CPT101-小豪的笔记

CPT101

1. Overview

1.1 Overview and history of computer architecture

Types of Computers

• Mainframe computers (1960s)

• Supercomputers (1970s)

• Workstations (1980s)

• Microcomputers (1980s)

• Personal computers (1980s)

• Microcontrollers（单片机） (1980s)

• Servers (1980s)

• Chip computer (?)

Computer Generations

• First (1944 to1958): vacuum tube(真空管;电子管)

• Second (1959 to1963): transistor(晶体管)

• Third generation (1964 to 1970): IC 集成电路（integrated circuit）

• Fourth generation (1971 to now: VLSI) 超大规模集成电路（very large-scale integration）

1.2 Computer Systems

Computer Hardware

• 5 categories
  • input
  • processing
  • output
  • storage
  • communications

Computer Software

• System software
• Applications software

Backward (Downward) Compatibility for new hardware

• Most software written for computers with old hardware can be run on computers with newer hardware

VDHL 极高速集成电路硬件描述语言（Very High Speed Integrated Circuits Hardware Description Language）

• A programming language to be used to specify both the structure and function of hardware circuits.
• Supports computer simulations as well as providing input to automatic layout packages which arranges the final circuits.

Hierarchy of Systems

Operating System

• Functionalities of hardware systems can be brought out by operating systems and thus offered to the user.
• The user’s programs interact with hardware systems through the functionalities provided by operating systems.

Hardware evolution: Moore’s Law

• A circuit designed 24 months ago can now be shrunk to fit into an area of half the size.
  • It is sometimes quoted as every 18 months.

2. Input-Process-Output Model

###2.1 Input-Process-Output

There are three components required for the implementation of Input-Process-Output and von Neumann model(s):

• Hardware.
• Software.
• Data that is being manipulated.

Hardware

• Central Processing Unit (CPU) is an active part which performs calculations and other operations.
• The main memory (primary storage or working storage), or RAM (for random access memory) holds data and programs for access by CPU.
• Memory is volatile…
• The secondary storage.
  • Long-term storage.
  • Holds programs and data.
  • Hard disk, CDs, DVD, etc.
• Input devices: keyboard, mouse, scanner, etc.
• Output devices: monitor, speaker, printer, etc.

Software

• The hardware of a computer (e.g. CPU) can carry out only very simple operations like adding numbers (very quickly).
• To make it perform useful tasks, these simple steps are combined in the form of programs, which are collectively known as software.

Machine instructions

• The CPU performs the execution of machine instructions.
• Every CPU has its own instruction set (100- 200 instructions, typically).
  • For a particular machine, this set is fixed.
• Although the instruction sets of different CPUs are similar, there is no standard instruction set.

Machine Instruction Categories

• Input-output: IN, OUT (Intel x86 and Pentium, but does not exist in some CPUs), …
• Data transfer and manipulations: MOV, ADD , MUL , AND , OR , …
• Transfer of program control: JMP, JC, …
• Machine control: can halt processing, reset the hardware, INT, HLT …

Machine Instructions and HLL

• High Level Programming Languages (HLLs) are more suitable for programming than the languages of machine instructions.
• The programs in HLL still have to be translated to the machine codes.

2.2 The von Neumann Model

• The idea was formulated by von Neumann (late 1940s).
  • The computer is a general-purpose machine controlled by an executable program.
• In this context:
  • A program is a list of instructions used to direct a task.
  • Both program and data are held in computer’s memory (store) and both represented by binary codes.
  • The fact that memory is re-writeable makes a von Neumann machine especially powerful.
  • A processor is an active part of the machine that executes the program instructions.

• Input device is for transmitting information from a user into the computer’s memory.
• Output device enables a user to see results of the program being performed.
• Von Neumann bottleneck.
  • CPU is continuously forced to wait for vital data (and instructions) to be transferred to or from memory.

Harvard architecture

• Separates data from programs.
• Requires different memories and access buses for programs and data.
• The intention is to increase transfer rates, improving throughput.

3. Machine instructions and HLL

Semantic gap

• The term expresses the enormous difference between the way human languages expressing ideas and actions and the way computer instructions representing data processing activities.

