[petsc-users] Constructing a MATNEST with blocks defined in different procs

Mark Adams mfadams at lbl.gov
Tue Apr 9 13:30:56 CDT 2019


On Tue, Apr 9, 2019 at 10:36 AM Diogo FERREIRA SABINO <
diogo.ferreira-sabino at univ-tlse3.fr> wrote:

> Hi Mark,
> Thank you for the quick answer.
>
> So, I defined each block of the Nest matrix as a MATMPIAIJ in the World
> communicator and I set the local sizes in such a way that A00 and A11 are
> in the processors 0 and 1, respectively.
> However, now I'm not able to set the off diagonal blocks of the Nest
> matrix. For example, if I try to set the block (0,1) of the nest with A00,
> the following error occurs:
>
> [0]PETSC ERROR: Arguments are incompatible
> [0]PETSC ERROR: Local sizes (2,2) of nested submatrix (0,1) do not agree
> with space defined by index sets (2,0)
>
> I'm using the following lines to generate the Nest matrix:
>
>   AfullNESTpointer[0]=Ablocks[0];
>   AfullNESTpointer[1]=Ablocks[0];  //Not working
>   // AfullNESTpointer[1]=NULL;       //It works
>   AfullNESTpointer[2]=NULL;
>   AfullNESTpointer[3]=Ablocks[1];
>
>   MatCreate(PETSC_COMM_WORLD,&AfullNEST);
>   MatSetSizes(AfullNEST,PETSC_DECIDE,PETSC_DECIDE,4,4);
>
> MatCreateNest(PETSC_COMM_WORLD,2,NULL,2,NULL,AfullNESTpointer,&AfullNEST);
>
>
This is not how you create a MatNest. See:

https://www.mcs.anl.gov/petsc/petsc-current/docs/manualpages/Mat/MatCreateNest.html

The C example look pretty much like what you want to do:

MatCreateNest <https://www.mcs.anl.gov/petsc/petsc-current/docs/manualpages/Mat/MatCreateNest.html#MatCreateNest>(PETSC_COMM_WORLD
<https://www.mcs.anl.gov/petsc/petsc-current/docs/manualpages/Sys/PETSC_COMM_WORLD.html#PETSC_COMM_WORLD>,
2, NULL, 2, NULL, s->subA, &s->A);




