/// <summary> Return active status of t.
/// Per-thread active status can only be accessed and
/// modified via synchronized method here in the group class.
///
/// </summary>
protected internal virtual bool GetActive(FJTaskRunner t)
{
lock (this)
{
return(t.active);
}
}
/* ------------ Methods called only by FJTaskRunners ------------- */
/// <summary> Return a task from entry queue, or null if empty.
/// Called only by FJTaskRunner.scan().
///
/// </summary>
protected internal virtual FJTask PollEntryQueue(FJTaskRunner runner)
{
try
{
object o = entryQueue.Poll(0);
if (o is FJTask)
{
return(o as FJTask);
}
else if (o is InvokableCandidateFJTask)
{
InvokableCandidateFJTask i = o as InvokableCandidateFJTask;
return(i.AsInvokableFJTask(runner));
}
else if (o is IRunnable)
{
return(new FJTask.Wrap(o as IRunnable));
}
else if (o == null)
{
return(null);
}
else
{
throw new ArgumentException("unexpected object in queue", o != null ? o.GetType().Name : "null");
}
}
catch (System.Threading.ThreadInterruptedException)
{
// ignore interrupts
ThreadClass.Current.Interrupt();
return(null);
}
}
/// <summary> Set active status of thread t to false, and
/// then wait until: (a) there is a task in the entry
/// queue, or (b) other threads are active, or (c) the current
/// thread is interrupted. Upon return, it
/// is not certain that there will be work available.
/// The thread must itself check.
/// <p>
/// The main underlying reason
/// for these mechanics is that threads do not
/// signal each other when they add elements to their queues.
/// (This would add to task overhead, reduce locality.
/// and increase contention.)
/// So we must rely on a tamed form of polling. However, tasks
/// inserted into the entry queue do result in signals, so
/// tasks can wait on these if all of them are otherwise idle.
/// </p>
/// </summary>
protected internal virtual void CheckActive(FJTaskRunner t, long scans)
{
lock (this)
{
Inactive = t;
try
{
// if nothing available, do a hard wait
if (activeCount_ == 0 && entryQueue.Peek() == null)
{
System.Threading.Monitor.Wait(this);
}
else
{
// If there is possibly some work,
// sleep for a while before rechecking
long msecs = scans / ScansPerSleep;
if (msecs > MaxSleepTime)
{
msecs = MaxSleepTime;
}
int nsecs = (msecs == 0)?1:0; // forces shortest possible sleep
System.Threading.Monitor.Wait(this, TimeSpan.FromMilliseconds(msecs + (nsecs / 1000)));
}
}
catch (System.Threading.ThreadInterruptedException)
{
System.Threading.Monitor.Pulse(this); // avoid lost notifies on interrupts
ThreadClass.Current.Interrupt();
}
}
}
/// <summary> Create all FJTaskRunner threads in this group.
///
/// </summary>
protected internal virtual void InitializeThreads()
{
for (int i = 0; i < threads.Length; ++i)
{
threads[i] = FJTaskRunner.New(this);
}
}
/// <summary>
/// Create a new <see cref="FJTaskRunner"/>, setting its thread
/// as needed
/// </summary>
public static FJTaskRunner New(FJTaskRunnerGroup runnerGroup)
{
FJTaskRunner runner = new FJTaskRunner(runnerGroup);
Thread t = new Thread(new ThreadStart(runner.DoStart));
t.Name = "FJTaskRunner thread #" + t.GetHashCode();
runner.SetThread(t);
return(runner);
}
public FJTask AsInvokableFJTask(FJTaskRunner runner)
{
lock (group)
{
if (invokableFJTask == null)
{
invokableFJTask = new InvokableFJTask(r, group);
}
return(invokableFJTask);
}
}
/// <summary> Prints various snapshot statistics to System.out.
/// <ul>
/// <li> For each FJTaskRunner thread (labeled as T<em>n</em>, for
/// <em>n</em> from zero to group size - 1):</li>
/// <ul>
/// <li> A star "*" is printed if the thread is currently active;
/// that is, not sleeping while waiting for work. Because
/// threads gradually enter sleep modes, an active thread
/// may in fact be about to sleep (or wake up).</li>
/// <li> <em>Q Cap</em> The current capacity of its task queue.</li>
/// <li> <em>Run</em> The total number of tasks that have been run.</li>
/// <li> <em>New</em> The number of these tasks that were
/// taken from either the entry queue or from other
/// thread queues; that is, the number of tasks run
/// that were <em>not</em> forked by the thread itself.</li>
/// <li> <em>Scan</em> The number of times other task</li>
/// queues or the entry queue were polled for tasks.
