num::lu_no_pivot_solve_transpose_multiple

Defined in header <linear/factorization/lu.hpp>

template <std::floating_point T>
void lu_no_pivot_solve_transpose_multiple(T *X, const T *LU, idx n,
                                          idx columns) noexcept;

Solves \(A^T x = b\) for several right-hand sides from an unpivoted LU factor.

For one right-hand side, \(U^T\) and \(L^T\) are applied a row of the factor at a time: once \(x_i\) is final, row \(i\) of the factor times \(x_i\) is subtracted from the entries still pending. That reads the factor with stride 1 and has no dependence between iterations, where the column-oriented form reads it with stride \(n\).