1 /* ---------------------------------------------------------------------------
3 * (c) The GHC Team, 1998-2004
7 * Different GHC ways use this scheduler quite differently (see comments below)
8 * Here is the global picture:
10 * WAY Name CPP flag What's it for
11 * --------------------------------------
12 * mp GUM PARALLEL_HASKELL Parallel execution on a distrib. memory machine
13 * s SMP SMP Parallel execution on a shared memory machine
14 * mg GranSim GRAN Simulation of parallel execution
15 * md GUM/GdH DIST Distributed execution (based on GUM)
17 * --------------------------------------------------------------------------*/
20 * Version with support for distributed memory parallelism aka GUM (WAY=mp):
22 The main scheduling loop in GUM iterates until a finish message is received.
23 In that case a global flag @receivedFinish@ is set and this instance of
24 the RTS shuts down. See ghc/rts/parallel/HLComms.c:processMessages()
25 for the handling of incoming messages, such as PP_FINISH.
26 Note that in the parallel case we have a system manager that coordinates
27 different PEs, each of which are running one instance of the RTS.
28 See ghc/rts/parallel/SysMan.c for the main routine of the parallel program.
29 From this routine processes executing ghc/rts/Main.c are spawned. -- HWL
31 * Version with support for simulating parallel execution aka GranSim (WAY=mg):
33 The main scheduling code in GranSim is quite different from that in std
34 (concurrent) Haskell: while concurrent Haskell just iterates over the
35 threads in the runnable queue, GranSim is event driven, i.e. it iterates
36 over the events in the global event queue. -- HWL
39 #include "PosixSource.h"
44 #include "BlockAlloc.h"
45 #include "OSThreads.h"
49 #define COMPILING_SCHEDULER
51 #include "StgMiscClosures.h"
52 #include "Interpreter.h"
53 #include "Exception.h"
61 #include "ThreadLabels.h"
62 #include "LdvProfile.h"
65 #include "Proftimer.h"
68 #if defined(GRAN) || defined(PARALLEL_HASKELL)
69 # include "GranSimRts.h"
71 # include "ParallelRts.h"
72 # include "Parallel.h"
73 # include "ParallelDebug.h"
78 #include "Capability.h"
81 #ifdef HAVE_SYS_TYPES_H
82 #include <sys/types.h>
96 // Turn off inlining when debugging - it obfuscates things
99 # define STATIC_INLINE static
103 #define USED_IN_THREADED_RTS
105 #define USED_IN_THREADED_RTS STG_UNUSED
108 #ifdef RTS_SUPPORTS_THREADS
109 #define USED_WHEN_RTS_SUPPORTS_THREADS
111 #define USED_WHEN_RTS_SUPPORTS_THREADS STG_UNUSED
114 /* Main thread queue.
115 * Locks required: sched_mutex.
117 StgMainThread *main_threads = NULL;
121 StgTSO* ActiveTSO = NULL; /* for assigning system costs; GranSim-Light only */
122 /* rtsTime TimeOfNextEvent, EndOfTimeSlice; now in GranSim.c */
125 In GranSim we have a runnable and a blocked queue for each processor.
126 In order to minimise code changes new arrays run_queue_hds/tls
127 are created. run_queue_hd is then a short cut (macro) for
128 run_queue_hds[CurrentProc] (see GranSim.h).
131 StgTSO *run_queue_hds[MAX_PROC], *run_queue_tls[MAX_PROC];
132 StgTSO *blocked_queue_hds[MAX_PROC], *blocked_queue_tls[MAX_PROC];
133 StgTSO *ccalling_threadss[MAX_PROC];
134 /* We use the same global list of threads (all_threads) in GranSim as in
135 the std RTS (i.e. we are cheating). However, we don't use this list in
136 the GranSim specific code at the moment (so we are only potentially
142 * Locks required: sched_mutex.
144 StgTSO *run_queue_hd = NULL;
145 StgTSO *run_queue_tl = NULL;
146 StgTSO *blocked_queue_hd = NULL;
147 StgTSO *blocked_queue_tl = NULL;
148 StgTSO *blackhole_queue = NULL;
149 StgTSO *sleeping_queue = NULL; /* perhaps replace with a hash table? */
153 /* The blackhole_queue should be checked for threads to wake up. See
154 * Schedule.h for more thorough comment.
156 rtsBool blackholes_need_checking = rtsFalse;
158 /* Linked list of all threads.
159 * Used for detecting garbage collected threads.
161 StgTSO *all_threads = NULL;
163 /* When a thread performs a safe C call (_ccall_GC, using old
164 * terminology), it gets put on the suspended_ccalling_threads
165 * list. Used by the garbage collector.
167 static StgTSO *suspended_ccalling_threads;
169 /* KH: The following two flags are shared memory locations. There is no need
170 to lock them, since they are only unset at the end of a scheduler
174 /* flag set by signal handler to precipitate a context switch */
175 int context_switch = 0;
177 /* if this flag is set as well, give up execution */
178 rtsBool interrupted = rtsFalse;
180 /* Next thread ID to allocate.
181 * Locks required: thread_id_mutex
183 static StgThreadID next_thread_id = 1;
186 * Pointers to the state of the current thread.
187 * Rule of thumb: if CurrentTSO != NULL, then we're running a Haskell
188 * thread. If CurrentTSO == NULL, then we're at the scheduler level.
191 /* The smallest stack size that makes any sense is:
192 * RESERVED_STACK_WORDS (so we can get back from the stack overflow)
193 * + sizeofW(StgStopFrame) (the stg_stop_thread_info frame)
194 * + 1 (the closure to enter)
196 * + 1 (spare slot req'd by stg_ap_v_ret)
198 * A thread with this stack will bomb immediately with a stack
199 * overflow, which will increase its stack size.
202 #define MIN_STACK_WORDS (RESERVED_STACK_WORDS + sizeofW(StgStopFrame) + 3)
209 /* This is used in `TSO.h' and gcc 2.96 insists that this variable actually
210 * exists - earlier gccs apparently didn't.
216 * Set to TRUE when entering a shutdown state (via shutdownHaskellAndExit()) --
217 * in an MT setting, needed to signal that a worker thread shouldn't hang around
218 * in the scheduler when it is out of work.
220 static rtsBool shutting_down_scheduler = rtsFalse;
222 #if defined(RTS_SUPPORTS_THREADS)
223 /* ToDo: carefully document the invariants that go together
224 * with these synchronisation objects.
226 Mutex sched_mutex = INIT_MUTEX_VAR;
227 Mutex term_mutex = INIT_MUTEX_VAR;
229 #endif /* RTS_SUPPORTS_THREADS */
231 #if defined(PARALLEL_HASKELL)
233 rtsTime TimeOfLastYield;
234 rtsBool emitSchedule = rtsTrue;
238 static char *whatNext_strs[] = {
248 /* -----------------------------------------------------------------------------
249 * static function prototypes
250 * -------------------------------------------------------------------------- */
252 #if defined(RTS_SUPPORTS_THREADS)
253 static void taskStart(void);
256 static void schedule( StgMainThread *mainThread USED_WHEN_RTS_SUPPORTS_THREADS,
257 Capability *initialCapability );
260 // These function all encapsulate parts of the scheduler loop, and are
261 // abstracted only to make the structure and control flow of the
262 // scheduler clearer.
264 static void schedulePreLoop(void);
265 static void scheduleStartSignalHandlers(void);
266 static void scheduleCheckBlockedThreads(void);
267 static void scheduleCheckBlackHoles(void);
268 static void scheduleDetectDeadlock(void);
270 static StgTSO *scheduleProcessEvent(rtsEvent *event);
272 #if defined(PARALLEL_HASKELL)
273 static StgTSO *scheduleSendPendingMessages(void);
274 static void scheduleActivateSpark(void);
275 static rtsBool scheduleGetRemoteWork(rtsBool *receivedFinish);
277 #if defined(PAR) || defined(GRAN)
278 static void scheduleGranParReport(void);
280 static void schedulePostRunThread(void);
281 static rtsBool scheduleHandleHeapOverflow( Capability *cap, StgTSO *t );
282 static void scheduleHandleStackOverflow( StgTSO *t);
283 static rtsBool scheduleHandleYield( StgTSO *t, nat prev_what_next );
284 static void scheduleHandleThreadBlocked( StgTSO *t );
285 static rtsBool scheduleHandleThreadFinished( StgMainThread *mainThread,
286 Capability *cap, StgTSO *t );
287 static rtsBool scheduleDoHeapProfile(rtsBool ready_to_gc);
288 static void scheduleDoGC(Capability *cap);
290 static void unblockThread(StgTSO *tso);
291 static rtsBool checkBlackHoles(void);
292 static SchedulerStatus waitThread_(/*out*/StgMainThread* m,
293 Capability *initialCapability
295 static void scheduleThread_ (StgTSO* tso);
296 static void AllRoots(evac_fn evac);
298 static StgTSO *threadStackOverflow(StgTSO *tso);
300 static void raiseAsync_(StgTSO *tso, StgClosure *exception,
301 rtsBool stop_at_atomically);
303 static void printThreadBlockage(StgTSO *tso);
304 static void printThreadStatus(StgTSO *tso);
306 #if defined(PARALLEL_HASKELL)
307 StgTSO * createSparkThread(rtsSpark spark);
308 StgTSO * activateSpark (rtsSpark spark);
311 /* ----------------------------------------------------------------------------
313 * ------------------------------------------------------------------------- */
315 #if defined(RTS_SUPPORTS_THREADS)
316 static rtsBool startingWorkerThread = rtsFalse;
321 ACQUIRE_LOCK(&sched_mutex);
322 startingWorkerThread = rtsFalse;
325 RELEASE_LOCK(&sched_mutex);
329 startSchedulerTaskIfNecessary(void)
331 if ( !EMPTY_RUN_QUEUE()
332 && !shutting_down_scheduler // not if we're shutting down
333 && !startingWorkerThread)
335 // we don't want to start another worker thread
336 // just because the last one hasn't yet reached the
337 // "waiting for capability" state
338 startingWorkerThread = rtsTrue;
339 if (!maybeStartNewWorker(taskStart)) {
340 startingWorkerThread = rtsFalse;
346 /* -----------------------------------------------------------------------------
347 * Putting a thread on the run queue: different scheduling policies
348 * -------------------------------------------------------------------------- */
351 addToRunQueue( StgTSO *t )
353 #if defined(PARALLEL_HASKELL)
354 if (RtsFlags.ParFlags.doFairScheduling) {
355 // this does round-robin scheduling; good for concurrency
356 APPEND_TO_RUN_QUEUE(t);
358 // this does unfair scheduling; good for parallelism
359 PUSH_ON_RUN_QUEUE(t);
362 // this does round-robin scheduling; good for concurrency
363 APPEND_TO_RUN_QUEUE(t);
367 /* ---------------------------------------------------------------------------
368 Main scheduling loop.
370 We use round-robin scheduling, each thread returning to the
371 scheduler loop when one of these conditions is detected:
374 * timer expires (thread yields)
379 Locking notes: we acquire the scheduler lock once at the beginning
380 of the scheduler loop, and release it when
382 * running a thread, or
383 * waiting for work, or
384 * waiting for a GC to complete.
387 In a GranSim setup this loop iterates over the global event queue.
388 This revolves around the global event queue, which determines what
389 to do next. Therefore, it's more complicated than either the
390 concurrent or the parallel (GUM) setup.
393 GUM iterates over incoming messages.
394 It starts with nothing to do (thus CurrentTSO == END_TSO_QUEUE),
395 and sends out a fish whenever it has nothing to do; in-between
396 doing the actual reductions (shared code below) it processes the
397 incoming messages and deals with delayed operations
398 (see PendingFetches).
399 This is not the ugliest code you could imagine, but it's bloody close.
401 ------------------------------------------------------------------------ */
404 schedule( StgMainThread *mainThread USED_WHEN_RTS_SUPPORTS_THREADS,
405 Capability *initialCapability )
409 StgThreadReturnCode ret;
412 #elif defined(PARALLEL_HASKELL)
415 rtsBool receivedFinish = rtsFalse;
417 nat tp_size, sp_size; // stats only
423 // Pre-condition: sched_mutex is held.
424 // We might have a capability, passed in as initialCapability.
425 cap = initialCapability;
427 #if !defined(RTS_SUPPORTS_THREADS)
428 // simply initialise it in the non-threaded case
429 grabCapability(&cap);
433 sched_belch("### NEW SCHEDULER LOOP (main thr: %p, cap: %p)",
434 mainThread, initialCapability);
439 // -----------------------------------------------------------
440 // Scheduler loop starts here:
442 #if defined(PARALLEL_HASKELL)
443 #define TERMINATION_CONDITION (!receivedFinish)
445 #define TERMINATION_CONDITION ((event = get_next_event()) != (rtsEvent*)NULL)
447 #define TERMINATION_CONDITION rtsTrue
450 while (TERMINATION_CONDITION) {
453 /* Choose the processor with the next event */
454 CurrentProc = event->proc;
455 CurrentTSO = event->tso;
458 IF_DEBUG(scheduler, printAllThreads());
460 #if defined(RTS_SUPPORTS_THREADS)
461 // Yield the capability to higher-priority tasks if necessary.
464 yieldCapability(&cap);
467 // If we do not currently hold a capability, we wait for one
470 waitForCapability(&sched_mutex, &cap,
471 mainThread ? &mainThread->bound_thread_cond : NULL);
474 // We now have a capability...
477 // Check whether we have re-entered the RTS from Haskell without
478 // going via suspendThread()/resumeThread (i.e. a 'safe' foreign
480 if (cap->r.rInHaskell) {
481 errorBelch("schedule: re-entered unsafely.\n"
482 " Perhaps a 'foreign import unsafe' should be 'safe'?");
487 // Test for interruption. If interrupted==rtsTrue, then either
488 // we received a keyboard interrupt (^C), or the scheduler is
489 // trying to shut down all the tasks (shutting_down_scheduler) in
493 if (shutting_down_scheduler) {
494 IF_DEBUG(scheduler, sched_belch("shutting down"));
495 releaseCapability(cap);
497 mainThread->stat = Interrupted;
498 mainThread->ret = NULL;
502 IF_DEBUG(scheduler, sched_belch("interrupted"));
507 #if defined(not_yet) && defined(SMP)
509 // Top up the run queue from our spark pool. We try to make the
510 // number of threads in the run queue equal to the number of
511 // free capabilities.
515 if (EMPTY_RUN_QUEUE()) {
516 spark = findSpark(rtsFalse);
518 break; /* no more sparks in the pool */
520 createSparkThread(spark);
522 sched_belch("==^^ turning spark of closure %p into a thread",
523 (StgClosure *)spark));
529 scheduleStartSignalHandlers();
531 // Only check the black holes here if we've nothing else to do.
532 // During normal execution, the black hole list only gets checked
533 // at GC time, to avoid repeatedly traversing this possibly long
534 // list each time around the scheduler.
535 if (EMPTY_RUN_QUEUE()) { scheduleCheckBlackHoles(); }
537 scheduleCheckBlockedThreads();
539 scheduleDetectDeadlock();
541 // Normally, the only way we can get here with no threads to
542 // run is if a keyboard interrupt received during
543 // scheduleCheckBlockedThreads() or scheduleDetectDeadlock().
544 // Additionally, it is not fatal for the
545 // threaded RTS to reach here with no threads to run.
547 // win32: might be here due to awaitEvent() being abandoned
548 // as a result of a console event having been delivered.
549 if ( EMPTY_RUN_QUEUE() ) {
550 #if !defined(RTS_SUPPORTS_THREADS) && !defined(mingw32_HOST_OS)
553 continue; // nothing to do
556 #if defined(PARALLEL_HASKELL)
557 scheduleSendPendingMessages();
558 if (EMPTY_RUN_QUEUE() && scheduleActivateSpark())
562 ASSERT(next_fish_to_send_at==0); // i.e. no delayed fishes left!
565 /* If we still have no work we need to send a FISH to get a spark
567 if (EMPTY_RUN_QUEUE()) {
568 if (!scheduleGetRemoteWork(&receivedFinish)) continue;
569 ASSERT(rtsFalse); // should not happen at the moment
571 // from here: non-empty run queue.
572 // TODO: merge above case with this, only one call processMessages() !
573 if (PacketsWaiting()) { /* process incoming messages, if
574 any pending... only in else
575 because getRemoteWork waits for
577 receivedFinish = processMessages();
582 scheduleProcessEvent(event);
586 // Get a thread to run
588 ASSERT(run_queue_hd != END_TSO_QUEUE);
591 #if defined(GRAN) || defined(PAR)
592 scheduleGranParReport(); // some kind of debuging output
594 // Sanity check the thread we're about to run. This can be
595 // expensive if there is lots of thread switching going on...
