https://hgpu.org/?p=6329
Using the High Productivity Language Chapel to Target GPGPU Architectures