unexpected results with intel primitive (when dividing by 1)

Question

I'm using Intel Integrated Performance Primitives (v2019, 64bit) and I came across unexpected results.

When I run the following code

    std::cout << std::setprecision(20);
    const int N = 8;
    const std::vector<double> one(N, 1.0);
    std::vector<double> d(N,0.014058305571221497);
    
    std::cout << "before division by 1 : " << d[0] << std::endl;
    ippsDiv_64f_I(one.data(), d.data(), N);
    std::cout << "after division by 1 : " << d[0] << std::endl;

I get the following output :

before division by 1 : 0.014058305571221497293 
after division by 1 : 0.014058305571221500763

1/ How dividing by 1 can change the value ?

2/ For any value of N between 1 and 7, dividing by 1 has no effect. Only for N>7 I get this. How can it be that the result of a division of doubles depend on the size of my array ?

Could it be a bug ?