Re: [PyCuda] test_driver failure (0.91.1)
by Andreas Klöckner
On Dienstag 30 Dezember 2008, Ian Cullinan wrote:
> $ python test_driver.py
> .....E....
> ======================================================================
> ERROR: test_mempool (__main__.TestCuda)
> ----------------------------------------------------------------------
> Traceback (most recent call last):
> File "test_driver.py", line 239, in test_mempool
> queue.append(pool.allocate(1<<e))
> MemoryError
You can ignore this. This just means that the memory pool tests allocate more
memory than is on your card. I'll fix this eventually, promise. :)
Andreas
10 years, 10 months
test_driver failure (0.91.1)
by Ian Cullinan
$ python test_driver.py
.....E....
======================================================================
ERROR: test_mempool (__main__.TestCuda)
----------------------------------------------------------------------
Traceback (most recent call last):
File "test_driver.py", line 239, in test_mempool
queue.append(pool.allocate(1<<e))
MemoryError
----------------------------------------------------------------------
Ran 10 tests in 0.181s
FAILED (errors=1)
$ python dump_properties.py
1 device(s) found.
Device #0: GeForce 8800 GTS
Compute Capability: 1.0
Total Memory: 654656 KB
CLOCK_RATE: 1188000
GPU_OVERLAP: 0
MAX_BLOCK_DIM_X: 512
MAX_BLOCK_DIM_Y: 512
MAX_BLOCK_DIM_Z: 64
MAX_GRID_DIM_X: 65535
MAX_GRID_DIM_Y: 65535
MAX_GRID_DIM_Z: 1
MAX_PITCH: 262144
MAX_REGISTERS_PER_BLOCK: 8192
MAX_SHARED_MEMORY_PER_BLOCK: 16384
MAX_THREADS_PER_BLOCK: 512
MULTIPROCESSOR_COUNT: 12
TEXTURE_ALIGNMENT: 256
TOTAL_CONSTANT_MEMORY: 65536
WARP_SIZE: 32
This is on Fedora 8 (i386) with driver 177.73, Toolkit 2.0, Python 2.5, gcc 4.1.2.
Is this error important or can it be safely ignored?
Cheers,
Ian Cullinan
10 years, 10 months
test_driver need gcc ?
by Jean-Christophe Penalva
Hello,
at the end of the install :
$ cd PyCuda-VERSION/test
$ python test_driver.py
need a gcc on the node with the GPU ?
Here, we have a cluster with login node (on wich i've compil+link) and
executions nodes with GPU (but whithout compiler).
Thank you.
######################################################################
Jean-Christophe Penalva
Centre Informatique National de l'Enseignement Superieur (CINES)
Montpellier, FRANCE
Tel : 33 4 67 141 414 Fax : 33 4 67 523 763
http://www.cines.fr/
10 years, 10 months
install witch python 2.4 anyone ?
by Jean-Christophe Penalva
Hello,
i try to install Pycuda on a host with python 2.4.
All is good (boost, numpy), until i fall on a problem during the make of
pycuda. There's a test on the version of python (for version >= 2.5 ...), and
that break my compil :
src/cpp/cuda.hpp: In member function ‘void
cuda::memcpy_2d::set_src_host(boost::python::api::object)’:
src/cpp/cuda.hpp:942: error: ‘Py_ssize_t’ was not declared in this scope
src/cpp/cuda.hpp:942: error: expected `;' before ‘len’
src/cpp/cuda.hpp:942: error: ‘len’ was not declared in this scope
Any idea ?
######################################################################
Jean-Christophe Penalva
Centre Informatique National de l'Enseignement Superieur (CINES)
Montpellier, FRANCE
Tel : 33 4 67 141 414 Fax : 33 4 67 523 763
http://www.cines.fr/
10 years, 11 months
no array module
by Randy Heiland
Maybe I've just caught the repo in the midst of changes, but in trying
to run the examples/plot_random_data.py, I get:
$ python plot_random_data.py
Traceback (most recent call last):
File "plot_random_data.py", line 9, in <module>
a = cuda.array(size).randn()
AttributeError: 'module' object has no attribute 'array'
Python 2.6.1 (r261:67515, Dec 17 2008, 15:56:40)
[GCC 4.2.1 (Apple Inc. build 5566)] on darwin
Type "help", "copyright", "credits" or "license" for more information.
