[CIG-SEISMO] Any experience running SPECFEM3D on a multicore machine?

[Steve Smith]

Does anyone have any experience building and running SPECFEM3D on a  
multicore machine?

I have a working implementation with IFORT and MPICH2 - but there are  
typically process communication problems (MPI_Abort(MPI_COMM_WORLD, 
30)) when running the mesher.

I have compiled and run other MPI codes on my Fedora 9 dual-quad core  
without problems. Therefore I suspect there may be SPECFEM3D  
optimizations/compile flags that may resolve the issue. I'd like to  
correspond with anyone who has a similar problem or knows a possible  
solution.

Thanks,

-Steve Smith
Ph.D. Student
Colorado School of Mines