Pseudocode:
pip install [pkg] --global-option="path/to/vecLib/headers"
Is this the "wrong" way to compile numpy (and eventually Scipy) against Apple's accelerate/vecLib on macOS M1? If so--why is this strategy worse than other commands to install numpy against Apple's Accelerate and vecLib libraries?
python3 -m pip install numpy --no-build-isolation --global-option="build_ext" --global-option="-I/System/Library/Frameworks/Accelerate.framework/Versions/Current/Frameworks/vecLib.framework/Headers" --global-option="-Wl,-framework" --global-option="-Wl,Accelerate"
i.) Is pointing to these local libraries explicitly in the 'pip install' command above not the most logical way to ensure they are found during installation?
ii.) If so: is this the correct way to do so? Alternatively, is the above command redundant, since the numpy version matching 'arm64' may automatically look for these libraries to link?
iii.) Assuming official support for Accelerate from Scipy and other modules, would this method hold true for other modules as well?
iv.) If the answers to the questions above are all yes, how could I do this with the '--config settings' flag, since pip 24.2 will enforce deprecation of --global-option?
When I then type into a python terminal:
import numpy as np
np.show_config(mode='dicts')['Build Dependencies']['blas']
I get:
{'name': 'accelerate', 'found': True, 'version': 'unknown', 'detection method': 'system', 'include directory': 'unknown', 'lib directory': 'unknown', 'openblas configuration': 'unknown', 'pc file directory': 'unknown'}
And in the following line:
np.show_config(mode='dicts')['Build Dependencies']['lapack']
gives:
{'name': 'dep4335520544', 'found': True, 'version': '1.26.4', 'detection method': 'internal', 'include directory': 'unknown', 'lib directory': 'unknown', 'openblas configuration': 'unknown', 'pc file directory': 'unknown'}
So it seems to have linked correctly? Thanks in advance for feedback--SO is invaluable to us amateurs.