|Copyright||(c) The University of Glasgow, 1994-2002|
|Portability||non-portable (GHC extensions)|
Basic concurrency stuff.
- data ThreadId = ThreadId ThreadId#
- forkIO :: IO () -> IO ThreadId
- forkIOWithUnmask :: ((forall a. IO a -> IO a) -> IO ()) -> IO ThreadId
- forkOn :: Int -> IO () -> IO ThreadId
- forkOnWithUnmask :: Int -> ((forall a. IO a -> IO a) -> IO ()) -> IO ThreadId
- numCapabilities :: Int
- getNumCapabilities :: IO Int
- setNumCapabilities :: Int -> IO ()
- getNumProcessors :: IO Int
- numSparks :: IO Int
- childHandler :: SomeException -> IO ()
- myThreadId :: IO ThreadId
- killThread :: ThreadId -> IO ()
- throwTo :: Exception e => ThreadId -> e -> IO ()
- par :: a -> b -> b
- pseq :: a -> b -> b
- runSparks :: IO ()
- yield :: IO ()
- labelThread :: ThreadId -> String -> IO ()
- mkWeakThreadId :: ThreadId -> IO (Weak ThreadId)
- data ThreadStatus
- data BlockReason
- threadStatus :: ThreadId -> IO ThreadStatus
- threadCapability :: ThreadId -> IO (Int, Bool)
- setAllocationCounter :: Int64 -> IO ()
- getAllocationCounter :: IO Int64
- enableAllocationLimit :: IO ()
- disableAllocationLimit :: IO ()
- newtype STM a = STM (State# RealWorld -> (#State# RealWorld, a#))
- atomically :: STM a -> IO a
- retry :: STM a
- orElse :: STM a -> STM a -> STM a
- throwSTM :: Exception e => e -> STM a
- catchSTM :: Exception e => STM a -> (e -> STM a) -> STM a
- alwaysSucceeds :: STM a -> STM ()
- always :: STM Bool -> STM ()
- data TVar a = TVar (TVar# RealWorld a)
- newTVar :: a -> STM (TVar a)
- newTVarIO :: a -> IO (TVar a)
- readTVar :: TVar a -> STM a
- readTVarIO :: TVar a -> IO a
- writeTVar :: TVar a -> a -> STM ()
- unsafeIOToSTM :: IO a -> STM a
- withMVar :: MVar a -> (a -> IO b) -> IO b
- modifyMVar_ :: MVar a -> (a -> IO a) -> IO ()
- setUncaughtExceptionHandler :: (SomeException -> IO ()) -> IO ()
- getUncaughtExceptionHandler :: IO (SomeException -> IO ())
- reportError :: SomeException -> IO ()
- reportStackOverflow :: IO ()
- sharedCAF :: a -> (Ptr a -> IO (Ptr a)) -> IO a
ThreadId is an abstract type representing a handle to a thread.
ThreadId is an instance of
Ord instance implements an arbitrary total ordering over
Show instance lets you convert an arbitrary-valued
ThreadId to string form; showing a
ThreadId value is occasionally
useful when debugging or diagnosing the behaviour of a concurrent
Note: in GHC, if you have a
ThreadId, you essentially have
a pointer to the thread itself. This means the thread itself can't be
garbage collected until you drop the
This misfeature will hopefully be corrected at a later date.
|Eq ThreadId #|
|Ord ThreadId #|
|Show ThreadId #|
Forking and suchlike
The new thread will be a lightweight, unbound thread. Foreign calls
made by this thread are not guaranteed to be made by any particular OS
thread; if you need foreign calls to be made by a particular OS
thread, then use
The new thread inherits the masked state of the parent (see
The newly created thread has an exception handler that discards the
ThreadKilled, and passes all other exceptions to the uncaught
forkIO, but the child thread is passed a function that can
be used to unmask asynchronous exceptions. This function is
typically used in the following way
... mask_ $ forkIOWithUnmask $ \unmask -> catch (unmask ...) handler
so that the exception handler in the child thread is established with asynchronous exceptions masked, meanwhile the main body of the child thread is executed in the unmasked state.
Note that the unmask function passed to the child thread should only be used in that thread; the behaviour is undefined if it is invoked in a different thread.
forkIO, but lets you specify on which capability the thread
should run. Unlike a
forkIO thread, a thread created by
will stay on the same capability for its entire lifetime (
threads can migrate between capabilities according to the scheduling
forkOn is useful for overriding the scheduling policy when
you know in advance how best to distribute the threads.
