Chapter 1 : Introduction

1
Chapter 1 Introduction

• CSIS 321
• Evans Adams

2
INTRODUCTION

• ARCHITECTURE
• The art and science of designing and
  constructing buildings, environments,
  communities, and structures.
• COMPUTER
• A digital computer is a machine that can solve
  problems for people by carrying out instructions
  given to it.

3

• A COMPUTER ARCHITECT is a person who
• should be a competent machine language programmer
  - preferably with experience in systems software.
• must know the nature of the computer systems
  building blocks (electronic circuits)
• should be exposed to problems which must be
  solved and how others have attempted to solve
  them.

4

• The study of computer architecture is the study
  of the organization and interconnection of
  components of computer systems.
• Computer architects construct computers from
  basic building blocks such as gates, memories,
  arithmetic logic units, serial and parallel
  interface adapters, buses, etc...

5

• From these building blocks, the computer
  architect can construct any of a number of
  different types of computers.
• These may range from the smallest, hand-held
  pocket calculator to the largest supercomputers.
• Thus, computer architecture is the discipline
  devoted to the design of highly specific and
  individual computers from basic building blocks.
• Computer organization and computer architecture
  shall be used as synonyms in this class.

6
LEVELS (See Figure 1-1)

• To communicate with a computer, we use the
  language of the machine - machine language.
• Most machine languages are very simple, indeed
  often only a few instructions such as add, test
  for zero, shift, transfer data between memory and
  registers.

7

• The instructions are simple because the expense
  of constructing such machines is correspondingly
  less expensive with respect to the complexity of
  the machine language.
• The problem is that humans are not very good at
  communication in these simple languages.
• Designers will develop a more convenient language
  for humans to use.

8

• The new language we will designate L2, and the
  machine language as L1.
• Since the machine can only execute instructions
  in L1, there must be a method to convert the
  instructions in L2 to L1. These are

• 1) Translation - the source program in L2 is
  translated in total to an equivalent program in
  L1.
• The original program in L2 may be discarded, and
  the object program (the equivalent program) in L1
  may now be executed.

9

• 2) Interpretation - the source program in L2
  is decoded on a line by line basis.
• The source program is considered as input to the
  interpreting program,
• each line is decoded and individually executed as
  a sequence of equivalent L1 instructions, and
• an equivalent machine language program is not
  generated.

10
VIRTUAL MACHINE

• A useful concept is the Virtual Machine or
  Hypothetical Computer whose machine language is
  L2.
• Although such a machine would be expensive to
  construct, people can write programs for virtual
  machines as though they existed.

11

• We view a computer as a series of virtual
  machines of layers, or levels, each one
  encompassing all others beneath it (See Figure
  1.1)
• level n ----- virtual machine Mn with machine
  language Ln
• level 1 ----- actual computer M1 with machine
  language L1
• level 0 ----- discrete electronic components

12

• Languages, levels, and virtual machines
• machine defines a language
• language defines a machine
• Any language may be implemented directly in
  hardware, but the cost may be prohibitive.

13

• A computer with n levels may be considered as n
  different virtual machines, each with a different
  machine language.
• The level implies the virtual machine, thus
  programs written directly in L1 can be directly
  carried out by the electronic circuits (or
  hardware).

14

• Programs written in L2, L3, , Ln must be
  translated or interpreted before they can be
  executed by the electronic circuits.
• To understand how computers really work, one must
  study all the levels.
• Structured computer organization comes from the
  concept of viewing a computer as a hierarchy of
  levels.

15
CONTEMPORARY MULTILEVEL MACHINES

• Most modern computers consist of two or more
  levels, we shall consider a six level machine
  (See Figure 1-2).
• Levels below 0 are the device level and solid
  state physics of the individual electronic
  devices. We shall consider

16

• Level 5 --gt Problem oriented language level
• - translation (compiler)
• Level 4 --gt Assembly language level
• - translation (assembler)
• Level 3 --gt Operating system machine level
• - partial interpretation
• (operating system)

17

• Level 2 --gt Instruction Set Architecture Level
• - interpretation (microcode program) or
• direct execution
• Level 1 --gt Microarchitecture Level
• - microprograms are executed
• directly by the hardware
• Level 0 --- Digital logic level
• - transistor, capacitors, etc
• --- Device/Solid State Physics Level

18

• Level 0 - Digital logic Level ? Gates
• Gates are digital (on/off, 0/1, true/false,
  high/low).
• Transistors (building blocks for gates) are
  analog (continuous values over a given range).
• Gates typically consist of a few transistors
• Registers

