Skip to content
This repository has been archived by the owner on Apr 8, 2024. It is now read-only.

(ADVANCED) setup an expipe server with GPU support

Jump to bottom
Alessio Buccino edited this page Mar 7, 2019 · 2 revisions

This page is a step-by-step guide to setup a computer with GPU support for the latest spike sorting software (Kilosort1-2, Ironclust) - date: 07/03/2019

Software versions (IMPORTANT):

  • Ubuntu 18.04
  • MATLAB 2018b
  • NVIDIA drivers
  • CUDA 9.1

Install MATLAB 2018b

mkdir ~/apps
sudo rsync  -a  app-sync.uio.no::rhel6-x86_64-matlab  ~/apps/matlab
sudo apt-get install -y matlab-support
sudo chmod -R 776 .matlab/
sudo chown -R $USER:$USER /home/$USER/.matlab/

Install NVIDIA drivers

sudo apt-add-repository ppa:graphics-drivers/ppa
sudo apt update
sudo apt install nvidia-drivers-415 (or latest)

Install CUDA 9.1

  • Download CUDA 9.1 (select Linux -> x86_64 -> Ubuntu-17.04 -> runfile)

run: sudo sh cuda_9.1.85_387.26_linux.run --override

And choose:

  • EULA: accept
  • unsupported: y
  • install NVIDIA drivers: n
  • install CUDA toolkit: y
  • location: default
  • sym: y
  • location: default

Add PATH=$PATH:/usr/bin/cuda-9.1/bin to .bashrc

Change CUDA source code for MATLAB compatibility

In order to have MATLAB run with the CUDA version, we have to make a slight change to the cuda codes:

cd /usr/local/cuda/bin/..//include/crt/
sudo gedit host_config.h

Change line 119 from:

#if __GNUC__ > 6

to

#if __GNUC__ > 8

If you have problems, contact Alessio Buccino

Clone this wiki locally