[Trilinos-Users] constructing rectangular crsmatrix

[Alberto F. Martín-Huertas]

 Dear Sunghwan,

 I already asked a similar question in the past.
 This was my question/the answer by Chris Baker:


 Dear all,

 the Epetra_CrsMatrix class reference explicitly mentions
 that Epetra_CrsMatrix matrices can be rectangular. I have thoroghly
 read Epetra_CrsMatrix class reference I did not manage to see
 any further reference on how to create rectangular Epetra_CrsMatrix.
 I tried with  piece of code below, which tries to create a 2x3
 Epetra_CrsMatrix without success (in particular
 it throws an uncatched exception on matrix->FillComplete()).

 I would greatly appreciate any hint on how to deal with
 rectangular.

 Thanks.

 Best regards,
 Alberto.

 RCP<Epetra_CrsMatrix> build2x3( double a00, double a01, double a02,
 double a10, double a11, double a12,
 Epetra_Comm & Comm)
 {
 RCP<Epetra_Map> map = rcp(new Epetra_Map(2,0,Comm));
 RCP<Epetra_CrsMatrix> matrix = rcp(new
 Epetra_CrsMatrix(Copy,*map,3));

 int MyPID = Comm.MyPID();
 int indicies[3] = { 0, 1, 2 };
 double values[3];

 int num_my_elements,   i ;
 int * my_global_elements ;

 num_my_elements    = map->NumMyElements();
 my_global_elements = map->MyGlobalElements();

 for ( i = 0 ; i < num_my_elements ; i++)
 {
 switch ( my_global_elements[i] )
 {
 case 0:
 values[0] = a00;
 values[1] = a01;
 values[2] = a02;
 break;
 case 1:
 values[0] = a10;
 values[1] = a11;
 values[2] = a12;
 break;
 }
 matrix->InsertGlobalValues(my_global_elements[i], 3,
 values,indicies);
 }

 // pack it up
 matrix->FillComplete();

 return matrix;
 }


 Hello Alberto,

 The shape of a sparse matrix, of course, isn't necessarily defined by 
 the
 entries. For Epetra, the shape of the matrix is defined by the range 
 and
 domain maps, I.e., the maps for the linear subspaces between which the
 matrix is an operator.

 What this means is that defining a rectangular Epetra_CrsMatrix 
 consists
 of specifying distinct domain and range maps the matrix. These are 
 passed
 to FillComplete(). In the case of your example, you can pass the row 
 map
 used to construct the matrix as the range map of the matrix, but you
 should pass a 3-dimensional map as the domain map. Documentation for 
 this
 version of FillComplete() is here:
 http://trilinos.sandia.gov/packages/docs/r10.6/packages/epetra/doc/html/cla
 ssEpetra__CrsMatrix.html#aac3c13275e8dbe80826e820abbb8e49c


 An example of this usage is available here:
 Trilinos/packages/epetra/test/CrsRectMatrix/cxx_main.cpp

 Chris


 I hope it helps.
 Best regards,
  Alberto.


 On Mon, 4 Mar 2013 21:35:53 +0900, Sunghwan Choi wrote:
> Hi,
>
> In CrsMatrix class reference, it is written that rectangular matrix
> could be constructed. However, I am not sure that not giving column
> values make crsmatrix rectangular. If I do not give values, crsmatrix
> recognize them as zero then operation that I want to do can be done
> however, by doing that, is there any disadvantage of performance? Is
> there different way to construct rectangular crsmatrix? If it is,
> please let me know
>
> Sunghwan Choi

-- 
 Alberto F. Martín-Huertas
 Centre Internacional de Mètodes Numèrics a l'Enginyeria (CIMNE)
 Parc Mediterrani de la Tecnologia, UPC
 Esteve Terradas 5, Building C3, Office 210,
 08860 Castelldefels (Barcelona, Spain)
 Tel.: (+34) 9341 34223
 e-mail: amartin at cimne.upc.edu

 ________________
 IMPORTANT NOTICE
 All personal data contained on this mail will be processed 
 confidentially and registered in a file property of CIMNE in
 order to manage corporate communications. You may exercise the rights 
 of access, rectification, erasure and object by
 letter sent to Ed. C1 Campus Norte UPC. Gran Capitán s/n Barcelona.