Topic: what proper iterative petsc solver for solving an indefinite system?
My system is indefinite as follows since Lagrange multipliers are employed.
[K C
C 0] δx = δr
I am using iterative solver with Schur complement preconditioner. The options for petsc is set as "-ksp_type fgmres -pc_type fieldsplit -pc_fieldsplit_detect_saddle_point -pc_fieldsplit_type schur -fieldsplit_0_ksp_type gmres -fieldsplit_0_pc_type gamg -ksp_view".
I make sure that zeros are present on the diagonals at the last. However, I got the following error:
No support for this operation for this object type!
This matrix type does not have a find zero diagonals defined!
What is the possible reason for this error? Did I issue the wrong commands for the petsc to work successfully? I searched online for solution. One potentially useful link is preconditioners in petsc. Unfortunately, those contents donot help me out.
https://scicomp.stackexchange.com/quest … ystems-sho
Anyone can share some experience when solving a class of saddle point problems?