num::seq::lu

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

template <std::floating_point T>
lu_result<T> lu(const mat<T> &A);

Blocked partial-pivoting LU through num::lu_factor_blocked: panel factorization, then a trsm and a gemm per panel, so the bulk of the work runs at gemm speed.

A pivot below singular_tol marks the result singular; the factorization still completes so the caller can inspect it.