[Trilinos-Users] [EXTERNAL] AZ_extract_comm_info: Received elements must be stored after.....

Raymond Tuminaro rstumin at sandia.gov
Mon May 23 11:12:31 EDT 2016


Ashesh,
    I believe this error message is connected to something that Aztec does not like in the matrix A as opposed to the 
vector b. Aztec is kind of old and has some requirements in terms of the numbering of columns. Are you explicitly 
specifying the column map or are you letting epetra's FillComplete() do it for you? If you are specifying the column 
map, then Aztec has some restrictions. I'm not sure if Belos has any restrictions ... so if you switch to Belos these 
errors may go away.


In Aztec, rows are assigned to processors. For the column entries within these rows with remote global ids (i.e. global 
ids that are not the same as the global ids of the rows assigned to this processor), the local ids of these columns must 
appear after the local ids of the non-remote global ids.  Not sure if that makes sense or not. The bottom line
is that the first NumMyElements entries of the column map should be the same as the first NumMyElements entries of the 
row map. I believe Aztec has some other restrictions on the column map, but before describing those ... I'd like to see 
if anything that I just said makes sense to you and if you are manually specifying the column map.

-Ray



On 05/22/16 22:44, Ashesh Chattopadhyay wrote:
> Dear all,
>
> I am trying to use AztecOO to solve a linear system AX=B. My B matrix has, for example, 24 elements divided into, say
> two processors,using a linear Map. All the elements of B are zero apart from 16-19 indices, which have values  of 10.
> To do that I am looping over NumMyElements and only updating values of B if they are at the correct global ids. When I
> use Aztecoo i get the error message :
> *AZ_extract_comm_info: Received elements must be stored after 12 local elements.*
>
>
> The code snippet in which I add elements to the b vector is as follows. I also understand from previous mail threads
> that this error is an issue with the r Epetra_Vector b. Please suggest as to how I can debug this issue.
>
> *for(int i=0;i<NumMyElements;i++)
> {
>
>    if(MyGlobalElements[i]==NumNodes)
>     {
>         for(int k=0;k<hor;k++)
>      {
>        b[i+k]=10;
>
>     }
> }
>
>
> }*
>
> Thanks
> Ashesh Chattopadhyay
> --
> Research Assistant, Multi-Scale Multi-Physics Computation Lab
> University of Texas at El Paso
> Texas, El Paso, USA
>
>
>
> This body part will be downloaded on demand.
>

-- 
Ray Tuminaro                      phone: (925) 294-2564
MS 9159                           fax:   (925) 294-2234
Sandia National Laboratories      email: rstumin at sandia.gov
PO Box 969                        http://www.cs.sandia.gov/~rstumin
Livermore, CA 94551




More information about the Trilinos-Users mailing list