596 IF_DEBUG(sanity,checkTSO(t));
599 #if defined(RTS_SUPPORTS_THREADS)
600 // Check whether we can run this thread in the current task.
601 // If not, we have to pass our capability to the right task.
603 StgMainThread *m = t->main;
610 sched_belch("### Running thread %d in bound thread", t->id));
611 // yes, the Haskell thread is bound to the current native thread
616 sched_belch("### thread %d bound to another OS thread", t->id));
617 // no, bound to a different Haskell thread: pass to that thread
618 PUSH_ON_RUN_QUEUE(t);
619 passCapability(&m->bound_thread_cond);
625 if(mainThread != NULL)
626 // The thread we want to run is bound.
629 sched_belch("### this OS thread cannot run thread %d", t->id));
630 // no, the current native thread is bound to a different
631 // Haskell thread, so pass it to any worker thread
632 PUSH_ON_RUN_QUEUE(t);
633 passCapabilityToWorker();
640 cap->r.rCurrentTSO = t;
642 /* context switches are now initiated by the timer signal, unless
643 * the user specified "context switch as often as possible", with
646 if ((RtsFlags.ConcFlags.ctxtSwitchTicks == 0
647 && (run_queue_hd != END_TSO_QUEUE
648 || blocked_queue_hd != END_TSO_QUEUE
649 || sleeping_queue != END_TSO_QUEUE)))
654 RELEASE_LOCK(&sched_mutex);
656 IF_DEBUG(scheduler, sched_belch("-->> running thread %ld %s ...",
657 (long)t->id, whatNext_strs[t->what_next]));
659 #if defined(PROFILING)
660 startHeapProfTimer();
663 // ----------------------------------------------------------------------
664 // Run the current thread
666 prev_what_next = t->what_next;
668 errno = t->saved_errno;
669 cap->r.rInHaskell = rtsTrue;
671 switch (prev_what_next) {
675 /* Thread already finished, return to scheduler. */
676 ret = ThreadFinished;
680 ret = StgRun((StgFunPtr) stg_returnToStackTop, &cap->r);
683 case ThreadInterpret:
684 ret = interpretBCO(cap);
688 barf("schedule: invalid what_next field");
691 // We have run some Haskell code: there might be blackhole-blocked
692 // threads to wake up now.
693 if ( blackhole_queue != END_TSO_QUEUE ) {
694 blackholes_need_checking = rtsTrue;
697 cap->r.rInHaskell = rtsFalse;
699 // The TSO might have moved, eg. if it re-entered the RTS and a GC
700 // happened. So find the new location:
701 t = cap->r.rCurrentTSO;
703 // And save the current errno in this thread.
704 t->saved_errno = errno;
706 // ----------------------------------------------------------------------
708 /* Costs for the scheduler are assigned to CCS_SYSTEM */
709 #if defined(PROFILING)
714 ACQUIRE_LOCK(&sched_mutex);
716 #if defined(RTS_SUPPORTS_THREADS)
717 IF_DEBUG(scheduler,debugBelch("sched (task %p): ", osThreadId()););
718 #elif !defined(GRAN) && !defined(PARALLEL_HASKELL)
719 IF_DEBUG(scheduler,debugBelch("sched: "););
722 schedulePostRunThread();
724 ready_to_gc = rtsFalse;
728 ready_to_gc = scheduleHandleHeapOverflow(cap,t);
732 scheduleHandleStackOverflow(t);
736 if (scheduleHandleYield(t, prev_what_next)) {
737 // shortcut for switching between compiler/interpreter:
743 scheduleHandleThreadBlocked(t);
748 if (scheduleHandleThreadFinished(mainThread, cap, t)) return;;
752 barf("schedule: invalid thread return code %d", (int)ret);
755 if (scheduleDoHeapProfile(ready_to_gc)) { ready_to_gc = rtsFalse; }
756 if (ready_to_gc) { scheduleDoGC(cap); }
757 } /* end of while() */
759 IF_PAR_DEBUG(verbose,
760 debugBelch("== Leaving schedule() after having received Finish\n"));
763 /* ----------------------------------------------------------------------------
764 * Setting up the scheduler loop
765 * ASSUMES: sched_mutex
766 * ------------------------------------------------------------------------- */
769 schedulePreLoop(void)
772 /* set up first event to get things going */
773 /* ToDo: assign costs for system setup and init MainTSO ! */
774 new_event(CurrentProc, CurrentProc, CurrentTime[CurrentProc],
776 CurrentTSO, (StgClosure*)NULL, (rtsSpark*)NULL);
779 debugBelch("GRAN: Init CurrentTSO (in schedule) = %p\n",
781 G_TSO(CurrentTSO, 5));
783 if (RtsFlags.GranFlags.Light) {
784 /* Save current time; GranSim Light only */
785 CurrentTSO->gran.clock = CurrentTime[CurrentProc];
790 /* ----------------------------------------------------------------------------
791 * Start any pending signal handlers
792 * ASSUMES: sched_mutex
793 * ------------------------------------------------------------------------- */
796 scheduleStartSignalHandlers(void)
798 #if defined(RTS_USER_SIGNALS) && !defined(RTS_SUPPORTS_THREADS)
799 if (signals_pending()) {
800 RELEASE_LOCK(&sched_mutex); /* ToDo: kill */
801 startSignalHandlers();
802 ACQUIRE_LOCK(&sched_mutex);
807 /* ----------------------------------------------------------------------------
808 * Check for blocked threads that can be woken up.
809 * ASSUMES: sched_mutex
810 * ------------------------------------------------------------------------- */
813 scheduleCheckBlockedThreads(void)
816 // Check whether any waiting threads need to be woken up. If the
817 // run queue is empty, and there are no other tasks running, we
818 // can wait indefinitely for something to happen.
820 if ( !EMPTY_QUEUE(blocked_queue_hd) || !EMPTY_QUEUE(sleeping_queue) )
822 #if defined(RTS_SUPPORTS_THREADS)
823 // We shouldn't be here...
824 barf("schedule: awaitEvent() in threaded RTS");
826 awaitEvent( EMPTY_RUN_QUEUE() && !blackholes_need_checking );
831 /* ----------------------------------------------------------------------------
832 * Check for threads blocked on BLACKHOLEs that can be woken up
833 * ASSUMES: sched_mutex
834 * ------------------------------------------------------------------------- */
836 scheduleCheckBlackHoles( void )
838 if ( blackholes_need_checking )
841 blackholes_need_checking = rtsFalse;
845 /* ----------------------------------------------------------------------------
846 * Detect deadlock conditions and attempt to resolve them.
847 * ASSUMES: sched_mutex
848 * ------------------------------------------------------------------------- */
851 scheduleDetectDeadlock(void)
854 * Detect deadlock: when we have no threads to run, there are no
855 * threads blocked, waiting for I/O, or sleeping, and all the
856 * other tasks are waiting for work, we must have a deadlock of
859 if ( EMPTY_THREAD_QUEUES() )
861 #if !defined(PARALLEL_HASKELL) && !defined(RTS_SUPPORTS_THREADS)
862 IF_DEBUG(scheduler, sched_belch("deadlocked, forcing major GC..."));
864 // Garbage collection can release some new threads due to
865 // either (a) finalizers or (b) threads resurrected because
866 // they are unreachable and will therefore be sent an
867 // exception. Any threads thus released will be immediately
869 GarbageCollect(GetRoots,rtsTrue);
870 if ( !EMPTY_RUN_QUEUE() ) return;
872 #if defined(RTS_USER_SIGNALS)
873 /* If we have user-installed signal handlers, then wait
874 * for signals to arrive rather then bombing out with a
877 if ( anyUserHandlers() ) {
879 sched_belch("still deadlocked, waiting for signals..."));
883 if (signals_pending()) {
884 RELEASE_LOCK(&sched_mutex);
885 startSignalHandlers();
886 ACQUIRE_LOCK(&sched_mutex);
889 // either we have threads to run, or we were interrupted:
890 ASSERT(!EMPTY_RUN_QUEUE() || interrupted);
894 /* Probably a real deadlock. Send the current main thread the
895 * Deadlock exception (or in the SMP build, send *all* main
896 * threads the deadlock exception, since none of them can make
902 switch (m->tso->why_blocked) {
903 case BlockedOnBlackHole:
904 case BlockedOnException:
906 raiseAsync(m->tso, (StgClosure *)NonTermination_closure);
909 barf("deadlock: main thread blocked in a strange way");
913 #elif defined(RTS_SUPPORTS_THREADS)
914 // ToDo: add deadlock detection in threaded RTS
915 #elif defined(PARALLEL_HASKELL)
916 // ToDo: add deadlock detection in GUM (similar to SMP) -- HWL
921 /* ----------------------------------------------------------------------------
922 * Process an event (GRAN only)
923 * ------------------------------------------------------------------------- */
927 scheduleProcessEvent(rtsEvent *event)
931 if (RtsFlags.GranFlags.Light)
932 GranSimLight_enter_system(event, &ActiveTSO); // adjust ActiveTSO etc
934 /* adjust time based on time-stamp */
935 if (event->time > CurrentTime[CurrentProc] &&
936 event->evttype != ContinueThread)
937 CurrentTime[CurrentProc] = event->time;
939 /* Deal with the idle PEs (may issue FindWork or MoveSpark events) */
940 if (!RtsFlags.GranFlags.Light)
943 IF_DEBUG(gran, debugBelch("GRAN: switch by event-type\n"));
945 /* main event dispatcher in GranSim */
946 switch (event->evttype) {
947 /* Should just be continuing execution */
949 IF_DEBUG(gran, debugBelch("GRAN: doing ContinueThread\n"));
950 /* ToDo: check assertion
951 ASSERT(run_queue_hd != (StgTSO*)NULL &&
952 run_queue_hd != END_TSO_QUEUE);
954 /* Ignore ContinueThreads for fetching threads (if synchr comm) */
955 if (!RtsFlags.GranFlags.DoAsyncFetch &&
956 procStatus[CurrentProc]==Fetching) {
957 debugBelch("ghuH: Spurious ContinueThread while Fetching ignored; TSO %d (%p) [PE %d]\n",
958 CurrentTSO->id, CurrentTSO, CurrentProc);
961 /* Ignore ContinueThreads for completed threads */
962 if (CurrentTSO->what_next == ThreadComplete) {
963 debugBelch("ghuH: found a ContinueThread event for completed thread %d (%p) [PE %d] (ignoring ContinueThread)\n",
964 CurrentTSO->id, CurrentTSO, CurrentProc);
967 /* Ignore ContinueThreads for threads that are being migrated */
968 if (PROCS(CurrentTSO)==Nowhere) {
969 debugBelch("ghuH: trying to run the migrating TSO %d (%p) [PE %d] (ignoring ContinueThread)\n",
970 CurrentTSO->id, CurrentTSO, CurrentProc);
973 /* The thread should be at the beginning of the run queue */
974 if (CurrentTSO!=run_queue_hds[CurrentProc]) {
975 debugBelch("ghuH: TSO %d (%p) [PE %d] is not at the start of the run_queue when doing a ContinueThread\n",
976 CurrentTSO->id, CurrentTSO, CurrentProc);
977 break; // run the thread anyway
980 new_event(proc, proc, CurrentTime[proc],
982 (StgTSO*)NULL, (StgClosure*)NULL, (rtsSpark*)NULL);
984 */ /* Catches superfluous CONTINUEs -- should be unnecessary */
985 break; // now actually run the thread; DaH Qu'vam yImuHbej
988 do_the_fetchnode(event);
989 goto next_thread; /* handle next event in event queue */
992 do_the_globalblock(event);
993 goto next_thread; /* handle next event in event queue */
996 do_the_fetchreply(event);
997 goto next_thread; /* handle next event in event queue */
999 case UnblockThread: /* Move from the blocked queue to the tail of */
1000 do_the_unblock(event);
1001 goto next_thread; /* handle next event in event queue */
1003 case ResumeThread: /* Move from the blocked queue to the tail of */
1004 /* the runnable queue ( i.e. Qu' SImqa'lu') */
1005 event->tso->gran.blocktime +=
1006 CurrentTime[CurrentProc] - event->tso->gran.blockedat;
1007 do_the_startthread(event);
1008 goto next_thread; /* handle next event in event queue */
1011 do_the_startthread(event);
1012 goto next_thread; /* handle next event in event queue */
1015 do_the_movethread(event);
1016 goto next_thread; /* handle next event in event queue */
1019 do_the_movespark(event);
1020 goto next_thread; /* handle next event in event queue */
1023 do_the_findwork(event);
1024 goto next_thread; /* handle next event in event queue */
1027 barf("Illegal event type %u\n", event->evttype);
1030 /* This point was scheduler_loop in the old RTS */
1032 IF_DEBUG(gran, debugBelch("GRAN: after main switch\n"));
1034 TimeOfLastEvent = CurrentTime[CurrentProc];
1035 TimeOfNextEvent = get_time_of_next_event();
1036 IgnoreEvents=(TimeOfNextEvent==0); // HWL HACK
1037 // CurrentTSO = ThreadQueueHd;
1039 IF_DEBUG(gran, debugBelch("GRAN: time of next event is: %ld\n",
1042 if (RtsFlags.GranFlags.Light)
1043 GranSimLight_leave_system(event, &ActiveTSO);
1045 EndOfTimeSlice = CurrentTime[CurrentProc]+RtsFlags.GranFlags.time_slice;
1048 debugBelch("GRAN: end of time-slice is %#lx\n", EndOfTimeSlice));
1050 /* in a GranSim setup the TSO stays on the run queue */
1052 /* Take a thread from the run queue. */
1053 POP_RUN_QUEUE(t); // take_off_run_queue(t);
1056 debugBelch("GRAN: About to run current thread, which is\n");
1059 context_switch = 0; // turned on via GranYield, checking events and time slice
1062 DumpGranEvent(GR_SCHEDULE, t));
1064 procStatus[CurrentProc] = Busy;
1068 /* ----------------------------------------------------------------------------
1069 * Send pending messages (PARALLEL_HASKELL only)
1070 * ------------------------------------------------------------------------- */
1072 #if defined(PARALLEL_HASKELL)
1074 scheduleSendPendingMessages(void)
1080 # if defined(PAR) // global Mem.Mgmt., omit for now
1081 if (PendingFetches != END_BF_QUEUE) {
1086 if (RtsFlags.ParFlags.BufferTime) {
1087 // if we use message buffering, we must send away all message
1088 // packets which have become too old...
1094 /* ----------------------------------------------------------------------------
1095 * Activate spark threads (PARALLEL_HASKELL only)
1096 * ------------------------------------------------------------------------- */
1098 #if defined(PARALLEL_HASKELL)
1100 scheduleActivateSpark(void)
1103 ASSERT(EMPTY_RUN_QUEUE());
1104 /* We get here if the run queue is empty and want some work.
1105 We try to turn a spark into a thread, and add it to the run queue,
1106 from where it will be picked up in the next iteration of the scheduler
1110 /* :-[ no local threads => look out for local sparks */
1111 /* the spark pool for the current PE */
1112 pool = &(cap.r.rSparks); // JB: cap = (old) MainCap
1113 if (advisory_thread_count < RtsFlags.ParFlags.maxThreads &&
1114 pool->hd < pool->tl) {
1116 * ToDo: add GC code check that we really have enough heap afterwards!!
1118 * If we're here (no runnable threads) and we have pending
1119 * sparks, we must have a space problem. Get enough space
1120 * to turn one of those pending sparks into a
1124 spark = findSpark(rtsFalse); /* get a spark */
1125 if (spark != (rtsSpark) NULL) {
1126 tso = createThreadFromSpark(spark); /* turn the spark into a thread */
1127 IF_PAR_DEBUG(fish, // schedule,
1128 debugBelch("==== schedule: Created TSO %d (%p); %d threads active\n",
1129 tso->id, tso, advisory_thread_count));
1131 if (tso==END_TSO_QUEUE) { /* failed to activate spark->back to loop */
1132 IF_PAR_DEBUG(fish, // schedule,
1133 debugBelch("==^^ failed to create thread from spark @ %lx\n",
1135 return rtsFalse; /* failed to generate a thread */
1136 } /* otherwise fall through & pick-up new tso */
1138 IF_PAR_DEBUG(fish, // schedule,
1139 debugBelch("==^^ no local sparks (spark pool contains only NFs: %d)\n",
1140 spark_queue_len(pool)));
1141 return rtsFalse; /* failed to generate a thread */
1143 return rtsTrue; /* success in generating a thread */
1144 } else { /* no more threads permitted or pool empty */
1145 return rtsFalse; /* failed to generateThread */
1148 tso = NULL; // avoid compiler warning only
1149 return rtsFalse; /* dummy in non-PAR setup */
1152 #endif // PARALLEL_HASKELL
1154 /* ----------------------------------------------------------------------------
1155 * Get work from a remote node (PARALLEL_HASKELL only)
1156 * ------------------------------------------------------------------------- */
1158 #if defined(PARALLEL_HASKELL)
1160 scheduleGetRemoteWork(rtsBool *receivedFinish)
1162 ASSERT(EMPTY_RUN_QUEUE());
1164 if (RtsFlags.ParFlags.BufferTime) {
1165 IF_PAR_DEBUG(verbose,
1166 debugBelch("...send all pending data,"));
1169 for (i=1; i<=nPEs; i++)
1170 sendImmediately(i); // send all messages away immediately
1174 //++EDEN++ idle() , i.e. send all buffers, wait for work
1175 // suppress fishing in EDEN... just look for incoming messages
1176 // (blocking receive)
1177 IF_PAR_DEBUG(verbose,
1178 debugBelch("...wait for incoming messages...\n"));
1179 *receivedFinish = processMessages(); // blocking receive...
