8:07 p.m.
Dear Python/OpenCL community,
I am pretty new (py)opencl and encountered a problem, maybe it a lack of understanding of
openCL, but I found strange python seg-faults:
test program:
#!/usr/bin/python
import numpy, pyopencl
ctx = pyopencl.create_some_context()
data=numpy.random.random((1024,1024)).astype(numpy.float32)
img = pyopencl.image_from_array(ctx, ary=data, mode="r", norm_int=False,
num_channels=1)
print img
System: debian sid: pyopencl2012.1 (the same code works with debian stable and v2011.2)
Here is the backtrace obtained with GDB:
0x0000000000000000 in ?? ()
(gdb) bt
#0 0x0000000000000000 in ?? ()
#1 0x00007ffff340c253 in pyopencl::create_image_from_desc(pyopencl::context const&,
unsigned long, _cl_image_format const&, _cl_image_desc&,
boost::python::api::object) () from /usr/lib/python2.7/dist-packages/pyopencl/_cl.so
#2 0x00007ffff342de36 in _object*
boost::python::detail::invoke<boost::python::detail::install_holder<pyopencl::image*>,
pyopencl::image* (*)(pyopencl::context const&, unsigned long, _cl_image_format
const&, _cl_image_desc&, boost::python::api::object),
boost::python::arg_from_python<pyopencl::context const&>,
boost::python::arg_from_python<unsigned long>,
boost::python::arg_from_python<_cl_image_format const&>,
boost::python::arg_from_python<_cl_image_desc&>,
boost::python::arg_from_python<boost::python::api::object>
>(boost::python::detail::invoke_tag_<false, false>,
boost::python::detail::install_holder<pyopencl::image*> const&, pyopencl::image*
(*&)(pyopencl::context const&, unsigned long, _cl_image_format const&,
_cl_image_desc&, boost::python::api::object),
boost::python::arg_from_python<pyopencl::context const&>&,
boost::python::arg_from_python<unsigned long>&,
boost::python::arg_from_python<_cl_image_format const&>&,
boost::python::arg_from_python<_cl_image_desc&>&,
boost::python::arg_from_python<boost::python::api::object>&) () from
/usr/lib/python2.7/dist-packages/pyopencl/_cl.so
#3 0x00007ffff342e06f in
boost::python::detail::caller_arity<5u>::impl<pyopencl::image*
(*)(pyopencl::context const&, unsigned long, _cl_image_format const&,
_cl_image_desc&, boost::python::api::object),
boost::python::detail::constructor_policy<boost::python::default_call_policies>,
boost::mpl::vector6<pyopencl::image*, pyopencl::context const&, unsigned long,
_cl_image_format const&, _cl_image_desc&, boost::python::api::object>
>::operator()(_object*, _object*) ()
from /usr/lib/python2.7/dist-packages/pyopencl/_cl.so
#4 0x00007ffff311715b in boost::python::objects::function::call(_object*, _object*) const
()
from /usr/lib/libboost_python-py27.so.1.49.0
#5 0x00007ffff3117378 in ?? () from /usr/lib/libboost_python-py27.so.1.49.0
#6 0x00007ffff3120593 in
boost::python::detail::exception_handler::operator()(boost::function0<void>
const&) const ()
from /usr/lib/libboost_python-py27.so.1.49.0
#7 0x00007ffff3445983 in
boost::detail::function::function_obj_invoker2<boost::_bi::bind_t<bool,
boost::python::detail::translate_exception<pyopencl::error, void (*)(pyopencl::error
const&)>, boost::_bi::list3<boost::arg<1>, boost::arg<2>,
boost::_bi::value<void (*)(pyopencl::error const&)> > >, bool,
boost::python::detail::exception_handler const&, boost::function0<void>
const&>::invoke(boost::detail::function::function_buffer&,
boost::python::detail::exception_handler const&, boost::function0<void>
const&) () from /usr/lib/python2.7/dist-packages/pyopencl/_cl.so
#8 0x00007ffff3120373 in
boost::python::handle_exception_impl(boost::function0<void>) ()
from /usr/lib/libboost_python-py27.so.1.49.0
#9 0x00007ffff3115635 in ?? () from /usr/lib/libboost_python-py27.so.1.49.0
Thanks for your help.
