Simplifying the rounding for Newton-Raphson algorithm with parallel remainder

This work presents a variable latency rounding method based on the calculation in parallel of a remainder estimation for Newton-Raphson algorithm. The value obtained is the remainder of the algorithm result without any transformation. Because of this a new rounding method based on this value has to be proposed. This method avoid the traditional remainder calculation in 91% of the cases for division and 88% of the cases for square root.