[SciPy-user] umfpack and building scipy from source

Nils Wagner nwagner at iam.uni-stuttgart.de
Fri Dec 1 04:30:20 EST 2006


Robert Cimrman wrote:
> Ryan Krauss wrote:
>   
>> Following Robert Kern's advice I am getting very close.  The API error
>> is now gone.
>>
>> There is my current scipy.test() output.
>>
>> In [4]: scipy.test()
>> Warning: FAILURE importing tests for <module 'scipy.linsolve.umfpack.umfpack' fr
>> om '...y\\linsolve\\umfpack\\umfpack.pyc'>
>> C:\Python24\Lib\site-packages\scipy\linsolve\umfpack\tests\test_umfpack.py:17: A
>> ttributeError: 'module' object has no attribute 'umfpack' (in ?)
>>     
>  > ...
>
> This is ok, since you do not have umfpack.
>
> r.
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
>   
I recommend UMFACK. It's faster than SuperLU.

SuperLU solution
Use minimum degree ordering on A'+A.
CPU time SuperLU 0.32002
Residual SuperLU 3.87138066926e-06

UMFPACK solution
CPU time UMFPACK 0.032002
Residual UMFPACK 2.88360989563e-07

Nils




More information about the SciPy-User mailing list