#11674: np.dot of two numeric matrices segfaults sage.
------------------------------+---------------------------------------------
Reporter: mister.wardrop | Owner: jason, jkantor
Type: defect | Status: new
Priority: blocker | Milestone: sage-4.7.1
Component: numerical | Keywords: ATLAS segfault matrix numpy
Work_issues: | Upstream: N/A
Reviewer: | Author:
Merged: | Dependencies:
------------------------------+---------------------------------------------
Changes (by leif):
* cc: vbraun (added)
Comment:
FWIW, I don't get the segfault when running the example with Valgrind
(3.6.1; 3.5.0 raises a `SIGILL` because it doesn't know some instructions)
and the binary distribution of Sage 4.7 (built on / for Ubuntu 10.04.1
x86_64), apparently because Valgrind replaces `malloc()` and `free()`
etc.:
{{{
... lots of messages ...
sage: import numpy as np
sage: ms = np.load("/home/leif/Sage/tmp/matrices.npy")
sage: time np.dot(ms[0],ms[1])
==25274== Invalid read of size 4
==25274== at 0x4EB960E: PyObject_Realloc (obmalloc.c:1154)
==25274== by 0x4EE1572: _PyUnicode_New (unicodeobject.c:262)
==25274== by 0x4EE1ED8: PyUnicodeUCS4_FromUnicode (unicodeobject.c:461)
==25274== by 0x4F1809E: PyEval_EvalFrameEx (ceval.c:4081)
==25274== by 0x4F1926C: PyEval_EvalCodeEx (ceval.c:2968)
==25274== by 0x4F1760C: PyEval_EvalFrameEx (ceval.c:3802)
==25274== by 0x4F1926C: PyEval_EvalCodeEx (ceval.c:2968)
==25274== by 0x4F1760C: PyEval_EvalFrameEx (ceval.c:3802)
==25274== by 0x4F1926C: PyEval_EvalCodeEx (ceval.c:2968)
==25274== by 0x4F1760C: PyEval_EvalFrameEx (ceval.c:3802)
==25274== by 0x4F17E87: PyEval_EvalFrameEx (ceval.c:3792)
==25274== by 0x4F17E87: PyEval_EvalFrameEx (ceval.c:3792)
==25274== Address 0x174f8020 is not stack'd, malloc'd or (recently)
free'd
==25274==
array([[ 0., 0., 0., ..., 0., -0., -0.],
[ 0., 0., 0., ..., 0., -0., -0.],
[ 0., 0., 0., ..., 0., -0., -0.],
...,
[ 0., 0., 0., ..., 0., 0., 0.],
[ 0., 0., 0., ..., 0., 0., 0.],
==25274== Invalid read of size 4
==25274== at 0x4EB9043: PyObject_Free (obmalloc.c:931)
==25274== by 0x4F15CA4: PyEval_EvalFrameEx (ceval.c:1632)
==25274== by 0x4F17E87: PyEval_EvalFrameEx (ceval.c:3792)
==25274== by 0x4F1926C: PyEval_EvalCodeEx (ceval.c:2968)
==25274== by 0x4E9F7CA: function_call (funcobject.c:524)
==25274== by 0x4E73162: PyObject_Call (abstract.c:2492)
==25274== by 0x4F15EC2: PyEval_EvalFrameEx (ceval.c:4019)
==25274== by 0x4F1926C: PyEval_EvalCodeEx (ceval.c:2968)
==25274== by 0x4E9F6CE: function_call (funcobject.c:524)
==25274== by 0x4E73162: PyObject_Call (abstract.c:2492)
==25274== by 0x4E8455E: instancemethod_call (classobject.c:2579)
==25274== by 0x4E73162: PyObject_Call (abstract.c:2492)
==25274== Address 0x363cc020 is 12,256 bytes inside a block of size
12,832 free'd
==25274== at 0x4C262A1: free (vg_replace_malloc.c:366)
==25274== by 0x11FFDEA8: ATL_dmmJIK (in /tmp/sage-4.7-linux-64bit-
ubuntu_10.04.1_lts-x86_64-Linux/local/lib/libatlas.so)
==25274== by 0x11FF5365: ATL_dgemm (in /tmp/sage-4.7-linux-64bit-
ubuntu_10.04.1_lts-x86_64-Linux/local/lib/libatlas.so)
==25274== by 0x32173B57: dotblas_matrixproduct (_dotblas.c:766)
==25274== by 0x4F17591: PyEval_EvalFrameEx (ceval.c:3706)
==25274== by 0x4F1926C: PyEval_EvalCodeEx (ceval.c:2968)
==25274== by 0x4F19341: PyEval_EvalCode (ceval.c:522)
==25274== by 0x4F18618: PyEval_EvalFrameEx (ceval.c:4401)
==25274== by 0x4F1926C: PyEval_EvalCodeEx (ceval.c:2968)
==25274== by 0x4F1760C: PyEval_EvalFrameEx (ceval.c:3802)
==25274== by 0x4F1926C: PyEval_EvalCodeEx (ceval.c:2968)
==25274== by 0x4F1760C: PyEval_EvalFrameEx (ceval.c:3802)
==25274==
[ 0., 0., 0., ..., 0., 0., 0.]])
Time: CPU 3.18 s, Wall: 5.07 s
sage:
}}}
I now really wonder whether the segfault would occur on redhawk (the
machine the binary distribution was built on) itself, and what GCC version
it uses...
--
Ticket URL: <http://trac.sagemath.org/sage_trac/ticket/11674#comment:19>
Sage <http://www.sagemath.org>
Sage: Creating a Viable Open Source Alternative to Magma, Maple, Mathematica,
and MATLAB
--
You received this message because you are subscribed to the Google Groups
"sage-trac" group.
To post to this group, send email to [email protected].
To unsubscribe from this group, send email to
[email protected].
For more options, visit this group at
http://groups.google.com/group/sage-trac?hl=en.
