Cluster der kan alt/Installation af CUDA

Installation af cuda

Follow this tutorial: http://www.quantstart.com/articles/Installing-Nvidia-CUDA-on-Ubuntu-14-04-for-Linux-GPU-Computing

Note: Suggest that you navigate to sub-directories in the samples directory and "make" individual projects. To "make" the whole samples directory (as the tutorial suggests) takes around an hour!

Note: .bash_profile is actually just called .profile

CUDA + OpenMPI install script:

#!/bin/bash
#####################################
# Made by Jesper Larsen, 28-10-2014 #
#####################################
#
# CUDA + OpenMPI installer
#
echo "Preparing for install..."
apt-get update
apt-get install -y build-essential
echo "Creating folders..."
mkdir ~/installdir
mkdir ~/installdir/dl
mkdir ~/installdir/dl/CUDA
# Downloading, installing deb.
if [ -f "/home/hpcadmin/installdir/dl/CUDA/cuda-repo-ubuntu1404_6.5-14_amd64.deb" ];
then
	echo "cuda6_5_rev14 .deb package found."
	echo "Installing: cuda65rev14..."
	dpkg -i /home/hpcadmin/installdir/dl/CUDA/cuda-repo-ubuntu1404_6.5-14_amd64.deb
else
	echo "cuda6_5_rev14 .deb package _NOT_ found."
	echo "Downloading and installing: cuda6_5_rev14..."
	wget http://developer.download.nvidia.com/compute/cuda/repos/ubuntu1404/x86_64/cuda-repo-ubuntu1404_6.5-14_amd64.deb /home/hpcadmin/installdir/dl/CUDA/
	dpkg -i /home/hpcadmin/installdir/dl/CUDA/cuda-repo-ubuntu1404_6.5-14_amd64.deb
fi
apt-get update
apt-get install -y cuda 
echo "export PATH=/usr/local/cuda-6.5/bin:$PATH" >> ~/.profile
echo "export LD_LIBRARY_PATH=/usr/local/cuda-6.5/lib64:$LD_LIBRARY_PATH" >> ~/.profile
source ~/.profile
#Installing OpenMPI
echo "Installing OpenMPI"
apt-get install -y libopenmpi-dev openmpi-bin openmpi-doc
echo "========================================="
echo "!!! CUDA + OpenMPI - INSTALL COMPLETE !!!"
echo "========================================="
echo "RESTART IN 20 SECONDS FOR DRIVERS TO WORK"
echo ""
echo "Before proceeding to test the GPU cards we will ensure that the drivers are"
echo "correctly installed. The following line will provide us with the driver version."
echo "Run this for testing after reboot: sudo cat /proc/driver/nvidia/version"
sleep 20
reboot now

CUDA samples compiling and test script:

#!/bin/bash
#####################################
# Made by Jesper Larsen, 28-10-2014 #
#####################################
#
# CUDA samples compile and test GPU
# 
# Getting: Hostname, IP, Current date
IP=$(hostname -I | awk '{print $1}')
Host=$(host $IP | awk '{print $5}' | cut -d '.' -f1)
DATE=`date +%Y_%m_%d-%H_%M`
#Script:
compile_cuda6_5_rev14="/home/hpcadmin/installdir/testing/CUDA/NVIDIA_CUDA-6.5_Samples/bin/"
if [ -d "/home/hpcadmin/installdir/testing/CUDA/NVIDIA_CUDA-6.5_Samples/bin/" ];
then
	echo "$compile_cuda6_5_rev14 found! Testing instead."
	echo "Testing communication with the GPU..."
	exec /home/hpcadmin/installdir/testing/CUDA/NVIDIA_CUDA-6.5_Samples/bin/x86_64/linux/release/deviceQuery > /home/hpcadmin/installdir/testing/CUDA/results/$Host-$DATE.txt
else
	echo "$compile_cuda6_5_rev14 _NOT_ found. Compiling then testing."
	echo "Copying CUDA samples to testing directory..."
	sh /usr/local/cuda-6.5/bin/cuda-install-samples-6.5.sh /home/hpcadmin/installdir/testing/CUDA/
	echo "Compiling CUDA samples - TAKES A LONG TIME 30-40 min..."
	make -C /home/hpcadmin/installdir/testing/CUDA/NVIDIA_CUDA-6.5_Samples/
	echo "Testing communication with the GPU..."
	exec /home/hpcadmin/installdir/testing/CUDA/NVIDIA_CUDA-6.5_Samples/bin/x86_64/linux/release/deviceQuery > /home/hpcadmin/installdir/testing/CUDA/results/$Host-$DATE.txt
fi