[CIG-LONG] problem with gale on a cluster

[Walter Landry]

js at cp.dias.ie wrote:
> 
> 
> Hello,
> 	firstly, I would like to thank you for your efforts in
> releasing and documenting Gale.
> 
> 	I recently downloaded both the stable (121) and svn versions
> and successfully ran the extension.xml model included in stable on
> both versions with a single processor.
> 
> 	I have just tried to run the same model on four processors,
> using gale_svn and it crashed about half way through with the following
> message,

Try this input file with gale_svn.  I have done a number of runs on
the cluster here, so it should work in parallel.  You may have to use
a direct solver.  If you did not build petsc with mumps, you can do it
by adding

  --download-mumps=yes --download-blacs=yes --download-scalapack=yes --with-blacs=yes --with-scalapack=yes

when configuring petsc.  Note that Gale is only tested with Petsc
2.3.2.  When running Gale, add

  -mat_type aijmumps -ksp_type preonly -pc_type lu

to the command line.  For example, something like

  ./gale_svn extension.xml -mat_type aijmumps -ksp_type preonly -pc_type lu

Cheers,
Walter Landry
walter at geodynamics.org
-------------- next part --------------
A non-text attachment was scrubbed...
Name: extension.xml
Type: text/xml
Size: 22877 bytes
Desc: not available
Url : http://geodynamics.org/pipermail/cig-long/attachments/20071211/f597c0f2/extension.bin