Stuff to know! - PowerPoint PPT Presentation

1 / 44
About This Presentation
Title:

Stuff to know!

Description:

Stuff to know! Introductions Instruction Set Architecture (ISA) – PowerPoint PPT presentation

Number of Views:118
Avg rating:3.0/5.0
Slides: 45
Provided by: Paul7289
Category:

less

Transcript and Presenter's Notes

Title: Stuff to know!


1
Stuff to know!
  • Introductions
  • Instruction Set Architecture (ISA)

2
Todays topics
  • Architecture overview
  • Machine instructions
  • Instruction Execution Cycle
  • CISC machines
  • Microprograms
  • RISC machines
  • Parallelism
  • Instruction-level
  • Processor-level
  • Internal representation
  • Limits of representation
  • External representation
  • Binary, octal, decimal, hexadecimal number systems

3
(No Transcript)
4
Terms
  • CPU Central Processing Unit
  • ALU Arithmetic/Logic Unit
  • Memory storage for data and programs (separate
    from CPU)
  • Register fast temporary storage inside the CPU
  • Bus parallel "wires" for transferring a set of
    electrical signals simultaneously
  • Internal Transfers signals among CPU components
  • Control Carries signals for memory and I/O
    operations
  • Address Links to specific memory locations
  • Data Carries data CPU ? memory
  • Microprogram sequence of micro-instructions
    required to execute a machine instruction
  • Cache temporary storage for faster access
  • Note caching takes place at many levels in a
    computer system

5
Registers
  • General/Temporary fast local memory inside the
    CPU
  • one type of cache
  • Control dictates current state of the machine
  • Status indicates error conditions
  • IR Instruction Register (holds current
    instruction)
  • IP Instruction Pointer (holds memory address of
    next instruction) often called Program Counter
    (PC)
  • MAR Memory Address Register (holds address of
    memory location currently referenced)
  • MDR Memory Data Register holds data being set
    to or retrieved from the memory address in the MAR

6
Machine instructions
  • Each computer architecture provides a set of
    machine-level instructions
  • Instruction Set Architecture (ISA)
  • Specific to one particular architecture
  • Like everything inside a computer, machine
    instructions are implemented electrically
  • Micro-instructions set the switches in the
    control register

7
(No Transcript)
8
Hypothetical CISC machine
  • Shows hardware components
  • Does not show digital logic level or
    microprograms.
  • Shows how machine-level instructions can be
    stored and executed.
  • Illustrates
  • Finite-state machine
  • CISC
  • Complex Instruction Set Computer
  • VonNeumann architecture
  • Instruction execution cycle

9
Real computers
  • Use the stored program concept
  • VonNeumann architecture
  • Program is stored in memory, and is executed
    under the control of the operating system
  • Operate using an Instruction Execution Cycle

10
Instruction Execution Cycle
  • Fetch next instruction (at address in IP) into
    IR.
  • Increment IP to point to next instruction.
  • Decode instruction in IR
  • If instruction requires memory access,
  • Determine memory address.
  • Fetch operand from memory into a CPU register, or
    send operand from a CPU register to memory.
  • Execute micro-program for instruction
  • Go to step 1.
  • Note default execution is sequential

11
Example CISC Instruction
ADD R1, mem1 (Add contents of memory location
mem1 to register R1)
Example ADD Microprogram(each microinstruction
executes in one clock cycle)
  1. Copy contents of R1 to ALU Operand_1
  2. Move address of mem1 to MAR
  3. Signal memory fetch (gets contents of memory
    address currently in MAR into MDR)
  4. Copy contents of MDR into ALU Operand_2
  5. Signal ALU addition
  6. Check Status Register
  7. Copy contents of ALU Result to R1

12
Improving CISC
  • CISC speed (and convenience) is increased by
  • more efficient microprograms
  • more powerful ISA level instructions
  • cache memory
  • more registers
  • wider buses
  • making it smaller
  • more processors
  • floating point instructions
  • Etc.

13
Clock Cycles
  • So how slow is this?
  • It isnt slow
  • Execution near light-speed
  • Clock cycle length determines CPU speed
  • (mostly)

14
Limitations of CISC
Improving a specific architecture requires
instructions to be backward compatible. So how
about a different architecture?
15
RISC machinesReduced Instruction Set Computer
  • Much smaller set of instructions at ISA level
  • Instructions are like CISC micro-instructions
  • RISC assembly level programs look much longer
    (more instructions) than CISC assembly level
    programs, but they execute faster. Why?