Int argument specifies a capability number (see
getNumCapabilities). Typically capabilities correspond to physical
processors, but the exact behaviour is implementation-dependent. The
value passed to
forkOn is interpreted modulo the total number of
capabilities as returned by
GHC note: the number of capabilities is specified by the
option when the program is started. Capabilities can be fixed to
actual processor cores with
+RTS -qa if the underlying operating
system supports that, although in practice this is usually unnecessary
(and may actually degrade performance in some cases - experimentation
the value passed to the
+RTS -N flag. This is the number of
Haskell threads that can run truly simultaneously at any given
time, and is typically set to the number of physical processor cores on
Strictly speaking it is better to use
the number of capabilities might vary at runtime.
Returns the number of Haskell threads that can run truly
simultaneously (on separate physical processors) at any given time. To change
this value, use
Set the number of Haskell threads that can run truly simultaneously
(on separate physical processors) at any given time. The number
forkOn is interpreted modulo this value. The initial
value is given by the
+RTS -N runtime flag.
This is also the number of threads that will participate in parallel garbage collection. It is strongly recommended that the number of capabilities is not set larger than the number of physical processor cores, and it may often be beneficial to leave one or more cores free to avoid contention with other processes in the machine.
throwTo raises an arbitrary exception in the target thread (GHC only).
Exception delivery synchronizes between the source and target thread:
throwTo does not return until the exception has been raised in the
target thread. The calling thread can thus be certain that the target
thread has received the exception. Exception delivery is also atomic
with respect to other exceptions. Atomicity is a useful property to have
when dealing with race conditions: e.g. if there are two threads that
can kill each other, it is guaranteed that only one of the threads
will get to kill the other.
Whatever work the target thread was doing when the exception was raised is not lost: the computation is suspended until required by another thread.
If the target thread is currently making a foreign call, then the
exception will not be raised (and hence
throwTo will not return)
until the call has completed. This is the case regardless of whether
the call is inside a
mask or not. However, in GHC a foreign call
can be annotated as
interruptible, in which case a
cause the RTS to attempt to cause the call to return; see the GHC
documentation for more details.
Important note: the behaviour of
throwTo differs from that described in
the paper "Asynchronous exceptions in Haskell"
In the paper,
throwTo is non-blocking; but the library implementation adopts
a more synchronous design in which
throwTo does not return until the exception
is received by the target thread. The trade-off is discussed in Section 9 of the paper.
Like any blocking operation,
throwTo is therefore interruptible (see Section 5.3 of
the paper). Unlike other interruptible operations, however,
is always interruptible, even if it does not actually block.
There is no guarantee that the exception will be delivered promptly,
although the runtime will endeavour to ensure that arbitrary
delays don't occur. In GHC, an exception can only be raised when a
thread reaches a safe point, where a safe point is where memory
allocation occurs. Some loops do not perform any memory allocation
inside the loop and therefore cannot be interrupted by a
If the target of
throwTo is the calling thread, then the behaviour
is the same as
throwIO, except that the exception
is thrown as an asynchronous exception. This means that if there is
an enclosing pure computation, which would be the case if the current
IO operation is inside
computation is not permanently replaced by the exception, but is
suspended as if it had received an asynchronous exception.
yield action allows (forces, in a co-operative multitasking
implementation) a context-switch to any other currently runnable
threads (if any), and is occasionally useful when implementing
labelThread stores a string as identifier for this thread if
you built a RTS with debugging support. This identifier will be used in
the debugging output to make distinction of different threads easier
(otherwise you only have the thread state object's address in the heap).
make a weak pointer to a
ThreadId. It can be important to do
this if you want to hold a reference to a
ThreadId while still
allowing the thread to receive the
BlockedIndefinitely family of
BlockedIndefinitelyOnMVar). Holding a normal
ThreadId reference will prevent the delivery of
BlockedIndefinitely exceptions because the reference could be
used as the target of
throwTo at any time, which would unblock
Weak ThreadId, on the other hand, will not prevent the
thread from receiving
BlockedIndefinitely exceptions. It is
still possible to throw an exception to a
Weak ThreadId, but the
caller must use
deRefWeak first to determine whether the thread
The current status of a thread
the thread is currently runnable or running
the thread has finished
the thread is blocked on some resource
the thread received an uncaught exception
|Eq ThreadStatus #|
|Ord ThreadStatus #|
|Show ThreadStatus #|
blocked on a computation in progress by another thread
currently in a foreign call
|Eq BlockReason #|
|Ord BlockReason #|
|Show BlockReason #|
returns the number of the capability on which the thread is currently
running, and a boolean indicating whether the thread is locked to
that capability or not. A thread is locked to a capability if it
was created with
Allocation counter and quota
Every thread has an allocation counter that tracks how much
memory has been allocated by the thread. The counter is
initialized to zero, and
setAllocationCounter sets the current
value. The allocation counter counts *down*, so in the absence of
a call to
setAllocationCounter its value is the negation of the
number of bytes of memory allocated by the thread.
