Denis Demidov

> How about just adding the example you had to the examples? I'll try to find time to do this soon. Or, if you are up to it, a PR...

Hello Dominic, I think that eigen choose option 3 actually: their matrices are compile-time convertible to arrays. Then, operator*() has linear algebra meaning for matrices and elementwise meaning for arrays....

Hi Pascal, What operations would you support for multidimensional arrays? Are there any having sense for multidimensional matrices, except for elementwise arithmetics, element access and probably slicing (I am not...

Ah, the number of dimensions is compile-time property of boost::multi_array. That makes things easier.

I think most of these would have to be implemented as standalone functions.

I think `vex::multi_array` should be single-device-only. This way we may easily introduce slicing operations, transpositions, etc. What do you think of the following interface: ``` C++ std::vector dims = {100,...

Yes, I like this better.

Here is an example of a dense matrix-matrix product implementation: ``` .cpp vex::vector A(ctx, n1 * n2); vex::vector B(ctx, n2 * n3); vex::vector C(ctx, n1 * n3); // C =...

Nice paper: [A C++11 implementation of arbitrary-rank tensors for high-performance computing](http://arxiv.org/abs/1209.1003).

I still don't understand what do you mean by `If we do take Eigen`. VexCL is not built on top of Eigen, and VexCL can not interface with Eigen (except...