1 /* ---------------------------------------------------------------------------
3 * (c) The GHC Team, 1998-2005
5 * The scheduler and thread-related functionality
7 * --------------------------------------------------------------------------*/
9 #include "PosixSource.h"
14 #include "BlockAlloc.h"
15 #include "OSThreads.h"
20 #include "StgMiscClosures.h"
21 #include "Interpreter.h"
22 #include "Exception.h"
24 #include "RtsSignals.h"
30 #include "ThreadLabels.h"
31 #include "LdvProfile.h"
34 #include "Proftimer.h"
37 #if defined(GRAN) || defined(PARALLEL_HASKELL)
38 # include "GranSimRts.h"
40 # include "ParallelRts.h"
41 # include "Parallel.h"
42 # include "ParallelDebug.h"
47 #include "Capability.h"
49 #include "AwaitEvent.h"
50 #if defined(mingw32_HOST_OS)
51 #include "win32/IOManager.h"
54 #ifdef HAVE_SYS_TYPES_H
55 #include <sys/types.h>
69 // Turn off inlining when debugging - it obfuscates things
72 # define STATIC_INLINE static
75 /* -----------------------------------------------------------------------------
77 * -------------------------------------------------------------------------- */
81 StgTSO* ActiveTSO = NULL; /* for assigning system costs; GranSim-Light only */
82 /* rtsTime TimeOfNextEvent, EndOfTimeSlice; now in GranSim.c */
85 In GranSim we have a runnable and a blocked queue for each processor.
86 In order to minimise code changes new arrays run_queue_hds/tls
87 are created. run_queue_hd is then a short cut (macro) for
88 run_queue_hds[CurrentProc] (see GranSim.h).
91 StgTSO *run_queue_hds[MAX_PROC], *run_queue_tls[MAX_PROC];
92 StgTSO *blocked_queue_hds[MAX_PROC], *blocked_queue_tls[MAX_PROC];
93 StgTSO *ccalling_threadss[MAX_PROC];
94 /* We use the same global list of threads (all_threads) in GranSim as in
95 the std RTS (i.e. we are cheating). However, we don't use this list in
96 the GranSim specific code at the moment (so we are only potentially
101 #if !defined(THREADED_RTS)
102 // Blocked/sleeping thrads
103 StgTSO *blocked_queue_hd = NULL;
104 StgTSO *blocked_queue_tl = NULL;
105 StgTSO *sleeping_queue = NULL; // perhaps replace with a hash table?
108 /* Threads blocked on blackholes.
109 * LOCK: sched_mutex+capability, or all capabilities
111 StgTSO *blackhole_queue = NULL;
114 /* The blackhole_queue should be checked for threads to wake up. See
115 * Schedule.h for more thorough comment.
116 * LOCK: none (doesn't matter if we miss an update)
118 rtsBool blackholes_need_checking = rtsFalse;
120 /* Linked list of all threads.
121 * Used for detecting garbage collected threads.
122 * LOCK: sched_mutex+capability, or all capabilities
124 StgTSO *all_threads = NULL;
126 /* flag set by signal handler to precipitate a context switch
127 * LOCK: none (just an advisory flag)
129 int context_switch = 0;
131 /* flag that tracks whether we have done any execution in this time slice.
132 * LOCK: currently none, perhaps we should lock (but needs to be
133 * updated in the fast path of the scheduler).
135 nat recent_activity = ACTIVITY_YES;
137 /* if this flag is set as well, give up execution
138 * LOCK: none (changes once, from false->true)
140 rtsBool interrupted = rtsFalse;
142 /* Next thread ID to allocate.
145 static StgThreadID next_thread_id = 1;
147 /* The smallest stack size that makes any sense is:
148 * RESERVED_STACK_WORDS (so we can get back from the stack overflow)
149 * + sizeofW(StgStopFrame) (the stg_stop_thread_info frame)
150 * + 1 (the closure to enter)
152 * + 1 (spare slot req'd by stg_ap_v_ret)
154 * A thread with this stack will bomb immediately with a stack
155 * overflow, which will increase its stack size.
157 #define MIN_STACK_WORDS (RESERVED_STACK_WORDS + sizeofW(StgStopFrame) + 3)
163 /* This is used in `TSO.h' and gcc 2.96 insists that this variable actually
164 * exists - earlier gccs apparently didn't.
170 * Set to TRUE when entering a shutdown state (via shutdownHaskellAndExit()) --
171 * in an MT setting, needed to signal that a worker thread shouldn't hang around
172 * in the scheduler when it is out of work.
174 rtsBool shutting_down_scheduler = rtsFalse;
177 * This mutex protects most of the global scheduler data in
178 * the THREADED_RTS runtime.
180 #if defined(THREADED_RTS)
184 #if defined(PARALLEL_HASKELL)
186 rtsTime TimeOfLastYield;
187 rtsBool emitSchedule = rtsTrue;
190 /* -----------------------------------------------------------------------------
191 * static function prototypes
192 * -------------------------------------------------------------------------- */
194 static Capability *schedule (Capability *initialCapability, Task *task);
197 // These function all encapsulate parts of the scheduler loop, and are
198 // abstracted only to make the structure and control flow of the
199 // scheduler clearer.
201 static void schedulePreLoop (void);
202 #if defined(THREADED_RTS)
203 static void schedulePushWork(Capability *cap, Task *task);
205 static void scheduleStartSignalHandlers (Capability *cap);
206 static void scheduleCheckBlockedThreads (Capability *cap);
207 static void scheduleCheckBlackHoles (Capability *cap);
208 static void scheduleDetectDeadlock (Capability *cap, Task *task);
210 static StgTSO *scheduleProcessEvent(rtsEvent *event);
212 #if defined(PARALLEL_HASKELL)
213 static StgTSO *scheduleSendPendingMessages(void);
214 static void scheduleActivateSpark(void);
215 static rtsBool scheduleGetRemoteWork(rtsBool *receivedFinish);
217 #if defined(PAR) || defined(GRAN)
218 static void scheduleGranParReport(void);
220 static void schedulePostRunThread(void);
221 static rtsBool scheduleHandleHeapOverflow( Capability *cap, StgTSO *t );
222 static void scheduleHandleStackOverflow( Capability *cap, Task *task,
224 static rtsBool scheduleHandleYield( Capability *cap, StgTSO *t,
225 nat prev_what_next );
226 static void scheduleHandleThreadBlocked( StgTSO *t );
227 static rtsBool scheduleHandleThreadFinished( Capability *cap, Task *task,
229 static rtsBool scheduleDoHeapProfile(rtsBool ready_to_gc);
230 static void scheduleDoGC(Capability *cap, Task *task, rtsBool force_major,
231 void (*get_roots)(evac_fn));
233 static void unblockThread(Capability *cap, StgTSO *tso);
234 static rtsBool checkBlackHoles(Capability *cap);
235 static void AllRoots(evac_fn evac);
237 static StgTSO *threadStackOverflow(Capability *cap, StgTSO *tso);
239 static void raiseAsync_(Capability *cap, StgTSO *tso, StgClosure *exception,
240 rtsBool stop_at_atomically, StgPtr stop_here);
242 static void deleteThread (Capability *cap, StgTSO *tso);
243 static void deleteRunQueue (Capability *cap);
246 static void printThreadBlockage(StgTSO *tso);
247 static void printThreadStatus(StgTSO *tso);
248 void printThreadQueue(StgTSO *tso);
251 #if defined(PARALLEL_HASKELL)
252 StgTSO * createSparkThread(rtsSpark spark);
253 StgTSO * activateSpark (rtsSpark spark);
257 static char *whatNext_strs[] = {
267 /* -----------------------------------------------------------------------------
268 * Putting a thread on the run queue: different scheduling policies
269 * -------------------------------------------------------------------------- */
272 addToRunQueue( Capability *cap, StgTSO *t )
274 #if defined(PARALLEL_HASKELL)
275 if (RtsFlags.ParFlags.doFairScheduling) {
276 // this does round-robin scheduling; good for concurrency
277 appendToRunQueue(cap,t);
279 // this does unfair scheduling; good for parallelism
280 pushOnRunQueue(cap,t);
283 // this does round-robin scheduling; good for concurrency
284 appendToRunQueue(cap,t);
288 /* ---------------------------------------------------------------------------
289 Main scheduling loop.
291 We use round-robin scheduling, each thread returning to the
292 scheduler loop when one of these conditions is detected:
295 * timer expires (thread yields)
301 In a GranSim setup this loop iterates over the global event queue.
302 This revolves around the global event queue, which determines what
303 to do next. Therefore, it's more complicated than either the
304 concurrent or the parallel (GUM) setup.
307 GUM iterates over incoming messages.
308 It starts with nothing to do (thus CurrentTSO == END_TSO_QUEUE),
309 and sends out a fish whenever it has nothing to do; in-between
310 doing the actual reductions (shared code below) it processes the
311 incoming messages and deals with delayed operations
312 (see PendingFetches).
313 This is not the ugliest code you could imagine, but it's bloody close.
315 ------------------------------------------------------------------------ */
318 schedule (Capability *initialCapability, Task *task)
322 StgThreadReturnCode ret;
325 #elif defined(PARALLEL_HASKELL)
328 rtsBool receivedFinish = rtsFalse;
330 nat tp_size, sp_size; // stats only
335 #if defined(THREADED_RTS)
336 rtsBool first = rtsTrue;
339 cap = initialCapability;
341 // Pre-condition: this task owns initialCapability.
342 // The sched_mutex is *NOT* held
343 // NB. on return, we still hold a capability.
346 sched_belch("### NEW SCHEDULER LOOP (task: %p, cap: %p)",
347 task, initialCapability);
352 // -----------------------------------------------------------
353 // Scheduler loop starts here:
355 #if defined(PARALLEL_HASKELL)
356 #define TERMINATION_CONDITION (!receivedFinish)
358 #define TERMINATION_CONDITION ((event = get_next_event()) != (rtsEvent*)NULL)
360 #define TERMINATION_CONDITION rtsTrue
363 while (TERMINATION_CONDITION) {
366 /* Choose the processor with the next event */
367 CurrentProc = event->proc;
368 CurrentTSO = event->tso;
371 #if defined(THREADED_RTS)
373 // don't yield the first time, we want a chance to run this
374 // thread for a bit, even if there are others banging at the
377 ASSERT_FULL_CAPABILITY_INVARIANTS(cap,task);
379 // Yield the capability to higher-priority tasks if necessary.
380 yieldCapability(&cap, task);
384 #if defined(THREADED_RTS)
385 schedulePushWork(cap,task);
388 // Check whether we have re-entered the RTS from Haskell without
389 // going via suspendThread()/resumeThread (i.e. a 'safe' foreign
391 if (cap->in_haskell) {
392 errorBelch("schedule: re-entered unsafely.\n"
393 " Perhaps a 'foreign import unsafe' should be 'safe'?");
394 stg_exit(EXIT_FAILURE);
398 // Test for interruption. If interrupted==rtsTrue, then either
399 // we received a keyboard interrupt (^C), or the scheduler is
400 // trying to shut down all the tasks (shutting_down_scheduler) in
405 #if defined(THREADED_RTS)
406 discardSparksCap(cap);
408 if (shutting_down_scheduler) {
409 IF_DEBUG(scheduler, sched_belch("shutting down"));
410 // If we are a worker, just exit. If we're a bound thread
411 // then we will exit below when we've removed our TSO from
413 if (task->tso == NULL && emptyRunQueue(cap)) {
417 IF_DEBUG(scheduler, sched_belch("interrupted"));
421 #if defined(THREADED_RTS)
422 // If the run queue is empty, take a spark and turn it into a thread.
