Invalid command Queue when using big data sets on nVidia
by Justin Heinermann
Dear all,
we are trying to implement a K nearest neighbor search on GPUs with
PyOpenCL. The goal of the algorithm is: For a given target point,
find the nearest points from a given set (training data). The distance
between two points is computed by the squared euclidean distance.
One of our implementations is a brute force approach, which aims
at processing big data sets in parallel, e.g. 1 million training data and
some millions of targets (test data). For every target point one kernel
instance is created which finds the k nearest points out of the
training points.
Our problem is the following. Everything works fine for small data sets
and the results are as expected on both GPU (GeForce GTX 650 with
nVidia Driver 313.09.) and CPU(Intel Core i5-3450 with AMD APP SDK)
running Ubuntu 12.10, PyOpenCL 2013.1-py2.7-linux-x86_64.
But if we increase the size of the data sets, the GPU version crashes
with the following error:
> File "brutegpu.py", line 65, in query
> cl.enqueue_copy(self.queue, d_min, self.d_min_buf).wait()
> File "/usr/local/lib/python2.7/dist-packages/
> pyopencl-2013.1-py2.7-linux-x86_64.egg/pyopencl/__init__.py",
> line 935, in enqueue_copy
> return _cl._enqueue_read_buffer(queue, src, dest, **kwargs)
> pyopencl.LogicError: clEnqueueReadBuffer failed: invalid command queue
The CPU-Version still works fine with 1 million training points
and 1 million of test points. Attached you can find the corresponding
source code as working minimal example, which consists of on
Host-Python-File
and one OpenCL-Kernel-File.
We would highly apprecriate any help - maybe we made a
mistake which is already known to you.
So the big question for us is: Why is it working on CPU and why isn't it
working on the GPU?
Are there nVidia-specific pitfalls for such big data sets?
The compiler says:
> ptxas info : Compiling entry function 'find_knn' for 'sm_30'
> ptxas info : Function properties for find_knn
> 0 bytes stack frame, 0 bytes spill stores, 0 bytes spill loads
> ptxas info : Used 17 registers, 336 bytes cmem[0], 4 bytes cmem[3]
Or are there any rules for using a kernel for big data sets such as setting
the work group sizes or maximum memory usage?
The error message "invalid command queue" is confusing and I wasn't able
to find any helpful information (except that oftentimes "invalid command
queue" means segfault, but i could not find any wrong array adress yet.)
Maybe one of you could have a look at our code and finds some stupid
mistake.
We would be very grateful for every hint.
Best regards,
Justin Heinermann,
University Oldenburg
5 years, 6 months
Problems with git repositories
by Tomasz Rybak
Hello.
What's going on with PyOpenCL repositories?
I cannot clone one on tiker.net:
$ git clone http://git.tiker.net/trees/pyopencl.git
Cloning into 'pyopencl'...
error: unable to open object pack
directory: /tmp/pyopencl/.git/objects/pack: Too many open files
fatal: failed to read object bff987ecd9aca7f6b2e1dc8a86af7086ee9cbcc2:
Too many open files
At the same time repository on gitbut contains old revision,
a2fb48462cf957590d2c9c6c8eb3776f781ca6b6 from
Wed Nov 27 14:40:35 2013 -0600, while the one from tiker.net
is a1d1603041ff90f3814dd263c33ee35c7de33ef9 from
Tue Feb 18 12:01:10 2014 -0600.
Short googling suggests the need to re-push or issuing
git gc on repository.
At the same time PyCUDA repositories seem OK.
--
Tomasz Rybak GPG/PGP key ID: 2AD5 9860
Fingerprint A481 824E 7DD3 9C0E C40A 488E C654 FB33 2AD5 9860
http://member.acm.org/~tomaszrybak
5 years, 11 months
Profiling OpenCL ...
by Jerome Kieffer
Dear all,
I wonder which profiler is mostly used by you developers ?
Since the switch to cuda5, the nvidia profiler "compute-prof" has been
replaced with "nvvp" which no more supports (py)opencl programs.
Has anybody tested to generate PTX from OpenCL code and profile the
corresponding PTX using nvvp ?
Cheers,
--
Jérôme Kieffer
tel +33 476 882 445
5 years, 11 months
Re: [PyOpenCL] Segmentation faults when using 'int_ptr' with other Python extension.
by Martin Rehr
Hi Bogdan,
Thanks for the tip, I will take a look at the Reikna package.
Cheers,
Martin
On 17/02/2014, at 11.03, Bogdan Opanchuk <mantihor(a)gmail.com> wrote:
> Hi Martin,
>
> Thank you, but as an author of this package I'd prefer that people did
> not mention it, and used Reikna instead (which is already featured in
> PyOpenCL docs, thanks to Andreas). It incorporates pyfft functionality
> and has some other useful stuff.
>
> On Mon, Feb 17, 2014 at 5:36 PM, Martin Rehr <rehr(a)nbi.ku.dk> wrote:
>>
>> On 16/02/2014, at 23.07, Andreas Kloeckner <lists(a)informa.tiker.net> wrote:
>>
>>> Brandon Amos <bdamos(a)vt.edu> writes:
>>>
>>>> Hi Andreas,
>>>>
>>>>
>>>>> peb does not build for me:
>>>>>
>>>> I think this is from incompatibilities between Python 3.3 and Python 2.7.