1181 // and reenter scheduling loop after having received something
1182 // (return rtsFalse below)
1184 # else /* activate SPARKS machinery */
1185 /* We get here, if we have no work, tried to activate a local spark, but still
1186 have no work. We try to get a remote spark, by sending a FISH message.
1187 Thread migration should be added here, and triggered when a sequence of
1188 fishes returns without work. */
1189 delay = (RtsFlags.ParFlags.fishDelay!=0ll ? RtsFlags.ParFlags.fishDelay : 0ll);
1191 /* =8-[ no local sparks => look for work on other PEs */
1193 * We really have absolutely no work. Send out a fish
1194 * (there may be some out there already), and wait for
1195 * something to arrive. We clearly can't run any threads
1196 * until a SCHEDULE or RESUME arrives, and so that's what
1197 * we're hoping to see. (Of course, we still have to
1198 * respond to other types of messages.)
1200 rtsTime now = msTime() /*CURRENT_TIME*/;
1201 IF_PAR_DEBUG(verbose,
1202 debugBelch("-- now=%ld\n", now));
1203 IF_PAR_DEBUG(fish, // verbose,
1204 if (outstandingFishes < RtsFlags.ParFlags.maxFishes &&
1205 (last_fish_arrived_at!=0 &&
1206 last_fish_arrived_at+delay > now)) {
1207 debugBelch("--$$ <%llu> delaying FISH until %llu (last fish %llu, delay %llu)\n",
1208 now, last_fish_arrived_at+delay,
1209 last_fish_arrived_at,
1213 if (outstandingFishes < RtsFlags.ParFlags.maxFishes &&
1214 advisory_thread_count < RtsFlags.ParFlags.maxThreads) { // send a FISH, but when?
1215 if (last_fish_arrived_at==0 ||
1216 (last_fish_arrived_at+delay <= now)) { // send FISH now!
1217 /* outstandingFishes is set in sendFish, processFish;
1218 avoid flooding system with fishes via delay */
1219 next_fish_to_send_at = 0;
1221 /* ToDo: this should be done in the main scheduling loop to avoid the
1222 busy wait here; not so bad if fish delay is very small */
1223 int iq = 0; // DEBUGGING -- HWL
1224 next_fish_to_send_at = last_fish_arrived_at+delay; // remember when to send
1225 /* send a fish when ready, but process messages that arrive in the meantime */
1227 if (PacketsWaiting()) {
1229 *receivedFinish = processMessages();
1232 } while (!*receivedFinish || now<next_fish_to_send_at);
1233 // JB: This means the fish could become obsolete, if we receive
1234 // work. Better check for work again?
1235 // last line: while (!receivedFinish || !haveWork || now<...)
1236 // next line: if (receivedFinish || haveWork )
1238 if (*receivedFinish) // no need to send a FISH if we are finishing anyway
1239 return rtsFalse; // NB: this will leave scheduler loop
1240 // immediately after return!
1242 IF_PAR_DEBUG(fish, // verbose,
1243 debugBelch("--$$ <%llu> sent delayed fish (%d processMessages); active/total threads=%d/%d\n",now,iq,run_queue_len(),advisory_thread_count));
1247 // JB: IMHO, this should all be hidden inside sendFish(...)
1249 sendFish(pe, thisPE, NEW_FISH_AGE, NEW_FISH_HISTORY,
1252 // Global statistics: count no. of fishes
1253 if (RtsFlags.ParFlags.ParStats.Global &&
1254 RtsFlags.GcFlags.giveStats > NO_GC_STATS) {
1255 globalParStats.tot_fish_mess++;
1259 /* delayed fishes must have been sent by now! */
1260 next_fish_to_send_at = 0;
1263 *receivedFinish = processMessages();
1264 # endif /* SPARKS */
1267 /* NB: this function always returns rtsFalse, meaning the scheduler
1268 loop continues with the next iteration;
1270 return code means success in finding work; we enter this function
1271 if there is no local work, thus have to send a fish which takes
1272 time until it arrives with work; in the meantime we should process
1273 messages in the main loop;
1276 #endif // PARALLEL_HASKELL
1278 /* ----------------------------------------------------------------------------
1279 * PAR/GRAN: Report stats & debugging info(?)
1280 * ------------------------------------------------------------------------- */
1282 #if defined(PAR) || defined(GRAN)
1284 scheduleGranParReport(void)
1286 ASSERT(run_queue_hd != END_TSO_QUEUE);
1288 /* Take a thread from the run queue, if we have work */
1289 POP_RUN_QUEUE(t); // take_off_run_queue(END_TSO_QUEUE);
1291 /* If this TSO has got its outport closed in the meantime,
1292 * it mustn't be run. Instead, we have to clean it up as if it was finished.
1293 * It has to be marked as TH_DEAD for this purpose.
1294 * If it is TH_TERM instead, it is supposed to have finished in the normal way.
1296 JB: TODO: investigate wether state change field could be nuked
1297 entirely and replaced by the normal tso state (whatnext
1298 field). All we want to do is to kill tsos from outside.
1301 /* ToDo: write something to the log-file
1302 if (RTSflags.ParFlags.granSimStats && !sameThread)
1303 DumpGranEvent(GR_SCHEDULE, RunnableThreadsHd);
1307 /* the spark pool for the current PE */
1308 pool = &(cap.r.rSparks); // cap = (old) MainCap
1311 debugBelch("--=^ %d threads, %d sparks on [%#x]\n",
1312 run_queue_len(), spark_queue_len(pool), CURRENT_PROC));
1315 debugBelch("--=^ %d threads, %d sparks on [%#x]\n",
1316 run_queue_len(), spark_queue_len(pool), CURRENT_PROC));
1318 if (RtsFlags.ParFlags.ParStats.Full &&
1319 (t->par.sparkname != (StgInt)0) && // only log spark generated threads
1320 (emitSchedule || // forced emit
1321 (t && LastTSO && t->id != LastTSO->id))) {
1323 we are running a different TSO, so write a schedule event to log file
1324 NB: If we use fair scheduling we also have to write a deschedule
1325 event for LastTSO; with unfair scheduling we know that the
1326 previous tso has blocked whenever we switch to another tso, so
1327 we don't need it in GUM for now
1329 IF_PAR_DEBUG(fish, // schedule,
1330 debugBelch("____ scheduling spark generated thread %d (%lx) (%lx) via a forced emit\n",t->id,t,t->par.sparkname));
1332 DumpRawGranEvent(CURRENT_PROC, CURRENT_PROC,
1333 GR_SCHEDULE, t, (StgClosure *)NULL, 0, 0);
1334 emitSchedule = rtsFalse;
1339 /* ----------------------------------------------------------------------------
1340 * After running a thread...
1341 * ASSUMES: sched_mutex
1342 * ------------------------------------------------------------------------- */
1345 schedulePostRunThread(void)
1348 /* HACK 675: if the last thread didn't yield, make sure to print a
1349 SCHEDULE event to the log file when StgRunning the next thread, even
1350 if it is the same one as before */
1352 TimeOfLastYield = CURRENT_TIME;
1355 /* some statistics gathering in the parallel case */
1357 #if defined(GRAN) || defined(PAR) || defined(EDEN)
1361 IF_DEBUG(gran, DumpGranEvent(GR_DESCHEDULE, t));
1362 globalGranStats.tot_heapover++;
1364 globalParStats.tot_heapover++;
1371 DumpGranEvent(GR_DESCHEDULE, t));
1372 globalGranStats.tot_stackover++;
1375 // DumpGranEvent(GR_DESCHEDULE, t);
1376 globalParStats.tot_stackover++;
1380 case ThreadYielding:
1383 DumpGranEvent(GR_DESCHEDULE, t));
1384 globalGranStats.tot_yields++;
1387 // DumpGranEvent(GR_DESCHEDULE, t);
1388 globalParStats.tot_yields++;
1395 debugBelch("--<< thread %ld (%p; %s) stopped, blocking on node %p [PE %d] with BQ: ",
1396 t->id, t, whatNext_strs[t->what_next], t->block_info.closure,
1397 (t->block_info.closure==(StgClosure*)NULL ? 99 : where_is(t->block_info.closure)));
1398 if (t->block_info.closure!=(StgClosure*)NULL)
1399 print_bq(t->block_info.closure);
1402 // ??? needed; should emit block before
1404 DumpGranEvent(GR_DESCHEDULE, t));
1405 prune_eventq(t, (StgClosure *)NULL); // prune ContinueThreads for t
1408 ASSERT(procStatus[CurrentProc]==Busy ||
1409 ((procStatus[CurrentProc]==Fetching) &&
1410 (t->block_info.closure!=(StgClosure*)NULL)));
1411 if (run_queue_hds[CurrentProc] == END_TSO_QUEUE &&
1412 !(!RtsFlags.GranFlags.DoAsyncFetch &&
1413 procStatus[CurrentProc]==Fetching))
1414 procStatus[CurrentProc] = Idle;
1417 //++PAR++ blockThread() writes the event (change?)
1421 case ThreadFinished:
1425 barf("parGlobalStats: unknown return code");
1431 /* -----------------------------------------------------------------------------
1432 * Handle a thread that returned to the scheduler with ThreadHeepOverflow
1433 * ASSUMES: sched_mutex
1434 * -------------------------------------------------------------------------- */
1437 scheduleHandleHeapOverflow( Capability *cap, StgTSO *t )
1439 // did the task ask for a large block?
1440 if (cap->r.rHpAlloc > BLOCK_SIZE) {
1441 // if so, get one and push it on the front of the nursery.
1445 blocks = (lnat)BLOCK_ROUND_UP(cap->r.rHpAlloc) / BLOCK_SIZE;
1448 debugBelch("--<< thread %ld (%s) stopped: requesting a large block (size %ld)\n",
1449 (long)t->id, whatNext_strs[t->what_next], blocks));
1451 // don't do this if it would push us over the
1452 // alloc_blocks_lim limit; we'll GC first.
1453 if (alloc_blocks + blocks < alloc_blocks_lim) {
1455 alloc_blocks += blocks;
1456 bd = allocGroup( blocks );
1458 // link the new group into the list
1459 bd->link = cap->r.rCurrentNursery;
1460 bd->u.back = cap->r.rCurrentNursery->u.back;
1461 if (cap->r.rCurrentNursery->u.back != NULL) {
1462 cap->r.rCurrentNursery->u.back->link = bd;
1465 ASSERT(g0s0->blocks == cap->r.rCurrentNursery &&
1466 g0s0->blocks == cap->r.rNursery);
1469 cap->r.rNursery = bd;
1471 cap->r.rCurrentNursery->u.back = bd;
1473 // initialise it as a nursery block. We initialise the
1474 // step, gen_no, and flags field of *every* sub-block in
1475 // this large block, because this is easier than making
1476 // sure that we always find the block head of a large
1477 // block whenever we call Bdescr() (eg. evacuate() and
1478 // isAlive() in the GC would both have to do this, at
1482 for (x = bd; x < bd + blocks; x++) {
1490 // don't forget to update the block count in g0s0.
1491 g0s0->n_blocks += blocks;
1493 // This assert can be a killer if the app is doing lots
1494 // of large block allocations.
1495 ASSERT(countBlocks(g0s0->blocks) == g0s0->n_blocks);
1498 // now update the nursery to point to the new block
1499 cap->r.rCurrentNursery = bd;
1501 // we might be unlucky and have another thread get on the
1502 // run queue before us and steal the large block, but in that
1503 // case the thread will just end up requesting another large
1505 PUSH_ON_RUN_QUEUE(t);
1506 return rtsFalse; /* not actually GC'ing */
1510 /* make all the running tasks block on a condition variable,
1511 * maybe set context_switch and wait till they all pile in,
1512 * then have them wait on a GC condition variable.
1515 debugBelch("--<< thread %ld (%s) stopped: HeapOverflow\n",
1516 (long)t->id, whatNext_strs[t->what_next]));
1519 ASSERT(!is_on_queue(t,CurrentProc));
1520 #elif defined(PARALLEL_HASKELL)
1521 /* Currently we emit a DESCHEDULE event before GC in GUM.
1522 ToDo: either add separate event to distinguish SYSTEM time from rest
1523 or just nuke this DESCHEDULE (and the following SCHEDULE) */
1524 if (0 && RtsFlags.ParFlags.ParStats.Full) {
1525 DumpRawGranEvent(CURRENT_PROC, CURRENT_PROC,
1526 GR_DESCHEDULE, t, (StgClosure *)NULL, 0, 0);
1527 emitSchedule = rtsTrue;
1531 PUSH_ON_RUN_QUEUE(t);
1533 /* actual GC is done at the end of the while loop in schedule() */
1536 /* -----------------------------------------------------------------------------
1537 * Handle a thread that returned to the scheduler with ThreadStackOverflow
1538 * ASSUMES: sched_mutex
1539 * -------------------------------------------------------------------------- */
1542 scheduleHandleStackOverflow( StgTSO *t)
1544 IF_DEBUG(scheduler,debugBelch("--<< thread %ld (%s) stopped, StackOverflow\n",
1545 (long)t->id, whatNext_strs[t->what_next]));
1546 /* just adjust the stack for this thread, then pop it back
1551 /* enlarge the stack */
1552 StgTSO *new_t = threadStackOverflow(t);
1554 /* This TSO has moved, so update any pointers to it from the
1555 * main thread stack. It better not be on any other queues...
1556 * (it shouldn't be).
1558 if (t->main != NULL) {
1559 t->main->tso = new_t;
1561 PUSH_ON_RUN_QUEUE(new_t);
1565 /* -----------------------------------------------------------------------------
1566 * Handle a thread that returned to the scheduler with ThreadYielding
1567 * ASSUMES: sched_mutex
1568 * -------------------------------------------------------------------------- */
1571 scheduleHandleYield( StgTSO *t, nat prev_what_next )
1573 // Reset the context switch flag. We don't do this just before
1574 // running the thread, because that would mean we would lose ticks
1575 // during GC, which can lead to unfair scheduling (a thread hogs
1576 // the CPU because the tick always arrives during GC). This way
1577 // penalises threads that do a lot of allocation, but that seems
1578 // better than the alternative.
