[petsc-users] Getting DMDA local coordinates

[Matthew Knepley]

On Mon, Oct 21, 2013 at 7:14 AM, Åsmund Ervik <asmund.ervik at ntnu.no> wrote:

> Thanks Jed,
>
> The following works fine (for future reference):
>
>   call DMDASetUniformCoordinates(SolScal,0.1,1.0,0.0,1.2,0.0,1.0,ierr)
>
>   call DMGetCoordinatesLocal(SolScal,coordVec,ierr)
>   call DMGetCoordinateDM(SolScal,CoordDM,ierr)
>   call DMDAVecGetArrayF90(CoordDM,coordVec,coords,ierr)
>
>   write(*,*) coords(:,0,0,0)
>   write(*,*) coords(:,imax-1,jmax-1,kmax-1)
> gives
>   0.1  0.0  0.0
>   1.0  1.2  1.0
>
> Perfect. Not very logical though, that we need this new DM, so perhaps
> add an example/something in the manual?
>

A DM describes a data layout over a topology. Here is the topology is a
Cartesian grid, and 'dm' describes the data layout of your solution vector.
If you want the layout of the coordinate vector then you need another
DM, namely 'coordDM'.

   Matt


> I'm by the way planning to send you a Fortran example solving van der
> Pol with a 3D DMDA (if you want it). I could include the above code in
> that example if desired.  The example is written with some utility
> functions (discussed in emails a few weeks ago) such that it is easy to
> reuse 95% of a legacy code solving van der Pol (or other PDE). The main
> motivation for adding this example (IMO) is that none of the current DM
> examples show how DMDAs are used for solving a PDE in a very simple
> case, particularly not for Fortran.
>
> One question in this regard: can you have examples split into two files?
> That makes it easier to use Fortran modules in this case.
>
>
> Åsmund
>
>
> On 21. okt. 2013 13:55, Jed Brown wrote:
> > Åsmund Ervik <asmund.ervik at ntnu.no> writes:
> >
> >> Dear all,
> >>
> >> I'm using a DMDA to manage my grid, and I want to have uniform
> >> coordinates, so I use DMDASetUniformCoordinates. Since I let PETSc
> >> decide the grid partitioning, I need to get the local coordinates, i.e.
> >> xmin,xmax for the local grid. I found the function
> >> DMGetCoordinatesLocal, but I can't seem to figure out how it works (not
> >> in the manual, examples are confusing). What I do is
> >>
> >>   call DMDASetUniformCoordinates(SolScal,0.1,1.0,0.0,1.2,0.0,1.0,ierr)
> >>   call DMGetCoordinatesLocal(SolScal,coordVec,ierr)
> >>   call DMDAVecGetArrayF90(SolScal,coordVec,coords,ierr)
> >>
> >> Then I have the "coords" F90 array, which has the same dimensions as my
> >> local part of the DA, i.e. dof:imax:jmax:kmax (all starting at 0).
> >
> > Actually, the dimensions are dim:imax:jmax:kmax.  Use
> >
> >   call DMGetCoordinateDM(SolScal,CoordDM,ierr)
> >   call DMDAVecGetArrayF90(CoordDM,coordVec,coords,ierr)
> >
> >> Where in this array can I find the values xmin,xmax etc? It looks
> >> really strange, e.g. a slice in i-direction gives 0.10 0.00 1.20 0.55
> >> 0.00
> >>
> >> I do realize I can compute these on my own, but I think using
> >> DMGetCoordinatesLocal has less potential for errors.
> >>
> >> Best regards,
> >> Åsmund Ervik
>



-- 
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
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20131021/f30bbce6/attachment-0001.html>