>>>>
>>>> Regardless, I'm probably going to remove `peb` from github and
>>>> just follow Gregor's gpyfft example for extending clblas.
>>>>
>>>> I don't know if this would benefit anybody else in the future,
>>>> but a reference to Gregor's gypfft library in the
>>>> "Interoperability with other OpenCL software" section in the documentation
>>>> would have really helped me!
>>>
>>> Done:
>>>
>>> http://documen.tician.de/pyopencl/#software-that-works-with-or-enhances-p...
>>>
>>> Andreas
>>>
>>> _______________________________________________
>>> PyOpenCL mailing list
>>> PyOpenCL(a)tiker.net
>>> http://lists.tiker.net/listinfo/pyopencl
>>
>> Another FFT package worth mentioning is pyfft: https://pypi.python.org/pypi/pyfft it works with both PyCUDA an PyOpenCL
>>
>> Cheers,
>> Martin
>> _______________________________________________
>> PyOpenCL mailing list
>> PyOpenCL(a)tiker.net
>> http://lists.tiker.net/listinfo/pyopencl
6 years
Re: [PyOpenCL] Segmentation faults when using 'int_ptr' with other Python extension.
by Jerome Kieffer
On Sun, 16 Feb 2014 16:42:09 -0500
Brandon Amos <bdamos(a)vt.edu> wrote:
> I don't know if this would benefit anybody else in the future,
> but a reference to Gregor's gypfft library in the
> "Interoperability with other OpenCL software" section in the documentation
> would have really helped me!
I was aware of this library and will test it as soon as I have some time.
How does it compare with cuFFT/scikit.cuda ? did anybody benchmark ?
Cheers,
--
Jérôme Kieffer
Data analysis unit - ESRF
6 years
Is there 2014 release plan?
by Tomasz Rybak
Hello.
I have question regarding potential dates of releasing
new versions of PyOpenCL (and PyCUDA). Is there any plan?
I'm asking as Debian Jessie will be frozen on 2014-11-05
which means that no new versions of packages will be accepted
after that date. To be sure that there is no ugly bugs
discovered at the last moment, I intend to upload latest
version in October 2014. Andreas - is that OK with you?
I can move upload earlier or later, but the latest day
I can upload and be sure that this particular version
reaches Jessie is 2014-10-22. Such late upload will be
risky though.
Best regards.
--
Tomasz Rybak GPG/PGP key ID: 2AD5 9860
Fingerprint A481 824E 7DD3 9C0E C40A 488E C654 FB33 2AD5 9860
http://member.acm.org/~tomaszrybak
6 years
Segmentation faults when using 'int_ptr' with other Python extension.
by Brandon Amos
Hi all, I'm really interested in the pyopencl project and on integrating
with clBLAS [0], as discussed in a previous thread [1],
and I'm currently working on building a Python C extension to
use pyopencl objects with clBLAS.
However, I'm very new to building Python C extensions,
and I'm new to writing Python C extensions, and
the interoperability documentation [2] mentions pyopencl objects
contain `from_int_ptr()` and `int_ptr`, but unfortunately
no examples are given.
So, as a first step to help me learn about build a Python C extension
to interoperate with pyopencl, and to help anybody else,
I've created a project on GitHub [3] to show a simple usage.
It's still very much a work in progress and needs a lot of cleaning.
I'm sending this message to ask for help with a segfault I'm getting when
calling `clGetContextInfo`.
`example-usage.py` [4] shows the example I'm using to obtain a pyopencl
context, queue, and buffer.
The `int_ptr` fields are passed to `peb.c::peb_example_method()` [5],
the context segfaults when passed to `clGetContextInfo`, which
I added as a check because `clCreateProgramWithSource` was segfaulting.
I've added `printf("Calling. %p %d\n", &m_context, m_context);`
to `pyopencl::wrap_cl.hpp::context::data` to print the address and
first 32 bits of the context that's being returned to python.
Then, in `peb.c` I added `printf("%p %d\n", ctx, (int) ctx);`.
This may be misleading, but the following output shows
the memory addresses differing (red flag), but the first 32 bits
being the same.
[peb::example-usage.py] 0x1d722b8
[pyopencl::wrap_cl.hpp::context::data] Calling. 0x1b55130 30876344
1
2
3
[peb::peb.c::peb_example_method()] 0x1d722b8 30876344
For more information about my configuration, I'm using x86_64 Linux,
the latest `pyopencl` from `git`, and the Intel OpenCL 1.2 SDK.
I've seen a few posts on the `pyopencl` mailing list, so
I hope this isn't an issue with the Intel SDK,
because I can't easily develop on any other OpenCL device.
Let me know if I can provide any more information.
Thanks,
Brandon.
[0]: https://github.com/clMathLibraries/clBLAS
[1]: http://lists.tiker.net/pipermail/pyopencl/2013-August/001573.html
[2]: http://documen.tician.de/pyopencl/misc.html#interoperability-with-other-o...
[3]: https://github.com/bamos/peb
[4]: https://github.com/bamos/peb/blob/master/example-usage.py
[5]: https://github.com/bamos/peb/blob/master/src/peb.c
6 years
Segmentation fault in pyopencl.image_from_array
by Jerome Kieffer
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
6 years
