Jean-Paul,

> I think that its a great idea to update the tutorials, but at the same
> time I feel that there's a lot of important information, tricks and
> understanding in doing fundamental processes one's self. That way,
> customisation that canhnot be done inside of the provided helper classes
> will not be a dificult task. Perhaps its not all that critical to most
> of deal.II users, but I think that without knowledge of- and familiarity
> with the lower-level operations it might be difficult to develop a
> program that goes beyond the norm of people use the software for.

I think we agree with that, and we don't plan to replace everything with a 
script from which a script automatically produces a program that when 
compiled with deal.II yields something where if you plug the right input 
file into it, the Laplace equation is solved, if you understand what I 
mean :-)

The question was more along the following lines: we still have the 
old-fashoined way of starting threads. We use it, for example, in 
step-9/13. But it's clumsy, and it doesn't scale very well in programs 
where other parts may also be run in parallel. So should we retain the way 
it is written right now, or replace it with the task-based interface that 
is described in the multithreading module? I think there's something to be 
said to not teach people something that we consider outdated already.

That said, I am a big fan of teaching basics before showing the complicated 
stuff, so the basic outline of tutorials isn't going to change soon, I 
believe. The things we've changed so far is:
- step-12 was rather clumsy in the way it assembled matrices, and the
  process of doing that over possibly refined faces was never convenient
  in the first place. Guido replaced that with a much more comfortable way
- step-16 showed how to use multigrid preconditioners, but so far multigrid
  only worked for globally refined meshes if continuous elements were used.
  Baerbel, Guido and I replaced step-16 with something that shows how to
  solve problems with multigrid on adaptively refined meshes now.

That said, we appreciate your input!
Best
 W.

-------------------------------------------------------------------------
Wolfgang Bangerth                email:            [email protected]
                                 www: http://www.math.tamu.edu/~bangerth/

_______________________________________________
dealii mailing list http://poisson.dealii.org/mailman/listinfo/dealii

Reply via email to