[petsc-users] Reuse symbolic factorization with petsc - mumps

[Thibault Faney]

Hi,

I am using the cray-petsc implementation on the NERSC system. Specifically,
I have implemented a runge kutta solver, and I need to solve a linear
system Ax = b several times per time step. I am using PETSC to interface
with the MUMPS solver for direct factorization.

The matrix A is constant over a whole time step, but changes at each time
step. What I do now is that at the beginning of each time step, I compute
the matrix A, use the MUMPS solver to compute a LU factorization of A, and
then use KSPSolve to solve each linear system during the same time step.
Here is an example of the succession of calls (inspired from example ex52.
c):

Once per time step:
call KSPSetOperators(KSP_A, A, A, DIFFERENT_NONZERO_PATTERN, ierr)
call KSPSetType(KSP_A, KSPPREONLY, ierr);
call KSPGetPC(KSP_A, PC_A, ierr)
call KSPSetTolerances(KSP_A, 1.d-20, PETSC_DEFAULT_DOUBLE_PRECISION,
PETSC_DEFAULT_DOUBLE_PRECISION, PETSC_DEFAULT_INTEGER, ierr)
call PCSetType(PC_A, PCLU, ierr)
call PCFactorSetMatSolverPackage(PC_A, MATSOLVERMUMPS, ierr)
call PCFactorSetUpMatSolverPackage(PC_A, ierr)
call PCFactorGetMatrix(PC_A, PC_A, ierr)
call MatMumpsSetIcntl(PC_A, 7, 5, ierr)
call PCSetup(PC_A, ierr)
call KSPSetUp(KSP_A, ierr)

Then several times per time step:
call KSPSolve( KSP_A, b, x, ierr)

This works fine.
However, I am able to precompute the maximal sparsity pattern of A before
the time iterations. Therefore, I would like to use this information by
precomputing the symbolic factorization before the time iterations start,
and only compute the numerical factorization at each time step. Is there a
way to use MUMPS in PETSC to do this ? I would appreciate any
help/reference on this topic.

Thank you,

Tibo
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20130314/48c3fc40/attachment.html>