[CIG-SEISMO] [Fwd: SPECFEM3D_GLOBE_V5.0.0 problem]
Jeroen Tromp
jtromp at princeton.edu
Tue Mar 16 07:33:40 PDT 2010
Hi Artie:
OK, thanks for letting us know this appears to be a compiler problem.
Jeroen
Arthur Rodgers wrote:
> Hi All -
>
> The problem I was having is related to the compilers on the machine
> I'm using. I tried to run the same 1-chunk case with version 4.0.4
> (which had run fine before) using the same compilation (FC=ifort,
> MPIFC=mpiifort, FFLAGS= -O2) and I got the same errors. In both cases
> version 4.0.4 and 5.0.0 I decreased the time step by 50% and got the
> same response. However, both cases crashed with a larger time step ("
> forward simulation became unstable and blew up in the solid").
>
> This indicates I have a compiler version problem, because I ran
> version 4.0.4 many, many times and never had to change the time step.
> Compilers at LLNL (Livermore Computing, LC) are complicated, the
> executables for ifort or mpifort are actually scripts whose behavior
> can be changed by invoking different environment variables or options.
> I will ask the LC helpdesk for assistance with compilation.
>
> Meanwhile, if you discover any problems with one-chunk problems or
> make any changes, please let me know.
>
> Cheers,
> Arthur
More information about the CIG-SEISMO
mailing list