XueLab

This page is read only. You can view the source, but not change it. Ask your administrator if you think this is wrong.
====== Install Nvidia graphic driver and CUDA ======
  * [[https://linuxconfig.org/how-to-disable-nouveau-nvidia-driver-on-ubuntu-18-04-bionic-beaver-linux | How to disable nouveau driver]]
  * Install kernel source in Debian 8 ([[http://www.tuicool.com/articles/yiU77v | Ref]])
<code>
# install kernel source
cd build
mkdir kernel
cd kernal
sudo apt-get install dpkg-dev
uname -r  #the output should be included in $(uname -r)

need to change the source of mirror into Tsinghua
vi /etc/apt/source.list
change deb-src into ??? #please remember to comment out security otherwise error would be reported.

#and then update
apt-get update

apt-get source linux-image-$(uname -r)
# finally copy source to system folder
mv linux-xxx /usr/src

# install kernel header
sudo apt-get install linux-headers-$(uname -r)
</code>

  * Install graphic driver
  * 20190819-log: this step has been incorporated in the installation of cuda
<code>
#download this one to the uda folder,then install it.
 * [[https://www.nvidia.com/en-us/drivers/unix/linux-amd64-display-archive  |  older version]]
sudo ./NVIDIA-Linux-x86_64-375.20.run
#accept license
#install binaries
#always continue/yes
</code>
    #pls noticed that this is step no longer need because we need latest CUDA which included latest NVIDIA already!

  * Install cuda
<code>
# create some soft links otherwise cuda installation
# will throw out warning message:
# Missing recommended library: libGLU.so
# Missing recommended library: libXi.so
# Missing recommended library: libXmu.so
#
# ref:  http://kmdouglass.github.io/stories/notes/cuda.html
#
# the following commands are likely not necessary!
#sudo apt-get install libglu1-mesa-dev
#sudo apt-get install libxi-dev
#sudo apt-get install libxmu-dev

cd /usr/lib
sudo ln -s x86_64-linux-gnu/libGLU.so.1.3.1 libGLU.so
sudo ln -s x86_64-linux-gnu/libXi.so.6.1.0 libXi.so
sudo ln -s x86_64-linux-gnu/libXmu.so.6.2.0 libXmu.so

# install cuda
# During installation, choose to skip driver installation (we did it as shown above)
sudo ./cuda_8.0.44_linux.run</code>
    *#accept previously read EULA
    *install on an unsupported configuration:yes
    *#install NVIDIA
    *#install OpenGL
    *#run NVidia-xconfig
    *#install CUDA 8.0|9.2 Toolkit
    *#enter toolkit location:/usr/local/cuda
    *#install a symbolic link
    *#install cuda toolkit
    *#toolkit location /usr/local/cuda-8.0
    *#install CUDA 8.0 samples
    *#CUDA sample Location

it turns out that cd /home/panying/NVIDIA_CUDA-8.0_samples/
<code>cd 1_Utilities/deviceQuery
make
</code>

<code>mv /home/panying/NVIDIA_CUDA-8.0_samples/ /home/usr/share/
cd NVIDIA_CUDA-8.0_samples/1_Utilities/deviceQuery
./deviceQuery
#pls check carefully and make sure that everything goes well with your GPU devices.
</code>

====== Troubleshooting ======
  * error: token ""__CUDACC_VER__ is no longer supported. SOLUTION: comment out relevant line in /usr/local/cuda/include/crt/common_functions.h. [[https://github.com/BVLC/caffe/issues/5994 | Ref]]
  * math_functions.hpp not found when using CUDA with Eigen [[https://stackoverflow.com/questions/43113508/math-functions-hpp-not-found-when-using-cuda-with-eigen | Ref]]  [[https://stackoverflow.com/questions/43113508/math-functions-hpp-not-found-when-using-cuda-with-eigen | aother possible solution]]
  * 20181228-panying，fix this bug in both orange and blue

========log========
* 20190820-log-install on ubuntu server 16.0 Xubuntu:https://askubuntu.com/questions/799184/how-can-i-install-cuda-on-ubuntu-16-04
XueLab

User Tools

Site Tools

Page Tools
