[SciPy-user] scipy 0.7.0.dev4373 + atlas FAILED (failures=2, errors=12)

[Xavier Gnata]

Robert Kern wrote:
> On Fri, May 16, 2008 at 4:58 PM, Xavier Gnata <xavier.gnata at gmail.com> wrote:
>
>   
>> Here it is. Ok it is long but I cannot cut it without removing an
>> interesting part :
>>     
>
> Well, the parts interesting to me are the ones starting here:
>
>   
>> ======================================================================
>> ERROR: Failure: ImportError
>> (/usr/lib/python2.5/site-packages/scipy/linalg/clapack.so: undefined
>> symbol: clapack_sgesv)
>> ----------------------------------------------------------------------
>>     
>
> This looks like an ATLAS build problem. Specifically, it looks like
> you didn't make a full LAPACK for ATLAS. I'm not current on the build
> process for recent ATLASes. I *think* that if you pass the correct
> --with-netlib-lapack to ./configure, it will correctly merge a full
> LAPACK. But it might not. Instructions for doing this manually are
> here:
>
>   http://svn.scipy.org/svn/scipy/trunk/INSTALL.txt
>
> This does not involve recompiling ATLAS.
>
> But you can probably save yourself a lot of trouble by just installing
> the atlas3-base (and possibly atlas3-sse or atlas3-sse2 if you have
> those capabilities) and their corresponding -dev packages. It's an old
> version, but provides a reasonable speedup for the effort expended.
>
>   
>> ======================================================================
>> FAIL: test_texture2 (test_segment.TestSegment)
>> ----------------------------------------------------------------------
>> Traceback (most recent call last):
>>  File
>> "/usr/lib/python2.5/site-packages/scipy/ndimage/tests/test_segment.py",
>> line 152, in test_texture2
>>    assert_array_almost_equal(tem0, truth_tem0, decimal=6)
>>  File "/usr/lib/python2.5/site-packages/numpy/testing/utils.py", line
>> 255, in assert_array_almost_equal
>>    header='Arrays are not almost equal')
>>  File "/usr/lib/python2.5/site-packages/numpy/testing/utils.py", line
>> 240, in assert_array_compare
>>    assert cond, msg
>> AssertionError:
>> Arrays are not almost equal
>>
>> (mismatch 66.6666666667%)
>>  x: array([  0.00000000e+00,   0.00000000e+00,   0.00000000e+00,
>>         0.00000000e+00,   0.00000000e+00,   0.00000000e+00,
>>         1.91816598e-01,   1.02515288e-01,   9.30087343e-02,...
>>  y: array([ 0.        ,  0.        ,  0.        ,  0.        ,  0.        ,
>>        0.        ,  0.13306101,  0.08511007,  0.05084148,  0.07550675,
>>        0.4334695 ,  0.03715914,  0.00289055,  0.02755581,  0.48142046,
>>        0.03137803,  0.00671277,  0.51568902,  0.01795249,  0.49102375,
>> 1.        ], dtype=float32)
>>     
>
> This might be an actual bug in the (heavily in-development) image
> segmentation code.
>
>   
>> ======================================================================
>> FAIL: test_pbdv (test_basic.TestCephes)
>> ----------------------------------------------------------------------
>> Traceback (most recent call last):
>>  File
>> "/usr/lib/python2.5/site-packages/scipy/special/tests/test_basic.py",
>> line 368, in test_pbdv
>>    assert_equal(cephes.pbdv(1,0),(0.0,0.0))
>>  File "/usr/lib/python2.5/site-packages/numpy/testing/utils.py", line
>> 139, in assert_equal
>>    assert_equal(actual[k], desired[k], 'item=%r\n%s' % (k,err_msg),
>> verbose)
>>  File "/usr/lib/python2.5/site-packages/numpy/testing/utils.py", line
>> 145, in assert_equal
>>    assert desired == actual, msg
>> AssertionError:
>> Items are not equal:
>> item=1
>>
>>  ACTUAL: 1.0
>>  DESIRED: 0.0
>>     
>
> Not sure about this one. Probably a real bug.
>
>   
Thanks for our support :)
I used to use the atlas unbutu package but I would like to understand 
how to compile ATLAS and to see if there is a real performance 
improvement or not. As you said,it should note be that large...

I have merge the lib using ar as explain in your link.
Now I get this :
FAILED (failures=5, errors=5)

****************************************************************
WARNING: clapack module is empty
-----------
See scipy/INSTALL.txt for troubleshooting.
Notes:
* If atlas library is not found by numpy/distutils/system_info.py,
  then scipy uses flapack instead of clapack.
****************************************************************

...........................FF.............NO ATLAS INFO AVAILABLE
.........................................
****************************************************************
WARNING: cblas module is empty

 It looks there still something wrong isn't it?

xavier