Class AsynchronousSteppable
- All Implemented Interfaces:
Serializable
,Stoppable
Like all multithreaded stuff, AsynchronousSteppables are inherently dangerous and not to be trifled with: they access data at the same time as other threads, and so must deal with locking. In general if you lock on the Schedule, you are guaranteed atomic access to the underlying simulation model. You'll need to do this for even basic things such as accessing the random number generator. Locking on the Schedule is fairly course-grained, however: the simulation model obtains a lock on the Schedule for the whole duration of a Schedule's step. Instead you might create your own lock shared between the AsynchronousSteppable and the main thread which allows access to some piece of data you both need in a more fine-grained fashion. In this case, make certain that the GUI isn't trying to read that data (to display it, say), or that the GUI obtains a lock when it needs to as well. If you have no idea what we're talking about: don't use an AsynchronousSteppable.
When an AsynchronousSteppable is stepped, it fires off a thread which performs the asynchronous task. This task could be an infinite loop (or otherwise very long process) or it could be a short one-shot thing which runs and ends. Infinite loops can be paused and resumed (for checkpointing) and they can be stopped entirely.
AsynchronousSteppables automatically register themselves to be stopped at the end of the simulation (and when stopped, they unregister themselves). They're also paused and upaused for checkpoints. But if the task is an infinite loop, it's possible you may wish to stop the loop before the simulation ends, perhaps at an agreed-upon point in the schedule. The easiest way to do this is to post a steppable on the Schedule which stops the AsynchronousSteppable, like this:
final AsynchronousSteppable s = ... Steppable stopper = new Steppable() { public void step(SimState state) { s.stop(); } } schedule.scheduleOnce(s....); schedule.scheduleOnce(stopper....);
If the task is a SHORT, one-shot process and the user can reasonably wait for the task to complete after he has presed the 'stop' button, then run(false) should perform the asynchronous task, run(true) should be set to do nothing, and halt(true) and halt(false) should both do nothing. Here's some code to show how to form such a beast.
AsynchronousSteppable s = new AsynchronousSteppable() { protected void run(boolean resuming, boolean restoringFromCheckpoint) { if (!resuming) { // do your stuff here } } };
If the task is an infinite loop or otherwise long process, it needs to be pausable, resumable, and haltable. In this case, run(false) should perform the asynchronous task, halt(false) and halt(true) should both cause the thread to die or trigger events which will soon lead to thread death halt(...) returns, and run(true) should fire up the task loop again after it had been halted with halt(true). The most common situation is where you don't distinguish between your thread being killed temporarily or being killed permanently, nor between starting and resuming. Here's some code for this situation:
AsynchronousSteppable s = new AsynchronousSteppable() { boolean shouldQuit = false; Object[] lock = new Object[0]; // an array is a unique, serializable object protected void run(boolean resuming, boolean restoringFromCheckpoint) { boolean quit = false; while(!quit) { // do your stuff here -- assuming it doesn't block... synchronized(lock) { quit = shouldQuit; shouldQuit = false; } } // we're quitting -- do cleanup here if you need to } protected void halt(boolean pausing) { synchronized(lock) { shouldQuit = val; } } };
It's possible you may need to distinguish between being started or being restarted (but pausing and quitting are considered the same). For example, if you were writing to a log and needed to know whether to open the log fresh or to append to it. You could do something along these lines:
AsynchronousSteppable s = new AsynchronousSteppable() { boolean shouldQuit = false; Object[] lock = new Object[0]; // an array is a unique, serializable object protected void run(boolean resuming, boolean restoringFromCheckpoint) { boolean quit = false; if (!resuming) { // we're starting fresh -- set up here if you have to } else { // we're resuming from a pause -- re-set up here if you have to } while(!quit) { // do your stuff here -- assuming it doesn't block... synchronized(lock) { quit = shouldQuit; shouldQuit = false; } } // we're quitting -- do cleanup here if you need to } protected void halt(boolean pausing) { synchronized(lock) { shouldQuit = val; } } };
Further, it's possible you may need to distinguish between being started or being restarted (but pausing and quitting are considered the same), and additionally you need to know if you're being restarted after the simulation has been restored from a checkpoint. For example, if you were writing to a log and needed to know whether to open the log fresh or to append to it; and if you're restoring from a an earlier checkpoint, you need to seek in the file to append in the right spot. You could do something along these lines:
