Function compute_kkt_jacobian 

pub fn compute_kkt_jacobian(
    q: &[Vec<f64>],
    g: &[Vec<f64>],
    a: &[Vec<f64>],
    x: &[f64],
    lam: &[f64],
    nu: &[f64],
) -> Vec<Vec<f64>>

Build the KKT Jacobian matrix ∂F/∂z for the QP:

min ½ x’Qx + c’x s.t. Gx ≤ h (m inequalities) Ax = b (p equalities)

The KKT conditions (with slacks absorbed into complementarity) are:

F₁ = Qx + c + G’diag(λ) + A’ν = 0 (stationarity, n eqs) F₂ = diag(λ)(Gx - h) = 0 (complementarity, m eqs) F₃ = Ax - b = 0 (primal equality, p eqs)

The Jacobian w.r.t. z = (x, λ, ν) is the (n+m+p) × (n+m+p) matrix:

┌ Q G’diag(λ)? A’ ┐ │ diag(λ)G diag(Gx-h) 0 │ └ A 0 0 ┘

For the complementarity row the correct linearisation is: ∂F₂/∂x = diag(λ) G ∂F₂/∂λ = diag(Gx - h) (= diag(s) where s = h - Gx ≥ 0 at optimum)

Arguments

• q – n*n cost matrix (row-major Vec<Vec<f64>>).
• g – m×n inequality constraint matrix.
• a – p×n equality constraint matrix.
• x – optimal primal (length n).
• lam – optimal inequality duals (length m).
• nu – optimal equality duals (length p).

Returns

The full KKT Jacobian as a flat (n+m+p) × (n+m+p) row-major matrix.