Fused Multiply Add

In many cases, an addition is required right after a multiplication. Thus, a dedicate fused instruction is provided to finish multiplication and addition c=a*b+c, where a, b, and c are floating-point numbers, with one final rounding only. The instruction is named fused multiply-add (FMA) or fused multiply-accumulate.

FMA was introduced in 1990 on the IBM RS/6000 processor. The instruction allows for faster and, in general, more accurate dot products, matrix multiplications, and polynomial evaluations[1].

References

[1] Muller JM. et al. (2010) The Fused Multiply-Add Instruction. In: Handbook of Floating-Point Arithmetic. Birkhäuser Boston [cache]