AsynchronousSteppable s = new AsynchronousSteppable() { boolean shouldQuit = false; Object[] lock = new Object[0]; // an array is a unique, serializable object protected void run(boolean resuming, boolean restoringFromCheckpoint) { boolean quit = false; if (!resuming) { // we're starting fresh -- set up here if you have to } else if (restoringFromCheckpoint) { // do anything you need when restoring from a checkpoint, like seeking in files } else { // we're resuming from a pause -- re-set up here if you have to } while(!quit) { // do your stuff here -- assuming it doesn't block... synchronized(lock) { quit = shouldQuit; shouldQuit = false; } } // we're quitting -- do cleanup here if you need to } protected void halt(boolean pausing) { synchronized(lock) { shouldQuit = val; } } };
Let's say the task also needs to distinguish between being paused and being quit as well. Here's some code for this situation:
AsynchronousSteppable s = new AsynchronousSteppable() { boolean shouldQuit = false; boolean shouldPause = false; Object[] lock = new Object[0]; // an array is a unique, serializable object protected void run(boolean resuming, boolean restoringFromCheckpoint) { boolean quit = false; boolean pause = false; if (!resuming) { // we're starting fresh -- set up here if you have to } else if (restoringFromCheckpoint) { // do anything you need when restoring from a checkpoint, like seeking in files } else { // we're resuming from a pause -- re-set up here if you have to } while(!quit invalid input: '&'invalid input: '&' !pause) { // do your stuff here -- assuming it doesn't block... synchronized(lock) { quit = shouldQuit; shouldQuit = false; pause = shouldPause; shouldPause = false; } } if (quit) { // we're quitting -- do cleanup here if you need to } else // if (pause) { // we're pausing -- do cleanup here if you need to } } protected void halt(boolean pausing) { synchronized(lock) { if (pausing) shouldPause = val; else shouldQuit = val; } } };
- See Also:
-
Field Summary
-
Constructor Summary
-
Method Summary
Modifier and TypeMethodDescriptionprotected void
halt
(boolean pausing) This method should cause the loop created in run(...) to die.final void
pause()
Requests that the AsynchronousSteppable shut down its thread (temporarily) and blocks until this occurs.final void
resume()
Deprecated.use resume(boolean)final void
resume
(boolean restoringFromCheckpoint) Fires up the AsynchronousSteppable after a pause().protected void
run
(boolean resuming) Deprecated.override run(resuming, restoringFromCheckpoint) instead.protected void
run
(boolean resuming, boolean restoringFromCheckpoint) This method should enter the parallel thread's loop.final void
Fires up the AsynchronousSteppable and registers it with the SimState.final void
stop()
Requests that the AsynchronousSteppable shut down its thread, and blocks until this occurs.final Steppable
stopper()
Deprecated.Will be deleted in the future.
-
Field Details
-
state
-
-
Constructor Details
-
AsynchronousSteppable
public AsynchronousSteppable()
-
-
Method Details
-
run
protected void run(boolean resuming, boolean restoringFromCheckpoint) This method should enter the parallel thread's loop. If resuming is true, then you may assume the parallel steppable is being resumed in the middle of a simulation after being paused (likely to checkpoint), as opposed to being started fresh. -
run
protected void run(boolean resuming) Deprecated.override run(resuming, restoringFromCheckpoint) instead. -
halt
protected void halt(boolean pausing) This method should cause the loop created in run(...) to die. If pausing is true, then you may assume the parallel steppable is being paused in the middle of a simulation (likely to checkpoint), as opposed to being entirely stopped due to the end of the simulation. -
step
Fires up the AsynchronousSteppable and registers it with the SimState. If it's already running, nothing happens. -
stop
public final void stop()Requests that the AsynchronousSteppable shut down its thread, and blocks until this occurs. If it's already stopped, nothing happens. -
pause
public final void pause()Requests that the AsynchronousSteppable shut down its thread (temporarily) and blocks until this occurs. If it's already paused or not running, nothing happens. -
resume
public final void resume()Deprecated.use resume(boolean)Fires up the AsynchronousSteppable after a pause(). If it's already unpaused or not running, nothing happens. -
resume
public final void resume(boolean restoringFromCheckpoint) Fires up the AsynchronousSteppable after a pause(). If it's already unpaused or not running, nothing happens. If 'restoringFromCheckpoint' is TRUE then resume(...) is called when MASON is being started up from checkpoint. Otherwise it is false. -
stopper
Deprecated.Will be deleted in the future.Call this method to get a Steppable, which when called, executes top() on the AsynchornousSteppable. You can then schedule this Steppable to occur at some point in the future on a schedule.
-