424 if (emptyRunQueue(cap)) {
426 spark = findSpark(cap);
429 sched_belch("turning spark of closure %p into a thread",
430 (StgClosure *)spark));
431 createSparkThread(cap,spark);
435 #endif // THREADED_RTS
437 scheduleStartSignalHandlers(cap);
439 // Only check the black holes here if we've nothing else to do.
440 // During normal execution, the black hole list only gets checked
441 // at GC time, to avoid repeatedly traversing this possibly long
442 // list each time around the scheduler.
443 if (emptyRunQueue(cap)) { scheduleCheckBlackHoles(cap); }
445 scheduleCheckBlockedThreads(cap);
447 scheduleDetectDeadlock(cap,task);
449 // Normally, the only way we can get here with no threads to
450 // run is if a keyboard interrupt received during
451 // scheduleCheckBlockedThreads() or scheduleDetectDeadlock().
452 // Additionally, it is not fatal for the
453 // threaded RTS to reach here with no threads to run.
455 // win32: might be here due to awaitEvent() being abandoned
456 // as a result of a console event having been delivered.
457 if ( emptyRunQueue(cap) ) {
458 #if !defined(THREADED_RTS) && !defined(mingw32_HOST_OS)
461 continue; // nothing to do
464 #if defined(PARALLEL_HASKELL)
465 scheduleSendPendingMessages();
466 if (emptyRunQueue(cap) && scheduleActivateSpark())
470 ASSERT(next_fish_to_send_at==0); // i.e. no delayed fishes left!
473 /* If we still have no work we need to send a FISH to get a spark
475 if (emptyRunQueue(cap)) {
476 if (!scheduleGetRemoteWork(&receivedFinish)) continue;
477 ASSERT(rtsFalse); // should not happen at the moment
479 // from here: non-empty run queue.
480 // TODO: merge above case with this, only one call processMessages() !
481 if (PacketsWaiting()) { /* process incoming messages, if
482 any pending... only in else
483 because getRemoteWork waits for
485 receivedFinish = processMessages();
490 scheduleProcessEvent(event);
494 // Get a thread to run
496 t = popRunQueue(cap);
498 #if defined(GRAN) || defined(PAR)
499 scheduleGranParReport(); // some kind of debuging output
501 // Sanity check the thread we're about to run. This can be
502 // expensive if there is lots of thread switching going on...
503 IF_DEBUG(sanity,checkTSO(t));
506 #if defined(THREADED_RTS)
507 // Check whether we can run this thread in the current task.
508 // If not, we have to pass our capability to the right task.
510 Task *bound = t->bound;
515 sched_belch("### Running thread %d in bound thread",
517 // yes, the Haskell thread is bound to the current native thread
520 sched_belch("### thread %d bound to another OS thread",
522 // no, bound to a different Haskell thread: pass to that thread
523 pushOnRunQueue(cap,t);
527 // The thread we want to run is unbound.
530 sched_belch("### this OS thread cannot run thread %d", t->id));
531 // no, the current native thread is bound to a different
532 // Haskell thread, so pass it to any worker thread
533 pushOnRunQueue(cap,t);
540 cap->r.rCurrentTSO = t;
542 /* context switches are initiated by the timer signal, unless
543 * the user specified "context switch as often as possible", with
546 if (RtsFlags.ConcFlags.ctxtSwitchTicks == 0
547 && !emptyThreadQueues(cap)) {
553 IF_DEBUG(scheduler, sched_belch("-->> running thread %ld %s ...",
554 (long)t->id, whatNext_strs[t->what_next]));
556 #if defined(PROFILING)
557 startHeapProfTimer();
560 // ----------------------------------------------------------------------
561 // Run the current thread
563 prev_what_next = t->what_next;
565 errno = t->saved_errno;
566 cap->in_haskell = rtsTrue;
570 recent_activity = ACTIVITY_YES;
572 switch (prev_what_next) {
576 /* Thread already finished, return to scheduler. */
577 ret = ThreadFinished;
583 r = StgRun((StgFunPtr) stg_returnToStackTop, &cap->r);
584 cap = regTableToCapability(r);
589 case ThreadInterpret:
590 cap = interpretBCO(cap);
595 barf("schedule: invalid what_next field");
598 cap->in_haskell = rtsFalse;
600 // The TSO might have moved, eg. if it re-entered the RTS and a GC
601 // happened. So find the new location:
602 t = cap->r.rCurrentTSO;
604 // We have run some Haskell code: there might be blackhole-blocked
605 // threads to wake up now.
606 // Lock-free test here should be ok, we're just setting a flag.
607 if ( blackhole_queue != END_TSO_QUEUE ) {
608 blackholes_need_checking = rtsTrue;
611 // And save the current errno in this thread.
612 // XXX: possibly bogus for SMP because this thread might already
613 // be running again, see code below.
614 t->saved_errno = errno;
617 // If ret is ThreadBlocked, and this Task is bound to the TSO that
618 // blocked, we are in limbo - the TSO is now owned by whatever it
619 // is blocked on, and may in fact already have been woken up,
620 // perhaps even on a different Capability. It may be the case
621 // that task->cap != cap. We better yield this Capability
622 // immediately and return to normaility.
623 if (ret == ThreadBlocked) {
625 sched_belch("--<< thread %d (%s) stopped: blocked\n",
626 t->id, whatNext_strs[t->what_next]));
631 ASSERT_FULL_CAPABILITY_INVARIANTS(cap,task);
633 // ----------------------------------------------------------------------
635 // Costs for the scheduler are assigned to CCS_SYSTEM
636 #if defined(PROFILING)
641 #if defined(THREADED_RTS)
642 IF_DEBUG(scheduler,debugBelch("sched (task %p): ", (void *)(unsigned long)(unsigned int)osThreadId()););
643 #elif !defined(GRAN) && !defined(PARALLEL_HASKELL)
644 IF_DEBUG(scheduler,debugBelch("sched: "););
647 schedulePostRunThread();
649 ready_to_gc = rtsFalse;
653 ready_to_gc = scheduleHandleHeapOverflow(cap,t);
657 scheduleHandleStackOverflow(cap,task,t);
661 if (scheduleHandleYield(cap, t, prev_what_next)) {
662 // shortcut for switching between compiler/interpreter:
668 scheduleHandleThreadBlocked(t);
672 if (scheduleHandleThreadFinished(cap, task, t)) return cap;
673 ASSERT_FULL_CAPABILITY_INVARIANTS(cap,task);
677 barf("schedule: invalid thread return code %d", (int)ret);
680 if (scheduleDoHeapProfile(ready_to_gc)) { ready_to_gc = rtsFalse; }
681 if (ready_to_gc) { scheduleDoGC(cap,task,rtsFalse,GetRoots); }
682 } /* end of while() */
684 IF_PAR_DEBUG(verbose,
685 debugBelch("== Leaving schedule() after having received Finish\n"));
688 /* ----------------------------------------------------------------------------
689 * Setting up the scheduler loop
690 * ------------------------------------------------------------------------- */
693 schedulePreLoop(void)
696 /* set up first event to get things going */
697 /* ToDo: assign costs for system setup and init MainTSO ! */
698 new_event(CurrentProc, CurrentProc, CurrentTime[CurrentProc],
700 CurrentTSO, (StgClosure*)NULL, (rtsSpark*)NULL);
703 debugBelch("GRAN: Init CurrentTSO (in schedule) = %p\n",
705 G_TSO(CurrentTSO, 5));
707 if (RtsFlags.GranFlags.Light) {
708 /* Save current time; GranSim Light only */
709 CurrentTSO->gran.clock = CurrentTime[CurrentProc];
714 /* -----------------------------------------------------------------------------
717 * Push work to other Capabilities if we have some.
718 * -------------------------------------------------------------------------- */
720 #if defined(THREADED_RTS)
722 schedulePushWork(Capability *cap USED_IF_THREADS,
723 Task *task USED_IF_THREADS)
725 Capability *free_caps[n_capabilities], *cap0;
728 // Check whether we have more threads on our run queue, or sparks
729 // in our pool, that we could hand to another Capability.
730 if ((emptyRunQueue(cap) || cap->run_queue_hd->link == END_TSO_QUEUE)
731 && sparkPoolSizeCap(cap) < 2) {
735 // First grab as many free Capabilities as we can.
736 for (i=0, n_free_caps=0; i < n_capabilities; i++) {
737 cap0 = &capabilities[i];
738 if (cap != cap0 && tryGrabCapability(cap0,task)) {
739 if (!emptyRunQueue(cap0) || cap->returning_tasks_hd != NULL) {
740 // it already has some work, we just grabbed it at
741 // the wrong moment. Or maybe it's deadlocked!
742 releaseCapability(cap0);
744 free_caps[n_free_caps++] = cap0;
749 // we now have n_free_caps free capabilities stashed in
750 // free_caps[]. Share our run queue equally with them. This is
751 // probably the simplest thing we could do; improvements we might
752 // want to do include:
754 // - giving high priority to moving relatively new threads, on
755 // the gournds that they haven't had time to build up a
756 // working set in the cache on this CPU/Capability.
758 // - giving low priority to moving long-lived threads
760 if (n_free_caps > 0) {
761 StgTSO *prev, *t, *next;
762 rtsBool pushed_to_all;
764 IF_DEBUG(scheduler, sched_belch("excess threads on run queue and %d free capabilities, sharing...", n_free_caps));
767 pushed_to_all = rtsFalse;
769 if (cap->run_queue_hd != END_TSO_QUEUE) {
770 prev = cap->run_queue_hd;
772 prev->link = END_TSO_QUEUE;
773 for (; t != END_TSO_QUEUE; t = next) {
775 t->link = END_TSO_QUEUE;
776 if (t->what_next == ThreadRelocated
777 || t->bound == task) { // don't move my bound thread
780 } else if (i == n_free_caps) {
781 pushed_to_all = rtsTrue;
787 IF_DEBUG(scheduler, sched_belch("pushing thread %d to capability %d", t->id, free_caps[i]->no));
788 appendToRunQueue(free_caps[i],t);
789 if (t->bound) { t->bound->cap = free_caps[i]; }
793 cap->run_queue_tl = prev;
796 // If there are some free capabilities that we didn't push any
797 // threads to, then try to push a spark to each one.
798 if (!pushed_to_all) {
800 // i is the next free capability to push to
801 for (; i < n_free_caps; i++) {
802 if (emptySparkPoolCap(free_caps[i])) {
803 spark = findSpark(cap);
805 IF_DEBUG(scheduler, sched_belch("pushing spark %p to capability %d", spark, free_caps[i]->no));
806 newSpark(&(free_caps[i]->r), spark);
812 // release the capabilities
813 for (i = 0; i < n_free_caps; i++) {
814 task->cap = free_caps[i];
815 releaseCapability(free_caps[i]);
818 task->cap = cap; // reset to point to our Capability.
822 /* ----------------------------------------------------------------------------
823 * Start any pending signal handlers
824 * ------------------------------------------------------------------------- */
826 #if defined(RTS_USER_SIGNALS) && (!defined(THREADED_RTS) || defined(mingw32_HOST_OS))
828 scheduleStartSignalHandlers(Capability *cap)
830 if (signals_pending()) { // safe outside the lock
831 startSignalHandlers(cap);
836 scheduleStartSignalHandlers(Capability *cap STG_UNUSED)
841 /* ----------------------------------------------------------------------------
842 * Check for blocked threads that can be woken up.
