This document is a guideline about how to install CUDA SDK in student's home directory for CS4803DGC spring 2011 in Linux machines in the States Lab.

You have to install NVIDIA CUDA SDK in your home directory. You can download it from NVIDIA.

You must download CUDA 2.3 (Choose Linux 32-bit, Redhat 5)
To install cudasdk_2.3_linux.run,
you need to type
sh cudasdk_2.3_linux.run

CUDA toolkit is located in /usr/local/cuda. To use CUDA, you have to set some environment variables such as PATH, LD_LIBRARY_PATH and CUDA_INSTALL_PATH like below:
In csh,
setenv PATH /usr/local/cuda/bin:$PATH
setenv LD_LIBRARY_PATH /usr/local/cuda/lib
setenv CUDA_INSTALL_PATH /usr/local/cuda

In bash,
export PATH="/usr/local/cuda/bin:$PATH"
export LD_LIBRARY_PATH="/usr/local/cuda/lib"
export CUDA_INSTALL_PATH="/usr/local/cuda"

After you install CUDA_SDK. You must create a symbolic link for glutlibrary.


ln -s /usr/lib/libglut.so.3.8.0 ~/NVIDIA_GPU_Computing_SDK/C/lib/libglut.so

Go to the NVIDIA_GPU_Computing_SDK/C/ directory, and type "make" then it starts to build. After you compile, the bin directory is created automatically. The bin directory is at ~/NVIDIA_GPU_Computing_SDK/C/bin/linux/release. You can execute the binaries in the release directory. There might be some compilation erros due to permission issues but it won't prevent from doing homeworks.