alu2 (2)

Embed Size (px)

Citation preview

  • 8/9/2019 alu2 (2)

    1/26

    SINGLE AND DOUBLE PRECISION

    FLOATING POINT

    MULTIPLICATION AND DIVISION

    ALU

    Reneesh C Zacharia

    Reg. no 2882122

    Internal Guide: Mr. T RAVI ME ,(Ph.D )

  • 8/9/2019 alu2 (2)

    2/26

    GOAL OF THE PROJECT

    Implementation of double precision floating

    point multiplication and division

    Implementation of pc interface of singleprecision floating point multiplication and

    division

  • 8/9/2019 alu2 (2)

    3/26

    INTRODUCTION

    The system is a 32-bit floating point ALU.

    The system can be expected to perform 32-bit

    multiplication and division operations. The system use four levels of pipelining.

    The floating point numbers are represented in

    IEEE-754 standard.

  • 8/9/2019 alu2 (2)

    4/26

    REPRESENTATIONREPRESENTATION OFOF FLOATINGFLOATING POINTPOINT

    NUMBERSNUMBERS ININ

    SINGLE PRECISIONSINGLE PRECISION IEEEIEEE 754754 STANDARDSTANDARD

    SIGN EXPONENTIAL MAGNITUDE

    22-030-2331

    IEEE 754 SINGLE PRECISIONIEEE 754 SINGLE PRECISION

    Bits of precision 24

    Unbiased exponent Emax 127

    Unbiased exponent Emin -126

    Exponent bias 127

  • 8/9/2019 alu2 (2)

    5/26

    REPRESENTATIONREPRESENTATION OFOF FLOATINGFLOATING POINTPOINT NUMBERSNUMBERS ININ

    SINGLE PRECISIONSINGLE PRECISION IEEEIEEE 754754 STANDARDSTANDARD

    Value = N = (-1)S X 2 E-127 X (1.M)

    SIGN EXPONENTIAL MAGNITUDE

    22-030-2331

  • 8/9/2019 alu2 (2)

    6/26

    REPRESENTATIONREPRESENTATION OFOF FLOATINGFLOATING POINTPOINT

    NUMBERSNUMBERS ININ

    DOUBLE PRECISIONDOUBLE PRECISION IEEEIEEE 754754 STANDARDSTANDARD

    SIGN EXPONENTIAL MAGNITUDE

    51-062-5263

    IEEE 754 DOUBLE PRECISIONIEEE 754 DOUBLE PRECISION

    Bits of precision 53

    Unbiased exponent Emax 1023

    Unbiased exponent Emin -1022

    Exponent bias 1023

  • 8/9/2019 alu2 (2)

    7/26

    REPRESENTATIONREPRESENTATION OFOF FLOATINGFLOATING POINTPOINT NUMBERSNUMBERS ININ

    DOUBLEDOUBLE PRECISIONPRECISION IEEEIEEE 754754 STANDARDSTANDARD

    Value = N = (-1)S X 2 E-1023X (1.M)

    SIGN EXPONENTIAL MAGNITUDE

    51-062-5263

  • 8/9/2019 alu2 (2)

    8/26

    ARITHMETIC UNIT DESIGN

    STRUCTURE

  • 8/9/2019 alu2 (2)

    9/26

    LOGIC UNIT DESIGN STRUCTURE

  • 8/9/2019 alu2 (2)

    10/26

    ADVANTAGES OF PIPELINING

    It can accept a new instruction every clock cycle

    The cycle time of the processor is reduced, thus

    increasing instruction issue-rate in most cases.

    Some combinational circuits such as adders or

    multipliers can be made faster by adding morecircuitry. If pipelining is used instead, it can save

    circuitry

  • 8/9/2019 alu2 (2)

    11/26

    DESCRIPTION-DOUBLE PRECISION (64-

    BIT)

    The 64-bit floating point numbers are stored in four

    pipelined registers.