1581 /* put the thread back on the run queue. Then, if we're ready to
1582 * GC, check whether this is the last task to stop. If so, wake
1583 * up the GC thread. getThread will block during a GC until the
1587 if (t->what_next != prev_what_next) {
1588 debugBelch("--<< thread %ld (%s) stopped to switch evaluators\n",
1589 (long)t->id, whatNext_strs[t->what_next]);
1591 debugBelch("--<< thread %ld (%s) stopped, yielding\n",
1592 (long)t->id, whatNext_strs[t->what_next]);
1597 //debugBelch("&& Doing sanity check on yielding TSO %ld.", t->id);
1599 ASSERT(t->link == END_TSO_QUEUE);
1601 // Shortcut if we're just switching evaluators: don't bother
1602 // doing stack squeezing (which can be expensive), just run the
1604 if (t->what_next != prev_what_next) {
1611 ASSERT(!is_on_queue(t,CurrentProc));
1614 //debugBelch("&& Doing sanity check on all ThreadQueues (and their TSOs).");
1615 checkThreadQsSanity(rtsTrue));
1622 /* add a ContinueThread event to actually process the thread */
1623 new_event(CurrentProc, CurrentProc, CurrentTime[CurrentProc],
1625 t, (StgClosure*)NULL, (rtsSpark*)NULL);
1627 debugBelch("GRAN: eventq and runnableq after adding yielded thread to queue again:\n");
1634 /* -----------------------------------------------------------------------------
1635 * Handle a thread that returned to the scheduler with ThreadBlocked
1636 * ASSUMES: sched_mutex
1637 * -------------------------------------------------------------------------- */
1640 scheduleHandleThreadBlocked( StgTSO *t
1641 #if !defined(GRAN) && !defined(DEBUG)
1648 debugBelch("--<< thread %ld (%p; %s) stopped, blocking on node %p [PE %d] with BQ: \n",
1649 t->id, t, whatNext_strs[t->what_next], t->block_info.closure, (t->block_info.closure==(StgClosure*)NULL ? 99 : where_is(t->block_info.closure)));
1650 if (t->block_info.closure!=(StgClosure*)NULL) print_bq(t->block_info.closure));
1652 // ??? needed; should emit block before
1654 DumpGranEvent(GR_DESCHEDULE, t));
1655 prune_eventq(t, (StgClosure *)NULL); // prune ContinueThreads for t
1658 ASSERT(procStatus[CurrentProc]==Busy ||
1659 ((procStatus[CurrentProc]==Fetching) &&
1660 (t->block_info.closure!=(StgClosure*)NULL)));
1661 if (run_queue_hds[CurrentProc] == END_TSO_QUEUE &&
1662 !(!RtsFlags.GranFlags.DoAsyncFetch &&
1663 procStatus[CurrentProc]==Fetching))
1664 procStatus[CurrentProc] = Idle;
1668 debugBelch("--<< thread %ld (%p; %s) stopped, blocking on node %p with BQ: \n",
1669 t->id, t, whatNext_strs[t->what_next], t->block_info.closure));
1672 if (t->block_info.closure!=(StgClosure*)NULL)
1673 print_bq(t->block_info.closure));
1675 /* Send a fetch (if BlockedOnGA) and dump event to log file */
1678 /* whatever we schedule next, we must log that schedule */
1679 emitSchedule = rtsTrue;
1682 /* don't need to do anything. Either the thread is blocked on
1683 * I/O, in which case we'll have called addToBlockedQueue
1684 * previously, or it's blocked on an MVar or Blackhole, in which
1685 * case it'll be on the relevant queue already.
1687 ASSERT(t->why_blocked != NotBlocked);
1689 debugBelch("--<< thread %d (%s) stopped: ",
1690 t->id, whatNext_strs[t->what_next]);
1691 printThreadBlockage(t);
1694 /* Only for dumping event to log file
1695 ToDo: do I need this in GranSim, too?
1701 /* -----------------------------------------------------------------------------
1702 * Handle a thread that returned to the scheduler with ThreadFinished
1703 * ASSUMES: sched_mutex
1704 * -------------------------------------------------------------------------- */
1707 scheduleHandleThreadFinished( StgMainThread *mainThread
1708 USED_WHEN_RTS_SUPPORTS_THREADS,
1712 /* Need to check whether this was a main thread, and if so,
1713 * return with the return value.
1715 * We also end up here if the thread kills itself with an
1716 * uncaught exception, see Exception.cmm.
1718 IF_DEBUG(scheduler,debugBelch("--++ thread %d (%s) finished\n",
1719 t->id, whatNext_strs[t->what_next]));
1722 endThread(t, CurrentProc); // clean-up the thread
1723 #elif defined(PARALLEL_HASKELL)
1724 /* For now all are advisory -- HWL */
1725 //if(t->priority==AdvisoryPriority) ??
1726 advisory_thread_count--; // JB: Caution with this counter, buggy!
1729 if(t->dist.priority==RevalPriority)
1733 # if defined(EDENOLD)
1734 // the thread could still have an outport... (BUG)
1735 if (t->eden.outport != -1) {
1736 // delete the outport for the tso which has finished...
1737 IF_PAR_DEBUG(eden_ports,
1738 debugBelch("WARNING: Scheduler removes outport %d for TSO %d.\n",
1739 t->eden.outport, t->id));
1742 // thread still in the process (HEAVY BUG! since outport has just been closed...)
1743 if (t->eden.epid != -1) {
1744 IF_PAR_DEBUG(eden_ports,
1745 debugBelch("WARNING: Scheduler removes TSO %d from process %d .\n",
1746 t->id, t->eden.epid));
1747 removeTSOfromProcess(t);
1752 if (RtsFlags.ParFlags.ParStats.Full &&
1753 !RtsFlags.ParFlags.ParStats.Suppressed)
1754 DumpEndEvent(CURRENT_PROC, t, rtsFalse /* not mandatory */);
1756 // t->par only contains statistics: left out for now...
1758 debugBelch("**** end thread: ended sparked thread %d (%lx); sparkname: %lx\n",
1759 t->id,t,t->par.sparkname));
1761 #endif // PARALLEL_HASKELL
1764 // Check whether the thread that just completed was a main
1765 // thread, and if so return with the result.
1767 // There is an assumption here that all thread completion goes
1768 // through this point; we need to make sure that if a thread
1769 // ends up in the ThreadKilled state, that it stays on the run
1770 // queue so it can be dealt with here.
1773 #if defined(RTS_SUPPORTS_THREADS)
1776 mainThread->tso == t
1780 // We are a bound thread: this must be our thread that just
1782 ASSERT(mainThread->tso == t);
1784 if (t->what_next == ThreadComplete) {
1785 if (mainThread->ret) {
1786 // NOTE: return val is tso->sp[1] (see StgStartup.hc)
1787 *(mainThread->ret) = (StgClosure *)mainThread->tso->sp[1];
1789 mainThread->stat = Success;
1791 if (mainThread->ret) {
1792 *(mainThread->ret) = NULL;
1795 mainThread->stat = Interrupted;
1797 mainThread->stat = Killed;
1801 removeThreadLabel((StgWord)mainThread->tso->id);
1803 if (mainThread->prev == NULL) {
1804 main_threads = mainThread->link;
1806 mainThread->prev->link = mainThread->link;
1808 if (mainThread->link != NULL) {
1809 mainThread->link->prev = NULL;
1811 releaseCapability(cap);
1812 return rtsTrue; // tells schedule() to return
1815 #ifdef RTS_SUPPORTS_THREADS
1816 ASSERT(t->main == NULL);
1818 if (t->main != NULL) {
1819 // Must be a main thread that is not the topmost one. Leave
1820 // it on the run queue until the stack has unwound to the
1821 // point where we can deal with this. Leaving it on the run
1822 // queue also ensures that the garbage collector knows about
1823 // this thread and its return value (it gets dropped from the
1824 // all_threads list so there's no other way to find it).
1825 APPEND_TO_RUN_QUEUE(t);
1831 /* -----------------------------------------------------------------------------
1832 * Perform a heap census, if PROFILING
1833 * -------------------------------------------------------------------------- */
1836 scheduleDoHeapProfile( rtsBool ready_to_gc STG_UNUSED )
1838 #if defined(PROFILING)
1839 // When we have +RTS -i0 and we're heap profiling, do a census at
1840 // every GC. This lets us get repeatable runs for debugging.
1841 if (performHeapProfile ||
1842 (RtsFlags.ProfFlags.profileInterval==0 &&
1843 RtsFlags.ProfFlags.doHeapProfile && ready_to_gc)) {
1844 GarbageCollect(GetRoots, rtsTrue);
1846 performHeapProfile = rtsFalse;
1847 return rtsTrue; // true <=> we already GC'd
1853 /* -----------------------------------------------------------------------------
1854 * Perform a garbage collection if necessary
1855 * ASSUMES: sched_mutex
1856 * -------------------------------------------------------------------------- */
1859 scheduleDoGC( Capability *cap STG_UNUSED )
1863 int n_capabilities = RtsFlags.ParFlags.nNodes - 1;
1864 // subtract one because we're already holding one.
1865 Capability *caps[n_capabilities];
1869 // In order to GC, there must be no threads running Haskell code.
1870 // Therefore, the GC thread needs to hold *all* the capabilities,
1871 // and release them after the GC has completed.
1873 // This seems to be the simplest way: previous attempts involved
1874 // making all the threads with capabilities give up their
1875 // capabilities and sleep except for the *last* one, which
1876 // actually did the GC. But it's quite hard to arrange for all
1877 // the other tasks to sleep and stay asleep.
1880 caps[n_capabilities] = cap;
1881 while (n_capabilities > 0) {
1882 IF_DEBUG(scheduler, sched_belch("ready_to_gc, grabbing all the capabilies (%d left)", n_capabilities));
1883 waitForReturnCapability(&sched_mutex, &cap);
1885 caps[n_capabilities] = cap;
1889 /* Kick any transactions which are invalid back to their
1890 * atomically frames. When next scheduled they will try to
1891 * commit, this commit will fail and they will retry.
1893 for (t = all_threads; t != END_TSO_QUEUE; t = t -> link) {
1894 if (t -> what_next != ThreadRelocated && t -> trec != NO_TREC && t -> why_blocked == NotBlocked) {
1895 if (!stmValidateTransaction (t -> trec)) {
1896 IF_DEBUG(stm, sched_belch("trec %p found wasting its time", t));
1898 // strip the stack back to the ATOMICALLY_FRAME, aborting
1899 // the (nested) transaction, and saving the stack of any
1900 // partially-evaluated thunks on the heap.
1901 raiseAsync_(t, NULL, rtsTrue);
1904 ASSERT(get_itbl((StgClosure *)t->sp)->type == ATOMICALLY_FRAME);
1910 // so this happens periodically:
1911 scheduleCheckBlackHoles();
1913 /* everybody back, start the GC.
1914 * Could do it in this thread, or signal a condition var
1915 * to do it in another thread. Either way, we need to
1916 * broadcast on gc_pending_cond afterward.
1918 #if defined(RTS_SUPPORTS_THREADS)
1919 IF_DEBUG(scheduler,sched_belch("doing GC"));
1921 GarbageCollect(GetRoots,rtsFalse);
1925 // release our stash of capabilities.
1927 for (i = 0; i < RtsFlags.ParFlags.nNodes-1; i++) {
1928 releaseCapability(caps[i]);
1934 /* add a ContinueThread event to continue execution of current thread */
1935 new_event(CurrentProc, CurrentProc, CurrentTime[CurrentProc],
1937 t, (StgClosure*)NULL, (rtsSpark*)NULL);
1939 debugBelch("GRAN: eventq and runnableq after Garbage collection:\n\n");
1945 /* ---------------------------------------------------------------------------
1946 * rtsSupportsBoundThreads(): is the RTS built to support bound threads?
1947 * used by Control.Concurrent for error checking.
1948 * ------------------------------------------------------------------------- */
1951 rtsSupportsBoundThreads(void)
1960 /* ---------------------------------------------------------------------------
1961 * isThreadBound(tso): check whether tso is bound to an OS thread.
1962 * ------------------------------------------------------------------------- */
1965 isThreadBound(StgTSO* tso USED_IN_THREADED_RTS)
1968 return (tso->main != NULL);
1973 /* ---------------------------------------------------------------------------
1974 * Singleton fork(). Do not copy any running threads.
1975 * ------------------------------------------------------------------------- */
1977 #ifndef mingw32_HOST_OS
1978 #define FORKPROCESS_PRIMOP_SUPPORTED
1981 #ifdef FORKPROCESS_PRIMOP_SUPPORTED
1983 deleteThreadImmediately(StgTSO *tso);
1986 forkProcess(HsStablePtr *entry
1987 #ifndef FORKPROCESS_PRIMOP_SUPPORTED
1992 #ifdef FORKPROCESS_PRIMOP_SUPPORTED
1998 IF_DEBUG(scheduler,sched_belch("forking!"));
1999 rts_lock(); // This not only acquires sched_mutex, it also
2000 // makes sure that no other threads are running
2004 if (pid) { /* parent */
2006 /* just return the pid */
2010 } else { /* child */
2013 // delete all threads
2014 run_queue_hd = run_queue_tl = END_TSO_QUEUE;
2016 for (t = all_threads; t != END_TSO_QUEUE; t = next) {
2019 // don't allow threads to catch the ThreadKilled exception
2020 deleteThreadImmediately(t);
2023 // wipe the main thread list
2024 while((m = main_threads) != NULL) {
2025 main_threads = m->link;
2026 # ifdef THREADED_RTS
2027 closeCondition(&m->bound_thread_cond);
2032 rc = rts_evalStableIO(entry, NULL); // run the action
2033 rts_checkSchedStatus("forkProcess",rc);
2037 hs_exit(); // clean up and exit
2040 #else /* !FORKPROCESS_PRIMOP_SUPPORTED */
2041 barf("forkProcess#: primop not supported, sorry!\n");
2046 /* ---------------------------------------------------------------------------
2047 * deleteAllThreads(): kill all the live threads.
2049 * This is used when we catch a user interrupt (^C), before performing
2050 * any necessary cleanups and running finalizers.
2052 * Locks: sched_mutex held.
2053 * ------------------------------------------------------------------------- */
2056 deleteAllThreads ( void )
2059 IF_DEBUG(scheduler,sched_belch("deleting all threads"));
2060 for (t = all_threads; t != END_TSO_QUEUE; t = next) {
2061 next = t->global_link;
2065 // The run queue now contains a bunch of ThreadKilled threads. We
2066 // must not throw these away: the main thread(s) will be in there
2067 // somewhere, and the main scheduler loop has to deal with it.
2068 // Also, the run queue is the only thing keeping these threads from
2069 // being GC'd, and we don't want the "main thread has been GC'd" panic.
2071 ASSERT(blocked_queue_hd == END_TSO_QUEUE);
2072 ASSERT(blackhole_queue == END_TSO_QUEUE);
2073 ASSERT(sleeping_queue == END_TSO_QUEUE);
2076 /* startThread and insertThread are now in GranSim.c -- HWL */
2079 /* ---------------------------------------------------------------------------
2080 * Suspending & resuming Haskell threads.
2082 * When making a "safe" call to C (aka _ccall_GC), the task gives back
2083 * its capability before calling the C function. This allows another
2084 * task to pick up the capability and carry on running Haskell
2085 * threads. It also means that if the C call blocks, it won't lock
2088 * The Haskell thread making the C call is put to sleep for the
2089 * duration of the call, on the susepended_ccalling_threads queue. We
2090 * give out a token to the task, which it can use to resume the thread
2091 * on return from the C function.
2092 * ------------------------------------------------------------------------- */
2095 suspendThread( StgRegTable *reg )
2099 int saved_errno = errno;
2101 /* assume that *reg is a pointer to the StgRegTable part
2104 cap = (Capability *)((void *)((unsigned char*)reg - sizeof(StgFunTable)));
2106 ACQUIRE_LOCK(&sched_mutex);
2109 sched_belch("thread %d did a _ccall_gc", cap->r.rCurrentTSO->id));
2111 // XXX this might not be necessary --SDM
2112 cap->r.rCurrentTSO->what_next = ThreadRunGHC;
2114 threadPaused(cap->r.rCurrentTSO);
2115 cap->r.rCurrentTSO->link = suspended_ccalling_threads;
2116 suspended_ccalling_threads = cap->r.rCurrentTSO;
2118 if(cap->r.rCurrentTSO->blocked_exceptions == NULL) {
2119 cap->r.rCurrentTSO->why_blocked = BlockedOnCCall;
2120 cap->r.rCurrentTSO->blocked_exceptions = END_TSO_QUEUE;
2122 cap->r.rCurrentTSO->why_blocked = BlockedOnCCall_NoUnblockExc;
2125 /* Use the thread ID as the token; it should be unique */
2126 tok = cap->r.rCurrentTSO->id;
2128 /* Hand back capability */
2129 cap->r.rInHaskell = rtsFalse;
2130 releaseCapability(cap);
2132 #if defined(RTS_SUPPORTS_THREADS)
2133 /* Preparing to leave the RTS, so ensure there's a native thread/task
2134 waiting to take over.
2136 IF_DEBUG(scheduler, sched_belch("worker (token %d): leaving RTS", tok));
2139 RELEASE_LOCK(&sched_mutex);
2141 errno = saved_errno;
2146 resumeThread( StgInt tok )
2148 StgTSO *tso, **prev;
2150 int saved_errno = errno;
2152 #if defined(RTS_SUPPORTS_THREADS)
2153 /* Wait for permission to re-enter the RTS with the result. */
2154 ACQUIRE_LOCK(&sched_mutex);
2155 waitForReturnCapability(&sched_mutex, &cap);
2157 IF_DEBUG(scheduler, sched_belch("worker (token %d): re-entering RTS", tok));
2159 grabCapability(&cap);
2162 /* Remove the thread off of the suspended list */
2163 prev = &suspended_ccalling_threads;
2164 for (tso = suspended_ccalling_threads;
2165 tso != END_TSO_QUEUE;
2166 prev = &tso->link, tso = tso->link) {
2167 if (tso->id == (StgThreadID)tok) {
2172 if (tso == END_TSO_QUEUE) {
2173 barf("resumeThread: thread not found");
2175 tso->link = END_TSO_QUEUE;
2177 if(tso->why_blocked == BlockedOnCCall) {
2178 awakenBlockedQueueNoLock(tso->blocked_exceptions);
2179 tso->blocked_exceptions = NULL;
2182 /* Reset blocking status */
2183 tso->why_blocked = NotBlocked;
2185 cap->r.rCurrentTSO = tso;
2186 cap->r.rInHaskell = rtsTrue;
2187 RELEASE_LOCK(&sched_mutex);
2188 errno = saved_errno;
2192 /* ---------------------------------------------------------------------------
2193 * Comparing Thread ids.
