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]
67 #ifdef __GLASGOW_HASKELL__
68 rtsSupportsBoundThreads,
74 -- * GHC's implementation of concurrency
76 -- |This section describes features specific to GHC's
77 -- implementation of Concurrent Haskell.
79 -- ** Terminating the program
90 import Control.Exception as Exception
92 #ifdef __GLASGOW_HASKELL__
94 import GHC.TopHandler ( reportStackOverflow, reportError )
95 import GHC.IOBase ( IO(..) )
96 import GHC.IOBase ( unsafeInterleaveIO )
97 import GHC.IOBase ( newIORef, readIORef, writeIORef )
100 import Foreign.StablePtr
101 import Foreign.C.Types ( CInt )
102 import Control.Monad ( when )
109 import Control.Concurrent.MVar
110 import Control.Concurrent.Chan
111 import Control.Concurrent.QSem
112 import Control.Concurrent.QSemN
113 import Control.Concurrent.SampleVar
121 The concurrency extension for Haskell is described in the paper
123 <http://www.haskell.org/ghc/docs/papers/concurrent-haskell.ps.gz>.
125 Concurrency is \"lightweight\", which means that both thread creation
126 and context switching overheads are extremely low. Scheduling of
127 Haskell threads is done internally in the Haskell runtime system, and
128 doesn't make use of any operating system-supplied thread packages.
130 However, if you want to interact with a foreign library that expects your
131 program to use the operating system-supplied thread package, you can do so
132 by using 'forkOS' instead of 'forkIO'.
134 Haskell threads can communicate via 'MVar's, a kind of synchronised
135 mutable variable (see "Control.Concurrent.MVar"). Several common
136 concurrency abstractions can be built from 'MVar's, and these are
137 provided by the "Control.Concurrent" library.
138 In GHC, threads may also communicate via exceptions.
143 Scheduling may be either pre-emptive or co-operative,
144 depending on the implementation of Concurrent Haskell (see below
145 for imformation related to specific compilers). In a co-operative
146 system, context switches only occur when you use one of the
147 primitives defined in this module. This means that programs such
151 > main = forkIO (write 'a') >> write 'b'
152 > where write c = putChar c >> write c
154 will print either @aaaaaaaaaaaaaa...@ or @bbbbbbbbbbbb...@,
155 instead of some random interleaving of @a@s and @b@s. In
156 practice, cooperative multitasking is sufficient for writing
157 simple graphical user interfaces.
161 Calling a foreign C procedure (such as @getchar@) that blocks waiting
162 for input will block /all/ threads, unless the @threadsafe@ attribute
163 is used on the foreign call (and your compiler \/ operating system
164 supports it). GHC's I\/O system uses non-blocking I\/O internally to
165 implement thread-friendly I\/O, so calling standard Haskell I\/O
166 functions blocks only the thread making the call.
169 -- Thread Ids, specifically the instances of Eq and Ord for these things.
170 -- The ThreadId type itself is defined in std/PrelConc.lhs.
172 -- Rather than define a new primitve, we use a little helper function
173 -- cmp_thread in the RTS.
175 #ifdef __GLASGOW_HASKELL__
176 id2TSO :: ThreadId -> ThreadId#
177 id2TSO (ThreadId t) = t
179 foreign import ccall unsafe "cmp_thread" cmp_thread :: ThreadId# -> ThreadId# -> Int
182 cmpThread :: ThreadId -> ThreadId -> Ordering
184 case cmp_thread (id2TSO t1) (id2TSO t2) of
189 instance Eq ThreadId where
191 case t1 `cmpThread` t2 of
195 instance Ord ThreadId where
198 foreign import ccall unsafe "rts_getThreadId" getThreadId :: ThreadId# -> Int
200 instance Show ThreadId where
202 showString "ThreadId " .
203 showsPrec d (getThreadId (id2TSO t))
206 This sparks off a new thread to run the 'IO' computation passed as the
207 first argument, and returns the 'ThreadId' of the newly created
210 The new thread will be a lightweight thread; if you want to use a foreign
211 library that uses thread-local storage, use 'forkOS' instead.
