
Square Root
381
In the multiplicative normalization method, instead of performing this multi-
plication at the end, it is possible to initialize q [0] = x and obtain Q instead
of R.
In floating-point units, the quotient has to be correctly rounded. The most
prevalent method to do this rounding (see Chapter 8) computes the remainder
produced by the approximation and performs a correction step. However, unlike
in digit recurrence methods, in these iterative methods, the remainder is not
obtained directly. Instead, at the end of operation, the product of the computed
quotient and the divisor is formed and subtracted ...