CS Computer Architecture Spring Lecture 01: Introduction

Transcription

1 CS Computer Architecture Spring 2008 Lecture 01: Introduction Created by Shannon Steinfadt Indicates slide was adapted from :Kevin Schaffer*, Mary Jane Irwinº, and from Computer Organization and Design (3rd Ed.), Patterson & Hennessy, 2005, UCB 1

2 Syllabus and Web Page Home page for the course: ex.html The Syllabus is located there at abus.html 2

3 What You Should Know Create, compile, and run C (C++, Java) programs Create, organize, and edit files and run programs on Unix/Linux 3

4 Classes of Computing Applications Different applications - different requirements Combine performance with stringent limitations on cost or power Desktop Computers Emphasize good performance for a single user Servers Modern form of mainframes, minicomputers and supercomputers Large workloads Greater emphasis on reliability Widest range in cost and performance 4

5 Classes of Computing Applications (con t) Embedded Computers Largest class of computers Widest range of applications and performance Usually have unique application requirements - Combine a minimum performance with stringent limitations on cost or power Despite low cost, least tolerance for failure In consumer-oriented applications - Do 1 function as perfectly as possible Which class (among 16- and 32-bit processors) is the fastest growing? 5

6 Where is the Market? Millions of Computers Embedded Desktop Servers 6

7 System Software * Computers can only perform simple tasks on their own; in order to accomplish any real work several layers are software are required Application software is the highest layer, this is the software that the user interacts with System software sits between application software and the hardware and provides support to the application software Fig. 1.3: simplified hierarchical abstraction 7

8 Below Your Program * An operating system is the interface to hardware resource and provides a number of services Input and output Memory management, file system Support for multiple programs running at once A compiler translates high-level language code into machine instructions 8

9 Bits and Bytes * Computer hardware operates using just two signals: on and off (or true and false) We represent these signals using binary numbers A single binary digit, or bit, can only be either a 0 or a 1, so multiple bits are needed to encode more complex information A sequence of 8 bits is called a byte (or octet) 32 bits are typically referred to as a word 9

10 Instructions * Binary numbers are also used to encode the instructions which the computer understands is an instruction that tells one computer to add two numbers The instructions of a computer, encoded in binary, make up the machine language of that computer The earliest programmers had no choice but to write software directly in machine language 10

11 Assembly Language * It didn't take long for those early programmers to find a way to avoid writing programs in machine language Assembly language replaces the binary numbers of machine language with symbols that are easier to understand Example: add A, B The computer only understands machine language; a program called an assembler is needed to translate assembly code to machine language 11

12 High-Level Languages * Assembly language statements still map directly to machine instructions High-level programming languages let you write code in a more natural way using algebraic notation, structured programming, etc. This means that a statement in a high-level language may map to several machine instructions A compiler translates the code into machine instructions 12

13 High-Level Languages * Advantages Easier to understand Fewer statements to accomplish a task Machine independence Disadvantage Some machine features can only be accessed through assembly/machine language 13

14 C Program Translation Figure 1.4: C program compiled into assembly language and then assembled into binary machine language 14

15 Computer Organization * Processor Control unit Datapath Memory system Main memory Cache Input/output (I/O) devices 15

16 I/O Devices * Keyboards Mice Video displays Sound cards Network interface cards Modems 17

17 Processor * The processor is where most of the work inside a computer is done Also known as a central processor unit (CPU) It is an integrated circuit (IC) or chip attached to the computer's motherboard Inside the processor, the control coordinates the processor's activities while the datapath follows the commands of the control unit 18

18 Memory * A computer's main memory is made up of dynamic random access memory (DRAM) DRAM chips are grouped together into modules called DIMMs (dual inline memory modules) DIMMs attach to the motherboard 19

19 Cache * Cache is a special type of fast memory that is used to hold data that the processor is currently using Cache is made from static random access memory (SRAM) Cache is typically connected directly to the motherboard or integrated into the processor 20

20 Instruction Set Architecture (ISA) ISA: An abstract interface between the hardware and the lowest level software of a machine Encompasses all the information necessary to write a machine language program that will run correctly, including instructions, registers, memory access, I/O, and so on. Enables implementations of varying cost and performance to run identical software - Implementation: hardware that obeys the architecture abstraction ABI (application binary interface): The basic instruction set and the operating system interface combined Provided for application programmers Defines a standard for binary portability across computers. 22

