需要一个更好的复杂划分算法

Question

我正在将Fortran转换为C＃，并且我已经使用双精度实现了我自己的Complex结构。在Fortran我有一个复杂的计算

c = a/b
a = (-6.756993318630313D-178,2.237434940548793D-178)    COMPLEX(8)
b = (1.381489403189045D-175,-3.000542776128005D-175)    COMPLEX(8)
c = (-1.470726933303390D-003,-1.574781630470373D-003)   COMPLEX(8) 

当我尝试进行此计算时，我的简单C＃实现会给出NaN。

public static Complex operator /(Complex complex1, Complex complex2)
{
    var denominator = complex2.Real * complex2.Real + complex2.Imaginary * complex2.Imaginary;
    return new Complex((complex1.Real * complex2.Real + complex1.Imaginary * complex2.Imaginary) / denominator,
            (-complex1.Real * complex2.Imaginary + complex2.Real * complex1.Imaginary) / denominator);
}

考虑到算法，获得NaN并不奇怪。编写英特尔Fortran编译器的人显然比我更聪明。任何人都可以提供与Fortran中使用的算法一样好的算法吗？