How do you think about use libfork for DAG tasks

[Frandy]

Libfork is great for parallel-for or nested job. How about DAG tasks ? We can maintain task dependency and call libfork for parallel. It is one kind of nested jobs, but the difference is it does not need to wait for child to finish. For DAG task parallel system, we can find taskflow and scheduling. For very large DAG, will it cause problem with too deep nest ?

[ConorWilliams]

Libfork implicitly models all tasks as a DAG (this is what enables so much of the theoretical analysis); libfork can handle arbitrarily deeply nested DAGs with its dynamic stacks. Waiting for child tasks to complete is one of the key constraints of strict fork-join parallelism, that allows the strong memory/time bounds, so we will unfortunetly not be loosening this constraint.