Hello all,

I was wondering if something is possible in an easy way.
I know that I can do it for LLT which I will describe below but not sure how to do it for LDLT
Say that I have a matrix that I decomposed in place.

Code: Select all

Eigen::MatrixXd N;
Eigen::LLT<Eigen::Ref<Eigen::MatrixXd>, Eigen::Lower> in_place_llt(N);


If my LLT object goes out of scope I can use the decomposed matrix to solve by

Code: Select all

typedef Eigen::internal::LLT_Traits<Eigen::MatrixXd, Eigen::Lower> TraitsType;
TraitsType::getL(N).solveInPlace(u);
TraitsType::getU(N).solveInPlace(u);


I am not sure if this is the best way but it works.
Otherwise I am not sure how I can set my current matrix in an LLT object such that the decomposition is initialized and m_matrix is set.

In LDLT things are a bit more complicated from what I see in the solve_impl.
Is there a way to achieve what I am describing for LDLT ?

I guess you already followed this tutorial: https://eigen.tuxfamily.org/dox-devel/g ... ition.html

In LDLT you really have to keep the LDLT object alive because it also stores the permutation matrix. For LLT, it is currently not possible to initialize it with an already existing factored matrix. We would need something like:

LLT<Ref<MatrixXd> > lltwrapper(A, AlreadyFactored);

Thanks ggael!