2195 * This is used from STG land in the implementation of the
2196 * instances of Eq/Ord for ThreadIds.
2197 * ------------------------------------------------------------------------ */
2200 cmp_thread(StgPtr tso1, StgPtr tso2)
2202 StgThreadID id1 = ((StgTSO *)tso1)->id;
2203 StgThreadID id2 = ((StgTSO *)tso2)->id;
2205 if (id1 < id2) return (-1);
2206 if (id1 > id2) return 1;
2210 /* ---------------------------------------------------------------------------
2211 * Fetching the ThreadID from an StgTSO.
2213 * This is used in the implementation of Show for ThreadIds.
2214 * ------------------------------------------------------------------------ */
2216 rts_getThreadId(StgPtr tso)
2218 return ((StgTSO *)tso)->id;
2223 labelThread(StgPtr tso, char *label)
2228 /* Caveat: Once set, you can only set the thread name to "" */
2229 len = strlen(label)+1;
2230 buf = stgMallocBytes(len * sizeof(char), "Schedule.c:labelThread()");
2231 strncpy(buf,label,len);
2232 /* Update will free the old memory for us */
2233 updateThreadLabel(((StgTSO *)tso)->id,buf);
2237 /* ---------------------------------------------------------------------------
2238 Create a new thread.
2240 The new thread starts with the given stack size. Before the
2241 scheduler can run, however, this thread needs to have a closure
2242 (and possibly some arguments) pushed on its stack. See
2243 pushClosure() in Schedule.h.
2245 createGenThread() and createIOThread() (in SchedAPI.h) are
2246 convenient packaged versions of this function.
2248 currently pri (priority) is only used in a GRAN setup -- HWL
2249 ------------------------------------------------------------------------ */
2251 /* currently pri (priority) is only used in a GRAN setup -- HWL */
2253 createThread(nat size, StgInt pri)
2256 createThread(nat size)
2263 /* First check whether we should create a thread at all */
2264 #if defined(PARALLEL_HASKELL)
2265 /* check that no more than RtsFlags.ParFlags.maxThreads threads are created */
2266 if (advisory_thread_count >= RtsFlags.ParFlags.maxThreads) {
2268 debugBelch("{createThread}Daq ghuH: refusing to create another thread; no more than %d threads allowed (currently %d)\n",
2269 RtsFlags.ParFlags.maxThreads, advisory_thread_count);
2270 return END_TSO_QUEUE;
2276 ASSERT(!RtsFlags.GranFlags.Light || CurrentProc==0);
2279 // ToDo: check whether size = stack_size - TSO_STRUCT_SIZEW
2281 /* catch ridiculously small stack sizes */
2282 if (size < MIN_STACK_WORDS + TSO_STRUCT_SIZEW) {
2283 size = MIN_STACK_WORDS + TSO_STRUCT_SIZEW;
2286 stack_size = size - TSO_STRUCT_SIZEW;
2288 tso = (StgTSO *)allocate(size);
2289 TICK_ALLOC_TSO(stack_size, 0);
2291 SET_HDR(tso, &stg_TSO_info, CCS_SYSTEM);
2293 SET_GRAN_HDR(tso, ThisPE);
2296 // Always start with the compiled code evaluator
2297 tso->what_next = ThreadRunGHC;
2299 tso->id = next_thread_id++;
2300 tso->why_blocked = NotBlocked;
2301 tso->blocked_exceptions = NULL;
2303 tso->saved_errno = 0;
2306 tso->stack_size = stack_size;
2307 tso->max_stack_size = round_to_mblocks(RtsFlags.GcFlags.maxStkSize)
2309 tso->sp = (P_)&(tso->stack) + stack_size;
2311 tso->trec = NO_TREC;
2314 tso->prof.CCCS = CCS_MAIN;
2317 /* put a stop frame on the stack */
2318 tso->sp -= sizeofW(StgStopFrame);
2319 SET_HDR((StgClosure*)tso->sp,(StgInfoTable *)&stg_stop_thread_info,CCS_SYSTEM);
2320 tso->link = END_TSO_QUEUE;
2324 /* uses more flexible routine in GranSim */
2325 insertThread(tso, CurrentProc);
2327 /* In a non-GranSim setup the pushing of a TSO onto the runq is separated
2333 if (RtsFlags.GranFlags.GranSimStats.Full)
2334 DumpGranEvent(GR_START,tso);
2335 #elif defined(PARALLEL_HASKELL)
2336 if (RtsFlags.ParFlags.ParStats.Full)
2337 DumpGranEvent(GR_STARTQ,tso);
2338 /* HACk to avoid SCHEDULE
2342 /* Link the new thread on the global thread list.
2344 tso->global_link = all_threads;
2348 tso->dist.priority = MandatoryPriority; //by default that is...
2352 tso->gran.pri = pri;
2354 tso->gran.magic = TSO_MAGIC; // debugging only
2356 tso->gran.sparkname = 0;
2357 tso->gran.startedat = CURRENT_TIME;
2358 tso->gran.exported = 0;
2359 tso->gran.basicblocks = 0;
2360 tso->gran.allocs = 0;
2361 tso->gran.exectime = 0;
2362 tso->gran.fetchtime = 0;
2363 tso->gran.fetchcount = 0;
2364 tso->gran.blocktime = 0;
2365 tso->gran.blockcount = 0;
2366 tso->gran.blockedat = 0;
2367 tso->gran.globalsparks = 0;
2368 tso->gran.localsparks = 0;
2369 if (RtsFlags.GranFlags.Light)
2370 tso->gran.clock = Now; /* local clock */
2372 tso->gran.clock = 0;
2374 IF_DEBUG(gran,printTSO(tso));
2375 #elif defined(PARALLEL_HASKELL)
2377 tso->par.magic = TSO_MAGIC; // debugging only
2379 tso->par.sparkname = 0;
2380 tso->par.startedat = CURRENT_TIME;
2381 tso->par.exported = 0;
2382 tso->par.basicblocks = 0;
2383 tso->par.allocs = 0;
2384 tso->par.exectime = 0;
2385 tso->par.fetchtime = 0;
2386 tso->par.fetchcount = 0;
2387 tso->par.blocktime = 0;
2388 tso->par.blockcount = 0;
2389 tso->par.blockedat = 0;
2390 tso->par.globalsparks = 0;
2391 tso->par.localsparks = 0;
2395 globalGranStats.tot_threads_created++;
2396 globalGranStats.threads_created_on_PE[CurrentProc]++;
2397 globalGranStats.tot_sq_len += spark_queue_len(CurrentProc);
2398 globalGranStats.tot_sq_probes++;
2399 #elif defined(PARALLEL_HASKELL)
2400 // collect parallel global statistics (currently done together with GC stats)
2401 if (RtsFlags.ParFlags.ParStats.Global &&
2402 RtsFlags.GcFlags.giveStats > NO_GC_STATS) {
2403 //debugBelch("Creating thread %d @ %11.2f\n", tso->id, usertime());
2404 globalParStats.tot_threads_created++;
2410 sched_belch("==__ schedule: Created TSO %d (%p);",
2411 CurrentProc, tso, tso->id));
2412 #elif defined(PARALLEL_HASKELL)
2413 IF_PAR_DEBUG(verbose,
2414 sched_belch("==__ schedule: Created TSO %d (%p); %d threads active",
2415 (long)tso->id, tso, advisory_thread_count));
2417 IF_DEBUG(scheduler,sched_belch("created thread %ld, stack size = %lx words",
2418 (long)tso->id, (long)tso->stack_size));
2425 all parallel thread creation calls should fall through the following routine.
2428 createThreadFromSpark(rtsSpark spark)
2430 ASSERT(spark != (rtsSpark)NULL);
2431 // JB: TAKE CARE OF THIS COUNTER! BUGGY
2432 if (advisory_thread_count >= RtsFlags.ParFlags.maxThreads)
2434 barf("{createSparkThread}Daq ghuH: refusing to create another thread; no more than %d threads allowed (currently %d)",
2435 RtsFlags.ParFlags.maxThreads, advisory_thread_count);
2436 return END_TSO_QUEUE;
2440 tso = createThread(RtsFlags.GcFlags.initialStkSize);
2441 if (tso==END_TSO_QUEUE)
2442 barf("createSparkThread: Cannot create TSO");
2444 tso->priority = AdvisoryPriority;
2446 pushClosure(tso,spark);
2448 advisory_thread_count++; // JB: TAKE CARE OF THIS COUNTER! BUGGY
2455 Turn a spark into a thread.
2456 ToDo: fix for SMP (needs to acquire SCHED_MUTEX!)
2460 activateSpark (rtsSpark spark)
2464 tso = createSparkThread(spark);
2465 if (RtsFlags.ParFlags.ParStats.Full) {
2466 //ASSERT(run_queue_hd == END_TSO_QUEUE); // I think ...
2467 IF_PAR_DEBUG(verbose,
2468 debugBelch("==^^ activateSpark: turning spark of closure %p (%s) into a thread\n",
2469 (StgClosure *)spark, info_type((StgClosure *)spark)));
2471 // ToDo: fwd info on local/global spark to thread -- HWL
2472 // tso->gran.exported = spark->exported;
2473 // tso->gran.locked = !spark->global;
2474 // tso->gran.sparkname = spark->name;
2480 /* ---------------------------------------------------------------------------
2483 * scheduleThread puts a thread on the head of the runnable queue.
2484 * This will usually be done immediately after a thread is created.
2485 * The caller of scheduleThread must create the thread using e.g.
2486 * createThread and push an appropriate closure
2487 * on this thread's stack before the scheduler is invoked.
2488 * ------------------------------------------------------------------------ */
2491 scheduleThread_(StgTSO *tso)
2493 // The thread goes at the *end* of the run-queue, to avoid possible
2494 // starvation of any threads already on the queue.
2495 APPEND_TO_RUN_QUEUE(tso);
2500 scheduleThread(StgTSO* tso)
2502 ACQUIRE_LOCK(&sched_mutex);
2503 scheduleThread_(tso);
2504 RELEASE_LOCK(&sched_mutex);
2507 #if defined(RTS_SUPPORTS_THREADS)
2508 static Condition bound_cond_cache;
2509 static int bound_cond_cache_full = 0;
2514 scheduleWaitThread(StgTSO* tso, /*[out]*/HaskellObj* ret,
2515 Capability *initialCapability)
2517 // Precondition: sched_mutex must be held
2520 m = stgMallocBytes(sizeof(StgMainThread), "waitThread");
2525 m->link = main_threads;
2527 if (main_threads != NULL) {
2528 main_threads->prev = m;
2532 #if defined(RTS_SUPPORTS_THREADS)
2533 // Allocating a new condition for each thread is expensive, so we
2534 // cache one. This is a pretty feeble hack, but it helps speed up
2535 // consecutive call-ins quite a bit.
2536 if (bound_cond_cache_full) {
2537 m->bound_thread_cond = bound_cond_cache;
2538 bound_cond_cache_full = 0;
2540 initCondition(&m->bound_thread_cond);
2544 /* Put the thread on the main-threads list prior to scheduling the TSO.
2545 Failure to do so introduces a race condition in the MT case (as
2546 identified by Wolfgang Thaller), whereby the new task/OS thread
2547 created by scheduleThread_() would complete prior to the thread
2548 that spawned it managed to put 'itself' on the main-threads list.
2549 The upshot of it all being that the worker thread wouldn't get to
2550 signal the completion of the its work item for the main thread to
2551 see (==> it got stuck waiting.) -- sof 6/02.
2553 IF_DEBUG(scheduler, sched_belch("waiting for thread (%d)", tso->id));
2555 APPEND_TO_RUN_QUEUE(tso);
2556 // NB. Don't call threadRunnable() here, because the thread is
2557 // bound and only runnable by *this* OS thread, so waking up other
2558 // workers will just slow things down.
2560 return waitThread_(m, initialCapability);
2563 /* ---------------------------------------------------------------------------
2566 * Initialise the scheduler. This resets all the queues - if the
2567 * queues contained any threads, they'll be garbage collected at the
2570 * ------------------------------------------------------------------------ */
2578 for (i=0; i<=MAX_PROC; i++) {
2579 run_queue_hds[i] = END_TSO_QUEUE;
2580 run_queue_tls[i] = END_TSO_QUEUE;
2581 blocked_queue_hds[i] = END_TSO_QUEUE;
2582 blocked_queue_tls[i] = END_TSO_QUEUE;
2583 ccalling_threadss[i] = END_TSO_QUEUE;
2584 blackhole_queue[i] = END_TSO_QUEUE;
2585 sleeping_queue = END_TSO_QUEUE;
2588 run_queue_hd = END_TSO_QUEUE;
2589 run_queue_tl = END_TSO_QUEUE;
2590 blocked_queue_hd = END_TSO_QUEUE;
2591 blocked_queue_tl = END_TSO_QUEUE;
2592 blackhole_queue = END_TSO_QUEUE;
2593 sleeping_queue = END_TSO_QUEUE;
2596 suspended_ccalling_threads = END_TSO_QUEUE;
2598 main_threads = NULL;
2599 all_threads = END_TSO_QUEUE;
2604 RtsFlags.ConcFlags.ctxtSwitchTicks =
2605 RtsFlags.ConcFlags.ctxtSwitchTime / TICK_MILLISECS;
2607 #if defined(RTS_SUPPORTS_THREADS)
2608 /* Initialise the mutex and condition variables used by
2610 initMutex(&sched_mutex);
2611 initMutex(&term_mutex);
2614 ACQUIRE_LOCK(&sched_mutex);
2616 /* A capability holds the state a native thread needs in
2617 * order to execute STG code. At least one capability is
2618 * floating around (only SMP builds have more than one).
2622 #if defined(RTS_SUPPORTS_THREADS)
2627 /* eagerly start some extra workers */
2628 startTasks(RtsFlags.ParFlags.nNodes, taskStart);
2631 #if /* defined(SMP) ||*/ defined(PARALLEL_HASKELL)
2635 RELEASE_LOCK(&sched_mutex);
2639 exitScheduler( void )
2641 interrupted = rtsTrue;
2642 shutting_down_scheduler = rtsTrue;
2643 #if defined(RTS_SUPPORTS_THREADS)
2644 if (threadIsTask(osThreadId())) { taskStop(); }
2649 /* ----------------------------------------------------------------------------
2650 Managing the per-task allocation areas.
2652 Each capability comes with an allocation area. These are
2653 fixed-length block lists into which allocation can be done.
2655 ToDo: no support for two-space collection at the moment???
2656 ------------------------------------------------------------------------- */
2658 static SchedulerStatus
2659 waitThread_(StgMainThread* m, Capability *initialCapability)
2661 SchedulerStatus stat;
2663 // Precondition: sched_mutex must be held.
2664 IF_DEBUG(scheduler, sched_belch("new main thread (%d)", m->tso->id));
2667 /* GranSim specific init */
2668 CurrentTSO = m->tso; // the TSO to run
2669 procStatus[MainProc] = Busy; // status of main PE
2670 CurrentProc = MainProc; // PE to run it on
2671 schedule(m,initialCapability);
2673 schedule(m,initialCapability);
2674 ASSERT(m->stat != NoStatus);
2679 #if defined(RTS_SUPPORTS_THREADS)
2680 // Free the condition variable, returning it to the cache if possible.
2681 if (!bound_cond_cache_full) {
2682 bound_cond_cache = m->bound_thread_cond;
2683 bound_cond_cache_full = 1;
2685 closeCondition(&m->bound_thread_cond);
2689 IF_DEBUG(scheduler, sched_belch("main thread (%d) finished", m->tso->id));
2692 // Postcondition: sched_mutex still held
2696 /* ---------------------------------------------------------------------------
2697 Where are the roots that we know about?
2699 - all the threads on the runnable queue
2700 - all the threads on the blocked queue
2701 - all the threads on the sleeping queue
2702 - all the thread currently executing a _ccall_GC
2703 - all the "main threads"
2705 ------------------------------------------------------------------------ */
2707 /* This has to be protected either by the scheduler monitor, or by the
2708 garbage collection monitor (probably the latter).
