Dear siesta users,

I have compiled siesta-3.2-pl-4 in parallel on cluster, it works well.

For compiling tbtrans, I used the command   *sh build_all.sh*  in
…/siesta-3.2-pl-4/Util
direction, So the tbtrans was built.

I run the program that has 51 Transport K-points.

  The problem is that the speed of calculation doesn’t increase in parallel
mode compared to the serial mode.

I submitted my job using the scripting files that include the below lines:

[fotoohi@cluster 0]$ cat test-tbtrans.pbs

#! /bin/sh

#PBS -N tbtrans-example

#PBS -j oe

#PBS -l nodes=1:ppn=8

#PBS -M [email protected]

#PBS -m abe

source /opt/intel/bin/compilervars.sh intel64

module load openmpi-x86_64

mpd &

cat $PBS_NODEFILE

cd  $PBS_O_WORKDIR

echo job start in `date`

mpiexec -np 8 /home/fotoohi/siesta-3.2-pl-4/Util/TBTrans/tbtrans <scat.fdf
> scat.out.tbt

echo job finish in `date`

I want to know whether tbtrans support parallel runs or not!! Please help
me.

Thanks in advance.

*Somayeh Fotoohi*

Responder a