The default installation uses the provided numarray.linear_algebra.lapack_lite implementation of these routines and this works without any further interaction.
Nevertheless if LAPACK is installed already or you are concerned about the performance of these routines you should consider installing numarray.linear_algebra to take advantage of the real LAPACK library. See the next section for instructions.
On some platforms, precompiled optimized versions of the LAPACK and BLAS libraries are preinstalled on the operating system, and the setup procedure needs to be modified to force the lapack_lite module to be linked against those rather than the builtin replacement functions.
Here's a recipe for building using LAPACK:
% setenv USE_LAPACK 1 % setenv LINALG_LIB <where your lapack, blas, atlas, etc are> % setenv LINALG_INCLUDE <where your lapack, blas, atlas headers are> % python setup.py install --selftest
For your particular system and library installations, you may need to edit addons.py and adjust the variables sourcelist, lapack_dirs, and lapack_libs.
Note: A frequent request is that somehow the maintainers of Numerical Python invent a procedure which will automatically find and use the best available versions of these libraries. We welcome any patches that provide the functionality in a simple, platform independent, and reliable way. The scipy project has done some work to provide such functionality, but is probably not mature enough for use by numarray yet.
Send comments to the NumArray community.