If you are not able to reproduce this bug, I should mention it to debian.
Cheers,
--
Jérôme Kieffer
Data analysis unit - ESRF
9:54 p.m.
Jerome Kieffer <Jerome.Kieffer(a)esrf.fr> writes:
Dear Python/OpenCL community,
I am pretty new (py)opencl and encountered a problem, maybe it a lack of understanding of
openCL, but I found strange python seg-faults:
test program:
#!/usr/bin/python
import numpy, pyopencl
ctx = pyopencl.create_some_context()
data=numpy.random.random((1024,1024)).astype(numpy.float32)
img = pyopencl.image_from_array(ctx, ary=data, mode="r", norm_int=False,
num_channels=1)
print img
Can you try the git version? Your code works fine for me with
that. (Also, on what implementation are you running?)
Andreas
9:56 p.m.
Jerome Kieffer <Jerome.Kieffer(a)esrf.fr> writes:
Dear Python/OpenCL community,
I am pretty new (py)opencl and encountered a problem, maybe it a lack of understanding of
openCL, but I found strange python seg-faults:
test program:
#!/usr/bin/python
import numpy, pyopencl
ctx = pyopencl.create_some_context()
data=numpy.random.random((1024,1024)).astype(numpy.float32)
img = pyopencl.image_from_array(ctx, ary=data, mode="r", norm_int=False,
num_channels=1)
print img
Ah, no--just remembered what the issue is. You'll have to recompile
PyOpenCL with
CL_PRETEND_VERSION = "1.1"
in siteconf.py until we put in a real fix for this.
Andreas
2:37 p.m.
On Mon, 15 Oct 2012 17:56:01 -0400
Andreas Kloeckner <lists(a)buster.tiker.net> wrote:
Jerome Kieffer <Jerome.Kieffer(a)esrf.fr> writes:
> Dear Python/OpenCL community,
>
> I am pretty new (py)opencl and encountered a problem, maybe it a lack of
understanding of openCL, but I found strange python seg-faults:
> test program:
>
> #!/usr/bin/python
> import numpy, pyopencl
> ctx = pyopencl.create_some_context()
> data=numpy.random.random((1024,1024)).astype(numpy.float32)
> img = pyopencl.image_from_array(ctx, ary=data, mode="r", norm_int=False,
num_channels=1)
> print img
Ah, no--just remembered what the issue is. You'll have to recompile
PyOpenCL with
CL_PRETEND_VERSION = "1.1"
in siteconf.py until we put in a real fix for this.
With this fix, it works ... thanks
May be we should warn the debian team ?
--
Jérôme Kieffer
Data analysis unit - ESRF
7:58 a.m.
Dnia 2012-10-15, pon o godzinie 22:07 +0200, Jerome Kieffer pisze:
Dear Python/OpenCL community,
I am pretty new (py)opencl and encountered a problem, maybe it a lack of understanding of
openCL, but I found strange python seg-faults:
test program:
Do you use Debian PyOpenCL packages, or have you compiled PyOpenCL
yourself?
From your description it looks like you are using NVIDIA drivers
(which do not support OpenCL 1.2) with Khronos OpenCL headers
(probably from Debian package) which define OpenCL 1.2 features.
Best regards.
--
Tomasz Rybak <tomasz.rybak(a)post.pl> GPG/PGP key ID: 2AD5 9860
Fingerprint A481 824E 7DD3 9C0E C40A 488E C654 FB33 2AD5 9860
http://member.acm.org/~tomaszrybak
10:25 p.m.
Hi Jerome,
Jerome Kieffer <Jerome.Kieffer(a)esrf.fr> writes:
I am pretty new (py)opencl and encountered a problem, maybe it a lack
of understanding of openCL, but I found strange python seg-faults:
test program:
#!/usr/bin/python
import numpy, pyopencl
ctx = pyopencl.create_some_context()
data=numpy.random.random((1024,1024)).astype(numpy.float32)
img = pyopencl.image_from_array(ctx, ary=data, mode="r", norm_int=False,
num_channels=1)
print img
System: debian sid: pyopencl2012.1 (the same code works with debian stable and v2011.2)
This should be fixed in current git.
