|On all platforms with threads support, Agar can be compiled with support for multithreading. Agar API calls, unless otherwise documented, then become free-threaded (safe to use from different threads without need for application-level synchronization).|
The Agar API documentation follows the convention that all functions are
free-threaded, unless mentioned otherwise.
Under some circumstances, application-level synchronization is required. The AG_Object(3) simplifies this task by providing a per-object lock (which is implicitely acquired in some contexts, such as event handler execution). For instance, the following code accesses a VFS in an unsafe manner:
The following code should be used instead:
|When compiled with threads support, Agar provides a portable, minimal interface to the operating system's native threads interface. These functions follow Agar's standard error-handling style (see AG_Intro(3)).|
Mutexes (MUTual EXclusion devices) are commonly used to protect shared
data structure against concurrent modifications.
The AG_MutexInit() function initializes a mutex structure. AG_MutexInitRecursive() initializes a recursive mutex (a mutex with a reference count), which allows nested AG_MutexLock() calls.
AG_MutexDestroy() frees all resources allocated for a mutex.
AG_MutexLock() and AG_MutexUnlock() respectively acquire and release a mutex.
AG_MutexTryLock() tries to acquire a mutex without blocking and immediately returns 0 on success. On failure, the function returns -1, but does not set any error message (so AG_GetError(3) should not be used).
AG_CondInit() initializes a condition variable structure.
AG_CondDestroy() releases resources allocated for a condition variable.
AG_CondBroadcast() unblock all threads which are currently blocked waiting on cv. AG_CondSignal() unblocks at least one thread currently blocked waiting on cv.
AG_CondWait() blocks the calling thread until cv is signaled. The AG_CondTimedWait() variant will not block for more than the specified amount of time.
All of these functions will raise a fatal condition if an error is encountered.
AG_ThreadCreate() creates a new thread executing fn. The optional argument arg is passed to fn.
The AG_ThreadCancel() routine requests that the specified thread be cancelled. If the given thread is invalid, a fatal error is raised.
The AG_ThreadJoin() function suspends the execution of the current thread until th terminates. When it does, the value passed to AG_ThreadExit() is made available in exitVal.
AG_ThreadExit() terminates the current thread. exitVal is an optional user pointer.
AG_ThreadKill() sends a signal to the specified thread.
AG_ThreadSelf() returns the identifier of the current (caller's) thread. AG_ThreadEqual() returns 1 if the identifiers a and b both refer to the same thread, or 0 if they differ.
AG_ThreadKeyCreate() initializes a key (i.e., a handle) to a thread-specific value. The handle itself is accessible to all threads. The thread-specific value (i.e., the value specified by AG_ThreadKeySet(), and which defaults to NULL) will persist only for the life of the thread. If an optional destructor is given, that function will be called (with the thread-specific value as its argument), when the thread exists.
The AG_ThreadKeyDelete() function releases resources allocated for a key.
AG_ThreadKeyGet() returns the thread-specific value associated with key.
AG_ThreadKeySet() sets a thread-specific value with key.
|The AG_Threads interface first appeared in Agar 1.0|