49 | | http://hdfpga.blogspot.com/2011/05/install-cuda-40-on-ubuntu-1104.html |
50 | | |
51 | | === [Optional] Using Synergy to control the whole cluster with only one keyboard and mouse === |
52 | | * Install synergy |
53 | | {{{ |
54 | | #!sh |
55 | | aptitude install synergy quicksynergy |
56 | | }}} |
57 | | |
58 | | howto: http://www.mattcutts.com/blog/how-to-configure-synergy-in-six-steps/ |
59 | | |
| 49 | Basically according to http://hdfpga.blogspot.com/2011/05/install-cuda-40-on-ubuntu-1104.html |
| 50 | |
| 51 | In Short: |
| 52 | * Stop X-Server with |
| 53 | {{{ |
| 54 | /etc/init.d/kdm stop |
| 55 | }}} |
| 56 | * Uninstall the auto Nvidia driver provided by Kubuntu. |
| 57 | * Download and install the newest Driver provided by Nvidia on its webpage. |
| 58 | * Start X-Server with |
| 59 | {{{ |
| 60 | /etc/init.d/kdm start |
| 61 | }}} |
| 62 | * Define GCC 4..4 as standart compiler for non-root: |
| 63 | {{{ |
| 64 | cd ~ |
| 65 | mkdir gcc44 |
| 66 | cd gcc4 |
| 67 | ln -s /usr/bin/cpp-4.4 cpp |
| 68 | ln -s /usr/bin/gcc-4.4 gcc |
| 69 | ln -s /usr/bin/g++-4.4 g++ |
| 70 | }}} |
| 71 | * Download and install CUDA: |
| 72 | * Download the following files and install them as root |
| 73 | {{{ |
| 74 | CUDA Toolkit for Ubuntu Linux 10.10 (sudo sh ...) : cudatoolkit_4.0.17_linux_64_ubuntu10.10.run |
| 75 | CUDA Tools SDK (sudo sh ...) : cudatools_4.0.17_linux_64.run |
| 76 | }}} |
| 77 | * Download the following file and install it as non-root |
| 78 | {{{ |
| 79 | GPU Computing SDK code samples (no sudo) : gpucomputingsdk_4.0.17_linux.run |
| 80 | }}} |
| 81 | * Announce CUDA to your build environment: |
| 82 | * Edit /usr/local/cuda/bin/nvcc.profile to look in the gcc44 directory (append the full path): |
| 83 | {{{ |
| 84 | compiler-bindir =/home/your-non-root-user/gcc44 |
| 85 | }}} |
| 86 | * Append in /root/.bashrc and /home/your-non-root-user/.bashrc |
| 87 | {{{ |
| 88 | export CUDA_HOME="/usr/local/cuda" |
| 89 | export LD_LIBRARY_PATH="${LD_LIBRARY_PATH}:${CUDA_HOME}/lib64" |
| 90 | export PATH=${CUDA_HOME}/bin:${PATH} |
| 91 | }}} |
| 92 | * Take changes in effect: |
| 93 | {{{ |
| 94 | sudo source /root/.bashrc |
| 95 | sudo source /home/your-non-root-user/.bashrc |
| 96 | sudo ldconfig |
| 97 | }}} |
| 98 | * Build GPU Computing SDK to check build CUDA client and sample programs |
| 99 | * run make in NVIDIA_GPU_Computing_SDK/C |
| 100 | {{{ |
| 101 | cd ~/NVIDIA_GPU_Computing_SDK/C |
| 102 | make |
| 103 | }}} |
| 104 | * Verify CUDA 4.0 installation by running the deviceQuery sample in the NVIDIA_GPU_Computing_SDK/C |
| 105 | {{{ |
| 106 | bin/linux/release/deviceQuery |
| 107 | }}} |
| 108 | * Download and install Nvidia Texture Tools 2 (NVTT) from svn ( stablerelease 2.0.8.1 won't work with CUDA 4.0) as non-root: |
| 109 | {{{ |
| 110 | cd ~ |
| 111 | svn checkout http://nvidia-texture-tools.googlecode.com/svn/branches/2.0/ nvidia-texture-tools-read-only |
| 112 | cd nvidia-texture-tools-read-only |
| 113 | # Use CMake instead of ./configure to ensure CMake looks for cuda/lib64 instead of cuda/lib. |
| 114 | make |
| 115 | sudo make install |
| 116 | }}} |