double fmin(double x, double y);
fmin determines the minimum of x and y.
fmin (NaN, y) is y. fmin (x, NaN) is x.