Skip to content

Latest commit

 

History

History
254 lines (171 loc) · 5.27 KB

guide-monoalg3d-fedora.md

File metadata and controls

254 lines (171 loc) · 5.27 KB

GUIDE TO CONFIGURE MONOALG_3D ON A FEDORA 40 MACHINE

This guide provides instructions to configure the MonoAlg3D solver in a fresh instalation of Fedora 40 with all its features working.

1) Download and build a USB boot drive

Download the Fedora 40 ISO image for your system:

Perform a system update:

$ sudo dnf update

Change the graphic mode to "GNOME over Xorg"

  • Close your current session
  • Click on your user
  • Click on the gear symbol and select the "GNOME over Xorg"
  • Begin your session again

Reboot your system

$ reboot

Install the GNOME-Tweak-Tools

$ sudo dnf install gnome-tweak-tool

Open Configurations

  • Disable energy saving options
  • Disable the suspend screen when notebook is closed

Install gedit

$ sudo dnf gedit

Install the Resources program to replace the System Monitor

  • Search for the 'Programs' software
  • Open 'Programs' and search for the Resources program
  • Install using the Flathub package

Install the Engrampa package manager

  • Get the 'Engrampa' software from Programs

2) Install RPMFusion

Add the packages to your system:

$ sudo dnf install https://download1.rpmfusion.org/free/fedora/rpmfusion-free-release-$(rpm -E %fedora).noarch.rpm https://download1.rpmfusion.org/nonfree/fedora/rpmfusion-nonfree-release-$(rpm -E %fedora).noarch.rpm
$ sudo dnf update
  • Reboot your system
$ reboot
  • Enter your BIOS by pressing F12 (check your computer key)
  • Disable the Secure Boot
  • Save your changes and exit

3) Install the NVIDIA driver

Run the following commands

$ sudo dnf update
$ sudo dnf install akmod-nvidia
$ sudo dnf install xorg-x11-drv-nvidia-cuda

Wait around 5 to 8 minutes in order to the NVIDIA kernels get built

  • Check if the kernels are ready
$ modinfo -F version nvidia
  • This should return the version of the driver such as 465.27

Reboot your system

$ reboot

Check if the NVIDIA driver is correctly installed

  • Go to Settings > System > About > System Details
  • Your NVIDIA Graphics Card should appear in Graphics field.

4) Install CUDA libraries

Download the CUDA runfile for your operating system

$ cd ~; mkdir CUDA-Install; cd CUDA-Install
$ wget -nc https://developer.download.nvidia.com/compute/cuda/12.6.3/local_installers/cuda_12.6.3_560.35.05_linux.run
  • Here, I am using the CUDA version 12.6.3.

Install dependencies

$ sudo dnf install gcc-c++ mesa-libGLU-devel libX11-devel libXi-devel libXmu-devel 
$ sudo dnf install freeglut freeglut-devel
$ sudo dnf install gcc13-c++-13.3.1-2.fc40.x86_64

Run the CUDA runfile

$ chmod +x cuda_12.6.3_560.35.05_linux.run
$ sudo ./cuda_12.6.3_560.35.05_linux.run

Configure the installation

  • Accept the terms
  • IMPORTANT! Disable the NVIDIA Driver installation
  • Enable the installation: CUDA Toolkit, CUDA Samples, CUDA Demo, CUDA Documentation
  • Install the package

Post installation steps

Log as root user

$ su -
  • If is the first time logging as root, you can set a password to the "root" user with:
$ sudo passwd root

Set the PATH and LD_LIBRARY_PATH by creating a shell configuration file for CUDA

cat << EOF > /etc/profile.d/cuda.sh
pathmunge /usr/local/cuda/bin before

if [ -z "${LD_LIBRARY_PATH}" ]; then
    LD_LIBRARY_PATH=/usr/local/cuda/lib64
else
    LD_LIBRARY_PATH=/usr/local/cuda/lib64:$LD_LIBRARY_PATH
fi

export PATH LD_LIBRARY_PATH
EOF

Logout and login again. Check the variables

$ echo $PATH
$ echo $LD_LIBRARY_PATH
  • The new paths must appear.

Test the CUDA library

$ cd /home/<username>/NVIDIA_CUDA-12.6_Samples/1_Utilities/deviceQuery
$ make
$ ./deviceQuery
  • This should print some information about your GPU hardware.

5) Configure MonoAlg3D_C

Download remaining dependencies

$ sudo dnf install libXcursor-devel libXrandr-devel libXinerama-devel

Make a symbolic link for GCC version 13

$ sudo ln -s /usr/bin/gcc-13 /usr/local/cuda/bin/gcc
$ sudo ln -s /usr/bin/g++-13 /usr/local/cuda/bin/g++

Build and compile MonoAlg3D

$ ./build.sh -f

IMPORTANT! If you stop here you already have a working MonoAlg3D binary executable in the /bin folder

[OPTIONAL] Install and configure MPI for batch simulations

Download OpenMPI

$ cd ~; mkdir OpenMPI-build; cd OpenMPI-build
$ wget -nc https://download.open-mpi.org/release/open-mpi/v5.0/openmpi-5.0.6.tar.gz

Build OpenMPI

$ gunzip -c openmpi-5.0.6.tar.gz | tar xf -
$ cd openmpi-5.0.6
$ ./configure --prefix=/usr/local
$ sudo make all install

Configure OpenMPI for MonoAlg3D

  • Change the following function in the 'bsbash/find_functions.sh' file:
FIND_MPI () {
	MPI_FOUND="y"
	MPI_LIBRARIES="mpi"
	MPI_LIBRARY_PATH=$OMPI_LIBRARY_PATH
	MPI_INCLUDE_PATH=$OMPI_INCLUDE_PATH
}
  • Rebuild MonoAlg3D
./build.sh -f
  • Check if you have the "MonoAlg3D_batch" executable inside the "/bin" folder.