[petsc-users] Gauss-Seidel

Jed Brown jedbrown at mcs.anl.gov
Mon Apr 23 20:42:45 CDT 2012


On Mon, Apr 23, 2012 at 20:40, tribur <burckhardt at itis.ethz.ch> wrote:

> Hi Jed,
>
> Thanks for the ultra fast response.
>
> Apart from the nonzero initial guess, is -ksp_type preonly -pc_type sor
> -pc_sor_its 10 right?
>

Yes.

You can use -ksp_initial_guess_nonzero -ksp_type richardson


>
> Am 24.04.2012 03:29, schrieb Jed Brown:
>
>> On Mon, Apr 23, 2012 at 20:27, tribur <burckhardt at itis.ethz.ch [1]>
>>
>> wrote:
>>
>>  I would like to do 10 iterations of a point-by-point Gauss-Seidel
>>> procedure for the purpose of testing.
>>>
>>> Is it correct if I type "mpiexec -n 1 ... -ksp_type preonly
>>> -pc_type sor -pc_sor_its 10"?
>>>
>>> How can I define the initial guess nonzero?
>>>
>>
>> With a nonzero initial guess, you are doing Richardson (not preonly).
>> You can set the nonzero initial guess as usual
>> (KSPSetInitialGuessNonzero()).
>>
>> Links:
>> ------
>> [1] mailto:burckhardt at itis.ethz.ch
>>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20120423/32380d26/attachment.htm>


More information about the petsc-users mailing list