Editing Software Development Platforms (section)

====Installing the CUDA SDK====

Again from the nVidia website, download the CUDA SDK package.  Install it.

 chmod +x gpucomputingsdk_x.x.xx_linux.run
 ./gpucomputingsdk_x.x.xx_linux.run

Notice that we don't need <tt>sudo</tt> this time, because root privileges are not required: the SDK will live in your own home directory somewhere.

To compile the examples, just <tt>cd</tt> to the SDK directory and run <tt>make</tt>.  My favorite example program is <tt>fluidsGL</tt>.

One complication however is that some SDK programs can only be built with gcc 4.4 and earlier, and the NVIDIA SDK doesn't respect the environment variables CC and CXX.  So, we're going to set up the Debian alternatives system to be able to switch at will, like this:

 sudo update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-4.6 60 --slave /usr/bin/g++ g++ /usr/bin/g++-4.6
 sudo update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-4.4 50 --slave /usr/bin/g++ g++ /usr/bin/g++-4.4

To switch, run this command and make your choice:

 sudo update-alternatives --config gcc

In addition, to build all the programs you will need these libraries:

 sudo apt-get install libxmu-dev libxi-dev

The linker fails to find libcuda as installed by Ubunutu, which can be fixed by running it like this:

 make LINKFLAGS=-L/usr/lib/nvidia-current

For trouble-shooting, run make with

 make verbose=1