843 * ------------------------------------------------------------------------- */
846 scheduleCheckBlockedThreads(Capability *cap USED_IF_NOT_THREADS)
848 #if !defined(THREADED_RTS)
850 // Check whether any waiting threads need to be woken up. If the
851 // run queue is empty, and there are no other tasks running, we
852 // can wait indefinitely for something to happen.
854 if ( !emptyQueue(blocked_queue_hd) || !emptyQueue(sleeping_queue) )
856 awaitEvent( emptyRunQueue(cap) && !blackholes_need_checking );
862 /* ----------------------------------------------------------------------------
863 * Check for threads blocked on BLACKHOLEs that can be woken up
864 * ------------------------------------------------------------------------- */
866 scheduleCheckBlackHoles (Capability *cap)
868 if ( blackholes_need_checking ) // check without the lock first
870 ACQUIRE_LOCK(&sched_mutex);
871 if ( blackholes_need_checking ) {
872 checkBlackHoles(cap);
873 blackholes_need_checking = rtsFalse;
875 RELEASE_LOCK(&sched_mutex);
879 /* ----------------------------------------------------------------------------
880 * Detect deadlock conditions and attempt to resolve them.
881 * ------------------------------------------------------------------------- */
884 scheduleDetectDeadlock (Capability *cap, Task *task)
887 #if defined(PARALLEL_HASKELL)
888 // ToDo: add deadlock detection in GUM (similar to THREADED_RTS) -- HWL
893 * Detect deadlock: when we have no threads to run, there are no
894 * threads blocked, waiting for I/O, or sleeping, and all the
895 * other tasks are waiting for work, we must have a deadlock of
898 if ( emptyThreadQueues(cap) )
900 #if defined(THREADED_RTS)
902 * In the threaded RTS, we only check for deadlock if there
903 * has been no activity in a complete timeslice. This means
904 * we won't eagerly start a full GC just because we don't have
905 * any threads to run currently.
907 if (recent_activity != ACTIVITY_INACTIVE) return;
910 IF_DEBUG(scheduler, sched_belch("deadlocked, forcing major GC..."));
912 // Garbage collection can release some new threads due to
913 // either (a) finalizers or (b) threads resurrected because
914 // they are unreachable and will therefore be sent an
915 // exception. Any threads thus released will be immediately
917 scheduleDoGC( cap, task, rtsTrue/*force major GC*/, GetRoots );
918 recent_activity = ACTIVITY_DONE_GC;
920 if ( !emptyRunQueue(cap) ) return;
922 #if defined(RTS_USER_SIGNALS) && (!defined(THREADED_RTS) || defined(mingw32_HOST_OS))
923 /* If we have user-installed signal handlers, then wait
924 * for signals to arrive rather then bombing out with a
927 if ( anyUserHandlers() ) {
929 sched_belch("still deadlocked, waiting for signals..."));
933 if (signals_pending()) {
934 startSignalHandlers(cap);
937 // either we have threads to run, or we were interrupted:
938 ASSERT(!emptyRunQueue(cap) || interrupted);
942 #if !defined(THREADED_RTS)
943 /* Probably a real deadlock. Send the current main thread the
944 * Deadlock exception.
947 switch (task->tso->why_blocked) {
949 case BlockedOnBlackHole:
950 case BlockedOnException:
952 raiseAsync(cap, task->tso, (StgClosure *)NonTermination_closure);
955 barf("deadlock: main thread blocked in a strange way");
963 /* ----------------------------------------------------------------------------
964 * Process an event (GRAN only)
965 * ------------------------------------------------------------------------- */
969 scheduleProcessEvent(rtsEvent *event)
973 if (RtsFlags.GranFlags.Light)
974 GranSimLight_enter_system(event, &ActiveTSO); // adjust ActiveTSO etc
976 /* adjust time based on time-stamp */
977 if (event->time > CurrentTime[CurrentProc] &&
978 event->evttype != ContinueThread)
979 CurrentTime[CurrentProc] = event->time;
981 /* Deal with the idle PEs (may issue FindWork or MoveSpark events) */
982 if (!RtsFlags.GranFlags.Light)
985 IF_DEBUG(gran, debugBelch("GRAN: switch by event-type\n"));
987 /* main event dispatcher in GranSim */
988 switch (event->evttype) {
989 /* Should just be continuing execution */
991 IF_DEBUG(gran, debugBelch("GRAN: doing ContinueThread\n"));
992 /* ToDo: check assertion
993 ASSERT(run_queue_hd != (StgTSO*)NULL &&
994 run_queue_hd != END_TSO_QUEUE);
996 /* Ignore ContinueThreads for fetching threads (if synchr comm) */
997 if (!RtsFlags.GranFlags.DoAsyncFetch &&
998 procStatus[CurrentProc]==Fetching) {
999 debugBelch("ghuH: Spurious ContinueThread while Fetching ignored; TSO %d (%p) [PE %d]\n",
1000 CurrentTSO->id, CurrentTSO, CurrentProc);
1003 /* Ignore ContinueThreads for completed threads */
1004 if (CurrentTSO->what_next == ThreadComplete) {
1005 debugBelch("ghuH: found a ContinueThread event for completed thread %d (%p) [PE %d] (ignoring ContinueThread)\n",
1006 CurrentTSO->id, CurrentTSO, CurrentProc);
1009 /* Ignore ContinueThreads for threads that are being migrated */
1010 if (PROCS(CurrentTSO)==Nowhere) {
1011 debugBelch("ghuH: trying to run the migrating TSO %d (%p) [PE %d] (ignoring ContinueThread)\n",
1012 CurrentTSO->id, CurrentTSO, CurrentProc);
1015 /* The thread should be at the beginning of the run queue */
1016 if (CurrentTSO!=run_queue_hds[CurrentProc]) {
1017 debugBelch("ghuH: TSO %d (%p) [PE %d] is not at the start of the run_queue when doing a ContinueThread\n",
1018 CurrentTSO->id, CurrentTSO, CurrentProc);
1019 break; // run the thread anyway
1022 new_event(proc, proc, CurrentTime[proc],
1024 (StgTSO*)NULL, (StgClosure*)NULL, (rtsSpark*)NULL);
1026 */ /* Catches superfluous CONTINUEs -- should be unnecessary */
1027 break; // now actually run the thread; DaH Qu'vam yImuHbej
1030 do_the_fetchnode(event);
1031 goto next_thread; /* handle next event in event queue */
1034 do_the_globalblock(event);
1035 goto next_thread; /* handle next event in event queue */
1038 do_the_fetchreply(event);
1039 goto next_thread; /* handle next event in event queue */
1041 case UnblockThread: /* Move from the blocked queue to the tail of */
1042 do_the_unblock(event);
1043 goto next_thread; /* handle next event in event queue */
1045 case ResumeThread: /* Move from the blocked queue to the tail of */
1046 /* the runnable queue ( i.e. Qu' SImqa'lu') */
1047 event->tso->gran.blocktime +=
1048 CurrentTime[CurrentProc] - event->tso->gran.blockedat;
1049 do_the_startthread(event);
1050 goto next_thread; /* handle next event in event queue */
1053 do_the_startthread(event);
1054 goto next_thread; /* handle next event in event queue */
1057 do_the_movethread(event);
1058 goto next_thread; /* handle next event in event queue */
1061 do_the_movespark(event);
1062 goto next_thread; /* handle next event in event queue */
1065 do_the_findwork(event);
1066 goto next_thread; /* handle next event in event queue */
1069 barf("Illegal event type %u\n", event->evttype);
1072 /* This point was scheduler_loop in the old RTS */
1074 IF_DEBUG(gran, debugBelch("GRAN: after main switch\n"));
1076 TimeOfLastEvent = CurrentTime[CurrentProc];
1077 TimeOfNextEvent = get_time_of_next_event();
1078 IgnoreEvents=(TimeOfNextEvent==0); // HWL HACK
1079 // CurrentTSO = ThreadQueueHd;
1081 IF_DEBUG(gran, debugBelch("GRAN: time of next event is: %ld\n",
1084 if (RtsFlags.GranFlags.Light)
1085 GranSimLight_leave_system(event, &ActiveTSO);
1087 EndOfTimeSlice = CurrentTime[CurrentProc]+RtsFlags.GranFlags.time_slice;
1090 debugBelch("GRAN: end of time-slice is %#lx\n", EndOfTimeSlice));
1092 /* in a GranSim setup the TSO stays on the run queue */
1094 /* Take a thread from the run queue. */
1095 POP_RUN_QUEUE(t); // take_off_run_queue(t);
1098 debugBelch("GRAN: About to run current thread, which is\n");
1101 context_switch = 0; // turned on via GranYield, checking events and time slice
1104 DumpGranEvent(GR_SCHEDULE, t));
1106 procStatus[CurrentProc] = Busy;
1110 /* ----------------------------------------------------------------------------
1111 * Send pending messages (PARALLEL_HASKELL only)
1112 * ------------------------------------------------------------------------- */
1114 #if defined(PARALLEL_HASKELL)
1116 scheduleSendPendingMessages(void)
1122 # if defined(PAR) // global Mem.Mgmt., omit for now
1123 if (PendingFetches != END_BF_QUEUE) {
1128 if (RtsFlags.ParFlags.BufferTime) {
1129 // if we use message buffering, we must send away all message
1130 // packets which have become too old...
1136 /* ----------------------------------------------------------------------------
1137 * Activate spark threads (PARALLEL_HASKELL only)
1138 * ------------------------------------------------------------------------- */
1140 #if defined(PARALLEL_HASKELL)
1142 scheduleActivateSpark(void)
1145 ASSERT(emptyRunQueue());
1146 /* We get here if the run queue is empty and want some work.
1147 We try to turn a spark into a thread, and add it to the run queue,
1148 from where it will be picked up in the next iteration of the scheduler
1152 /* :-[ no local threads => look out for local sparks */
1153 /* the spark pool for the current PE */
1154 pool = &(cap.r.rSparks); // JB: cap = (old) MainCap
1155 if (advisory_thread_count < RtsFlags.ParFlags.maxThreads &&
1156 pool->hd < pool->tl) {
1158 * ToDo: add GC code check that we really have enough heap afterwards!!
