Accurate floating-point calculation method and device

摘要

A method is provided to narrow down the exponent range throughout most part of the division and square root calculations, to make both software assistance and precision extension unnecessary. The method adjusts the exponent at the end of the calculation to reach IEEE-754 results.

主权项

1. A computer implemented method for providing a floating point operation; wherein a quotient is provided by dividing an original dividend value by an original divisor value, the computer implemented method comprises:
obtaining utilizing a first hardware element a new dividend by reducing an exponent range of the original dividend value and a new divisor by reducing an exponent range of the original divisor value; obtaining utilizing a second hardware element an approximate reciprocal value of the new divisor value; providing an approximate quotient value and one or more error values based upon the new dividend and the approximate reciprocal value of the new divisor value; utilizing the original dividend value and the original divisor value to generate an adjustment value by a third hardware element; and utilizing a fourth hardware element to provide an accurate quotient based upon the one or more error values and the approximate quotient value.