19

• Level 1 - Microarchitecture Level (Fig 2-2)
• True machine language level, instructions are
  executed directly on the hardware.
• A collection of registers that form a local
  memory and,
• a circuit called an ALU, and
• a data path connecting ALU and the registers
• Data Path may be controlled by a microprogram or
  directly by hardware

20

• Level 2 - Instruction Set Architecture Level
• The instructions carried out interpretatively by
  a microprogram or directly by hardware execution
  circuits at Level 1
• This is the language defined by the vendors
  instruction set, i.e., the Machine Language
  Reference Manual
• Manufacturers may provide multiple microprogram
  interpreters, implying multiple instruction sets
  for the same machine

21

• Level 3 - Operating System Level
• This is a hybrid level because most instructions
  are identical to the level 2 instructions.
• The new added facilities have, historically, been
  called an OPERATING SYSTEM. Thus, only a partial
  interpretation is done at this level.
• These features include such capabilities as file
  management, virtual I/O, memory management, task
  management, etc...

22

• Level 4 - Assembly language level
• The lowest three levels are intended primarily
  for running the interpreters and translators
  needed to support the higher levels.
• Levels 2 and 3 are always interpreted, and the
  programs are written by people called systems
  programmers who specialize in writing new virtual
  machines.
• Level 4 and above are intended for applications
  programmers.
• The upper levels are usually supported by
  translation, but may be supported by
  interpretation.

23

• The primary distinction between the upper levels
  and the lower levels is that the UPPER LEVELS use
  MNEMONICS (symbolic) whereas the LOWER LEVELS use
  NUMERICS.
• Level 4 is a symbolic form of one of the
  underlying levels.
• The assembly language programs are first
  translated to one of levels 1, 2 or 3, and
• the resulting program is interpreted by the
  appropriate virtual or actual machine.
• An assembler is a translator!

24

• Level 5 - Problem-oriented language level
• High-level languages, problem-oriented languages
• Ada ALGOL68 JAVA
• BASIC C COBOL
• FORTRAN LISP MODULA-2
• Pascal PROLOG SNOBOL
• If TRANSLATED, it is a COMPILER, some are
  INTERPRETED such as BASIC, LISP and many
  scripting languages.

25

• Level 6 ??
• It is possible to imagine virtual machines
  specifically tailored for specific applications
  such as computer design, manufacturing,
  administration, etc...

26

• The Set of
• Data Types,
• Operations, and
• Features
• available at each level is called its
  Architecture
• The Architecture deals with those aspects that
  are visible to the user of a level
• The study of how to design those parts of a
  computer system that are visible to programmers
  at different levels is called Computer
  Architecture.

27
HISTORICAL EVOLUTION OF MULTILEVEL MACHINES

• 1951
• M. V. Wilkes suggested a 3 level machine to
  reduce hardware complexity.
• Prior to that, digital computers were 2 level
  machines.
• Wilkes is the father of microprogramming

28

• The 1960s
• Operating systems developed to automate the
  operators job.
• Both BATCH systems and TIME SHARING systems were
  developed during this time period, and both are
  still in use today.
• By 1970
• The idea of having a microprogram to interpret
  the conventional machine language was widespread.

29

• Microprograms tended to get slower as new
  instructions and new features were added
• Researchers began to
• Eliminate the microprogram
• Reduce the instruction set, and
• have the remaining instructions executed
  directly by the hardware
• to speed up machines using RISC design principles

30
HARDWARE, SOFTWARE AND MULTILEVEL MACHINES

• HARDWARE consists of TANGIBLE OBJECTS - ICs,
  printed circuit boards, power supplies, memories,
  printers, etc
• SOFTWARE consists of algorithms and the data
  structures which support them.
• It is usually represented on some medium, but
  the instructions and data are really the
  SOFTWARE, an INTANGIBLE OBJECT.

31

• Intermediate between software and hardware is
  FIRMWARE, which is software embedded in
  electronic devices (ROM).
• Firmware is used for rarely changing software and
  when permanent retention of software is required.

32

• In important concept of this course is that
  SOFTWARE AND HARDWARE ARE TOPOLOGICALLY
  EQUIVALENT.
• An instruction can be performed by software or
  built into hardware. The designer must consider
• cost
• speed
• reliability
• expected frequency of changes

33

• There are no definite guidelines which determine
  what should be done in hardware and what should
  be done in software.
• Thus boundaries between the two should be
  considered fluid.
• (See Fig 1-7 IBM 360 Product Line)

34
PROCESSES

• Another fundamental concept in this course is the
  PROCESS.
• A PROCESS (often called a sequential process) is,
  essentially, a program in execution.
• It is an active entity capable of causing events
  to occur.

