https://hgpu.org/?p=12770
Characterization and Analysis of Dynamic Parallelism in Unstructured GPU Applications