[petsc-users] Use block Jacobi preconditioner with SNES

[Jed Brown]

Ali Reza Khaz'ali <arkhazali at cc.iut.ac.ir> writes:

>> It looks like you have forgotten to destroy all of your objects.  If you
>> clean up after yourself, you'll also see the totaly memory used by each
>> of the PETSc classes that you use.
>
> That is because of PetscLogView was placed before the destroy functions. 
> I moved it after them, and here is the new log: (additionally, this is 
> the log of one time step only, so all of the objects won't be destroyed 
> before all time steps have been finished)

You can just use the run-time option and let PetscFinalize handle this
output.

> Memory usage is given in bytes:
>
> Object Type          Creations   Destructions     Memory Descendants' Mem.
> Reports information only for process 0.
>
> --- Event Stage 0: Main Stage
>
>                  SNES     1              1         1372     0.
>                DMSNES     1              1          672     0.
>        SNESLineSearch     1              1         1000     0.
>                Vector    20             20     17313120     0.
>                Matrix     1              1    105843212     0.

As estimated previously, PETSc uses a little over 100 MB in total.

>      Distributed Mesh     2              2         9504     0.
>     Star Forest Graph     4              4         3200     0.
>       Discrete System     2              2         1856     0.
>         Krylov Solver     1              1        18416     0.
>       DMKSP interface     1              1          656     0.
>        Preconditioner     1              1          832     0.
>                Viewer     1              0            0     0.