Menu
▾
▴
Re: [Qucs-devel] About the EM simulator
|
From: Bastien R. <rou...@gm...> - 2008-02-21 14:41:17
|
Le jeudi 21 février 2008, Fredy konig a écrit : > Hi all, > > I will go quiet and concentrate on getFEM and MEEP for some time. You could add for solver; SUPER LU (BSD) includig MPI http://crd.lbl.gov/~xiaoye/SuperLU/ UMFPACK (mono) http://www.cise.ufl.edu/research/sparse/umfpack/ > I summarised my goal for the EM simulator in my blog. > http://emctheory.blogspot.com/ Please comment her and send your mind by mail. I do not think for instance that FEM is the best solution. For radiation stuff MOM or hybrid is better. BTW nec is not Mom it is a hybrid code based on MoM Code aster and fenic project have not yet a EM solver For git use tuxfamily. please do not use svn. BTW tuxfamily is better managed than source forge. You are freer. I could open a project on tuxfamily in less than ten minutes. Regards Bastien > I ll keep you posted when I ll get some good results. > regards > Fredy > > > Message du 16/02/08 à 22h45 > > De : "Bastien ROUCARIES" <rou...@gm...> > > A : quc...@li..., fre...@vo... > > Copie à : "dany" <da...@gm...> > > Objet : Re: [Qucs-devel] About the EM simulator > > > > Le samedi 16 février 2008, Fredy konig a écrit : > > > Hi all, > > > > > > I agree with you about doing things in an organised manner. Before I > > > start planning the work I would like to ask few questions: > > > > > > First we should decide which road to take Qt3 or Qt4 > > > I am myself a supporter of using qt4, as I already start to play with > > > it but it may not be the best option. Gopala, Stefan, Bastien and other > > > qucser may think otherwise, please let me know. Qt3 may be easier for a > > > start? but Qt4 may be best for the future of the project? 1) is it > > > better to start on Qt4? or Qt3? do both would be time consuming? is > > > there plan to release soon a stable qt4 version of qucs? > > > > No need to use graphical stuff. The worst and harder thing to get right > > is the core not the graphical stuff. It will be the last thing what you > > will do! > > > > BTW Personnally I prefer to code my geometry, and not draw it. For > > instance I prefer to type I want a sphere radius 1 center 0,0,0, than > > drawing it. And I am not alone. > > > > I think something like a povray like syntax will be really good suited. > > Why not reusing kpovmodeler and use special instruction for EM property ? > > > > > Second, if we use Qt4 the first way to make executable is with Qmake > > > and make. or the autoconfig/automake used in the current version. but I > > > read Gopala input about Cmake, I also read some nice thing about Cmake > > > on a linux journal saying that it is the way forward but never try. > > > 2)CMake ? would it be a good way forward when using Qt4? > > > > Idem. Keep a good insulation between graphical stuff and core. > > For the core autoconf/automake is the way to do. > > > > And BTW use modern development tool like git for creating the project. > > > > > Third, looking at the code of qucs I can see that there is differrent > > > source diretories defining specific block for qucs-attenuattor , > > > qucs-filter, etc... 3)should we add an qucs-ems (electromagnetic > > > simulator) directory and start this as a single block.?is there any > > > restriction or rule to follow to be compatible with the rest of qucs.? > > > > No need tobe compatible. Respect the unix philosphy one program, one > > task. We could always create translator program. > > > > > Forth, About the display of EM output we will probably have to use the > > > openGL libraries 4) Is anybody know where to find good Qt4 openGL > > > examples? > > > > Do not waste your time with gui at this time of project, please. We could > > use something like paraview > > http://www.paraview.org/New/index.html. > > > > Qt support is directly included and it is well suported by big science > > project. It is in the way to be debianized :-) > > > > > Fith, about the EM engine we need to select the computing method TLM, > > > FEM , FDTD, MoMs etc... I suggested to start with c++lib such as nec++ > > > using MoM, other possibilities are YATPAC lib in c using TLM etc... > > > 5)This will need to be discuss more, and I will try to combine all > > > suggestions? but ideas are welcome. > > > > MoM like in momentum is a best thing to do. Nec++ is a good starting > > point but code does not include (to my best knowledge) planar circuit > > computation.It will be good to do that. I strongly think we need to > > develop moment solution. > > > > About FEM, I know some free tool. I expect a lot about fenic project. > > > > http://home.gna.org/getfem/ > > http://www.fenics.org/wiki/FEniCS_Project > > > > FTDT package exist use MEEP http://ab-initio.mit.edu/wiki/index.php/Meep > > > > > 6)Should we also plan for parallel computing like they do on commercial > > > package? I am not sure what open source lib are available for this.. > > > > They exist package what could work. Some of the guys in the uni building > > next to me have coded mumps. > > http://mumps.enseeiht.fr/. It work well and is public domain. Packaging > > could be hard, but I could try to get some contact with the authors. > > > > BTW if we could use this kind of stuff for qucs it will be good :-) > > > > > 7)Finally one important step is to collect related imformation, papers > > > or related project. if you send me your idea I try to publish it in > > > coordinated manner. > > > > I can also publish some paper. I have the chance to have direct access to > > some fellow members. But I will prefer to wait and see. > > About paper I can also collect. > > > > Regards Bastien > > > > > Thanks all > > > and please be patient as it is a big task. > > > Fredy > > > > > > > Message du 15/02/08 à 19h21 > > > > De : "dany" <da...@gm...> > > > > A : quc...@li... > > > > Copie à : > > > > Objet : [Qucs-devel] About the EM simulator > > > > > > > > Hi > > > > > > > > My name is Dany and I've been playing around with qucs quite some > > > > time now. I decided that I want to help push this project forward. > > > > > > > > Specifically I'm interested in the EM simulator, the challenge looks > > > > interesting. > > > > > > > > Markus, Fredy, I'm ready to start on your cue. Do you have something > > > > already? Or is it necessary to start from scratch? > > > > In any case, I think we should try and do it in an organized manner, > > > > meaning, first to draw an outline of what needs to be done, and not > > > > go all rogue programming :) > > > > > > > > Looking forward to hearing from you soon, > > > > Dany > > > > > > > > --------------------------------------------------------------------- > > > >---- This SF.net email is sponsored by: Microsoft > > > > Defy all challenges. Microsoft(R) Visual Studio 2008. > > > > http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ > > > > _______________________________________________ > > > > Qucs-devel mailing list > > > > Quc...@li... > > > > https://lists.sourceforge.net/lists/listinfo/qucs-devel > > > > > > Fredy Konig > > > > > > > > > > > > ----------------------------------------------------------------------- > > >-- This SF.net email is sponsored by: Microsoft > > > Defy all challenges. Microsoft(R) Visual Studio 2008. > > > http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ > > > _______________________________________________ > > > Qucs-devel mailing list > > > Quc...@li... > > > https://lists.sourceforge.net/lists/listinfo/qucs-devel > > > > -- > > > > "ROUCARIES Bastien" > > > > rou...@gm... > > ------------------------------------------------------------------------- > >------ DO NOT WRITE TO rou...@gm... OR BE > > BLACKLISTED > > Fredy Konig -- "ROUCARIES Bastien" rou...@gm... ------------------------------------------------------------------------------- DO NOT WRITE TO rou...@gm... OR BE BLACKLISTED |