used by those algorithms (aka "second level").
This is a row import : we copy/paste the files from cminpack and make
very few changes :
* template<Scalar> them (replace double)
* dpmpar() replaced by c++ standard equivalent
* abs/fabs/sqrt/min/max replaced by ei_* or std::*
* use eigen norms instead of enorm()
Important Notes:
* The use of stableNorm() was not enough in some cases, but using
blueNorm() instead fixed the problems (some tests gave bad results,
either in number of iterations or precision of the results)
* As a whole, the only test that changed is testNistMGH17() : it now takes
some few steps less to get the same result. So this is a small improvement.
After this commit, the only remaining dependency from the cminpack
static library is 'covar', only used from the tests.
My initial fix was incorrect, the libraries must be quoted when being
passed to the add test macro, but must be unquoted when passed to the
target_link_libraries function.
- R-SVD preconditioning now done with meta selectors to avoid compiling useless code
- SVD options now honored, with options to hint "at least as many rows as cols" etc...
- fix compilation in bad cases (rectangular and fixed-size)
- the check for termination is now done on the fly, no more goto (should have done that earlier!)
and since it was my first try of the patch queue feature I did not
managed to apply it with a good commit message, so here you go:
* Add a ComplexSchur decomposition class built on top of HessenbergDecomposition
* Add a ComplexEigenSolver built on top of ComplexSchur
There are still a couple of FIXME but at least they work for any reasonable matrices,
still have to extend the unit tests to stress them with nasty matrices...
- support complex numbers
- big rewrite of the 2x2 kernel, much more robust
* Jacobi:
- fix weirdness in initial design, e.g. applyJacobiOnTheRight actually did the inverse transformation
- fully support complex numbers
- fix logic to decide whether to vectorize
- remove several clumsy methods
fix for complex numbers
