[Wien] Re: [Wien]

[Peter Blaha]

> >From this test, I realized 3 processors do not compatible for BLACS.

Right! Probably they must be multiples of 2 only.

> Then I changed the number of processors, from 3 to 4 processors.
> #mpirun -np 4 -machinefile m /home/wien2k/WIEN2k_03_3/SRC_lapw1/
> lapw1_mpi lapw1.def
>  Using            4  processors, My ID =            0
>  Using            4  processors, My ID =            2
>  Using            4  processors, My ID =            1
>  Using            4  processors, My ID =            3
> FORTRAN STOP LOPW - Error
> FORTRAN STOP LOPW - Error
> FORTRAN STOP FORTRAN STOP LOPW - Error
> LOPW - Error
>
> --------------------
> Content of the file "m" is changed to
> earth46
> earth47
> earth48
> earth49
> --------------------
>
> I have any more ideas to test the fine grained parallel executions.
> Could anyone help me? Any suggestions are very welcome.

Does lapw1 run in single mode ? LOPW errors often occur when your struct
file is wrong and some atoms are specified twice.

If yes, is your lapw1-version a recent one ? We have constantly updated
the parallel lapw1 and fixed small bugs showing up only for some compilers.

PS: I've got a new parallel hns.F from R.Dohmen with better parallelization
and will put it on the web soon.

                                      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/
--------------------------------------------------------------------------