I'm using unix server (OS: centos5) and I'm using several clusters.
Some clusters have 8 cores, other are 24 cores and etc.
I execute jobs using this command.
qsub -q cluster_name -d $PWD Program_name
To efficiently use clusters, I'm using OPENMP for parallelization.
Problem is, in most times, my program use only 3 cores at the same time.
It means, I'm wasting computational power a lot.
When I use my PC (Ubuntu) not clusters,
I maximize efficiency, by using in the following way.
OPEN TERMINAL
export OPM_NUM_THREAD=A
./program_name
However, I don't know how to apply this method to the clusters.
Therefore, I want to use clusters in the following way.
./Execute.sh
And Execute.sh is,
SET_NUMBER_OF_CORE_FOR_PROGRAM1
qsub -q cluster_name -d $PWD Program_name1&
SET_NUMBER_OF_CORE_FOR_PROGRAM2
qsub -q cluster_name -d $PWD Program_name2&
...
SET_NUMBER_OF_CORE_FOR_PROGRAMn
qsub -q cluster_name -d $PWD Program_namen&
at least I hope I can set same number of core for all programs.
If you know how to do this, please let me know.
Thank you for reading,
user2268721.
Aucun commentaire:
Enregistrer un commentaire