Re: [eigen] Using Cholmods LDLT factorization with eigen.

[ Thread Index | Date Index | More lists.tuxfamily.org/eigen Archives ]




On Sun, Sep 18, 2016 at 3:39 PM, Brad Bell <bradbell@xxxxxxxxxx> wrote:
I have a case where I need both the LDLT factorization and permutation corresponding to the factorization; i.e., I need eigen matrices corresponding to L, D, and P in
    P * A * P' = L * D * L'
In addition, I wish to ensure that factorization corresponds to the corresponding cholmod routine.

Looking at the specifications for
https://eigen.tuxfamily.org/dox/classEigen_1_1SimplicialLDLT.html
I see that  the Eigen simplicial LDLT routine can return the necessary information. But looking at
https://eigen.tuxfamily.org/dox/classEigen_1_1CholmodSimplicialLDLT.html
I do not see member functions for obtaining L, D, and P. On the other hand I do see the text
    'This simplicial variant is equivalent to Eigens's built-in SimplicialLDLT class.'

The word "equivalent" here is probably too strong. It does not mean that one is calling the other, just that the respective underlying algorithms follow the same logic and level of sophistication. Or put differently, that there should not be any advantage in calling CholmodSimplicialLDLT instead of the built-in SimplicialLDLT.
 
Are the L, D, and P matrices accessible from CholmodSimplicialLDLT ?

I guess those accessors could be added easily.. Patch welcome.
 
If not, under what conditions can I be certain that cholmod is being used for the factorization when I use the SimplicialLDLT class ?

As I said, SimplicialLDLT will never call cholmod.

cheers,
gael
 



Mail converted by MHonArc 2.6.19+ http://listengine.tuxfamily.org/