Class FJTaskRunner
- All Implemented Interfaces:
Runnable
Each FJTaskRunner keeps FJTasks in a double-ended queue (DEQ). Double-ended queues support stack-based operations push and pop, as well as queue-based operations put and take. Normally, threads run their own tasks. But they may also steal tasks from each others DEQs.
The algorithms are minor variants of those used in Cilk and Hood, and to a lesser extent Filaments, but are adapted to work in Java.
The two most important capabilities are:
- Fork a FJTask:
Push task onto DEQ
- Get a task to run (for example within taskYield)
If DEQ is not empty, Pop a task and run it. Else if any other DEQ is not empty, Take ("steal") a task from it and run it. Else if the entry queue for our group is not empty, Take a task from it and run it. Else if current thread is otherwise idling If all threads are idling Wait for a task to be put on group entry queue Else Yield or Sleep for a while, and then retry
Implementations of the underlying representations and operations are geared for use on JVMs operating on multiple CPUs (although they should of course work fine on single CPUs as well).
A possible snapshot of a FJTaskRunner's DEQ is:
0 1 2 3 4 5 6 ... +-----+-----+-----+-----+-----+-----+-----+-- | | t | t | t | t | | | ... deq array +-----+-----+-----+-----+-----+-----+-----+-- ^ ^ base top (incremented (incremented on take, on push decremented decremented on put) on pop)
FJTasks are held in elements of the DEQ.
They are maintained in a bounded array that
works similarly to a circular bounded buffer. To ensure
visibility of stolen FJTasks across threads, the array elements
must be volatile
.
Using volatile rather than synchronizing suffices here since
each task accessed by a thread is either one that it
created or one that has never seen before. Thus we cannot
encounter any staleness problems executing run methods,
although FJTask programmers must be still sure to either synch or use
volatile for shared data within their run methods.
However, since there is no way to declare an array of volatiles in Java, the DEQ elements actually hold VolatileTaskRef objects, each of which in turn holds a volatile reference to a FJTask. Even with the double-indirection overhead of volatile refs, using an array for the DEQ works out better than linking them since fewer shared memory locations need to be touched or modified by the threads while using the DEQ. Further, the double indirection may alleviate cache-line sharing effects (which cannot otherwise be directly dealt with in Java).
The indices for the base
and top
of the DEQ
are declared as volatile. The main contention point with
multiple FJTaskRunner threads occurs when one thread is trying
to pop its own stack while another is trying to steal from it.
This is handled via a specialization of Dekker's algorithm,
in which the popping thread pre-decrements top
,
and then checks it against base
.
To be conservative in the face of JVMs that only partially
honor the specification for volatile, the pop proceeds
without synchronization only if there are apparently enough
items for both a simultaneous pop and take to succeed.
It otherwise enters a
synchronized lock to check if the DEQ is actually empty,
if so failing. The stealing thread
does almost the opposite, but is set up to be less likely
to win in cases of contention: Steals always run under synchronized
locks in order to avoid conflicts with other ongoing steals.
They pre-increment base
, and then check against
top
. They back out (resetting the base index
and failing to steal) if the
DEQ is empty or is about to become empty by an ongoing pop.
A push operation can normally run concurrently with a steal. A push enters a synch lock only if the DEQ appears full so must either be resized or have indices adjusted due to wrap-around of the bounded DEQ. The put operation always requires synchronization.
When a FJTaskRunner thread has no tasks of its own to run, it tries to be a good citizen. Threads run at lower priority while scanning for work.
If the task is currently waiting via yield, the thread alternates scans (starting at a randomly chosen victim) with Thread.yields. This is well-behaved so long as the JVM handles Thread.yield in a sensible fashion. (It need not. Thread.yield is so underspecified that it is legal for a JVM to treat it as a no-op.) This also keeps things well-behaved even if we are running on a uniprocessor JVM using a simple cooperative threading model.
If a thread needing work is is otherwise idle (which occurs only in the main runloop), and there are no available tasks to steal or poll, it instead enters into a sleep-based (actually timed wait(msec)) phase in which it progressively sleeps for longer durations (up to a maximum of FJTaskRunnerGroup.MAX_SLEEP_TIME, currently 100ms) between scans. If all threads in the group are idling, they further progress to a hard wait phase, suspending until a new task is entered into the FJTaskRunnerGroup entry queue. A sleeping FJTaskRunner thread may be awakened by a new task being put into the group entry queue or by another FJTaskRunner becoming active, but not merely by some DEQ becoming non-empty. Thus the MAX_SLEEP_TIME provides a bound for sleep durations in cases where all but one worker thread start sleeping even though there will eventually be work produced by a thread that is taking a long time to place tasks in DEQ. These sleep mechanics are handled in the FJTaskRunnerGroup class.
