Dismiss
InnovationQ will be updated on Sunday, Oct. 22, from 10am ET - noon. You may experience brief service interruptions during that time.
Browse Prior Art Database

Ensuring Absolute Difference Result in Floating Point Adder

IP.com Disclosure Number: IPCOM000105621D
Original Publication Date: 1993-Aug-01
Included in the Prior Art Database: 2005-Mar-20
Document File: 2 page(s) / 62K

Publishing Venue

IBM

Related People

Atkins, MG: AUTHOR

Abstract

Most prediction circuits in a floating point adder are complicated by the difference operation. This is because the use of sign magnitude numbers results in two possible paths that could be taken by the fractions of the operands. These paths are dependent upon whether the result is positive or negative. Therefore, any predication circuit must essentially take into account three possible paths; the add, and the tow difference paths. The aim of the proposed methodology is to eliminate one of the possible difference paths, and manipulate the sign to cover the other path.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 53% of the total text.

Ensuring Absolute Difference Result in Floating Point Adder

      Most prediction circuits in a floating point adder are
complicated by the difference operation.  This is because the use of
sign magnitude numbers results in two possible paths that could be
taken by the fractions of the operands.  These paths are dependent
upon whether the result is positive or negative.  Therefore, any
predication circuit must essentially take into account three possible
paths; the add, and the tow difference paths.  The aim of the
proposed methodology is to eliminate one of the possible difference
paths, and manipulate the sign to cover the other path.

      By forcing the smaller of the two operands to be used as the
subtrahend, the intermediate result from the adder will be the
absolute difference between the two numbers.  The sign can be
calculated in parallel with the intermediate adder result.

      The proposed methodology's data flow for the fractions in a
floating point adder is illustrated in the Figure.

      The following steps are taken to ensure that the smallest
number is introduced to the adder as the subtrahend:

1.  Determine what type of operation is to be performed

    a.  Sum

        1)  Add op-code with operands of the same sign.
        2)  Subtract op-code with operands of differing signs.

    b.  Difference

        1)  Add op-code with operands of the same sign.
        2)  Subtract op-code with operands of the same sign.

2.  Compare the two operands

    a.  Compare the exponents
    b.  Compare the fractions

3.    Unnormalize the operand...