Weslley S. Pereira

Hi @SzymonNowakowski. Could you try to use my input (https://github.com/Reference-LAPACK/lapack/files/9057573/crashes_svd2.zip) in your machines? Thanks

The matrix I provided before, [crashes_svd2.zip](https://github.com/Reference-LAPACK/lapack/files/9057573/crashes_svd2.zip), had an error. I switched `m` with `n`. Here follows the output using the correct pencil [crashes_svd3.zip](https://github.com/Reference-LAPACK/lapack/files/9059133/crashes_svd3.zip) : ```sh $ gcc test_dgesdd.c -o test_dgesdd...

More tests: 1. Using my C code compiled with gcc version 9.4.0 (Ubuntu 9.4.0-1ubuntu1~20.04.1): - I couldn't reproduce the issue with LAPACK 3.10.0. - I could reproduce the issue in...

More information: I still couldn't reproduce this problem on Windows. On my Linux machine, using your data https://github.com/SzymonNowakowski/DMRnet/blob/testing_branch/data/crashes_svd.RData or, equivalently, https://github.com/Reference-LAPACK/lapack/files/9059133/crashes_svd3.zip: - I can generate D and E (bidiagonal form)...

> Thank you so much for your work with this. So basically you are saying that (apart from my oldish Windows 7 machine) on LAPACK 3.10.0 it is a non-issue....

Thanks for the follow up! I will repeat my tests on those matrices and see if I can get any sense of what is happening.

> This build failure on MacOS seems to be bogus … I did not touch the Makefile build, anyway. Btw., if you are not intending to replace the Makefile build...

> Oh. The -frecursive thing creeps me out. LAPACK is not thread-safe without it? Or are you just not sure? It is a significant difference in build systems if one...

I have some comments about the changes related to "consistent logic for USE_OPTIMIZED_BLAS and USE_OPTIMIZED_LAPACK". As I understood, and please correct me if I am wrong, there are at least...

Maybe we should split your contributions into different PRs to ease discussion with the community. "LAPACK option to be able to just build BLAS" and "consistent logic for USE_OPTIMIZED_BLAS and...