Re: OOT - Parallel FEM Language

[David Meeker <dmeeker@xxxxxxxx>]

David Meeker wrote:

> Hi,
>
> I'm sorry for this OOT mail, but I really need your help...
>
> I'm CS graduate student. Currently I'm doing my master theses, it is 
> about to design a parallel FEM language. Unfortunately until now I had 
> not tried any FEM software (maybe next month). My goal is to design an 
> explicitly (not a "heroic" one) parallel language that is suitable for 
> solving FEM problems and convenient enough for programmers. You can 
> give me suggestions, and/or answer the questions below:
> - If there is a good parallel FEM language, what features/spec that 
> language should have ?

http://www.google.com/search?q=parallel%20finite%20element%20language

> - What is the most important feature FEM problem solver 
> application/languageshould have ?

The most important feature (other than being bug-free and providing 
accurate results) is the general ease-of-use. It should be "easy" to 
set up a geometry, automatically generate a mesh, and post-process the 
solution to get some desired type of result. If you are writing a 
parallel finite element language, the user should be insulated from 
having to manage low-level details of how the parallel solver is 
implemented. Anyhow, there are a lot of finite element programs, 
languages, and libraries out there (e.g. see Internet Finite Element 
Resources 
<http://www.engr.usask.ca/%7Emacphed/finite/fe_resources/fe_resources.html>). 
Some specific examples of finite element languages that you might look 
at are:
GetDP http://www.geuz.org/getdp/
FreeFEM++ http://www.ann.jussieu.fr/~hecht/freefem++.htm
FluxPDE http://www.pdesolutions.com/
Although these aren't really parallel implementations, they'll give you 
an idea of what people have put into languages or script interpreters 
meant for solving finite element programs.

> - What features of current FEM Application 
> (AnSys/Calculix/Nastran/OpenAlgor/Lisa/ DesignSpace or others) you 
> consideredvery nice/helpful/important ?

All of these applications either come with an elaborate graphical front 
end or can be plugged into a generic graphical pre/post processor like 
FEMAP or GiD. The detailed process of setting up the finite element 
problem are largely masked from the user. From the user's point of 
view, this is pretty desirable--you just want do draw your part and get 
an analysis with the push of a button. However, creating a good user 
interface is sort of orthogonal to creating a good solver... 

Anyhow, other desirable features, which may require some scripting on 
the part of the user, are the ability to define problems parametrically, 
and the ability to somehow plug into / interface with other analysis 
programs.

Dave.
--
David Meeker
email: dmeeker@xxxxxxxx
www: http://femm.berlios.de/dmeeker