1160 * If we're here (no runnable threads) and we have pending
1161 * sparks, we must have a space problem. Get enough space
1162 * to turn one of those pending sparks into a
1166 spark = findSpark(rtsFalse); /* get a spark */
1167 if (spark != (rtsSpark) NULL) {
1168 tso = createThreadFromSpark(spark); /* turn the spark into a thread */
1169 IF_PAR_DEBUG(fish, // schedule,
1170 debugBelch("==== schedule: Created TSO %d (%p); %d threads active\n",
1171 tso->id, tso, advisory_thread_count));
1173 if (tso==END_TSO_QUEUE) { /* failed to activate spark->back to loop */
1174 IF_PAR_DEBUG(fish, // schedule,
1175 debugBelch("==^^ failed to create thread from spark @ %lx\n",
1177 return rtsFalse; /* failed to generate a thread */
1178 } /* otherwise fall through & pick-up new tso */
1180 IF_PAR_DEBUG(fish, // schedule,
1181 debugBelch("==^^ no local sparks (spark pool contains only NFs: %d)\n",
1182 spark_queue_len(pool)));
1183 return rtsFalse; /* failed to generate a thread */
1185 return rtsTrue; /* success in generating a thread */
1186 } else { /* no more threads permitted or pool empty */
1187 return rtsFalse; /* failed to generateThread */
1190 tso = NULL; // avoid compiler warning only
1191 return rtsFalse; /* dummy in non-PAR setup */
1194 #endif // PARALLEL_HASKELL
1196 /* ----------------------------------------------------------------------------
1197 * Get work from a remote node (PARALLEL_HASKELL only)
1198 * ------------------------------------------------------------------------- */
1200 #if defined(PARALLEL_HASKELL)
1202 scheduleGetRemoteWork(rtsBool *receivedFinish)
1204 ASSERT(emptyRunQueue());
1206 if (RtsFlags.ParFlags.BufferTime) {
1207 IF_PAR_DEBUG(verbose,
1208 debugBelch("...send all pending data,"));
1211 for (i=1; i<=nPEs; i++)
1212 sendImmediately(i); // send all messages away immediately
1216 //++EDEN++ idle() , i.e. send all buffers, wait for work
1217 // suppress fishing in EDEN... just look for incoming messages
1218 // (blocking receive)
1219 IF_PAR_DEBUG(verbose,
1220 debugBelch("...wait for incoming messages...\n"));
1221 *receivedFinish = processMessages(); // blocking receive...
1223 // and reenter scheduling loop after having received something
1224 // (return rtsFalse below)
1226 # else /* activate SPARKS machinery */
1227 /* We get here, if we have no work, tried to activate a local spark, but still
1228 have no work. We try to get a remote spark, by sending a FISH message.
1229 Thread migration should be added here, and triggered when a sequence of
1230 fishes returns without work. */
1231 delay = (RtsFlags.ParFlags.fishDelay!=0ll ? RtsFlags.ParFlags.fishDelay : 0ll);
1233 /* =8-[ no local sparks => look for work on other PEs */
1235 * We really have absolutely no work. Send out a fish
1236 * (there may be some out there already), and wait for
1237 * something to arrive. We clearly can't run any threads
1238 * until a SCHEDULE or RESUME arrives, and so that's what
1239 * we're hoping to see. (Of course, we still have to
1240 * respond to other types of messages.)
1242 rtsTime now = msTime() /*CURRENT_TIME*/;
1243 IF_PAR_DEBUG(verbose,
1244 debugBelch("-- now=%ld\n", now));
1245 IF_PAR_DEBUG(fish, // verbose,
1246 if (outstandingFishes < RtsFlags.ParFlags.maxFishes &&
1247 (last_fish_arrived_at!=0 &&
1248 last_fish_arrived_at+delay > now)) {
1249 debugBelch("--$$ <%llu> delaying FISH until %llu (last fish %llu, delay %llu)\n",
1250 now, last_fish_arrived_at+delay,
1251 last_fish_arrived_at,
1255 if (outstandingFishes < RtsFlags.ParFlags.maxFishes &&
1256 advisory_thread_count < RtsFlags.ParFlags.maxThreads) { // send a FISH, but when?
1257 if (last_fish_arrived_at==0 ||
1258 (last_fish_arrived_at+delay <= now)) { // send FISH now!
1259 /* outstandingFishes is set in sendFish, processFish;
1260 avoid flooding system with fishes via delay */
1261 next_fish_to_send_at = 0;
1263 /* ToDo: this should be done in the main scheduling loop to avoid the
1264 busy wait here; not so bad if fish delay is very small */
1265 int iq = 0; // DEBUGGING -- HWL
1266 next_fish_to_send_at = last_fish_arrived_at+delay; // remember when to send
1267 /* send a fish when ready, but process messages that arrive in the meantime */
1269 if (PacketsWaiting()) {
1271 *receivedFinish = processMessages();
1274 } while (!*receivedFinish || now<next_fish_to_send_at);
1275 // JB: This means the fish could become obsolete, if we receive
1276 // work. Better check for work again?
1277 // last line: while (!receivedFinish || !haveWork || now<...)
1278 // next line: if (receivedFinish || haveWork )
1280 if (*receivedFinish) // no need to send a FISH if we are finishing anyway
1281 return rtsFalse; // NB: this will leave scheduler loop
1282 // immediately after return!
1284 IF_PAR_DEBUG(fish, // verbose,
1285 debugBelch("--$$ <%llu> sent delayed fish (%d processMessages); active/total threads=%d/%d\n",now,iq,run_queue_len(),advisory_thread_count));
1289 // JB: IMHO, this should all be hidden inside sendFish(...)
1291 sendFish(pe, thisPE, NEW_FISH_AGE, NEW_FISH_HISTORY,
1294 // Global statistics: count no. of fishes
1295 if (RtsFlags.ParFlags.ParStats.Global &&
1296 RtsFlags.GcFlags.giveStats > NO_GC_STATS) {
1297 globalParStats.tot_fish_mess++;
1301 /* delayed fishes must have been sent by now! */
1302 next_fish_to_send_at = 0;
1305 *receivedFinish = processMessages();
1306 # endif /* SPARKS */
1309 /* NB: this function always returns rtsFalse, meaning the scheduler
1310 loop continues with the next iteration;
1312 return code means success in finding work; we enter this function
1313 if there is no local work, thus have to send a fish which takes
1314 time until it arrives with work; in the meantime we should process
1315 messages in the main loop;
1318 #endif // PARALLEL_HASKELL
1320 /* ----------------------------------------------------------------------------
1321 * PAR/GRAN: Report stats & debugging info(?)
1322 * ------------------------------------------------------------------------- */
1324 #if defined(PAR) || defined(GRAN)
1326 scheduleGranParReport(void)
1328 ASSERT(run_queue_hd != END_TSO_QUEUE);
1330 /* Take a thread from the run queue, if we have work */
1331 POP_RUN_QUEUE(t); // take_off_run_queue(END_TSO_QUEUE);
1333 /* If this TSO has got its outport closed in the meantime,
1334 * it mustn't be run. Instead, we have to clean it up as if it was finished.
1335 * It has to be marked as TH_DEAD for this purpose.
1336 * If it is TH_TERM instead, it is supposed to have finished in the normal way.
1338 JB: TODO: investigate wether state change field could be nuked
1339 entirely and replaced by the normal tso state (whatnext
1340 field). All we want to do is to kill tsos from outside.
1343 /* ToDo: write something to the log-file
1344 if (RTSflags.ParFlags.granSimStats && !sameThread)
1345 DumpGranEvent(GR_SCHEDULE, RunnableThreadsHd);
1349 /* the spark pool for the current PE */
1350 pool = &(cap.r.rSparks); // cap = (old) MainCap
1353 debugBelch("--=^ %d threads, %d sparks on [%#x]\n",
1354 run_queue_len(), spark_queue_len(pool), CURRENT_PROC));
1357 debugBelch("--=^ %d threads, %d sparks on [%#x]\n",
1358 run_queue_len(), spark_queue_len(pool), CURRENT_PROC));
1360 if (RtsFlags.ParFlags.ParStats.Full &&
1361 (t->par.sparkname != (StgInt)0) && // only log spark generated threads
1362 (emitSchedule || // forced emit
1363 (t && LastTSO && t->id != LastTSO->id))) {
1365 we are running a different TSO, so write a schedule event to log file
1366 NB: If we use fair scheduling we also have to write a deschedule
1367 event for LastTSO; with unfair scheduling we know that the
1368 previous tso has blocked whenever we switch to another tso, so
1369 we don't need it in GUM for now
1371 IF_PAR_DEBUG(fish, // schedule,
1372 debugBelch("____ scheduling spark generated thread %d (%lx) (%lx) via a forced emit\n",t->id,t,t->par.sparkname));
1374 DumpRawGranEvent(CURRENT_PROC, CURRENT_PROC,
1375 GR_SCHEDULE, t, (StgClosure *)NULL, 0, 0);
1376 emitSchedule = rtsFalse;
1381 /* ----------------------------------------------------------------------------
1382 * After running a thread...
1383 * ------------------------------------------------------------------------- */
1386 schedulePostRunThread(void)
1389 /* HACK 675: if the last thread didn't yield, make sure to print a
1390 SCHEDULE event to the log file when StgRunning the next thread, even
1391 if it is the same one as before */
1393 TimeOfLastYield = CURRENT_TIME;
1396 /* some statistics gathering in the parallel case */
1398 #if defined(GRAN) || defined(PAR) || defined(EDEN)
1402 IF_DEBUG(gran, DumpGranEvent(GR_DESCHEDULE, t));
1403 globalGranStats.tot_heapover++;
1405 globalParStats.tot_heapover++;
1412 DumpGranEvent(GR_DESCHEDULE, t));
1413 globalGranStats.tot_stackover++;
1416 // DumpGranEvent(GR_DESCHEDULE, t);
1417 globalParStats.tot_stackover++;
1421 case ThreadYielding:
1424 DumpGranEvent(GR_DESCHEDULE, t));
1425 globalGranStats.tot_yields++;
1428 // DumpGranEvent(GR_DESCHEDULE, t);
1429 globalParStats.tot_yields++;
1436 debugBelch("--<< thread %ld (%p; %s) stopped, blocking on node %p [PE %d] with BQ: ",
1437 t->id, t, whatNext_strs[t->what_next], t->block_info.closure,
1438 (t->block_info.closure==(StgClosure*)NULL ? 99 : where_is(t->block_info.closure)));
1439 if (t->block_info.closure!=(StgClosure*)NULL)
1440 print_bq(t->block_info.closure);
1443 // ??? needed; should emit block before
1445 DumpGranEvent(GR_DESCHEDULE, t));
1446 prune_eventq(t, (StgClosure *)NULL); // prune ContinueThreads for t
1449 ASSERT(procStatus[CurrentProc]==Busy ||
1450 ((procStatus[CurrentProc]==Fetching) &&
1451 (t->block_info.closure!=(StgClosure*)NULL)));
1452 if (run_queue_hds[CurrentProc] == END_TSO_QUEUE &&
1453 !(!RtsFlags.GranFlags.DoAsyncFetch &&
1454 procStatus[CurrentProc]==Fetching))
1455 procStatus[CurrentProc] = Idle;
1458 //++PAR++ blockThread() writes the event (change?)
1462 case ThreadFinished:
1466 barf("parGlobalStats: unknown return code");
1472 /* -----------------------------------------------------------------------------
1473 * Handle a thread that returned to the scheduler with ThreadHeepOverflow
1474 * -------------------------------------------------------------------------- */
1477 scheduleHandleHeapOverflow( Capability *cap, StgTSO *t )
1479 // did the task ask for a large block?
1480 if (cap->r.rHpAlloc > BLOCK_SIZE) {
1481 // if so, get one and push it on the front of the nursery.
1485 blocks = (lnat)BLOCK_ROUND_UP(cap->r.rHpAlloc) / BLOCK_SIZE;
1488 debugBelch("--<< thread %ld (%s) stopped: requesting a large block (size %ld)\n",
1489 (long)t->id, whatNext_strs[t->what_next], blocks));
1491 // don't do this if the nursery is (nearly) full, we'll GC first.
1492 if (cap->r.rCurrentNursery->link != NULL ||
1493 cap->r.rNursery->n_blocks == 1) { // paranoia to prevent infinite loop
1494 // if the nursery has only one block.
