https://hgpu.org/?p=3120
Performance study of mapping irregular computations on GPUs