21 By the architecture of a system, I mean the complete and detailed specification of the user interface. As Blaauw has said, Where architecture tells what happens, implementation tells how it is made to happen. The Mythical Man-Month, Brooks, pg 45 23

22 ISA Type Sales Other Millions of Processor SPARC Hitachi SH PowerPC Motorola 68K MIPS IA-32 ARM PowerPoint comic bar chart with approximate values (see Figure 1.2 in text for correct values) 24

23 Storage Volatile storage like DRAM and SRAM only retain data while receiving power Nonvolatile storage like magnetic disks maintain data even without power (Figure 1.11) Optical disks (CDs, DVDs) are a form of removable, nonvolatile storage FLASH-based (USB/Jump drives) are another form of removable, nonvolatile storage Typically a computer has primary (or main) memory which is volatile and secondary memory which is nonvolatile 25

24 Communicating with Other Computers: Computer Networks Major advantages to being networked: Communication at high speeds Resource sharing (such as I/O devices) Nonlocal access allows remote use of machine 26

25 Integrated Circuits * ICs are built from electrically controlled switches called transistors Transistors are an on/off switch controlled by electricity The feature size of an IC tells us how small the transistors are and so a smaller feature size means more transistors in the same area Improvement in processor performance is due in large part to shrinking transistor sizes VLSI, very large scale integrated circuits - millions of transistors or better 27

26 Moore s Law In 1965, Gordon Moore predicted that the number of transistors that can be integrated on a die would double every 18 to 24 months (i.e., grow exponentially with time). Amazingly visionary million transistor/chip barrier was crossed in the 1980 s transistors, 1 MHz clock (Intel 4004) Million transistors (Ultra Sparc III) 42 Million transistors, 2 GHz clock (Intel Xeon) Million transistors, 3 GHz, 130nm technology, 250mm 2 die (Intel Pentium 4) Million transistor (HP PA-8500) 1.72 billion transistors, 1.6GHz, 90nm tech. 596 mm² die (Dual-Core Intel Itanium 2) 28

27 Processor Performance Increase Performance (SPEC Int) MIPS M2000 SUN-4/260 MIPS M/120 1 Intel Pentium 4/3000 DEC Alpha 21264A/667 DEC Alpha 21264/600 Intel Xeon/2000 DEC Alpha 5/500 DEC Alpha 4/266 DEC Alpha 5/300 DEC AXP/500 IBM POWER 100 HP 9000/750 IBM RS Year 29

28 DRAM Capacity Growth Kbit capacity M M 128M256M 16M M 1M K 64K K Year of introduction 30

29 Chip Fabrication * An silicon ingot is sliced into wafers Patterns are etched onto the wafers by repeatedly treating them with chemicals and exposing them to ultraviolet light Wafers are cut into dies Dies are bonded to packages 31

30 Chip Fabrication Figure 1.14: The chip manufacturing process 32

31 Chip Fabrication Die photo of an Intel 45 nm shuttle test chip including 153 Mbit SRAM and logic test circuits Intel 300 mm wafer with 45 nm shuttle test chips 33

32 Power * Power consumption is becoming an increasingly important issue in chip design Embedded devices typically run on batteries and so power must be conserved Even in desktop computers, power is a concern since the more power a chip uses the more heat it generates 34

33 Impacts of Advancing Technology Processor logic capacity: performance: increases about 30% per year 2x every 1.5 years ClockCycle = 1/ClockRate 500 MHz ClockRate = 2 nsec ClockCycle 1 GHz ClockRate = 1 nsec ClockCycle 4 GHz ClockRate = 250 psec ClockCycle Memory Disk DRAM capacity: 4x every 3 years, now 2x every 2 years memory speed: 1.5x every 10 years cost per bit: decreases about 25% per year capacity: increases about 60% per year 35

34 Example Machine Organization Workstation design target 25% of cost on processor 25% of cost on memory (minimum memory size) Rest on I/O devices, power supplies, box Computer CPU Memory Devices Control Input Datapath Output 36

35 Inside the Pentium 4 Processor Chip 37