[petsc-users] How to efficiently fill in, in parallel, a PETSc matrix from a COO sparse matrix?

[Matthew Knepley]

On Tue, Jun 20, 2023 at 10:55 AM Diego Magela Lemos via petsc-users <
petsc-users at mcs.anl.gov> wrote:

> Considering, for instance, the following COO sparse matrix format, with
> repeated indices:
>
> std::vector<size_t> rows{0, 0, 1, 2, 3, 4};
> std::vector<size_t> cols{0, 0, 1, 2, 3, 4};
> std::vector<double> values{2, -1, 2, 3, 4, 5};
>
> that represents a 5x5 diagonal matrix A.
>
> So far, the code that I have is:
>
> // fill_in_matrix.cc
> static char help[] = "Fill in a parallel COO format sparse matrix.";
> #include <petsc.h>#include <vector>
> int main(int argc, char **args){
>     Mat A;
>     PetscInt m = 5, i, Istart, Iend;
>
>     PetscCall(PetscInitialize(&argc, &args, NULL, help));
>
>     PetscCall(MatCreate(PETSC_COMM_WORLD, &A));
>     PetscCall(MatSetSizes(A, PETSC_DECIDE, PETSC_DECIDE, m, m));
>     PetscCall(MatSetFromOptions(A));
>     PetscCall(MatSetUp(A));
>     PetscCall(MatGetOwnershipRange(A, &Istart, &Iend));
>
>     std::vector<PetscInt> II{0, 0, 1, 2, 3, 4};
>     std::vector<PetscInt> JJ{0, 0, 1, 2, 3, 4};
>     std::vector<PetscScalar> XX{2, -1, 2, 3, 4, 5};
>
>     for (i = Istart; i < Iend; i++)
>         PetscCall(MatSetValues(A, 1, &II.at(i), 1, &JJ.at(i), &XX.at(i), ADD_VALUES));
>
>     PetscCall(MatAssemblyBegin(A, MAT_FINAL_ASSEMBLY));
>     PetscCall(MatAssemblyEnd(A, MAT_FINAL_ASSEMBLY));
>     PetscCall(MatView(A, PETSC_VIEWER_STDERR_WORLD));
>
>     PetscCall(MatDestroy(&A));
>     PetscCall(PetscFinalize());
>     return 0;
> }
>
> When running it with
>
> petscmpiexec -n 4 ./fill_in_matrix
>
>
> I get
>
>
>  Mat Object: 4 MPI processes
>
>   type: mpiaij
> row 0: (0, 1.)
> row 1: (1, 2.)
> row 2: (2, 3.)
> row 3: (3, 4.)
> row 4:
>
>
> Which is missing the entry of the last row.
>
> What am I missing? Even better, which would be the best way to fill in this matrix?
>
> We have a new interface for this:

  https://petsc.org/main/manualpages/Mat/MatSetValuesCOO/

  Thanks,

     Matt

-- 
What most experimenters take for granted before they begin their
experiments is infinitely more interesting than any results to which their
experiments lead.
-- Norbert Wiener

https://www.cse.buffalo.edu/~knepley/ <http://www.cse.buffalo.edu/~knepley/>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20230620/8a48b549/attachment-0001.html>