1497 bd = allocGroup( blocks );
1499 cap->r.rNursery->n_blocks += blocks;
1501 // link the new group into the list
1502 bd->link = cap->r.rCurrentNursery;
1503 bd->u.back = cap->r.rCurrentNursery->u.back;
1504 if (cap->r.rCurrentNursery->u.back != NULL) {
1505 cap->r.rCurrentNursery->u.back->link = bd;
1507 #if !defined(THREADED_RTS)
1508 ASSERT(g0s0->blocks == cap->r.rCurrentNursery &&
1509 g0s0 == cap->r.rNursery);
1511 cap->r.rNursery->blocks = bd;
1513 cap->r.rCurrentNursery->u.back = bd;
1515 // initialise it as a nursery block. We initialise the
1516 // step, gen_no, and flags field of *every* sub-block in
1517 // this large block, because this is easier than making
1518 // sure that we always find the block head of a large
1519 // block whenever we call Bdescr() (eg. evacuate() and
1520 // isAlive() in the GC would both have to do this, at
1524 for (x = bd; x < bd + blocks; x++) {
1525 x->step = cap->r.rNursery;
1531 // This assert can be a killer if the app is doing lots
1532 // of large block allocations.
1533 IF_DEBUG(sanity, checkNurserySanity(cap->r.rNursery));
1535 // now update the nursery to point to the new block
1536 cap->r.rCurrentNursery = bd;
1538 // we might be unlucky and have another thread get on the
1539 // run queue before us and steal the large block, but in that
1540 // case the thread will just end up requesting another large
1542 pushOnRunQueue(cap,t);
1543 return rtsFalse; /* not actually GC'ing */
1548 debugBelch("--<< thread %ld (%s) stopped: HeapOverflow\n",
1549 (long)t->id, whatNext_strs[t->what_next]));
1551 ASSERT(!is_on_queue(t,CurrentProc));
1552 #elif defined(PARALLEL_HASKELL)
1553 /* Currently we emit a DESCHEDULE event before GC in GUM.
1554 ToDo: either add separate event to distinguish SYSTEM time from rest
1555 or just nuke this DESCHEDULE (and the following SCHEDULE) */
1556 if (0 && RtsFlags.ParFlags.ParStats.Full) {
1557 DumpRawGranEvent(CURRENT_PROC, CURRENT_PROC,
1558 GR_DESCHEDULE, t, (StgClosure *)NULL, 0, 0);
1559 emitSchedule = rtsTrue;
1563 pushOnRunQueue(cap,t);
1565 /* actual GC is done at the end of the while loop in schedule() */
1568 /* -----------------------------------------------------------------------------
1569 * Handle a thread that returned to the scheduler with ThreadStackOverflow
1570 * -------------------------------------------------------------------------- */
1573 scheduleHandleStackOverflow (Capability *cap, Task *task, StgTSO *t)
1575 IF_DEBUG(scheduler,debugBelch("--<< thread %ld (%s) stopped, StackOverflow\n",
1576 (long)t->id, whatNext_strs[t->what_next]));
1577 /* just adjust the stack for this thread, then pop it back
1581 /* enlarge the stack */
1582 StgTSO *new_t = threadStackOverflow(cap, t);
1584 /* The TSO attached to this Task may have moved, so update the
1587 if (task->tso == t) {
1590 pushOnRunQueue(cap,new_t);
1594 /* -----------------------------------------------------------------------------
1595 * Handle a thread that returned to the scheduler with ThreadYielding
1596 * -------------------------------------------------------------------------- */
1599 scheduleHandleYield( Capability *cap, StgTSO *t, nat prev_what_next )
1601 // Reset the context switch flag. We don't do this just before
1602 // running the thread, because that would mean we would lose ticks
1603 // during GC, which can lead to unfair scheduling (a thread hogs
1604 // the CPU because the tick always arrives during GC). This way
1605 // penalises threads that do a lot of allocation, but that seems
1606 // better than the alternative.
1609 /* put the thread back on the run queue. Then, if we're ready to
1610 * GC, check whether this is the last task to stop. If so, wake
1611 * up the GC thread. getThread will block during a GC until the
1615 if (t->what_next != prev_what_next) {
1616 debugBelch("--<< thread %ld (%s) stopped to switch evaluators\n",
1617 (long)t->id, whatNext_strs[t->what_next]);
1619 debugBelch("--<< thread %ld (%s) stopped, yielding\n",
1620 (long)t->id, whatNext_strs[t->what_next]);
1625 //debugBelch("&& Doing sanity check on yielding TSO %ld.", t->id);
1627 ASSERT(t->link == END_TSO_QUEUE);
1629 // Shortcut if we're just switching evaluators: don't bother
1630 // doing stack squeezing (which can be expensive), just run the
1632 if (t->what_next != prev_what_next) {
1637 ASSERT(!is_on_queue(t,CurrentProc));
1640 //debugBelch("&& Doing sanity check on all ThreadQueues (and their TSOs).");
1641 checkThreadQsSanity(rtsTrue));
1645 addToRunQueue(cap,t);
1648 /* add a ContinueThread event to actually process the thread */
1649 new_event(CurrentProc, CurrentProc, CurrentTime[CurrentProc],
1651 t, (StgClosure*)NULL, (rtsSpark*)NULL);
1653 debugBelch("GRAN: eventq and runnableq after adding yielded thread to queue again:\n");
1660 /* -----------------------------------------------------------------------------
1661 * Handle a thread that returned to the scheduler with ThreadBlocked
1662 * -------------------------------------------------------------------------- */
1665 scheduleHandleThreadBlocked( StgTSO *t
1666 #if !defined(GRAN) && !defined(DEBUG)
1673 debugBelch("--<< thread %ld (%p; %s) stopped, blocking on node %p [PE %d] with BQ: \n",
1674 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)));
1675 if (t->block_info.closure!=(StgClosure*)NULL) print_bq(t->block_info.closure));
1677 // ??? needed; should emit block before
1679 DumpGranEvent(GR_DESCHEDULE, t));
1680 prune_eventq(t, (StgClosure *)NULL); // prune ContinueThreads for t
1683 ASSERT(procStatus[CurrentProc]==Busy ||
1684 ((procStatus[CurrentProc]==Fetching) &&
1685 (t->block_info.closure!=(StgClosure*)NULL)));
1686 if (run_queue_hds[CurrentProc] == END_TSO_QUEUE &&
1687 !(!RtsFlags.GranFlags.DoAsyncFetch &&
1688 procStatus[CurrentProc]==Fetching))
1689 procStatus[CurrentProc] = Idle;
1693 debugBelch("--<< thread %ld (%p; %s) stopped, blocking on node %p with BQ: \n",
1694 t->id, t, whatNext_strs[t->what_next], t->block_info.closure));
1697 if (t->block_info.closure!=(StgClosure*)NULL)
1698 print_bq(t->block_info.closure));
1700 /* Send a fetch (if BlockedOnGA) and dump event to log file */
1703 /* whatever we schedule next, we must log that schedule */
1704 emitSchedule = rtsTrue;
1708 // We don't need to do anything. The thread is blocked, and it
1709 // has tidied up its stack and placed itself on whatever queue
1710 // it needs to be on.
1712 #if !defined(THREADED_RTS)
1713 ASSERT(t->why_blocked != NotBlocked);
1714 // This might not be true under THREADED_RTS: we don't have
1715 // exclusive access to this TSO, so someone might have
1716 // woken it up by now. This actually happens: try
1717 // conc023 +RTS -N2.
1721 debugBelch("--<< thread %d (%s) stopped: ",
1722 t->id, whatNext_strs[t->what_next]);
1723 printThreadBlockage(t);
1726 /* Only for dumping event to log file
1727 ToDo: do I need this in GranSim, too?
1733 /* -----------------------------------------------------------------------------
1734 * Handle a thread that returned to the scheduler with ThreadFinished
1735 * -------------------------------------------------------------------------- */
1738 scheduleHandleThreadFinished (Capability *cap STG_UNUSED, Task *task, StgTSO *t)
1740 /* Need to check whether this was a main thread, and if so,
1741 * return with the return value.
1743 * We also end up here if the thread kills itself with an
1744 * uncaught exception, see Exception.cmm.
1746 IF_DEBUG(scheduler,debugBelch("--++ thread %d (%s) finished\n",
1747 t->id, whatNext_strs[t->what_next]));
1750 endThread(t, CurrentProc); // clean-up the thread
1751 #elif defined(PARALLEL_HASKELL)
1752 /* For now all are advisory -- HWL */
1753 //if(t->priority==AdvisoryPriority) ??
1754 advisory_thread_count--; // JB: Caution with this counter, buggy!
1757 if(t->dist.priority==RevalPriority)
1761 # if defined(EDENOLD)
1762 // the thread could still have an outport... (BUG)
1763 if (t->eden.outport != -1) {
1764 // delete the outport for the tso which has finished...
1765 IF_PAR_DEBUG(eden_ports,
1766 debugBelch("WARNING: Scheduler removes outport %d for TSO %d.\n",
1767 t->eden.outport, t->id));
1770 // thread still in the process (HEAVY BUG! since outport has just been closed...)
1771 if (t->eden.epid != -1) {
1772 IF_PAR_DEBUG(eden_ports,
1773 debugBelch("WARNING: Scheduler removes TSO %d from process %d .\n",
1774 t->id, t->eden.epid));
1775 removeTSOfromProcess(t);
1780 if (RtsFlags.ParFlags.ParStats.Full &&
1781 !RtsFlags.ParFlags.ParStats.Suppressed)
1782 DumpEndEvent(CURRENT_PROC, t, rtsFalse /* not mandatory */);
1784 // t->par only contains statistics: left out for now...
1786 debugBelch("**** end thread: ended sparked thread %d (%lx); sparkname: %lx\n",
1787 t->id,t,t->par.sparkname));
1789 #endif // PARALLEL_HASKELL
1792 // Check whether the thread that just completed was a bound
1793 // thread, and if so return with the result.
1795 // There is an assumption here that all thread completion goes
1796 // through this point; we need to make sure that if a thread
1797 // ends up in the ThreadKilled state, that it stays on the run
1798 // queue so it can be dealt with here.
1803 if (t->bound != task) {
1804 #if !defined(THREADED_RTS)
1805 // Must be a bound thread that is not the topmost one. Leave
1806 // it on the run queue until the stack has unwound to the
1807 // point where we can deal with this. Leaving it on the run
1808 // queue also ensures that the garbage collector knows about
1809 // this thread and its return value (it gets dropped from the
1810 // all_threads list so there's no other way to find it).
1811 appendToRunQueue(cap,t);
1814 // this cannot happen in the threaded RTS, because a
1815 // bound thread can only be run by the appropriate Task.
1816 barf("finished bound thread that isn't mine");
1820 ASSERT(task->tso == t);
1822 if (t->what_next == ThreadComplete) {
1824 // NOTE: return val is tso->sp[1] (see StgStartup.hc)
1825 *(task->ret) = (StgClosure *)task->tso->sp[1];
1827 task->stat = Success;
1830 *(task->ret) = NULL;
1833 task->stat = Interrupted;
1835 task->stat = Killed;
1839 removeThreadLabel((StgWord)task->tso->id);
1841 return rtsTrue; // tells schedule() to return
1847 /* -----------------------------------------------------------------------------
1848 * Perform a heap census, if PROFILING
1849 * -------------------------------------------------------------------------- */
1852 scheduleDoHeapProfile( rtsBool ready_to_gc STG_UNUSED )
1854 #if defined(PROFILING)
1855 // When we have +RTS -i0 and we're heap profiling, do a census at
1856 // every GC. This lets us get repeatable runs for debugging.
