I tried to solve linear systems with PETSC.
There are two linear systems.
We got a solution of one of them.
For another one, the PETSC solution does not converge.
Thus, I out put the eigenvalues of these two coefficient matrixes (in the
attachments)
in order to find if there is a difference between them.
Maybe, these eigenvalues can help me to know why that system doesn't converge.
However, I can not find a difference.
Do you have some suggestions about solving this non-convergence linear system?
Here is my code:
PetscReal rtol=(1.e-4)*pow(h,4.0); PetscReal stol=(1.e-3)*pow(h,4.0);
KSPSetTolerances(ksp, rtol, stol, PETSC_DEFAULT, PETSC_DEFAULT);
KSPGetPC(ksp, &pc); PCSetType(pc, PCGAMG); KSPSetFromOptions(ksp);
KSPSetUp(ksp); KSPSolve(ksp, load_vect, m_solut);
KSPGetConvergedReason(ksp, &reason);
Thanks,
Meng
------------------ Original ------------------
From: "Matthew Knepley";<[email protected]>;
Send time: Friday, Apr 25, 2014 5:53 AM
To: "Oo "<[email protected]>;
Cc: "petsc-users"<[email protected]>;
Subject: Re: [petsc-users] Convergence_Eigenvalues_k=3
On Thu, Apr 24, 2014 at 12:57 PM, Oo <[email protected]> wrote:
Hi,
For analysis the convergence of linear solver,
I meet a problem.
One is the list of Eigenvalues whose linear system which has a convergence
solution.
The other is the list of Eigenvalues whose linear system whose solution does
not converge (convergenceReason=-3).
These are just lists of numbers. It does not tell us anything about the
computation. What is the problem you are having?
Matt
Do you know what kind of method can be used to obtain a convergence solution
for our non-convergence case?
Thanks,
Meng
--
What most experimenters take for granted before they begin their experiments is
infinitely more interesting than any results to which their experiments lead.
-- Norbert Wiener
