[petsc-dev] MatSetValuesBlocked not inserting values
Matthew Knepley
knepley at gmail.com
Wed Apr 8 18:53:36 CDT 2020
I would just get in the debugger, break at the MatSetValuesBlocked() line,
and step through until its about to insert a value and
see what goes wrong. That would only take about 5 minutes.
Thanks,
Matt
On Wed, Apr 8, 2020 at 6:17 PM Jacob Faibussowitsch <jacob.fai at gmail.com>
wrote:
> No I also printed those out and checked them. They return 0->nGlobalAtoms.
> The globalID-1 is because the other library which is assigning these ID's
> (LAMMPS) doesn’t use 0-based indexing for the global indices for whatever
> reason.
>
> Best regards,
>
> Jacob Faibussowitsch
> (Jacob Fai - booss - oh - vitch)
> Cell: (312) 694-3391
>
> On Apr 8, 2020, at 5:14 PM, Jed Brown <jed at jedbrown.org> wrote:
>
> Sure, but could you have computed them as a negative number here?
>
> Jacob Faibussowitsch <jacob.fai at gmail.com> writes:
>
> Supposed to be the block row and column indices. So they are the atom
> global IDs.
>
> Best regards,
>
> Jacob Faibussowitsch
> (Jacob Fai - booss - oh - vitch)
> Cell: (312) 694-3391
>
> On Apr 8, 2020, at 4:29 PM, Jed Brown <jed at jedbrown.org> wrote:
>
> What are idxm[0] and idxn[0] in this code?
>
> Jacob Faibussowitsch <jacob.fai at gmail.com> writes:
>
> Hello All,
>
> I am using MATSBAIJ to make a symmetric hamiltonian matrix, but for some
> reason MatSetValuesBlocked isn’t inserting anything.
>
> Setup:
>
> I want a 4n x 4n symmetric matrix, where each MatSetValuesBlocked inserts
> a 4x4 sub-matrix of values into the global matrix, positioned by the global
> ID of two atoms that I am extracting. So the 4x4 interaction sub matrix
> between atom 1 and atom 4 will be in rows 0-4 and cols 16-20.
>
> My non-working implementation:
>
> PetscInt bs = 4;
>
> ierr = MatCreateSBAIJ(comm, bs, bs*nLocalAtoms, bs*nLocalAtoms,
> PETSC_DETERMINE, PETSC_DETERMINE, 1, NULL, 1, NULL, H);CHKERRQ(ierr);
> ierr = MatSetOption(*H, MAT_SYMMETRIC, PETSC_TRUE);CHKERRQ(ierr);
> ierr = MatSetUp(*H);CHKERRQ(ierr);
> .
> .
> .
> START LOOP {
>
> Calculations...
>
> const PetscScalar valmat[16] = {Es[0], Es[1], Es[2], Es[3], Ex[0], Ex[1],
> Ex[2], Ex[3], Ey[0], Ey[1], Ey[2], Ey[3], Ez[0], Ez[1], Ez[2], Ez[3]};
> PetscInt idxm[1], idxn[1];
> PetscInt m = 1, n = 1;
>
> if (globalIDs[i] < globalIDs[neighidx[0]]-1) {
> // Insert in upper triangular
> idxm[0] = globalIDs[i]-1;
> idxn[0] = globalIDs[neighidx[0]]-1;
> } else {
> // swap so we insert on upper triangular
>
> idxn[0] = globalIDs[i]-1;
> idxm[0] = globalIDs[neighidx[0]]-1;
> }
> ierr = MatSetValuesBlocked(*H, m, (const PetscInt *) idxm, n, (const
> PetscInt *) idxn, valmat, INSERT_VALUES);CHKERRQ(ierr);
> } END LOOP
>
> ierr = MatAssemblyBegin(*H, MAT_FINAL_ASSEMBLY);CHKERRQ(ierr);
> ierr = MatAssemblyEnd(*H, MAT_FINAL_ASSEMBLY);CHKERRQ(ierr);
>
> RETURN H
>
> Output from -mat_view:
>
> Mat Object: 2 MPI processes
> type: mpisbaij
> row 0:
> row 1:
> row 2:
> row 3:
> row 4:
> row 5:
> row 6:
> row 7:
> row 8:
> row 9:
> row 10:
> row 11:
> row 12:
> row 13:
> row 14:
> row 15:
>
> I also checked -info :mat: and it also reports X unneeded storage space
> but more importantly, 0 used. I am not sure what is going wrong here.
>
> Best regards,
>
> Jacob Faibussowitsch
> (Jacob Fai - booss - oh - vitch)
> Cell: (312) 694-3391
>
>
>
--
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-dev/attachments/20200408/6e4a8884/attachment.html>
More information about the petsc-dev
mailing list