Installing via APT

We can install LAMMPS using apt.

sudo apt install lammps

The example files can be found in /usr/share/lammps.

cp /usr/share/lammps/examples/melt ~/Documents
cd ~/Documents/melt
lmp -in in.melt

Hard to believe it’s this easy to get LAMMPS running—what a time to be alive.

Building from Source

If we want to use the GPU package, we’ll need to build LAMMPS from source.

It requires installing several dependencies, and I got it working after a fair bit of trial and error.

git, cmake, cuda, Required Packages

We’ll need to install git, cmake, cuda, and various development libraries.

sudo apt install git cmake nvidia-cuda-toolkit build-essential libopenmpi-dev libjpeg-dev libpng-dev ffmpeg

git clone

git clone -b release https://github.com/lammps/lammps.git lammps

Building

Refer to the official guide:

https://docs.lammps.org/Build_cmake.html ⧉

cd lammps
mkdir build; cd build
cmake -C ../cmake/presets/basic.cmake -D PKG_GPU=yes -D GPU_API=cuda -D GPU_ARCH=sm_89 -D PKG_OPENMP=yes -D WITH_JPEG=yes -D WITH_PNG=yes -D WITH_FFMPEG=yes ../cmake
cmake --build .
make install

GPU_ARCH depends on your GPU model. You can look it up here:

https://arnon.dk/matching-sm-architectures-arch-and-gencode-for-various-nvidia-cards/ ⧉

Some guides suggest including -DBIN2C=/usr/local/cuda/bin/bin2c, but I was able to build it without that. Not sure what difference it makes—worth investigating.

Add to PATH

Add the appropriate paths to your .bashrc.

export PATH=/usr/username/.local/bin:PATH

Running

cd ~/Documents/melt
lmp -pk gpu 1 -in in.melt

If you see an error about OMP_NUM_THREADS not being set, add something like the following to your .bashrc (adjust the number to suit your environment):

export OMP_NUM_THREADS=2