[petsc-users] Least Squares options in Petsc

Barry Smith bsmith at mcs.anl.gov
Thu Dec 22 10:15:23 CST 2011


On Dec 21, 2011, at 9:01 PM, Mengda Wu wrote:

> Hi all,
> 
>    I am trying to solve an overdetermined system using Petsc. I am using un-preconditioned LSQR but it seems that it converges slowly. I am woer ndering what preconditioner may improve the convergence. Or
> are there any other ksp I can use?

   What are the typical sizes of the matrix?

   Note that when running with preconditioning you need to call KSPSetOperators(ksp,A,B,DIFFERENT_NONZERO_PATTERN) where B is A'*A which can be obtained with 
MatMatMultTranspose(). (in petsc-3.2). 

   Then you can try -pc_type lu   -pc_type ilu  or any other preconditioner.

   Barry


   
> 
> Thanks,
> Mengda



More information about the petsc-users mailing list