When installing CUDA on Ubuntu 20.04, you may encounter an “unsupported compiler version” error. This typically happens when the version of the gcc compiler installed on your system is not compatible with the CUDA version you’re trying to install. In this article, we will guide you through the steps to fix this error.
To fix the "unsupported compiler version" error when installing CUDA on Ubuntu 20.04, you need to install gcc version 8 and update the gcc and g++ alternatives to use version 8. After that, you can run the CUDA installation command with the –override argument to force the installation. Finally, add the necessary symlinks for CUDA and update your PATH and LD_LIBRARY_PATH environment variables.
- Prerequisites
- Step 1: Make the Runfile Executable
- Step 2: Install GCC-8
- Step 3: Update GCC and G++ Alternatives
- Step 4: Verify the GCC Version
- Step 5: Run the CUDA Installation Command
- Step 6: Check the Installation Log
- Step 7: Add Necessary Symlinks for CUDA
- Step 8: Update Your PATH and LD_LIBRARY_PATH
- Step 9: Apply the Changes
- Conclusion
Prerequisites
Before we start, make sure you have downloaded the CUDA installation runfile from the official Nvidia website.
Step 1: Make the Runfile Executable
Open a terminal and navigate to the directory where the runfile is located. Run the following command to make the runfile executable:
chmod +x cuda_10.2.89_440.33.01_linux.runIn this command, chmod +x is used to change the permission of the file and make it executable. The cuda_10.2.89_440.33.01_linux.run is the name of the runfile you downloaded from the Nvidia website.
Step 2: Install GCC-8
CUDA 10.2 requires gcc version 8. To install it, run the following command:
sudo apt install gcc-8The sudo command is used to run the command as the root user, apt install is used to install new packages, and gcc-8 is the package name for the gcc version 8.
Step 3: Update GCC and G++ Alternatives
Next, update the gcc and g++ alternatives to use version 8 by running the following commands:
sudo update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-8 8
sudo update-alternatives --install /usr/bin/g++ g++ /usr/bin/g++-8 8The update-alternatives command is used to maintain symbolic links determining default commands. The --install option is used to add a group of alternatives to the system. /usr/bin/gcc and /usr/bin/g++ are the master links, gcc and g++ are the names of their groups, /usr/bin/gcc-8 and /usr/bin/g++-8 are the paths to the alternative versions, and 8 is the priority of the alternatives.
Step 4: Verify the GCC Version
To verify that the gcc version has been updated, run:
gcc -vThe output should show the installed version as 8.x.x.
Step 5: Run the CUDA Installation Command
Now, run the CUDA installation command with the --override argument to force the installation:
sudo sh cuda_10.2.89_440.33.01_linux.run --overrideThe --override option forces the installer to ignore the unsupported compiler error and proceed with the installation.
Step 6: Check the Installation Log
If the installation fails, you can check the log file for more details by running:
cat /var/log/cuda-installer.logStep 7: Add Necessary Symlinks for CUDA
To add the necessary symlinks for CUDA, run the following commands:
sudo ln -s /usr/bin/gcc-8 /usr/local/cuda-10.1/bin/gcc
sudo ln -s /usr/bin/gcc-ar-8 /usr/local/cuda-10.1/bin/ar
sudo ln -s /usr/bin/gcc-ranlib-8 /usr/local/cuda-10.1/bin/ranlib
sudo ln -s /usr/bin/gcc-nm-8 /usr/local/cuda-10.1/bin/nmThe ln -s command is used to create symbolic links. This step ensures that CUDA uses the correct version of gcc.
Step 8: Update Your PATH and LD_LIBRARY_PATH
Update your PATH environment variable to include the CUDA binaries by adding the following line to your .profile or .bashrc file:
export PATH=/usr/local/cuda-10.1/bin:$PATHUpdate your LD_LIBRARY_PATH environment variable to include the CUDA libraries by adding the following line to your .profile or .bashrc file:
export LD_LIBRARY_PATH=/usr/local/cuda-10.1/lib64:$LD_LIBRARY_PATHThe export command sets the environment variables. PATH is used to specify directories where executable programs are located, and LD_LIBRARY_PATH is used to specify directories where libraries should be searched for first.
Step 9: Apply the Changes
Save the file and run the following command to apply the changes:
source ~/.profileThe source command reads and executes commands from the file specified as its argument in the current shell environment.
Conclusion
Now, you should be able to run the make files on the CUDA samples and get a working demo. If you encounter any further issues, please provide more details for further assistance. This process should help you to overcome the “unsupported compiler version” error when installing CUDA on Ubuntu 20.04.
No, CUDA requires a specific version of gcc for compatibility. Updating the gcc version is necessary to fix the "unsupported compiler version" error.
You can download the CUDA installation runfile from the official Nvidia website at https://developer.nvidia.com/cuda-downloads.
It is recommended to install the required gcc version (in this case, gcc-8) alongside your existing version. You can have multiple versions of gcc installed simultaneously without conflicts.
You can check the current version of gcc by running the command gcc -v in the terminal. The output will display the installed version.
If the installation fails, you can check the installation log file for more details. Run the command cat /var/log/cuda-installer.log to view the log and troubleshoot the issue.
Yes, updating the PATH and LD_LIBRARY_PATH environment variables is necessary to ensure that CUDA is correctly detected and utilized by your system. Follow the steps mentioned in the article to update these variables.
Yes, you can apply a similar fix for other versions of CUDA by adjusting the corresponding version numbers in the commands and paths mentioned in the article. However, make sure to check the specific requirements for the CUDA version you intend to install.
