[Trilinos-Users] [EXTERNAL] ml with Anasazi

[Thornquist, Heidi K]

Hi Sunghwan,

Anasazi's preconditioned eigensolvers, like LOBPCG and Block Davidson, can work directly with ML.  For block Krylov-Schur, the ML preconditioner can be used with a spectral transformation, like shift-invert, to get the smallest eigenpairs.

Your example showing how the ML preconditioner is constructed is missing one crucial element that allows the Epetra_Operator generated by ML to work with Anasazi, this operator needs to be wrapped in an inverse operator so that calls to the Epetra_Operator method Apply actually call the ML preconditioner's ApplyInverse method.  Without that, Anasazi will call the ML preconditioner's Apply method, which is not defined for this interface.  See below.

Teuchos::RCP < Anasazi::BasicEigenproblem<double,MV,OP> > MyProblem =Teuchos::rcp( new Anasazi::BasicEigenproblem<double,MV,OP>(matrix, eigenvectors) );
Teuchos::RCP<ML_Epetra::MultiLevelPreconditioner> prec;
<
…
construct ML preconditioner, prec, as below
…
>
Teuchos::RCP<Epetra_Operator> PrecOp;
PrecOp = Teuchos::rcp( new Epetra_InvOperator(prec.get()) );
MyProblem->setPrec(PrecOp);

Thanks,
Heidi

--

Heidi K. Thornquist
Electrical Models & Simulation
Sandia National Laboratories
Albuquerque, NM  87185-1323


From: SungHwan Choi <sunghwanchoi91 at gmail.com<mailto:sunghwanchoi91 at gmail.com>>
Date: Thursday, July 17, 2014 7:15 AM
To: "trilinos-users at software.sandia.gov<mailto:trilinos-users at software.sandia.gov>" <trilinos-users at software.sandia.gov<mailto:trilinos-users at software.sandia.gov>>
Subject: [EXTERNAL] [Trilinos-Users] ml with Anasazi

Dear all,
Hi, I am an Anasazi user. I want to use a multilevel preconditioner. As far as, I know AztecOO work well with ML package. However, I am not sure multigrid like preconditioner can be applied on Anasazi.
The article titled "A comparison of eigensolvers for large-scale 3D modal analysis using AMG-preconditioned iterative methods" said that ML package can work with Anasazi solver but I couldn't find how ML can work with Anasazi.

Thus, applying my pre-knowledge, I coded it but It does not work well. Would you help me?


            Teuchos::RCP < Anasazi::BasicEigenproblem<double,MV,OP> > MyProblem =Teuchos::rcp( new Anasazi::BasicEigenproblem<double,MV,OP>(matrix, eigenvectors) );
            Teuchos::RCP<Epetra_Operator> PrecOp;
            Teuchos::ParameterList MLList;
            ML_Epetra::SetDefaults("DD",MLList);
            MLList.set("smoother: pre or post", "post");
            MLList.set("PDE equations", 1);
            MLList.set("smoother: type","IFPACK");
            MLList.set("smoother: ifpack type", "ILU");
            MLList.set("smoother: ifpack overlap", 1);
            MLList.sublist("smoother: ifpack list").set("fact: level-of-fill", 5);
            ML_Epetra::MultiLevelPreconditioner* MLPrec =new ML_Epetra::MultiLevelPreconditioner(*matrix.get(), MLList, true);
            PrecOp=Teuchos::rcp( MLPrec);
            MyProblem->setPrec(PrecOp);


Sunghwan Choi
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://software.sandia.gov/pipermail/trilinos-users/attachments/20140717/5f4b749c/attachment.html>