[CIG-LONG] 2008 Benchmarks files

Walter Landry walter at geodynamics.org
Thu May 28 13:54:45 PDT 2009


Leonardo  Cruz <leocruz at stanford.edu> wrote:
> Hello,
> 
> Has anyone ran the benchmark 2008 input files in the version 1.4.1?
> When I run them (i.e., geomod_2008_unstable.xml), it stops after 41
> time steps due to non convergence error . See error below. I got
> similar errors with the other files.
> 
> Any ideas?

Are you using the direct solver?  You can use it by adding

  -mat_type aijmumps -ksp_type preonly -pc_type lu

on the command line.  The default solver can have problems like this
when there is yielding.

Cheers,
Walter Landry
walter at geodynamics.org


More information about the CIG-LONG mailing list