Bug#581761: libatlas3gf-base illegal instructions

[Sylvestre Ledru]

Hello,

> 
> I'm being affected by this bug in lenny, with libatlas3gf-base
> v3.6.0-22. It's making octave unusable for me -
I am sorry but this bug is unlikely to be fixed since Squeeze should be
released pretty soon and that Atlas packages have changed a lot since
this release.

Sylvestre





-- 
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org

Bug#581761: libatlas3gf-base illegal instructions

[grev...@gmail.com]

Hello

I'm being affected by this bug in lenny, with libatlas3gf-base
v3.6.0-22. It's making octave unusable for me -

gdb octave
run
octave:1> A=[2 4;3 1];
octave:2> b = [5;2];
octave:3> A\b

Program received signal SIGILL, Illegal instruction.
[Switching to Thread 0xb53cdab0 (LWP 2415)]
0xb5b0546e in ATL_diamax_xp1yp0aXbX () from /usr/lib/atlas/libblas.so.3gf
(gdb) x/4xb 0xb5b0546e
0xb5b0546e : 0xdf0xf10x770x0c

A quick check in the latest Intel IA-32 instruction set manual
(p3-366) shows that DF F1 is an FCOMIP instruction for the x87.

A google search reveals this instruction was introduced in the Pentium
Pro architecture. My laptop is a Pentium 1 MMX -

cat /proc/cpuinfo
processor   : 0
vendor_id   : GenuineIntel
cpu family  : 5
model   : 4
model name  : Pentium MMX
stepping: 4
cpu MHz : 165.788
cache size  : 0 KB
fdiv_bug: no
hlt_bug : no
f00f_bug: yes
coma_bug: no
fpu : yes
fpu_exception   : yes
cpuid level : 1
wp  : yes
flags   : fpu vme de pse tsc msr mce cx8 mmx
bogomips: 332.52
clflush size: 32
power management:

I'm running the i486 kernel, and don't have problems with the rest of
Lenny. Is this a code generation configuration issue?



-- 
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org

Bug#581761: libatlas3gf-base: Illegal instructions .... dgemm

[Yaroslav Halchenko]

For those interested -- I've rebuilt atlas manually in a clean cow chroot
on a native architecture -- illegal instruction issue is gone!

On Mon, 26 Jul 2010, Yaroslav Halchenko wrote:

> Package: libatlas3gf-base
> Version: 3.8.3-24
> Severity: normal
-- 
  .-.
=--   /v\  =
Keep in touch// \\ (yoh@|www.)onerussian.com
Yaroslav Halchenko  /(   )\   ICQ#: 60653192
   Linux User^^-^^[17]




signature.asc
Description: Digital signature

Bug#581761: libatlas3gf-base: Illegal instructions .... dgemm

[Yaroslav Halchenko]

and here is an easy way to reproduce it with even different function as
a victim:

