https://hgpu.org/?p=4900
Effective Dynamic Scheduling on Heterogeneous Multi/Manycore Desktop Platforms