You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
This should be fairly easy to do on Linux by updating the manylinux image. Likewise on macOS, we can likely compile something at build time (compile something because otherwise we might end up with mismatched MACOSX_TARGET_VERSION from a brew-installed version). Windows might be a little tougher.
The text was updated successfully, but these errors were encountered:
For Linux, where would the BLAS/LAPACK library come from at runtime? I guess this would require bundling the BLAS/LAPACK shared libraries with Cantera or explicitly linking to a static version of BLAS/LAPACK? Or perhaps piggybacking off of whatever library Numpy bundles, assuming contains all the functions we need.
For macOS, I think the easiest and best option is to just link to the Accelerate framework.
where would the BLAS/LAPACK library come from at runtime
Wheels are always "repaired" so that links are rewritten to a bundled, uniquely named, copy of non-libc and non-libpython dependencies. This is how we're already including libhdf, libfmt, etc.
This should be fairly easy to do on Linux by updating the manylinux image. Likewise on macOS, we can likely compile something at build time (compile something because otherwise we might end up with mismatched
MACOSX_TARGET_VERSION
from a brew-installed version). Windows might be a little tougher.The text was updated successfully, but these errors were encountered: