Hello.
I attach patch updating pycuda.tools.DeviceData and
pycuda.tools.OccupancyRecord
to take new devices into consideration. I have tried to maintain "style" of
those classes
and introduced changes only when necessary. I have done changes using my old
notes
and NVIDIA Occupancy Calculator. Unfortunately I currently do not have
access to Fermi
to test those fully.
Best regards.
Tomasz Rybak
Hi all,
I'm observing the following behavior with latest (git-fetched today)
pycuda and opencl versions on Snow Leopard 10.6.4:
$ python
>>> import pycuda.driver
>>> import pyopencl
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "/Library/Python/2.6/site-packages/pyopencl-0.92beta-py2.6-macosx-10.6-i386.egg/pyopencl/__init__.py",
line 3, in <module>
import pyopencl._cl as _cl
AttributeError: 'NoneType' object has no attribute '__dict__'
$ python
>>> import pyopencl
>>> import pycuda.driver
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "/Library/Python/2.6/site-packages/pycuda-0.94rc-py2.6-macosx-10.6-i386.egg/pycuda/driver.py",
line 1, in <module>
from _driver import *
AttributeError: 'NoneType' object has no attribute '__dict__'
This worked with stable versions. Does anyone know why is this happening?
(One may ask why do I need both libraries in the same program. I have
the set of tests for my module, which can use both Cuda and OpenCL,
and it is convenient to run all the tests using the single file.
Although it is not a critical issue, I'm just curious).
Best regards,
Bogdan
Hi,
I want to run element-wise computations on different parts of an
array. Loading each part of the array to device mem when needed turned
out to use up a lot of time and not really speed things up compared to
cpu. Instead, I want to once load the data array into device mem and
provide pointers to which elements to look at (I do have the numpy
view/slice of the array). I looked into different ways of doing this
but can't seem to find the right approach, any help would be
appreciated.
ElementwiseKernel seems to support range and slicing now, however, my
code is (cuda) c and I import it as a SourceModule which probably
means I can't use the ElementwiseKernel approach.
-Thomas
On Sun, 29 May 2011 16:08:28 +1000, <Brett.Bryan(a)csiro.au> wrote:
> Just wondering if there are plans/timelines for implementing the Numpy
> multidimensional and fancy indexing/slicing in PyCUDA? We use this
> feature a lot to do repetitive data processing. There is not much of a
> performance improvement from indexing on CPU then passing to GPU,
> processing, then passing back. It seems to me that passing a single
> multidimensional array to GPU memory then indexing/slicing and
> processing entirely on the GPU would provide substantially greater
> performance improvement?
This is being worked on--see the recent announcement regarding the
'gpundarray' list. No precise time line, but potentially soon--faster if
you help. :)
Andreas
Hi All
Just wondering if there are plans/timelines for implementing the Numpy multidimensional and fancy indexing/slicing in PyCUDA? We use this feature a lot to do repetitive data processing. There is not much of a performance improvement from indexing on CPU then passing to GPU, processing, then passing back. It seems to me that passing a single multidimensional array to GPU memory then indexing/slicing and processing entirely on the GPU would provide substantially greater performance improvement?
Thanks
Regards
Brett
Just trying a new install on fedora 15.
pycuda-0.94.2
cudatoolkit_4.0.17_linux_64_fedora13.run
I think cuda is not usable on this setup. If I run make test in pycuda, I see:
E CompileError: nvcc compilation of /tmp/tmp8bom2S/kernel.cu failed
E [command: nvcc --cubin -arch sm_11 -I/usr/lib64/python2.7/site-
packages/pycuda-0.94.2-py2.7-linux-x86_64.egg/pycuda/../include/pycuda
kernel.cu]
E [stderr:
E In file included from /opt/cuda/bin/../include/cuda_runtime.h:59:0,
E from <command-line>:0:
E /opt/cuda/bin/../include/host_config.h:82:2: error: #error --
unsupported GNU version! gcc 4.5 and up are not supported!Traceback (most recent
call last):
But I don't think an older gcc4 is available - the only old compat packages are
for gcc34.
On Sun, 17 Apr 2011 17:10:30 +0200, "=?UTF-8?B?VG9tYXN6IFJ5YmFr?=" <bogomips(a)post.pl> wrote:
> Dnia 2011-04-16 00:48 Andreas Kloeckner napisał(a):
>
> >Hi Tomasz,
> >
> >any progress on the things below? Has maybe another email gone missing?
> >I'd like to release PyCUDA 2011.1 soon.
>
>
> I have send the patch to the mailing list on 2011-03-18 and 2011-03-24.
> Please check the archives.
Weird. I'm really not sure what happened there--I could've sworn I never
saw these emails.
In any case, I've merged your patch into the
curand-wrapper-v2-from-tomasz branch, which I've also brought up to
current master.
Looks like we're almost done here--only initialization is missing.
Thanks for your work,
Andreas
On Sun, 22 May 2011 11:33:47 -0400, Nicolas Pinto <nicolas.pinto(a)gmail.com> wrote:
> >
> > Also, is there a public issue tracker for PyCUDA and PyOpenCL? I
> > personally find keeping track of e-mail threads to be a nightmare. I
> > hate to be a pain but like everyone else I'm +1 on putting these
> > projects on GitHub.
> >
>
> +1 on (re-)opening the issue tracker on github:
> https://github.com/inducer/pyopencl
Well, fine. I've reopened the issue trackers for PyOpenCL and PyCUDA.
Andreas
hi all, what is the exact format of the --ldflags option of configure.py? On
my setup the logical link libcuda.so is in /usr/lib/nvidia-current,however
whatever i try i get "cannot find lcuda" as response to make install, what
am i doing wrong?
Thanks in advance
joop