Andreas
4:32 p.m.
On Wed, 30 Jan 2013 17:25:57 -0500
Andreas Kloeckner <lists(a)buster.tiker.net> wrote:
This should be fixed in current git.
Hi Andreas,
I was wondering if this bug re-appeared ...
I just checked out v2013.2 and I did not manage to have a system with all my devices
operational:
ii amd-opencl-icd-legacy:amd64 8.97.100.7-3~bpo70+1 amd64
AMD OpenCL ICD (legacy)
ii intel-opencl-icd 3.2.1.16712-2 amd64
OpenCL* runtime for Intel® CPU device
ii mic-opencl-icd 3.2.1.16712-2 amd64
OpenCL* runtime for Intel® Xeon Phi™ coprocessor device
ii nvidia-opencl-common 331.20-1 amd64
NVIDIA OpenCL driver
ii nvidia-opencl-icd:amd64 331.20-1 amd64
NVIDIA OpenCL ICD
ii ocl-icd-libopencl1:amd64 1.3-3 amd64
Generic OpenCL ICD Loader
ii opencl-headers 1.2-2012.04.18a-1 all
OpenCL (Open Computing Language) header files
ii python-pyopencl 2013.2-1 amd64
Python wrapper for OpenCL
Is your trick of CL_PRETEND_VERSION = '1.1' still working ?
I have the opencl version 1.2 for both the header & library on the
computer.
All devices (AMD, Intel, MIC) can be used except the Nvidia one which
crashes with a segfault when one creates a context on it.
Do you have any suggestion? thanks in advance.
Cheers,
--
Jérôme Kieffer
tel +33 476 882 445
8:39 p.m.
I am answering myself ...
I downgraded the OpenCL headers to 1.1 and installed the nvidia libopencl1 library which
is limited to v1.1.
But now I am no more able to compile pyopencl which fails at the link time with:
gcc -pthread -fno-strict-aliasing -fwrapv -Wall -DNDEBUG -fPIC -DPYGPU_PACKAGE=pyopencl
-DPYGPU_PYOPENCL=1 -DPYOPENCL_PRETEND_CL_VERSION=4112 -DPYOPENCL_USE_DEVICE_FISSION=1
-DHAVE_GL=1 -I/usr/lib/pymodules/python2.7/numpy/core/include -I/usr/include/python2.7 -c
src/wrapper/bitlog.cpp -o build/temp.linux-x86_64-2.7/src/wrapper/bitlog.o
g++ -pthread -shared -Wl,-Bsymbolic-functions -Wl,-z,relro
build/temp.linux-x86_64-2.7/src/wrapper/wrap_cl.o
build/temp.linux-x86_64-2.7/src/wrapper/wrap_cl_part_1.o
build/temp.linux-x86_64-2.7/src/wrapper/wrap_cl_part_2.o
build/temp.linux-x86_64-2.7/src/wrapper/wrap_constants.o
build/temp.linux-x86_64-2.7/src/wrapper/wrap_mempool.o
build/temp.linux-x86_64-2.7/src/wrapper/bitlog.o -lboost_python-py27 -lOpenCL -o
build/lib.linux-x86_64-2.7/pyopencl/_cl.so
/usr/bin/ld: cannot find -lOpenCL
collect2: error: ld returned 1 exit status
error: command 'g++' failed with exit status 1
Now clinfo (which uses this same library) works a bit and complains before segfaulting:
$ clinfo
clinfo: /usr/lib/x86_64-linux-gnu/libOpenCL.so.1: no version information available
(required by clinfo)
Number of platforms: 3
Platform Profile: FULL_PROFILE
Platform Version: OpenCL 1.1 CUDA 6.0.1
Platform Name: NVIDIA CUDA
Platform Vendor: NVIDIA Corporation
Platform Extensions: cl_khr_byte_addressable_store cl_khr_icd cl_khr_gl_sharing
cl_nv_compiler_options cl_nv_device_attribute_query cl_nv_pragma_unroll
Platform Profile: FULL_PROFILE
Platform Version: OpenCL 1.2 AMD-APP (937.2)
Platform Name: AMD Accelerated Parallel Processing
Platform Vendor: Advanced Micro Devices, Inc.
