SYNOPSIS

#include <tcl.h>

---

void
Tcl_ConditionNotify(condPtr)

---

void
Tcl_ConditionWait(condPtr, mutexPtr, timePtr)

---

void
Tcl_ConditionFinalize(condPtr)

---

Void *
Tcl_GetThreadData(keyPtr, size)

---

void
Tcl_MutexLock(mutexPtr)

---

void
Tcl_MutexUnlock(mutexPtr)

---

void
Tcl_MutexFinalize(mutexPtr)

---

int
Tcl_CreateThread(idPtr, threadProc, clientData, stackSize, flags)

---

int
Tcl_JoinThread(id, result)

ARGUMENTS

Tcl_Condition    *condPtr    (in)
A condition variable, which must be associated with a mutex lock.

Tcl_Mutex    *mutexPtr    (in)
A mutex lock.

Tcl_Time    *timePtr    (in)
A time limit on the condition wait. NULL to wait forever. Note that a polling value of 0 seconds doesn't make much sense.

Tcl_ThreadDataKey    *keyPtr    (in)
This identifies a block of thread local storage. The key should be static and process-wide, yet each thread will end up associating a different block of storage with this key.

int    *size    (in)
The size of the thread local storage block. This amount of data is allocated and initialized to zero the first time each thread calls Tcl_GetThreadData.

Tcl_ThreadId    *idPtr    (out)
The referred storage will contain the id of the newly created thread as returned by the operating system.

Tcl_ThreadId    id    (in)
Id of the thread waited upon.

Tcl_ThreadCreateProc    threadProc    (in)
This procedure will act as the main() of the newly created thread. The specified clientData will be its sole argument.

ClientData    clientData    (in)
Arbitrary information. Passed as sole argument to the threadProc.

int    stackSize    (in)
The size of the stack given to the new thread.

int    flags    (in)
Bitmask containing flags allowing the caller to modify behaviour of the new thread.

int    *result    (out)
The referred storage is used to place the exit code of the thread waited upon into it.

INTRODUCTION

Beginning with the 8.1 release, the Tcl core is thread safe, which allows you to incorporate Tcl into multithreaded applications without customizing the Tcl core. To enable Tcl multithreading support, you must include the --enable-threads option to configure when you configure and compile your Tcl core.

An important constraint of the Tcl threads implementation is that only the thread that created a Tcl interpreter can use that interpreter. In other words, multiple threads can not access the same Tcl interpreter. (However, as was the case in previous releases, a single thread can safely create and use multiple interpreters.)

Tcl does provide Tcl_CreateThread for creating threads. The caller can determine the size of the stack given to the new thread and modify the behaviour through the supplied flags. The value TCL_THREAD_STACK_DEFAULT for the stackSize indicates that the default size as specified by the operating system is to be used for the new thread. As for the flags, currently are only the values TCL_THREAD_NOFLAGS and TCL_THREAD_JOINABLE defined. The first of them invokes the default behaviour with no specialties. Using the second value marks the new thread as joinable. This means that another thread can wait for the such marked thread to exit and join it.

Restrictions: On some unix systems the pthread-library does not contain the functionality to specify the stacksize of a thread. The specified value for the stacksize is ignored on these systems. Both Windows and Macintosh currently do not support joinable threads. This flag value is therefore ignored on these platforms.

Tcl does provide Tcl_ExitThread and Tcl_FinalizeThread for terminating threads and invoking optional per-thread exit handlers. See the Tcl_Exit page for more information on these procedures.

The Tcl_JoinThread function is provided to allow threads to wait upon the exit of another thread, which must have been marked as joinable through usage of the TCL_THREAD_JOINABLE-flag during its creation via Tcl_CreateThread.

Trying to wait for the exit of a non-joinable thread or a thread which is already waited upon will result in an error. Waiting for a joinable thread which already exited is possible, the system will retain the necessary information until after the call to Tcl_JoinThread. This means that not calling Tcl_JoinThread for a joinable thread will cause a memory leak.

Tcl provides Tcl_ThreadQueueEvent and Tcl_ThreadAlert for handling event queueing in multithreaded applications. See the Notifier manual page for more information on these procedures.

In this release, the Tcl language itself provides no support for creating multithreaded scripts (for example, scripts that could spawn a Tcl interpreter in a separate thread). If you need to add this feature at this time, see the tclThreadTest.c file in the Tcl source distribution for an experimental implementation or use the Tcl "Threading Extension" package implementing thread creation and management commands at the script level.

DESCRIPTION

A mutex is a lock that is used to serialize all threads through a piece of code by calling Tcl_MutexLock and Tcl_MutexUnlock. If one thread holds a mutex, any other thread calling Tcl_MutexLock will block until Tcl_MutexUnlock is called. A mutex can be destroyed after its use by calling Tcl_MutexFinalize. The result of locking a mutex twice from the same thread is undefined. On some platforms it will result in a deadlock. The Tcl_MutexLock, Tcl_MutexUnlock and Tcl_MutexFinalize procedures are defined as empty macros if not compiling with threads enabled. For declaration of mutexes the TCL_DECLARE_MUTEX macro should be used. This macro assures correct mutex handling even when the core is compiled without threads enabled.

A condition variable is used as a signaling mechanism: a thread can lock a mutex and then wait on a condition variable with Tcl_ConditionWait. This atomically releases the mutex lock and blocks the waiting thread until another thread calls Tcl_ConditionNotify. The caller of Tcl_ConditionNotify should have the associated mutex held by previously calling Tcl_MutexLock, but this is not enforced. Notifying the condition variable unblocks all threads waiting on the condition variable, but they do not proceed until the mutex is released with Tcl_MutexUnlock. The implementation of Tcl_ConditionWait automatically locks the mutex before returning.

The caller of Tcl_ConditionWait should be prepared for spurious notifications by calling Tcl_ConditionWait within a while loop that tests some invariant.

A condition variable can be destroyed after its use by calling Tcl_ConditionFinalize.

The Tcl_ConditionNotify, Tcl_ConditionWait and Tcl_ConditionFinalize procedures are defined as empty macros if not compiling with threads enabled.

The Tcl_GetThreadData call returns a pointer to a block of thread-private data. Its argument is a key that is shared by all threads and a size for the block of storage. The storage is automatically allocated and initialized to all zeros the first time each thread asks for it. The storage is automatically deallocated by Tcl_FinalizeThread.

INITIALIZATION

All of these synchronization objects are self initializing. They are implemented as opaque pointers that should be NULL upon first use. The mutexes and condition variables are either cleaned up by process exit handlers (if living that long) or explicitly by calls to Tcl_MutexFinalize or Tcl_ConditionFinalize. Thread local storage is reclaimed during Tcl_FinalizeThread.

CREATING THREADS

The API to create threads is not finalized at this time. There are private facilities to create threads that contain a new Tcl interpreter, and to send scripts among threads. Dive into tclThreadTest.c and tclThread.c for examples.

KEYWORDS

thread, mutex, condition variable, thread local storage