1

We're aiming to compile SciPy (and UMFPACK) with the Intel Compiler to use Intel's (multi-threaded) MKL BLASs, having already done this successfully for NumPy. With the help of various websites, we've achieved this for UMFPACK and SciPy without any apparent error messages. However, we suspect that the resulting SciPy is not finding our UMFPACK and instead is silently using much slower alternative functions (SuperLU).

Could anyone help us with (a) how to determine whether SciPy is finding our UMFPACK, and (b) how to point SciPy, whether at the compilation or execution stage, to our UMFPACK library?

user2736013
  • 11
  • 1
  • would you mind sharing details on how you got the NumPy compilation to work? I've been at it for several days, and still can't successfully `import numpy as np; np.test()`... – MattDMo Sep 17 '13 at 15:48
  • These instructions worked for me http://software.intel.com/en-us/articles/numpyscipy-with-intel-mkl. In site.cfg be sure to edit the path to your Intel MKL installation if it is not at the default location. Also, I am using numpy-1.7.0 and found that it was not necessary to edit intel.py. – user2736013 Sep 23 '13 at 07:30

0 Answers0