1857 if (performHeapProfile ||
1858 (RtsFlags.ProfFlags.profileInterval==0 &&
1859 RtsFlags.ProfFlags.doHeapProfile && ready_to_gc)) {
1860 GarbageCollect(GetRoots, rtsTrue);
1862 performHeapProfile = rtsFalse;
1863 return rtsTrue; // true <=> we already GC'd
1869 /* -----------------------------------------------------------------------------
1870 * Perform a garbage collection if necessary
1871 * -------------------------------------------------------------------------- */
1874 scheduleDoGC (Capability *cap, Task *task USED_IF_THREADS,
1875 rtsBool force_major, void (*get_roots)(evac_fn))
1879 static volatile StgWord waiting_for_gc;
1880 rtsBool was_waiting;
1885 // In order to GC, there must be no threads running Haskell code.
1886 // Therefore, the GC thread needs to hold *all* the capabilities,
1887 // and release them after the GC has completed.
1889 // This seems to be the simplest way: previous attempts involved
1890 // making all the threads with capabilities give up their
1891 // capabilities and sleep except for the *last* one, which
1892 // actually did the GC. But it's quite hard to arrange for all
1893 // the other tasks to sleep and stay asleep.
1896 was_waiting = cas(&waiting_for_gc, 0, 1);
1899 IF_DEBUG(scheduler, sched_belch("someone else is trying to GC..."));
1900 if (cap) yieldCapability(&cap,task);
1901 } while (waiting_for_gc);
1905 for (i=0; i < n_capabilities; i++) {
1906 IF_DEBUG(scheduler, sched_belch("ready_to_gc, grabbing all the capabilies (%d/%d)", i, n_capabilities));
1907 if (cap != &capabilities[i]) {
1908 Capability *pcap = &capabilities[i];
1909 // we better hope this task doesn't get migrated to
1910 // another Capability while we're waiting for this one.
1911 // It won't, because load balancing happens while we have
1912 // all the Capabilities, but even so it's a slightly
1913 // unsavoury invariant.
1916 waitForReturnCapability(&pcap, task);
1917 if (pcap != &capabilities[i]) {
1918 barf("scheduleDoGC: got the wrong capability");
1923 waiting_for_gc = rtsFalse;
1926 /* Kick any transactions which are invalid back to their
1927 * atomically frames. When next scheduled they will try to
1928 * commit, this commit will fail and they will retry.
1933 for (t = all_threads; t != END_TSO_QUEUE; t = next) {
1934 if (t->what_next == ThreadRelocated) {
1937 next = t->global_link;
1938 if (t -> trec != NO_TREC && t -> why_blocked == NotBlocked) {
1939 if (!stmValidateNestOfTransactions (t -> trec)) {
1940 IF_DEBUG(stm, sched_belch("trec %p found wasting its time", t));
1942 // strip the stack back to the
1943 // ATOMICALLY_FRAME, aborting the (nested)
1944 // transaction, and saving the stack of any
1945 // partially-evaluated thunks on the heap.
1946 raiseAsync_(&capabilities[0], t, NULL, rtsTrue, NULL);
1949 ASSERT(get_itbl((StgClosure *)t->sp)->type == ATOMICALLY_FRAME);
1957 // so this happens periodically:
1958 if (cap) scheduleCheckBlackHoles(cap);
1960 IF_DEBUG(scheduler, printAllThreads());
1962 /* everybody back, start the GC.
1963 * Could do it in this thread, or signal a condition var
1964 * to do it in another thread. Either way, we need to
1965 * broadcast on gc_pending_cond afterward.
1967 #if defined(THREADED_RTS)
1968 IF_DEBUG(scheduler,sched_belch("doing GC"));
1970 GarbageCollect(get_roots, force_major);
1972 #if defined(THREADED_RTS)
1973 // release our stash of capabilities.
1974 for (i = 0; i < n_capabilities; i++) {
1975 if (cap != &capabilities[i]) {
1976 task->cap = &capabilities[i];
1977 releaseCapability(&capabilities[i]);
1988 /* add a ContinueThread event to continue execution of current thread */
1989 new_event(CurrentProc, CurrentProc, CurrentTime[CurrentProc],
1991 t, (StgClosure*)NULL, (rtsSpark*)NULL);
1993 debugBelch("GRAN: eventq and runnableq after Garbage collection:\n\n");
1999 /* ---------------------------------------------------------------------------
2000 * rtsSupportsBoundThreads(): is the RTS built to support bound threads?
2001 * used by Control.Concurrent for error checking.
2002 * ------------------------------------------------------------------------- */
2005 rtsSupportsBoundThreads(void)
2007 #if defined(THREADED_RTS)
2014 /* ---------------------------------------------------------------------------
2015 * isThreadBound(tso): check whether tso is bound to an OS thread.
2016 * ------------------------------------------------------------------------- */
2019 isThreadBound(StgTSO* tso USED_IF_THREADS)
2021 #if defined(THREADED_RTS)
2022 return (tso->bound != NULL);
2027 /* ---------------------------------------------------------------------------
2028 * Singleton fork(). Do not copy any running threads.
2029 * ------------------------------------------------------------------------- */
2031 #if !defined(mingw32_HOST_OS)
2032 #define FORKPROCESS_PRIMOP_SUPPORTED
2035 #ifdef FORKPROCESS_PRIMOP_SUPPORTED
2037 deleteThreadImmediately(Capability *cap, StgTSO *tso);
2040 forkProcess(HsStablePtr *entry
2041 #ifndef FORKPROCESS_PRIMOP_SUPPORTED
2046 #ifdef FORKPROCESS_PRIMOP_SUPPORTED
2052 #if defined(THREADED_RTS)
2053 if (RtsFlags.ParFlags.nNodes > 1) {
2054 errorBelch("forking not supported with +RTS -N<n> greater than 1");
2055 stg_exit(EXIT_FAILURE);
2059 IF_DEBUG(scheduler,sched_belch("forking!"));
2061 // ToDo: for SMP, we should probably acquire *all* the capabilities
2066 if (pid) { // parent
2068 // just return the pid
2074 // delete all threads
2075 cap->run_queue_hd = END_TSO_QUEUE;
2076 cap->run_queue_tl = END_TSO_QUEUE;
2078 for (t = all_threads; t != END_TSO_QUEUE; t = next) {
2081 // don't allow threads to catch the ThreadKilled exception
2082 deleteThreadImmediately(cap,t);
2085 // wipe the task list
2086 ACQUIRE_LOCK(&sched_mutex);
2087 for (task = all_tasks; task != NULL; task=task->all_link) {
2088 if (task != cap->running_task) discardTask(task);
2090 RELEASE_LOCK(&sched_mutex);
2092 cap->suspended_ccalling_tasks = NULL;
2094 #if defined(THREADED_RTS)
2095 // wipe our spare workers list.
2096 cap->spare_workers = NULL;
2097 cap->returning_tasks_hd = NULL;
2098 cap->returning_tasks_tl = NULL;
2101 cap = rts_evalStableIO(cap, entry, NULL); // run the action
2102 rts_checkSchedStatus("forkProcess",cap);
2105 hs_exit(); // clean up and exit
2106 stg_exit(EXIT_SUCCESS);
2108 #else /* !FORKPROCESS_PRIMOP_SUPPORTED */
2109 barf("forkProcess#: primop not supported on this platform, sorry!\n");
2114 /* ---------------------------------------------------------------------------
2115 * Delete the threads on the run queue of the current capability.
2116 * ------------------------------------------------------------------------- */
2119 deleteRunQueue (Capability *cap)
2122 for (t = cap->run_queue_hd; t != END_TSO_QUEUE; t = next) {
2123 ASSERT(t->what_next != ThreadRelocated);
2125 deleteThread(cap, t);
2129 /* startThread and insertThread are now in GranSim.c -- HWL */
2132 /* -----------------------------------------------------------------------------
2133 Managing the suspended_ccalling_tasks list.
2134 Locks required: sched_mutex
2135 -------------------------------------------------------------------------- */
2138 suspendTask (Capability *cap, Task *task)
2140 ASSERT(task->next == NULL && task->prev == NULL);
2141 task->next = cap->suspended_ccalling_tasks;
2143 if (cap->suspended_ccalling_tasks) {
2144 cap->suspended_ccalling_tasks->prev = task;
2146 cap->suspended_ccalling_tasks = task;
2150 recoverSuspendedTask (Capability *cap, Task *task)
2153 task->prev->next = task->next;
2155 ASSERT(cap->suspended_ccalling_tasks == task);
2156 cap->suspended_ccalling_tasks = task->next;
2159 task->next->prev = task->prev;
2161 task->next = task->prev = NULL;
2164 /* ---------------------------------------------------------------------------
2165 * Suspending & resuming Haskell threads.
2167 * When making a "safe" call to C (aka _ccall_GC), the task gives back
2168 * its capability before calling the C function. This allows another
2169 * task to pick up the capability and carry on running Haskell
2170 * threads. It also means that if the C call blocks, it won't lock
2173 * The Haskell thread making the C call is put to sleep for the
2174 * duration of the call, on the susepended_ccalling_threads queue. We
2175 * give out a token to the task, which it can use to resume the thread
2176 * on return from the C function.
2177 * ------------------------------------------------------------------------- */
2180 suspendThread (StgRegTable *reg)
2183 int saved_errno = errno;
2187 /* assume that *reg is a pointer to the StgRegTable part of a Capability.
2189 cap = regTableToCapability(reg);
2191 task = cap->running_task;
2192 tso = cap->r.rCurrentTSO;
2195 sched_belch("thread %d did a safe foreign call", cap->r.rCurrentTSO->id));
2197 // XXX this might not be necessary --SDM
2198 tso->what_next = ThreadRunGHC;
2200 threadPaused(cap,tso);
2202 if(tso->blocked_exceptions == NULL) {
2203 tso->why_blocked = BlockedOnCCall;
2204 tso->blocked_exceptions = END_TSO_QUEUE;
2206 tso->why_blocked = BlockedOnCCall_NoUnblockExc;
2209 // Hand back capability
2210 task->suspended_tso = tso;
2212 ACQUIRE_LOCK(&cap->lock);
2214 suspendTask(cap,task);
2215 cap->in_haskell = rtsFalse;
2216 releaseCapability_(cap);
2218 RELEASE_LOCK(&cap->lock);
2220 #if defined(THREADED_RTS)
2221 /* Preparing to leave the RTS, so ensure there's a native thread/task
2222 waiting to take over.
2224 IF_DEBUG(scheduler, sched_belch("thread %d: leaving RTS", tso->id));
2227 errno = saved_errno;
2232 resumeThread (void *task_)
2236 int saved_errno = errno;
2240 // Wait for permission to re-enter the RTS with the result.
2241 waitForReturnCapability(&cap,task);
2242 // we might be on a different capability now... but if so, our
2243 // entry on the suspended_ccalling_tasks list will also have been
2246 // Remove the thread from the suspended list
2247 recoverSuspendedTask(cap,task);
2249 tso = task->suspended_tso;
2250 task->suspended_tso = NULL;
2251 tso->link = END_TSO_QUEUE;
2252 IF_DEBUG(scheduler, sched_belch("thread %d: re-entering RTS", tso->id));
2254 if (tso->why_blocked == BlockedOnCCall) {
2255 awakenBlockedQueue(cap,tso->blocked_exceptions);
2256 tso->blocked_exceptions = NULL;
2259 /* Reset blocking status */
2260 tso->why_blocked = NotBlocked;
2262 cap->r.rCurrentTSO = tso;
2263 cap->in_haskell = rtsTrue;
2264 errno = saved_errno;
2266 /* We might have GC'd, mark the TSO dirty again */
2272 /* ---------------------------------------------------------------------------
2273 * Comparing Thread ids.
