[petsc-users] GMRES pc right fails while pc left succeeds

[Barry Smith]

   In theory if you use petsc-3.3 (release any day) switching to use the PCFIELDSPLIT solver (with appropriate options) should be able to reproduce your algorithm (I think) with very little coding on your part. This would make it easier for us to help you understand the issue since we know exactly what PCFIELDSPLIT does and how it is suppose to work and the -ksp_view option (or -snes_view) tells exactly the algorithm being used.

    Barry

On Jun 4, 2012, at 9:47 AM, Klaij, Christiaan wrote:

>>> Both approximations are in the preconditioner.
>> 
>> 
>> I'm afraid I don't have an explanation based on the information you have
>> sent. If you can set up a test case that shows the issue, I'll investigate
>> further.
> 
> Thanks for offering, unfortunately I don't have a small testcase that
> shows the same behaviour. It only happens in the odd industrial case,
> convergence is fine for most similar cases and perfect for small
> academic cases.
> 
>> 
>> I would think about null space issues. For example, if the Schur complement
>> has constants in its null space, but the right hand side is not consistent
>> and/or the KSP was not informed of the null space.
> 
> Definitely not the constant, but there could be a near null space...
> 
> 
> dr. ir. Christiaan Klaij
> CFD Researcher
> Research & Development
> E mailto:C.Klaij at marin.nl
> T +31 317 49 33 44
> 
> MARIN
> 2, Haagsteeg, P.O. Box 28, 6700 AA Wageningen, The Netherlands
> T +31 317 49 39 11, F +31 317 49 32 45, I www.marin.nl
>