2713 GetRoots( evac_fn evac )
2718 for (i=0; i<=RtsFlags.GranFlags.proc; i++) {
2719 if ((run_queue_hds[i] != END_TSO_QUEUE) && ((run_queue_hds[i] != NULL)))
2720 evac((StgClosure **)&run_queue_hds[i]);
2721 if ((run_queue_tls[i] != END_TSO_QUEUE) && ((run_queue_tls[i] != NULL)))
2722 evac((StgClosure **)&run_queue_tls[i]);
2724 if ((blocked_queue_hds[i] != END_TSO_QUEUE) && ((blocked_queue_hds[i] != NULL)))
2725 evac((StgClosure **)&blocked_queue_hds[i]);
2726 if ((blocked_queue_tls[i] != END_TSO_QUEUE) && ((blocked_queue_tls[i] != NULL)))
2727 evac((StgClosure **)&blocked_queue_tls[i]);
2728 if ((ccalling_threadss[i] != END_TSO_QUEUE) && ((ccalling_threadss[i] != NULL)))
2729 evac((StgClosure **)&ccalling_threads[i]);
2736 if (run_queue_hd != END_TSO_QUEUE) {
2737 ASSERT(run_queue_tl != END_TSO_QUEUE);
2738 evac((StgClosure **)&run_queue_hd);
2739 evac((StgClosure **)&run_queue_tl);
2742 if (blocked_queue_hd != END_TSO_QUEUE) {
2743 ASSERT(blocked_queue_tl != END_TSO_QUEUE);
2744 evac((StgClosure **)&blocked_queue_hd);
2745 evac((StgClosure **)&blocked_queue_tl);
2748 if (sleeping_queue != END_TSO_QUEUE) {
2749 evac((StgClosure **)&sleeping_queue);
2753 if (blackhole_queue != END_TSO_QUEUE) {
2754 evac((StgClosure **)&blackhole_queue);
2757 if (suspended_ccalling_threads != END_TSO_QUEUE) {
2758 evac((StgClosure **)&suspended_ccalling_threads);
2761 #if defined(PARALLEL_HASKELL) || defined(GRAN)
2762 markSparkQueue(evac);
2765 #if defined(RTS_USER_SIGNALS)
2766 // mark the signal handlers (signals should be already blocked)
2767 markSignalHandlers(evac);
2771 /* -----------------------------------------------------------------------------
2774 This is the interface to the garbage collector from Haskell land.
2775 We provide this so that external C code can allocate and garbage
2776 collect when called from Haskell via _ccall_GC.
2778 It might be useful to provide an interface whereby the programmer
2779 can specify more roots (ToDo).
2781 This needs to be protected by the GC condition variable above. KH.
2782 -------------------------------------------------------------------------- */
2784 static void (*extra_roots)(evac_fn);
2789 /* Obligated to hold this lock upon entry */
2790 ACQUIRE_LOCK(&sched_mutex);
2791 GarbageCollect(GetRoots,rtsFalse);
2792 RELEASE_LOCK(&sched_mutex);
2796 performMajorGC(void)
2798 ACQUIRE_LOCK(&sched_mutex);
2799 GarbageCollect(GetRoots,rtsTrue);
2800 RELEASE_LOCK(&sched_mutex);
2804 AllRoots(evac_fn evac)
2806 GetRoots(evac); // the scheduler's roots
2807 extra_roots(evac); // the user's roots
2811 performGCWithRoots(void (*get_roots)(evac_fn))
2813 ACQUIRE_LOCK(&sched_mutex);
2814 extra_roots = get_roots;
2815 GarbageCollect(AllRoots,rtsFalse);
2816 RELEASE_LOCK(&sched_mutex);
2819 /* -----------------------------------------------------------------------------
2822 If the thread has reached its maximum stack size, then raise the
2823 StackOverflow exception in the offending thread. Otherwise
2824 relocate the TSO into a larger chunk of memory and adjust its stack
2826 -------------------------------------------------------------------------- */
2829 threadStackOverflow(StgTSO *tso)
2831 nat new_stack_size, stack_words;
2836 IF_DEBUG(sanity,checkTSO(tso));
2837 if (tso->stack_size >= tso->max_stack_size) {
2840 debugBelch("@@ threadStackOverflow of TSO %ld (%p): stack too large (now %ld; max is %ld)\n",
2841 (long)tso->id, tso, (long)tso->stack_size, (long)tso->max_stack_size);
2842 /* If we're debugging, just print out the top of the stack */
2843 printStackChunk(tso->sp, stg_min(tso->stack+tso->stack_size,
2846 /* Send this thread the StackOverflow exception */
2847 raiseAsync(tso, (StgClosure *)stackOverflow_closure);
2851 /* Try to double the current stack size. If that takes us over the
2852 * maximum stack size for this thread, then use the maximum instead.
2853 * Finally round up so the TSO ends up as a whole number of blocks.
2855 new_stack_size = stg_min(tso->stack_size * 2, tso->max_stack_size);
2856 new_tso_size = (lnat)BLOCK_ROUND_UP(new_stack_size * sizeof(W_) +
2857 TSO_STRUCT_SIZE)/sizeof(W_);
2858 new_tso_size = round_to_mblocks(new_tso_size); /* Be MBLOCK-friendly */
2859 new_stack_size = new_tso_size - TSO_STRUCT_SIZEW;
2861 IF_DEBUG(scheduler, debugBelch("== sched: increasing stack size from %d words to %d.\n", tso->stack_size, new_stack_size));
2863 dest = (StgTSO *)allocate(new_tso_size);
2864 TICK_ALLOC_TSO(new_stack_size,0);
2866 /* copy the TSO block and the old stack into the new area */
2867 memcpy(dest,tso,TSO_STRUCT_SIZE);
2868 stack_words = tso->stack + tso->stack_size - tso->sp;
2869 new_sp = (P_)dest + new_tso_size - stack_words;
2870 memcpy(new_sp, tso->sp, stack_words * sizeof(W_));
2872 /* relocate the stack pointers... */
2874 dest->stack_size = new_stack_size;
2876 /* Mark the old TSO as relocated. We have to check for relocated
2877 * TSOs in the garbage collector and any primops that deal with TSOs.
2879 * It's important to set the sp value to just beyond the end
2880 * of the stack, so we don't attempt to scavenge any part of the
2883 tso->what_next = ThreadRelocated;
2885 tso->sp = (P_)&(tso->stack[tso->stack_size]);
2886 tso->why_blocked = NotBlocked;
2888 IF_PAR_DEBUG(verbose,
2889 debugBelch("@@ threadStackOverflow of TSO %d (now at %p): stack size increased to %ld\n",
2890 tso->id, tso, tso->stack_size);
2891 /* If we're debugging, just print out the top of the stack */
2892 printStackChunk(tso->sp, stg_min(tso->stack+tso->stack_size,
2895 IF_DEBUG(sanity,checkTSO(tso));
2897 IF_DEBUG(scheduler,printTSO(dest));
2903 /* ---------------------------------------------------------------------------
2904 Wake up a queue that was blocked on some resource.
2905 ------------------------------------------------------------------------ */
2909 unblockCount ( StgBlockingQueueElement *bqe, StgClosure *node )
2912 #elif defined(PARALLEL_HASKELL)
2914 unblockCount ( StgBlockingQueueElement *bqe, StgClosure *node )
2916 /* write RESUME events to log file and
2917 update blocked and fetch time (depending on type of the orig closure) */
2918 if (RtsFlags.ParFlags.ParStats.Full) {
2919 DumpRawGranEvent(CURRENT_PROC, CURRENT_PROC,
2920 GR_RESUMEQ, ((StgTSO *)bqe), ((StgTSO *)bqe)->block_info.closure,
2921 0, 0 /* spark_queue_len(ADVISORY_POOL) */);
2922 if (EMPTY_RUN_QUEUE())
2923 emitSchedule = rtsTrue;
2925 switch (get_itbl(node)->type) {
2927 ((StgTSO *)bqe)->par.fetchtime += CURRENT_TIME-((StgTSO *)bqe)->par.blockedat;
2932 ((StgTSO *)bqe)->par.blocktime += CURRENT_TIME-((StgTSO *)bqe)->par.blockedat;
2939 barf("{unblockOneLocked}Daq Qagh: unexpected closure in blocking queue");
2946 static StgBlockingQueueElement *
2947 unblockOneLocked(StgBlockingQueueElement *bqe, StgClosure *node)
2950 PEs node_loc, tso_loc;
2952 node_loc = where_is(node); // should be lifted out of loop
2953 tso = (StgTSO *)bqe; // wastes an assignment to get the type right
2954 tso_loc = where_is((StgClosure *)tso);
2955 if (IS_LOCAL_TO(PROCS(node),tso_loc)) { // TSO is local
2956 /* !fake_fetch => TSO is on CurrentProc is same as IS_LOCAL_TO */
2957 ASSERT(CurrentProc!=node_loc || tso_loc==CurrentProc);
2958 CurrentTime[CurrentProc] += RtsFlags.GranFlags.Costs.lunblocktime;
2959 // insertThread(tso, node_loc);
2960 new_event(tso_loc, tso_loc, CurrentTime[CurrentProc],
2962 tso, node, (rtsSpark*)NULL);
2963 tso->link = END_TSO_QUEUE; // overwrite link just to be sure
2966 } else { // TSO is remote (actually should be FMBQ)
2967 CurrentTime[CurrentProc] += RtsFlags.GranFlags.Costs.mpacktime +
2968 RtsFlags.GranFlags.Costs.gunblocktime +
2969 RtsFlags.GranFlags.Costs.latency;
2970 new_event(tso_loc, CurrentProc, CurrentTime[CurrentProc],
2972 tso, node, (rtsSpark*)NULL);
2973 tso->link = END_TSO_QUEUE; // overwrite link just to be sure
2976 /* the thread-queue-overhead is accounted for in either Resume or UnblockThread */
2978 debugBelch(" %s TSO %d (%p) [PE %d] (block_info.closure=%p) (next=%p) ,",
2979 (node_loc==tso_loc ? "Local" : "Global"),
2980 tso->id, tso, CurrentProc, tso->block_info.closure, tso->link));
2981 tso->block_info.closure = NULL;
2982 IF_DEBUG(scheduler,debugBelch("-- Waking up thread %ld (%p)\n",
2985 #elif defined(PARALLEL_HASKELL)
2986 static StgBlockingQueueElement *
2987 unblockOneLocked(StgBlockingQueueElement *bqe, StgClosure *node)
2989 StgBlockingQueueElement *next;
2991 switch (get_itbl(bqe)->type) {
2993 ASSERT(((StgTSO *)bqe)->why_blocked != NotBlocked);
2994 /* if it's a TSO just push it onto the run_queue */
2996 ((StgTSO *)bqe)->link = END_TSO_QUEUE; // debugging?
2997 APPEND_TO_RUN_QUEUE((StgTSO *)bqe);
2999 unblockCount(bqe, node);
3000 /* reset blocking status after dumping event */
3001 ((StgTSO *)bqe)->why_blocked = NotBlocked;
3005 /* if it's a BLOCKED_FETCH put it on the PendingFetches list */
3007 bqe->link = (StgBlockingQueueElement *)PendingFetches;
3008 PendingFetches = (StgBlockedFetch *)bqe;
3012 /* can ignore this case in a non-debugging setup;
3013 see comments on RBHSave closures above */
3015 /* check that the closure is an RBHSave closure */
3016 ASSERT(get_itbl((StgClosure *)bqe) == &stg_RBH_Save_0_info ||
3017 get_itbl((StgClosure *)bqe) == &stg_RBH_Save_1_info ||
3018 get_itbl((StgClosure *)bqe) == &stg_RBH_Save_2_info);
3022 barf("{unblockOneLocked}Daq Qagh: Unexpected IP (%#lx; %s) in blocking queue at %#lx\n",
3023 get_itbl((StgClosure *)bqe), info_type((StgClosure *)bqe),
3027 IF_PAR_DEBUG(bq, debugBelch(", %p (%s)\n", bqe, info_type((StgClosure*)bqe)));
3031 #else /* !GRAN && !PARALLEL_HASKELL */
3033 unblockOneLocked(StgTSO *tso)
3037 ASSERT(get_itbl(tso)->type == TSO);
3038 ASSERT(tso->why_blocked != NotBlocked);
3039 tso->why_blocked = NotBlocked;
3041 tso->link = END_TSO_QUEUE;
3042 APPEND_TO_RUN_QUEUE(tso);
3044 IF_DEBUG(scheduler,sched_belch("waking up thread %ld", (long)tso->id));
3049 #if defined(GRAN) || defined(PARALLEL_HASKELL)
3050 INLINE_ME StgBlockingQueueElement *
3051 unblockOne(StgBlockingQueueElement *bqe, StgClosure *node)
3053 ACQUIRE_LOCK(&sched_mutex);
3054 bqe = unblockOneLocked(bqe, node);
3055 RELEASE_LOCK(&sched_mutex);
3060 unblockOne(StgTSO *tso)
3062 ACQUIRE_LOCK(&sched_mutex);
3063 tso = unblockOneLocked(tso);
3064 RELEASE_LOCK(&sched_mutex);
3071 awakenBlockedQueue(StgBlockingQueueElement *q, StgClosure *node)
3073 StgBlockingQueueElement *bqe;
3078 debugBelch("##-_ AwBQ for node %p on PE %d @ %ld by TSO %d (%p): \n", \
3079 node, CurrentProc, CurrentTime[CurrentProc],
3080 CurrentTSO->id, CurrentTSO));
3082 node_loc = where_is(node);
3084 ASSERT(q == END_BQ_QUEUE ||
3085 get_itbl(q)->type == TSO || // q is either a TSO or an RBHSave
3086 get_itbl(q)->type == CONSTR); // closure (type constructor)
3087 ASSERT(is_unique(node));
3089 /* FAKE FETCH: magically copy the node to the tso's proc;
3090 no Fetch necessary because in reality the node should not have been
3091 moved to the other PE in the first place
3093 if (CurrentProc!=node_loc) {
3095 debugBelch("## node %p is on PE %d but CurrentProc is %d (TSO %d); assuming fake fetch and adjusting bitmask (old: %#x)\n",
3096 node, node_loc, CurrentProc, CurrentTSO->id,
3097 // CurrentTSO, where_is(CurrentTSO),
3098 node->header.gran.procs));
3099 node->header.gran.procs = (node->header.gran.procs) | PE_NUMBER(CurrentProc);
3101 debugBelch("## new bitmask of node %p is %#x\n",
3102 node, node->header.gran.procs));
3103 if (RtsFlags.GranFlags.GranSimStats.Global) {
3104 globalGranStats.tot_fake_fetches++;
3109 // ToDo: check: ASSERT(CurrentProc==node_loc);
3110 while (get_itbl(bqe)->type==TSO) { // q != END_TSO_QUEUE) {
3113 bqe points to the current element in the queue
3114 next points to the next element in the queue
3116 //tso = (StgTSO *)bqe; // wastes an assignment to get the type right
3117 //tso_loc = where_is(tso);
3119 bqe = unblockOneLocked(bqe, node);
3122 /* if this is the BQ of an RBH, we have to put back the info ripped out of
3123 the closure to make room for the anchor of the BQ */
3124 if (bqe!=END_BQ_QUEUE) {
3125 ASSERT(get_itbl(node)->type == RBH && get_itbl(bqe)->type == CONSTR);
3127 ASSERT((info_ptr==&RBH_Save_0_info) ||
3128 (info_ptr==&RBH_Save_1_info) ||
3129 (info_ptr==&RBH_Save_2_info));
3131 /* cf. convertToRBH in RBH.c for writing the RBHSave closure */
3132 ((StgRBH *)node)->blocking_queue = (StgBlockingQueueElement *)((StgRBHSave *)bqe)->payload[0];
3133 ((StgRBH *)node)->mut_link = (StgMutClosure *)((StgRBHSave *)bqe)->payload[1];
3136 debugBelch("## Filled in RBH_Save for %p (%s) at end of AwBQ\n",
3137 node, info_type(node)));
3140 /* statistics gathering */
3141 if (RtsFlags.GranFlags.GranSimStats.Global) {
3142 // globalGranStats.tot_bq_processing_time += bq_processing_time;
3143 globalGranStats.tot_bq_len += len; // total length of all bqs awakened
3144 // globalGranStats.tot_bq_len_local += len_local; // same for local TSOs only
3145 globalGranStats.tot_awbq++; // total no. of bqs awakened
3148 debugBelch("## BQ Stats of %p: [%d entries] %s\n",
3149 node, len, (bqe!=END_BQ_QUEUE) ? "RBH" : ""));
3151 #elif defined(PARALLEL_HASKELL)
3153 awakenBlockedQueue(StgBlockingQueueElement *q, StgClosure *node)
3155 StgBlockingQueueElement *bqe;
3157 ACQUIRE_LOCK(&sched_mutex);
3159 IF_PAR_DEBUG(verbose,
3160 debugBelch("##-_ AwBQ for node %p on [%x]: \n",
3164 if(get_itbl(q)->type == CONSTR || q==END_BQ_QUEUE) {
3165 IF_PAR_DEBUG(verbose, debugBelch("## ... nothing to unblock so lets just return. RFP (BUG?)\n"));
3170 ASSERT(q == END_BQ_QUEUE ||
3171 get_itbl(q)->type == TSO ||
3172 get_itbl(q)->type == BLOCKED_FETCH ||
3173 get_itbl(q)->type == CONSTR);
3176 while (get_itbl(bqe)->type==TSO ||
3177 get_itbl(bqe)->type==BLOCKED_FETCH) {
3178 bqe = unblockOneLocked(bqe, node);
3180 RELEASE_LOCK(&sched_mutex);
3183 #else /* !GRAN && !PARALLEL_HASKELL */
3186 awakenBlockedQueueNoLock(StgTSO *tso)
3188 while (tso != END_TSO_QUEUE) {
3189 tso = unblockOneLocked(tso);
3194 awakenBlockedQueue(StgTSO *tso)
3196 ACQUIRE_LOCK(&sched_mutex);
3197 while (tso != END_TSO_QUEUE) {
3198 tso = unblockOneLocked(tso);
3200 RELEASE_LOCK(&sched_mutex);
3204 /* ---------------------------------------------------------------------------
3206 - usually called inside a signal handler so it mustn't do anything fancy.