> A simple test is given below, lunching it with: mpirun -n 2
> ./Main_petsc.exe
>
> Thanks in advance,
> Diogo
>
>
>
> static char help[]     = "Create MPI Nest";
> #include <petscmat.h>
> #undef __FUNCT__
> #define __FUNCT__ "main"
> int main(int argc,char **argv)
> {
>   PetscInitialize(&argc,&argv,(char*)0,help);
>
> //////////////////////////////////////////////////////////////////////////////
>   PetscErrorCode  ierr;
>   PetscInt        i,PetscIntTEMP;
>
> //////////////////////////////////////////////////////////////////////////////
>   PetscMPIInt MPIrank,MPIsize;
>   MPI_Comm_rank(PETSC_COMM_WORLD,&MPIrank);
>   MPI_Comm_size(PETSC_COMM_WORLD,&MPIsize);
>   ///////////////////////////////////////////////////////////////   Create
> Mats:
>   Mat Ablocks[MPIsize];
>
>   for(i=0;i<MPIsize;i++){
>     MatCreate(PETSC_COMM_WORLD,&Ablocks[i]);
>
>     PetscIntTEMP = (i==MPIrank) ? 2 : 0;
>
> MatSetSizes(Ablocks[i],PetscIntTEMP,PetscIntTEMP,PETSC_DECIDE,PETSC_DECIDE);
>     MatSetType(Ablocks[i],MATMPIAIJ);
>
> MatMPIAIJSetPreallocation(Ablocks[i],PetscIntTEMP,NULL,PetscIntTEMP,NULL);
>
>     PetscInt ISstart,ISend;
>     MatGetOwnershipRange(Ablocks[i],&ISstart,&ISend); //Check info
>     PetscSynchronizedPrintf(PETSC_COMM_WORLD,"proc(%d): Ablocks[%d]:(%d to
> %d) \n",MPIrank,i,ISstart,ISend);
>   }
>   PetscSynchronizedFlush(PETSC_COMM_WORLD,PETSC_STDOUT);
>
>   MatSetValue(Ablocks[MPIrank],0,0,5+MPIrank,INSERT_VALUES);
>   MatSetValue(Ablocks[MPIrank],0,1,10+MPIrank,INSERT_VALUES);
>   MatSetValue(Ablocks[MPIrank],1,0,15+MPIrank,INSERT_VALUES);
>   MatSetValue(Ablocks[MPIrank],1,1,20+MPIrank,INSERT_VALUES);
>
>   for(i=0;i<MPIsize;i++){
>     MatAssemblyBegin(Ablocks[i],MAT_FINAL_ASSEMBLY);
> MatAssemblyEnd(Ablocks[i],MAT_FINAL_ASSEMBLY);
>     // MatView(Ablocks[i],PETSC_VIEWER_STDOUT_WORLD);
>   }
>   ///////////////////////////////////////////////////////////////   Create
> Nest:
>   Mat       AfullNEST, *AfullNESTpointer;
>   Mat       AfullConverted;
>
>   PetscMalloc1(4,&AfullNESTpointer);
>   AfullNESTpointer[0]=Ablocks[0];
>   AfullNESTpointer[1]=Ablocks[0];  //Not working
>   // AfullNESTpointer[1]=NULL;       //It works
>   AfullNESTpointer[2]=NULL;
>   AfullNESTpointer[3]=Ablocks[1];
>
>   MatCreate(PETSC_COMM_WORLD,&AfullNEST);
>   // MatSetSizes(AfullNEST,2,2,PETSC_DECIDE,PETSC_DECIDE);
>   MatSetSizes(AfullNEST,PETSC_DECIDE,PETSC_DECIDE,4,4);
>
> MatCreateNest(PETSC_COMM_WORLD,2,NULL,2,NULL,AfullNESTpointer,&AfullNEST);
>
>   MatView(AfullNEST,PETSC_VIEWER_STDOUT_WORLD);
>   MatConvert(AfullNEST,MATMPIAIJ,MAT_INITIAL_MATRIX,&AfullConverted);
>   MatView(AfullConverted,PETSC_VIEWER_STDOUT_WORLD);
>
>   ierr = PetscFinalize(); CHKERRQ(ierr);  return 0;
> }
> April 5, 2019 2:38:46 PM CEST Mark Adams <mfadams at lbl.gov> wrote:
>
>
>
> On Fri, Apr 5, 2019 at 7:19 AM Diogo FERREIRA SABINO via petsc-users <
> petsc-users at mcs.anl.gov> wrote:
>
> Hi,
> I'm new in petsc and I'm trying to construct a MATNEST in two procs, by
> setting each block of the nested matrix with a MATMPIAIJ matrix defined in
> each proc.
> I'm trying to use MatCreateNest or MatNestSetSubMats, but I'm not being
> able to do it.
> Using MatNestSetSubMats, I'm trying to construct the MATNEST, giving a