###3.1 Translation

• Translation is done by special programs such as：
  • Compilers, translating HLL instructions into machine code (sequence of instructions) before the code can be run on the machine.
  • Assemblers, translating mnemonic form of machine instructions (like MOV, ADD, etc) into their binary codes.
  • Interpreters, translating HLL instructions into machine code on-the-fly (while the program is running).

compilers and assemblers

Linking

• Big programs usually are divided into several separate parts or modules.
• Each module has to be designed, coded and compiled.
• There are frequent occasions when code in one module needs to reference data or subroutines in another module.
• A compiler can translate a module into binary codes, but it cannot resolve those references to other modules.
• Those external references remain symbolic after the compilation, until the linker gets to work.
  • The linker is to join together all the binary parts.
  • The linker will report errors if it cannot find the module or code referred to by those external references.

Library files

• Translated object code.
• Provide many functions for programmers, but are only usable if linked into your code.
• In Unix:
  • Directories /lib and /usr/lib/.
• In Windows:
  • DLL files.

3.2 Interpreters

alternative way of running HLL programs

• Instructions are converted into an intermediate form, consisting of tokens.
  • In Java, tokens such as: static, boolean, file, string, void, return
• Tokens are then passed to the decoder, which selects appropriate routines for execution.
• Compilers.
  • Take a program and translate it as a whole into machine code.
  • The processes of translation and execution are separate.
• Interpreters.
  • Take an instruction, one at a time, translate and execute it.
  • The processes of translation and execution are interlaced.

C program compilation, linking & execution

• C language source code --> compiler ( program) --> assembly language --> assembler --> machine code
• Once we have machine code:
• machine code --> linking and loading (program) --> program code execution (program)

Java

Java source code --> compiler (program) --> Java “byte codes” --> Java interpreter (program)

Interpreters vs. Compilers

• Execution of compiled code is much faster than execution of interpreted code.
• Interpreters are more suitable for rapid prototyping and for other situations when a program is frequently modified.
  • Interpreters are more accurate in terms of error reporting.
  • Interpretation can provide uniform execution environment across several diverse computers. (Portable)

Interpreters as Virtual Machines

• Interpreters are somewhat similar to the computer hardware (CPU)
  • take one instruction at a time and execute it.
• Because of that sometimes they are referred to as a virtual machine
  • Example: JVM, Java Virtual Machine

###3.3 Code sharing and reuse

• How to reuse existing proven software when developing new systems?

  • Source-level subroutines and macro libraries.

  • Pre-translated re-locatable binary libraries.

  • Dynamic libraries and dynamic linking.

How code can be shared

Source-level subroutines and macro libraries

• Intention.
  • Take copies of the library routines.
  • Edit them into your new code.
  • Translate the whole together.
• Disadvantages.
  • Who owned the code?
  • Who should maintain it?

Pre-translated relocatable binary libraries

• Intention.
  • Libraries are pre-translated into relocatable binary code.
  • Can be linked into your new code, but not altered.
• Acceptance.
  • Successful, and still essential for all software development undertaken today.
• Disadvantage.
  • Each program is to have a private copy of the subroutines, wasting valuable memory space, and swapping time, in a multitasking system.

Dynamic libraries and dynamic linking

• Intention.

  • Load a program which uses “public” routines already loaded into memory.
  • The memory-resident libraries are mapped, through the memory management system to control access and avoid multiple code copies.

• Acceptance.

  • Successful through Microsoft’s ActiveX standard.

4. Data, Information and Knowledge

• Data – raw facts, figures, measurements, …
  • 1.00001, 1.00000010, 2.0000101, 3.0000102, 5.000…
• Information
  • data organized into useful representation
  • 1.000, 1.000, 2.000, 3.000, 5.000, …
• Knowledge
  • application of reasoned analysis of information
  • ‘data are in increasing order’, ‘data can be derived based on Fibonacci sequencing’, etc

4.1 Alphanumeric codes

• The majority of the data originally comes in the form of letters in alphabet, numbers and punctuation (alphanumeric data).
  • They are represented in computers by binary numbers.

bit

A bit is the most basic unit of information possible: it contains the information necessary to distinguish two alternatives (1 or 0, YES or NOT, etc.).

• ASCII code ( American Standard Code for Information interchange (7-bit code) and its extensions (8-bit codes) (well-established).
• EBCDIC code ( Extended Binary Coded Decimal Interchange Code) 8-bit code. (IBM mainframe computers)
• Unicode. Recent 16-bit standard. (Up to 216 characters can be encoded)

ASCII code table

• Only half of possible byte (8-bits) patterns is used.
• The table is divided into two classes of codes:
  • Printing characters.
  • Control characters.
• Printing characters produce output on the screen or on a printer.
• Control characters are used:
  • To control the position of output on the screen or paper (e.g. ‘HT’).
  • To cause some action to occur (e.g. ‘BEL’).
  • To communicate status between the computer and an I/O device (e.g. ‘Control-C’ combination).

