1 -----------------------------------------------------------------------------
3 -- Module : Control.Concurrent
4 -- Copyright : (c) The University of Glasgow 2001
5 -- License : BSD-style (see the file libraries/base/LICENSE)
7 -- Maintainer : libraries@haskell.org
8 -- Stability : experimental
9 -- Portability : non-portable (concurrency)
11 -- A common interface to a collection of useful concurrency
14 -----------------------------------------------------------------------------
16 module Control.Concurrent (
17 -- * Concurrent Haskell
21 -- * Basic concurrency operations
24 #ifdef __GLASGOW_HASKELL__
29 #ifdef __GLASGOW_HASKELL__
43 #ifdef __GLASGOW_HASKELL__
45 threadDelay, -- :: Int -> IO ()
46 threadWaitRead, -- :: Int -> IO ()
47 threadWaitWrite, -- :: Int -> IO ()
50 -- * Communication abstractions
52 module Control.Concurrent.MVar,
53 module Control.Concurrent.Chan,
54 module Control.Concurrent.QSem,
55 module Control.Concurrent.QSemN,
56 module Control.Concurrent.SampleVar,
58 -- * Merging of streams
60 mergeIO, -- :: [a] -> [a] -> IO [a]
61 nmergeIO, -- :: [[a]] -> IO [a]
65 #ifdef __GLASGOW_HASKELL__
68 rtsSupportsBoundThreads,
75 -- * GHC's implementation of concurrency
77 -- |This section describes features specific to GHC's
78 -- implementation of Concurrent Haskell.
80 -- ** Terminating the program
91 import Control.Exception as Exception
93 #ifdef __GLASGOW_HASKELL__
94 import GHC.Conc ( ThreadId(..), myThreadId, killThread, yield,
95 threadDelay, threadWaitRead, threadWaitWrite )
96 import GHC.TopHandler ( reportStackOverflow, reportError )
97 import GHC.IOBase ( IO(..) )
98 import GHC.IOBase ( unsafeInterleaveIO )
99 import GHC.IOBase ( newIORef, readIORef, writeIORef )
102 import Foreign.StablePtr
103 import Foreign.C.Types ( CInt )
104 import Control.Monad ( when )
111 import Control.Concurrent.MVar
112 import Control.Concurrent.Chan
113 import Control.Concurrent.QSem
114 import Control.Concurrent.QSemN
115 import Control.Concurrent.SampleVar
123 The concurrency extension for Haskell is described in the paper
125 <http://www.haskell.org/ghc/docs/papers/concurrent-haskell.ps.gz>.
127 Concurrency is \"lightweight\", which means that both thread creation
128 and context switching overheads are extremely low. Scheduling of
129 Haskell threads is done internally in the Haskell runtime system, and
130 doesn't make use of any operating system-supplied thread packages.
132 However, if you want to interact with a foreign library that expects your
133 program to use the operating system-supplied thread package, you can do so
134 by using 'forkOS' instead of 'forkIO'.
136 Haskell threads can communicate via 'MVar's, a kind of synchronised
137 mutable variable (see "Control.Concurrent.MVar"). Several common
138 concurrency abstractions can be built from 'MVar's, and these are
139 provided by the "Control.Concurrent" library.
140 In GHC, threads may also communicate via exceptions.
145 Scheduling may be either pre-emptive or co-operative,
146 depending on the implementation of Concurrent Haskell (see below
147 for information related to specific compilers). In a co-operative
148 system, context switches only occur when you use one of the
149 primitives defined in this module. This means that programs such
153 > main = forkIO (write 'a') >> write 'b'
154 > where write c = putChar c >> write c
156 will print either @aaaaaaaaaaaaaa...@ or @bbbbbbbbbbbb...@,
157 instead of some random interleaving of @a@s and @b@s. In
158 practice, cooperative multitasking is sufficient for writing
159 simple graphical user interfaces.
163 Calling a foreign C procedure (such as @getchar@) that blocks waiting
164 for input will block /all/ threads, unless the @threadsafe@ attribute
165 is used on the foreign call (and your compiler \/ operating system
166 supports it). GHC's I\/O system uses non-blocking I\/O internally to
167 implement thread-friendly I\/O, so calling standard Haskell I\/O
168 functions blocks only the thread making the call.
