The pthread_create() function shall create a new thread, with attributes
specified by attr, within a process. If attr is NULL, the
default attributes shall be used. If the attributes specified by attr
are modified later, the thread's attributes shall not be affected. Upon
successful completion, pthread_create() shall store the ID of the
created thread in the location referenced by thread.
The thread is created executing start_routine with arg as its sole
argument. If the start_routine returns, the effect shall be as if there
was an implicit call to pthread_exit() using the return value of
start_routine as the exit status. Note that the thread in which
main() was originally invoked differs from this. When it returns from
main(), the effect shall be as if there was an implicit call to
exit() using the return value of main() as the exit status.
The signal state of the new thread shall be initialized as follows:
The signal mask shall be inherited from the creating
The set of signals pending for the new thread shall be
The alternate stack shall not be inherited.
The floating-point environment shall be inherited from the creating thread.
If pthread_create() fails, no new thread is created and the contents of
the location referenced by thread are undefined.
If _POSIX_THREAD_CPUTIME is defined, the new thread shall have a CPU-time clock
accessible, and the initial value of this clock shall be set to zero.
A suggested alternative to pthread_create() would be to define two
separate operations: create and start. Some applications would find such
behavior more natural. Ada, in particular, separates the "creation"
of a task from its "activation".
Splitting the operation was rejected by the standard developers for many
The number of calls required to start a thread would
increase from one to two and thus place an additional burden on
applications that do not require the additional synchronization. The
second call, however, could be avoided by the additional complication of a
start-up state attribute.
An extra state would be introduced: "created but not
started". This would require the standard to specify the behavior of
the thread operations when the target has not yet started executing.
For those applications that require such behavior, it is
possible to simulate the two separate steps with the facilities that are
currently provided. The start_routine() can synchronize by waiting
on a condition variable that is signaled by the start operation.
An Ada implementor can choose to create the thread at either of two points in
the Ada program: when the task object is created, or when the task is
activated (generally at a "begin"). If the first approach is
adopted, the start_routine() needs to wait on a condition variable to
receive the order to begin "activation". The second approach
requires no such condition variable or extra synchronization. In either
approach, a separate Ada task control block would need to be created when the
task object is created to hold rendezvous queues, and so on.
An extension of the preceding model would be to allow the state of the thread to
be modified between the create and start. This would allow the thread
attributes object to be eliminated. This has been rejected because:
All state in the thread attributes object has to be able to
be set for the thread. This would require the definition of functions to
modify thread attributes. There would be no reduction in the number of
function calls required to set up the thread. In fact, for an application
that creates all threads using identical attributes, the number of
function calls required to set up the threads would be dramatically
increased. Use of a thread attributes object permits the application to
make one set of attribute setting function calls. Otherwise, the set of
attribute setting function calls needs to be made for each thread
Depending on the implementation architecture, functions to
set thread state would require kernel calls, or for other implementation
reasons would not be able to be implemented as macros, thereby increasing
the cost of thread creation.
The ability for applications to segregate threads by class
would be lost.
Another suggested alternative uses a model similar to that for process creation,
such as "thread fork". The fork semantics would provide more
flexibility and the "create" function can be implemented simply by
doing a thread fork followed immediately by a call to the desired "start
routine" for the thread. This alternative has these problems:
For many implementations, the entire stack of the calling
thread would need to be duplicated, since in many architectures there is
no way to determine the size of the calling frame.
Efficiency is reduced since at least some part of the stack
has to be copied, even though in most cases the thread never needs the
copied context, since it merely calls the desired start routine.
Portions of this text are reprinted and reproduced in electronic form from IEEE
Std 1003.1, 2003 Edition, Standard for Information Technology -- Portable
Operating System Interface (POSIX), The Open Group Base Specifications Issue
6, Copyright (C) 2001-2003 by the Institute of Electrical and Electronics
Engineers, Inc and The Open Group. In the event of any discrepancy between
this version and the original IEEE and The Open Group Standard, the original
IEEE and The Open Group Standard is the referee document. The original
Standard can be obtained online at http://www.opengroup.org/unix/online.html