Platform Extensions: cl_khr_icd cl_amd_event_callback cl_amd_offline_devices
Platform Profile: FULL_PROFILE
Platform Version: OpenCL 1.2 LINUX
Platform Name: Intel(R) OpenCL
Platform Vendor: Intel(R) Corporation
Platform Extensions: cl_khr_icd cl_khr_global_int32_base_atomics
cl_khr_global_int32_extended_atomics cl_khr_local_int32_base_atomics
cl_khr_local_int32_extended_atomics cl_khr_byte_addressable_store cl_khr_spir cl_khr_fp64
Platform Name: NVIDIA CUDA
Number of devices: 1
Device Type: CL_DEVICE_TYPE_GPU
Device ID: 4318
[...]
Global memory size: 5032706048
Constant buffer size: 65536
Max number of constant args: 9
Local memory type: Scratchpad
Local memory size: 49152
Segmentation fault
If I replace this library with the one from AMD or the open-source one, pyopencl compiles
but seg-faults when creating an Nvidia context.
--
Jérôme Kieffer
Data analysis unit - ESRF
8:01 p.m.
Dnia 2014-02-13, czw o godzinie 21:39 +0100, Jerome Kieffer pisze:
I am answering myself ...
I downgraded the OpenCL headers to 1.1 and installed the nvidia libopencl1 library which
is limited to v1.1.
But now I am no more able to compile pyopencl which fails at the link time with:
gcc -pthread -fno-strict-aliasing -fwrapv -Wall -DNDEBUG -fPIC -DPYGPU_PACKAGE=pyopencl
-DPYGPU_PYOPENCL=1 -DPYOPENCL_PRETEND_CL_VERSION=4112 -DPYOPENCL_USE_DEVICE_FISSION=1
-DHAVE_GL=1 -I/usr/lib/pymodules/python2.7/numpy/core/include -I/usr/include/python2.7 -c
src/wrapper/bitlog.cpp -o build/temp.linux-x86_64-2.7/src/wrapper/bitlog.o
g++ -pthread -shared -Wl,-Bsymbolic-functions -Wl,-z,relro
build/temp.linux-x86_64-2.7/src/wrapper/wrap_cl.o
build/temp.linux-x86_64-2.7/src/wrapper/wrap_cl_part_1.o
build/temp.linux-x86_64-2.7/src/wrapper/wrap_cl_part_2.o
build/temp.linux-x86_64-2.7/src/wrapper/wrap_constants.o
build/temp.linux-x86_64-2.7/src/wrapper/wrap_mempool.o
build/temp.linux-x86_64-2.7/src/wrapper/bitlog.o -lboost_python-py27 -lOpenCL -o
build/lib.linux-x86_64-2.7/pyopencl/_cl.so
/usr/bin/ld: cannot find -lOpenCL
collect2: error: ld returned 1 exit status
error: command 'g++' failed with exit status 1
This might be stupid question, but have you installed nvidia-opencl-dev,
or only nvidia-libopencl1 when compiling PyOpenCL with NVIDIA library?
--
Tomasz Rybak GPG/PGP key ID: 2AD5 9860
Fingerprint A481 824E 7DD3 9C0E C40A 488E C654 FB33 2AD5 9860
http://member.acm.org/~tomaszrybak
9:38 p.m.
This might be stupid question, but have you installed
nvidia-opencl-dev,
or only nvidia-libopencl1 when compiling PyOpenCL with NVIDIA library?
The Nvidia driver in debian experimental (I know it means something is
broken...) has no nvidia-opencl-dev in it. It does not provide the
include-headers files for OpenCL.
I included the header from opencl-headers version 1.1
(1.1-2010.07.15-2). I found them in the archives of Ubuntu as Debian
ships only headers for version 1.2 :(
Do I answer your question ?
--
Jérôme Kieffer
Data analysis unit - ESRF
2011
days inactive
2500
days old
9 comments
3 participants
participants (3)
-
Andreas Kloeckner -
Jerome Kieffer -
Tomasz Rybak