Thanks for your reply. Turns out I had an old NVIDIA driver, when I
replaced it and started using Cuda 5.5 everything was fine.
On Wed, Aug 7, 2013 at 2:28 PM, Andreas Kloeckner
Trevor Cickovski <movingpictures83(a)gmail.com> writes:
> I am running pycuda using Python 2.6.5, Cuda 5.0 on Ubuntu 10. My
> card is an NVIDIA GTX680.
> Whenever I do 'import pycuda.autoinit', everything hangs (requires manual
> kill). I have traced it to this statement in autoinit.py:
> context = make_default_context()
> Which in turn freezes on this line in tools.py:
> return ctx_maker(dev)
> This in turns hangs on the return statement here:
> if ctx_maker is None:
> def ctx_maker(dev):
> return dev.make_context()
> dev is type Device, which appears to be in _driver.so.
> Has anyone had this problem before? I tried Cuda 4.0 as well with the
Can you run 'regular' CUDA C programs? (I suspect not, but I might be
wrong.) In any case, it seems to me that something's up with your CUDA
installation, more so than with PyCUDA.