Limitations of ASCII code

• The limitations of the well-established 8 bit ASCII codes.
  • Too limited for the display requirements of modern Windows-based word-processors.
  • The requirement of global software market for handling international character sets.

Unicode

• Even 8-bit extensions of ASCII code table is capable to code only up to 256 characters.
• Unicode Standard (1991) is an 16-bit international encoding system for information interchange.
  • Code values are available for more than 65,000 characters.

Representation of numbers

• Two’s complement(补码) as a method of representing and manipulating negative integers.

5 = 00000101
-5 = 11111011 

Representation of real numbers

• IEEE 754 standard.
  • The most widely-used standard for floatingpoint computation.
  • defines formats for representing floating-point numbers, special values, and a set of floatingpoint operations that operate on these values.

Declaration of variables in programs

• What happens when you declare variables in a program?
  • You are telling the compiler to reserve the correct amount of memory space to hold the variable.
  • You are also telling the compiler what encoding/decoding/representation scheme to be used.

5. Operating systems

5.1 Operating systems: examples

• OS/360 for IBM System/360, 1960s.
• Unix, 1970s.
• MS-DOS for IBM PC and Mac OS for Apple Macintosh, 1980s.
• Windows 95, 98, NT, 1990s.
  • NT served as the basis for Microsoft’s desktop operating system line starting in 2001.
• Apple rebuilt their operating systems on top of a Unix core as Mac OS X, released in 2001.
• Linux, BSD Unix…

5.2 Onion ring model

• Core of operating system: dealing directly with the hardware.
  • Kernel: device drivers, memory allocator…
  • CLI(命令行界面) : provide user accessibilities to the system

5.3 Interaction with operating system

• CLI (command line interpreter.)
  • DOS: type a command in a command line.
  • Unix/Linux: shell scripts (sequences of instructions).
  • Windows/Mac OS X: click with mouse on icons.

Computer Networks

• Perhaps the most far-reaching changes ever produced to von Neumann’s original blueprint.
• Operating system usually provides access to network facilities. (via networking API, e.g. socket interface)
• Computer network is an interconnected collection of autonomous computers to facilitate fast information exchange.

5.4 Client-server computing

• Client(客户端): The originator of a request.
• Server(服务器): The supplier of the service.

Client-server interaction

• Client starts the interaction by sending a request message to the server.
• Server responds by sending replies back…

6. Principal components of a computer

• These are the minimum set of components for a working digital computer.

6.1 Motherboard

• Three principal subsystems:
  • CPU
  • main memory
  • input-output units

###6.2 Processor and Registers

• Processor
  • arithmetic/logic unit (ALU) (运算器)
  • control unit(控制单元): part of a CPU responsible for performing the machine cycle-fetch, decode, execute, store
• Registers
  • Program counter (PC)(程序计数器): contains the address of the next instruction to execute
  • Instruction register (IR)(指令寄存器): part of a CPU control unit that stores an instruction

Coprocessors: Assistants to the CPU

• Coprocessors: microprocessors performing specialized functions that CPU cannot perform or cannot perform as well and as quickly
  • math
  • graphics

6.3 Buses

• On the motherboard, all the components are interconnected by buses (“signal highways”).
• A bus is a bundle of conductors, wires, or tracks.
• Typically, there are address, data and control buses, each including several signal lines.
  • Intel 8086: 20 shared address/data lines, and a further 17 lines for control.
  • Intel Pentium: data bus 64 lines, and the address bus 32 lines.

• Each hardware unit is connected to all these buses.
  • A simple way of building up complex systems in which each unit can communicate with each other.
  • Little disruption when plugging in new units and swapping out failed units.

6.4 Two parts of CPU

6.5 Registers

• CPU registers: small block of fast memory.
  • Temporarily store for data and address variables.
• Some CPU registers:
  • Instruction Pointer (IP) or Program Counter (PC).
    • Stores the address of the next instruction.
  • Accumulator (AX, EAX in Pentium).
    • General purpose data register.
  • Instruction Register (IR).
    • Stores the instruction that is being executed.
• Memory address register (MAR).
  • Temporarily holds address of the memory location during a bus transfer.
• MBR

6.6 Instruction Set

• The collection of machine language instructions that a particular processor understands
• machine language instructions
  • instructions for a specific CPU
  • designed to be executed by a computer without being translated
  • Also called machine code
  • Operations like: ADD, SUB, INC, DEC, etc.

