Installing CUDA Toolkit 5.0 on CentOS 6.4

1、Verify You Have a CUDA-Capable GPU
   lspci | grep -i nvidia
2、Verify You Have a Supported Version of Linux
   uname -m && cat /etc/*release
3、Verify the System Has gcc Installed
   gcc --version
   --or 
   yum groupinstall "Development Tools"
   --or
   yum install gcc gcc-c++ kernel-devel

4、Install OpenGL driver
   yum install freeglut
   ln -s /usr/lib/x86_64-linux-gnu/libglut.so.3 /usr/lib/libglut.so
   如果不安裝，在summary部分就會出現：
   Samples: Installation Failed. Missing required libraries.
   Missing required library libglut.so
5、Download the NVIDIA CUDA Toolkit
   http://www.nvidia.com/content/cuda/cuda-downloads.html.
6、Install the NVIDIA CUDA Toolkit
   1）Exit the GUI if you are in a GUI environment
   /sbin/init 3
   2) Install the CUDA Toolkit by running the downloaded .run file as a superuser.
   sudo sh cuda_5.0.35_linux_64_rhel6.x-1.run
   3) Define the environment variables.
   The typical way to place these values in your environment is with the following commands:

   export PATH=/usr/local/cuda-5.0/bin:$PATH

   export LD_LIBRARY_PATH=/usr/local/cuda-5.0/lib:$LD_LIBRARY_PATH
   for 32-bit operating systems, or

   export LD_LIBRARY_PATH=/usr/local/cuda-5.0/lib:/usr/local/cuda-5.0/lib64:$LD_LIBRARY_PATH
   for 64-bit operating systems.

   To make such settings permanent
  
   vi ~/.bashrc     //如果需要多個用戶共同使用，修改/etc/profile:此文件爲系統的每個用戶設置環境信息,當用戶第一次登錄時,該文件被執行.

   export PATH=$PATH:/usr/local/cuda/bin:$PATH

   export  LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/cuda/lib64:$LD_LIBRARY_PATH
   4) CUDA SDK Samples
   ~/NVIDIA_CUDA-5.0_Samples
   make

   If everything goes well, you should be able to verify your CUDA installation by running the deviceQuery sample in bin/linux/release.

參考文獻

1.http://docs.nvidia.com/cuda/cuda-getting-started-guide-for-linux/index.html

2. http://www.xue5.com/Developer/C++/492704.html

3.http://blog.sina.com.cn/s/blog_705c12e501017wvx.html