16
RISC design principles
  • Instructions executed directly by hardware (no
    microprograms).
  • Maximize rate of fetching instructions.
  • Instruction cache
  • Instructions easy to decode
  • Fetching operands, etc.
  • Only LOAD and STORE instructions reference
    memory.
  • Plenty of registers

17
More speed improvement
  • Minimize memory and I/O accesses
  • Cache
  • Separate I/O unit (buffers/processing)
  • Separate network communication unit (NIC)
  • Etc.
  • Parallel processing

18
Parallelism (overview)
  • Instruction-level parallelism
  • pipeline
  • cache
  • Processor-level parallelism
  • multiprocessor (multiple CPUs, common memory)
  • multicomputer (multiple CPUs, each with own
    memory)

19
Pipelining
20
Pipelining Equations!
  • For k execution stages, n instructions require k
    (n 1)

S1 S2 S3 S4 S5 S6
1 I-1
2 I-1
3 I-1
4 I-1
5 I-1
6 I-1
7 I-2
8 I-2
9 I-2
10 I-2
11 I-2
12 I-2
S1 S2 S3 S4 S5 S6
1 I-1
2 I-2 I-1
3 I-2 I-1
4 I-2 I-1
5 I-2 I-1
6 I-2 I-1
7 I-2
No Pipelining 6-stage pipeline
21
Instruction Caching
  • Hardware provides area for multiple instructions
    in the CPU
  • Reduces number of memory accesses
  • Instructions are available for immediate
    execution
  • Might cause problems with decision, repetition,
    and procedure structures in programs

22
Multiprocessor (shared memory)
23
Multicomputer (distributed memory)
24
Comparisons
  • Cache and Pipelining
  • Implemented in hardware
  • Multiprocessor
  • Difficult to build
  • Relatively easy to program
  • Multicomputer
  • Easy to build (given networking technology)
  • Extremely difficult to program

25
Other types of parallelism
  • Hybrid systems
  • Scalable architectures
  • Add more processors (nodes), without having to
    re-invent the system
  • Simulated parallelism
  • Super-Scalar

26
Applications of Parallelism
  • Multi-user systems
  • Networks
  • Internet
  • Speed up single processes
  • Chess example
  • Other AI applications

27
Questions?
  • Parallelism
  • more later
  • Internal representation (tomorrow!)
  • Data
  • Instructions
  • Addresses

28
Internal representation
  • Just like everything else in a computer, the
    representation of numbers is implemented
    electrically
  • switches set to off or on
  • with open(transparent)/closed(opaque) gates.
  • There are two states for each gate
  • The binary number system uses two digits (0 and
    1)
  • In order to simplify discussion, we use the
    standard shorthand to transcribe the computer
    representation
  • off is written as digit 0
  • on is written as digit 1

29
External representation
  • Use the binary number system to represent numeric
    values electrically.
  • Switches (gates) are grouped into bytes, words,
    etc., to represent the digits of a binary number.
  • Note The number of gates in a group depends on
    the computer architecture and the type of data
    represented. E.G.,
  • For Intel-based architectures
  • byte 8-bits, word 2 bytes (16 bits)
  • integers use 2, 4, 8, or 10 bytes

30
Binary number system
  • has 2 digits 0 and 1 (binary digit)
  • has places and place values determined by powers
    of 2.
  • (in theory) can uniquely represent any integer
    value
  • A binary representation is just another way of
    writing a number that we are accustomed to seeing
    in decimal form.
  • (in practice, inside the computer) representation
    is finite
  • Representations with too many digits get chopped.

31
Internal representation
  • Place values (right-to-left) are 20,21,22,23,24,
    etc.
  • Bits are numbered (right-to-left) starting at 0
  • Place value depends on number of "bits" defined
    for the type.
  • Example
  • A 16-bit integer might be (red is "on")


15 14 13 12 11 10 9 8 7 6 5
4 3 2 1 0 (bit numbers) transcribed by
a human as 0000000010110010 To convert to its
familiar decimal representation, just add up the
place values of the places that are "on".
32
Converting binary to decimal
215 32768 214 16384 213 8192 212 4096 211 2048 210 1024 29 512 28 256 27 128 26 64 25 32 24 16 23 8 22 4 21 2 20 1
0 0 0 0 0 0 0 0 1 0 1 1 0 0 1
0 in decimal form 128 32 16 2 178 How
many different codes (integers) can be
represented using 16 bits? What is the largest
(unsigned) integer that can be represented using
16 bits? What is the largest (unsigned) integer
that can be represented using 32 bits? Prove that
for n-bit representation, number of codes is 2n,
largest unsigned integer is 2n 1, and largest
signed integer is 2n-1 - 1
33
Converting decimal to binary
  • Method 1 Removing largest powers of 2
  • Method 2 Successive division by 2