/// </ul>
/// <li> <em>Execute</em> The total number of tasks entered
/// (but not necessarily yet run) via execute or invoke.</li>
/// <li> <em>Time</em> Time in seconds since construction of this
/// FJTaskRunnerGroup.</li>
/// <li> <em>Rate</em> The total number of tasks processed
/// per second across all threads. This
/// may be useful as a simple throughput indicator
/// if all processed tasks take approximately the
/// same time to run.</li>
/// </ul>
/// <p>
/// Cautions: Some statistics are updated and gathered
/// without synchronization,
/// so may not be accurate. However, reported counts may be considered
/// as lower bounds of actual values.
/// Some values may be zero if classes are compiled
/// with COLLECT_STATS set to false. (FJTaskRunner and FJTaskRunnerGroup
/// classes can be independently compiled with different values of
/// COLLECT_STATS.) Also, the counts are maintained as ints so could
/// overflow in exceptionally long-lived applications.
/// </p>
/// <p>
/// These statistics can be useful when tuning algorithms or diagnosing
/// problems. For example:
/// </p>
/// <ul>
/// <li> High numbers of scans may mean that there is insufficient
/// parallelism to keep threads busy. However, high scan rates
/// are expected if the number
/// of Executes is also high or there is a lot of global
/// synchronization in the application, and the system is not otherwise
/// busy. Threads may scan
/// for work hundreds of times upon startup, shutdown, and
/// global synch points of task sets.</li>
/// <li> Large imbalances in tasks run across different threads might
/// just reflect contention with unrelated threads on a system
/// (possibly including JVM threads such as GC), but may also
/// indicate some systematic bias in how you generate tasks.</li>
/// <li> Large task queue capacities may mean that too many tasks are being
/// generated before they can be run.
/// Capacities are reported rather than current numbers of tasks
/// in queues because they are better indicators of the existence
/// of these kinds of possibly-transient problems.
/// Queue capacities are
/// resized on demand from their initial value of 4096 elements,
/// which is much more than sufficient for the kinds of
/// applications that this framework is intended to best support.</li>
/// </ul>
/// </summary>
public virtual void Stats(TextWriter writer)
{
long time = Utils.CurrentTimeMillis - initTime;
//TODO: WARNING: Narrowing conversions may produce unexpected results in C#. 'ms-help://MS.VSCC.2003/commoner/redir/redirect.htm?keyword="jlca1042"'
double secs = ((double)time) / 1000.0;
long totalRuns = 0;
long totalScans = 0;
long totalSteals = 0;
writer.Write("Thread" + "\tQ Cap" + "\tScans" + "\tNew" + "\tRuns" + "\n");
for (int i = 0; i < threads.Length; ++i)
{
FJTaskRunner t = threads[i];
int truns = t.runs;
totalRuns += truns;
int tscans = t.scans;
totalScans += tscans;
int tsteals = t.steals;
totalSteals += tsteals;
System.String star = (GetActive(t))?"*":" ";
writer.Write("T" + i + star + "\t" + t.deqSize() + "\t" + tscans + "\t" + tsteals + "\t" + truns + "\n");
}
writer.Write("Total" + "\t " + "\t" + totalScans + "\t" + totalSteals + "\t" + totalRuns + "\n");
writer.Write("Execute: " + entries);
writer.Write("\tTime: " + secs);
long rps = 0;
if (secs != 0)
{
//TODO: WARNING: Narrowing conversions may produce unexpected results in C#. 'ms-help://MS.VSCC.2003/commoner/redir/redirect.htm?keyword="jlca1042"'
rps = (long)System.Math.Round((double)(totalRuns) / secs);
}
writer.WriteLine("\tRate: " + rps);
}
/// <summary> Set active status of thread t to true, and notify others
/// that might be waiting for work.