How instructions are executed?

• The basic operation, known as the fetchexecute cycle or machine cycle.
  • The sequence whereby each instruction of the program is executed:
    • Read from the memory.
    • Decoded.
    • Executed.

Machine Cycle

• Fetch the instruction from memory. This step brings the instruction into the instruction register, a circuit that holds the instruction so that it can be decoded and executed
• Decode the instruction
• [Read the effective address from memory if the instruction has an indirect address ]
• Execute the instruction
• [Store the results]

The fetch phase of the cycle

• The address in IP register is copied onto the address bus and further to MAR register.
• IP is incremented ready for the next cycle. IP now points to the next location in the program memory.
• Memory selects location and copies the content onto the data bus.
• CPU copies the instruction code from the data bus into IR.
• Decoding of the instruction starts.

• A Pentium instruction: 10111000 00000000 00000001 • Assembly code: MOV AX 0x100

• Note that the content of a memory cell is different from its address (not shown in the figure).

The execution phase of the cycle

• Execute phase depends on the type of instruction.
• Example: the execution of MOV AX,256 instruction includes:
  • IP is copied to address bus and latched into memory.
  • IP is incremented.
  • The value selected in memory is copied onto the data bus.
  • CPU copies the value from the data bus into AX.

6.7 CISC & RISC

• CISC (“sisk”)
  • complex instruction set
  • most mainframes and PCs
• RISC (“risk”)
  • reduced instruction set (精简指令集)
  • cheaper and faster
  • shift some work to software

CISC vs RISC

• In RISC an instruction usually consists of a single word but in CISC an instruction may be several words long, requiring several fetches

RISC is faster because …

• The vacated area of the chip can be used to accelerate the performance of more commonly used instructions, rather than compensating for those rarely used instructions
• Easier to optimize the design
• Simplifies translation from high-level languages into the smaller instruction set that the hardware understands, resulting in more efficient programs

7. Hardware

7.1 Output Hardware

• Hardcopy output
  • graphics
  • letters
• Softcopy output
  • video
  • audio

7.2 Screen Clarity

• Standard screen resolutions
  • 640 x 480
  • 800 x 600
  • 1024 x 768
  • 1280 x 1024
  • 1600 x 1200

7.3 Communications Hardware

• Facilitate networks
  • modems
  • hubs and other components of a network

7.4 Ports

connecting peripherals to the computers

• Parallel port (IEEE 1284)
  • printers, some scanners
• Serial port (RS-232)
  • modems, scanners, mice

7.5 USB (Universal Serial Bus)

• USB
  • industry standard developed in the mid-1990s that defines the cables, connectors and protocols used for connection, communication and power supply between computers and electronic devices
  • standardized the connection of computer peripherals, such as keyboards, pointing devices, digital cameras, printers, portable media players, disk drives and network adapters to PCs
  • replaced earlier interfaces, such as serial and parallel ports, as well as separate power chargers for portable devices

7.6 Connectors

7.7 Power supply

• Power supply
  • protected by power surge protector or
  • uninterrupted power supply unit (UPS)

8. Data codes – numeric and character

• To store numbers we need an encoding scheme, which would allow us to encode:
  • The algebraic sign of numbers (+/-).
  • Decimal point that might be associated with a fractional number.

8.1 Unsigned integers: BCD

Each decimal digit is individually converted to binary.

• This requires 4 bits per digit (not all 4-bits patterns are used).

8.2 Sign-and-magnitude representation

• It is representation of signed integers by a plus or minus sign and a value.
• Agreement. – Left-most bit represent a sign,
  • e.g., 0 stands for + and 1 stands for -.
  • 8-bits can represent the numbers from -127 to 127 (0 being represented twice).

8.3 10’s complementary coding

8.4 Floating Point Numbers

Single-precision 32 bit IEEE 754

Double-precision 64 bit IEEE 754

9. Data storage

• Storage is the capacity of a device to hold and retain data.
• Two main types of storage in a computer:
  • Main memory.
  • Mass storage.

9.1 Main memory

• It refers to physical memory that is internal to the computer.
• The computer can manipulate only data that is inside the main memory.

9.2 RAM

随机存取器 (random access memory)

• The memory can be seen as a set of numbered storage elements, called words, each of which contains some information.
• Each word is numbered with its address.
• Any word of memory can be accessed “without touching” the preceding words (Random Access).
• Access time is the same for all the stored items.

