[Ipopt] IPOPT Binary Status and linear solver choices.

Ruhollah Tavakoli rohtav at gmail.com
Fri Feb 20 08:31:30 EST 2009


First of all, if you are insisted to use windows, why do not use cygwin or
mingw, ipopt works well under these shells, and also there are some trick to
use precompiled libraries them into MSVC projects (just notice that using
cygwin has performance penalty, in my case ~ 50% slower). Also notice that
under windows you do not have chance of using Pardiso, as there is
neither any cygwin/mingw version nor native windows complers like
MSVC/intel. It seems that tis developper do not attend to it (i personally
send couple of request for it, but without at least any response ;) )

Also, not that there are lot of nonlinear programming codes works
cross-platform, so IPOPT may not be the only solution. But i should admit
that i found IPOPT the most *robust* (not efficient) amoung various codes i
have used (drawback is at least memory which is matter for large problems).
e.g. ALGENCAN (fortran 77) is also of value to try.

2  your question is ambigious for me, but take a look at this page (i may
helps), "sintef":

3  regarding to qudrature i do not understand your enquiry well, but to my
understanding, "Deal II" library is an OOP FEM library with arbitrary order
quadrature on logically rectangular (hypercube) grid (support qudtree/octree
refinement), Also FEniCS project
(http://www.fenics.org/wiki/FEniCS_Project) provide
an automatic variational compiler and of course support qudrature
integration (support only *affine* mappings). Both of these not easy to use
with native win, but easy with cygwin.

hope be helpful


On Fri, Feb 20, 2009 at 4:16 PM, Jesse Perla <jesseperla at gmail.com> wrote:

>  > Maybe I need to bite the bullet and do my own build on windows...
>> Also good idea, if you can find a fortran compiler...
> I will give it a shot with Intel Fortran.  But the instructions in the
> visual studio project files say to use f2c if I remember?  Am I better off
> just compiling that part of the solution with Fortran?
>> >> > 2) If it is linked in, any chance of getting a binary build with the
>> >> No chance ;-). It's a headache to build on Windows, if I find time,
>> >> maybe I'll do it.So why you use it? ;-)
> A very reasonable question.
> Answer 1: I am addicted to the visual studio debugger/dev environment and
> can't find anything even remotely close for linux.  gdb is utterly useless
> for mediocre developers like me.
> Answer 2: Now that I am using cmake for crossplatform, I don't need to use
> the idiotic visual studio project files and figure I can do development on a
> different platform than the cluster I will deploy on.
> Answer 3: A lot of my fellow economists are using Windows, and trying to
> sell coauthors/etc. using my libraries to move to linux is a hard sell.
> Another question:
> I am planning to get BonMin up and running as well at some point.  Looking
> at the Bonmin-1.0.1 distribution, it looks like they are distributing an
> older copy of ipopt.  Can I setup ipopt-3.5.5 and then use it with this
> version of bonmin?  Or should I build the one they have in the distribution.
> And if people aren't sick of me yet, can I ask a few general optimization
> questions?:
> 1) What open source optimizer do you guys suggest for minimizing nonlinear
> constrained continuous problems where: I cannot calculate a derivative
> analytically, nor with auto-differentiation, it probably won't be convex,
> and my function will probably be expensive to calculate?  The number of
> policy variables would be in the order of 2-50.  The DFO on COIN doesn't
> seem to be very active and is in Fortran.  Is it still the right one to use?
>  What about the algorithms in
> http://www-unix.mcs.anl.gov/tao/docs/manual/manual.html or http://trilinos.sandia.gov/packages/moocho/
>   <http://trilinos.sandia.gov/packages/moocho/>And for some of my
> problems, the calculation of the function will be REALLY expensive (~30
> seconds per evaluation), so if there is a good optimizer out there that can
> distribute function evaluations and anything else intelligent to overcome
> this, I would love it.
> 2)  Do you guys have any idea of a good open source interpolator
> (cubicsplines hopefully) written for arbitrary dimension N?  I don't mind if
> it is C or Fortran since I have basically given up on a C++ one.  This seems
> like something people have done a million times and I can't find out for the
> life of me.
> 3) On the same vein, do you guys know of a good open general quadrature
> library for arbitrary dimension N?  C or Fortran is fine here as well.
> Thanks for all your help and these great libraries.
> _______________________________________________
> Ipopt mailing list
> Ipopt at list.coin-or.org
> http://list.coin-or.org/mailman/listinfo/ipopt
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://list.coin-or.org/pipermail/ipopt/attachments/20090220/05976baa/attachment-0001.html 

More information about the Ipopt mailing list