[eigen] Why not Matrix<T, const int rows, const int cols> ??

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


Why is EIGEN restricted to square matrices?
why not support non-square matrices? does it really produce too much
complexity??? i have nonsquare matrices in my small vecmath class (
http://www.coreloop.com/dump/vecmath.h ). its not nearly as mature and
well tested as eigen, but anyway - i have nonsquare matrices and i would
miss them. i really would like to switch to eigen, though...

you could still keep efficiency and algorithms for square matrices via
partial template specialisation ala

template<class T> class Matrix<T,4,4>
{
// all these loop-unrolled fast functions go here
};

while having general functions via for loops in

template<class T, const int rows, const int cols> Matrix
{
...
...
};

in all other aspects i really like eigen. its exactly the complexity
level i like. no template meta magic like in tvmet, just straight but
fast vector matrix stuff.

by the way - is there any code that shows how to implement
pseudo-inverse calculation using eigen?



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