Dear CCX enthusiasts,
as part of a recently merged change to our use of third party code in CCX, the fallback LAPACK routines previously distributed with CCX for users who want to build without providing their own LAPACK library has been removed.
If you build CCX on your own and previously relied on the fallbacks, you will have to adapt your build scripts to link an existing LAPACK installation on your machine, see also Guido’s adaption of his Makefile. Note that there are optimized implementations available from the major hardware vendors’ math libraries, e.g. as part of MKL for Intel or ROCm for AMD, as well as a reference implementation on netlib. Linux users may also easily install LAPACK from their distribution’s package repositories.
The rationale behind the removal of the fallbacks is primarily reducing maintenance and code deduplication on our part as well as encouraging (the arguably small number of) developers who rely on them to link against optimized implementations.
If you encounter issues with your builds I invite you to ask for assistance in the “Code development” section of the forum.
Best regards
Christoph