Re: [eigen] questions and remarks to automatic differentiation based on adol-c

[ Thread Index | Date Index | More lists.tuxfamily.org/eigen Archives ]


Hi,

Sorry for not having replied to your email earlier! Sometimes emails
get lost in the near-constant traffic we get here. Don't hesitate to
ping us much sooner --- no need to wait that long.

It's not clear from your e-mail whether you are talking about 2.0 or
the development branch. The adol-c module is only in the development
branch, but strangely enough, some of your errors below refer to 2.0 !

> A) Using the LU module:
> --------------------------------------------------------------------
> error message:
> <PATH TO EIGEN>/Eigen/src/LU/LU.h:514: error: no matching function for call
> to ‘ei_isMuchSmallerThan(const adtl::adouble&, adtl::adouble&, const
> adtl::adouble&)’

Ah, so here you are talking about Eigen 2.0. In Eigen 2, you have to
specialize all the math functions, including ei_isMuchSmallerThan, for
your scalar type. Since Eigen2 doesn't have any adol-c support you're
on your own :-)

>
>
> B) Using the Choleskey module:
> --------------------------------------------------------------------
> there was missing something, too

2.0 or development branch?

>
>
> C) in AdolcForward:
> --------------------------------------------------------------------
> 1)..................................
> in
> template<> struct NumTraits<adtl::adouble>
> {
> typedef adtl::adouble Real;
> typedef adtl::adouble NonInteger;
> typedef adtl::adouble Nested;
> enum {
> ...... (**)
> };
> };
>
> " HasFloatingPoint = 1" is missing at (**)

HasFloatingPoint has been replaced by IsInteger (which means the opposite).

>
> 2)..................................
> The comment suggests (line 75)
> #include <unsupported/Eigen/AdolcSupport>
> but AdolCSupport does not exist.

Aaaah... it's getting clear now what happened: you've been trying to
use our adolc-support from Eigen3 (development branch) against Eigen2.

Benoit

>
> ______________________________________________________________________________________________________________________________________
>
> --
> Christian Reinl                       reinl@xxxxxxxxxxxxxxxxxxx
> Simulation, Systems Optimization      Phone:  +49 (0) 6151-16-5212
> and Robotics                          Fax:    +49 (0) 6151-16-6648
> Technische Universitaet Darmstadt
> Hochschulstr. 10
> D-64289 Darmstadt, Germany            http://www.sim.tu-darmstadt.de
>
>
>
>



Mail converted by MHonArc 2.6.19+ http://listengine.tuxfamily.org/