About The Project

RBMD (Random Batch Molecular Dynamics) is a GPU-CPU heterogeneous accelerating molecular dynamics software baed on random batch series algorithms.

(back to top)

Installation

Prerequisites

On Linux

The third-party library vtk-m needs to be installed. First, clone the VTK-m from github on your_folder

git clone https://gitlab.kitware.com/vtk/vtk-m.git

Then switch to the branch: v1.9.0 (cat version.txt to check the version of vtkm for install)

git checkout -b 1.9.0 v1.9.0

Finally, compile VTK-m with cmake (CUDA support required) , CMake/3.20, GCC/9.0, and CUDA/11.0 have been confirmed to be necessary for compiling VTK-m. Furthermore, the compilation process is related to the GPU driver, and if deploying on a server, it needs to be compiled on the compute nodes.

cmake .. -DBUILD_SHARED_LIBS=OFF -DCMAKE_BUILD_TYPE=Release -DVTKm_ENABLE_CUDA=ON -DVTKm_USE_64BIT_IDS=OFF -DVTKm_ENABLE_TESTING=OFF -DVTKm_ENABLE_RENDERING=OFF -DBUILD_TESTING=OFF -DCMAKE_INSTALL_PREFIX=your_folder/vtkm_out

And install in the vtkm-out

make -j$(nproc)
make install

For rbmd compilation,

1. Install the source code on your_folder and switch to the dev branch

   git clone https://github.com/randbatch-md/rbmd.git
   git checkout dev

2. cmake RBMD based on the same packages :CMake/3.22, GCC/9.3, and CUDA/11.8 in the build folder

   cmake .. -DCMAKE_BUILD_TYPE=Release -DCMAKE_LINKER=your_folder/rbmd/framework/tools/Linux/Release/jsoncpp/lib64

3. compile the executable file rbmd and then rbmd is on the framework folder

   make -j$(nproc)

Remark

1. If the step for make rbmd is wrong in the first time, the Linux folder needs to be deleted

   rm -rf your_folder/rbmd/framework/tools/Linux

   and complie again.

2. If in the end of make, the error cannot find -ljsoncpp is occured, the environment variables needs to be specified in the ~/.bashrc

   export LD_LIBRARY_PATH=your_folder/rbmd/framework/tools/Linux/Release/jsoncpp/lib64:$LD_LIBRARY_PATH
   export LIBRARY_PATH=your_folder/rbmd/framework/tools/Linux/Release/jsoncpp/lib64:$LIBRARY_PATH

Docker Image

You can also pull the docker image and run docker dircetly

docker pull ghcr.io/randbatch-md/rbmd:1.0.0
docker run --rm --gpus all -it -v $PWD:/app rbmd  /bin/bash -c "rbmd -j rbmd.json"

(back to top)

Getting started

run the example json file rbmd.json in framework, then rbmd is executed

./rbmd -j rbmd.json

Usage

For more examples, please refer to the https://www.randbatch.com/guide/

(back to top)

(back to top)

Contributing

(back to top)

License

Distributed under the GPL-3.0 License. See LICENSE.txt for more information.

(back to top)

Contact

Weizhu scientific computing platform Link: https://github.com/randbatch-md

(back to top)

Acknowledgments

(back to top)