213 forkIO :: IO () -> IO ThreadId
214 forkIO action = IO $ \ s ->
215 case (fork# action_plus s) of (# s1, id #) -> (# s1, ThreadId id #)
217 action_plus = Exception.catch action childHandler
219 childHandler :: Exception -> IO ()
220 childHandler err = Exception.catch (real_handler err) childHandler
222 real_handler :: Exception -> IO ()
225 -- ignore thread GC and killThread exceptions:
226 BlockedOnDeadMVar -> return ()
227 AsyncException ThreadKilled -> return ()
229 -- report all others:
230 AsyncException StackOverflow -> reportStackOverflow False
231 ErrorCall s -> reportError False s
232 other -> reportError False (showsPrec 0 other "\n")
234 #endif /* __GLASGOW_HASKELL__ */
240 mergeIO :: [a] -> [a] -> IO [a]
241 nmergeIO :: [[a]] -> IO [a]
244 -- The 'mergeIO' and 'nmergeIO' functions fork one thread for each
245 -- input list that concurrently evaluates that list; the results are
246 -- merged into a single output list.
248 -- Note: Hugs does not provide these functions, since they require
249 -- preemptive multitasking.
252 = newEmptyMVar >>= \ tail_node ->
253 newMVar tail_node >>= \ tail_list ->
254 newQSem max_buff_size >>= \ e ->
255 newMVar 2 >>= \ branches_running ->
259 forkIO (suckIO branches_running buff ls) >>
260 forkIO (suckIO branches_running buff rs) >>
261 takeMVar tail_node >>= \ val ->
266 = (MVar (MVar [a]), QSem)
268 suckIO :: MVar Int -> Buffer a -> [a] -> IO ()
270 suckIO branches_running buff@(tail_list,e) vs
272 [] -> takeMVar branches_running >>= \ val ->
274 takeMVar tail_list >>= \ node ->
276 putMVar tail_list node
278 putMVar branches_running (val-1)
281 takeMVar tail_list >>= \ node ->
282 newEmptyMVar >>= \ next_node ->
284 takeMVar next_node >>= \ y ->
286 return y) >>= \ next_node_val ->
287 putMVar node (x:next_node_val) >>
288 putMVar tail_list next_node >>
289 suckIO branches_running buff xs
295 newEmptyMVar >>= \ tail_node ->
296 newMVar tail_node >>= \ tail_list ->
297 newQSem max_buff_size >>= \ e ->
298 newMVar len >>= \ branches_running ->
302 mapIO (\ x -> forkIO (suckIO branches_running buff x)) lss >>
303 takeMVar tail_node >>= \ val ->
307 mapIO f xs = sequence (map f xs)
308 #endif /* __HUGS__ */
310 -- ---------------------------------------------------------------------------
315 Support for multiple operating system threads and bound threads as described
316 below is currently only available in the GHC runtime system when the runtime system
317 has been compiled using a special option.
319 When recompiling GHC, use ./configure --enable-threaded-rts to enable this.
320 To find your GHC has already been compiled that way, use
321 'rtsSupportsBoundThreads' from GHCi.
323 Other Haskell systems do not currently support multiple operating system threads.
325 A bound thread is a haskell thread that is /bound/ to an operating system
326 thread. While the bound thread is still scheduled by the Haskell run-time
327 system, the operating system thread takes care of all the foreign calls made
330 To a foreign library, the bound thread will look exactly like an ordinary
331 operating system thread created using OS functions like @pthread_create@
334 Bound threads can be created using the 'forkOS' function below. All foreign
335 exported functions are run in a bound thread (bound to the OS thread that
336 called the function). Also, the @main@ action of every Haskell program is
337 run in a bound thread.
339 Why do we need this? Because if a foreign library is called from a thread
340 created using 'forkIO', it won't have access to any /thread-local state/ -
341 state variables that have specific values for each OS thread
342 (see POSIX's @pthread_key_create@ or Win32's @TlsAlloc@). Therefore, some
343 libraries (OpenGL, for example) will not work from a thread created using
344 'forkIO'. They work fine in threads created using 'forkOS' or when called
345 from @main@ or from a @foreign export@.
351 -- | 'True' if bound threads are supported.
352 -- If @rtsSupportsBoundThreads@ is 'False', 'isCurrentThreadBound'
353 -- will always return 'False' and both 'forkOS' and 'runInBoundThread' will
355 foreign import ccall rtsSupportsBoundThreads :: Bool
359 Like 'forkIO', this sparks off a new thread to run the 'IO' computation passed as the
360 first argument, and returns the 'ThreadId' of the newly created
363 However, @forkOS@ uses operating system-supplied multithreading support to create
364 a new operating system thread. The new thread is /bound/, which means that
365 all foreign calls made by the 'IO' computation are guaranteed to be executed
366 in this new operating system thread; also, the operating system thread is not
367 used for any other foreign calls.