171 -- Thread Ids, specifically the instances of Eq and Ord for these things.
172 -- The ThreadId type itself is defined in std/PrelConc.lhs.
174 -- Rather than define a new primitve, we use a little helper function
175 -- cmp_thread in the RTS.
177 #ifdef __GLASGOW_HASKELL__
178 id2TSO :: ThreadId -> ThreadId#
179 id2TSO (ThreadId t) = t
181 foreign import ccall unsafe "cmp_thread" cmp_thread :: ThreadId# -> ThreadId# -> Int
184 cmpThread :: ThreadId -> ThreadId -> Ordering
186 case cmp_thread (id2TSO t1) (id2TSO t2) of
191 instance Eq ThreadId where
193 case t1 `cmpThread` t2 of
197 instance Ord ThreadId where
200 foreign import ccall unsafe "rts_getThreadId" getThreadId :: ThreadId# -> Int
202 instance Show ThreadId where
204 showString "ThreadId " .
205 showsPrec d (getThreadId (id2TSO t))
208 This sparks off a new thread to run the 'IO' computation passed as the
209 first argument, and returns the 'ThreadId' of the newly created
212 The new thread will be a lightweight thread; if you want to use a foreign
213 library that uses thread-local storage, use 'forkOS' instead.
215 forkIO :: IO () -> IO ThreadId
216 forkIO action = IO $ \ s ->
217 case (fork# action_plus s) of (# s1, id #) -> (# s1, ThreadId id #)
219 action_plus = Exception.catch action childHandler
221 childHandler :: Exception -> IO ()
222 childHandler err = Exception.catch (real_handler err) childHandler
224 real_handler :: Exception -> IO ()
227 -- ignore thread GC and killThread exceptions:
228 BlockedOnDeadMVar -> return ()
229 BlockedIndefinitely -> return ()
230 AsyncException ThreadKilled -> return ()
232 -- report all others:
233 AsyncException StackOverflow -> reportStackOverflow
234 other -> reportError other
236 #endif /* __GLASGOW_HASKELL__ */
242 mergeIO :: [a] -> [a] -> IO [a]
243 nmergeIO :: [[a]] -> IO [a]
246 -- The 'mergeIO' and 'nmergeIO' functions fork one thread for each
247 -- input list that concurrently evaluates that list; the results are
248 -- merged into a single output list.
250 -- Note: Hugs does not provide these functions, since they require
251 -- preemptive multitasking.
254 = newEmptyMVar >>= \ tail_node ->
255 newMVar tail_node >>= \ tail_list ->
256 newQSem max_buff_size >>= \ e ->
257 newMVar 2 >>= \ branches_running ->
261 forkIO (suckIO branches_running buff ls) >>
262 forkIO (suckIO branches_running buff rs) >>
263 takeMVar tail_node >>= \ val ->
268 = (MVar (MVar [a]), QSem)
270 suckIO :: MVar Int -> Buffer a -> [a] -> IO ()
272 suckIO branches_running buff@(tail_list,e) vs
274 [] -> takeMVar branches_running >>= \ val ->
276 takeMVar tail_list >>= \ node ->
278 putMVar tail_list node
280 putMVar branches_running (val-1)
283 takeMVar tail_list >>= \ node ->
284 newEmptyMVar >>= \ next_node ->
286 takeMVar next_node >>= \ y ->
288 return y) >>= \ next_node_val ->
289 putMVar node (x:next_node_val) >>
290 putMVar tail_list next_node >>
291 suckIO branches_running buff xs
297 newEmptyMVar >>= \ tail_node ->
298 newMVar tail_node >>= \ tail_list ->
299 newQSem max_buff_size >>= \ e ->
300 newMVar len >>= \ branches_running ->
304 mapIO (\ x -> forkIO (suckIO branches_running buff x)) lss >>
305 takeMVar tail_node >>= \ val ->
309 mapIO f xs = sequence (map f xs)
310 #endif /* __HUGS__ */
312 #ifdef __GLASGOW_HASKELL__
313 -- ---------------------------------------------------------------------------
318 Support for multiple operating system threads and bound threads as described
319 below is currently only available in the GHC runtime system if you use the
320 /-threaded/ option when linking.
