[Trilinos-Users] Semantics of Epetra_vector reduction operations distributed conformally with overlapping Epetra_Maps

Heroux, Mike MHeroux at csbsju.edu
Fri Jun 17 12:34:43 MDT 2011


Alberto,

The semantics of summation collectives are undefined for vectors with
overlapping GIDs.  There are a few ways to get around this.  Probably most
appropriate is to define a new map that does not have overlap, and associate
your vector with this map.  You can make the association by using the
Epetra_Import facilities, or if you put all of the overlapping terms at the
tail (or head) of your vector, you can make the non-overlapping vector as a
"view" of the overlapped vector.  Then the tail will not be visible in the
non-overlapping vector.

I hope this helps.

Mike

On 6/17/11 1:03 PM, "Alberto F. Martín-Huertas" <amartin at cimne.upc.edu>
wrote:

>  Dear all,
> 
>  Assume that you have an Epetra_Map with overlapping
>  GIDS defined, e.g., accordingly to the following piece of code
> 
>    num_my_elements = 3;
>    if (MyPID == 0)
>    {
>      my_global_elements [0] = 0;
>      global_values [0] = 0.0;
>      my_global_elements [1] = 1;
>      global_values [1] = 1.0;
>      my_global_elements [2] = 2;
>      global_values [2] = 2.0;
>    }
>    else
>    {
>      my_global_elements [0] = 1;
>      global_values [0] = 1.0;
>      my_global_elements [1] = 2;
>      global_values [1] = 2.0;
>      my_global_elements [2] = 3;
>      global_values [2] = 3.0;
>    }
> 
>    Epetra_Map mapeb ( -1,
>                       num_my_elements,
>                       my_global_elements, 0, Comm);
> 
>  and an Epetra_Vector v, distributed conformally with mapeb, i.e.,
> 
>    Epetra_Vector v ( View, mapeb, global_values ) ;
> 
>  How are the semantics of vector reduction operations
>  like, e.g., Norm2, or Norm1, defined for vectors with their
>  distribution defined accordingly to overlapping maps ?
> 
>  Thanks in advance.
> 
>  Best regards,
>   Alberto.
> 
> 
> _______________________________________________
> Trilinos-Users mailing list
> Trilinos-Users at software.sandia.gov
> http://software.sandia.gov/mailman/listinfo/trilinos-users
> 





More information about the Trilinos-Users mailing list