369 This means that you can use all kinds of foreign libraries from this thread
370 (even those that rely on thread-local state), without the limitations of 'forkIO'.
372 forkOS :: IO () -> IO ThreadId
374 foreign export ccall forkOS_entry
375 :: StablePtr (IO ()) -> IO ()
377 foreign import ccall "forkOS_entry" forkOS_entry_reimported
378 :: StablePtr (IO ()) -> IO ()
380 forkOS_entry stableAction = do
381 action <- deRefStablePtr stableAction
384 foreign import ccall forkOS_createThread
385 :: StablePtr (IO ()) -> IO CInt
388 | rtsSupportsBoundThreads = do
390 let action_plus = Exception.catch action childHandler
391 entry <- newStablePtr (myThreadId >>= putMVar mv >> action_plus)
392 err <- forkOS_createThread entry
393 when (err /= 0) $ fail "Cannot create OS thread."
397 | otherwise = fail "RTS not built to support multiple OS threads."
399 -- | Returns 'True' if the calling thread is /bound/, that is, if it is
400 -- safe to use foreign libraries that rely on thread-local state from the
402 isCurrentThreadBound :: IO Bool
403 isCurrentThreadBound = IO $ \ s# ->
404 case isCurrentThreadBound# s# of
405 (# s2#, flg #) -> (# s2#, not (flg ==# 0#) #)
409 Run the 'IO' computation passed as the first argument. If the calling thread
410 is not /bound/, a bound thread is created temporarily. @runInBoundThread@
411 doesn't finish until the 'IO' computation finishes.
413 You can wrap a series of foreign function calls that rely on thread-local state
414 with @runInBoundThread@ so that you can use them without knowing whether the
415 current thread is /bound/.
417 runInBoundThread :: IO a -> IO a
419 runInBoundThread action
420 | rtsSupportsBoundThreads = do
421 bound <- isCurrentThreadBound
425 ref <- newIORef undefined
426 let action_plus = Exception.try action >>= writeIORef ref
428 bracket (newStablePtr action_plus)
430 (\cEntry -> forkOS_entry_reimported cEntry >> readIORef ref)
431 case resultOrException of
432 Left exception -> Exception.throw exception
433 Right result -> return result
434 | otherwise = fail "RTS not built to support multiple OS threads."
437 Run the 'IO' computation passed as the first argument. If the calling thread
438 is /bound/, an unbound thread is created temporarily using 'forkIO'.
439 @runInBoundThread@ doesn't finish until the 'IO' computation finishes.
441 Use this function /only/ in the rare case that you have actually observed a
442 performance loss due to the use of bound threads. A program that
443 doesn't need it's main thread to be bound and makes /heavy/ use of concurrency
444 (e.g. a web server), might want to wrap it's @main@ action in
445 @runInUnboundThread@.
447 runInUnboundThread :: IO a -> IO a
449 runInUnboundThread action = do
450 bound <- isCurrentThreadBound
454 forkIO (E.try action >>= putMVar mv)
455 takeMVar mv >>= \either -> case either of
456 Left exception -> E.throw exception
457 Right result -> return result
462 -- ---------------------------------------------------------------------------
467 In a standalone GHC program, only the main thread is
468 required to terminate in order for the process to terminate.
469 Thus all other forked threads will simply terminate at the same
470 time as the main thread (the terminology for this kind of
471 behaviour is \"daemonic threads\").
473 If you want the program to wait for child threads to
474 finish before exiting, you need to program this yourself. A
475 simple mechanism is to have each child thread write to an
476 'MVar' when it completes, and have the main
477 thread wait on all the 'MVar's before
480 > myForkIO :: IO () -> IO (MVar ())
482 > mvar \<- newEmptyMVar
483 > forkIO (io \`finally\` putMVar mvar ())
486 Note that we use 'finally' from the
487 "Control.Exception" module to make sure that the
488 'MVar' is written to even if the thread dies or
489 is killed for some reason.
491 A better method is to keep a global list of all child
492 threads which we should wait for at the end of the program:
494 > children :: MVar [MVar ()]
495 > children = unsafePerformIO (newMVar [])
497 > waitForChildren :: IO ()
498 > waitForChildren = do
499 > (mvar:mvars) \<- takeMVar children
500 > putMVar children mvars
504 > forkChild :: IO () -> IO ()
506 > mvar \<- newEmptyMVar
507 > forkIO (p \`finally\` putMVar mvar ())
508 > childs \<- takeMVar children
509 > putMVar children (mvar:childs)
511 > later = flip finally
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 pathalogical 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 :-)