Incomplete installation - CUDA 6.5 Cross-compile


I would like to cross-compile on my host Ubuntu 14.04 LTS i386 running on Virtual Machine (VMware Player) to my Jetson board R21.
I’m trying to install CUDA 6.5 using this file: from the cuda download website.

Here the steps I’m doing:

  • Log off
  • CTRL + ALT + F1
  • Log in
  • sudo service lightdm stop
  • sudo ./
  • I accept the EULA
  • YES to "You are attempting to install on an unsupported configuration. Do you wish to continue ?"
  • NO to "Install NVIDIA Accelerated Graphics Driver for Linux-x86 340.29?"
  • YES to "Install the CUDA 6.5 Toolkit?"
  • /usr/local/cuda-6.5 to "Enter Toolkit Location:"
  • YES to "Do you want to install a symbolic link at /usr/local/cuda?"
  • And here the results of the installation:

    = Summary =
    Driver:   Not Selected
    Toolkit:  Installed in /usr/local/cuda-6.5
    Samples:  Not Supported
    Please make sure that
     -   PATH includes /usr/local/cuda-6.5/bin
     -   LD_LIBRARY_PATH includes /usr/local/cuda-6.5/lib, or, add /usr/local/cuda-6.5/lib to /etc/ and run ldconfig as root
    To uninstall the CUDA Toolkit, run the uninstall script in /usr/local/cuda-6.5/bin
    To uninstall the NVIDIA Driver, run nvidia-uninstall
    ***WARNING: Incomplete installation! This installation did not install the CUDA Driver. A driver of version at least 340.00 is required for CUDA 6.5 functionality to work.
    To install the driver using this installer, run the following command, replacing <CudaInstaller> with the name of this run file:
        sudo <CudaInstaller>.run -silent -driver
    Logfile is /tmp/cuda_install_5754.log

    When I am trying “sudo ./ -silent -driver”, nothing happens on the screen (it’s doing something as I can not type things on the screen for several seconds) and when I try again to run the, I’m having the same warning. It seems like the installation of the driver did not work correctly (but nothing can tell me if it works or not).

    What am I doing something wrong ? What am I missing ?
    Is it possible to cross-compile using a 32bit host system ?
    Is it the good .run file that am I using ?

    Thank you to everyone who will read me !


    Well all the warning tells you is that that particular installation did not install the driver, not that the driver isn’t installed.

    What happens if you go to Terminal and run this?

    cat /proc/driver/nvidia/version


    Thank you for your reply.
    I got “No such file or directory” as the answer of your command.

    As I was not able to find a solution with ubuntu 14.04 32bit, I decided to install an ubuntu 14.04 64bit.
    Everything is now working perfectly so I supposed that the 32bit version was “kind of” not supported.

    Sorry for others who have the same problem than me, I do not have another solution than installing a 64bit version.

    Are you still using VMWare player?

    I assume the VM did not have a GPU in it. And you did not install the driver. So it’s no surprise that the installer reports no driver installed. A GPU in the host should not be necessary for cross-compiling anyway. And it’s also no surprise that cat /proc/driver/nvidia/version reports nothing, because you have no GPU in the VMWare player VM (I assume. Otherwise, please clarify.)

    It’s also not clear what you mean now that “everything is now working”. Are you still installing the 64bit Ubuntu OS in a VMWare player VM? Or are you now using a baremetal config (no VM/Hypervisor). If you are now in a baremetal config, and an NVIDIA GPU is installed in your host machine, then I would expect the driver behavior and reporting to be different.

    Having said all that, although ubuntu 14.04 32bit OS is supported (although “deprecated”):

    that support does not officially include cross-compile/cross-build targetting ARMv7 (i.e. Jetson).

    So using the Ubuntu 14.04 x86_64 path is preferred.