Good question.

To a coroutine library, there is two kinds of CPU pinning,

• Bind the thread to CPU.
• Bind the coroutine(in libcsp it's called process) to CPU.

Currently libcsp only implements the first one(that's why it's so fast). Sometimes it may be a bad idea to bind a task process to a CPU since other CPUs may starve. What we can do is to try the best to do it without guarantee. Anyway I'll try to implement the second one.

I would personally be interested indeed in the second one. Let the OS decides about which process starves since this would be the user's choice. However, it would perhaps be best to allow/enable users with ways to query the current scheduling state done by libcsp. That's mostly for debugging concerns.

Yeah, it's important to add the related debugging tools. Libcsp will support it in the future.