue.(Dec. 10/17) 8:30-10:00, 10:10-11:40, 14:00-15:30 – Wed.(Dec.11/18) 16:00-17:30 – Thu.(Dec. 12/19) 8:30-10:00, 10:10-11:40 – Fri.(Dec. 13/20) 8:30-10:00, 10:10-11:40 •Due date (TBA) – Lab & report (within two weeks from class ending date) – Homework assignment (within two weeks from class ending date)
高级计算机体系结构设计及其在数据中心和云计算的应用Course TopicsIntro/FundamentalsMemory Hierarchy·PipeliningInstruction-LevelParallelismOut-of-Order and Speculative Execution CoreMulti-[socket(SMP,DSM) /thread(SMT,CMT)/core(CMP)Throughput-Oriented Architecture and GPUsData Center as a ComputerSustainable Computing for Future Clouds & Big DataWill devote most timeto itemsinbold
高级计算机体系结构设计及其在数据中心和云计算的应 用 Course Topics • Intro/Fundamentals • Memory Hierarchy • Pipelining • Instruction-Level Parallelism • Out-of-Order and Speculative Execution Core • Multi-[socket(SMP,DSM)|thread(SMT,CMT)|core(CMP)] • Throughput-Oriented Architecture and GPUs • Data Center as a Computer • Sustainable Computing for Future Clouds & Big Data Will devote most time to items in bold
高级计算机体系结构设计及其在数据中心和云计算的应用QuantitativeDesignand AnalysisUnderstanding and Measuring PerformanceMemory LocalityPower and EnergyParallelism and Critical PathsInstructionSetArchitectureBasic Processor Organization
高级计算机体系结构设计及其在数据中心和云计算的应 用 Quantitative Design and Analysis • Understanding and Measuring Performance • Memory Locality • Power and Energy • Parallelism and Critical Paths • Instruction Set Architecture • Basic Processor Organization
高级计算机体系结构设计及其在数据中心和云计算的应用Amdahl's LawSpeedup = timewithout enhancement / timewith enhancementAnenhancement speedsupfractionfof ataskbyfactorStimenew = timeorig:( (1-f) + f/S )SSoverall = 1 / ( (1-f) + f/S )timeorigf(1 - f)timenew(1 - f)f/Sf/S
高级计算机体系结构设计及其在数据中心和云计算的应 用 1 timeorig (1 - f) f timeorig (1 - f) f timeorig Amdahl’s Law Speedup = timewithout enhancement / timewith enhancement An enhancement speeds up fraction f of a task by factor S timenew = timeorig·( (1-f) + f/S ) Soverall = 1 / ( (1-f) + f/S ) (1 - f) timenew (1 - f) f/S timenew f/S
高级计算机体系结构设计及其在数据中心和云计算的应用The Iron Law of Processor PerformanceTimeTimeInstructionsCyclesXXCycleProgramProgramInstructionTotalWorkCPIorI/IPCI/f (frequency)InProgramAlgorithms,Microarchitecture,Compilers,MicroarchitectureProcessTechISA ExtensionsArchitects target CPl,but mustunderstand theothers
高级计算机体系结构设计及其在数据中心和云计算的应 用 The Iron Law of Processor Performance Cycle Time Instruction Cycles Program Instructions Program Time Architects target CPI, but must understand the others Total Work In Program CPI or 1/IPC 1/f (frequency) Algorithms, Compilers, ISA Extensions Microarchitecture Microarchitecture, Process Tech