Scribd Logo
Upload

Welcome to Scribd!

  • 60 views

    Floating Point Numbers - Representation & Arithmetic: Dr. Arunachalam V Associate Professor, SENSE

    Uploaded by

    Anil Kumar Gorantala
    Floating point numbers represent values as the product of a signed mantissa and a power of the base, with the exponent expressed as an offset from a bias. Common formats include half precision which uses 1 sign bit, 5 exponent bits, and 10 mantissa bits. Floating point addition first aligns the mantissas by shifting, then adds the significands and exponents while tracking special values and exception flags.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd

    Floating Point Numbers - Representation & Arithmetic: Dr. Arunachalam V Associate Professor, SENSE

    Uploaded by

    Anil Kumar Gorantala
    60 views14 pages
    Floating point numbers represent values as the product of a signed mantissa and a power of the base, with the exponent expressed as an offset from a bias. Common formats include half precision which uses 1 sign bit, 5 exponent bits, and 10 mantissa bits. Floating point addition first aligns the mantissas by shifting, then adds the significands and exponents while tracking special values and exception flags.

    Original Title

    15_Floating_point_number_representation_add_sub.pdf

    Copyright

    © © All Rights Reserved

    Available Formats

    PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    Floating point numbers represent values as the product of a signed mantissa and a power of the base, with the exponent expressed as an offset from a bias. Common formats include half precision which uses 1 sign bit, 5 exponent bits, and 10 mantissa bits. Floating point addition first aligns the mantissas by shifting, then adds the significands and exponents while tracking special values and exception flags.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    60 views14 pages

    Floating Point Numbers - Representation & Arithmetic: Dr. Arunachalam V Associate Professor, SENSE

    Uploaded by

    Anil Kumar Gorantala
    Floating point numbers represent values as the product of a signed mantissa and a power of the base, with the exponent expressed as an offset from a bias. Common formats include half precision which uses 1 sign bit, 5 exponent bits, and 10 mantissa bits. Floating point addition first aligns the mantissas by shifting, then adds the significands and exponents while tracking special values and exception flags.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    of 14

    Floating point numbers –

    representation & Arithmetic


    Dr. Arunachalam V
    Associate Professor, SENSE
    Introduction

    =± ×
    A floating point number has four components:
    1. The sign, s (+ by 0/- by 1), usually it represents sign of mantissa.
    2. The exponent base b (usually 2 for digital).
    3. The significand or Mantissa, S or M. Usually normalized by shifting, so that
    the MSB becomes nonzero. With base 2, the fixed leading 1 can be removed
    to save one bit; this bit is called as “ hidden 1”.
    4. The exponent, e (usually biased representation, = + , =
    2 − 1 for k bit exponent)
    5. For example, k = 4 bits, =2 − 1 = 8 − 1 = 7 ; e = + 2, e’ = 2+7 =9
    and e = - 2, e’ = - 2 + 7 = 5.
    Subranges and special values

    = × = ×

    • Range [-max, max] increases and precision decreases as the base b increase.
    • In most practical cases, b = 2.
    • Devoting more bits to exponent part widens the range but reduces the
    precision.
    • In most practical cases, exponent uses lesser number of bits than the
    mantissa.
    ANSI / IEEE Std 754 - 1985
    Half precision
    • In computing, half precision is a binary floating-point computer number
    format that occupies 16 bits (two bytes in modern computers) in computer
    memory.
    • In the IEEE 754-2008 standard, the 16-bit base 2 format is referred to as
    binary16. It is intended for storage of floating-point values in applications
    where higher precision is not essential for performing arithmetic computations.
    • Sign bit: 1 bit
    • Exponent width: 5 bits
    • Significand precision: 11 bits (10 explicitly stored)
    Addition algorithm
    1. Subtract exponents (d =e1 - e2)
    2. Align the mantissas :
    a) Shift right d positions the mantissa of the operand with the smallest exponent.
    b) Select the largest exponent as exponent of the result.
    3. Add (sub) significands and produce sign of result. The effective operation
    (EOP):

    Floating point operation Sign of the operands EOP


    Add Equal Add
    Add Different Sub
    Sub Equal Sub
    Sub Different Add
    Addition algorithm
    4. Normalize:
    a) Effective operation addition: there is no over flow of the significand., i.e.
    result is already normalized: no action is needed
    b) Effective operation addition: there might be an over flow of the
    significand. Normalize:
    i. Shift right the significand one position
    ii. Increment by one the exponent
    c) Effective operation subtraction: the result might have leading zeros.
    Normalize:
    i. Shift left the significand by a number of positions corresponding to the number of
    leading zeros.
    ii. Decrement the exponent by the number of leading zeros.
    Addition algorithm
    5. Round. According to the specified mode. Might require an addition. If
    over flow occurs, normalize by a right shift and increment the exponent.
    6. Determine exception flags and special values : exponent over flow
    (special value ± infinity), exponent under flow (special value gradual
    under- flow), inexact, and the special value zero.
    Multiplication algorithm
    Input: x and y - normalized operands represented by ( : : )&( : : )
    1. Multiply significands ( = ∗× ∗ ).

    2. Add exponents = + − ,
    3. Determine sign ( = ⨁ )
    4. Normalize and update exponent
    5. Round (to nearest , to zero and to ±∞)
    6. Determine exception Flags and special values
    Reference
    1. Chapters 17 of Behrooz Parhami, “Computer Arithmetic: Algorithms and
    Hardware Design”, (2/e) Oxford University Press 2015.
    Next Class

    FLOATING POINT ARITHMETIC PROBLEM


    SOLVING

    You might also like