Yes, that is correct.
First, I think that LAPACK95 interface is no longer supported. But more importantly, it does not allow the full features of the language to be used. The explicit interfaces only allow a subset of the functionality that could be achieved if LAPACK were written in modern fortran. That will probably always be the case as long as LAPACK is intended to be used from multiple languages. That might be the compromise that we have to live with, but it is a compromise nonetheless.
Suggested solution?
If anyone wants to work with me to use Modern Fortran (with MoA) to formulate a tensor SVD we could
start to help folks know that they have something built on formality and modern design.
For instance?
I don’t think this exists in LAPACK.
1 Like
Correct.
That’s why I want to help create it.
Though not sure, I guess some tensor network libraries in modern Fortran might have such a routine, for example this one (I’ve just found it from Google, so not sure if it suits the purpose). Table 1 seems to have “SVDtensor”.
TNSPackage: A Fortran2003 library designed for tensor network state methods
Thanks, I’ll check it out.