[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[ff3d-users] Re: Negative values, memory=none
From: |
Stephane Del Pino |
Subject: |
[ff3d-users] Re: Negative values, memory=none |
Date: |
Fri, 28 Nov 2003 19:52:21 +0100 |
User-agent: |
KMail/1.5.4 |
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
On Thursday 27 November 2003 10:19, Joerg R. Weimar wrote:
> I have a few questions:
>
> 1. Does ff3d have a bug preventing it from using negative values?
>
> The attached problem "negtest.txt" should generate a linear solution
> ranging from -1 to 1, but somehow generates 1-0-1. This is not just in
> the output, but also in the data, as can be seen from the integral
> evaluation.
I have been fighting a terrible bug few weeks ago and I just forgot to remove
some debugging code and this leads to some unstabilities...
Thanks to you I removed the an absolute value function call that I forgot. It
should work properly by now...
> 2. When selecting option memory(matrix=none), one must also set
> method(type=penalty)
> if dirichlet boundary conditions are present. This is not the default,
> contrary to the information in the docs. Is this correct?
You are right. Default method is *always* elimination. It is easier to use.
With penalty you may have to choose the value of epsilon... The documentation
is quite obsolete in some ways. I have to finish some implementation and then
I will go back to the documentation ... Then I think ff3d will be ready for
version 1.
> 3. Are any primitives beyond box,sphere,cylinder supported (or planned)?
> I would like to see cone, surface of revolution, extrusion.
Yes. I tried to implement all the primitives from POV-Ray.
cone, torus and plane are already available. If you need surface of
revolution and extrusion quickly, I can try to provide them to you as soon as
possible but only with linear interpolation in a first time. Please tell me.
> 4. I am trying to solve a combined problem in two domains:
> first solve on mesh M, then solve a similar problem on mesh M2, using
> the first solution as boundary condition. I cannot figure out how to do
> this correctly (See combined.txt).
I looked at you file. The method seems correct to me but it crashes since your
second mesh M2 has a volume that is 0: b and c are both of the form (1,1,x).
I should check that volume is not 0...
Does this answer to your question?
Best regards,
Stephane.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.3 (GNU/Linux)
iD8DBQE/x5lmbZyhKhZwdc0RAqVvAJ4++IjvlJBdTkx5bV/unhSW9RDtjACdHlJI
YlpLKp5zgYKCTHKc8kYYg6A=
=hBiO
-----END PGP SIGNATURE-----