fortran interface for MatCreateSeqAIJWithArrays

Barry Smith bsmith at
Mon Aug 24 16:37:47 CDT 2009

On Aug 24, 2009, at 4:19 PM, tsjb00 wrote:

> Hi, there!
> I have a quick question on the Fortran interface of  
> MatCreateSeqAIJWithArrays.
> The documentation indicates 'The i and j indices are 0 based'.

    This means the values in i and j start with zero; it does not mean  
you have to declare them as PetscInt i(0:a big number) you can declare  
them either as
(0:a big number minus 1) or (1:a big number)

> Is it so when called in Fortran? Should the array be 0 based too?

    That does not matter.

> Any precautions to be taken to avoid segmentation error?

    This routine does NOT copy the values in the arrays. This means  
that the arrays that YOU pass in, i,j,a must exist until you finish  
using the matrix. If you allocate them
then don't deallocate them until you destroy the matrix. If you  
declare them, for example PetscScalar a(somebignumber) then make sure  
this variable does not go out of
scope until the matrix is destroyed.

> Another question is about the storage of vector/arrays when using  
> Fortran. In the PETSc examples, say a vector X_v is created using  
> DACreateXXXVector, then the values are accessed with array X_a using  
> DAVecGetArray. To access individual value, the loop is coded as:
>     for (k=zs; k<zs+zm; k++) {
>       for (j=ys; j<ys+ym; j++) {
>     for (i=xs; i<xs+xm; i++) {
>       for (idof=0;idof<dof;idof++){
>                   X_a[k][j][i][idof]=.............
> ...............................................
> &nb sp;         }
>         }
>      }
>  }
> If in Fortran, how should the loop be structured? Please check if  
> the following is right.
>      do k=zs,zs+zm
>      do j=ys,ys+ym
>      do i=xs,xs+xm
>      do iof=1,dof
>            X_a(iof,i,j,k)=..............
>       enddo
> .............



> Many thanks in advance!
> 使用新一代 Windows Live Messenger 轻松交流和共享! 立刻下载!

More information about the petsc-users mailing list