34
Converting decimal to binary
  • Example 157
  • Method 1 Removing largest powers of 2
  • 157 128 29
  • 29 16 13
  • 13 8 5
  • 5 4 1
  • 1 1 0
  • 1 0 0 1 1 1 0 1
  • Method 2 Successive division by 2
  • 157 2 78 R 1
  • 78 2 39 R 0
  • 39 2 19 R 1
  • 19 2 9 R 1
  • 9 2 4 R 1
  • 4 2 2 R 0
  • 2 2 1 R 0
  • 1 2 0 R 1
  • 1 0 0 1 1 1 0 1

35
Numeric representation
  • We will show (later) exactly how an electrical
    operation can be performed on two electrical
    numeric representations to give an electrical
    result that is consistent with the rules of
    arithmetic.

36
but not quite consistent
  • Since the number of gates in each group (byte,
    word, etc.) is finite, computers can represent
    numbers with finite precision only.
  • Example
  • Suppose that signed integer data is represented
    using 16 gates. Then the largest integer that
    can be represented is 65535. What happens if we
    add 1 ?
  • If necessary, representations are truncated
    overflow / underflow can occur, and the Status
    Register will be set

37
Representing negative integers
  • Must specify size!
  • Specify n number of bits (8, 16, 32, etc.)
  • There are 2n possible "codes"
  • Separate the "codes" so that half of them
    represent negative numbers.
  • Note that exactly half of the codes have 1 in the
    "leftmost" bit.)

38
Binary form of negative numbers
  • Several methods, each with disadvantages.
  • We will focus on twos-complement form
  • For a negative number x
  • specify number of bits
  • start with binary representation of x
  • change every bit to its opposite, then add 1 to
    the result.

39
Binary form of negative numbers
  • Example -13 in 16-bit twos-complement
  • -13 13 0000000000001101
  • ones-complement is 1111111111110010
  • add 1 to get 1111111111110011 -13
  • Note that -(-13) should give 13. Try it.
  • Hexadecimal representation?
  • Convert binary to hex in the usual way
  • -13 1111111111110011 FFF3 H 0xfff3
  • Note For byte, word, etc., if the first hex
    digit is greater than or equal to 8, the value is
    negative.
  • Convert negative binary to decimal?
  • Find twos complement, convert, and prepend a
    minus sign.

40
Signed numbers using 4-bit twos-complement
form
-8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7
1000 1001 1010 1011 1100 1101 1110 1111 0000 0001 0010 0011 0100 0101 0110 0111
  • Notice that all of the negative numbers have 1 in
    the leftmost bit. All of the non-negative
    numbers have 0 in the leftmost bit.
  • For this reason, the leftmost bit is called the
    sign bit
  • Note Nobody uses 4-bit representations
    (nibble), but theres not enough room to show
    8-bit representations here.
  • You can extend this diagram to 8-bit, 16-bit, etc.

41
n-bit twos-complement form
  • The 2n possible codes give
  • all zero
  • 2n-1 - 1 positive numbers
  • 2n-1 negative numbers
  • Note zero is its own complement
  • Note there is one weird number
  • 01111111 1 10000000
  • 127 1 -128
  • (inconsistent with rules of arithmetic)
  • 127 is the largest number that can be represented
    in 8 bits. This means that -(-128) cannot be
    represented with 8 bits.
  • i.e., the 2's-complement of 10000000 is 10000000

42
Signed or Unsigned?
  • A 16-bit representation could be used for signed
    or unsigned numbers
  • 16-bit signed range is -32768 .. 32767
  • 16-bit unsigned range is 0 .. 65535
  • Both forms use the same 216 codes
  • Example
  • 1010101010101010 unsigned is 43690 decimal
  • 1010101010101010 signed is -21846
  • Programmer must tell the computer which form is
    being used.

43
Other representations
  • Every integer number has a unique representation
    in each "base" ? 2
  • Hexadecimal is commonly used for easily
    converting binary to a more manageable form.
  • example 16-bit binary to hexadecimal
  • Binary 0001 0111 1011 1101
  • Hexadecimal 1 7 B
    D
  • Write it as 0x17BD or 17BDh

44
Questions?
Read Irvine Chapter 17.1
Write a Comment
User Comments (0)
About PowerShow.com