Haskell Hierarchical Libraries (base package)ContentsIndex
Control.Concurrent
Portability non-portable (concurrency)
Stability experimental
Maintainer libraries@haskell.org
Contents
Concurrent Haskell
Basic concurrency operations
Scheduling
Blocking
Waiting
Communication abstractions
Merging of streams
Bound Threads
GHC's implementation of concurrency
Terminating the program
Pre-emption
Description
A common interface to a collection of useful concurrency abstractions.
Synopsis
data ThreadId
myThreadId :: IO ThreadId
forkIO :: IO () -> IO ThreadId
killThread :: ThreadId -> IO ()
throwTo :: ThreadId -> Exception -> IO ()
yield :: IO ()
threadDelay :: Int -> IO ()
threadWaitRead :: Int -> IO ()
threadWaitWrite :: Int -> IO ()
module Control.Concurrent.MVar
module Control.Concurrent.Chan
module Control.Concurrent.QSem
module Control.Concurrent.QSemN
module Control.Concurrent.SampleVar
mergeIO :: [a] -> [a] -> IO [a]
nmergeIO :: [[a]] -> IO [a]
rtsSupportsBoundThreads :: Bool
forkOS :: IO () -> IO ThreadId
isCurrentThreadBound :: IO Bool
runInBoundThread :: IO a -> IO a
runInUnboundThread :: IO a -> IO a
Concurrent Haskell

The concurrency extension for Haskell is described in the paper Concurrent Haskell http://www.haskell.org/ghc/docs/papers/concurrent-haskell.ps.gz.

Concurrency is "lightweight", which means that both thread creation and context switching overheads are extremely low. Scheduling of Haskell threads is done internally in the Haskell runtime system, and doesn't make use of any operating system-supplied thread packages.

However, if you want to interact with a foreign library that expects your program to use the operating system-supplied thread package, you can do so by using forkOS instead of forkIO.

Haskell threads can communicate via MVars, a kind of synchronised mutable variable (see Control.Concurrent.MVar). Several common concurrency abstractions can be built from MVars, and these are provided by the Control.Concurrent library. In GHC, threads may also communicate via exceptions.

Basic concurrency operations
data ThreadId

A ThreadId is an abstract type representing a handle to a thread. ThreadId is an instance of Eq, Ord and Show, where the Ord instance implements an arbitrary total ordering over ThreadIds. The 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 program.

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 ThreadId. This misfeature will hopefully be corrected at a later date.

Note: Hugs does not provide any operations on other threads; it defines ThreadId as a synonym for ().

Instances
Eq ThreadId
Ord ThreadId
Show ThreadId
myThreadId :: IO ThreadId
Returns the ThreadId of the calling thread (GHC only).
forkIO :: IO () -> IO ThreadId

This sparks off a new thread to run the IO computation passed as the first argument, and returns the ThreadId of the newly created thread.

The new thread will be a lightweight thread; if you want to use a foreign library that uses thread-local storage, use forkOS instead.

killThread :: ThreadId -> IO ()

killThread terminates the given thread (GHC only). Any work already done by the thread isn't lost: the computation is suspended until required by another thread. The memory used by the thread will be garbage collected if it isn't referenced from anywhere. The killThread function is defined in terms of throwTo:

 killThread tid = throwTo tid (AsyncException ThreadKilled)
throwTo :: ThreadId -> Exception -> IO ()

throwTo raises an arbitrary exception in the target thread (GHC only).

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. This is a useful property to know when dealing with race conditions: eg. 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.

Scheduling

Scheduling may be either pre-emptive or co-operative, depending on the implementation of Concurrent Haskell (see below for imformation related to specific compilers). In a co-operative system, context switches only occur when you use one of the primitives defined in this module. This means that programs such as:

   main = forkIO (write 'a') >> write 'b'
     where write c = putChar c >> write c

will print either aaaaaaaaaaaaaa... or bbbbbbbbbbbb..., instead of some random interleaving of as and bs. In practice, cooperative multitasking is sufficient for writing simple graphical user interfaces.

yield :: IO ()
The 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 concurrency abstractions.
Blocking
Calling a foreign C procedure (such as getchar) that blocks waiting for input will block all threads, unless the threadsafe attribute is used on the foreign call (and your compiler / operating system supports it). GHC's I/O system uses non-blocking I/O internally to implement thread-friendly I/O, so calling standard Haskell I/O functions blocks only the thread making the call.
Waiting
threadDelay :: Int -> IO ()

Suspends the current thread for a given number of microseconds (GHC only).

Note that the resolution used by the Haskell runtime system's internal timer is 1/50 second, and threadDelay will round its argument up to the nearest multiple of this resolution.

There is no guarantee that the thread will be rescheduled promptly when the delay has expired, but the thread will never continue to run earlier than specified.

threadWaitRead :: Int -> IO ()
Block the current thread until data is available to read on the given file descriptor (GHC only).
threadWaitWrite :: Int -> IO ()
Block the current thread until data can be written to the given file descriptor (GHC only).
Communication abstractions
module Control.Concurrent.MVar
module Control.Concurrent.Chan
module Control.Concurrent.QSem
module Control.Concurrent.QSemN
module Control.Concurrent.SampleVar
Merging of streams
mergeIO :: [a] -> [a] -> IO [a]
nmergeIO :: [[a]] -> IO [a]

The mergeIO and nmergeIO functions fork one thread for each input list that concurrently evaluates that list; the results are merged into a single output list.

Note: Hugs does not provide these functions, since they require preemptive multitasking.

