PC GAMESS-related discussion club

Alex Granovsky
gran@classic.chem.msu.su

please ask your friend to add the following command line switch to the pc gamess command line options:

pcgamess ...some other switches... -ex < path to directory where pcgp2p.ex resides >

E.g., if you have PC GAMESS' runtime extensions (pcgp2p.ex and fastdiag.ex) in the /opt/pcgamess71 folder on all nodes, use the following switch:

 -ex /opt/pcgamess71

Then, add the following line to your input:

 $p2p p2p=.t. dlb=.t. $end

Best regards,
Alex


On Mon Dec 10 '07, Zoran Matovic wrote
--------------------------------------
>Thanks for answer Alex. I must admitt that I'm not some expert for linux and writter of scripts; however my colleagues wrote a script for me to be able to start pcgamess in parallel using PBS torgue as a job manager. the scripts is given bellow:
>#!/bin/sh

>### Set the job name
>#PBS -N batch

>### Declare myprogram non-rerunable
>#PBS -r n

>### Have PBS mail you results
>##PBS -m ae

>#PBS -q batch

>### Specify the number of cpus for your job.  This example will run on 16 cpus
>### using 8 nodes with 2 processes per node.  Set this to a max of 2 and 2 if you
>### are using the "bigmem" queue.
>##PBS -l nodes=1:ppn=4
>#PBS -l nodes=3:ppn=4

>INPUT="newpdda"
>TPMDIRLIST=""
>echo Working directory is $PBS_O_WORKDIR
>cd $PBS_O_WORKDIR
>echo Running on host `hostname`
>echo Time is `date`
>echo Directory is `pwd`
>echo This jobs runs on the following processors:
>echo `cat $PBS_NODEFILE`
>NPROCS=`wc -l < $PBS_NODEFILE`
>echo This job has allocated $NPROCS cpus

>NODES=`cat $PBS_NODEFILE`

>for host in ${NODES}
>do
>  N=$RANDOM
>  echo "Copying input file to $host..."
>  ssh -v $host rm -rf /tmp/pcgamess_$N/*
>  ssh -v $host mkdir /tmp/pcgamess_$N
>  ssh -v $host cp $PWD/$INPUT /tmp/pcgamess_$N/input
>  TMPDIRLIST=$TMPDIRLIST' '/tmp/pcgamess_$N
>done

>echo $TMPDIRLIST

>mpirun -np $NPROCS -machinefile $PBS_NODEFILE pcgamess $TMPDIRLIST > $INPUT.log
>

>It is not quite clear for me where to put -ex /home/zoran/pcgamess (in my case; is it written OK?). Will you help me how to handle this to activate p2p. When I activate p2p and wrote -ex home at the and of above scripts the pcgamess abort with a message: Loading P2P interface library... error loading library!

>Thanks.

>zoran
>
>
>
>On Mon Dec 10 '07, Alex Granovsky wrote
>---------------------------------------
>>Hi,

>>possibly you are bandwidth limited, if your job is not large enough.

>>Are you using ch_p4 device and running PC GAMESS in DLB mode over P2P interface?

>>Best regards,
>>Alex Granovsky
>>
>>
>>On Mon Dec 10 '07, Zoran Matovic wrote
>>--------------------------------------
>>>Dear pcgamess users,
>>>Till couple days ago I ran pcgamess on 8cpu cluster. now I extended it for 4cpu more. However, when ran jobs with 12cpus it costs me two times more time than with 8 cpu. Why is that so? Can anyone help?
>>>Thanks in advance
>>>zoran matovic

>>>cluster: three machines connected through gigabite switch.
>>>each machine:
>>>Processor: CPU INTEL CORE 2 QUAD Q6600 2.4GHz
>>>MB: Gigabyte GA-P35-DS3P
>>>RAM:  MEMORIJA DDR2 transcdent 1024MB 800MHz 4GB
>>>HDD 160 GB Western Digital WD1600AAJS S-ATA II 2
>>>Video: VGA MSI PCI-E NX7300LE-TD256E 2
>>>Optics:  DVD RW PIONEER DVR-112DBk 2
>>>Case: CHIEFTEC 550W
>>>
>>>
>>>OS: Scientific Linux 4.5  PBS Torgue  (use pcgamess 7.01)
>>>