>>> import pycuda.gpuarray as cuda
>>> dir(cuda)
['DiskDict', 'GPUArray', '__builtins__', '__doc__', '__file__',
'__name__', '__package__', '_kernel', '_splay_backend', 'arange',
'division', 'drv', 'empty', 'memoize', 'numpy', 'random', 'splay',
'to_gpu', 'zeros']
>>>
Unrelated, and off-topic, does anyone happen to use git inside
eclipse? I'm trying to figure it out and would welcome advice.
Please contact me off-list.
thanks, Randy
10 years, 11 months
Re: [PyCuda] test_driver.py problem
by Andreas Klöckner
Hi Jyh-Shyong,
first off, for your own benefit and that of "posterity", please keep the list
cc'd on your replies.
On Sonntag 14 Dezember 2008, you wrote:
> Thanks for your reply, I removed the boost and boost-devel packages
> which come with
> OpenSuSE11.0 and installed boost_1.37 from source. I also downloaded and
> installed
> PyCuda 0.91.1. Here is the output of command
>
> ldd
> /usr/local/lib64/python2.5/site-packages/pycuda-0.91-py2.5-linux-x86_64.egg
>/pycuda/_driver.so
>
> libboost_python-gcc43-mt-1_37.so.1.37.0 =>
> /usr/local/lib/libboost_python-gcc43-mt-1_37.so.1.37.0 (0x00007fbbbf9a3000)
[snip]
> File
> "/usr/local/lib64/python2.5/site-packages/pycuda-0.91-py2.5-linux-x86_64.eg
>g/pycuda/driver.py", line 1, in <module>
> from _driver import *
> ImportError:
> /usr/local/lib64/python2.5/site-packages/pycuda-0.91-py2.5-linux-x86_64.egg
>/pycuda/_driver.so: undefined symbol:
> _ZN5boost6python9converter8registry9push_backEPFPvP7_objectEPFvS5_PNS1_30rv
>alue_from_python_stage1_dataEENS0_9type_infoE
>
> What might be the problem? I also replaced Cuda 2.1 beta with Cuda 2.0
> since we have S1060 and S1070
> GPU systems installed.
Weird. Quick request before we move into debugging mode: Please do 'rm -Rf
build' in the PyCuda directory, then reinstall and try again. Distutils is not
exactly good at figuring out when it needs to recompile stuff.
Andreas
10 years, 11 months
test_driver.py problem
by Jyh-Shyong Ho
Hi,
I just installed pycuda-0.90.2 on my computer and I got the following
error message when I ran the program test_drive.py:
python test_drive.py
Traceback (most recent call last):
File "test_driver.py", line 2, in <module>
import pycuda.driver as drv
File "/usr/local/lib64/python2.5/site-packages/pycuda-0.90.2-py2.5-linux-x86_64.egg/pycuda/driver.py", line 1, in <module>
from _driver import *
ImportError: /usr/local/lib64/python2.5/site-packages/pycuda-0.90.2-py2.5-linux-x86_64.egg/pycuda/_driver.so: undefined symbol: _ZN5boost6python9converter8registry9push_backEPFPvP7_objectEPFvS5_PNS1_30rvalue_from_python_stage1_dataEENS0_9type_infoE
The OS is OpenSuSE 11.0, and Cuda 2.1 Beta is installed.
Do I miss something?
Thanks.
Jyh-Shyong Ho, Ph.D.
Research Scientist
National Center for High Performance Computing
Hsinchu, Taiwan, ROC
10 years, 11 months
newbie building on OSX 10.5
by Randy Heiland
Greetings,
I'm trying to build pycuda for my first time and have hit a snag
already. Thought I'd ask this list while I go off and search for
solutions on my own also. The first obvious problem is that my
'ctags' doesn't have a "-R" option. I'm on OSX 10.5.5 and using gcc
4.2 (although the system python 2.5.1 was built using 4.0.1, fwiw).
