Re: [Libmesh-users] More questions about libmesh linking and others

[Daniel D. <d.d...@tu...>]

On Sun, 7 Mar 2004, Yuna wrote:

> Hi Ahmed,
>
>   Thanks for your advise.
>
>   I set the environment variables in my .cshrc file, now PETSC directory has
> been found by the configure. Good!
>
>   I am now try to run Make again! I still have several things I can's get
> clearly:
>
> 1). What is the different when running make on "OPTIMAL MODE" and "DEBUG
> MODE"? Which one is better? And generally what method should we use?


That's what i suggest:
- 95% of your time spending on libmesh/coding, use debug
mode, and use small-sized problems that still reflect everything you
want to do later.  The neat thing with METHOD=dbg, you can use
debuggers to trace what happened, and also get some more hints from
libmesh itself...

- Only when you are perfectly sure that your code is debugged and gives
correct answers, use optimized mode.  _Very_ seldom it may happen that
you get an error when using opt mode (which didn't occur in dbg), but it's
generally harder to track down.

I also heavily recommend to consult the PETSc documentation.

>
> 2). I remember the input file can be IDEAS .unv file? Am I right? How do I
> use it?

Use some common FE preprocessor.  the Universal format is, as the name
says, pretty universal. ;)

libmesh has both .unv input and output, and also offers some mesh
-generation and -modification tools.  With this you can also play around
with .unv-files:  create some mesh in libmesh, write to .unv, edit the
.unv-file, read back to libmesh, see whether mesh is ok using the gmv
viewer, etc...

Rather recently, there is also an interface for tetgen-produced files
(.node and .ele).  Tetgen is freely available (tetgen.berlios.de or so).

>
> 3). I have a project for finite element analysis for heterogenous material
> objects. So I have to include material information in both of my element and
> node classes. Isn't that a good or bad way to modify the origianl libmesh
> library or can I extend the original mesh and node classes, or do you have

With the MeshData class, you already have full support of arbitrary nodal
data: You can associate a std::vector<Number> with any node in the mesh,
this should be sufficient...  Consult the documentation in there, and also
the example dwelling on the MeshData class:

  http://libmesh.sourceforge.net/ex12.php

Prior to modifying libmesh, i'd suggest to go and test out this class.
I'd say it already offers everything you need.

However to be honest, file I/O in .unv format currently only supports
nodal data.  Coding the I/O of element-associated data is straightforward,
but not yet done (no need for it).  compare mesh_data_unv_support.C.

File I/O in tetgen format currently supports only read method, but for
_both_ nodes and elements, i think.


-Daniel


>
> -------------------------------------------------------
> This SF.Net email is sponsored by: IBM Linux Tutorials
> Free Linux tutorial presented by Daniel Robbins, President and CEO of
> GenToo technologies. Learn everything from fundamentals to system
> administration.http://ads.osdn.com/?ad_id=1470&alloc_id=3638&op=click
> _______________________________________________
> Libmesh-users mailing list
> Lib...@li...
> https://lists.sourceforge.net/lists/listinfo/libmesh-users
>