[eigen] Eigen 3.2-beta1 ??

[Gael Guennebaud <gael.guennebaud@xxxxxxxxx>]

Hi list,

I know that none of the major features initially planed for 3.2 are
here (evaluators, indexed sub-matrices, AVX) but 1) it has been 8
months since 3.1, and 2) the devel branch already contains many new
cool and very useful features that, from my point of view, justify a
new release of Eigen. Here is a short list of them:

- Core: The Ref<> class allowing to write non templated function
taking various kind of Eigen dense objects without copies.
- Eigenvalues: RealQZ factorization and GeneralizedEigenSolver
- SparseLU: built-in sparse LU with super-nodes and numerical column pivoting.
- SparseQR: rank-revealing sparse QR factorization (numerical column pivoting).
- OrderingMethods: extended with COLAMD and a unified Ordering API.
- MetisSupport: wrapper to the famous graph partitioning library.
- SPQRSupport: wrapper to suitesparse's supernodal QR solver.
- plus many other small improvements and fixes that could not be backported.

So I think we should make beta release asap. To give us the time to
agree on this and on the list of features that should or not make it,
as well as to perform the necessary API reviews, I propose to schedule
the release on Tuesday 5th of March. During this period everybody is
welcome to:
- discuss this release plan in general (timing, features that should
make it, etc.)
- review the APIs and documentations of these new features (very important)
- help identifying regressions by running the unit tests on exotic
platforms or even testing our code against the devel branch
- and of course documentation and fix patches are always welcome !

I'll start the discussion by saying that in the above list I'm tempted
to move MetisSupport and SPQRSupport in unsupported/ because I
realized that maintaining such wrappers to externals libraries is
pretty time consuming on the long run.


cheers,
Gael