[Gfs-users] orthogonal curvilinear coordinates

[Stephane P. <s.p...@gm...>]

Dear all,

The latest version of Gerris (2011-10-25) includes a new GfsMetric
object which allows to define analytical curvilinear coordinates. See:

http://gfs.sourceforge.net/wiki/index.php/GfsMetric

If you want a more graphical idea of what this is all about, have a
look at this new test case:

http://gerris.dalembert.upmc.fr/gerris/tests/tests/lid.html#metric

Note however that not everything works yet. In particular z-coordinate
transformation is not operational yet.

As you can see in the test cases, this seems to work fine for
Navier-Stokes with simple non-uniform stretching. This will probably
not work yet for Navier-Stokes and more complex coordinates (e.g.
polar, spherical etc...).

In the general (orthogonal) case:

* It should work OK for pure advection of scalars, see e.g. (with an
other type of metric):

http://gerris.dalembert.upmc.fr/gerris/tests/tests/cosine.html

* It should also work OK for Poisson problems, see e.g:

http://gerris.dalembert.upmc.fr/gerris/tests/tests/harmonic.html
http://gerris.dalembert.upmc.fr/gerris/tests/tests/gaussian.html

* And also for the GfsRiver Saint-Venant solver:

http://gerris.dalembert.upmc.fr/gerris/tests/tests/lonlat.html

Things which could work but have not been tested yet include:

- Pure scalar diffusion problems
- VOF advection

If you have ideas on how to further test this first implementation,
your contributions would be most welcome.

enjoy,

Stephane