Scribd Logo
Upload

Welcome to Scribd!

  • 14 views

    Lec 03

    Uploaded by

    gdeepthi
    lect 3

    Copyright:

    Attribution Non-Commercial (BY-NC)
    Download as PDF, TXT or read online from Scribd

    Lec 03

    Uploaded by

    gdeepthi
    14 views16 pages
    lect 3

    Original Title

    lec03

    Copyright

    © Attribution Non-Commercial (BY-NC)

    Available Formats

    PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    lect 3

    Copyright:

    Attribution Non-Commercial (BY-NC)
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    14 views16 pages

    Lec 03

    Uploaded by

    gdeepthi
    lect 3

    Copyright:

    Attribution Non-Commercial (BY-NC)
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    of 16

    LECTURE - 03

    Amdahl's Law
    e
    Amdahl's law:

    Diminishing returns

    Limit on overall speedup


    1-F
    F
    1-F
    F/Speedup
    e
    Corollary: make the
    common case fast
    Overall speedup=
    (1-F)+F
    (1-F)+
    F
    Speedup
    Illustrating Amdahl's Law
    e
    Example: implement cache, or faster ALU?

    Cache improves performance by 10x

    ALU improves performance by 3x


    e
    Depends on fraction of instructions

    Suppose
    F
    mem
    =0.2, F
    alu
    =0.5, F
    other
    =0.3
    Speedup withcache=
    1
    0.8+0.2/10
    =1.22
    Speedup with faster ALU=
    1
    0.5+0.5/ 3
    =1.5
    Example continued...
    e
    Fixing for what value of is
    adding a cache better?
    F
    alu
    =0.5 F
    mem
    1
    1-F
    mem
    +F
    mem
    /10
    1.5
    =F
    mem

    10
    27
    =0.36
    The CPU Performance Equation
    CPUtime=Num. clock cyclesClock cycletime
    CPUtime=Num. of clock cycles-Clock rate
    OR
    CPUtime=ICCPI Cycle time
    Putting these together
    Num. of clock cycles
    =InstructionCount Cycles Per Instruction
    =ICCPI
    For a program,
    More on the Equation
    e
    This form is convenient

    Involves many relevant parameters


    e
    Remembering is easy
    CPUtime=
    Seconds
    Program
    =
    Seconds
    Clock cycle

    Clock cycles
    Instruction

    Instructions
    Program
    e
    With CPI as the independent variable
    CPI=
    CPU time
    Clock cycletimeIC
    Other Convenient Forms of the
    Equation
    e
    Number of clock cycles can be counted as:
    CPUclock cycles=
    _
    i=1
    n
    CPI
    i
    IC
    i
    Hence , CPUtime=(
    _
    i=1
    n
    CPI
    i
    IC
    i
    )Clock cycletime
    e
    Calculating in terms of
    CPI
    CPI
    i
    CPI =
    CPU time
    Clock cycletimeIC
    =
    _
    i=1
    n
    CPI
    i
    (
    IC
    i
    IC
    )
    Usefulness of the Equation
    e
    easier to measure than

    Equivalently, is measured through


    e
    Equation includes relevant parameters such
    as the cycle time
    IC
    i
    F
    i
    F
    i
    IC
    i
    Measuring the Parameters for
    the Equation
    e
    Clock cycle time:

    Easy for existing architectures

    Needs to be estimated in the design process


    e
    Instruction Count:

    Requires a compiler

    And, simulator/interpreter, or instrumentation code


    e
    CPI for each instruction type:

    Easy for simple architectures

    Pipelines, caches introduce complications

    Need to simulate and measure average CPI


    A Design Example
    e
    A design choice for conditional branch
    instructions:

    Choice 1: condition code is set by a compare


    instruction, checked by the next (branch)
    instruction
    e
    20% instructions are branches, and another 20% are
    compares
    e
    2 cycles per branch, 1 cycle for all others
    e
    Clock-rate is 25% faster

    Choice 2: single instruction for compare and


    branch
    e
    Which choice is better?
    Solution for Design Example
    CPUtime
    1
    =
    IC
    1
    |(0.81)+(0.22)
    1.25C
    =
    IC
    1
    C

    1.2
    1.25
    CPUtime
    2
    =
    IC
    1
    |(0.61)+(0.22)
    C
    =
    IC
    1
    C
    Detailed Example: Using Caches
    e
    Thumb rule in hardware design:

    Smaller is faster
    e
    Signal propagation delay is lesser
    e
    More power per memory cell
    e
    Observation w.r.t. software:

    Locality of reference

    Spatial as well as temporal


    The Memory Hierarchy
    CPU
    Registers
    Cache Memory
    I/O Devices
    Slower
    5ns 10ns 100ns O(10ms)
    Larger
    200B 512KB 512MB O(10-100GB)
    Cheaper
    Modifying the CPU Performance
    Equation
    e
    Caches involved hits and misses
    e
    Cache miss ==> memory stalls
    CPU time=(CPU clock cycles+Memory stall cycles)
    Clock cycle
    Memory stall cycles=Num. missesMiss penalty
    =ICMisses per instructionMiss penalty
    =ICMem. refs. per instructionMiss rateMiss penalty
    e
    Equation in the final form is useful: parameters
    can be measured
    Some Numerics...
    Fractionof memory access instructions=0.4
    CPI for memoryinstructions(hits)=2
    CPI for other instructions=1
    Choice1: 0.04 missrate , 25cycle penalty
    Choice2: 0.02miss rate , 50cycle penalty
    Which is a better choice?
    What is the overall average CPI?
    CPI
    avg
    =(0.61)+(0.42)+(1+0.4)0.0250=2.8
    Next week...
    e
    Instruction set architecture
    e
    Pipelining
    e
    Pipelining hazards

    You might also like