322 Other Haskell systems do not currently support multiple operating system threads.
324 A bound thread is a haskell thread that is /bound/ to an operating system
325 thread. While the bound thread is still scheduled by the Haskell run-time
326 system, the operating system thread takes care of all the foreign calls made
329 To a foreign library, the bound thread will look exactly like an ordinary
330 operating system thread created using OS functions like @pthread_create@
333 Bound threads can be created using the 'forkOS' function below. All foreign
334 exported functions are run in a bound thread (bound to the OS thread that
335 called the function). Also, the @main@ action of every Haskell program is
336 run in a bound thread.
338 Why do we need this? Because if a foreign library is called from a thread
339 created using 'forkIO', it won't have access to any /thread-local state/ -
340 state variables that have specific values for each OS thread
341 (see POSIX's @pthread_key_create@ or Win32's @TlsAlloc@). Therefore, some
342 libraries (OpenGL, for example) will not work from a thread created using
343 'forkIO'. They work fine in threads created using 'forkOS' or when called
344 from @main@ or from a @foreign export@.
347 -- | 'True' if bound threads are supported.
348 -- If @rtsSupportsBoundThreads@ is 'False', 'isCurrentThreadBound'
349 -- will always return 'False' and both 'forkOS' and 'runInBoundThread' will
351 foreign import ccall rtsSupportsBoundThreads :: Bool
355 Like 'forkIO', this sparks off a new thread to run the 'IO' computation passed as the
356 first argument, and returns the 'ThreadId' of the newly created
359 However, @forkOS@ uses operating system-supplied multithreading support to create
360 a new operating system thread. The new thread is /bound/, which means that
361 all foreign calls made by the 'IO' computation are guaranteed to be executed
362 in this new operating system thread; also, the operating system thread is not
363 used for any other foreign calls.
365 This means that you can use all kinds of foreign libraries from this thread
366 (even those that rely on thread-local state), without the limitations of 'forkIO'.
368 forkOS :: IO () -> IO ThreadId
370 foreign export ccall forkOS_entry
371 :: StablePtr (IO ()) -> IO ()
373 foreign import ccall "forkOS_entry" forkOS_entry_reimported
374 :: StablePtr (IO ()) -> IO ()
376 forkOS_entry stableAction = do
377 action <- deRefStablePtr stableAction
380 foreign import ccall forkOS_createThread
381 :: StablePtr (IO ()) -> IO CInt
383 failNonThreaded = fail $ "RTS doesn't support multiple OS threads "
384 ++"(use ghc -threaded when linking)"
387 | rtsSupportsBoundThreads = do
389 let action_plus = Exception.catch action childHandler
390 entry <- newStablePtr (myThreadId >>= putMVar mv >> action_plus)
391 err <- forkOS_createThread entry
392 when (err /= 0) $ fail "Cannot create OS thread."
396 | otherwise = failNonThreaded
398 -- | Returns 'True' if the calling thread is /bound/, that is, if it is
399 -- safe to use foreign libraries that rely on thread-local state from the
401 isCurrentThreadBound :: IO Bool
402 isCurrentThreadBound = IO $ \ s# ->
403 case isCurrentThreadBound# s# of
404 (# s2#, flg #) -> (# s2#, not (flg ==# 0#) #)
408 Run the 'IO' computation passed as the first argument. If the calling thread
409 is not /bound/, a bound thread is created temporarily. @runInBoundThread@
410 doesn't finish until the 'IO' computation finishes.
412 You can wrap a series of foreign function calls that rely on thread-local state
413 with @runInBoundThread@ so that you can use them without knowing whether the
414 current thread is /bound/.