///
/// </summary>
protected internal virtual void SetActive(FJTaskRunner t)
{
lock (this)
{
if (!t.active)
{
t.active = true;
++activeCount_;
if (nstarted < threads.Length)
{
threads[nstarted++].Start();
}
else
{
System.Threading.Monitor.PulseAll(this);
}
}
}
}
/// <summary>
/// Create a new <see cref="FJTaskRunner"/>, setting its thread
/// as needed
/// </summary>
public static FJTaskRunner New(FJTaskRunnerGroup runnerGroup)
{
FJTaskRunner runner = new FJTaskRunner(runnerGroup);
Thread t = new Thread(new ThreadStart(runner.DoStart));
t.Name = "FJTaskRunner thread #" + t.GetHashCode();
runner.SetThread(t);
return runner;
}
/// <summary> Construct and return a FJTask object that, when executed, will
/// invoke task1 and task2, in parallel
///
/// </summary>
public static FJTask NewPar(FJTask task1, FJTask task2, FJTaskRunner fjTaskRunner)
{
return(new Par2(task1, task2));
}
/// <summary> Construct and return a FJTask object that, when executed, will
/// invoke the tasks in the tasks array in parallel using coInvoke
///
/// </summary>
public static FJTask NewPar(FJTask[] tasks, FJTaskRunner runner)
{
return(new Par(tasks));
}
/// <summary> Construct and return a FJTask object that, when executed, will
/// invoke the tasks in the tasks array in array order
/// </summary>
public static FJTask NewSeq(FJTask[] tasks, FJTaskRunner fjTaskRunner)
{
return(new Seq(tasks));
}
/// <summary> Fork all tasks in array, and await their completion.
/// Behaviorally equivalent to:
/// <pre>
/// for (int i = 0; i < tasks.length; ++i) tasks[i].fork();
/// for (int i = 0; i < tasks.length; ++i) tasks[i].join();
/// </pre>
///
/// </summary>
public void CoInvoke(FJTask[] tasks)
{
FJTaskRunner.coInvoke(tasks);
}
/* ------------ Scheduling ------------------- */
/// <summary> Do all but the pop() part of yield or join, by
/// traversing all DEQs in our group looking for a task to
/// steal. If none, it checks the entry queue.
/// <p>
/// Since there are no good, portable alternatives,
/// we rely here on a mixture of Thread.yield and priorities
/// to reduce wasted spinning, even though these are
/// not well defined. We are hoping here that the JVM
/// does something sensible.
/// </p>
/// </summary>
/// <param name="waitingFor">if non-null, the current task being joined
///
/// </param>
protected internal virtual void scan(FJTask waitingFor)
{
FJTask task = null;
// to delay lowering priority until first failure to steal
bool lowered = false;
/*
* Circularly traverse from a random start index.
*
* This differs slightly from cilk version that uses a random index
* for each attempted steal.
* Exhaustive scanning might impede analytic tractablity of
* the scheduling policy, but makes it much easier to deal with
* startup and shutdown.
*/
FJTaskRunner[] ts = group_.Array;
int idx = victimRNG.Next(ts.Length);
for (int i = 0; i < ts.Length; ++i)
{
FJTaskRunner t = ts[idx];
if (++idx >= ts.Length)
{
idx = 0; // circularly traverse
}
if (t != null && t != this)
{
if (waitingFor != null && waitingFor.Done)
{
break;
}
else
{
if (CollectStats)
{
++scans;
}
task = t.take();
if (task != null)
{
if (CollectStats)
{
++steals;
}
break;
}
else
{
if (Interrupted)
{
break;
}
else if (!lowered)
{
// if this is first fail, lower priority
lowered = true;
Priority = (ThreadPriority)scanPriority_;
}
else
{
// otherwise we are at low priority; just yield
Thread.Sleep(0);
}
}
}
}
}
if (task == null)
{
if (CollectStats)
{
++scans;
}
task = group_.PollEntryQueue(this);
if (CollectStats)
{
if (task != null)
{
++steals;
}
}
}
if (lowered)
{
Priority = (ThreadPriority)runPriority_;
}
if (task != null && !task.Done)
{
if (CollectStats)
{
++runs;
}
task.Run();
task.SetDone();
}
}
/// <summary> Set active status of thread t to true, and notify others
/// that might be waiting for work.
///
/// </summary>
protected internal virtual void SetActive(FJTaskRunner t)
{
lock (this)
{
if (!t.active)
{
t.active = true;
++activeCount_;
if (nstarted < threads.Length)
threads[nstarted++].Start();
else
System.Threading.Monitor.PulseAll(this);
}
}
}
/// <summary> Return active status of t.
/// Per-thread active status can only be accessed and
/// modified via synchronized method here in the group class.
///
/// </summary>
protected internal virtual bool GetActive(FJTaskRunner t)
{
lock (this)
{
return t.active;
}
}
/// <summary> Construct and return a FJTask object that, when executed, will
/// invoke task1 and task2, in order
///
/// </summary>
public static FJTask NewSeq(FJTask task1, FJTask task2, FJTaskRunner fjTaskRunner)
{
return new Seq2(task1, task2);
}
/// <summary> Construct and return a FJTask object that, when executed, will
/// invoke the tasks in the tasks array in parallel using coInvoke
///
/// </summary>
public static FJTask NewPar(FJTask[] tasks, FJTaskRunner runner)
{
return new Par(tasks);
}
/// <summary> Same as scan, but called when current thread is idling.
