Data processing apparatus and method for multiplying floating point operands

Abstract

A data processing apparatus and method are provided for multiplying first and second normalised floating point operands in order to generate a result, each normalised floating point operand comprising a significand and an exponent. Exponent determination circuitry is used to compute a result exponent for a normalised version of the result, and rounding value generation circuitry then generates a rounding value by shifting a rounding constant in a first direction by a shift amount that is dependent on the result exponent. Partial product generation circuitry multiplies the significands of the first and second normalised floating point operands to generate the first and second partial products, and the first and second partial products are then added together, along with the rounding value, in order to generate a normalised result significand. Thereafter, the normalised result significand is shifted in a second direction opposite to the first direction, by the shift amount, in order to generate a rounded result significand. This provides a particularly efficient mechanism for multiplying floating point numbers, whilst correctly rounding the result in situations where the result is subnormal.

Description

technical field [0001] The present invention relates to a data processing device and method for multiplying floating point operands. Background technique [0002] The ability of data processing systems to manipulate floating point numbers is well known. Hardware mechanisms for performing floating-point multiplication typically use multipliers and multiplicands to generate many partial products. Various forms of carry-save adders are then used to reduce these partial products to two partial products. The final two partial products are added and then rounded to generate the final product result. [0003] The problem with the above is what to do with denormalized results. The result of denormalization is that the value has the minimum allowed value and the significand (also called the mantissa) is less than one, eg, 0.001101. The nature of denormalized numbers is well known to those skilled in the art. One way to handle denormalized numbers is to treat them as exceptions a...

AI Technical Summary

Problems solved by technology

However, while this approach simplifies handling of rounding issues within denormalized results, it requires an additional shift stage for partial products, which adversely affects performance

Images