Function ndtr_impl 

pub fn ndtr_impl<P, E, V>(x: V) -> V
where
    P: Policy,
    E: FloatElement,
    V: FloatVector<Element = E> + SpecializedSpecialMath<E>,

ndtr(x) = erfc(-x/sqrt 2)/2, the standard normal CDF.

The erfc kernel handles the reflection to the right side itself, so this is the whole function. The 1/sqrt 2 scaling costs one rounding in the argument, which the tail amplifies by x^2. That is the function’s own condition number, not the kernel’s.