There are two things that you can do with this counter:
- Use it as a simple profiling mechanism, with
- Use it as a resource limit. See
Allocation accounting is accurate only to about 4Kbytes.
Return the current value of the allocation counter for the current thread.
Enables the allocation counter to be treated as a limit for the
current thread. When the allocation limit is enabled, if the
allocation counter counts down below zero, the thread will be sent
AllocationLimitExceeded asynchronous exception. When this
happens, the counter is reinitialised (by default
to 100K, but tunable with the
+RTS -xq option) so that it can handle
the exception and perform any necessary clean up. If it exhausts
this additional allowance, another
is sent, and so forth. Like other asynchronous exceptions, the
AllocationLimitExceeded exception is deferred while the thread is inside
mask or an exception handler in
Note that memory allocation is unrelated to live memory, also known as heap residency. A thread can allocate a large amount of memory and retain anything between none and all of it. It is better to think of the allocation limit as a limit on CPU time, rather than a limit on memory.
Compared to using timeouts, allocation limits don't count time spent blocked or in foreign calls.
Disable allocation limit processing for the current thread.
A monad supporting atomic memory transactions.
|Monad STM #|
|Functor STM #|
|Applicative STM #|
|MonadPlus STM #|
|Alternative STM #|
Perform a series of STM actions atomically.
You cannot use
atomically inside an
Any attempt to do so will result in a runtime error. (Reason: allowing
this would effectively allow a transaction inside a transaction, depending
on exactly when the thunk is evaluated.)
Retry execution of the current memory transaction because it has seen values in TVars which mean that it should not continue (e.g. the TVars represent a shared buffer that is now empty). The implementation may block the thread until one of the TVars that it has read from has been udpated. (GHC only)
Compose two alternative STM actions (GHC only). If the first action completes without retrying then it forms the result of the orElse. Otherwise, if the first action retries, then the second action is tried in its place. If both actions retry then the orElse as a whole retries.
Throwing an exception in
STM aborts the transaction and propagates the
throw e `seq` x ===> throw e throwSTM e `seq` x ===> x
The first example will cause the exception
e to be raised,
whereas the second one won't. In fact,
throwSTM will only cause
an exception to be raised when it is used within the
throwSTM variant should be used in preference to
raise an exception within the
STM monad because it guarantees
ordering with respect to other
STM operations, whereas
Exception handling within STM actions.
alwaysSucceeds adds a new invariant that must be true when passed to alwaysSucceeds, at the end of the current transaction, and at the end of every subsequent transaction. If it fails at any of those points then the transaction violating it is aborted and the exception raised by the invariant is propagated.
always is a variant of alwaysSucceeds in which the invariant is expressed as an STM Bool action that must return True. Returning False or raising an exception are both treated as invariant failures.
Shared memory locations that support atomic memory transactions.
Return the current value stored in a TVar. This is equivalent to
readTVarIO = atomically . readTVar
but works much faster, because it doesn't perform a complete
transaction, it just reads the current value of the
Unsafely performs IO in the STM monad. Beware: this is a highly dangerous thing to do.
- The STM implementation will often run transactions multiple times, so you need to be prepared for this if your IO has any side effects.
- The STM implementation will abort transactions that are known to
be invalid and need to be restarted. This may happen in the middle
unsafeIOToSTM, so make sure you don't acquire any resources that need releasing (exception handlers are ignored when aborting the transaction). That includes doing any IO using Handles, for example. Getting this wrong will probably lead to random deadlocks.
- The transaction may have seen an inconsistent view of memory when
the IO runs. Invariants that you expect to be true throughout
your program may not be true inside a transaction, due to the
way transactions are implemented. Normally this wouldn't be visible
to the programmer, but using
unsafeIOToSTMcan expose it.