35

• In contrast, a program is a passive entity.
• At any time, a process is in a specific STATE.
• The state of a process tells how far the process
  is in its execution. A process consists of the
  following
• program
• pointer to the next instruction to be executed
• values of the programs data
• status of all I/O devices
• status of internal flags and, possibly, other
  status information

36

• If we prohibit processes that alter the program
  during execution (a very poor programming
  practice!), then a process has two parts
• program (which may not change)
• state vector (which changes during execution).
• A PROCESSOR then is a machine (actual or virtual)
  which advances a process from one state to the
  next.

37

• Thus a process passes through a time-ordered
  sequence of states. This gives rise to some
  important properties of a process
• The effect of a process is independent of its
  execution speed.
• If a process executes again with the same data,
  it goes through precisely the same sequence of
  states and produces the same results.
• THE PROCESSOR CAN, ITSELF, BE ANOTHER PROCESS.

38

• An interpreter in execution may be considered a
  processor.
• An interpreter is a program that, when executed
• pushes the state vector of some process, P,
• through the sequence of states required by Ps
  program.
• Hardware designers often build software
  interpreters to simulate new hardware features

39
MILESTONESin Computer Architecture

• Generation 0 - Mechanical Computers
• Pascal (1642) - Addition and Subtraction
• Leibniz (1672) - , -, , /

40

• Babbage (1830s) - Difference Engine and
  Analytical Engine
• Same Basic Design as Modern Computers
• Store (Memory)
• Mill (Computation Unit)
• Input Section (Punched Card Reader)
• Output Section (Punched and Printed Output)

41

• Analytic Engine was the first General Purpose
  Computer
• By punching different programs on the input
  cards, it performs different computations.
• Hardware Technology of the period was inadequate
  to fully construct the machine.

42

• Zuse (1930s) - built a series of computers using
  electromagnetic relays (his work was destroyed by
  an Allied bombing in 1944).
• Atanasoff, Stibbitz, Aiken (early 1940s) used
  electromagnetic relays and capacitors to design
  computing machines of limited capability.

43

• Generation 1 - Vacuum Tubes (1945-55)
• COLLOSUS (1943) - first electronic digital
  computer
• Built by British Intelligence to decode German
  messages (Alan Turing helped design this
  machine).
• Classified as a military secret for 30 years.

44

• ENIAC (1946) - John Mauchley and J. Presper
  Eckert - US Army contract for calculating range
  tables for firing heavy artillery.
• 18,000 vacuum tubes and 1500 relays
• Weighed 30 tons and consumed 140 Kilowatts of
  power
• 20 registers which held 10-digit decimal numbers

45

• Programmed by setting up to 6000 multiposition
  switches and connecting sockets with jumper
  cables.
• Not built until the war was over, but work was
  presented at a summer institute which sparked
  interest in building large digital computers.
• Eckert-Mauchley Computer Corporation which
  eventually became the UNISYS Corporation was
  founded based upon this research and produced
  UNIVAC (the first commercially available
  computer).

46

• EDSAC and IAS
• John Von Neumann - a genius and the most eminent
  mathematician in the world - expanded upon the
  work of Eckert/Mauchley.
• First Stored-Program Computer (eliminated the
  need to flip switches and connect sockets)
• Basic design still used today (see Figure 1-5).
• IBM 701, 704, 709 - scientific computers based
  upon vacuum tube technology.

47

• Generation 2 - Transistors (1955-65)
• Bardeen, Bratain and Schockley developed the
  transistor at Bell Labs in 1948 (won Nobel Prize
  in 1956).
• Caused vacuum tube computers to be obsolete.
• TX-0 - first transistorized computer built at MIT.

48

• Kenneth Olsen left MIT to form DEC in 1957 to
  market the PDP-1 (based upon TX-0 design).
• DEC venture capitalists believed there was no
  market for computers and mainly marketed small
  circuit boards.
• PDP-1 was finally marketed in 1961 (first
  minicomputer) for 120,000.
• IBM mainframes (7090 was fastest at this time)
  cost millions.

49

• PDP-8 developed a few years later sold for
  16,000 and was more powerful than PDP-1.
• Major innovation was a single bus (omnibus)
  (Figure 1.6).
• 50,000 were sold making DEC the leader in
  minicomputers.

50

• CDC 6600 (1964) - multiple processor scientific
  computer.
• Burroughs B5000.
• Instruction set built specifically to execute
  Algol 60 programs.
• First machine optimized for software.