2275 * This is used from STG land in the implementation of the
2276 * instances of Eq/Ord for ThreadIds.
2277 * ------------------------------------------------------------------------ */
2280 cmp_thread(StgPtr tso1, StgPtr tso2)
2282 StgThreadID id1 = ((StgTSO *)tso1)->id;
2283 StgThreadID id2 = ((StgTSO *)tso2)->id;
2285 if (id1 < id2) return (-1);
2286 if (id1 > id2) return 1;
2290 /* ---------------------------------------------------------------------------
2291 * Fetching the ThreadID from an StgTSO.
2293 * This is used in the implementation of Show for ThreadIds.
2294 * ------------------------------------------------------------------------ */
2296 rts_getThreadId(StgPtr tso)
2298 return ((StgTSO *)tso)->id;
2303 labelThread(StgPtr tso, char *label)
2308 /* Caveat: Once set, you can only set the thread name to "" */
2309 len = strlen(label)+1;
2310 buf = stgMallocBytes(len * sizeof(char), "Schedule.c:labelThread()");
2311 strncpy(buf,label,len);
2312 /* Update will free the old memory for us */
2313 updateThreadLabel(((StgTSO *)tso)->id,buf);
2317 /* ---------------------------------------------------------------------------
2318 Create a new thread.
2320 The new thread starts with the given stack size. Before the
2321 scheduler can run, however, this thread needs to have a closure
2322 (and possibly some arguments) pushed on its stack. See
2323 pushClosure() in Schedule.h.
2325 createGenThread() and createIOThread() (in SchedAPI.h) are
2326 convenient packaged versions of this function.
2328 currently pri (priority) is only used in a GRAN setup -- HWL
2329 ------------------------------------------------------------------------ */
2331 /* currently pri (priority) is only used in a GRAN setup -- HWL */
2333 createThread(nat size, StgInt pri)
2336 createThread(Capability *cap, nat size)
2342 /* sched_mutex is *not* required */
2344 /* First check whether we should create a thread at all */
2345 #if defined(PARALLEL_HASKELL)
2346 /* check that no more than RtsFlags.ParFlags.maxThreads threads are created */
2347 if (advisory_thread_count >= RtsFlags.ParFlags.maxThreads) {
2349 debugBelch("{createThread}Daq ghuH: refusing to create another thread; no more than %d threads allowed (currently %d)\n",
2350 RtsFlags.ParFlags.maxThreads, advisory_thread_count);
2351 return END_TSO_QUEUE;
2357 ASSERT(!RtsFlags.GranFlags.Light || CurrentProc==0);
2360 // ToDo: check whether size = stack_size - TSO_STRUCT_SIZEW
2362 /* catch ridiculously small stack sizes */
2363 if (size < MIN_STACK_WORDS + TSO_STRUCT_SIZEW) {
2364 size = MIN_STACK_WORDS + TSO_STRUCT_SIZEW;
2367 stack_size = size - TSO_STRUCT_SIZEW;
2369 tso = (StgTSO *)allocateLocal(cap, size);
2370 TICK_ALLOC_TSO(stack_size, 0);
2372 SET_HDR(tso, &stg_TSO_info, CCS_SYSTEM);
2374 SET_GRAN_HDR(tso, ThisPE);
2377 // Always start with the compiled code evaluator
2378 tso->what_next = ThreadRunGHC;
2380 tso->why_blocked = NotBlocked;
2381 tso->blocked_exceptions = NULL;
2382 tso->flags = TSO_DIRTY;
2384 tso->saved_errno = 0;
2387 tso->stack_size = stack_size;
2388 tso->max_stack_size = round_to_mblocks(RtsFlags.GcFlags.maxStkSize)
2390 tso->sp = (P_)&(tso->stack) + stack_size;
2392 tso->trec = NO_TREC;
2395 tso->prof.CCCS = CCS_MAIN;
2398 /* put a stop frame on the stack */
2399 tso->sp -= sizeofW(StgStopFrame);
2400 SET_HDR((StgClosure*)tso->sp,(StgInfoTable *)&stg_stop_thread_info,CCS_SYSTEM);
2401 tso->link = END_TSO_QUEUE;
2405 /* uses more flexible routine in GranSim */
2406 insertThread(tso, CurrentProc);
2408 /* In a non-GranSim setup the pushing of a TSO onto the runq is separated
2414 if (RtsFlags.GranFlags.GranSimStats.Full)
2415 DumpGranEvent(GR_START,tso);
2416 #elif defined(PARALLEL_HASKELL)
2417 if (RtsFlags.ParFlags.ParStats.Full)
2418 DumpGranEvent(GR_STARTQ,tso);
2419 /* HACk to avoid SCHEDULE
2423 /* Link the new thread on the global thread list.
2425 ACQUIRE_LOCK(&sched_mutex);
2426 tso->id = next_thread_id++; // while we have the mutex
2427 tso->global_link = all_threads;
2429 RELEASE_LOCK(&sched_mutex);
2432 tso->dist.priority = MandatoryPriority; //by default that is...
2436 tso->gran.pri = pri;
2438 tso->gran.magic = TSO_MAGIC; // debugging only
2440 tso->gran.sparkname = 0;
2441 tso->gran.startedat = CURRENT_TIME;
2442 tso->gran.exported = 0;
2443 tso->gran.basicblocks = 0;
2444 tso->gran.allocs = 0;
2445 tso->gran.exectime = 0;
2446 tso->gran.fetchtime = 0;
2447 tso->gran.fetchcount = 0;
2448 tso->gran.blocktime = 0;
2449 tso->gran.blockcount = 0;
2450 tso->gran.blockedat = 0;
2451 tso->gran.globalsparks = 0;
2452 tso->gran.localsparks = 0;
2453 if (RtsFlags.GranFlags.Light)
2454 tso->gran.clock = Now; /* local clock */
2456 tso->gran.clock = 0;
2458 IF_DEBUG(gran,printTSO(tso));
2459 #elif defined(PARALLEL_HASKELL)
2461 tso->par.magic = TSO_MAGIC; // debugging only
2463 tso->par.sparkname = 0;
2464 tso->par.startedat = CURRENT_TIME;
2465 tso->par.exported = 0;
2466 tso->par.basicblocks = 0;
2467 tso->par.allocs = 0;
2468 tso->par.exectime = 0;
2469 tso->par.fetchtime = 0;
2470 tso->par.fetchcount = 0;
2471 tso->par.blocktime = 0;
2472 tso->par.blockcount = 0;
2473 tso->par.blockedat = 0;
2474 tso->par.globalsparks = 0;
2475 tso->par.localsparks = 0;
2479 globalGranStats.tot_threads_created++;
2480 globalGranStats.threads_created_on_PE[CurrentProc]++;
2481 globalGranStats.tot_sq_len += spark_queue_len(CurrentProc);
2482 globalGranStats.tot_sq_probes++;
2483 #elif defined(PARALLEL_HASKELL)
2484 // collect parallel global statistics (currently done together with GC stats)
2485 if (RtsFlags.ParFlags.ParStats.Global &&
2486 RtsFlags.GcFlags.giveStats > NO_GC_STATS) {
2487 //debugBelch("Creating thread %d @ %11.2f\n", tso->id, usertime());
2488 globalParStats.tot_threads_created++;
2494 sched_belch("==__ schedule: Created TSO %d (%p);",
2495 CurrentProc, tso, tso->id));
2496 #elif defined(PARALLEL_HASKELL)
2497 IF_PAR_DEBUG(verbose,
2498 sched_belch("==__ schedule: Created TSO %d (%p); %d threads active",
2499 (long)tso->id, tso, advisory_thread_count));
2501 IF_DEBUG(scheduler,sched_belch("created thread %ld, stack size = %lx words",
2502 (long)tso->id, (long)tso->stack_size));
2509 all parallel thread creation calls should fall through the following routine.
2512 createThreadFromSpark(rtsSpark spark)
2514 ASSERT(spark != (rtsSpark)NULL);
2515 // JB: TAKE CARE OF THIS COUNTER! BUGGY
2516 if (advisory_thread_count >= RtsFlags.ParFlags.maxThreads)
2518 barf("{createSparkThread}Daq ghuH: refusing to create another thread; no more than %d threads allowed (currently %d)",
2519 RtsFlags.ParFlags.maxThreads, advisory_thread_count);
2520 return END_TSO_QUEUE;
2524 tso = createThread(RtsFlags.GcFlags.initialStkSize);
2525 if (tso==END_TSO_QUEUE)
2526 barf("createSparkThread: Cannot create TSO");
2528 tso->priority = AdvisoryPriority;
2530 pushClosure(tso,spark);
2532 advisory_thread_count++; // JB: TAKE CARE OF THIS COUNTER! BUGGY
2539 Turn a spark into a thread.
2540 ToDo: fix for SMP (needs to acquire SCHED_MUTEX!)
2544 activateSpark (rtsSpark spark)
2548 tso = createSparkThread(spark);
2549 if (RtsFlags.ParFlags.ParStats.Full) {
2550 //ASSERT(run_queue_hd == END_TSO_QUEUE); // I think ...
2551 IF_PAR_DEBUG(verbose,
2552 debugBelch("==^^ activateSpark: turning spark of closure %p (%s) into a thread\n",
2553 (StgClosure *)spark, info_type((StgClosure *)spark)));
2555 // ToDo: fwd info on local/global spark to thread -- HWL
2556 // tso->gran.exported = spark->exported;
2557 // tso->gran.locked = !spark->global;
2558 // tso->gran.sparkname = spark->name;
2564 /* ---------------------------------------------------------------------------
2567 * scheduleThread puts a thread on the end of the runnable queue.
2568 * This will usually be done immediately after a thread is created.
2569 * The caller of scheduleThread must create the thread using e.g.
2570 * createThread and push an appropriate closure
2571 * on this thread's stack before the scheduler is invoked.
2572 * ------------------------------------------------------------------------ */
2575 scheduleThread(Capability *cap, StgTSO *tso)
2577 // The thread goes at the *end* of the run-queue, to avoid possible
2578 // starvation of any threads already on the queue.
2579 appendToRunQueue(cap,tso);
2583 scheduleWaitThread (StgTSO* tso, /*[out]*/HaskellObj* ret, Capability *cap)
2587 // We already created/initialised the Task
2588 task = cap->running_task;
2590 // This TSO is now a bound thread; make the Task and TSO
2591 // point to each other.
2596 task->stat = NoStatus;
2598 appendToRunQueue(cap,tso);
2600 IF_DEBUG(scheduler, sched_belch("new bound thread (%d)", tso->id));
2603 /* GranSim specific init */
2604 CurrentTSO = m->tso; // the TSO to run
2605 procStatus[MainProc] = Busy; // status of main PE
2606 CurrentProc = MainProc; // PE to run it on
2609 cap = schedule(cap,task);
2611 ASSERT(task->stat != NoStatus);
2612 ASSERT_FULL_CAPABILITY_INVARIANTS(cap,task);
2614 IF_DEBUG(scheduler, sched_belch("bound thread (%d) finished", task->tso->id));
2618 /* ----------------------------------------------------------------------------
2620 * ------------------------------------------------------------------------- */
2622 #if defined(THREADED_RTS)
2624 workerStart(Task *task)
2628 // See startWorkerTask().
