[Numpy-discussion] Building Windows binaries on OS X

David Cournapeau cournape@gmail....
Mon Feb 8 19:54:15 CST 2010

On Mon, Feb 8, 2010 at 9:14 PM, Ralf Gommers
<ralf.gommers@googlemail.com> wrote:
> Hi David and all,
> I have a few questions on setting up the build environment on OS X for
> Windows binaries. I have Wine installed with Python 2.5 and 2.6, MakeNsis
> and MinGW. The first question is what is meant in the Paver script by "cpuid
> plugin". Wine seems to know what to do with a cpuid instruction, but I can
> not find a plugin. Searching for "cpuid plugin" turns up nothing except the
> NumPy pavement.py file. What is this?

That's a small NSIS plugin to detect at install time the exact
capabilities of the CPU (SSE2, SSE3, etc...). The sources are found in
tools/win32build/cpucaps, and should be built with mingw (Visual
Studio is not supported, it uses gcc-specific inline assembly). You
then copy the dll into the plugin directory of nsis.

> Second question is about Fortran. It's needed for SciPy at least, so I may
> as well get it right now. MinGW only comes with g77, and this page:
> http://www.scipy.org/Installing_SciPy/Windows says that this is the default
> compiler. So Fortran 77 on Windows and Fortran 95 on OS X as defaults, is
> that right? No need for g95/gfortran at all?

gc 4.x is still not officially supported by MinGW. Gfortran is
incompatible with g77, so care should be taken if we change it for
NumPy (every extension using f2py will likely be broken as a result on

Gfortran is the only option on win64, so I was thinking about making
the transition to gfortran once we manage to build Numpy and Scipy
with it on win64.

> Final question is about Atlas and friends. Is 3.8.3 the best version to
> install? Does it compile out of the box under Wine? Is this page
> http://www.scipy.org/Installing_SciPy/Windows still up-to-date with regard
> to the Lapack/Atlas info and does it apply for Wine?

Atlas 3.9.x should not be used, it is too unstable IMO (it is a dev
version after all, and windows receives little testing compared to
unix). I will put the Atlas binaries I am using somewhere - building
Atlas is already painful, but building it with a limited architecture
on windows takes it to a whole new level (it is not supported by
atlas, you have to patch the build system by yourself).



More information about the NumPy-Discussion mailing list