Bound Threads

Support for multiple operating system threads and bound threads as described below is currently only available in the GHC runtime system when the runtime system has been compiled using a special option.

When recompiling GHC, use ./configure --enable-threaded-rts to enable this. To find your GHC has already been compiled that way, use rtsSupportsBoundThreads from GHCi.

Other Haskell systems do not currently support multiple operating system threads.

A bound thread is a haskell thread that is bound to an operating system thread. While the bound thread is still scheduled by the Haskell run-time system, the operating system thread takes care of all the foreign calls made by the bound thread.

To a foreign library, the bound thread will look exactly like an ordinary operating system thread created using OS functions like pthread_create or CreateThread.

Bound threads can be created using the forkOS function below. All foreign exported functions are run in a bound thread (bound to the OS thread that called the function). Also, the main action of every Haskell program is run in a bound thread.

Why do we need this? Because if a foreign library is called from a thread created using forkIO, it won't have access to any thread-local state - state variables that have specific values for each OS thread (see POSIX's pthread_key_create or Win32's TlsAlloc). Therefore, some libraries (OpenGL, for example) will not work from a thread created using forkIO. They work fine in threads created using forkOS or when called from main or from a foreign export.

rtsSupportsBoundThreads :: Bool
True if bound threads are supported. If rtsSupportsBoundThreads is False, isCurrentThreadBound will always return False and both forkOS and runInBoundThread will fail.
forkOS :: IO () -> IO ThreadId

Like forkIO, this sparks off a new thread to run the IO computation passed as the first argument, and returns the ThreadId of the newly created thread.

However, forkOS uses operating system-supplied multithreading support to create a new operating system thread. The new thread is bound, which means that all foreign calls made by the IO computation are guaranteed to be executed in this new operating system thread; also, the operating system thread is not used for any other foreign calls.

This means that you can use all kinds of foreign libraries from this thread (even those that rely on thread-local state), without the limitations of forkIO.

isCurrentThreadBound :: IO Bool
Returns True if the calling thread is bound, that is, if it is safe to use foreign libraries that rely on thread-local state from the calling thread.
runInBoundThread :: IO a -> IO a

Run the IO computation passed as the first argument. If the calling thread is not bound, a bound thread is created temporarily. runInBoundThread doesn't finish until the IO computation finishes.

You can wrap a series of foreign function calls that rely on thread-local state with runInBoundThread so that you can use them without knowing whether the current thread is bound.

runInUnboundThread :: IO a -> IO a

Run the IO computation passed as the first argument. If the calling thread is bound, an unbound thread is created temporarily using forkIO. runInBoundThread doesn't finish until the IO computation finishes.

Use this function only in the rare case that you have actually observed a performance loss due to the use of bound threads. A program that doesn't need it's main thread to be bound and makes heavy use of concurrency (e.g. a web server), might want to wrap it's main action in runInUnboundThread.

GHC's implementation of concurrency
This section describes features specific to GHC's implementation of Concurrent Haskell.
Terminating the program

In a standalone GHC program, only the main thread is required to terminate in order for the process to terminate. Thus all other forked threads will simply terminate at the same time as the main thread (the terminology for this kind of behaviour is "daemonic threads").

If you want the program to wait for child threads to finish before exiting, you need to program this yourself. A simple mechanism is to have each child thread write to an MVar when it completes, and have the main thread wait on all the MVars before exiting:

   myForkIO :: IO () -> IO (MVar ())
   myForkIO io = do
     mvar \<- newEmptyMVar
     forkIO (io \`finally\` putMVar mvar ())
     return mvar

Note that we use finally from the Control.Exception module to make sure that the MVar is written to even if the thread dies or is killed for some reason.

A better method is to keep a global list of all child threads which we should wait for at the end of the program:

     children :: MVar [MVar ()]
     children = unsafePerformIO (newMVar [])
     
     waitForChildren :: IO ()
     waitForChildren = do
     	(mvar:mvars) \<- takeMVar children
     	putMVar children mvars
     	takeMVar mvar
     	waitForChildren
     
     forkChild :: IO () -> IO ()
     forkChild io = do
     	 mvar \<- newEmptyMVar
     	 forkIO (p \`finally\` putMVar mvar ())
     	 childs \<- takeMVar children
     	 putMVar children (mvar:childs)
     
     later = flip finally
     
     main =
     	later waitForChildren $
     	...

The main thread principle also applies to calls to Haskell from outside, using foreign export. When the foreign exported function is invoked, it starts a new main thread, and it returns when this main thread terminates. If the call causes new threads to be forked, they may remain in the system after the foreign exported function has returned.

Pre-emption

GHC implements pre-emptive multitasking: the execution of threads are interleaved in a random fashion. More specifically, a thread may be pre-empted whenever it allocates some memory, which unfortunately means that tight loops which do no allocation tend to lock out other threads (this only seems to happen with pathalogical benchmark-style code, however).

The rescheduling timer runs on a 20ms granularity by default, but this may be altered using the -i<n> RTS option. After a rescheduling "tick" the running thread is pre-empted as soon as possible.

One final note: the aaaa bbbb example may not work too well on GHC (see Scheduling, above), due to the locking on a Handle. Only one thread may hold the lock on a Handle at any one time, so if a reschedule happens while a thread is holding the lock, the other thread won't be able to run. The upshot is that the switch from aaaa to bbbbb happens infrequently. It can be improved by lowering the reschedule tick period. We also have a patch that causes a reschedule whenever a thread waiting on a lock is woken up, but haven't found it to be useful for anything other than this example :-)

Produced by Haddock version 0.6