RE: [Libmesh-users] direct solver

[Kirk, B. \(JSC-EG\) <ben...@na...>]

I'd be happy to help here too, but there is another option you could
consider.

PETSc says they can interface directly with UMFPACK.  In that case, you
simply need to build PETSc such that it uses UMFPACK internally and then
you should have access to (at leas a subset of) the UMFPACK sparse
directly.  The standard PETSc allows for "-pc_type direct" which does a
dense direct LU factorization and then uses it as the "preconditioner"
in an iterative method (which will then converge in 1 iteration).  This
only works on serial problems, mind you, but it looks like UMFPACK is
serial as well.

If you install PETSc with UMFPACK support I would guess you'd get
something like "-pc_type umfpack" which would then build the
"preconditioner" using UMFPACK's Unsymmetric sparse multifrontal method.


The only reason I mention it is because it may be the path of least
resistance, especially if UMFPACK is relegated to serial problems only.
Adding a new solver interface is fairly straightforward but time
consuming, and it is a lot of work to go through for a solver which will
only work on serial problems.

-Ben


-----Original Message-----
From: lib...@li...
[mailto:lib...@li...] On Behalf Of John
Peterson
Sent: Thursday, February 23, 2006 8:47 AM
To: Blome Mark
Cc: Lib...@li...
Subject: [Libmesh-users] direct solver

Blome  Mark writes:
 >
 > I would like to use a direct sparse matrix solver package within  >
libmesh (for example umfpack, pardiso or thelike).
 > Am I right that I have to implement an interface for the solver  >
package by deriving from the class LinearSolver (linear_solver.h)?
 > Are there any pitfalls I should be aware of ? How do I include the  >
solver package into the libmesh Makefile so it will get linked  >
correctly? Did anybody alreay implement a direct solver interface ?

You might get some hints from the Petsc and Laspack implementations of
the LinearSolver interface.  Some things you will probably need are:

1.) UMF_Matrix derived from SparseMatrix
2.) UMF_Vector derived from NumericVector
3.) UMF_Interface as you mentioned

This sounds like an interesting project.  If UMF is open source, and
small enough, we can distribute it directly in libmesh's contrib
directory.  Otherwise, you will need

4.) configure tests, etc.

I think we can help with 1.) -- 4.) if you do end up trying to do this.

-J


-------------------------------------------------------
This SF.Net email is sponsored by xPML, a groundbreaking scripting
language that extends applications into web and mobile media. Attend the
live webcast and join the prime developer group breaking into this new
coding territory!
http://sel.as-us.falkag.net/sel?cmd=3Dlnk&kid=3D110944&bid=3D241720&dat=3D=
121642
_______________________________________________
Libmesh-users mailing list
Lib...@li...
https://lists.sourceforge.net/lists/listinfo/libmesh-users