Joinable and Detached Threads
http://www.domaigne.com/blog/computing/joinable-and-detached-threads/
Introduction
By default, created threads are joinable. That means, we can wait for its completion in any other thread using the function pthread_join()
:
#include int pthread_join( pthread_t thread, // thread to join void **value_ptr // store value returned by thread );
This function shall suspend the calling thread until the targeted thread
terminates. When value_ptr
is non-NULL, then value_ptr
shall contain the value
returned by thread. We saw how to use pthread_join()
already in our article Pthreads argument passing.
Joinable Thread and Consequences
A thread may have already terminated before we join it. As a result, the Pthreads system must retain some information when a thread is joinable: namely, at least the thread ID and the returned value[1]. Indeed this information is needed for joining the thread.
Actually, the Pthreads system can retain also other resources
associated with the joinable thread, like for instance the thread’s
stack. This is perfectly legal from a POSIX standpoint. In fact, pthread_join()
guarantees to reclaim whatever storage is associated to the joined thread.
Detached Threads
What should I do when my thread doesn’t return anything useful, and I don’t have to wait for its completion? Do I have to call pthread_join()
anyway just for clean-up purpose?
Fortunately, not. Pthreads offers a mechanism to tell the system: I am starting this thread, but I am not interested about joining it. Please perform any clean-up action for me, once the thread has terminated. This operation is called detaching a thread. We can detach a thread as follows:
- During thread creation using the
detachstate
thread attribute. - From any thread, using
pthread_detach()
.
Let’s start with the second form, which is the easiest.
#include int pthread_detach( pthread_t thread, // thread to detach );