[molpro-user] fail to run parallel jobs thru PBS

[Kirk Peterson]

Hi,

just sort the file $PBS_NODEFILE and then pass this sorted nodefile  
to molpro:

sort -n \$PBS_NODEFILE  > nodefile.$$
set np=\`cat nodefile.$$ | wc -l\`

/home/users/kipeters/molpro2002.6/bin/molpro -n \$np  --mpirun- 
machinefile nodefile.$$   inputfile


-Kirk

On Dec 9, 2005, at 7:49 PM, ST Wong (ITSC) wrote:

> Dear all,
>
> I run molpro job thru PBS like following:
>
> --------------------------------------------
> #!/bin/bash
> #PBS -q   q4n16g
> #PBS -N   test4n
> #PBS -lnodes=4:ppn=2
> #PBS -m   bea
>
> cd /tmp/molpro/demo
> /usr/local/bin/molpro -v -g -o 1-methyl-cis-radical.8.out 1-methyl- 
> cis-radical.in
> --------------------------------------------
>
> but got following error:
>
> --------------------------------------------
> It appears that tasks allocated on the same host machine do not have
> consecutive message-passing IDs/numbers. This is not acceptable
> to the ARMCI library as it prevents SMP optimizations and would
> lead to poor resource utilization.
>
> Please contact your System Administrator or, if you can, modify the  
> TCGMSG
> message-passing job startup configuration.
>
> 0:Cannot run: improper task to host mapping!: 0
> --------------------------------------------
>
> Found that PBS assign jobs "alternatively", e.g. on 2 2-cpu nodes,  
> jobs are run in this order: A,B,A,B.   Seems Molpro jobs running on  
> the same node must have consecutive message-passing ID.  I wonder  
> if this is the cause.  Is there there is any solution ?
>
> Thanks a lot.
> Best Regards,
> ST Wong
>
>
>
>
>
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.molpro.net/pipermail/molpro-user/attachments/20051211/a08aadfb/attachment.html>