You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hi, The svd functions may return fully wrong values on the svn trunk (Here: 0.9.9.2800).
It worked with 0.9.8, still.
Short script to trigger the problem:
from numpy import *
from numpy.linalg import svd
from numpy.random import rand
K = rand(10003 * 2).reshape(10003, 2).astype(float32)
K3 = K[:,0:1]
r1 = svd(K3.copy(), 0)[0]
r2 = svd(K3, 0)[0]
(r1 == r2).all()
The last line should return True, but returns False presently. Note that strangely enough, it starts failing for me at sizes of about 10010. Thanks
The text was updated successfully, but these errors were encountered:
This looks similar to the problem mentioned in #763. The following error from Valgrind 3.1.0 seems to indicate the main culprit (tested with r2813):
==7522== Conditional jump or move depends on uninitialised value(s)
==7522== at 0x4632701: PyArray_Scalar (arrayobject.c:1289)
==7522== by 0x4632EF5: PyArray_Return (arrayobject.c:1444)
==7522== by 0x4696EF6: PyUFunc_GenericReduction (ufuncobject.c:2559)
==7522== by 0x5F15CFC: PyCFunction_Call (in /usr/lib/libpython2.4.so.1.0)
==7522== by 0x5EEB586: PyObject_Call (in /usr/lib/libpython2.4.so.1.0)
==7522== by 0x463D318: PyArray_GenericReduceFunction (arrayobject.c:3211)
==7522== by 0x46669E3: PyArray_All (multiarraymodule.c:781)
==7522== by 0x4666A77: array_all (arraymethods.c:1380)
==7522== by 0x5F15CFC: PyCFunction_Call (in /usr/lib/libpython2.4.so.1.0)
==7522== by 0x5F50E2A: PyEval_EvalFrame (in /usr/lib/libpython2.4.so.1.0)
...
Original ticket http://projects.scipy.org/numpy/ticket/178 on 2006-07-13 by @bthyreau, assigned to unknown.
Hi, The svd functions may return fully wrong values on the svn trunk (Here: 0.9.9.2800).
It worked with 0.9.8, still.
Short script to trigger the problem:
The last line should return True, but returns False presently. Note that strangely enough, it starts failing for me at sizes of about 10010. Thanks
The text was updated successfully, but these errors were encountered: