https://docs.nvidia.com/cuda/cuda-programming-guide/04-special-topics/cluster-launch-control.html
4.13. Work Stealing with Cluster Launch Control — CUDA Programming Guide
cuda programming guidework stealinglaunch controlcluster
https://ayende.com/Blog/170625/work-stealing-in-the-presence-of-startup-shutdown-costs
Work stealing in the presence of startup / shutdown costs - Ayende @ Rahien
Work stealing in thread pools: reduce startup and shutdown costs, improve parallel indexing performance, and dynamically allocate cores.
work stealingthe presencestartupshutdowncosts