obenauer
March 31st, 2009, 03:56 PM
I recently installed Nvidia's CUDA programming environment and software development kit for doing general purpose computing on their graphics cards. I installed it on a base Ubuntu 8.04 (Hardy Heron) system, 32-bit version. They don't have a download package for later versions of Ubuntu (as of March 2009). In case it benefits other people, here are the steps I followed:
1. First, I went to this URL:
http://www.nvidia.com/object/cuda_get.html
I selected my operating system (Linux 32-bit), Linux version (Ubuntu 8.04), and downloaded the CUDA 2.1 driver, toolkit, and software development kit (SDK). These are their filenames:
NVIDIA-Linux-x86-180.06-pkg1.run
cuda-linux-rel-nightly-2.1.1635-3065709.run
cuda-sdk-linux-2.10.1126.1520-3141441.run
(The numbers in the filenames may be different as they make updates.)
2. In order to change the Nvidia driver to one that supports the most recent CUDA features, I needed to exit X windows:
sudo /etc/init.d/gdm stop
(Or if you're using KDM, substitute "kdm" instead of "gdm" in the command.)
3. I hit Ctrl-Alt-F1 to start a terminal session, and logged in.
4. I removed the Nvidia driver and related files that came with the default Ubuntu installation:
sudo apt-get remove nvidia-glx-new nvidia-kernel-common ubuntu-restricted-extras linux-restricted-modules-2.6.24-19-generic linux-restricted-modules-2.6.24-22-generic linux-restricted-modules-2.6.24-23-generic linux-restricted-modules-generic
5. Install development tools:
sudo apt-get install build-essential libglut3-dev
6. I navigated to the Nvidia CUDA driver setup file that I had downloaded above (cd projects/cuda/installation in my case), and ran the command:
sudo sh NVIDIA-Linux-x86-180.06-pkg1.run
At this stage, the file /usr/lib/libcuda.so should exist.
7. I disabled the default "nv" driver:
sudo vi /etc/default/linux-restricted-modules-common
Edit DISABLED_MODULES="" to say DISABLED_MODULES="nv"
8. I rebooted the computer:
sudo reboot
Upon restarting, X windows started successfully and was using the new Nvidia driver. (Usually this is visible because the screen resolution is higher than with the open source "nv" driver.)
9. I installed the CUDA toolkit:
sudo sh cuda-linux-rel-nightly-2.1.1635-3065709.run
This installs CUDA in the default path, /usr/local/cuda. To uninstall it, delete this directory (rm -rf /usr/local/cuda).
10. I added CUDA directories to the default path and library path system variables, by adding these lines to ~/.bashrc:
# CUDA
export PATH=$PATH:/usr/local/cuda/bin
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/cuda/lib
The terminal needs to be restarted for the change to take effect.
11. I installed the CUDA SDK:
sh cuda-sdk-linux-2.10.1126.1520-3141441.run
This installs the SDK to ~/NVIDIA_CUDA_SDK.
To try out the installation, I compiled the program "fluidsGL" that comes with the SDK. First I needed some libraries it depends on:
sudo apt-get install libxi-dev libxmu-dev
12. All the example programs can then be compiled by typing "make" in the ~/NVIDIA_CUDA_SDK directory. To run the programs, navigate to ~/NVIDIA_CUDA_SDK/bin/linux/release, and type their filenames with the location "./" as a prefix. I ran fluidsGL ("./fluidsGL"), and it opened a window showing a green fluid. Pointing and dragging the mouse makes the fluid move. The other programs in the SDK also compiled and ran successfully.
1. First, I went to this URL:
http://www.nvidia.com/object/cuda_get.html
I selected my operating system (Linux 32-bit), Linux version (Ubuntu 8.04), and downloaded the CUDA 2.1 driver, toolkit, and software development kit (SDK). These are their filenames:
NVIDIA-Linux-x86-180.06-pkg1.run
cuda-linux-rel-nightly-2.1.1635-3065709.run
cuda-sdk-linux-2.10.1126.1520-3141441.run
(The numbers in the filenames may be different as they make updates.)
2. In order to change the Nvidia driver to one that supports the most recent CUDA features, I needed to exit X windows:
sudo /etc/init.d/gdm stop
(Or if you're using KDM, substitute "kdm" instead of "gdm" in the command.)
3. I hit Ctrl-Alt-F1 to start a terminal session, and logged in.
4. I removed the Nvidia driver and related files that came with the default Ubuntu installation:
sudo apt-get remove nvidia-glx-new nvidia-kernel-common ubuntu-restricted-extras linux-restricted-modules-2.6.24-19-generic linux-restricted-modules-2.6.24-22-generic linux-restricted-modules-2.6.24-23-generic linux-restricted-modules-generic
5. Install development tools:
sudo apt-get install build-essential libglut3-dev
6. I navigated to the Nvidia CUDA driver setup file that I had downloaded above (cd projects/cuda/installation in my case), and ran the command:
sudo sh NVIDIA-Linux-x86-180.06-pkg1.run
At this stage, the file /usr/lib/libcuda.so should exist.
7. I disabled the default "nv" driver:
sudo vi /etc/default/linux-restricted-modules-common
Edit DISABLED_MODULES="" to say DISABLED_MODULES="nv"
8. I rebooted the computer:
sudo reboot
Upon restarting, X windows started successfully and was using the new Nvidia driver. (Usually this is visible because the screen resolution is higher than with the open source "nv" driver.)
9. I installed the CUDA toolkit:
sudo sh cuda-linux-rel-nightly-2.1.1635-3065709.run
This installs CUDA in the default path, /usr/local/cuda. To uninstall it, delete this directory (rm -rf /usr/local/cuda).
10. I added CUDA directories to the default path and library path system variables, by adding these lines to ~/.bashrc:
# CUDA
export PATH=$PATH:/usr/local/cuda/bin
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/cuda/lib
The terminal needs to be restarted for the change to take effect.
11. I installed the CUDA SDK:
sh cuda-sdk-linux-2.10.1126.1520-3141441.run
This installs the SDK to ~/NVIDIA_CUDA_SDK.
To try out the installation, I compiled the program "fluidsGL" that comes with the SDK. First I needed some libraries it depends on:
sudo apt-get install libxi-dev libxmu-dev
12. All the example programs can then be compiled by typing "make" in the ~/NVIDIA_CUDA_SDK directory. To run the programs, navigate to ~/NVIDIA_CUDA_SDK/bin/linux/release, and type their filenames with the location "./" as a prefix. I ran fluidsGL ("./fluidsGL"), and it opened a window showing a green fluid. Pointing and dragging the mouse makes the fluid move. The other programs in the SDK also compiled and ran successfully.