Hi all,
I need to solve a large scale potentially ill-conditioned (I can overcome this with scaling) LINEAR system. Previously I was using parFEAP8.2--PETSC2.3.3. Through linear solver schemes, I had to have PCLU for the preconditioner (PCBJACOBI, etc does not give accurate results) and convergence was achieved using most of Krylov solvers, nonetheless the computational time is expensive for the MatConvert operation as a necessity for PCLU operation. This issue has been resolved in later PETSC versions. I am trying to install a compatible parFEAP platform now.
Could you comment on the state of parFEAPs being released and the compatibility with PETSC 3.3 and above?
Could you also comment on an optimal way of solving such systems in PETSC?
Regards,
Reza
