It's an asymmetric coroutine library (like lua).
You can use coroutine_open to open a schedule first, and then create coroutine in that schedule.
You should call coroutine_resume in the thread that you call coroutine_open, and you can't call it in a coroutine in the same schedule.
Coroutines in the same schedule share the stack , so you can create many coroutines without worry about memory.
But switching context will copy the stack the coroutine used.
Read source for detail.
Chinese blog : http://blog.codingnow.com/2012/07/c_coroutine.html