hello Garth
thanks, that works fine with the demo and but switches to a krylov
iterative solver because i only have umfpack via petsc. the krylov does
not converge on the dolfin-grade2, a LU is favored but petsc obviously
inhibits me to use periodic bc, what about another way to access another
LU solver? point dolfin to the umfpack (in the externalpackages folder
of my petsc) directly or use trilinois?
patrick
Garth N. Wells wrote:
>
>
> Patrick Riesen wrote:
>> Hello Andy
>>
>> Thanks for your reply! I am using dolfin 0.8.1 together with the
>> dolfin-grade2 library. I did not yet update to keep consistent with
>> the grade2lib. otherwise there might be a possibility to switch of the
>> petsc backend in the periodic demo or do i have to recompile dolfin to
>> not using petsc?
>>
>
> Try
>
> dolfin_set("linear algebra backend", "uBLAS");
>
> Garth
>
>> patrick
>>
>>
>> Andy Ray Terrel wrote:
>>> Hi Patrick,
>>>
>>> Are you using the development version of FEniCS or still 0.8.1 with
>>> dolfin-grade2? I'm trying to update my library at the moment and the
>>> periodic demo works without petsc on the dev version, so you might be
>>> able to work on it with that. It looks like somewhere in the code is
>>> using MatRestoreRow with a modified array instead of MatSetValues, but
>>> not having the dev environment setup with Petsc I haven't found the
>>> problem yet.
>>>
>>> -- Andy
>>>
>>> On Wed, Jan 21, 2009 at 7:56 AM, Patrick Riesen
>>> <[email protected]> wrote:
>>>> hello!
>>>>
>>>> i observed some strange behavior when i run the periodic poisson demo
>>>> (/pde/periodic). i use PETSC as backend and when it comes to applying
>>>> the bc's i get strange messages all over std output, here is a cut-out
>>>> from that:
>>>>
>>>>
>>>> [0]PETSC ERROR:
>>>> ------------------------------------------------------------------------
>>>>
>>>> [0]PETSC ERROR: MatRestoreRow() line 242 in src/mat/interface/matrix.c
>>>> [0]PETSC ERROR: --------------------- Error Message
>>>> ------------------------------------
>>>> [0]PETSC ERROR: Object is in wrong state!
>>>> [0]PETSC ERROR: Not for unassembled matrix!
>>>> [0]PETSC ERROR:
>>>> ------------------------------------------------------------------------
>>>>
>>>> [0]PETSC ERROR: Petsc Release Version 2.3.3, Patch 15, Tue Sep 23
>>>> 10:02:49 CDT 2008 HG revision: 31306062cd1a6f6a2496fccb4878f485c9b91760
>>>> [0]PETSC ERROR: See docs/changes/index.html for recent updates.
>>>> [0]PETSC ERROR: See docs/faq.html for hints about trouble shooting.
>>>> [0]PETSC ERROR: See docs/index.html for manual pages.
>>>> [0]PETSC ERROR:
>>>> ------------------------------------------------------------------------
>>>>
>>>> [0]PETSC ERROR: Unknown Name on a x64-ompi named jokull by priesen Wed
>>>> Jan 21 14:37:50 2009
>>>> [0]PETSC ERROR: Libraries linked from
>>>> /home/priesen/softw/petsc-2.3.3-p15/lib/x64-ompi
>>>> [0]PETSC ERROR: Configure run at Fri Nov 28 17:20:38 2008
>>>> [0]PETSC ERROR: Configure options --with-shared
>>>> --with-mpi-dir=/home/priesen/num/openmpi
>>>> --with-blas-lapack-dir=/usr/pack/atlas-3.7.24-ds --download-ml
>>>> --download-hypre --download-umfpack --download-mumps
>>>> --download-scalapack --download-blacs
>>>> [0]PETSC ERROR:
>>>>
>>>> and and and...
>>>>
>>>> this stuff occurs exactly between the "Applying periodic boundary
>>>> conditions" and the "Applying periodic boundary conditions (finished)"
>>>> and then the program finishes successfully, though i do not trust the
>>>> solution.
>>>> Do you ever had something similar to this?
>>>> Is there a problem that i can not use periodic boundary conditions
>>>> together with PETSC or did i possibly miss-configured either PETSC or
>>>> dolfin?
>>>>
>>>>
>>>> Thanks for your help,
>>>>
>>>> Patrick
>>>>
>>>>
>>>>
>>>> _______________________________________________
>>>> DOLFIN-dev mailing list
>>>> [email protected]
>>>> http://www.fenics.org/mailman/listinfo/dolfin-dev
>>>>
>>
>> _______________________________________________
>> DOLFIN-dev mailing list
>> [email protected]
>> http://www.fenics.org/mailman/listinfo/dolfin-dev
>
>
_______________________________________________
DOLFIN-dev mailing list
[email protected]
http://www.fenics.org/mailman/listinfo/dolfin-dev
