EN105 : Computer architecture. Course overview J. CRENNE 2015/2016

Transcription

1 EN105 : Computer architecture Course overview J. CRENNE 2015/2016

2 Schedule Cours Cours Cours Cours Cours Cours Cours Cours Cours Cours 2

3 CM 1 - Warmup CM 2 - Computer architecture CM 3 - CISC2RISC CM 4 - Pipelining 1 CM 5 - Pipelining 2 CM 6 - Memory systems basics CM 7 - Cache microarchitecture CM 8 - Network topology CM 9 - VLIW processors CM 10 - Vector and MT processors 3

4 What is computer architecture?

5 The computer system stack Application Gap too large to bridge in one step (but there are exceptions, e.g., a magnetic compass) Technology In its broadest definition, computer engineering is the development of the abstraction/implementation layers that allow us to execute information processing applications efficiently using available manufacturing technologies 5

6 The computer system stack 6

7 The computer system stack Technology 6

8 The computer system stack Devices Technology 6

9 The computer system stack Circuits Devices Technology 6

10 The computer system stack Gate level Circuits Devices Technology 6

11 The computer system stack Register-transfer level Gate level Circuits Devices Technology 6

12 The computer system stack Microarchitecture Register-transfer level Gate level Circuits Devices Technology 6

13 The computer system stack Instruction set architecture Microarchitecture Register-transfer level Gate level Circuits Devices Technology 6

14 The computer system stack Operating system Instruction set architecture Microarchitecture Register-transfer level Gate level Circuits Devices Technology 6

15 The computer system stack Programming language Operating system Instruction set architecture Microarchitecture Register-transfer level Gate level Circuits Devices Technology 6

16 The computer system stack Algorithm Programming language Operating system Instruction set architecture Microarchitecture Register-transfer level Gate level Circuits Devices Technology 6

17 The computer system stack Application Algorithm Programming language Operating system Instruction set architecture Microarchitecture Register-transfer level Gate level Circuits Devices Technology 6

18 Computer architecture The computer system stack Application Operating system Instruction set architecture Microarchitecture Register-transfer level Gate level Circuits Devices Technology Sort an array of numbers 2, 6, 3, 8, 4, 5 -> 2, 3, 4, 5, 6, 8 Insertion sort algorithm 1. Find minimum number in input array Algorithm 2. Move minimum number into output array 3. Repeat steps 1 and 2 until finished Programming language C implementation of insertion sort void isort( int b[], int a[], int n ) { for ( int idx, k = 0; k < n; k++ ) { int min = 100; for ( int i = 0; i < n; i++ ) { if ( a[i] < min ) { min = a[i]; idx = i; } b[k] = min; a[idx] = 100; } } 7

19 The computer system stack MIPS32 instruction set Instructions that machine executes blez move li move move li lw addiu slt movn addiu movn $a2, $a7, $t4, $a4, $v1, $a3, $a5, $a4, $a6, $v0, $v1, $a3, done $zero 99 $a1 $zero 99 0($a4) $a4, 4 $a5, $a3 $v1, $a6, $v1, 1 $a3, $a5, $a6 Application Algorithm Programming language Operating system Instruction set architecture Microarchitecture Register-transfer level Gate level Circuits Devices Technology Mac OS X, Windows, Linux Handles low-level hardware management 8

20 The computer system stack Application Algorithm Programming language Operating system Instruction set architecture Microarchitecture Register-transfer level Gate level Circuits Devices Technology 9 How data flows through system Boolean logic gates and functions Combining devices to do useful work Transistors and wires Silicon process technology

21 Application requirements vs. technology constraints Application Algorithm Programming language Operating system Instruction set architecture Microarchitecture Register-transfer level Gate level Circuits Devices Technology Application requirements Suggest how to improve architecture Provide revenue to fund development Computer architects provide feedback to guide application and technology research directions Technology constraints Restrict what can be done efficiently New technologies make new arch possible 10

