[molpro-user] Problems with Biorth run in parallel

[Laura McKemmish]

Hi all, 

I am having a lot of trouble calculating transition moments using the biorth command when I run Molpro in parallel (single node, 4 cores, UCL Legion clusters). It works okay when I use ‘noexc’ in the ci wave functions, but not when I calculate the ci wave functions in full. 

The input is: 

******
{matrop
read,xstate,file=x.orb
save,xstate,3001.2,orbitals}

{ci;orbit,3002.2;
wf,30,4,2;state,1;
save,4002.2;}

{matrop
read,xstate,file=x.orb
save,xstate,3001.2,orbitals}

{ci;orbit,3001.2;
wf,30,1,2;state,1;
wf,30,4,2;state,1;
save,4001.2;}

{ci;trans,4001.2,4002.2,biorth}
*****

My error message is: 

   Transition moment calculation
   =============================
 =============================================================================
 |  Orbitals are different, transformation to biorthogonal orbitals is used |
 =============================================================================
 1PROGRAM * CITRA (MRCI wavefunction orbital transformation)  Author : A.O.Mitrushenkov, 2005

 Ket wavefunction transformed from  4002.2 to  4002.4 in         0.8 sec
 Bra wavefunction transformed from  4001.2 to  4001.4 in         0.8 sec

 Ket wavefunction restored from record 4002.4 to file 8
--------------------------------------------------------------------------
mpirun noticed that process rank 2 with PID 48958 on node node-t00a-001 exited on signal 11 (Segmentation fault).
—————————————————————————————————————

It seems to be a memory problem when the computer tries to reload the bra wavefunction. 

The command I am using to run Molpro is: 
molpro -t 4 -n 4 -I $TMPDIR -d $TMPDIR -W $TMPDIR < $pwd/${fname}.com > $pwd/${fname}.out

and I am submitting the job using the command below:
qsub -P AllUsers -N $job -e $job.e -o $job.o -l h_rt=03:59:59 -l mem=$MEM -pe smp 4 -l tmpfs=150G  -wd $pwd $pwd/r-iterat_jobs.csh $1 $2 $pwd $script $5 $


Thanks for any advice/ assistance with this, 
Laura


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.molpro.net/pipermail/molpro-user/attachments/20151127/02f06964/attachment.html>