getfem-users
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Getfem-users] solvers won't solve


From: Michael . Walker
Subject: Re: [Getfem-users] solvers won't solve
Date: Wed, 16 Sep 2009 13:50:51 +0200 (CEST)
User-agent: SquirrelMail/1.4.8-5.el5_3.7

Dear Luis,
I have removed getfem from my system and tried to install the latest
version. I am having the same problem I had back in March, namely that the
gmake goes looking for numarray and calls an error because I have numpy,
which it doesn't find. The first time I installed it Yves Renard emailed
me an archive which was what I built the installation from. So that is the
reason that my installation was not the stable release.

So, what do I need to do to make the latest stable release install and
work properly with numpy, rather than numarray?

Regards

Michael

> Michael,
>
> the 2931 release is not the stable release, and its python
> interface contains a bug (fixed in current release).
> Could you use the latest version and try again?
>
> regards,
> Luis
>
>
> address@hidden escribió:
>> Thanks Luis,
>> According to the getfem.py file it is  __version__ = "$Revision: 2931
>> $".
>> I'm pretty sure its version 3.1, the latest stable release when I
>> downloaded it in early March of this year.
>> Below is the relevant part of the code. I have left out the construction
>> of the mesh m, which is working properly.
>> thanks
>> Michael
>>
>> fem = MeshFem(m,1)
>> fem.set_fem(Fem('FEM_PK(2,1)'))
>>
>> ##################
>> #
>> print "integration method"
>> #
>> ##################
>> mim = MeshIm(m)
>> mim.set_integ(Integ('IM_TRIANGLE(1)'))
>>
>> ##################
>> #
>> print "bricks"
>> #
>> ##################
>> #B = MdBrick('bilaplacian',mim,fem)
>> B = MdBrick('generic_elliptic',mim,fem)
>>
>> m.set_region(41,m.faces_from_cvid(face_trans[50]))
>> m.set_region(42,m.faces_from_cvid(face_trans[51]))
>>
>> C = MdBrick('dirichlet',B,41,fem,'penalized')
>> C.set_param('R',array([0.]) )
>>
>> D = MdBrick('dirichlet',C,42,fem,'penalized')
>> D.set_param('R',array([0.]) )
>>
>>
>> mds = MdState('real')
>> D.solve(mds)
>> state = mds.state()
>>
>>
>>> Hello Michael,
>>>
>>> Could you show me the code of your python-script and your version of
>>> getfem?
>>>
>>> regards,
>>> Luis.
>>>
>>> address@hidden escribió:
>>>
>>>> Hello again,
>>>> I've been poking around the workings of the python interface and found
>>>> that the solvers that getfem couldn't find aren't supported by getfem,
>>>> which is fine but those it can find
>>>> a) just won't solve a simple diffusion equation
>>>> b) but solve it without difficulty on a colleague's computer. The main
>>>> differences between his system and mine are 1) numarray on his and
>>>> numpy
>>>> on mine (yes I did make the necessary changes to the import statement)
>>>> and
>>>> 2) ubuntu on his and fedora on mine.
>>>>
>>>> I have yet to verify if the solvers work outside of getfem, but in the
>>>> meantime, does anyone have any ideas about what may cause a solver to
>>>> fail?
>>>>
>>>> superlu - fails to solve and program crashes with error 'SuperLU
>>>> failed
>>>> to
>>>> solve info=9
>>>> cg/ildlt - doesn't crash but doesn't solve either
>>>> gmres/ilu[t[p]] - program hangs
>>>>
>>>>
>>>> Michael Walker
>>>>
>>>> Plant Modelling Group
>>>> CIRAD, Montpellier
>>>> 04 67 61 57 27
>>>>
>>>> _______________________________________________
>>>> Getfem-users mailing list
>>>> address@hidden
>>>> https://mail.gna.org/listinfo/getfem-users
>>>>
>>>>
>>> _______________________________________________
>>> Getfem-users mailing list
>>> address@hidden
>>> https://mail.gna.org/listinfo/getfem-users
>>>
>>>
>>
>>
>> Michael Walker
>>
>> Plant Modelling Group
>> CIRAD, Montpellier
>> 04 67 61 57 27
>>
>
>


Michael Walker

Plant Modelling Group
CIRAD, Montpellier
04 67 61 57 27



reply via email to

[Prev in Thread] Current Thread [Next in Thread]