Function phase_half_x2 

pub fn phase_half_x2<P, E, V>(x: V) -> V
where
    E: FloatElement,
    V: FloatVector<Element = E>,
    P: Policy,

x^2/2 mod 2, the argument for sincos_pi, to full precision for every x whose square is finite.

See the module docs for why the two words are reduced separately. Below Average the residual is dropped entirely and this is the naive x*x/2, which is accurate only while x^2 is exact.