[lakeat]

Hi Forum,

Bcoz I have difficulties to insert "-parallel" option to my supercomputer (They did not let me execute mpirun command directly), so I am trying the possible solution like this:

Suppose I have a 4-cores Intel CPU personal computer,
Can I execute the mpirun command without "-parallel" option (without decomposePar first), like this:
mpirun -np 4 `which icoFoam` -case $FOAM_RUN/tutorials/icoFoam/cylinder > log &


Is this meaningful? Please help.. Thank you.

[kbr]

Hi,

this means you run your case as a single case 4 times on 4 CPUs. I think that's not what you want to do.

Kerstin

[lakeat]

Okay, how then can I achieve this goal??
I am using a super computer managed by lsf system, instead execute

mpirun -np 4 icoFoam -parallel > log &;

everytime I need to do:

bsub ./c.lsf

to submit my case to the queue. where, c.lsf file is:

APP_NAME=snode
NP=16
RUN="`which icoFoam` -case /home/users/temp/OpenFOAM/temp-1.5/run/tutorials/icoFoam/cylinder -parallel > log"

and then it stops with complains as shown in log file:

attempt to run parallel on 1 processor.....FOAM aborting...

This error is very annoying! Since the lsf system will dynamically appoint 16 nodes to do the job (so I think It's not available to appoint host names by "-hostfile machines"), and lsf cannot correctly pass the "-parallel" option, regarding this as an attempt to run parallel on 1 processor!!!!!!

What can I do? Please
The question is now "How to use lsf system to manage the foam Job"

[lakeat]

I know in Fluent, there's option "-lsf" and "-mpi=", eg.
bsub -I fluent 3d -lsf
but OpenFOAM, only gives a "-parallel" option, is is possible to provide such more convenient option?