[mpich-discuss] Problem with MPI_File_write_all in Fortran90?

[Rajeev Thakur]

It's the 2D array allocation problem in C. If you declare the array as int local_array[2][2], and initialize it as local_array[i][j] = blah, not local_array[i,j] = blah, it will work.

Rajeev

On Jul 24, 2012, at 5:16 AM, Angel de Vicente wrote:

> Hi,
> 
> Rajeev Thakur <thakur at mcs.anl.gov> writes:
>> Even in C, the file size should be 40K for a 100*100 integer array, not 118K. 9.8T is generous.
> 
> I guess if the system says that "restart is required" after installing a
> bunch of new libraries, perhaps it is really required :-) After
> restarting, the Fortran code (simplified code, which I attach) behaves
> as expected, but now it is the C version that is giving me funny
> results.
> 
> angelv$$ mpicc -o parallel_i_oc parallel_i_o.c
> angelv$$ mpif90 -o parallel_i_of parallel_i_o.f90
> angelv$$ rm datafile_c
> angelv$$ rm datafile_f
> 
> angelv$$ mpiexec -n 4 ./parallel_i_oc
> angelv$$ mpiexec -n 4 ./parallel_i_of
> 
> angelv$$ od -i datafile_c
> 0000000        2001        2002      102001      102002
> 0000020 -2038304232       32707 -1525747176       32711
> 0000040      202001      202002      302001      302002
> 0000060  -892304872       32686   -13377000       32668
> 0000100
> 
> angelv$$ od -i datafile_f
> 0000000        1001        2001      201001      202001
> 0000020        1002        2002      201002      202002
> 0000040      101001      102001      301001      302001
> 0000060      101002      102002      301002      302002
> 0000100
> angelv$$ 
> 
> Do you see any problem with the C code?
> 
> #include "mpi.h" 
> 
> int main( int argc, char *argv[] ) 
> { 
>  int          gsizes[2], distribs[2], dargs[2], psizes[2], rank, size, m, n; 
>  MPI_Datatype filetype; 
>  int          i,j;
>  MPI_File     fh;  
>  int          *local_array;  
>  MPI_Status   status; 
> 
>  MPI_Init( &argc, &argv ); 
>  MPI_Comm_rank(MPI_COMM_WORLD, &rank); 
>  MPI_Comm_size( MPI_COMM_WORLD, &size ); 
> 
>  local_array = (int *)malloc( 2 * 2 * sizeof(int) ); 
> 
>  for (i=0;i<2;i++) {
>    for (j=0;j<2;j++) {
>      local_array[i,j] = rank * 100000 + (i+1)*1000 + (j+1);
>    }
>  }
> 
>  gsizes[0] = 4;
>  gsizes[1] = 4;
> 
>  distribs[0] = MPI_DISTRIBUTE_BLOCK;
>  distribs[1] = MPI_DISTRIBUTE_BLOCK;
> 
>  dargs[0] = MPI_DISTRIBUTE_DFLT_DARG;
>  dargs[1] = MPI_DISTRIBUTE_DFLT_DARG;
> 
>  psizes[0] = 2;
>  psizes[1] = 2;
> 		
> 
>  MPI_Type_create_darray(4, rank, 2, gsizes, distribs, dargs,  
> 			 psizes, MPI_ORDER_C, MPI_INT, &filetype); 
>  MPI_Type_commit(&filetype); 
> 
>  MPI_File_open(MPI_COMM_WORLD, "datafile_c",  
> 		MPI_MODE_CREATE | MPI_MODE_WRONLY,  
> 		MPI_INFO_NULL, &fh); 
> 
>  MPI_File_set_view(fh, 0, MPI_INT, filetype, "native",  
> 		    MPI_INFO_NULL); 
> 
>  MPI_File_write_all(fh, local_array, 4,  
> 		     MPI_INT, &status); 
> 
>  MPI_File_close(&fh); 
> 
>  MPI_Finalize(); 
>  return 0; 
> } 
> <parallel_i_o.f90>
> Thanks,
> -- 
> Ùngel de Vicente
> http://www.iac.es/galeria/angelv/          
> 
> ---------------------------------------------------------------------------------------------
> ADVERTENCIA: Sobre la privacidad y cumplimiento de la Ley de Protección de Datos, acceda a http://www.iac.es/disclaimer.php
> WARNING: For more information on privacy and fulfilment of the Law concerning the Protection of Data, consult http://www.iac.es/disclaimer.php?lang=en
> _______________________________________________
> mpich-discuss mailing list     mpich-discuss at mcs.anl.gov
> To manage subscription options or unsubscribe:
> https://lists.mcs.anl.gov/mailman/listinfo/mpich-discuss