*$> gdb  --args /usr/bin/python2.6 -c 'import scipy; scipy.test()'
GNU gdb (GDB) 7.1-debian
Copyright (C) 2010 Free Software Foundation, Inc.
License GPLv3+: GNU GPL version 3 or later 
This is free software: you are free to change and redistribute it.
There is NO WARRANTY, to the extent permitted by law.  Type "show copying"
and "show warranty" for details.
This GDB was configured as "x86_64-linux-gnu".
For bug reporting instructions, please see:
...
Reading symbols from /usr/bin/python2.6...Reading symbols from 
/usr/lib/debug/usr/bin/python2.6...done.
done.
(gdb) c
The program is not being run.
(gdb) r
Starting program: /usr/bin/python2.6 -c import\ scipy\;\ scipy.test\(\)
[Thread debugging using libthread_db enabled]
Running unit tests for scipy
NumPy version 1.4.1
NumPy is installed in /usr/lib/pymodules/python2.6/numpy
SciPy version 0.7.2
SciPy is installed in /usr/lib/python2.6/dist-packages/scipy
Python version 2.6.5+ (release26-maint, Jul  6 2010, 12:58:20) [GCC 4.4.4]
nose version 0.11.1
[New Thread 0x7fffee0a8710 (LWP 23988)]
/usr/lib/python2.6/dist-packages/scipy/linsolve/__init__.py:4: 
DeprecationWarning: scipy.linsolve has moved to scipy.sparse.linalg.dsolve
  warn('scipy.linsolve has moved to scipy.sparse.linalg.dsolve', 
DeprecationWarning)
.
Program received signal SIGILL, Illegal instruction.
0x75a9e327 in ATL_zaxpy_xp1yp1aXbX () from /usr/lib/libblas.so.3gf
(gdb) bt 10
#0  0x75a9e327 in ATL_zaxpy_xp1yp1aXbX () from /usr/lib/libblas.so.3gf
#1  0x7fffec37b38b in zgesl_ () from 
/usr/lib/python2.6/dist-packages/scipy/integrate/vode.so
#2  0x7fffec36ad9a in zvsol_ () from 
/usr/lib/python2.6/dist-packages/scipy/integrate/vode.so
#3  0x7fffec36f815 in zvnlsd_ () from 
/usr/lib/python2.6/dist-packages/scipy/integrate/vode.so
#4  0x7fffec372f61 in zvstep_ () from 
/usr/lib/python2.6/dist-packages/scipy/integrate/vode.so
#5  0x7fffec37a6ec in zvode_ () from 
/usr/lib/python2.6/dist-packages/scipy/integrate/vode.so
#6  0x7fffec34d838 in f2py_rout_vode_zvode (capi_self=, capi_args=, capi_keywds=, 
f2py_func=)
at build/src.linux-x86_64-2.6/scipy/integrate/vodemodule.c:1424
#7  0x0041ef47 in PyObject_Call (func=, 
arg=, kw=) at 
../Objects/abstract.c:2492
#8  0x004a6215 in ext_do_call (f=
Frame 0x280b8b0, for file 
/usr/lib/python2.6/dist-packages/scipy/integrate/ode.py, line 554, in run 
(self=, success=1, first_step=, ml=None, iwork=, 
rwork=, nsteps=500, mu=None, 
with_jacobian=0, rtol=, zwork=, call_args=[, , 1, 1, , , , 11], atol=, min_step=, meth=1, order=12) at remote 
0x27c6590>, args=(, , , , 
, (), (...))), throwflag=) at 
../Python/ceval.c:4063
#9  PyEval_EvalFrameEx (f=
Frame 0x280b8b0, for file 
/usr/lib/python2.6/dist-packages/scipy/integrate/ode.py, line 554, in run 
(self=, success=1, first_step=, ml=None, iwork=, 
rwork=, nsteps=500, mu=None, 
with_jacobian=0, rtol=, zwork=, call_args=[, , 1, 1, , , , 11], atol=, min_step=, meth=1, order=12) at remote 
0x27c6590>, args=(, , , , 
, (), (...))), throwflag=) at 
../Python/ceval.c:2452
(More stack frames follow...)


On Mon, 26 Jul 2010, Yaroslav Halchenko wrote:

> ah... and test_plt.py provided in original report works fine for me (via
> forwarded X to suspect box) ...
> so it could be a separate issue... sorry for not testing right away

-- 
  .-.
=--   /v\  =
Keep in touch// \\ (yoh@|www.)onerussian.com
Yaroslav Halchenko  /(   )\   ICQ#: 60653192
   Linux User^^-^^[17]





-- 
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org

Bug#581761: libatlas3gf-base: Illegal instructions .... dgemm

[Yaroslav Halchenko]

ah... and test_plt.py provided in original report works fine for me (via
forwarded X to suspect box) ...
so it could be a separate issue... sorry for not testing right away

