[petsc-users] CFD code with immersed boundary method

Matthew Knepley knepley at gmail.com
Thu Nov 8 10:11:51 CST 2012

On Thu, Nov 8, 2012 at 9:19 AM, Arun Kamath <arun.kamath.m at gmail.com> wrote:
> The fact is I do use a uniform cartesian grid, but the data structure in the
> code is in "AIJ" instead of the "IJK" format.

I do not understand what you mean by AIJ and IJK. AIJ is a storage
format. IJK does
not mean anything to me.

> This was done because the ghost cells (from the immersed boundary method)
> were updated from multiple directions and this cannot be handled using the
> "ijk" format.
> The computational grid is a staggered grid with velocities at the cell edges
> and pressure at cell centres.
> I guess , that puts the DMDA out of the solution then.

Okay, the default DMDA defines variables only at vertices. You can try telling
DA that they are at vertices, but calculating like they are at edges,
which is what
SNES ex30 does. However, then you have to make the DA bigger and handle
BCs yourself.


> So, though I am on a structured uniform cartesian grid (planning on local
> mesh refinement in the future), the boundary treatment doesn't allow me to
> use the procedure that would be normally followed.
> Am I right in this assessment? So then I wait for the manual on DMComplex
> and get pets-dev.
> What do you suggest?
> Thanks a lot!
> --
> Arun Kamath
> PhD Stipendiat
> NTNU, Trondheim

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

More information about the petsc-users mailing list