[petsc-users] Fwd: A quick question on 'un-symmetric graph'

[Jed Brown]

Mark, this is a reasonably reduced test case.

mpiexec -n 3 ./ex45 -da_grid_x 19 -da_grid_y 13 -da_grid_z 23 -pc_type gamg
-pc_gamg_agg_nsmooths 1 -ksp_monitor

This is a symmetric graph, but not a symmetric matrix, though the asymmetry
is only due to boundary conditions (boundary rows have only a diagonal
nonzero; the other zeros are still stored explicitly). Do you have a
preference about how this is handled? (I don't think we should need extra
options to handle this extremely common case.)

On Wed, Jan 2, 2013 at 5:27 PM, Zhenglun (Alan) Wei
<zhenglun.wei at gmail.com>wrote:

>  Hello all,
>     I did some other tests. Now, it is narrow down to this situation:
>     The domain and grid size is: 163*81*41 with 0.0125 = dx=dy=dz. The
> boundary condition is still the same.
>     The code runs well with 4 processes while the same problem occurs when
> I was trying to use more than 4 processes (e.g. 6 or 8 processes).
>
>      Hope this helps to detect the problem.
>
> thanks,
> Alan
>
>
>
> -------- Original Message --------  Subject: A quick question on
> 'un-symmetric graph'  Date: Wed, 02 Jan 2013 15:41:47 -0600  From: Zhenglun
> (Alan) Wei <zhenglun.wei at gmail.com> <zhenglun.wei at gmail.com>  To: PETSc
> users list <petsc-users at mcs.anl.gov> <petsc-users at mcs.anl.gov>
>
> Dear folks,
>      Here I came across a problem.
>
> [0]PETSC ERROR: Petsc has generated inconsistent data!
> [0]PETSC ERROR: Have un-symmetric graph (apparently). Use
> '-pc_gamg_sym_graph true' to symetrize the graph or '-pc_gamg_threshold
> 0.0' if the matrix is structurally symmetric.!
>
>     My code basically uses PETSc /src/ksp/ksp/example/tutorial/ex45.c
> to solve the Poisson equation with the Dirichlet BC in x-direction and
> the Periodic BC in y- and z- direction.
>     The executable file is:
>
> mpiexec -f $PBS_NODEFILE -np 32 ./ex45 -pc_type gamg -ksp_type cg
> -pc_gamg_agg_nsmooths 1 -mg_levels_ksp_max_it 1 -mg_levels_ksp_type
> richardson -ksp_rtol 1.0e-7
>
>
>     There is no problem for my code when I use small computational
> domain (83*41*21 with single core and even 163*81*41 with 4 processes).
> However, when I increase the domain size (323*161*81 with 32 processes),
> the error comes up. I wonder what's the possible reason of this kind of
> problem. Do you need any of my output in order to do some further
> inspection? or I just need to blindly add '-pc_gamg_sym_graph true' in
> PETSc option?
>
> thanks,
> Alan
>
>
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20130105/59782800/attachment.html>