User level threads and kernel level threads

Threading model

One to one thread model

Many to one threading model

N Yes M Thread model of

There are two ways to implement threads

User-level threads

Implement thread instances in user mode , Use one TCB(Thread Control Block) To describe a user thread , The TCB Stored in user mode .

1. The kernel can't see the user level thread , For the kernel , You can only see the current process , Threads created in the current process cannot be scheduled by the operating system ;
2. User level thread creation , The destruction , And switching is extremely fast , There is no need to switch back and forth between user mode and kernel mode ;
3. When a user thread blocks or gets stuck in the kernel , Now the whole process will be blocked , Other user threads will not be able to use ( Consider using asynchronous invocation to mitigate );
4. All user threads in the process need to actively give up CPU, Otherwise, it will CPU The time slice is gone , Because for the kernel , You can only see the process , Only time slices can be assigned to the current process ;
5. Each process can customize the thread scheduling algorithm .

User level threads are implemented by thread libraries in some applications .

Kernel level threads

Implement thread instances in kernel mode , Use one TCB(Thread Control Block) To describe a user thread , The TCB Stored in kernel mode , The kernel can directly perform scheduling management .

1. Thread creation , The overhead of termination and switching is large , It is all done by the kernel ;
2. The kernel is aware of threads , Managed by the operating system , Scheduled by the operating system ;
3. The allocation of kernel time slice is based on threads , The kernel allocates an entire slice of time to threads .

coroutines

coroutines = User-level threads + Scheduler