22 Logic, state and interconnect Digital systems are implemented with three basic building blocks Logic to process data State to store data Interconnect to move data Application Algorithm Programming language Operating system Instruction set architecture Microarchitecture Register-transfer level Gate level Circuits Devices Technology Logic State Logic State Logic State Logic State Logic State 11

23 Processors, memories, and networks Application Algorithm Computer architecture basic building blocks Processors for computation Memories for storage Networks for communication Programming language Operating system Instruction set architecture Microarchitecture Register-transfer level Gate level Circuits Devices Input data Processor Network Memory Output data Compute data Move data Store data Technology 12

24 Computer architecture artifacts 13

25 Trends in computer architecture

26 Application requirements vs. technology constraints Application Algorithm Programming language Operating system Instruction set architecture Microarchitecture Register-transfer level Gate level Circuits Devices Technology Traditional application requirements As much processor compute as possible As much memory capacity as possible As much network bandwidth as possible Traditional technology constraints Exponential scaling of resources 15

27 Exponential scaling for processor computation 16

28 Exponential scaling for memory capacity 17

29 Exponential scaling for network bandwidth 18

30 Key trends in the application requirements and technology constraints over the past decade have resulted in a radical rethinking of the processors, memories, and networks used in modern computing systems Three key trends in computer engineering 1. Growing diversity in application requirements motivate growing diversity in computing systems 2. Energy and power constraints motivate transition to multiple processors integrated onto a single chip 3. Technology scaling challenges motivate new emerging processor, memory, and net work device technologies 19

31 Trend 1: Bell s law Roughly every decade a new, lower priced computer class forms based on a new programming platform resulting in new usage and industries 20

32 Trend 1: Growing diversity in apps & systems 21

33 Trend 2: Energy/power constraints all modern systems Power Energy = = Second Energy Ops x Ops Second Power Chip packaging Chip cooling System noise Case temperature Data-center air conditioning Energy Battery life Electricity bill Mobile device weight 22

34 Trend 2: Power constraints single-processor scaling 23

35 Trend 2: Transition to multicore processors Intel Pentium 4 Single monolithic processor (2000) P N M 24

36 Trend 2: Transition to multicore processors AMD Quad-core Opteron Four cores on the same die (2003) P P P P N M M M M 25

37 Trend 2: Transition to multicore processors Cray XT3 supercomputer 1024 single-core processors (2004) P P P N M M M 26

38 Trend 2: Transition to multicore processors IBM Blue Gene Q supercomputer Thousands of 18-core processors (2011) P P P P N M M M M 27

39 Trend 2: Energy and performance of multicores 28

40 Trend 2: Multicore performance scaling 29

41 Trend 3: Emerging device technologies 30

42 Computer architecture design

43 What do computer architects actually do? General Science Discover truths about nature Ask question about nature Construct hypothesis Test with experiment Analyze results and draw conclusions Computer engineering Explore design space for a given system Design and build initial system Ask question about system Modify systems or build/design alternatives Test with experiment to compare alternatives Analyze results and draw conclusions 32

44 What do computer architects actually do? Computer engineering Explore design space for a given system Design and build initial system Ask question about system Modify systems or build/design alternatives Test with experiment to compare alternatives Analyze results and draw conclusions Fighter airplane: ~100, 000 parts Intel Sandy Bridge E: 2.27 Billion transistors 33

45 Design principles Abstraction - Levels of modeling Modularity - Separate design into components w/ well-defined interfaces Hierarchy - Recursively apply modularity principle Encapsulation - Hide implementation details from interfaces Regularity - Leverage structure Extensibility - Include mechanisms/hooks for future changes Design pattern Processors, memories, networks Control/datapath split Single-cycle, FSM, pipelines control Raw port, message, method interfaces Design methodologies Test-driven development Incremental development Automated development 34

46 Take-away points Computer architecture is the process of building computing systems to meet given application requirements within physical technology constraints We are entering an exciting new era of computer architecture with growing diversity in applications and systems, a remarkable industrial shift towards mainstream parallel processing, and significant technology scaling challenges This era offers tremendous challenges and opportunities, which makes it a wonderful time to study and contribute to the field of computer architecture 35

47 Thanks!