Re: [eigen] compile-time multidimensional arrays

[ Thread Index | Date Index | More Archives ]

Hi Christoph,

thanks for pointing me at the TensorFixedSize class. The proposed multidim has more power on slicing/viewing the multidimensional data manipulating size/steps pairs. At the same time multidim storage is based on Eigen Matrix while Tensor starts from scratch (C++ array of T).

Also for the offset computation I played a bit with metaprogramming, while in Tensor it seems to be relying in inlining via instantiation of std::array (linearizedIndex).

Best Regards,


Il 21/07/15 11:10, Christoph Hertzberg ha scritto:

I did not have a look at your source yet, but can you tell if it significantly differs from the existing (unsupported) Tensor code (namely TensorFixedSize)? The existing Tensor code is far from being feature-complete or stable at the moment, so any feature requests or API changes are at least open for debate.


On 21.07.2015 at 09:37, Emanuele Ruffaldi wrote:

Dear Eigen members,

I have drafted a compile-time multidimensional array class based on
C++11 parameter packs, using Eigen for underlying math operations.

It would be interesting to have some comments on the approach taken:



Best Regards,
Emanuele Ruffaldi

Emanuele Ruffaldi, Ph.D.     Assistant Professor
PERCRO, TeCIP Institute      Scuola Superiore Sant'Anna
via Luigi Alamanni 13D       San Giuliano Terme 56010 (PI), Italy
mob.: +39 340 46 72 468      tel.: +39 050 882 508
fax.: +39 050 882 564        skype: pititaly

Mail converted by MHonArc 2.6.19+