Debian Cuda packages

[Andreas Beckmann]

Hi Brent,

On 2011-10-09 21:24, Brent Busby wrote:
> Sorry for the rather basic question...
> 
> With the new Cuda packages in wheezy/sid, is it possible or recommended
> to install the SDK by simply running the upstream ".run" installer from
> nVidia over the top of the install provided by the packages?  I didn't
> want to ask, but there's no README.Debian or any other best practices
> docs, and I know that Debian's way of installing the actual nVidia
> driver often doesn't get along with nVidia's ".run" installer very well.
> 
> By the way, thanks for making these packages.  We have some clusters
> here that are doing protein-folding research which are running Debian,
> and I didn't want to have to convert any nodes to Ubuntu or Fedora just
> to get Cuda running!

I'll add the following /usr/share/doc/nvidia-cuda-toolkit/README.Debian
to the next upload. Does this clarify your question regarding the SDK?

nvidia-cuda-toolkit for Debian
------------------------------

Installing the GPU Computing SDK
================================

To install the GPU Computing SDK (a collection of code examples for CUDA
and OpenCL), just download and run gpucomputingsdk_<VERSION>_linux.run
from NVIDIA - it will extract into ~/NVIDIA_GPU_Computing_SDK/ by default.

The following Debian packages are needed to build the examples:
  * freeglut3-dev
  * libgl1-mesa-dev
  * libglu1-mesa-dev
  * libx11-dev
  * libxi-dev
  * libxmu-dev

To build the example programs, change to the directory where the SDK was
installed (by default: ~/NVIDIA_GPU_Computing_SDK/) and run

    make CUDA_INSTALL_PATH=/usr

For more information on the SDK, please see
http://developer.nvidia.com/gpu-computing-sdk

 -- Andreas Beckmann <debian at abeckmann.de>  Mon, 10 Oct 2011 18:52:42 +0200