https://hgpu.org/?p=26891
Modeling GPU Dynamic Parallelism for Self Similar Density Workloads