[Numpy-discussion] Installing numpy with MKL

[Søren Gammelmark]

I feel a bit silly :-)
The compile now appear to work. Sort of. I discovered I had sneaken in a 
"pthreads" in the mkl_libs line, which is obviously not in the 
mkl-directory (I had stolen this line from a stand-alone program I wrote 
at some point). The error message is a bit confusing though - the python 
setup.py-script seems to claim that none of the libraries are there.

In any case the compile works now. However, when I try to run the python 
tests, python crashes because it cannot find libmkl_def.so. This is 
weird because LD_LIBRARY_PATH is set correctly. I also tried compiling 
with LD_RUN_PATH=$MKLROOT/lib/em64t but to no avail. I then tried adding 
mkl_def to the mkl_libs-line and now it works. However, I seem to recall 
reading somewhere that this is not a good idea. I guess that this is 
related to what your referred post was about? Did you end up using the 
patch or have you found a less invasive method since?

Now on to SciPy... :-)

Thankyou for your help
Søren

On 03-08-2010 18:33, Dag Sverre Seljebotn wrote:
> Søren Gammelmark wrote:
>    
>> Hi everyone
>>
>> I realize that this e-mail has been written a lot of times before. I
>> have, however, not been able to make it work even though I have looked
>> through a lot of e-mails on the mailinglist. I apologize if the
>> solution is written neatly somewhere.
>>      
> In my experience, it is not.
>
> I got this working, my config etc. is in this thread (site.cfg is in my
> second post):
>
> http://www.mail-archive.com/numpy-discussion@scipy.org/msg23151.html
>
> If you try that config and post the exact error you get if it does not
> work, perhaps we'll get somewhere...
>
> Dag Sverre
> _______________________________________________
> NumPy-Discussion mailing list
> NumPy-Discussion at scipy.org
> http://mail.scipy.org/mailman/listinfo/numpy-discussion
>