https://hgpu.org/?p=2115
A massively multicore parallelization of the Kohn-Sham energy gradients