I attempted/get the following:
$ ./configure --boost-inc-dir=/usr/local/include/boost-1_37 --cuda-
root=/usr/local/cuda
$ make
ctags -R src || true
ctags: illegal option -- R
usage: ctags [-BFadtuwvx] [-f tagsfile] file .../Library/Frameworks/
Python.framework/Versions/2.5/Resources/Python.app/Contents/MacOS/
Python setup.py build
running build
running build_py
creating build
creating build/lib.macosx-10.3-fat-2.5
creating build/lib.macosx-10.3-fat-2.5/pycuda
copying src/python/__init__.py -> build/lib.macosx-10.3-fat-2.5/pycuda
copying src/python/_kernel.py -> build/lib.macosx-10.3-fat-2.5/pycuda
copying src/python/cumath.py -> build/lib.macosx-10.3-fat-2.5/pycuda
copying src/python/driver.py -> build/lib.macosx-10.3-fat-2.5/pycuda
copying src/python/gpuarray.py -> build/lib.macosx-10.3-fat-2.5/pycuda
copying src/python/rt.py -> build/lib.macosx-10.3-fat-2.5/pycuda
copying src/python/tools.py -> build/lib.macosx-10.3-fat-2.5/
pycudarunning build_extbuilding '_driver' extensioncreating build/
temp.macosx-10.3-fat-2.5
creating build/temp.macosx-10.3-fat-2.5/src
creating build/temp.macosx-10.3-fat-2.5/src/cpp
creating build/temp.macosx-10.3-fat-2.5/src/wrapper
gcc -isysroot /Developer/SDKs/MacOSX10.4u.sdk -fno-strict-aliasing -
Wno-long-double -no-cpp-precomp -mno-fused-madd -fno-common -dynamic -
O3 -DNDEBUG -Isrc/cpp -I/usr/local/include/boost-1_37 -I/usr/local/
cuda/include -I/Library/Frameworks/Python.framework/Versions/2.5/lib/
python2.5/site-packages/numpy/core/include -I/Library/Frameworks/
Python.framework/Versions/2.5/include/python2.5 -c src/cpp/cuda.cpp -o
build/temp.macosx-10.3-fat-2.5/src/cpp/cuda.o -arch i386
cc1plus: error: unrecognized command line option "-Wno-long-double"
error: command 'gcc' failed with exit status 1
thanks,
Randy
10 years, 11 months
Source Modules in emulation mode?
by Brad Zima
Hello,
I noticed in the pycuda documentation that it is possible to pass
command-line parameters to nvcc for compilation of source modules. My
question is whether it is possible to use the -deviceemu flag to have a
source module run in emulation mode? If so, what would be the syntax for
the source module, maybe using this as an example:
mod = drv.SourceModule("""
__global__ void multiply_them(float *dest, float *a, float *b)
{
const int i = threadIdx.x;
dest[i] = a[i] * b[i];
}
""")
Brad Zima
10 years, 11 months
Re: [PyCuda] pycuda on os x, "version mismatch"?
by Bill Punch
Thanks very much. boost had indeed linked against the default Leopard
python version (2.5.1) instead of the later update I have been using
(2.5.2). Truth is, I had forgotten there was an "original" version and
that's what threw me off (/System/Library/Frameworks vs
/Library/Frameworks for the newer stuff from python.org). It shouldn't
have mattered but it did. Works now. Thanks again.
>>>bill<<<
Nicholas Piël wrote:
>
> On Dec 4, 2008, at 8:18 PM, Bill Punch wrote:
>
>> I've given this a pretty good try but no luck. Any help appreciated. I
>> have a
>>
>> Macbook Pro (Model Identifier MacBookPro3,1). 2.4 GHz, 4Gb, with a
>> GeForce 8600M GT and 256 Mb. Running Leopard 10.5.5
>>
>>>>
>>> Python 2.5.2 (r252:60911, Feb 22 2008, 07:57:53)
>>> [GCC 4.0.1 (Apple Computer, Inc. build 5363)] on darwin
>>> Type "help", "copyright", "credits" or "license" for more information.
>>>>>> import pycuda.driver
>>> Fatal Python error: Interpreter not initialized (version mismatch?)
>>> Abort trap
>>
>> The wisdom I've seen is that this is a mismatch of version (of python,
>> of gcc, something) but I'm at a loss as to what given what is shown
>> above. Any help appreciated.
>
> The problem is probably that the boost library is linked against a
> different Python version. You can use 'otool -L library'
>
> fe:
> $ otool -L /usr/local/lib/libboost_python-xgcc40-mt.dylib
>
> Now check if this is the same Python version you are currently running
> or Pycuda is linked against. It could very well be that boost compiled
> against your DarwinPorts/Fink Python and pycuda is using your
> Framework version.
>
> Good luck!
> Nicholas
10 years, 11 months
