Christoph Conrads

> You'd need to add an "int" after the "info" parameter for each "char*" argument in the list. This depends on the compiler and in the case of gfortran also...

> I do not know what to do about it since I do not know whether the code relies on the P or not. The computed beta values are used...

> I see... If one wants to fix all of them, we would need to rewrite a number of algorithms with no real improvement in practice. I couldn't think of...

> I see... If one wants to fix all of them, we would need to rewrite a number of algorithms with no real improvement in practice. I couldn't think of...

> Thanks, @christoph-conrads! That was easy to implement. I followed your suggestions and created a new commit. I am now applying the fixes to xORBDB1, and then I will look...

Hi @epsilon-0, > Some things which I had in mind for allowing this to co-exist with the standard installation - name the libraries to libblas64.so, libcblas64.so, liblapack64.so, liblapacke64.so, this way...

Here is the motivation for the new GSVD solver from #63: > The generalized singular value decomposition can be computed directly with xGGSVD, xGGSVD3 (see [LAWN 46](http://www.netlib.org/lapack/lawnspdf/lawn46.pdf)) or by computing...

> I would like to have feedback on the computation of the matrix R. Currently it is always computed but this is expensive because of an additional matrix-matrix multiplication followed...

> TODO: > * [x] Re-read [LAWN 46](http://www.netlib.org/lapack/lawnspdf/lawn46.pdf) and understand `xGGSVD3()`algorithm The whole theory behind `xGGSVD3()` is based on simultaneous transformations of a pair of upper triangular matrices.

The latest `xGGQRCS` code contains row scaling (`{s,d}LASRTR`) in an attempt to handle matrix pencils where the input matrices `A`, `B` differ significantly in norm. This does not even yield...