https://hgpu.org/?p=8939
Developing Performance-Portable Molecular Dynamics Kernels in OpenCL