416 runInBoundThread :: IO a -> IO a
418 runInBoundThread action
419 | rtsSupportsBoundThreads = do
420 bound <- isCurrentThreadBound
424 ref <- newIORef undefined
425 let action_plus = Exception.try action >>= writeIORef ref
427 bracket (newStablePtr action_plus)
429 (\cEntry -> forkOS_entry_reimported cEntry >> readIORef ref)
430 case resultOrException of
431 Left exception -> Exception.throw exception
432 Right result -> return result
433 | otherwise = failNonThreaded
436 Run the 'IO' computation passed as the first argument. If the calling thread
437 is /bound/, an unbound thread is created temporarily using 'forkIO'.
438 @runInBoundThread@ doesn't finish until the 'IO' computation finishes.
440 Use this function /only/ in the rare case that you have actually observed a
441 performance loss due to the use of bound threads. A program that
442 doesn't need it's main thread to be bound and makes /heavy/ use of concurrency
443 (e.g. a web server), might want to wrap it's @main@ action in
444 @runInUnboundThread@.
446 runInUnboundThread :: IO a -> IO a
448 runInUnboundThread action = do
449 bound <- isCurrentThreadBound
453 forkIO (Exception.try action >>= putMVar mv)
454 takeMVar mv >>= \either -> case either of
455 Left exception -> Exception.throw exception
456 Right result -> return result
459 #endif /* __GLASGOW_HASKELL__ */
461 -- ---------------------------------------------------------------------------
466 In a standalone GHC program, only the main thread is
467 required to terminate in order for the process to terminate.
468 Thus all other forked threads will simply terminate at the same
469 time as the main thread (the terminology for this kind of
470 behaviour is \"daemonic threads\").
472 If you want the program to wait for child threads to
473 finish before exiting, you need to program this yourself. A
474 simple mechanism is to have each child thread write to an
475 'MVar' when it completes, and have the main
476 thread wait on all the 'MVar's before
479 > myForkIO :: IO () -> IO (MVar ())
481 > mvar <- newEmptyMVar
482 > forkIO (io `finally` putMVar mvar ())
485 Note that we use 'finally' from the
486 "Control.Exception" module to make sure that the
487 'MVar' is written to even if the thread dies or
488 is killed for some reason.
490 A better method is to keep a global list of all child
491 threads which we should wait for at the end of the program:
493 > children :: MVar [MVar ()]
494 > children = unsafePerformIO (newMVar [])
496 > waitForChildren :: IO ()
497 > waitForChildren = do
498 > cs <- takeMVar children
502 > putMVar children ms
506 > forkChild :: IO () -> IO ()
508 > mvar <- newEmptyMVar
509 > childs <- takeMVar children
510 > putMVar children (mvar:childs)
511 > forkIO (io `finally` putMVar mvar ())
514 > later waitForChildren $
517 The main thread principle also applies to calls to Haskell from
518 outside, using @foreign export@. When the @foreign export@ed
519 function is invoked, it starts a new main thread, and it returns
520 when this main thread terminates. If the call causes new
521 threads to be forked, they may remain in the system after the
522 @foreign export@ed function has returned.
527 GHC implements pre-emptive multitasking: the execution of
528 threads are interleaved in a random fashion. More specifically,
529 a thread may be pre-empted whenever it allocates some memory,
530 which unfortunately means that tight loops which do no
531 allocation tend to lock out other threads (this only seems to
532 happen with pathological benchmark-style code, however).
534 The rescheduling timer runs on a 20ms granularity by
535 default, but this may be altered using the
536 @-i\<n\>@ RTS option. After a rescheduling
537 \"tick\" the running thread is pre-empted as soon as
541 @aaaa@ @bbbb@ example may not
542 work too well on GHC (see Scheduling, above), due
543 to the locking on a 'System.IO.Handle'. Only one thread
544 may hold the lock on a 'System.IO.Handle' at any one
545 time, so if a reschedule happens while a thread is holding the
546 lock, the other thread won't be able to run. The upshot is that
547 the switch from @aaaa@ to
548 @bbbbb@ happens infrequently. It can be
549 improved by lowering the reschedule tick period. We also have a
550 patch that causes a reschedule whenever a thread waiting on a
551 lock is woken up, but haven't found it to be useful for anything
552 other than this example :-)