A floating-point fused add-subtract unit

Abstract

A floating-point fused add-subtract unit is described that performs simultaneous floating-point add and subtract operations on a common pair of single-precision data in about the same time that it takes to perform a single addition with a conventional floating-point adder. When placed and routed in a 45 nm process, the fused add-subtract unit is only about 40% larger than a conventional floating-point adder.