[IPython-dev] Numpy with OpenBLAS vs. IPython
Florian M. Wagner
wagnerfl at student.ethz.ch
Wed Oct 30 04:46:11 EDT 2013
I reproduced the error with the latest IPython source:
fwagner at fwagner-notebook ~/Downloads/ipython $ ipython
Python 2.7.4 (default, Sep 26 2013, 03:20:26)
Type "copyright", "credits" or "license" for more information.
IPython 2.0.0-dev -- An enhanced Interactive Python.
? -> Introduction and overview of IPython's features.
%quickref -> Quick reference.
help -> Python's own help system.
object? -> Details about 'object', use 'object??' for extra details.
In [1]: A = np.random.randn(50,50)
In [2]: A[0,0]
Out[2]: 0.62268531821589967
In [3]: A.T.dot(A)
Speicherzugriffsfehler
Compared to:
Python 2.7.4 (default, Sep 26 2013, 03:20:26)
[GCC 4.7.3] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> import numpy as np
>>> A = np.random.randn(50,50)
>>> A.T.dot(A)
array([[ 4.61866860e+01, 1.97714469e+00, 8.78842173e+00, ...,
-9.65143259e+00, 4.74312877e+00, -1.85163266e+01],
[ 1.97714469e+00, 4.50113072e+01, -1.13105045e+01, ...,
2.31604581e+00, 4.93068512e+00, 1.29316267e+01],
[ 8.78842173e+00, -1.13105045e+01, 6.02859864e+01, ...,
-2.65040891e+00, 3.98374437e+00, -1.54860323e+01],
...,
[ -9.65143259e+00, 2.31604581e+00, -2.65040891e+00, ...,
3.81774984e+01, 1.63768677e+00, 1.16681961e+01],
[ 4.74312877e+00, 4.93068512e+00, 3.98374437e+00, ...,
1.63768677e+00, 5.11159138e+01, 6.56039753e-02],
[ -1.85163266e+01, 1.29316267e+01, -1.54860323e+01, ...,
1.16681961e+01, 6.56039753e-02, 6.61061915e+01]])
Am 30.10.2013 09:33, schrieb Florian M. Wagner:
> Thanks guys,
>
> @Julian: Reducing the thread number did not help
> @Thomas: dap is mpl-basemap dependency, fixed the import issue, but the
> error still occurs
>
> It is really weird that it is occuring only when printing the array, or
> slicing the array. And whats more weird, that it works in plain Python
> as I understood that IPython has no numpy dependency?
>
> Thanks and regards
> Florian
>
> Am 29.10.2013 19:39, schrieb Julian Taylor:
>> does this also happen if you do this before starting ipython?
>>
>> export OPENBLAS_NUM_THREADS=1
>> (assuming bash shell)
>>
>> On 29.10.2013 17:19, Florian M. Wagner wrote:
>>> Dear all,
>>>
>>> I recently compiled numpy with OpenBlas for faster matrix-matrix
>>> multiplication. Everything works fine, but IPython seems to dislike it.
>>> I get a Segmentation Fault error when trying this (I reinstalled IPython
>>> 1.1.0 after numpy compilation):
>>>
>>> fwagner at fwagner-notebook ~ $ ipython -c "import numpy as np; print
>>> np.__version__; import IPython; print IPython.__version__; A =
>>> np.random.randn(600, 600); print A[20,53]; print A"
>>> /usr/local/bin/ipython:5: UserWarning: Module dap was already
>>> imported from None, but /usr/lib/python2.7/dist-packages is being
>>> added to sys.path
>>> from pkg_resources import load_entry_point
>>> 1.9.0.dev-54d3559
>>> 1.1.0
>>> -2.1004055886
>>> Speicherzugriffsfehler
>>>
>>> In comparison, this works fine (just python instead of ipython):
>>>
>>> fwagner at fwagner-notebook ~ $ python -c "import numpy as np; print
>>> np.__version__; import IPython; print IPython.__version__; A =
>>> np.random.randn(600, 600); print A[20,53]; print A"
>>> 1.9.0.dev-54d3559
>>> 1.1.0
>>> -1.10282239744
>>> [[ 0.13798084 -0.39195796 -0.15309986 ..., 1.47471232 -0.9158455
>>> 0.97060671]
>>> [ 0.25720169 0.51895478 0.55175787 ..., -0.02010573 0.78047754
>>> -1.07891961]
>>> [ 1.59478156 0.81245486 -0.03959768 ..., -0.04816659 -0.03356907
>>> -0.11860379]
>>> ...,
>>> [-0.06299817 -0.0956477 -0.32987518 ..., -1.47714386 1.66736916
>>> -0.47704276]
>>> [-0.93769589 1.50006542 0.97782803 ..., -0.5034154 1.31592011
>>> -0.74482995]
>>> [-0.24661336 -0.98979706 -0.39366363 ..., 0.20507632 0.8701609
>>> -1.10449957]]
>>>
>>> Any ideas what might cause this problem?
>>>
>>> Thanks,
>>>
>>> Florian
>>>
>>>
>>>
>>> _______________________________________________
>>> IPython-dev mailing list
>>> IPython-dev at scipy.org
>>> http://mail.scipy.org/mailman/listinfo/ipython-dev
>>>
> _______________________________________________
> IPython-dev mailing list
> IPython-dev at scipy.org
> http://mail.scipy.org/mailman/listinfo/ipython-dev
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/ipython-dev/attachments/20131030/d2cd662f/attachment.html>
More information about the IPython-dev
mailing list