Speeding up an overrelaxation method of division in Radix-2n machine

For normalized floating point division, digital computers can take advantage of a division process that uses an iterative multiplying operation instead of repeated subtractions. An improvement of this division process by using accelerating constants in the overrelaxation has previously been proposed. Multiplication by a chosen accelerating constant accelerates the process of generating accurate digits of a quotient in division. We propose a further improvement by generalizing the accelerating constants in the overrelaxation method. Two benefits resulting from this improvement promise to yield faster division in digital computers.

View this article in the ACM Digital Library.