Re: Demo

[Adam C Powell IV <hazelsct@mit.edu>]

Heiko Bauke wrote:

> Hi,
>
> I' locking for a nice demo program to demonstrate the power of
> beowulf computers. It should give some graphical output that's easy
> to understand for people who are not mathematician or physicists.
> Any idea? Rendering Mandelbrot sets is too easy.
>
> We are using Linux and MPI (LAM).

Not to toot my own horn, but PETScGraphics
(http://lyre.mit.edu/~powell/petscgraphics.html) has some neat graphics with it,
though distributed triangulation of isoquant surfaces won't be done for about a
week.  (Also, it exposes a bug in Geomview, which makes it a touch
"unprofessional", just type "aT" to turn off transparency and make the bug go
away.)

The demo I package with it ("make check" or "./chts") solves the time-dependent
Cahn-Hilliard equation using finite differences in 3-D with periodic boundary
conditions, starting with a cube at C=1 as the initial condition.  (Maybe I
should replace this with small random fluctuations around C=0.5 to watch the
pretty spinodal decomposition patterns...)  When distributed triangulation is
done, I'll package and upload version 0.2 into Debian.

You can easily rebuild the PETSc package with lam (default is mpich), just
"debian/rules PETSC_MPI=lam binary".  I'm not sure if this will require
rebuilding PETScGraphics with lam too, we'll see.

Now if I can just get some time to check Eray's METIS package, then after that
I'll finish PEScGraphics 0.2...

Zeen,

-Adam P.

GPG fingerprint: D54D 1AEE B11C CE9B A02B  C5DD 526F 01E8 564E E4B6

              Welcome to the best software in the world today cafe!