Hello,
I've read in the documentation here at the bottom
http://eigen.tuxfamily.org/dox/QuickRefPage.html
that I can convert a full square symmetric matrix with

selfadjointView<Eigen::Upper>()

into a triangular upper matrix. Is there a way to define a matrix from the beginning on as a triangular matrix, so that I don't have to initialize al it's entries, just the diagonal and the upper part?
In my case I have a (constant) 3x3 symmetric Matrix and I guess it's not worth to declare it as MatrixX3d when I don't need all entries.
Thanks
matse

There is no compact triangular storage yet. Anyway, for such small matrices a compact storage would more likely slowdown the performance without clear gain on the memory footprint (unless you have to store million of them...)

OK, thanks for your reply, just wanted to be sure that I don't waste even a single bit of memory space, but since this really just a small matrix it does not matter.
matse

I am actually using thousands of symetrical MatrixXd (around 50*50)
Do you plan to write such a compact triangular storage to gain memory?

There is currently no concrete plan to implement it, little time for too many things to do. So unless someone (you?) step by to start it, it won't be implemented soon.