KSPComputeEigenvaluesExplicitly#

Computes all of the eigenvalues of the preconditioned operator using LAPACK.

Synopsis#

#include "petscksp.h" 
PetscErrorCode KSPComputeEigenvaluesExplicitly(KSP ksp, PetscInt nmax, PetscReal r[], PetscReal c[])

Collective

Input Parameters#

ksp - iterative context obtained from KSPCreate()
n - size of arrays r and c

Output Parameters#

r - real part of computed eigenvalues, provided by user with a dimension at least of n
c - complex part of computed eigenvalues, provided by user with a dimension at least of n

Notes#

This approach is very slow but will generally provide accurate eigenvalue estimates. This routine explicitly forms a dense matrix representing the preconditioned operator, and thus will run only for relatively small problems, say n < 500.

Many users may just want to use the monitoring routine KSPMonitorSingularValue() (which can be set with option -ksp_monitor_singular_value) to print the singular values at each iteration of the linear solve.

The preconditioner operator, rhs vector, solution vectors should be set before this routine is called. i.e use KSPSetOperators(), KSPSolve()