51

• Generation 3 - Integrated Circuits (1965-80)
• Technology allowed dozens of transistors to be
  put on a single chip and building computers that
  are smaller, faster and cheaper.

52

• IBM System 360 - a single product line that
  combined features for scientific and business
  computing.
• First family of computers with the same
  assembly language and increasing size and power
  (see Figure 1-7).
• Software for a smaller model runs on a larger
  model without modifications.

53

• Downward compatible with older IBM machines.
• First to introduce MULTIPROGRAMMING (several
  programs in memory at once so that one could
  execute while another was waiting for I/O).
• IBM 370, 4300, 3080 and 3090 series all used this
  same architecture.

54

• DEC PDP-11 succeeded PDP-8 and became very
  popular at universities.
• IBM and DEC dominated mainframe and minicomputer
  markets.

55

• Generation 4 - Personal Computers
• and VLSI (1980-?)
• Very Large Scale Integration (VLSI) allows
  currently millions of transistors on a single
  chip leading to smaller and faster computers.
• Personal computer completed the cycle from large
  companies to departments to individuals being
  able to afford computer technology.
• Current Categories of Computers (See Figure 1.9).

56

• Moores Law
• An empirical observation about how fast solid
  state physicists and process engineers are
  advancing the state of the art
• Physical limitations indicate that it will
  probably hold until about 2020 (See Figure 1-8)
• Nathans Law
• Software is a gas. It expands to fill the
  container holding it.
• Software with additional features drives the
  demand for faster processors, larger memories and
  increased I/O capacity
• The Computer Spectrum (see Figure 1-9)

57
Example Architectures

• Pentium II
• a traditional CISC (Complex Instruction Set
  Computer) using modern superscalar (pipeline)
  technology
• UltraSPARC II
• a RISC (Reduced Instruction Set Computer) with
  superscalar technology
• picoJava II
• a dedicated Java chip for use in embedded systems

58

• Intel Chips
• 4004 - First single-chip CPU (4-Bit) to support
  hand-held calculators (1970)
• 8008 - 8-bit version of 4004 (1972)
• 8088 - 16-bit CPU selected by IBM for its first
  PC
• All Intel chips are backward compatible wrt
  software
• Great for running older software
• A nuisance to chip designers who would prefer a
  simple, clean, modern architecture unencumbered
  by mistakes and technology from the past
• (see Figure 1-10)

59

• SPARC Architecture
• Andy Bechtolsheim - a Stanford Graduate student
  built a personal UNIX workstation out of
  off-the-shelf parts called the SUN-1
• SUN-1 evolved into the first commercial product
  of Sun Microsystems (using Motorola CPUs)
• Suns initial products were powerful
  workstations with built-in networking support
• Sun developed the SPARC (Scalable Processor
  ARChitecture) based upon the RISC II design of UC
  Berkeley
• Sun decided not to manufacture the SPARC chips
  itself, but to license several manufacturers to
  produce them
• Sun hoped to build competition to the Intel
  world by making SPARC was an open-architecture

60

• The SPARC architecture is a specification of
• An instruction set, and
• other programmer-visible features
• Various implementations of the SPARC
  architecture have been developed by different
  hardware vendors
• We will mainly study the generic SPARC
  architecture specification, except, when
  discussing CPU chips, a specific SPARC chip used
  in Sun Workstations will be used
• Specifically, we use the 64-bit V9 UltraSPARC
  II architecture

61

• picoJava II Architecture
• Java Virtual Machine Architecture
• 32-bit memory words and 226 instructions
• Suns Java compiler compiles into JVM code
  (byte-code)
• A JVM interpreter then executes JVM code
• The JVM interpreter is written in C and can be
  compiled on any machine with a C compiler
• JVM code is shipped over the Internet
• JVM interpreter on a local machine then
  interprets JVM code leading to Javas portability
• Most browsers contain a JVM interpeter

62

• Since interpretation is relatively slow, Sun
  (and other companies) have designed a JVM chip
  which can directly execute JVM programs
• picoJava II is an architecture specification
  which is the basis for a number of commercial
  chips such as the Sun microJava 701
• picoJava II has two optional units (a cache and
  a floating-point unit) which a chip manufacturer
  can either include or remove

63

• Computer Architecture - Main Topics
• The overall design of the level (and why it is
  designed that way).
• The kinds of instructions available.
• The kinds of data used.
• Mechanisms available for altering the flow of
  control.
• Memory organization and addressing.
• Relationship between instruction set and memory
  organization.
• Method by which each level is implemented.

64
(No Transcript)