PETSc problem during mpirun

[Jan Ludvig Vinningland]

Hi,

when running the attached script (a modified version 

of an example from the nsbench-project) I get the following error:

*** -------------------------------------------------------------------------

*** Error:   Unable to successfully call PETSc function 'KSPSolve'.

*** Reason:  PETSc error code is: 92.

*** Where:   This error was encountered inside /tmp/dolfin20160311-76839-1f8s8p1/dolfin-1.6.0/dolfin/la/PETScKrylovSolver.cpp.

*** Process: 0

*** 

*** DOLFIN version: 1.6.0

*** Git changeset:  

*** -------------------------------------------------------------------------

The script runs perfect without mpi (or mpi with one process). The error message 

is from a OSX run, but I also get an similar error under Ubuntu with code 56 instead of 92.

Any suggestions on how to resolve this is very welcome...

Best regards,

Jan Ludvig Vinningland

[Garth N. Wells]

Remove the argument "ilu" to the solver. PETSc does not have a parallel ILU preconditioner.

Garth

--
You received this message because you are subscribed to the Google Groups "fenics-support" group.
To unsubscribe from this group and stop receiving emails from it, send an email to fenics-suppor...@googlegroups.com.
To post to this group, send email to fenics-...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/fenics-support/33e153a8-8f57-430b-9491-ef018751d5e5%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Anders Logg]

I has the same problem recently. Shouldn't we check for this?

--
Anders

fre 18 mars 2016 kl. 17:11 skrev Garth N. Wells <gn...@cam.ac.uk>:

Remove the argument "ilu" to the solver. PETSc does not have a parallel ILU preconditioner.

Garth

On 18 March 2016 at 15:40, Jan Ludvig Vinningland <janl...@gmail.com> wrote:

--

You received this message because you are subscribed to the Google Groups "fenics-support" group.
To unsubscribe from this group and stop receiving emails from it, send an email to fenics-suppor...@googlegroups.com.
To post to this group, send email to fenics-...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/fenics-support/33e153a8-8f57-430b-9491-ef018751d5e5%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to the Google Groups "fenics-support" group.
To unsubscribe from this group and stop receiving emails from it, send an email to fenics-suppor...@googlegroups.com.
To post to this group, send email to fenics-...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/fenics-support/CAEymZaebXbvtdHi4745QkYZesM2XExO6epo52FK--CaGLZg_uQ%40mail.gmail.com.

[Jan Ludvig Vinningland]

That was the trick! Thanks!

Jan Ludvig

[Garth N. Wells]

On 18 March 2016 at 16:41, Anders Logg <ander...@gmail.com> wrote:

I has the same problem recently. Shouldn't we check for this?

We could reasonably check if we reduce the number of preconditioners that we support via name strings. It we have too many strings, we'll inevitably get inconsistencies with PETSc. 

Garth

 

To view this discussion on the web visit https://groups.google.com/d/msgid/fenics-support/CA%2BZarhP34ioVtvfjVSU5AQq0wbczLueuWzWUR2fLsLQb20LfEw%40mail.gmail.com.