https://hgpu.org/?p=2301
FPGA-based acceleration of CHARMM-potential minimization