Method and system for performing floating-point division using selected approximation values

In a floating-point unit having multiply and add functions, a system for performing floating-point division of a dividend by a divisor for producing a quotient having a mantissa, said system comprising:

• means for selecting a first table value from a table lookup using a linear approximation, wherein said first table value approximates two/divisor;
• means for selecting a second table value from said table lookup wherein said second table value approximates the reciprocal of the divisor squared and said first table value and said second table value operate as linear correction terms;
• means for determining an overall reciprocal approximation value by subtracting from said first table value the product of said divisor and said second table value;
• means for determining a table approximation error value as being 1.0 minus the product of the divisor and the overall reciprocal approximation value;
• means for determining a first approximation quotient by multiplying the dividend by the overall reciprocal approximation value;
• means for determining a second reciprocal approximation value by adding the overall reciprocal approximation value to the product of the overall reciprocal approximation value and the table approximation error value;
• means for generating a second table approximation error value by subtractinq from the dividend the product of the divisor and the first approximation quotient;
• means for determining a resultant quotient by adding said first approximation quotient to the product of said second reciprocal approximation value and said second table approximation error value.