/// It repeatedly scans other threads for tasks,
/// sleeping while none are available.
/// <p>
/// This differs from scan mainly in that
/// since there is no reason to return to recheck any
/// condition, we iterate until a task is found, backing
/// off via sleeps if necessary.
/// </p>
///
/// </summary>
protected internal virtual void scanWhileIdling()
{
FJTask task = null;
bool lowered = false;
long iters = 0;
FJTaskRunner[] ts = group_.Array;
int idx = victimRNG.Next(ts.Length);
do
{
for (int i = 0; i < ts.Length; ++i)
{
FJTaskRunner t = ts[idx];
if (++idx >= ts.Length)
{
idx = 0; // circularly traverse
}
if (t != null && t != this)
{
if (CollectStats)
{
++scans;
}
task = t.take();
if (task != null)
{
if (CollectStats)
{
++steals;
}
if (lowered)
{
Priority = (ThreadPriority)runPriority_;
}
group_.SetActive(this);
break;
}
}
}
if (task == null)
{
if (Interrupted)
{
return;
}
if (CollectStats)
{
++scans;
}
task = group_.PollEntryQueue(this);
if (task != null)
{
if (CollectStats)
{
++steals;
}
if (lowered)
{
Priority = (ThreadPriority)runPriority_;
}
group_.SetActive(this);
}
else
{
++iters;
// Check here for yield vs sleep to avoid entering group synch lock
if (iters >= FJTaskRunnerGroup.ScansPerSleep)
{
group_.CheckActive(this, iters);
if (Interrupted)
{
return;
}
}
else if (!lowered)
{
lowered = true;
Priority = (ThreadPriority)scanPriority_;
}
else
{
Thread.Sleep(0);
}
}
}
}while (task == null);
if (!task.Done)
{
if (CollectStats)
{
++runs;
}
task.Run();
task.SetDone();
}
}
/// <summary> Construct and return a FJTask object that, when executed, will
/// invoke the tasks in the tasks array in array order
/// </summary>
public static FJTask NewSeq(FJTask[] tasks, FJTaskRunner fjTaskRunner)
{
return new Seq(tasks);
}
/// <summary> Arrange for execution of a strictly dependent task.
/// The task that will be executed in
/// procedure-call-like LIFO order if executed by the
/// same worker thread, but is FIFO with respect to other tasks
/// forked by this thread when taken by other worker threads.
/// That is, earlier-forked
/// tasks are preferred to later-forked tasks by other idle workers.
/// <p>
/// Fork() is noticeably
/// faster than start(). However, it may only
/// be used for strictly dependent tasks -- generally, those that
/// could logically be issued as straight method calls without
/// changing the logic of the program.
/// The method is optimized for use in parallel fork/join designs
/// in which the thread that issues one or more forks
/// cannot continue until at least some of the forked
/// threads terminate and are joined.
/// </p>
/// </summary>
public virtual void Fork()
{
FJTaskRunner.push(this);
}
/// <summary> Set active status of thread t to false, and
/// then wait until: (a) there is a task in the entry
/// queue, or (b) other threads are active, or (c) the current
/// thread is interrupted. Upon return, it
/// is not certain that there will be work available.
/// The thread must itself check.
/// <p>
/// The main underlying reason
/// for these mechanics is that threads do not
/// signal each other when they add elements to their queues.
/// (This would add to task overhead, reduce locality.
/// and increase contention.)
/// So we must rely on a tamed form of polling. However, tasks
/// inserted into the entry queue do result in signals, so
/// tasks can wait on these if all of them are otherwise idle.
/// </p>
/// </summary>
protected internal virtual void CheckActive(FJTaskRunner t, long scans)
{
lock (this)
{
Inactive = t;
try
{
// if nothing available, do a hard wait
if (activeCount_ == 0 && entryQueue.Peek() == null)
{
System.Threading.Monitor.Wait(this);
}
else
{
// If there is possibly some work,
// sleep for a while before rechecking
long msecs = scans / ScansPerSleep;
if (msecs > MaxSleepTime)
msecs = MaxSleepTime;
int nsecs = (msecs == 0)?1:0; // forces shortest possible sleep
System.Threading.Monitor.Wait(this, TimeSpan.FromMilliseconds(msecs + (nsecs / 1000)));
}
}
catch (System.Threading.ThreadInterruptedException)
{
System.Threading.Monitor.Pulse(this); // avoid lost notifies on interrupts
ThreadClass.Current.Interrupt();
}
}
}
/// <summary> Allow the current underlying FJTaskRunner thread to process other tasks.
