Having @v_unnop (Dew) guide me step-by-step, the installation was quick and easy. (A lot of thanks to him ^^)
Note: this is for computers that do not use NVIDIA graphic cards.
1. Download Packages
- Navigate to http://developer.nvidia.com/object/cuda_2_3_downloads.html
- Download
- CUDA Toolkit for Fedora 10 -> 32 bit
- CUDA SDK code samples and more (optional)
- In case your computer has NVIDIA graphics card, I think you’ll need this too:
- Developer Drivers for Linux (190.42) -> 32bit
2. chmod +x those downloaded files (.run) and run them as root.
- cudatoolkit_2.3_linux_32_fedora10.run
- cudasdk_2.3_linux.run
3. Configure environment variables
- add the following lines into your /etc/profile or .bashrc
- PATH = $PATH:/usr/local/cuda/bin
- export PATH
- LD_LIBRARY_PATH = $LD_LIBRARY_PATH:/usr/local/cuda/lib
- export LD_LIBRARY_PATH
4. Test the SDK
- cd ~/NVIDIA_GPU_Computing_SDK/C
- sudo make emu=1
- If everything is OK, all the example code will be successfully build
- cd bin/linux/emurelease
- run some example program you just compiled
At this step, If you’re lucky, you’re done.
If not, you may have to deal with SELinux
If you cannot run the program because of this error:
./<program_name> error while loading shared libraries: /usr/local/cuda/lib/libcudart.so.2: cannot restore segment prot after reloc: Permission denied
This is an SELinux problem, it blocks libcudart.so.2, which the program needed to use.
You can add the exception to SELinux list with this command:
sudo chcon -t textrel_shlib_t ‘/usr/local/cuda/lib/libcudart.so.2.3′
(libcudart.so.2 is actually a link to libcudart.so.2.3)
Try running the program again, and, We’re done!!!
