First, my setup: I'm running pyopencl out of git (last update Nov 27) on
an Ubuntu 12.04.1 LTS system with CUDA 5.0 and an NVIDIA GTX 560 Ti
graphics card.
I have no problems running the examples, and I have written some nearly
trivial code that works just fine as well. When I move on to more
sophisticated solutions, I get the dreaded "out of resources" result which
really means I have a segmentation fault or something in my code.
My normal mode of debugging problems like this is to use a whole lot of
printf statements. That approach doesn't work with OpenCL for obvious
reasons. How do you guys debug your OpenCL code if you can't use printf?
The only debuggers I've seen for OpenCL have been for Visual Studio and
other C++ solutions. :-(
I got the bright idea of trying an OpenCL simulator which would ideally let
me trace one individual core and watch what's going on inside. I
downloaded GPGPU-Sim and built it with CUDA 5.0 but it does not currently
work with PyOpenCL. I get this error when I try to import the modules:
Traceback (most recent call last):
File "/home/jmt/git/pyopencl/examples/demo.py", line 1, in <module>
import pyopencl as cl
File
"/usr/local/lib/python2.7/dist-packages/pyopencl-2012.2-py2.7-linux-x86_64.egg/pyopencl/__init__.py",
line 4, in <module>
import pyopencl._cl as _cl
ImportError:
/usr/local/lib/python2.7/dist-packages/pyopencl-2012.2-py2.7-linux-x86_64.egg/pyopencl/_cl.so:
undefined symbol: clEnqueueWriteImage
So using that simulator is out. Are there any simulators that are known to
work with PyOpenCL and CUDA 5.0?
Thanks in advance! :-)
Jack.
--
mathuin at gmail dot com
_______________________________________________
PyOpenCL mailing list
[email protected]
http://lists.tiker.net/listinfo/pyopencl
