https://hgpu.org/?p=2920
Data parallel loop statement extension to CUDA: GpuC