Composite operations such as taskJoin include heavy manual inlining of the most time-critical operations (mainly FJTask.invoke). This opens up a few opportunities for further hand-optimizations. Until Java compilers get a lot smarter, these tweaks improve performance significantly enough for task-intensive programs to be worth the poorer maintainability and code duplication.
Because they are so fragile and performance-sensitive, nearly all methods are declared as final. However, nearly all fields and methods are also declared as protected, so it is possible, with much care, to extend functionality in subclasses. (Normally you would also need to subclass FJTaskRunnerGroup.)
None of the normal java.lang.Thread class methods should ever be called on FJTaskRunners. For this reason, it might have been nicer to declare FJTaskRunner as a Runnable to run within a Thread. However, this would have complicated many minor logistics. And since no FJTaskRunner methods should normally be called from outside the FJTask and FJTaskRunnerGroup classes either, this decision doesn't impact usage.
You might think that layering this kind of framework on top of Java threads, which are already several levels removed from raw CPU scheduling on most systems, would lead to very poor performance. But on the platforms tested, the performance is quite good.
- See Also:
-
Nested Class Summary
Nested ClassesModifier and TypeClassDescriptionprotected static final class
An object holding a single volatile reference to a FJTask.Nested classes/interfaces inherited from class java.lang.Thread
Thread.State, Thread.UncaughtExceptionHandler
-
Field Summary
FieldsModifier and TypeFieldDescriptionprotected boolean
Record whether current thread may be processing a task (i.e., has been started and is not in an idle wait).protected final Object
An extra object to synchronize on in order to achieve a memory barrier.protected int
Current base of DEQ.protected FJTaskRunner.VolatileTaskRef[]
The DEQ array.protected final FJTaskRunnerGroup
The group of which this FJTaskRunner is a memberprotected static final int
FJTasks are held in an array-based DEQ with INITIAL_CAPACITY elements.protected static final int
The maximum supported DEQ capacity.protected int
Priority to use while running tasksprotected int
Total number of tasks runprotected int
Priority to use while scanning for workprotected int
Total number of queues scanned for workprotected int
Total number of tasks obtained via scanprotected int
Current top of DEQ.protected final Random
Random starting point generator for scan()Fields inherited from class java.lang.Thread
MAX_PRIORITY, MIN_PRIORITY, NORM_PRIORITY
-
Constructor Summary
ConstructorsModifierConstructorDescriptionprotected
Constructor called only during FJTaskRunnerGroup initialization -
Method Summary
Modifier and TypeMethodDescriptionprotected void
Adjust top and base, and grow DEQ if necessary.protected final void
Array-based version of coInvokeprotected final void
A specialized expansion ofw.fork(); invoke(v); w.join();
protected final FJTask
confirmPop
(int provisionalTop) Check under synch lock if DEQ is really empty when doing pop.protected FJTask
confirmTake
(int oldBase) double-check a potential takeprotected int
deqSize()
Current size of the task DEQprotected final FJTaskRunnerGroup
getGroup()
Return the FJTaskRunnerGroup of which this thread is a memberprotected final FJTask
pop()
Return a popped task, or null if DEQ is empty.protected final void
Push a task onto DEQ.protected final void
Enqueue task at base of DEQ.void
run()
Main runloopprotected void
Do all but the pop() part of yield or join, by traversing all DEQs in our group looking for a task to steal.protected void
Same as scan, but called when current thread is idling.protected void
setRunPriority
(int pri) Set the priority to use while running tasks.protected void
setScanPriority
(int pri) Set the priority to use while scanning.protected void
slowCoInvoke
(FJTask[] tasks) Backup to handle atypical or noninlinable cases of coInvokeprotected void
slowCoInvoke
(FJTask w, FJTask v) Backup to handle noninlinable cases of coInvokeprotected void
Handle slow case for pushprotected final FJTask
take()
Take a task from the base of the DEQ.protected final void
Process tasks until w is done.protected final void
Execute a task in this thread.Methods inherited from class java.lang.Thread
activeCount, checkAccess, clone, countStackFrames, currentThread, dumpStack, enumerate, getAllStackTraces, getContextClassLoader, getDefaultUncaughtExceptionHandler, getId, getName, getPriority, getStackTrace, getState, getThreadGroup, getUncaughtExceptionHandler, holdsLock, interrupt, interrupted, isAlive, isDaemon, isInterrupted, join, join, join, onSpinWait, resume, setContextClassLoader, setDaemon, setDefaultUncaughtExceptionHandler, setName, setPriority, setUncaughtExceptionHandler, sleep, sleep, start, stop, suspend, toString, yield
