Author Topic: Turbomole parallel error  (Read 1059 times)

kshatresh

  • Newbie
  • *
  • Posts: 1
  • Karma: +0/-0
Turbomole parallel error
« on: February 01, 2018, 11:33:33 am »
Dear Users,

I am using Turbomole 7.2.1 for QM/MM using chemshell interface. The script runs fine for serial version but it gives an error during mpi run, as below:

Parallel program dscf_mpi will be taken out of the TURBODIR directory.
STARTING dscf ON 48 PROCESSORS!
RUNNING PROGRAM /hpc2n/eb/software/Core/Turbomole/7.2.1-mpi/bin/em64t-unknown-linux-gnu_mpi/dscf_mpi.
PLEASE WAIT UNTIL dscf HAS FINISHED.
Look for the output in slave1.output.
MACHINEFILE is /pfs/nobackup/home/k/kddubey/HisA/QM-MM-setup/kdd-MD/SNAP-12621/TURBOMOLE/TS-OPT/B3LYP/FREQ/machinelist.used
 this is node-proc. number 1 running on node b-cn0313.hpc2n.umu.se
 the total number of node-proc. spawned is  50
  parallel platform: MPP or cluster with fast interconnect

 data group $actual step is not empty
 due to the abend of dscf

Error: statistics runs not possible with MPI binaries...
       delete $statistics from control file or use sequential binary

========================
 internal module stack:
------------------------
    dscf
    turbo_setup
========================

 fatal error for MPI rank    1

 statistics run not possible with MPI binaries!
 dscf ended abnormally

Please help me to fix this issue.

Regards
Kshatresh

uwe

  • Global Moderator
  • Sr. Member
  • *****
  • Posts: 425
  • Karma: +0/-0
Re: Turbomole parallel error
« Reply #1 on: February 01, 2018, 03:56:34 pm »
Hello,

Quote
Error: statistics runs not possible with MPI binaries...
       delete $statistics from control file or use sequential binary

chemshell probably uses a very old feature which pre-calculates the task distribution of a parallel job in a separate step. This is called the 'statistics run'. Perhaps chemshell has an option to avoid this unnecessary step.

I'd recommend to contact the Turbomole support and send the input files of the Turbomole job to check if that is the problem and how it can be solved. Or ask the ChemShell people how to switch this statistics run off.

Regards,

Uwe