[CIG-SEISMO] SPECFEM3D_GLOBE_V5.0.0 problem
Arthur Rodgers
rodgers7 at llnl.gov
Mon Mar 15 04:51:45 PDT 2010
Hello Jeroen -
Greetings from Grenoble. I'm settled in an started trying to run
SPECFEM3D_GLOBE_V5.0.0.
I have had several cases crash with the solution blowing up in the
fluid. I started with a 3D case (s29ea) to compare to previous
results (one-chunk, 45-degree width, north korean explosion). I also
tried with a 1D model (1D_iasp91) and it blew up as well.
I have incrementally decreased the time step by as much as 25%, by
changing the DT in read_compute_parameters.f90 as below:
! takes a 5% safety margin on the maximum stable time step
! which was obtained by trial and error
! DT = DT * (1.d0 - 0.05d0)
DT = DT * (1.d0 - 0.25d0)
I'm running on atlas, a linux cluster running CHAOS 4.2 on AMD
Opteron cpu's and the code was compiled with both mpiifort. I had
problems when using mpif90 also.
I have not had to make any changes in DT with the version 4.0.4 of the code.
Could there be problems with running a chunk < 90 degrees?
Regards,
Artie
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://geodynamics.org/pipermail/cig-seismo/attachments/20100315/597eefee/attachment.htm
More information about the CIG-SEISMO
mailing list