3207 ------------------------------------------------------------------------ */
3210 interruptStgRts(void)
3216 /* -----------------------------------------------------------------------------
3219 This is for use when we raise an exception in another thread, which
3221 This has nothing to do with the UnblockThread event in GranSim. -- HWL
3222 -------------------------------------------------------------------------- */
3224 #if defined(GRAN) || defined(PARALLEL_HASKELL)
3226 NB: only the type of the blocking queue is different in GranSim and GUM
3227 the operations on the queue-elements are the same
3228 long live polymorphism!
3230 Locks: sched_mutex is held upon entry and exit.
3234 unblockThread(StgTSO *tso)
3236 StgBlockingQueueElement *t, **last;
3238 switch (tso->why_blocked) {
3241 return; /* not blocked */
3244 // Be careful: nothing to do here! We tell the scheduler that the thread
3245 // is runnable and we leave it to the stack-walking code to abort the
3246 // transaction while unwinding the stack. We should perhaps have a debugging
3247 // test to make sure that this really happens and that the 'zombie' transaction
3248 // does not get committed.
3252 ASSERT(get_itbl(tso->block_info.closure)->type == MVAR);
3254 StgBlockingQueueElement *last_tso = END_BQ_QUEUE;
3255 StgMVar *mvar = (StgMVar *)(tso->block_info.closure);
3257 last = (StgBlockingQueueElement **)&mvar->head;
3258 for (t = (StgBlockingQueueElement *)mvar->head;
3260 last = &t->link, last_tso = t, t = t->link) {
3261 if (t == (StgBlockingQueueElement *)tso) {
3262 *last = (StgBlockingQueueElement *)tso->link;
3263 if (mvar->tail == tso) {
3264 mvar->tail = (StgTSO *)last_tso;
3269 barf("unblockThread (MVAR): TSO not found");
3272 case BlockedOnBlackHole:
3273 ASSERT(get_itbl(tso->block_info.closure)->type == BLACKHOLE_BQ);
3275 StgBlockingQueue *bq = (StgBlockingQueue *)(tso->block_info.closure);
3277 last = &bq->blocking_queue;
3278 for (t = bq->blocking_queue;
3280 last = &t->link, t = t->link) {
3281 if (t == (StgBlockingQueueElement *)tso) {
3282 *last = (StgBlockingQueueElement *)tso->link;
3286 barf("unblockThread (BLACKHOLE): TSO not found");
3289 case BlockedOnException:
3291 StgTSO *target = tso->block_info.tso;
3293 ASSERT(get_itbl(target)->type == TSO);
3295 if (target->what_next == ThreadRelocated) {
3296 target = target->link;
3297 ASSERT(get_itbl(target)->type == TSO);
3300 ASSERT(target->blocked_exceptions != NULL);
3302 last = (StgBlockingQueueElement **)&target->blocked_exceptions;
3303 for (t = (StgBlockingQueueElement *)target->blocked_exceptions;
3305 last = &t->link, t = t->link) {
3306 ASSERT(get_itbl(t)->type == TSO);
3307 if (t == (StgBlockingQueueElement *)tso) {
3308 *last = (StgBlockingQueueElement *)tso->link;
3312 barf("unblockThread (Exception): TSO not found");
3316 case BlockedOnWrite:
3317 #if defined(mingw32_HOST_OS)
3318 case BlockedOnDoProc:
3321 /* take TSO off blocked_queue */
3322 StgBlockingQueueElement *prev = NULL;
3323 for (t = (StgBlockingQueueElement *)blocked_queue_hd; t != END_BQ_QUEUE;
3324 prev = t, t = t->link) {
3325 if (t == (StgBlockingQueueElement *)tso) {
3327 blocked_queue_hd = (StgTSO *)t->link;
3328 if ((StgBlockingQueueElement *)blocked_queue_tl == t) {
3329 blocked_queue_tl = END_TSO_QUEUE;
3332 prev->link = t->link;
3333 if ((StgBlockingQueueElement *)blocked_queue_tl == t) {
3334 blocked_queue_tl = (StgTSO *)prev;
3340 barf("unblockThread (I/O): TSO not found");
3343 case BlockedOnDelay:
3345 /* take TSO off sleeping_queue */
3346 StgBlockingQueueElement *prev = NULL;
3347 for (t = (StgBlockingQueueElement *)sleeping_queue; t != END_BQ_QUEUE;
3348 prev = t, t = t->link) {
3349 if (t == (StgBlockingQueueElement *)tso) {
3351 sleeping_queue = (StgTSO *)t->link;
3353 prev->link = t->link;
3358 barf("unblockThread (delay): TSO not found");
3362 barf("unblockThread");
3366 tso->link = END_TSO_QUEUE;
3367 tso->why_blocked = NotBlocked;
3368 tso->block_info.closure = NULL;
3369 PUSH_ON_RUN_QUEUE(tso);
3373 unblockThread(StgTSO *tso)
3377 /* To avoid locking unnecessarily. */
3378 if (tso->why_blocked == NotBlocked) {
3382 switch (tso->why_blocked) {
3385 // Be careful: nothing to do here! We tell the scheduler that the thread
3386 // is runnable and we leave it to the stack-walking code to abort the
3387 // transaction while unwinding the stack. We should perhaps have a debugging
3388 // test to make sure that this really happens and that the 'zombie' transaction
3389 // does not get committed.
3393 ASSERT(get_itbl(tso->block_info.closure)->type == MVAR);
3395 StgTSO *last_tso = END_TSO_QUEUE;
3396 StgMVar *mvar = (StgMVar *)(tso->block_info.closure);
3399 for (t = mvar->head; t != END_TSO_QUEUE;
3400 last = &t->link, last_tso = t, t = t->link) {
3403 if (mvar->tail == tso) {
3404 mvar->tail = last_tso;
3409 barf("unblockThread (MVAR): TSO not found");
3412 case BlockedOnBlackHole:
3414 last = &blackhole_queue;
3415 for (t = blackhole_queue; t != END_TSO_QUEUE;
3416 last = &t->link, t = t->link) {
3422 barf("unblockThread (BLACKHOLE): TSO not found");
3425 case BlockedOnException:
3427 StgTSO *target = tso->block_info.tso;
3429 ASSERT(get_itbl(target)->type == TSO);
3431 while (target->what_next == ThreadRelocated) {
3432 target = target->link;
3433 ASSERT(get_itbl(target)->type == TSO);
3436 ASSERT(target->blocked_exceptions != NULL);
3438 last = &target->blocked_exceptions;
3439 for (t = target->blocked_exceptions; t != END_TSO_QUEUE;
3440 last = &t->link, t = t->link) {
3441 ASSERT(get_itbl(t)->type == TSO);
3447 barf("unblockThread (Exception): TSO not found");
3451 case BlockedOnWrite:
3452 #if defined(mingw32_HOST_OS)
3453 case BlockedOnDoProc:
3456 StgTSO *prev = NULL;
3457 for (t = blocked_queue_hd; t != END_TSO_QUEUE;
3458 prev = t, t = t->link) {
3461 blocked_queue_hd = t->link;
3462 if (blocked_queue_tl == t) {
3463 blocked_queue_tl = END_TSO_QUEUE;
3466 prev->link = t->link;
3467 if (blocked_queue_tl == t) {
3468 blocked_queue_tl = prev;
3474 barf("unblockThread (I/O): TSO not found");
3477 case BlockedOnDelay:
3479 StgTSO *prev = NULL;
3480 for (t = sleeping_queue; t != END_TSO_QUEUE;
3481 prev = t, t = t->link) {
3484 sleeping_queue = t->link;
3486 prev->link = t->link;
3491 barf("unblockThread (delay): TSO not found");
3495 barf("unblockThread");
3499 tso->link = END_TSO_QUEUE;
3500 tso->why_blocked = NotBlocked;
3501 tso->block_info.closure = NULL;
3502 APPEND_TO_RUN_QUEUE(tso);
3506 /* -----------------------------------------------------------------------------
3509 * Check the blackhole_queue for threads that can be woken up. We do
3510 * this periodically: before every GC, and whenever the run queue is
3513 * An elegant solution might be to just wake up all the blocked
3514 * threads with awakenBlockedQueue occasionally: they'll go back to
3515 * sleep again if the object is still a BLACKHOLE. Unfortunately this
3516 * doesn't give us a way to tell whether we've actually managed to
3517 * wake up any threads, so we would be busy-waiting.
3519 * -------------------------------------------------------------------------- */
3522 checkBlackHoles( void )
3525 rtsBool any_woke_up = rtsFalse;
3528 IF_DEBUG(scheduler, sched_belch("checking threads blocked on black holes"));
3530 // ASSUMES: sched_mutex
3531 prev = &blackhole_queue;
3532 t = blackhole_queue;
3533 while (t != END_TSO_QUEUE) {
3534 ASSERT(t->why_blocked == BlockedOnBlackHole);
3535 type = get_itbl(t->block_info.closure)->type;
3536 if (type != BLACKHOLE && type != CAF_BLACKHOLE) {
3537 t = unblockOneLocked(t);
3539 any_woke_up = rtsTrue;
3549 /* -----------------------------------------------------------------------------
3552 * The following function implements the magic for raising an
3553 * asynchronous exception in an existing thread.
3555 * We first remove the thread from any queue on which it might be
3556 * blocked. The possible blockages are MVARs and BLACKHOLE_BQs.
3558 * We strip the stack down to the innermost CATCH_FRAME, building
3559 * thunks in the heap for all the active computations, so they can
3560 * be restarted if necessary. When we reach a CATCH_FRAME, we build
3561 * an application of the handler to the exception, and push it on
3562 * the top of the stack.
3564 * How exactly do we save all the active computations? We create an
3565 * AP_STACK for every UpdateFrame on the stack. Entering one of these
3566 * AP_STACKs pushes everything from the corresponding update frame
3567 * upwards onto the stack. (Actually, it pushes everything up to the
3568 * next update frame plus a pointer to the next AP_STACK object.
3569 * Entering the next AP_STACK object pushes more onto the stack until we
3570 * reach the last AP_STACK object - at which point the stack should look
3571 * exactly as it did when we killed the TSO and we can continue
3572 * execution by entering the closure on top of the stack.
3574 * We can also kill a thread entirely - this happens if either (a) the
3575 * exception passed to raiseAsync is NULL, or (b) there's no
3576 * CATCH_FRAME on the stack. In either case, we strip the entire
3577 * stack and replace the thread with a zombie.
3579 * Locks: sched_mutex held upon entry nor exit.
3581 * -------------------------------------------------------------------------- */
3584 deleteThread(StgTSO *tso)
3586 if (tso->why_blocked != BlockedOnCCall &&
3587 tso->why_blocked != BlockedOnCCall_NoUnblockExc) {
3588 raiseAsync(tso,NULL);
3592 #ifdef FORKPROCESS_PRIMOP_SUPPORTED
3594 deleteThreadImmediately(StgTSO *tso)
3595 { // for forkProcess only:
3596 // delete thread without giving it a chance to catch the KillThread exception
3598 if (tso->what_next == ThreadComplete || tso->what_next == ThreadKilled) {
3602 if (tso->why_blocked != BlockedOnCCall &&
3603 tso->why_blocked != BlockedOnCCall_NoUnblockExc) {
3607 tso->what_next = ThreadKilled;
3612 raiseAsyncWithLock(StgTSO *tso, StgClosure *exception)
3614 /* When raising async exs from contexts where sched_mutex isn't held;
3615 use raiseAsyncWithLock(). */
3616 ACQUIRE_LOCK(&sched_mutex);
3617 raiseAsync(tso,exception);
3618 RELEASE_LOCK(&sched_mutex);
3622 raiseAsync(StgTSO *tso, StgClosure *exception)
3624 raiseAsync_(tso, exception, rtsFalse);
3628 raiseAsync_(StgTSO *tso, StgClosure *exception, rtsBool stop_at_atomically)
3630 StgRetInfoTable *info;
3633 // Thread already dead?
3634 if (tso->what_next == ThreadComplete || tso->what_next == ThreadKilled) {
3639 sched_belch("raising exception in thread %ld.", (long)tso->id));
3641 // Remove it from any blocking queues
3646 // The stack freezing code assumes there's a closure pointer on
3647 // the top of the stack, so we have to arrange that this is the case...
3649 if (sp[0] == (W_)&stg_enter_info) {
3653 sp[0] = (W_)&stg_dummy_ret_closure;
3659 // 1. Let the top of the stack be the "current closure"
3661 // 2. Walk up the stack until we find either an UPDATE_FRAME or a
3664 // 3. If it's an UPDATE_FRAME, then make an AP_STACK containing the
3665 // current closure applied to the chunk of stack up to (but not
3666 // including) the update frame. This closure becomes the "current
3667 // closure". Go back to step 2.
3669 // 4. If it's a CATCH_FRAME, then leave the exception handler on
3670 // top of the stack applied to the exception.
3672 // 5. If it's a STOP_FRAME, then kill the thread.
3674 // NB: if we pass an ATOMICALLY_FRAME then abort the associated
3681 info = get_ret_itbl((StgClosure *)frame);
3683 while (info->i.type != UPDATE_FRAME
3684 && (info->i.type != CATCH_FRAME || exception == NULL)
3685 && info->i.type != STOP_FRAME
3686 && (info->i.type != ATOMICALLY_FRAME || stop_at_atomically == rtsFalse))
3688 if (info->i.type == CATCH_RETRY_FRAME || info->i.type == ATOMICALLY_FRAME) {
3689 // IF we find an ATOMICALLY_FRAME then we abort the
3690 // current transaction and propagate the exception. In
3691 // this case (unlike ordinary exceptions) we do not care
3692 // whether the transaction is valid or not because its
3693 // possible validity cannot have caused the exception
3694 // and will not be visible after the abort.
3696 debugBelch("Found atomically block delivering async exception\n"));
3697 stmAbortTransaction(tso -> trec);
3698 tso -> trec = stmGetEnclosingTRec(tso -> trec);
3700 frame += stack_frame_sizeW((StgClosure *)frame);
3701 info = get_ret_itbl((StgClosure *)frame);
3704 switch (info->i.type) {
3706 case ATOMICALLY_FRAME:
3707 ASSERT(stop_at_atomically);
3708 ASSERT(stmGetEnclosingTRec(tso->trec) == NO_TREC);
3709 stmCondemnTransaction(tso -> trec);
3713 // R1 is not a register: the return convention for IO in
3714 // this case puts the return value on the stack, so we
3715 // need to set up the stack to return to the atomically
3716 // frame properly...
3717 tso->sp = frame - 2;
3718 tso->sp[1] = (StgWord) &stg_NO_FINALIZER_closure; // why not?
3719 tso->sp[0] = (StgWord) &stg_ut_1_0_unreg_info;
3721 tso->what_next = ThreadRunGHC;
3725 // If we find a CATCH_FRAME, and we've got an exception to raise,
3726 // then build the THUNK raise(exception), and leave it on
3727 // top of the CATCH_FRAME ready to enter.
3731 StgCatchFrame *cf = (StgCatchFrame *)frame;
3735 // we've got an exception to raise, so let's pass it to the
3736 // handler in this frame.
3738 raise = (StgClosure *)allocate(sizeofW(StgClosure)+1);
3739 TICK_ALLOC_SE_THK(1,0);
3740 SET_HDR(raise,&stg_raise_info,cf->header.prof.ccs);
3741 raise->payload[0] = exception;
3743 // throw away the stack from Sp up to the CATCH_FRAME.
