Automating toolkit install (avoiding entering install path)

Accelerated Computing CUDA CUDA Programming and Performance
1

I would like to include the CUDA toolkit as part of an automated package installation script I am writing for my lab. However, the toolkit installation requires that the user enter a installation path. It seems that the run script is encoded, so there is no way for me to edit it myself.

Is it possible to use a command line argument or other linux magic to specify the installation path for the toolkit, so that the user does not need to manually enter the path? I would like the installation process to require no user interaction. I have tried, for example, redirecting input, as in:

echo /my/path > inputfile
./toolkit.run < inputfile

… but that doesn’t work.

Any suggestions?

Thanks,
S.

2

./toolkit.run --help
will show you all the flags you can pass to the installer.

3

Thanks, I should have seen that.

For the record, I ended up doing something like “./toolkit.run – --prefix=/my/path” … I found the “prefix” command by extracting the install script and snooping around the code. The extract command was revealed by the “–help” option.