[Wien] Problems with lapw1_mpi in 64-bit

[Iglesias Roberto]

Hi!
 
SCALAPACK comes from the AMD library amdscalapack-3-1-0-pgi-64bit.

  RP  RP_LIB(SCALAPACK+PBLAS): -L/opt/scalapack/amdscalapack-3.1-pgi-64/pgi64/lib -lscalapack -lblacsCinit -lblacs -lblacsCinit
/opt/acml/acml-3.1-pgi-64/pgi64/lib/libacml.a -L/opt/mpi/mpich-1.2.7-pgi-6.1-opteron/lib -lmpich
      FP  FPOPT(par.comp.options): -Mfreeform -fast -Kieee
      MP  MPIRUN commando        : mpirun -np _NP_ -machinefile _HOSTS_ _EXEC_
 
 
What was your version?
 
Kind regards
 
Roberto


________________________________

De: wien-bounces at zeus.theochem.tuwien.ac.at en nombre de Peter Blaha
Enviado el: mié 28/06/2006 10:31
Para: A Mailing list for WIEN2k users
Asunto: Re: [Wien] Problems with lapw1_mpi in 64-bit



Hi,

I run lapw1_mpi in 64bit Xeon machines without problems.

Have you tried a "smaller problem" ? Could be a memory problem.

mpi seems to work (as also your test say, or try lapw0_mpi), but
maybe there is a problem with SCALAPACK/BLACS. Where does your
SCALAPCK  come from ??

Regards

> We're experiencing problems when trying to run our Wien2k installation
> on a cluster of dual core AMD Opteron processors, with 64-bit PGI 6.1
> compilers, MPICH-1.2.6, Scientific Linux 4 (Red Hat 4) and AMD
> libraries. The version of the program is Wien2k_06.2.
>
> We could build and run a test F90 MPICH code in the same environment.
>
> But when running a test, it crashes in lapw1_mpi in the following
> sequence of calls:
>
>    run_lapw -p -cc 0.0001
>    ...
>    /home3/iglesias/w2k_par/x -f TiC lapw1 -p
>    ...
>    /home3/iglesias/w2k_par/lapw1para lapw1.def
>    ...
>    mpirun -np 4 -machinefile .machine1 /home3/iglesias/w2k_par/lapw1_mpi
> lapw1_1.def
>
>   Using             4  processors, My ID =             0
> 2 - <NO ERROR MESSAGE> : Could not convert index 1140850688 into a pointer
> The index may be an incorrect argument.
> Possible sources of this problem are a missing "include 'mpif.h'",
> a misspelled MPI object (e.g., MPI_COM_WORLD instead of MPI_COMM_WORLD)
> or a misspelled user variable for an MPI object (e.g.,
> com instead of comm).
> p0_10792:  p4_error: interrupt SIGx: 13
> Killed by signal 2.
>
> I have not been able to find any solution to this in the mailing list,
> neither in the FAQ. There were messages about compilation errors with a
> 64-bit build, but has someone ever suffered similar problems with lapw1_mpi?
>
> Thanks a lot
>
> Roberto Iglesias
> _______________________________________________
> Wien mailing list
> Wien at zeus.theochem.tuwien.ac.at
> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>


                                      P.Blaha
--------------------------------------------------------------------------
Peter BLAHA, Inst.f. Materials Chemistry, TU Vienna, A-1060 Vienna
Phone: +43-1-58801-15671             FAX: +43-1-58801-15698
Email: blaha at theochem.tuwien.ac.at    WWW: http://info.tuwien.ac.at/theochem/
--------------------------------------------------------------------------
_______________________________________________
Wien mailing list
Wien at zeus.theochem.tuwien.ac.at
http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien


-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/ms-tnef
Size: 7089 bytes
Desc: not available
Url : http://zeus.theochem.tuwien.ac.at/pipermail/wien/attachments/20060628/ee6abd91/attachment.bin