Solve several problems at once using DGGLSE

[rbulle]

Hello,

I am currently working on constrained minimization problems and I'm particularly interested in solving a set of problems as follow

        min || Ax_i - b_i ||_2
subject to: Bx_i = d_i         forall i in [[1,...,N]]

where {b_i}_i and {d_i}_i are families of N vectors and A and B are independent of i.

I was wondering: is it possible to solve such a set of problems using a single generalized RQ decomposition with LAPACK ?
I know that LAPACK DGGLSE can do it for a single problem only (N=1). I took a look at the code of DGGLSE and I could'nt see where the limitation preventing $N$ to be greater than one was but I am not very familiar with Fortran so maybe I missed something...

Thank you for your wonderful work!
Best,
Raphaël

[langou]

I guess we were waiting on someone to request the feature. :)

I think you are right: it should not be too hard to get DGGLSE to work on multiple right hand sides.

That would be an interface change, so maybe we call the new routine xGGLSE_NRHS or something like this.

[rbulle]

That would be perfect! Thanks for the quick response! :)