dolfin team mailing list archive

[Anders Logg <logg@xxxxxxxxx>]

On Thu, Jul 17, 2008 at 03:36:18PM +0100, Nuno David Lopes wrote:
> On Wednesday 16 July 2008, Anders Logg wrote:
> > On Wed, Jul 16, 2008 at 12:24:56AM +0200, Anders Logg wrote:
> > > > By the way about the smooth( ) function,
> > >
> > > I don't know if there's an easy fix.
> >
> > This should be fixed now.
> 
> When testing  the Cylinder i've found that the new smooth function does a less 
> effective job, (i guess), please take a look at the pictures in:
> http://ptmat.ptmat.fc.ul.pt/~ndl/other/dolfin.dir/zct/png.tgz
> 
> I've used the following code (for the circle images):
> //---------------------------------------------------
> UnitCircle mesh(40,UnitCircle::left,UnitCircle::rotsumn);
> for (int i = 0; i < 5; i++)
>     mesh.smooth();
>   File fmesh("circle.pvd");
>   fmesh<<mesh;
> (with "UnitCircle mesh(40)" is a little less obvious) 
> //-----------------------------------------------------------------
> 
> Another problem, and probably a more complicated one to solve:
> For the 3D case like the cylinder, where the mesh is done by the circle 
> transformation (in y,z for instance and maintaining x), 
> if we don't let  the boundary points to be  smoothed then, at least near the 
> (y,z) boundary circle , the cells will be very bad.
> So i'm  really not so certain that this is  a good approach.     
> 
> Another little thing that smooth function could have is an argument  for the 
> number of passes: 
> mesh.smooth( number ) instead of  the repeated calling of  smooth. 

I don't have time to come up with a better algorithm right now, but
adding an argument that allows a variable number of smoothings is a
good suggestion. I'll add it.

-- 
Anders

Attachment: signature.asc
Description: Digital signature