No longer maintained. This is the wrong repo, do not use it.
Build numpy wheels with MSVC and pre-built ATLAS binaries.
See appveyor.yml
for the build process as applied on Appveyor.
If you are doing the build manually:
Make a virtualenv to work in, and activate it;
pip install -r requirements.txt
;To build a wheel with old-style Windows platform tags (more compatible with old versions of pip), do
pip install wheel==0.26
;Clone the numpy github repo, checkout the version you want to build, and run the build script:
git clone https://github.com/numpy/numpy.git cd numpy git checkout v1.11.0 python ..\build_numpy.py
This should build you a numpy wheel with the ATLAS binary packaged inside. It
will be in the dist
subdirectory of the numpy source tree.
The appveyor.yml
script uploades the wheels to a Rackspace container at:
- http://58688808cd85529d4031-38dee5dca2544308e91131f21428d924.r12.cf2.rackcdn.com
- https://84c1a9a06db6836f5a98-38dee5dca2544308e91131f21428d924.ssl.cf2.rackcdn.com
The contents via the HTTPS URL seems to get updated more slowly than via the HTTP URL, so if you need the binaries quickly, prefer the HTTP URL.
Also, the Rackspace CDN takes some time to update, so leave at least 15 minutes after the end of a build, before fetching the built binary from the Rackspace URLs above.
ATLAS libraries built laboriously on a dedicated machine with Clint Whaley's help.
lib
and exp
files generated by MSVC lib.exe
tool, as described
here:
http://stackoverflow.com/questions/9946322/how-to-generate-an-import-library-lib-file-from-a-dll