Christoph Conrads
Christoph Conrads
@julielangou It's done. Given a pair of matrices `(A, B)`, `xGGQRCS` computes the GSVD such that `A = U1 D1 X`, `B = U2 D2 X`. The tests can be...
The branch `qr+cs-gsvd` with the C++ tests and approximately 90% line coverage for xGGQRCS can be found here: https://travis-ci.org/github/christoph-conrads/lapack/branches
# Benchmarks This post compares the performance of the GSVD solvers xGGSVD3 and xGGQRCS. ## Introduction Of course, the comparison will be very limited because of the huge number of...
> I would worry about one thing: i'm pretty sure the SVD is a little better at identifying the numerical rank than RRQR. I haven't looked into this, but is...
@thijssteel Please let me know if you are satisfied by the changes.
> # Benchmarks > > This post compares the performance of the GSVD solvers xGGSVD3 and xGGQRCS. > > [snip] > > ## Plots > > The plots below compare...
> Nice job @christoph-conrads ! I think we should add tests for the new routines. Maybe we can just replicate the tests for xCGGSVD3. What do you think? There are...
> Would it be reliable enough to specify a convenience function (e.g. for SV-only) that switches between xGGSVD3 and xGGQRCS in a way that (mostly) avoids the doubled time for...
> > There are tests in the christoph-conrads/qr+cs-gsvd branch in [`TESTING/cpp/xGGQRCS_tests.cpp`](https://github.com/christoph-conrads/lapack/blob/qr%2Bcs-gsvd/TESTING/cpp/xGGQRCS_tests.cpp) and the [tests are passing in Travis CI](https://travis-ci.org/github/christoph-conrads/lapack/jobs/759890624#L1241). The tests are not part of the pull request because they...
> > > There are tests in the christoph-conrads/qr+cs-gsvd branch in [`TESTING/cpp/xGGQRCS_tests.cpp`](https://github.com/christoph-conrads/lapack/blob/qr%2Bcs-gsvd/TESTING/cpp/xGGQRCS_tests.cpp) and the [tests are passing in Travis CI](https://travis-ci.org/github/christoph-conrads/lapack/jobs/759890624#L1241). The tests are not part of the pull request because...