https://hgpu.org/?p=3115
Task management for irregular-parallel workloads on the GPU