I have found details about the non-restoring division algorithm, but from what I found it assumes that the dividend is greater than the divisor. Does this have to be true?
I am asking because I want to use the algorithm for designing a floating point division unit in hardware.