[ARPSSUPPORT] (no subject)

[Kevin W. Thomas]

>Hello,
>    I am a user of ARPS, now I can run the serial verion of ARPS successfuly. When I run the parallel version "arps_mpi" with LSF, I only find one way :"mpirun.lsf -np xx ./arps_mpi". Most of the time this way works well, but sometimes does not. When I using the "bsub" command :" bsub -n xx -i input/arps.input ./arps_mpi", the program can't get the correct number of proccesors. When I run the other models (WRF), the "bsub" command works well. How can I run the  parallel version "arps_mpi" using a "bsub" command? Could you give any suggestions?
> Thank you!
> 
>Sincerely,
>
>Wangpeng

Hi...

The "mpirun.lsf" command accepts the "-np xx" syntax, however, it then ignores
it, and runs the job with the full number of processors allocated to it.  ARPS
software will check, and if the number of processors gives it doesn't match
what it is started with, it's a fatal error.

This means the job will only work when "nproc_x times nproc_y" is "xx".

You need to look into submit matching scripts ("bsub < batch_script", not
"bsub batchscript").

	Kevin W. Thomas
	Center for Analysis and Prediction of Storms
	University of Oklahoma
	Norman, Oklahoma
	Email:  kwthomas at ou.edu