[MITgcm-support] 24 processor simulation

Chris Hill cnh at mit.edu
Sun Jan 9 20:59:07 EST 2005


Daniel,

  I tried and things seemed to work.

  Did you do

  \rm blanklist.txt

  (see bottom of - http://mitgcm.org/sealion/online_documents/node238.html )

Chris
Daniel Enderton wrote:
> I have talked to Ed and Jean-Michel about this and neither of them can 
> see the problem thus far.  I want to do a 24 processor simulation of the 
> aim atmosphere experiment.  I generate the appropriate exch2 files 
> (w2_e2setup.F, W2_EXCH2_TOPOLOGY.h) with driver.m by specifiying
> 
> nr=32; nb=32; ng=32;
> :
> tnx=16;tny=16;
> 
> I set SIZE.h to:
> 
>      &           sNx =  16,
>      &           sNy =  16,
>      &           OLx =   2,
>      &           OLy =   2,
>      &           nSx =   1,
>      &           nSy =   1,
>      &           nPx =   24,
>      &           nPy =   1,
>      &           Nx  = sNx*nSx*nPx,
>      &           Ny  = sNy*nSy*nPy,
>      &           Nr  =   5)
> 
> Then my compiling commands look like:
> 
> ../../../tools/genmake2 -mods=../code/ 
> -of=../../../tools/build_options/linux_ia32_ifc+mpi_cg01
> make depend
> make
> 
> 
> The resulting PBS error file has just a bunch of these lines:
> FORTRAN STOP ERROR: W2_EEBOOT number of tiles is not divisible by nPx*nSx
> 
> Other than that there are 24 empty STDERR files and 24 short STDOUT 
> files.  Note that if I look at STDOUT.0000 it has:
> (PID.TID 0000.0001)  North neighbor = processor 0000
> (PID.TID 0000.0001)  South neighbor = processor 0000
> (PID.TID 0000.0001)   East neighbor = processor 0001
> (PID.TID 0000.0001)   West neighbor = processor 0023
> 
> And STDOUT.0001 has:
> (PID.TID 0001.0001)  North neighbor = processor 0001
> (PID.TID 0001.0001)  South neighbor = processor 0001
> (PID.TID 0001.0001)   East neighbor = processor 0002
> (PID.TID 0001.0001)   West neighbor = processor 0000
> 
> This appears incorrect.
> 
> The directory with the results is at /s06/enderton/Atm24 on cg01.
> 
> Any ideas would be helpful.
> 
> Daniel
> _______________________________________________
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org
> http://dev.mitgcm.org/mailman/listinfo/mitgcm-support
> 




More information about the MITgcm-support mailing list