Dynamic RAM (DRAM)

• Cheaper, but slower.

• Implemented via capacitors.

• DRAM needs to be refreshed.

Static RAM (SRAM)

• Faster, but more expensive.
• Implemented via flip-flops.
• No need for refreshing.

Both types of RAM are volatile

They lose their contents when the power is turned off.

9.3 ROM

• Read-only memory (ROM)
• Software stored inside also known as firmware
• Helps boot up the system
• BIOS – Basic Input Output System

9.4 Other Forms of Memory

Cache memory

• quick access memory, internal or external to the processor
• bridge between the processor and RAM
• including simultaneous read/write

Video memory

VRAM

Mass storage

• It refers to various techniques and devices for storing a large amount of data.
• Unlike main memory, mass storage devices retain data even when the computer is turned off.

Types of mass storage

• Hard disks.
• Optical disks: CD-ROM, CD-RW, DVD, etc.
• USB disks, Floppy disks.

Hard Disk Drives (HDD)

硬盘驱动器

• Hard disk drives are the most important types of permanent storage used in computers (esp. PCs).
• Hard disks differ from the other mass storage devices in three ways:
  • Size (usually larger).
  • Speed (usually faster).
  • Permanence (usually fixed in computer and not removable).

10. Memory

• Any memory location in main memory has its own address.

• It follows then the more memory the larger addresses are needed.

• Maximal memory length depends on address width

10.1 Address width

• Address width is determined by:
  • The number of bits in the CPU address registers such as IP, MAR.
  • The number of lines in the address bus.

10.2 Memory modules

###10.3 Memory mapping

• When the CPU sends out an address:
  • A part of the address locates the correct chip.
  • Another part specifies an address within the correct chip.
• How actually the addresses are mapped to the memory locations is defined by memory maps.

Memory map for a small system

10.4 Memory address decoding

• Memory chips are not normally matched to the width of the address bus. For example:
  • CPU may send 32-bit address.
  • RAM may receive directly 24-bit address.
• Special Memory Address Decoding circuit implements necessary decoding

10.5 Registers

• Registers are the memory cells which are core part of the processor itself.
• It has very fast access (a few nanoseconds).
• Not that much memory: tens of 32-, 64-, 80- bit registers (typically).

10.6 Cache memory

• A memory (more expensive, but faster SRAM) placed between CPU and main memory.
• Contains a copy of the portion of main memory.
• The aim is to maintain in fast cache the currently active sections of code and data.
• Processor when needs some information first checks cache.
• If not found in cache, the block of memory containing the needed information is moved into the cache.

Levels of cache

• Typically Level 1 cache has the size 8-64 KB.
• Typically Level 2 cache has the size 128-512KB

10.7 Localisation of access

• The idea of cache memory exploits Localisation of Memory Access principle:
  • Computers tend to spend periods of time accessing the same locality of memory.
  • A portion of code or data which require access needs to be loaded into the fastest memory nearest to CPU.
  • Other sections of the program and data can be held in readiness lower down the memory hierarchy.

Why localisation of access works?

• Partly due to the programmer clustering related data items together in arrays or records.
• Partly due to the repeating patterns in a program (i.e. loops)
• Partly due to the compiler attempting to organise the code in an efficient manner.

10.8 Cache memory and cache control unit

10.9 Memory hierarchy

• Going down the hierarchy:
  • Increased capacity.
  • Increased access time.
  • Decreased frequency of access of the memory by the processor.
  • Decreased cost per bit.

11. Hard disk drives

• Hard disk drives are the most important type of permanent storage used in computers (esp. PCs).

Schematic diagram of hard disk

Storage Technology

• Retrieving files into RAM is called reading

  • loading an application
  • opening a file
  • files can be programs or documents

• Copying data from RAM onto a secondary storage device is called writing

11.1 Virtual memory

• Virtual memory is a technique, in a sense, opposite to caching:

  • It is the use of low-level memory (i.e. hard disk) to ‘expand’ high-level (main) memory.
  • It provides a convenient expansion of main memory by ‘overflowing’ data and program code onto magnetic disk.

• The area on disk reserved for this purpose is known as the swap area.

11.2 Memory Management

• Virtual memory
  • hard disk space
  • when processor needs more RAM space, swaps unused data onto designated hard disk space
  • improves flexibility but is slower than RAM to which the processor has direct access

12. Digital systems

12.1 Boolean gates

12.2 Flip-flops(触发器), or latches

12.3 Use of D flip-flops - Copying data