Hi there,

sometimes Voigt notation it is useful to express self-adjoint symmetric 3x3 matrix (2nd rank symmetric tensor) as a 6-vector, so that it can be multiplied easily with self-adjoint 6x6 matrix (representing 4th rank symmetric tensor).

I am impatiently waiting for self-adjoint matrices with compact storage; Voigt view would be perhaps a nice extra thing that I could try to implement when that part is stabilized (as far as I understand SpecialMatrix, this is work in progress right now)? Would it fit in the framework? The weight factor for some elements could be perhaps introduced via a template parameter of the view (like 1 or Sqrt2).

What do you think? Or would I be better off using n-dimensional arrays (perhaps from ndarray, so that 4th-rank tensor would be 3x3x3x3 "symmetric" matrix, 2nd rank 3x3 symmetric etc? How about compact storage then?

Cheers, Vaclav

Hi,

the compact storage we are planing to implement won't help much at all in your case because we plan to implement the rectangular full packed format (http://netlib.org/lapack/lapack-3.2.htm ... ked_format) which allows for efficient "blas level 3" (aka matrix-matrix) operations.