wrathematics / float

Single precision (float) matrices for R.
Other
46 stars 12 forks source link

Do not hardcode LAPACK any flags #51

Closed jeroen closed 1 month ago

jeroen commented 8 months ago

This should fix errors like this: https://github.com/r-universe/cran/actions/runs/7619266054/job/20756957057. The problem here is that the LAPACK configuration on the server that has built float could be different than on the user machine.

The portable way for an R package to link to LAPACK/BLAS is PKG_LIBS= $(LAPACK_LIBS) $(BLAS_LIBS) $(FLIBS) in the Makevars, verbatim. This works on any platform. There is no for float to try to resolve this.