/// <p>
/// Spinloops based on yield() are well behaved so long
/// as the event or condition being waited for is produced via another
/// FJTask. Additionally, you must never hold a lock
/// while performing a yield or join. (This is because
/// multiple FJTasks can be run by the same Thread during
/// a yield. Since java locks are held per-thread, the lock would not
/// maintain the conceptual exclusion you have in mind.)
/// </p>
/// <p>
/// Otherwise, spinloops using
/// yield are the main construction of choice when a task must wait
/// for a condition that it is sure will eventually occur because it
/// is being produced by some other FJTask. The most common
/// such condition is built-in: join() repeatedly yields until a task
/// has terminated after producing some needed results. You can also
/// use yield to wait for callbacks from other FJTasks, to wait for
/// status flags to be set, and so on. However, in all these cases,
/// you should be confident that the condition being waited for will
/// occur, essentially always because it is produced by
/// a FJTask generated by the current task, or one of its subtasks.
/// </p>
/// </summary>
public void Yield()
{
FJTaskRunner.taskYield();
}
/* ------------ Methods called only by FJTaskRunners ------------- */
/// <summary> Return a task from entry queue, or null if empty.
/// Called only by FJTaskRunner.scan().
///
/// </summary>
protected internal virtual FJTask PollEntryQueue(FJTaskRunner runner)
{
try
{
object o = entryQueue.Poll(0);
if (o is FJTask)
{
return o as FJTask;
}
else if (o is InvokableCandidateFJTask)
{
InvokableCandidateFJTask i = o as InvokableCandidateFJTask;
return i.AsInvokableFJTask(runner);
}
else if (o is IRunnable)
{
return new FJTask.Wrap(o as IRunnable);
}
else if (o == null)
{
return null;
}
else
{
throw new ArgumentException("unexpected object in queue", o != null ? o.GetType().Name : "null");
}
}
catch (System.Threading.ThreadInterruptedException)
{
// ignore interrupts
ThreadClass.Current.Interrupt();
return null;
}
}
/// <summary> Yield until this task isDone.
/// Equivalent to <code>while(!isDone()) yield(); </code>
/// </summary>
public virtual void Join()
{
FJTaskRunner.taskJoin(this);
}
public FJTask AsInvokableFJTask(FJTaskRunner runner)
{
lock (group)
{
if (invokableFJTask == null)
{
invokableFJTask = new InvokableFJTask(r, group);
}
return invokableFJTask;
}
}
/// <summary> Fork both tasks and then wait for their completion. It behaves as:
/// <pre>
/// task1.fork(); task2.fork(); task2.join(); task1.join();
/// </pre>
/// As a simple classic example, here is
/// a class that computes the Fibonacci function:
/// <pre>
/// public class Fib extends FJTask {
///
/// // Computes fibonacci(n) = fibonacci(n-1) + fibonacci(n-2); for n> 1
/// // fibonacci(0) = 0;
/// // fibonacci(1) = 1.
///
/// // Value to compute fibonacci function for.
/// // It is replaced with the answer when computed.
/// private volatile int number;
///
/// public Fib(int n) { number = n; }
///
/// public int getAnswer() {
/// if (!isDone()) throw new Error("Not yet computed");
/// return number;
/// }
///
/// public void run() {
/// int n = number;
/// if (n > 1) {
/// Fib f1 = new Fib(n - 1);
/// Fib f2 = new Fib(n - 2);
///
/// coInvoke(f1, f2); // run these in parallel
///
/// // we know f1 and f2 are computed, so just directly access numbers
/// number = f1.number + f2.number;
/// }
/// }
///
/// public static void main(String[] args) { // sample driver
/// try {
/// int groupSize = 2; // 2 worker threads
/// int num = 35; // compute fib(35)
/// FJTaskRunnerGroup group = new FJTaskRunnerGroup(groupSize);
/// Fib f = new Fib(num);
/// group.invoke(f);
/// int result = f.getAnswer();
/// System.out.println(" Answer: " + result);
/// }
/// catch (InterruptedException ex) {
/// System.out.println("Interrupted");
/// }
/// }
/// }
/// </pre>
///
/// </summary>
public void CoInvoke(FJTask task1, FJTask task2)
{
FJTaskRunner.coInvoke(task1, task2);
}
