[SciPy-User] NumPy/SciPy compilation VC10/Intel Fortran/MKL - Architecture question

[David Cournapeau]

On Thu, Dec 4, 2014 at 9:42 AM, Relja Ljubobratovic <
relja.ljubobratovic at 3lateral.com> wrote:

> The first thing I would look at is ensureing all the .lib/.obj linked
>> together in the failed linking step are at the same architecture. It seems
>> that you are using the correct ifort and cl compilers, but you never know.
>
>
> Thank you, David, for providing information and advice.
>
> I did thorough inspection of all libraries and objects that are involved
> in failed linking and it seems that they are all compiled using same
> architecture setup (x64). Although I found that missing symbols needed for
> linking of object files dct.obj and dst.obj. What troubles me now, is that
> these external symbols used here, are named with the underscore (e.g.
> dsinqb_), but those symbols are present in fftpack.lib as without
> underscore. This is my first encounter with fortran compilation, so I'm
> unsure of this, but I've read that fortran functions that are wrapped to c,
> receive the following underscore (like lapacke functions, but I'm only
> guessing). Is it possible that some sort of name mangling is needed for
> compilation or linking of those objects?
>

Yes, that may be the cause. I suspect you are one of the first person to
look into compiling with those tools.

The fortran mangling is unfortunately compiler-dependent: either the
mangling in recent version of ifort changed (unlikely), or we do not detect
it correctly with this version of ifort in numpy/scipy (more likely).

David
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20141204/2c57d66f/attachment.html>