    There are four pipelined units. They perform

    following functions in parallel.

    1. 0 operation number check

    2. Exponent addition and subtraction operation

    3. Fraction multiplication and division operation4. Result normalization and rounding

  • 8/9/2019 alu2 (2)

    12/26

    BASICBASIC FLOATINGFLOATING POINTPOINT MULTIPLICATIONMULTIPLICATION

    ALGORITHMALGORITHM

    Assuming that the operands are already in the IEEE

    754 format, performing floating point multiplication:

    Result = R = X * Y = (-1)Xs (Xm x 2Xe) *

    (-1)Ys (Ym x 2Ye)

    involves the following steps:

    (1) If one or both operands is equal to zero, returnthe result as zero, otherwise:

  • 8/9/2019 alu2 (2)

    13/26

    (2) Compute the sign of the result Xs XOR Ys

    (3) Compute the mantissa of the result:

    Multiply the mantissas: Xm * Ym

    Round the result to the allowed number of mantissa bits

    (4) Compute the exponent of the result:Result exponent = biased exponent (X) + biased exponent

    (Y) - bias

    (5) Normalize if needed, by shifting mantissa right,incrementing result exponent.

  • 8/9/2019 alu2 (2)

    14/26

    MULTIPLICATION FLOW CHART

  • 8/9/2019 alu2 (2)

    15/26

    BASICBASIC FLOATINGFLOATING POINTPOINT DIVISIONDIVISION

    ALGORITHMALGORITHM

    Assuming that the operands are already in the IEEE

    754 format, performing floating point multiplication:

    Result = R = X / Y= (-1)Xs (Xm x 2Xe)/

    (-1)Ys (Ym x 2Ye)

    involves the following steps:

    (1) If the divisor Y is zero return Infinity, if both are zeroreturn NaN

  • 8/9/2019 alu2 (2)

    16/26

    (2) Compute the sign of the result Xs XOR Ys

    (3) Compute the mantissa of the result:

    The dividend mantissa is extended to 48 bits by adding 0's to

    the right of the least significant bit.

    When divided by a 24 bit divisor Ym, a 24 bit quotient is produced.

    (4) Compute the exponent of the result:

    Result exponent = [biased exponent (X) - biased exponent (Y)]

    + bias

  • 8/9/2019 alu2 (2)

    17/26

    DIVISION FLOW CHART

  • 8/9/2019 alu2 (2)

    18/26

    FIRST STAGE OFCONNECTION

    GRAPH

    OPERAND CHECKING MODULE

  • 8/9/2019 alu2 (2)

    19/26

    SECOND STAGE

    EXPONENT ADDITION AND SUBTRACTION MODULE

  • 8/9/2019 alu2 (2)

    20/26

    THIRD STAGE

    FRACTION MULTIPLICATION AND DIVISION MODULE

  • 8/9/2019 alu2 (2)

    21/26

    FOURTH STAGE

    NORMALIZATIONAND ROUNDING MODULE

  • 8/9/2019 alu2 (2)

    22/26

    PC INTERFACE

    F NCTIONA OC OF PC INTERFACE

  • 8/9/2019 alu2 (2)

    23/26

    RESULTS-32 BIT

  • 8/9/2019 alu2 (2)

    24/26

    RESULT-64 BIT

  • 8/9/2019 alu2 (2)

    25/26

    REFERENCES

    Computer Organization by Carl Hamacher

    Verilog HDL Samir Palnikar

    VLSI Digital Signal ProcessingSystems-Keshab K.parhi

    Fundamental of Digital Design Floyed

    IEEE Standard of Binary Floating-Point

    Arithmetic" IEEEStandard754,IEEE Computer

    Society,1985 Design of Double Precision IEEE-754 Floating-Point

    Units by Michael Kennedy B.I.T -Griffith University

  • 8/9/2019 alu2 (2)

    26/26

    THANK YOU