[petsc-users] MATELEMENTAL

[Hong Zhang]

Shuangshuang,
In ex38.c, m and n are local block sizes, not global. When running np=2 on
a matrix with global sizes 12 by 12, you should give
mpiexec -n 2 ./ex38 -Cexp_view_ownership 1 -m 6 -n 6

Hong



On Tue, Jun 11, 2013 at 3:22 PM, Jin, Shuangshuang <
Shuangshuang.Jin at pnnl.gov> wrote:

>  Hello, I’m trying to set my Jacobian matrix to MATELEMENTAL type.
>
> According to an Elemental matrix example code:
> /src/mat/examples/tests/ex38.c, I wrote my piece of code as following:
>
>   ierr = MatCreate(PETSC_COMM_WORLD, &J); CHKERRQ(ierr); // J: Jacobian
> matrix
>   ierr = MatSetSizes(J, 12, 12, PETSC_DECIDE, PETSC_DECIDE); CHKERRQ(ierr);
>   ierr = MatSetType(J, MATELEMENTAL); CHKERRQ(ierr);
>   ierr = MatSetFromOptions(J); CHKERRQ(ierr);
>   ierr = MatSetUp(J); CHKERRQ(ierr);
>
>   PetscInt       nrows,ncols;
>   const PetscInt *rows,*cols;
>   IS             isrows,iscols;
>   PetscScalar    *v;
>   ierr = MatGetOwnershipIS(J, &isrows, &iscols); CHKERRQ(ierr);
>
>   // Set local matrix entries
>   ierr = ISGetLocalSize(isrows, &nrows); CHKERRQ(ierr);
>   ierr = ISGetIndices(isrows, &rows); CHKERRQ(ierr);
>   ierr = ISGetLocalSize(iscols, &ncols); CHKERRQ(ierr);
>   ierr = ISGetIndices(iscols, &cols); CHKERRQ(ierr);
>   ierr = PetscMalloc(nrows*ncols*sizeof(*v), &v); CHKERRQ(ierr);
>   printf("nrows=%d, ncols=%d\n",nrows,ncols);
>   for (i = 0; i < nrows; i++) {
>     for (j = 0; j < ncols; j++) {
>       v[i*ncols+j] = jacVal[i][j]; // jacVal stores the value of the 12*12
> Jacobian matrix
>     }
>   }
>   ierr = MatSetValues(J, nrows, rows, ncols, cols, v, INSERT_VALUES);
> CHKERRQ(ierr);
>   ierr = ISRestoreIndices(isrows, &rows); CHKERRQ(ierr);
>   ierr = ISRestoreIndices(iscols, &cols); CHKERRQ(ierr);
>   ierr = MatAssemblyBegin(J, MAT_FINAL_ASSEMBLY); CHKERRQ(ierr);
>   ierr = MatAssemblyEnd(J, MAT_FINAL_ASSEMBLY); CHKERRQ(ierr);
>   ierr = ISDestroy(&isrows); CHKERRQ(ierr);
>   ierr = ISDestroy(&iscols); CHKERRQ(ierr);
>
> This works for one processor. When it goes to multiple processor, it will
> fail.
>
> I notice that when the # of processor equals to 1, the nrows and ncols
> will be 12 each, which are the dimension of the Jacobian matrix.
> However, when the # of processor goes up, for example, to 2, nrows=24,
> ncols=12, then it will fail because the index of my jac[i][j] are withn [0,
> 11].
>
> My question is, to use MATELEMENTAL, what is the rule to fit the values of
> my 12x12 matrix into a 24X12 arrays by the two level for loops?
>
> And I don’t understand what the set means here:
> [d3m956 at olympus tests]$ mpiexec -n 1 ex38 -Cexp_view_ownership 1 -m 12 -n
> 12
> Ownership of explicit C:
> Row index set:
> Number of indices in set 12
> 0 0
> 1 1
> 2 2
> 3 3
> 4 4
> 5 5
> 6 6
> 7 7
> 8 8
> 9 9
> 10 10
> 11 11
> Column index set:
> Number of indices in set 12
> 0 0
> 1 1
> 2 2
> 3 3
> 4 4
> 5 5
> 6 6
> 7 7
> 8 8
> 9 9
> 10 10
> 11 11
> nrows=12, ncols=12
>
> [d3m956 at olympus tests]$ mpiexec -n 2 ex38 -Cexp_view_ownership 1 -m 12 -n
> 12
> Ownership of explicit C:
> Row index set:
> [0] Number of indices in set 24
> [0] 0 0
> [0] 1 12
> [0] 2 1
> [0] 3 13
> [0] 4 2
> [0] 5 14
> [0] 6 3
> [0] 7 15
> [0] 8 4
> [0] 9 16
> [0] 10 5
> [0] 11 17
> [0] 12 6
> [0] 13 18
> [0] 14 7
> [0] 15 19
> [0] 16 8
> [0] 17 20
> [0] 18 9
> [0] 19 21
> [0] 20 10
> [0] 21 22
> [0] 22 11
> [0] 23 23
> nrows=24, ncols=12
> [1] Number of indices in set 24
> [1] 0 0
> [1] 1 12
> [1] 2 1
> [1] 3 13
> [1] 4 2
> [1] 5 14
> [1] 6 3
> [1] 7 15
> [1] 8 4
> [1] 9 16
> [1] 10 5
> [1] 11 17
> [1] 12 6
> [1] 13 18
> [1] 14 7
> [1] 15 19
> [1] 16 8
> [1] 17 20
> [1] 18 9
> [1] 19 21
> [1] 20 10
> [1] 21 22
> [1] 22 11
> [1] 23 23
> Column index set:
> [0] Number of indices in set 12
> [0] 0 0
> [0] 1 1
> [0] 2 2
> [0] 3 3
> [0] 4 4
> [0] 5 5
> [0] 6 6
> [0] 7 7
> [0] 8 8
> [0] 9 9
> [0] 10 10
> [0] 11 11
> [1] Number of indices in set 12
> [1] 0 12
> [1] 1 13
> [1] 2 14
> [1] 3 15
> [1] 4 16
> [1] 5 17
> [1] 6 18
> [1] 7 19
> [1] 8 20
> [1] 9 21
> [1] 10 22
> [1] 11 23
> nrows=24, ncols=12
>
> Thanks,
> Shuangshuang
>
>
>
>
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20130611/26203ff9/attachment.html>