[Pw_forum] LSF problem?

Charles polynmr at physics.unc.edu
Thu Apr 24 21:41:00 CEST 2008


Dear PWSCF users

Recently I compiled QE4.0cvs2 on SGI Altix with 128 Itanium2 CPU running 
Linux. MPI is used. Job need to be submitted to the cluster via LSF, and 
the cluster also provides 8 CPU for testing jobs.

I did some calculation with pw.x on the example silicon system. Here is 
the strange I found

1. run locally on the 8 CPU cluster, requesting 4 CPUs for the job

the reported time used by PWSCF is 4.38s CPU time and  2.02s wall time.

It does not look bad to me.

2. run on 128-CPU cluster, requesting 4 CPUs for the job

    PWSCF reports"
    Number of processors in use:       4
    R & G space division:  proc/pool =    4
    PWSCF        : 35m48.41s CPU time,     3m14.98s wall time

This is too slow I guess and the system is overloaded if I run some big 
system.

It looks there is something wrong with the system. But the system 
administrator can not provide useful support on this case.

Could anybody have the similar system tell me how to tweak the compilation?

Thanks!

Charles




More information about the users mailing list