How to install NVIDIA CUDA Toolkit on Fedora 29 Linux - LinuxConfig.org

LinuxConfig.org Articles
1
The Nvidia CUDA toolkit is an extension of GPU parallel computing platform and programming model.
This is a companion discussion topic for the original entry at https://linuxconfig.org/how-to-install-nvidia-cuda-toolkit-on-fedora-29-linux
2

Thank you very much for the commands you provided.
Using $ sudo dnf --releasever=27 install gcc-7.3.1-6.fc27 is indeed really dangerous. A way around is to install a seperate compiler and give a link to that compiler.

To hack gcc compiler problem use negativo’s repo.
sudo dnf config-manager --add-repo=https://negativo17.org/repos/fedora-nvidia.repo
Install compilers
sudo dnf install cuda-gcc.x86_64 cuda-gcc-c++.x86_64

Change gcc and g++ binary paths for cuda:

sudo ln -s /usr/bin/cuda-gcc /usr/local/cuda/bin/gcc
sudo ln -s /usr/bin/cuda-g++ /usr/local/cuda/bin/g++

before running a build command provide CC and CXX flags. for example:
CC=cuda-gcc CXX=cuda-g++ python setup.py build develop

That worked for me I hope it works for someone else also.

1 Like
3

mahiratmis
Ok, but not works cuda-gcc Torch7 install (github nagadomi/distro)

CC=cuda-gcc CXX=cuda-g++ ./install.sh
…

/usr/libexec/gcc/x86_64-redhat-linux/8.3.0/cc1: error while loading shared libraries: libmpfr.so.4: cannot open shared object file: No such file or directory
gmake[1]: *** [CMakeFiles/cmTC_286af.dir/build.make:83: CMakeFiles/cmTC_286af.dir/testCCompiler.c.o] Error 1
gmake[1]: Leaving directory ‘/home/dankahazi/torch/build/CMakeFiles/CMakeTmp’
gmake: *** [Makefile:138: cmTC_286af/fast] Error 2

4

Hi Idanka,

Welcome to our forums.

This error message suggests you are missing that libmpfr.so.4 object file, which is in the mpfr package. Have you tried to install it?

5

Hi!
thx, but omg i not add negativo17 - org repo
i use rpmfusion - org

i will try the other package