Installing the CUDA toolkit results in the following instructions being printed to the console.
Please make sure your LD_LIBRARY_PATH
for 64-bit Linux distributions includes /usr/local/cuda-5.0/lib64:/usr/local/cuda-5.0/lib
for 64-bit Linux distributions add /usr/local/cuda-5.0/lib64 and /usr/local/cuda-5.0/lib
to /etc/ld.so.conf and run ldconfig as root
The following code in
/etc/profile had no effect.
if [ -z "$LD_LIBRARY_PATH" ]; then LD_LIBRARY_PATH=/usr/local/cuda-5.0/lib64:/usr/local/cuda-5.0/lib else LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/cuda-5.0/lib64:/usr/local/cuda-5.0/lib fi export LD_LIBRARY_PATH
That is, rebooting and issuing
echo $LD_LIBRARY_PATH showed the variable was not defined.
To try the alternative suggestion, I added the two lines to the file
/etc/ld.so.conf so my file looks like this
include /etc/ld.so.conf.d/*.conf /usr/local/cuda-5.0/lib64 /usr/local/cuda-5.0/lib
Then I issued:
Still the environment variable was not set. How do I comply with the CUDA installation instructions shown above?
Add a file with the
/etc/ld.so.conf.d/that contains the paths to the libraries and then run ldconfig. Be sure to set the permissions and ownership of the file to match the other files in the directory.
This is a system wide solution as opposed to the user specific solution of modifying .bashrc.
On my system I made
/etc/ld.so.conf.d/. The file contains the lines:
If you create the file as
sudothen your permissions should be good to go, but my
nvidia.confis owner/group root and rw-r--r-- (or 644).