NumPy is a general-purpose array-processing package designed to
efficiently manipulate large multi-dimensional arrays of arbitrary
records without sacrificing too much speed for small multi-dimensional
arrays. NumPy is built on the Numeric code base and adds features
introduced by numarray as well as an extended C-API and the ability
to create arrays of arbitrary type which also makes NumPy suitable for
interfacing with general-purpose data-base applications.
There are also basic facilities for discrete fourier transform, basic
linear algebra and random number generation.
If you need to build numpy for debugging, set DEBUG=y. If you use
software which is having problems with numpy's new relaxed strides
checking, set NPY_RSC=0.
It is highly recommended to install libraries implementing BLAS and
LAPACK before installing numpy. You may choose between:
a) BLAS and LAPACK (reference but unoptimized and thus slow)
b) OpenBLAS (optimized, provides LAPACK too)
c) ATLAS and LAPACK (optimized), good to read README.ATLAS
All these are available on SlackBuilds.org.
If you want to use the UMFPACK library instead of SuperLU to solve
unsymmetric sparse linear systems, then run this Slackbuild with
NO_UMFPACK set to "no" and then install scikit-umfpack on top of
scipy. In this context, UMFPACK is an optional dependency for
numpy. Nevertheless, note that presently scikit-umfpack is not
available on SlackBuilds.org while its dependencies are.
NOTE: If you use this SlackBuild, numpy will run with the python version
provided by Slackware Linux, which is presently 2.7.xx. If you'd
like to use python 3.x then you have to install numpy with the
numpy-legacy3 SlackBuild.
IMPORTANT: This version, 1.8.2, is the latest to include the oldnumeric
and numarray compatibility modules. Starting with version
1.9.0 these modules got removed by the numpy developers.
THUS: This SlackBuild conflicts with the numpy SlackBuild
which installs versions >= 1.9.0!