On Mon, 26 Jul 2010, Yaroslav Halchenko wrote:

> In both cases (different python modules tested) it seems to boil to (sorry 
> that
> I used python2.5 at first... the same with 2.6)

> #1  0x73303626 in ATL_dmmJIK2 () from /usr/lib/libblas.so.3gf
> #2  0x73304082 in ATL_dmmJIK () from /usr/lib/libblas.so.3gf
> #3  0x732f7ad4 in ATL_dgemm () from /usr/lib/libblas.so.3gf
> #4  0x73703ae8 in ?? () from 
> /usr/lib/pymodules/python2.5/numpy/core/_dotblas.so

-- 
  .-.
=--   /v\  =
Keep in touch// \\ (yoh@|www.)onerussian.com
Yaroslav Halchenko  /(   )\   ICQ#: 60653192
   Linux User^^-^^[17]





-- 
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org

Bug#581761: libatlas3gf-base: Illegal instructions .... dgemm

[Yaroslav Halchenko]

Package: libatlas3gf-base
Version: 3.8.3-24
Severity: normal


I think I started to experience similar simptoms recently.  Our testbox started
throwing "illegal instructions".  It started to happen after recent upgrades.
Before providing this information I've updated to current sid but nothing
changed... installed also libatlas3gf-amd64sse3 which became used... the
same... 

In both cases (different python modules tested) it seems to boil to (sorry that
I used python2.5 at first... the same with 2.6)

#1  0x73303626 in ATL_dmmJIK2 () from /usr/lib/libblas.so.3gf
#2  0x73304082 in ATL_dmmJIK () from /usr/lib/libblas.so.3gf
#3  0x732f7ad4 in ATL_dgemm () from /usr/lib/libblas.so.3gf
#4  0x73703ae8 in ?? () from 
/usr/lib/pymodules/python2.5/numpy/core/_dotblas.so

Ill Ins started to pop out after following updates which included atlas:

2010-07-15 09:59:22 upgrade fsl-atlases 4.1.5-1 4.1.6-1
2010-07-15 09:59:22 status half-configured fsl-atlases 4.1.5-1
2010-07-15 09:59:22 status unpacked fsl-atlases 4.1.5-1
2010-07-15 09:59:22 status half-installed fsl-atlases 4.1.5-1
2010-07-15 09:59:24 status half-installed fsl-atlases 4.1.5-1
2010-07-15 09:59:29 status unpacked fsl-atlases 4.1.6-1
2010-07-15 09:59:29 status unpacked fsl-atlases 4.1.6-1
2010-07-15 10:00:43 upgrade libatlas3gf-base 3.6.0-24 3.8.3-24
2010-07-15 10:00:43 status half-configured libatlas3gf-base 3.6.0-24
2010-07-15 10:00:43 status unpacked libatlas3gf-base 3.6.0-24
2010-07-15 10:00:44 status half-installed libatlas3gf-base 3.6.0-24
2010-07-15 10:00:44 status half-installed libatlas3gf-base 3.6.0-24
2010-07-15 10:00:45 status unpacked libatlas3gf-base 3.8.3-24
2010-07-15 10:00:45 status unpacked libatlas3gf-base 3.8.3-24
2010-07-15 10:54:30 configure fsl-atlases 4.1.6-1 4.1.6-1
2010-07-15 10:54:30 status unpacked fsl-atlases 4.1.6-1
2010-07-15 10:54:30 status half-configured fsl-atlases 4.1.6-1
2010-07-15 10:54:30 status installed fsl-atlases 4.1.6-1
2010-07-15 10:54:41 configure libatlas3gf-base 3.8.3-24 3.8.3-24
2010-07-15 10:54:41 status unpacked libatlas3gf-base 3.8.3-24
2010-07-15 10:54:41 status half-configured libatlas3gf-base 3.8.3-24
2010-07-15 10:54:41 update-alternatives: run with --install 
/usr/lib/libblas.so.3gf libblas.so.3gf /usr/lib/atlas-base/atlas/libblas.so.3gf 
35 --slave /usr/lib/libcblas.so.3gf libcblas.so.3gf 
/usr/lib/atlas-base/libcblas.so.3gf --slave /usr/lib/libf77blas.so.3gf 
libf77blas.so.3gf /usr/lib/atlas-base/libf77blas.so.3gf --slave 
/usr/lib/libatlas.so.3gf libatlas.so.3gf /usr/lib/atlas-base/libatlas.so.3gf 
--slave /usr/lib/liblapack_atlas.so.3gf liblapack_atlas.so.3gf 
/usr/lib/atlas-base/liblapack_atlas.so.3gf
2010-07-15 10:54:41 update-alternatives: link group libblas.so.3gf updated to 
point to /usr/lib/atlas-base/atlas/libblas.so.3gf
2010-07-15 10:54:41 update-alternatives: run with --install 
/usr/lib/liblapack.so.3gf liblapack.so.3gf 
/usr/lib/atlas-base/atlas/liblapack.so.3gf 35
2010-07-15 10:54:41 update-alternatives: link group liblapack.so.3gf updated to 
point to /usr/lib/atlas-base/atlas/liblapack.so.3gf
2010-07-15 10:54:41 status installed libatlas3gf-base 3.8.3-24

