One can use this techniques to optimize the multiplier inside the FP FMA unit. However this cannot be used to multiply two floating point numbers as FP arithmetic is not associative.