num::lu_result

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

template <std::floating_point T>
struct lu_result;

The factorization \(PA = LU\), in precision T.

Both factors share the matrix LU, the layout LAPACK getrf returns: \(U\) on and above the diagonal, \(L\) strictly below it. \(L\) has a unit diagonal, which is not stored. lower() and upper() copy the two factors out.

\(P\) is stored as the row swaps that built it: step k exchanged rows k and swaps[k]. swaps is empty when the factorization did not pivot, so \(P = I\).

Members

member	description
mat<T> LU
array<idx> swaps
bool singular = false	True when a pivot fell below tolerance.
idx size() const