3747 /* Ensure that async excpetions are blocked now, so we don't get
3748 * a surprise exception before we get around to executing the
3751 if (tso->blocked_exceptions == NULL) {
3752 tso->blocked_exceptions = END_TSO_QUEUE;
3755 /* Put the newly-built THUNK on top of the stack, ready to execute
3756 * when the thread restarts.
3759 sp[-1] = (W_)&stg_enter_info;
3761 tso->what_next = ThreadRunGHC;
3762 IF_DEBUG(sanity, checkTSO(tso));
3771 // First build an AP_STACK consisting of the stack chunk above the
3772 // current update frame, with the top word on the stack as the
3775 words = frame - sp - 1;
3776 ap = (StgAP_STACK *)allocate(PAP_sizeW(words));
3779 ap->fun = (StgClosure *)sp[0];
3781 for(i=0; i < (nat)words; ++i) {
3782 ap->payload[i] = (StgClosure *)*sp++;
3785 SET_HDR(ap,&stg_AP_STACK_info,
3786 ((StgClosure *)frame)->header.prof.ccs /* ToDo */);
3787 TICK_ALLOC_UP_THK(words+1,0);
3790 debugBelch("sched: Updating ");
3791 printPtr((P_)((StgUpdateFrame *)frame)->updatee);
3792 debugBelch(" with ");
3793 printObj((StgClosure *)ap);
3796 // Replace the updatee with an indirection - happily
3797 // this will also wake up any threads currently
3798 // waiting on the result.
3800 // Warning: if we're in a loop, more than one update frame on
3801 // the stack may point to the same object. Be careful not to
3802 // overwrite an IND_OLDGEN in this case, because we'll screw
3803 // up the mutable lists. To be on the safe side, don't
3804 // overwrite any kind of indirection at all. See also
3805 // threadSqueezeStack in GC.c, where we have to make a similar
3808 if (!closure_IND(((StgUpdateFrame *)frame)->updatee)) {
3809 // revert the black hole
3810 UPD_IND_NOLOCK(((StgUpdateFrame *)frame)->updatee,
3813 sp += sizeofW(StgUpdateFrame) - 1;
3814 sp[0] = (W_)ap; // push onto stack
3819 // We've stripped the entire stack, the thread is now dead.
3820 sp += sizeofW(StgStopFrame);
3821 tso->what_next = ThreadKilled;
3832 /* -----------------------------------------------------------------------------
3833 raiseExceptionHelper
3835 This function is called by the raise# primitve, just so that we can
3836 move some of the tricky bits of raising an exception from C-- into
3837 C. Who knows, it might be a useful re-useable thing here too.
3838 -------------------------------------------------------------------------- */
3841 raiseExceptionHelper (StgTSO *tso, StgClosure *exception)
3843 StgClosure *raise_closure = NULL;
3845 StgRetInfoTable *info;
3847 // This closure represents the expression 'raise# E' where E
3848 // is the exception raise. It is used to overwrite all the
3849 // thunks which are currently under evaluataion.
3853 // LDV profiling: stg_raise_info has THUNK as its closure
3854 // type. Since a THUNK takes at least MIN_UPD_SIZE words in its
3855 // payload, MIN_UPD_SIZE is more approprate than 1. It seems that
3856 // 1 does not cause any problem unless profiling is performed.
3857 // However, when LDV profiling goes on, we need to linearly scan
3858 // small object pool, where raise_closure is stored, so we should
3859 // use MIN_UPD_SIZE.
3861 // raise_closure = (StgClosure *)RET_STGCALL1(P_,allocate,
3862 // sizeofW(StgClosure)+1);
3866 // Walk up the stack, looking for the catch frame. On the way,
3867 // we update any closures pointed to from update frames with the
3868 // raise closure that we just built.
3872 info = get_ret_itbl((StgClosure *)p);
3873 next = p + stack_frame_sizeW((StgClosure *)p);
3874 switch (info->i.type) {
3877 // Only create raise_closure if we need to.
3878 if (raise_closure == NULL) {
3880 (StgClosure *)allocate(sizeofW(StgClosure)+MIN_UPD_SIZE);
3881 SET_HDR(raise_closure, &stg_raise_info, CCCS);
3882 raise_closure->payload[0] = exception;
3884 UPD_IND(((StgUpdateFrame *)p)->updatee,raise_closure);
3888 case ATOMICALLY_FRAME:
3889 IF_DEBUG(stm, debugBelch("Found ATOMICALLY_FRAME at %p\n", p));
3891 return ATOMICALLY_FRAME;
3897 case CATCH_STM_FRAME:
3898 IF_DEBUG(stm, debugBelch("Found CATCH_STM_FRAME at %p\n", p));
3900 return CATCH_STM_FRAME;
3906 case CATCH_RETRY_FRAME:
3915 /* -----------------------------------------------------------------------------
3916 findRetryFrameHelper
3918 This function is called by the retry# primitive. It traverses the stack
3919 leaving tso->sp referring to the frame which should handle the retry.
3921 This should either be a CATCH_RETRY_FRAME (if the retry# is within an orElse#)
3922 or should be a ATOMICALLY_FRAME (if the retry# reaches the top level).
3924 We skip CATCH_STM_FRAMEs because retries are not considered to be exceptions,
3925 despite the similar implementation.
3927 We should not expect to see CATCH_FRAME or STOP_FRAME because those should
3928 not be created within memory transactions.
3929 -------------------------------------------------------------------------- */
3932 findRetryFrameHelper (StgTSO *tso)
3935 StgRetInfoTable *info;
3939 info = get_ret_itbl((StgClosure *)p);
3940 next = p + stack_frame_sizeW((StgClosure *)p);
3941 switch (info->i.type) {
3943 case ATOMICALLY_FRAME:
3944 IF_DEBUG(stm, debugBelch("Found ATOMICALLY_FRAME at %p during retrry\n", p));
3946 return ATOMICALLY_FRAME;
3948 case CATCH_RETRY_FRAME:
3949 IF_DEBUG(stm, debugBelch("Found CATCH_RETRY_FRAME at %p during retrry\n", p));
3951 return CATCH_RETRY_FRAME;
3953 case CATCH_STM_FRAME:
3955 ASSERT(info->i.type != CATCH_FRAME);
3956 ASSERT(info->i.type != STOP_FRAME);
3963 /* -----------------------------------------------------------------------------
3964 resurrectThreads is called after garbage collection on the list of
3965 threads found to be garbage. Each of these threads will be woken
3966 up and sent a signal: BlockedOnDeadMVar if the thread was blocked
3967 on an MVar, or NonTermination if the thread was blocked on a Black
3970 Locks: sched_mutex isn't held upon entry nor exit.
3971 -------------------------------------------------------------------------- */
3974 resurrectThreads( StgTSO *threads )
3978 for (tso = threads; tso != END_TSO_QUEUE; tso = next) {
3979 next = tso->global_link;
3980 tso->global_link = all_threads;
3982 IF_DEBUG(scheduler, sched_belch("resurrecting thread %d", tso->id));
3984 switch (tso->why_blocked) {
3986 case BlockedOnException:
3987 /* Called by GC - sched_mutex lock is currently held. */
3988 raiseAsync(tso,(StgClosure *)BlockedOnDeadMVar_closure);
3990 case BlockedOnBlackHole:
3991 raiseAsync(tso,(StgClosure *)NonTermination_closure);
3994 raiseAsync(tso,(StgClosure *)BlockedIndefinitely_closure);
3997 /* This might happen if the thread was blocked on a black hole
3998 * belonging to a thread that we've just woken up (raiseAsync
3999 * can wake up threads, remember...).
4003 barf("resurrectThreads: thread blocked in a strange way");
4008 /* ----------------------------------------------------------------------------
4009 * Debugging: why is a thread blocked
4010 * [Also provides useful information when debugging threaded programs
4011 * at the Haskell source code level, so enable outside of DEBUG. --sof 7/02]
4012 ------------------------------------------------------------------------- */
4015 printThreadBlockage(StgTSO *tso)
4017 switch (tso->why_blocked) {
4019 debugBelch("is blocked on read from fd %ld", tso->block_info.fd);
4021 case BlockedOnWrite:
4022 debugBelch("is blocked on write to fd %ld", tso->block_info.fd);
4024 #if defined(mingw32_HOST_OS)
4025 case BlockedOnDoProc:
4026 debugBelch("is blocked on proc (request: %ld)", tso->block_info.async_result->reqID);
4029 case BlockedOnDelay:
4030 debugBelch("is blocked until %ld", tso->block_info.target);
4033 debugBelch("is blocked on an MVar");
4035 case BlockedOnException:
4036 debugBelch("is blocked on delivering an exception to thread %d",
4037 tso->block_info.tso->id);
4039 case BlockedOnBlackHole:
4040 debugBelch("is blocked on a black hole");
4043 debugBelch("is not blocked");
4045 #if defined(PARALLEL_HASKELL)
4047 debugBelch("is blocked on global address; local FM_BQ is %p (%s)",
4048 tso->block_info.closure, info_type(tso->block_info.closure));
4050 case BlockedOnGA_NoSend:
4051 debugBelch("is blocked on global address (no send); local FM_BQ is %p (%s)",
4052 tso->block_info.closure, info_type(tso->block_info.closure));
4055 case BlockedOnCCall:
4056 debugBelch("is blocked on an external call");
4058 case BlockedOnCCall_NoUnblockExc:
4059 debugBelch("is blocked on an external call (exceptions were already blocked)");
4062 debugBelch("is blocked on an STM operation");
4065 barf("printThreadBlockage: strange tso->why_blocked: %d for TSO %d (%d)",
4066 tso->why_blocked, tso->id, tso);
4071 printThreadStatus(StgTSO *tso)
4073 switch (tso->what_next) {
4075 debugBelch("has been killed");
4077 case ThreadComplete:
4078 debugBelch("has completed");
4081 printThreadBlockage(tso);
4086 printAllThreads(void)
4091 char time_string[TIME_STR_LEN], node_str[NODE_STR_LEN];
4092 ullong_format_string(TIME_ON_PROC(CurrentProc),
4093 time_string, rtsFalse/*no commas!*/);
4095 debugBelch("all threads at [%s]:\n", time_string);
4096 # elif defined(PARALLEL_HASKELL)
4097 char time_string[TIME_STR_LEN], node_str[NODE_STR_LEN];
4098 ullong_format_string(CURRENT_TIME,
4099 time_string, rtsFalse/*no commas!*/);
4101 debugBelch("all threads at [%s]:\n", time_string);
4103 debugBelch("all threads:\n");
4106 for (t = all_threads; t != END_TSO_QUEUE; t = t->global_link) {
4107 debugBelch("\tthread %d @ %p ", t->id, (void *)t);
4110 void *label = lookupThreadLabel(t->id);
4111 if (label) debugBelch("[\"%s\"] ",(char *)label);
4114 printThreadStatus(t);
4122 Print a whole blocking queue attached to node (debugging only).
4124 # if defined(PARALLEL_HASKELL)
4126 print_bq (StgClosure *node)
4128 StgBlockingQueueElement *bqe;
4132 debugBelch("## BQ of closure %p (%s): ",
4133 node, info_type(node));
4135 /* should cover all closures that may have a blocking queue */
4136 ASSERT(get_itbl(node)->type == BLACKHOLE_BQ ||
4137 get_itbl(node)->type == FETCH_ME_BQ ||
4138 get_itbl(node)->type == RBH ||
4139 get_itbl(node)->type == MVAR);
4141 ASSERT(node!=(StgClosure*)NULL); // sanity check
4143 print_bqe(((StgBlockingQueue*)node)->blocking_queue);
4147 Print a whole blocking queue starting with the element bqe.
4150 print_bqe (StgBlockingQueueElement *bqe)
4155 NB: In a parallel setup a BQ of an RBH must end with an RBH_Save closure;
4157 for (end = (bqe==END_BQ_QUEUE);
4158 !end; // iterate until bqe points to a CONSTR
4159 end = (get_itbl(bqe)->type == CONSTR) || (bqe->link==END_BQ_QUEUE),
4160 bqe = end ? END_BQ_QUEUE : bqe->link) {
4161 ASSERT(bqe != END_BQ_QUEUE); // sanity check
4162 ASSERT(bqe != (StgBlockingQueueElement *)NULL); // sanity check
4163 /* types of closures that may appear in a blocking queue */
4164 ASSERT(get_itbl(bqe)->type == TSO ||
4165 get_itbl(bqe)->type == BLOCKED_FETCH ||
4166 get_itbl(bqe)->type == CONSTR);
4167 /* only BQs of an RBH end with an RBH_Save closure */
4168 //ASSERT(get_itbl(bqe)->type != CONSTR || get_itbl(node)->type == RBH);
4170 switch (get_itbl(bqe)->type) {
4172 debugBelch(" TSO %u (%x),",
4173 ((StgTSO *)bqe)->id, ((StgTSO *)bqe));
4176 debugBelch(" BF (node=%p, ga=((%x, %d, %x)),",
4177 ((StgBlockedFetch *)bqe)->node,
4178 ((StgBlockedFetch *)bqe)->ga.payload.gc.gtid,
4179 ((StgBlockedFetch *)bqe)->ga.payload.gc.slot,
4180 ((StgBlockedFetch *)bqe)->ga.weight);
4183 debugBelch(" %s (IP %p),",
4184 (get_itbl(bqe) == &stg_RBH_Save_0_info ? "RBH_Save_0" :
4185 get_itbl(bqe) == &stg_RBH_Save_1_info ? "RBH_Save_1" :
4186 get_itbl(bqe) == &stg_RBH_Save_2_info ? "RBH_Save_2" :
4187 "RBH_Save_?"), get_itbl(bqe));
4190 barf("Unexpected closure type %s in blocking queue", // of %p (%s)",
4191 info_type((StgClosure *)bqe)); // , node, info_type(node));
4197 # elif defined(GRAN)
4199 print_bq (StgClosure *node)
4201 StgBlockingQueueElement *bqe;
4202 PEs node_loc, tso_loc;
4205 /* should cover all closures that may have a blocking queue */
4206 ASSERT(get_itbl(node)->type == BLACKHOLE_BQ ||
4207 get_itbl(node)->type == FETCH_ME_BQ ||
4208 get_itbl(node)->type == RBH);
4210 ASSERT(node!=(StgClosure*)NULL); // sanity check
4211 node_loc = where_is(node);
4213 debugBelch("## BQ of closure %p (%s) on [PE %d]: ",
4214 node, info_type(node), node_loc);
4217 NB: In a parallel setup a BQ of an RBH must end with an RBH_Save closure;
4219 for (bqe = ((StgBlockingQueue*)node)->blocking_queue, end = (bqe==END_BQ_QUEUE);
4220 !end; // iterate until bqe points to a CONSTR
4221 end = (get_itbl(bqe)->type == CONSTR) || (bqe->link==END_BQ_QUEUE), bqe = end ? END_BQ_QUEUE : bqe->link) {
4222 ASSERT(bqe != END_BQ_QUEUE); // sanity check
4223 ASSERT(bqe != (StgBlockingQueueElement *)NULL); // sanity check
4224 /* types of closures that may appear in a blocking queue */
4225 ASSERT(get_itbl(bqe)->type == TSO ||
4226 get_itbl(bqe)->type == CONSTR);
4227 /* only BQs of an RBH end with an RBH_Save closure */
4228 ASSERT(get_itbl(bqe)->type != CONSTR || get_itbl(node)->type == RBH);
4230 tso_loc = where_is((StgClosure *)bqe);
4231 switch (get_itbl(bqe)->type) {
4233 debugBelch(" TSO %d (%p) on [PE %d],",
4234 ((StgTSO *)bqe)->id, (StgTSO *)bqe, tso_loc);
4237 debugBelch(" %s (IP %p),",
4238 (get_itbl(bqe) == &stg_RBH_Save_0_info ? "RBH_Save_0" :
4239 get_itbl(bqe) == &stg_RBH_Save_1_info ? "RBH_Save_1" :
4240 get_itbl(bqe) == &stg_RBH_Save_2_info ? "RBH_Save_2" :
4241 "RBH_Save_?"), get_itbl(bqe));
4244 barf("Unexpected closure type %s in blocking queue of %p (%s)",
4245 info_type((StgClosure *)bqe), node, info_type(node));
4253 #if defined(PARALLEL_HASKELL)
4260 for (i=0, tso=run_queue_hd;
4261 tso != END_TSO_QUEUE;
4270 sched_belch(char *s, ...)
4274 #ifdef RTS_SUPPORTS_THREADS
4275 debugBelch("sched (task %p): ", osThreadId());
4276 #elif defined(PARALLEL_HASKELL)
4279 debugBelch("sched: ");