ATM they look like following while testing nitime

Program received signal SIGILL, Illegal instruction.
0x732e1899 in ATL_dJIK52x52x52TN52x52x0_a1_b0 () from 
/usr/lib/libblas.so.3gf
(gdb) bt 30
#0  0x732e1899 in ATL_dJIK52x52x52TN52x52x0_a1_b0 () from 
/usr/lib/libblas.so.3gf
#1  0x73303626 in ATL_dmmJIK2 () from /usr/lib/libblas.so.3gf
#2  0x73304082 in ATL_dmmJIK () from /usr/lib/libblas.so.3gf
#3  0x732f7ad4 in ATL_dgemm () from /usr/lib/libblas.so.3gf
#4  0x73703ae8 in ?? () from 
/usr/lib/pymodules/python2.5/numpy/core/_dotblas.so
#5  0x0048fef6 in call_function (f=0x24990f0, throwflag=) at ../Python/ceval.c:3612
#6  PyEval_EvalFrameEx (f=0x24990f0, throwflag=) at 
../Python/ceval.c:2304
#7  0x004915f9 in PyEval_EvalCodeEx (co=0xe85eb8, globals=, locals=, args=, 
argcount=21632, kws=0x0, kwcount=0, defs=0x0, defcount=0, closure=0x0)
at ../Python/ceval.c:2875
#8  0x004dac19 in function_call (func=0xe8c398, arg=0x2457a28, kw=0x0) 
at ../Objects/funcobject.c:517
#9  0x00418493 in PyObject_Call (func=0x2552f60, arg=0x1a0, 
kw=0x25337e0) at ../Objects/abstract.c:1861
#10 0x0042009f in instancemethod_call (func=0xe8c398, arg=0x2457a28, 
kw=0x0) at ../Objects/classobject.c:2519
#11 0x00418493 in PyObject_Call (func=0x2552f60, arg=0x1a0, 
kw=0x25337e0) at ../Objects/abstract.c:1861
#12 0x004617a8 in call_maybe (o=, name=, nameobj=, format=0x4f28fa "(O)") at 
../Objects/typeobject.c:967
#13 0x00462eb4 in slot_nb_multiply (self=0x2486d70, other=0x2486668) at 
../Objects/typeobject.c:4393
#14 0x0041913f in binary_op1 (v=0x2486d70, w=0x2486668, op_slot=16) at 
../Objects/abstract.c:398
#15 0x00419c68 in PyNumber_Multiply (v=0x2552f60, w=0x1a0) at 
../Objects/abstract.c:669
#16