-
Field Details
-
group
The group of which this FJTaskRunner is a member -
INITIAL_CAPACITY
protected static final int INITIAL_CAPACITYFJTasks are held in an array-based DEQ with INITIAL_CAPACITY elements. The DEQ is grown if necessary, but default value is normally much more than sufficient unless there are user programming errors or questionable operations generating large numbers of Tasks without running them. Capacities must be a power of two.- See Also:
-
MAX_CAPACITY
protected static final int MAX_CAPACITYThe maximum supported DEQ capacity. When exceeded, FJTaskRunner operations throw Errors- See Also:
-
deq
The DEQ array. -
top
protected volatile int topCurrent top of DEQ. Generally acts just like a stack pointer in an array-based stack, except that it circularly wraps around the array, as in an array-based queue. The value is NOT always kept within0 ... deq.length
though. The current top element is always attop & (deq.length-1)
. To avoid integer overflow, top is reset down within bounds whenever it is noticed to be out out bounds; at worst when it is at2 * deq.length
. -
base
protected volatile int baseCurrent base of DEQ. Acts like a take-pointer in an array-based bounded queue. Same bounds and usage as top. -
barrier
An extra object to synchronize on in order to achieve a memory barrier. -
active
protected boolean activeRecord whether current thread may be processing a task (i.e., has been started and is not in an idle wait). Accessed, under synch, ONLY by FJTaskRunnerGroup, but the field is stored here for simplicity. -
victimRNG
Random starting point generator for scan() -
scanPriority
protected int scanPriorityPriority to use while scanning for work -
runPriority
protected int runPriorityPriority to use while running tasks -
runs
protected int runsTotal number of tasks run -
scans
protected int scansTotal number of queues scanned for work -
steals
protected int stealsTotal number of tasks obtained via scan
-
-
Constructor Details
-
FJTaskRunner
Constructor called only during FJTaskRunnerGroup initialization
-
-
Method Details
-
getGroup
Return the FJTaskRunnerGroup of which this thread is a member -
deqSize
protected int deqSize()Current size of the task DEQ -
setScanPriority
protected void setScanPriority(int pri) Set the priority to use while scanning. We do not bother synchronizing access, since by the time the value is needed, both this FJTaskRunner and its FJTaskRunnerGroup will necessarily have performed enough synchronization to avoid staleness problems of any consequence. -
setRunPriority
protected void setRunPriority(int pri) Set the priority to use while running tasks. Same usage and rationale as setScanPriority. -
push
Push a task onto DEQ. Called ONLY by current thread. -
slowPush
Handle slow case for push -
put
Enqueue task at base of DEQ. Called ONLY by current thread. This method is currently not called from class FJTask. It could be used as a faster way to do FJTask.start, but most users would find the semantics too confusing and unpredictable. -
pop
Return a popped task, or null if DEQ is empty. Called ONLY by current thread.This is not usually called directly but is instead inlined in callers. This version differs from the cilk algorithm in that pop does not fully back down and retry in the case of potential conflict with take. It simply rechecks under synch lock. This gives a preference for threads to run their own tasks, which seems to reduce flailing a bit when there are few tasks to run.
-
confirmPop
Check under synch lock if DEQ is really empty when doing pop. Return task if not empty, else null. -
take
Take a task from the base of the DEQ. Always called by other threads via scan() -
confirmTake
double-check a potential take -
checkOverflow
protected void checkOverflow()Adjust top and base, and grow DEQ if necessary. Called only while DEQ synch lock being held. We don't expect this to be called very often. In most programs using FJTasks, it is never called. -
scan
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.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.
- Parameters:
waitingFor
- if non-null, the current task being joined
-
scanWhileIdling
protected void scanWhileIdling()Same as scan, but called when current thread is idling. It repeatedly scans other threads for tasks, sleeping while none are available.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.
-
run
public void run()Main runloop -
taskYield
protected final void taskYield()Execute a task in this thread. Generally called when current task cannot otherwise continue. -
taskJoin
Process tasks until w is done. Equivalent towhile(!w.isDone()) taskYield();
-
coInvoke
A specialized expansion ofw.fork(); invoke(v); w.join();
-
slowCoInvoke
Backup to handle noninlinable cases of coInvoke -
coInvoke
Array-based version of coInvoke -
slowCoInvoke
Backup to handle atypical or noninlinable cases of coInvoke
-