The AsyncTaskChain is a subset of the AsyncTaskManager. Each chain maintains a separate list of tasks, and will execute them with its own set of threads. Each chain may thereby operate independently of the other chains.
The AsyncTaskChain will spawn a specified number of threads (possibly 0) to serve the tasks. If there are no threads, you must call
poll()from time to time to serve the tasks in the main thread. Normally this is done by calling
Each task will run exactly once each epoch. Beyond that, the tasks’ sort and priority values control the order in which they are run: tasks are run in increasing order by sort value, and within the same sort value, they are run roughly in decreasing order by priority value, with some exceptions for parallelism. Tasks with different sort values are never run in parallel together, but tasks with different priority values might be (if there is more than one thread).
setTickClock(tick_clock: bool) → None¶
Sets the tick_clock flag. When this is true, get_clock()->tick() will be called automatically at each task epoch. This is false by default.
setNumThreads(num_threads: int) → None¶
Changes the number of threads for this task chain. This may require stopping the threads if they are already running.
getNumThreads() → int¶
Returns the number of threads that will be servicing tasks for this chain. Also see
getNumRunningThreads() → int¶
Returns the number of threads that have been created and are actively running. This will return 0 before the threads have been started; it will also return 0 if thread support is not available.
setThreadPriority(priority: ThreadPriority) → None¶
Changes the priority associated with threads that serve this task chain. This may require stopping the threads if they are already running.
getThreadPriority() → ThreadPriority¶
Returns the priority associated with threads that serve this task chain.
- Return type
setFrameBudget(frame_budget: float) → None¶
Sets the maximum amount of time per frame the tasks on this chain are granted for execution. If this is less than zero, there is no limit; if it is >= 0, it represents a maximum amount of time (in seconds) that will be used to execute tasks. If this time is exceeded in any one frame, the task chain will stop executing tasks until the next frame, as defined by the TaskManager’s clock.
getFrameBudget() → float¶
Returns the maximum amount of time per frame the tasks on this chain are granted for execution. See
setFrameSync(frame_sync: bool) → None¶
Sets the frame_sync flag. When this flag is true, this task chain will be forced to sync with the TaskManager’s clock. It will run no faster than one epoch per clock frame.
When this flag is false, the default, the task chain will finish all of its tasks and then immediately start from the first task again, regardless of the clock frame. When it is true, the task chain will finish all of its tasks and then wait for the clock to tick to the next frame before resuming the first task.
This only makes sense for threaded task chains. Non-threaded task chains are automatically synchronous.
setTimeslicePriority(timeslice_priority: bool) → None¶
Sets the timeslice_priority flag. This changes the interpretation of priority, and the number of times per epoch each task will run.
When this flag is true, some tasks might not run in any given epoch. Instead, tasks with priority higher than 1 will be given precedence, in proportion to the amount of time they have already used. This gives higher-priority tasks more runtime than lower-priority tasks. Each task gets the amount of time proportional to its priority value, so a task with priority 100 will get five times as much processing time as a task with priority 20. For these purposes, priority values less than 1 are deemed to be equal to 1.
When this flag is false (the default), all tasks are run exactly once each epoch, round-robin style. Priority is only used to determine which task runs first within tasks of the same sort value.
getTimeslicePriority() → bool¶
Returns the timeslice_priority flag. This changes the interpretation of priority, and the number of times per epoch each task will run. See
stopThreads() → None¶
startThreads() → None¶
Starts any requested threads to service the tasks on the queue. This is normally not necessary, since adding a task will start the threads automatically.
isStarted() → bool¶
Returns true if the thread(s) have been started and are ready to service requests, false otherwise. If this is false, the next call to add() or add_and_do() will automatically start the threads.
hasTask(task: AsyncTask) → bool¶
Returns true if the indicated task has been added to this AsyncTaskChain, false otherwise.
waitForTasks() → None¶
Blocks until the task list is empty.
getNumTasks() → int¶
Returns the number of tasks that are currently active or sleeping within the task chain.
getTasks() → AsyncTaskCollection¶
Returns the set of tasks that are active or sleeping on the task chain, at the time of the call.
- Return type
getActiveTasks() → AsyncTaskCollection¶
Returns the set of tasks that are active (and not sleeping) on the task chain, at the time of the call.
- Return type
getSleepingTasks() → AsyncTaskCollection¶
Returns the set of tasks that are sleeping (and not active) on the task chain, at the time of the call.
- Return type
poll() → None¶
Runs through all the tasks in the task list, once, if the task chain is running in single-threaded mode (no threads available). This method does nothing in threaded mode, so it may safely be called in either case.
Normally, you would not call this function directly; instead, call
AsyncTaskManager.poll(), which polls all of the task chains in sequence.
getNextWakeTime() → float¶
Returns the scheduled time (on the manager’s clock) of the next sleeping task, on any task chain, to awaken. Returns -1 if there are no sleeping tasks.
output(out: ostream) → None¶
write(out: ostream, indent_level: int) → None¶