62

I have installed cuda8.0, the latest version. I followed the procedure provided by Nvidia; but, when I type the command nvcc --version it says nvcc is not installed!

What do I do now?

10
  • do a which nvcc see if it returns any thing... Feb 21, 2017 at 4:45
  • No it didnt return Feb 21, 2017 at 4:46
  • That means it didn't install well or is not in your PATH, find where it is with whereis nvcc.. Feb 21, 2017 at 4:49
  • it only displayed "nvcc: " Feb 21, 2017 at 4:56
  • Close and launch terminal and try nvcc --version to be sure Feb 21, 2017 at 5:01

7 Answers 7

114

The problem is [ based on the link you provided ] you haven't added it in the .bashrc. file so it can be seen:

  1. From the terminal:

     nano /home/username/.bashrc
    
     # or
    
     nano /home/$USER/.bashrc
    
  2. Inside there add the following: (replace cuda-8.0 with your version)

     export PATH="/usr/local/cuda-8.0/bin:$PATH"
     export LD_LIBRARY_PATH="/usr/local/cuda-8.0/lib64:$LD_LIBRARY_PATH"
    
  3. Then do the following to save and close the editor:

     On you keyboard press the following: 
    
     ctrl + o             --> save 
     enter or return key  --> accept changes
     ctrl + x             --> close editor
    
  4. Now either do source .bashrc or close and open another terminal

  5. Now run nvcc --version

Information:

  • .bashrc: is the file read by the terminal before opening and its found in the /home/$USER diretory of the user in question.
  • the . before the file means its hidden from view unless you instruct your file manager to show hidden files
1
15

The above solution by @George Udosen is fine. If you want to save the manual procedure, you can automate it by the following:

  1. Create a file add_to_bashrc and add the following to it:

    export CUDA_HOME=/usr/local/cuda
    export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/cuda/lib64:/usr/local/cuda/extras/CUPTI/lib64
    export PATH=$PATH:$CUDA_HOME/bin
    
  2. Create a shell script automate.sh:

    (... some installation procedure ...)
    
    cat add_to_bashrc >> ~/.bashrc
    . ~/.bashrc
    
  3. Then you just need to run your shell script:

    sh automate.sh
    
  4. Don't forget to check if the CUDA's shortcut (symLink) works correctly. Simply execute:

    ls /usr/local/cuda
    
0
7

For later versions of CUDA (11.7 for me), I followed the below steps:

  1. Go to https://developer.nvidia.com/cuda-downloads and follow the installation instructions.

  2. Inside the ~/.bashrc file, update the path as follows:

    export PATH="/usr/local/cuda-11.7/bin:$PATH"`
    export LD_LIBRARY_PATH="/usr/local/cuda-11.7/lib64:$LD_LIBRARY_PATH"
    
  3. The output of nvcc -V then looks like:

    nvcc: NVIDIA (R) Cuda compiler driver
    Copyright (c) 2005-2022 NVIDIA Corporation
    Built on Wed_Jun__8_16:49:14_PDT_2022
    Cuda compilation tools, release 11.7, V11.7.99
    Build cuda_11.7.r11.7/compiler.31442593_0
    
6

The answer from @George Udosen is perfect.

Just for increment it, you can also export to /usr/local/cuda which is a symbolic link to /usr/local/cuda-10.1, based on this answer. So, you can also write:

export PATH=/usr/local/cuda/bin${PATH:+:${PATH}}$ 
export LD_LIBRARY_PATH=/usr/local/cuda/lib64${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}}
1

Terminal:

sudo gedit /home/$USER/.bashrc

Add at the end of the file

export CUDA_HOME=/usr/local/cuda
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/cuda/lib64:/usr/local/cuda/extras/CUPTI/lib64
export PATH=$PATH:$CUDA_HOME/bin

Then terminal:

source .bashrc

Now run

nvcc --version

2
  • This is not an answer. All you have here is a repeat of the accepted answer.
    – David
    Apr 29, 2023 at 8:38
  • I don't deny that, but I believe my answer can help many people struggling to edit the path correctly. Dec 3, 2023 at 16:46
0

The above solution by @Agile Bean is fine. If you want to you can use the env var CUDA_HOME instead of /usr/local/cuda in the LD_LIBRARY_PATH and edit your .bashrc or .zshrc file or the resource file for your shell with emacs or vi or nano or your favorite editor:

  1. emacs $HOME/.zshrc

  2. Add the following lines to the .zshrc file export CUDA_HOME=/usr/local/cuda

export LD_LIBRARY_PATHlrwxrwxrwx 1 root root 20 Sep 11 19:59 =$LD_LIBRARY_PATH:$CUDA_HOME/lib64:$CUDA_HOME/extras/CUPTI/lib64

export PATH=$PATH:$CUDA_HOME/bin

  1. Then open a new shell or source your shell's resource file (e.g. source .zshrc) from the current shell.

(base) ➜ ~ nvcc --version
nvcc: NVIDIA (R) Cuda compiler driver Copyright (c) 2005-2022 NVIDIA Corporation Built on Wed_Jun__8_16:49:14_PDT_2022 Cuda compilation tools, release 11.7, V11.7.99 Build cuda_11.7.r11.7/compiler.31442593_0 (base) ➜ ~ which nvcc
/usr/local/cuda/bin/nvcc

Where is cuda really located? Let's follow the symbolic links to find out.

(base) ➜ ~ ls -ld /usr/local/cuda

/usr/local/cuda -> /etc/alternatives/cuda

(base) ➜ ~ ls -ld /etc/alternatives/cuda

/etc/alternatives/cuda -> /usr/local/cuda-11.7

0

I am using Asus VivoBook X540UBR (date: 01/08/2020) which comes with entry level NVIDIA GeForce MX110.

2 years ago, I abandoned Windows 10 and re-installed the notebook with Linux Mint 21.2 Cinnamon.

To my surprise, Linux has properly installed and maintained NVIDIA drivers via

$ sudo apt update
$ sudo apt dist-upgrade

I don't need to explicitly install Cuda toolkit from https://developer.nvidia.com/cuda-downloads?target_os=Linux&target_arch=x86_64&Distribution=Ubuntu&target_version=20.04&target_type=deb_local nor explicitly configure the environment path.

However, I installed Cuda toolkit as hinted by apt via

$ sudo apt install nvidia-cuda-toolkit

Then, my GPU environment is ready.

To check the Cuda version in Ubuntu/Mint:

$ nvidia-smi
+--------------------------------------------------------------------------+
| NVIDIA-SMI 535.161.07  Driver Version: 535.161.07  CUDA Version: 12.2    |
|----------------------------+----------------------+----------------------+
... (ignore the rest) ...

To check whether Cuda toolkit is installed:

$ nvcc --version
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2021 NVIDIA Corporation
Built on Thu_Nov_18_09:45:30_PST_2021
Cuda compilation tools, release 11.5, V11.5.119
Build cuda_11.5.r11.5/compiler.30672275_0

I am in the middle of studying https://youtu.be/Z_ikDlimN6A?t=4906 and installed pytorch=1.10.0.

>>> import torch
>>> print(torch.cuda.is_available())
>>> print(torch.__version__)

True
1.10.0

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .