TURBOMOLE Users Forum

Installation and usage of TURBOMOLE => Parallel Runs => Topic started by: icorral on July 02, 2010, 04:22:38 PM

Title: problem with parallel geometry optimization
Post by: icorral on July 02, 2010, 04:22:38 PM

Hi,

I am trying to run a parallel DFT geometry optimization of a 225 atoms molecule. The dscf module ends normally but the calculation crashes when calculating the gradient and I get the following error message:

 dscf ended normally
0 -------------------------------------FORTRAN server ends
 dscf ended normally
 dscf ended normally
OPTIMIZATION CYCLE 1
/usr/local/TURBOMOLE/bin/x86_64-unknown-linux-gnu_mpi/grad: line 155: 10682 Segmentation fault  $SEQPROG > grad.statistics.parallel
<HP-MPI>: MPI_Send: MPI calls must be made after MPI_Init and before MPI_Finalize
<HP-MPI>: MPI_Send: MPI calls must be made after MPI_Init and before MPI_Finalize
<HP-MPI>: MPI_Send: MPI calls must be made after MPI_Init and before MPI_Finalize
MPI Application rank 1 exited before MPI_Init() with status 15


I have not been able to find in the outputs what is the problem with my calculation.
Is any one familiar with this problem?
Attached are the job.last and grad.statistics.parallel files in case they help,

Thanks a lot in advance,

Regards,

Ines

Title: Re: problem with parallel geometry optimization
Post by: christof.haettig on July 20, 2010, 05:55:25 PM

Did you report this problem to your TURBOMOLE reseller?

Regards,
Christof

Title: Re: problem with parallel geometry optimization
Post by: uwe on August 06, 2010, 08:55:46 PM

Hi,

there is a bug in the parallel statistics run of grad when having Cs symmetry together with some other keywords. A bug fix is available from the Turbomole Support. I think that icorral did receive the fixed binary.

Regards,

Uwe