> pointer to the correct matrices depending on the MPIrank of that proc.
> I'm obtaining the error message for the line
> :MatNestSetSubMats(AfullNEST,1,&IS_ROW,2,&IS_COL,AfullNESTpointer);
> [0]PETSC ERROR: Invalid argument
> [0]PETSC ERROR: Wrong type of object: Parameter # 5
>
> Is there a way of doing it, or all the blocks of the MATNEST have to exist
> in the same communicator as the MATNEST matrix?
>
>
> Yes, they must all have the same communicator. A matrix can be empty on a
> process, so you just create them with the global communicator, set the
> local sizes that you want (eg, 0 on some procs).
>
>
> A simple test is given below, lunching it with: mpirun -n 2
> ./Main_petsc.exe
>
> static char help[]     = "Create MPI Nest";
> #include <petscmat.h>
>
> #undef __FUNCT__
> #define __FUNCT__ "main"
> int main(int argc,char **argv)
> {
>   PetscInitialize(&argc,&argv,(char*)0,help);
>
> //////////////////////////////////////////////////////////////////////////////
>   PetscErrorCode  ierr;
>   PetscMPIInt MPIrank,MPIsize;
>   MPI_Comm_rank(PETSC_COMM_WORLD,&MPIrank);
>   MPI_Comm_size(PETSC_COMM_WORLD,&MPIsize);
>
>   ////////////////////////////////////////////////////////   Create Each
> Matrix:
>   Mat Adiag;
>
>   //Create a Adiag different on each proc:
>   ierr = MatCreate(PETSC_COMM_SELF,&Adiag);
>  CHKERRQ(ierr);
>   ierr = MatSetSizes(Adiag,2,2,PETSC_DECIDE,PETSC_DECIDE);
> CHKERRQ(ierr);
>   ierr = MatSetType(Adiag,MATMPIAIJ);
>  CHKERRQ(ierr);
>   ierr = MatSetFromOptions(Adiag);
> CHKERRQ(ierr);
>   ierr = MatMPIAIJSetPreallocation(Adiag,2,NULL,2,NULL);
> CHKERRQ(ierr);
>
>   MatSetValue(Adiag,0,0,(MPIrank+5),INSERT_VALUES);
>   MatSetValue(Adiag,0,1,(MPIrank+10),INSERT_VALUES);
>   MatSetValue(Adiag,1,0,(MPIrank+15),INSERT_VALUES);
>   MatSetValue(Adiag,1,1,(MPIrank+20),INSERT_VALUES);
>   MatAssemblyBegin(Adiag,MAT_FINAL_ASSEMBLY);
>  MatAssemblyEnd(Adiag,MAT_FINAL_ASSEMBLY);
>
>   ///////////////////////////////////////////////////////////////   Create
> Nest:
>   MPI_Barrier(PETSC_COMM_WORLD);
>   Mat       AfullNEST, *AfullNESTpointer;
>
>   PetscMalloc1(2,&AfullNESTpointer);
>   AfullNESTpointer[0]=NULL;
>   AfullNESTpointer[1]=NULL;
>   AfullNESTpointer[MPIrank]=Adiag;
>   // Rank=0 --> AfullNESTpointer[0]=Adiag; AfullNESTpointer[1]=NULL;
>   // Rank=1 --> AfullNESTpointer[0]=NULL;  AfullNESTpointer[1]=Adiag;
>
>   IS        IS_ROW,IS_COL;
>   ISCreateStride(PETSC_COMM_SELF,1,MPIrank,0,&IS_ROW);
>   ISCreateStride(PETSC_COMM_SELF,2,0,1,&IS_COL);
>   // Rank=0 --> IS_ROW= [ 0 ] ; IS_COL= [ 0, 1 ] ;
>   // Rank=1 --> IS_ROW= [ 1 ] ; IS_COL= [ 0, 1 ] ;
>
>   MatCreate(PETSC_COMM_WORLD,&AfullNEST);
>   MatSetSizes(AfullNEST,2,2,PETSC_DECIDE,PETSC_DECIDE);
>   // MatSetSizes(AfullNEST,PETSC_DECIDE,PETSC_DECIDE,4,4);
>   //
> MatCreateNest(PETSC_COMM_WORLD,1,&IS_ROW,1,&IS_COL,AfullNESTpointer,&AfullNEST);
>     ierr =
> MatNestSetSubMats(AfullNEST,1,&IS_ROW,2,&IS_COL,AfullNESTpointer);
> CHKERRQ(ierr);
>
>   ierr = PetscFinalize(); CHKERRQ(ierr);
>   return 0;
> }
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20190409/0f81e107/attachment.html>


More information about the petsc-users mailing list