2629 ACQUIRE_LOCK(&task->lock);
2631 RELEASE_LOCK(&task->lock);
2633 // set the thread-local pointer to the Task:
2636 // schedule() runs without a lock.
2637 cap = schedule(cap,task);
2639 // On exit from schedule(), we have a Capability.
2640 releaseCapability(cap);
2645 /* ---------------------------------------------------------------------------
2648 * Initialise the scheduler. This resets all the queues - if the
2649 * queues contained any threads, they'll be garbage collected at the
2652 * ------------------------------------------------------------------------ */
2659 for (i=0; i<=MAX_PROC; i++) {
2660 run_queue_hds[i] = END_TSO_QUEUE;
2661 run_queue_tls[i] = END_TSO_QUEUE;
2662 blocked_queue_hds[i] = END_TSO_QUEUE;
2663 blocked_queue_tls[i] = END_TSO_QUEUE;
2664 ccalling_threadss[i] = END_TSO_QUEUE;
2665 blackhole_queue[i] = END_TSO_QUEUE;
2666 sleeping_queue = END_TSO_QUEUE;
2668 #elif !defined(THREADED_RTS)
2669 blocked_queue_hd = END_TSO_QUEUE;
2670 blocked_queue_tl = END_TSO_QUEUE;
2671 sleeping_queue = END_TSO_QUEUE;
2674 blackhole_queue = END_TSO_QUEUE;
2675 all_threads = END_TSO_QUEUE;
2680 RtsFlags.ConcFlags.ctxtSwitchTicks =
2681 RtsFlags.ConcFlags.ctxtSwitchTime / TICK_MILLISECS;
2683 #if defined(THREADED_RTS)
2684 /* Initialise the mutex and condition variables used by
2686 initMutex(&sched_mutex);
2689 ACQUIRE_LOCK(&sched_mutex);
2691 /* A capability holds the state a native thread needs in
2692 * order to execute STG code. At least one capability is
2693 * floating around (only THREADED_RTS builds have more than one).
2699 #if defined(THREADED_RTS) || defined(PARALLEL_HASKELL)
2703 #if defined(THREADED_RTS)
2705 * Eagerly start one worker to run each Capability, except for
2706 * Capability 0. The idea is that we're probably going to start a
2707 * bound thread on Capability 0 pretty soon, so we don't want a
2708 * worker task hogging it.
2713 for (i = 1; i < n_capabilities; i++) {
2714 cap = &capabilities[i];
2715 ACQUIRE_LOCK(&cap->lock);
2716 startWorkerTask(cap, workerStart);
2717 RELEASE_LOCK(&cap->lock);
2722 RELEASE_LOCK(&sched_mutex);
2726 exitScheduler( void )
2728 interrupted = rtsTrue;
2729 shutting_down_scheduler = rtsTrue;
2731 #if defined(THREADED_RTS)
2736 ACQUIRE_LOCK(&sched_mutex);
2737 task = newBoundTask();
2738 RELEASE_LOCK(&sched_mutex);
2740 for (i = 0; i < n_capabilities; i++) {
2741 shutdownCapability(&capabilities[i], task);
2743 boundTaskExiting(task);
2749 /* ---------------------------------------------------------------------------
2750 Where are the roots that we know about?
2752 - all the threads on the runnable queue
2753 - all the threads on the blocked queue
2754 - all the threads on the sleeping queue
2755 - all the thread currently executing a _ccall_GC
2756 - all the "main threads"
2758 ------------------------------------------------------------------------ */
2760 /* This has to be protected either by the scheduler monitor, or by the
2761 garbage collection monitor (probably the latter).
2766 GetRoots( evac_fn evac )
2773 for (i=0; i<=RtsFlags.GranFlags.proc; i++) {
2774 if ((run_queue_hds[i] != END_TSO_QUEUE) && ((run_queue_hds[i] != NULL)))
2775 evac((StgClosure **)&run_queue_hds[i]);
2776 if ((run_queue_tls[i] != END_TSO_QUEUE) && ((run_queue_tls[i] != NULL)))
2777 evac((StgClosure **)&run_queue_tls[i]);
2779 if ((blocked_queue_hds[i] != END_TSO_QUEUE) && ((blocked_queue_hds[i] != NULL)))
2780 evac((StgClosure **)&blocked_queue_hds[i]);
2781 if ((blocked_queue_tls[i] != END_TSO_QUEUE) && ((blocked_queue_tls[i] != NULL)))
2782 evac((StgClosure **)&blocked_queue_tls[i]);
2783 if ((ccalling_threadss[i] != END_TSO_QUEUE) && ((ccalling_threadss[i] != NULL)))
2784 evac((StgClosure **)&ccalling_threads[i]);
2791 for (i = 0; i < n_capabilities; i++) {
2792 cap = &capabilities[i];
2793 evac((StgClosure **)&cap->run_queue_hd);
2794 evac((StgClosure **)&cap->run_queue_tl);
2796 for (task = cap->suspended_ccalling_tasks; task != NULL;
2798 evac((StgClosure **)&task->suspended_tso);
2802 #if !defined(THREADED_RTS)
2803 evac((StgClosure **)(void *)&blocked_queue_hd);
2804 evac((StgClosure **)(void *)&blocked_queue_tl);
2805 evac((StgClosure **)(void *)&sleeping_queue);
2809 // evac((StgClosure **)&blackhole_queue);
2811 #if defined(THREADED_RTS) || defined(PARALLEL_HASKELL) || defined(GRAN)
2812 markSparkQueue(evac);
2815 #if defined(RTS_USER_SIGNALS)
2816 // mark the signal handlers (signals should be already blocked)
2817 markSignalHandlers(evac);
2821 /* -----------------------------------------------------------------------------
2824 This is the interface to the garbage collector from Haskell land.
2825 We provide this so that external C code can allocate and garbage
2826 collect when called from Haskell via _ccall_GC.
2828 It might be useful to provide an interface whereby the programmer
2829 can specify more roots (ToDo).
2831 This needs to be protected by the GC condition variable above. KH.
2832 -------------------------------------------------------------------------- */
2834 static void (*extra_roots)(evac_fn);
2837 performGC_(rtsBool force_major, void (*get_roots)(evac_fn))
2839 Task *task = myTask();
2842 ACQUIRE_LOCK(&sched_mutex);
2843 task = newBoundTask();
2844 RELEASE_LOCK(&sched_mutex);
2845 scheduleDoGC(NULL,task,force_major, get_roots);
2846 boundTaskExiting(task);
2848 scheduleDoGC(NULL,task,force_major, get_roots);
2855 performGC_(rtsFalse, GetRoots);
2859 performMajorGC(void)
2861 performGC_(rtsTrue, GetRoots);
2865 AllRoots(evac_fn evac)
2867 GetRoots(evac); // the scheduler's roots
2868 extra_roots(evac); // the user's roots
2872 performGCWithRoots(void (*get_roots)(evac_fn))
2874 extra_roots = get_roots;
2875 performGC_(rtsFalse, AllRoots);
2878 /* -----------------------------------------------------------------------------
2881 If the thread has reached its maximum stack size, then raise the
2882 StackOverflow exception in the offending thread. Otherwise
2883 relocate the TSO into a larger chunk of memory and adjust its stack
2885 -------------------------------------------------------------------------- */
2888 threadStackOverflow(Capability *cap, StgTSO *tso)
2890 nat new_stack_size, stack_words;
2895 IF_DEBUG(sanity,checkTSO(tso));
2896 if (tso->stack_size >= tso->max_stack_size) {
2899 debugBelch("@@ threadStackOverflow of TSO %ld (%p): stack too large (now %ld; max is %ld)\n",
2900 (long)tso->id, tso, (long)tso->stack_size, (long)tso->max_stack_size);
2901 /* If we're debugging, just print out the top of the stack */
2902 printStackChunk(tso->sp, stg_min(tso->stack+tso->stack_size,
2905 /* Send this thread the StackOverflow exception */
2906 raiseAsync(cap, tso, (StgClosure *)stackOverflow_closure);
2910 /* Try to double the current stack size. If that takes us over the
2911 * maximum stack size for this thread, then use the maximum instead.
2912 * Finally round up so the TSO ends up as a whole number of blocks.
2914 new_stack_size = stg_min(tso->stack_size * 2, tso->max_stack_size);
2915 new_tso_size = (lnat)BLOCK_ROUND_UP(new_stack_size * sizeof(W_) +
2916 TSO_STRUCT_SIZE)/sizeof(W_);
2917 new_tso_size = round_to_mblocks(new_tso_size); /* Be MBLOCK-friendly */
2918 new_stack_size = new_tso_size - TSO_STRUCT_SIZEW;
2920 IF_DEBUG(scheduler, sched_belch("increasing stack size from %ld words to %d.\n", (long)tso->stack_size, new_stack_size));
2922 dest = (StgTSO *)allocate(new_tso_size);
2923 TICK_ALLOC_TSO(new_stack_size,0);
2925 /* copy the TSO block and the old stack into the new area */
2926 memcpy(dest,tso,TSO_STRUCT_SIZE);
2927 stack_words = tso->stack + tso->stack_size - tso->sp;
2928 new_sp = (P_)dest + new_tso_size - stack_words;
2929 memcpy(new_sp, tso->sp, stack_words * sizeof(W_));
2931 /* relocate the stack pointers... */
2933 dest->stack_size = new_stack_size;
2935 /* Mark the old TSO as relocated. We have to check for relocated
2936 * TSOs in the garbage collector and any primops that deal with TSOs.
2938 * It's important to set the sp value to just beyond the end
2939 * of the stack, so we don't attempt to scavenge any part of the
2942 tso->what_next = ThreadRelocated;
2944 tso->sp = (P_)&(tso->stack[tso->stack_size]);
2945 tso->why_blocked = NotBlocked;
2947 IF_PAR_DEBUG(verbose,
2948 debugBelch("@@ threadStackOverflow of TSO %d (now at %p): stack size increased to %ld\n",
2949 tso->id, tso, tso->stack_size);
2950 /* If we're debugging, just print out the top of the stack */
2951 printStackChunk(tso->sp, stg_min(tso->stack+tso->stack_size,
2954 IF_DEBUG(sanity,checkTSO(tso));
2956 IF_DEBUG(scheduler,printTSO(dest));
2962 /* ---------------------------------------------------------------------------
2963 Wake up a queue that was blocked on some resource.
2964 ------------------------------------------------------------------------ */
2968 unblockCount ( StgBlockingQueueElement *bqe, StgClosure *node )
2971 #elif defined(PARALLEL_HASKELL)
2973 unblockCount ( StgBlockingQueueElement *bqe, StgClosure *node )
2975 /* write RESUME events to log file and
2976 update blocked and fetch time (depending on type of the orig closure) */
2977 if (RtsFlags.ParFlags.ParStats.Full) {
2978 DumpRawGranEvent(CURRENT_PROC, CURRENT_PROC,
2979 GR_RESUMEQ, ((StgTSO *)bqe), ((StgTSO *)bqe)->block_info.closure,
2980 0, 0 /* spark_queue_len(ADVISORY_POOL) */);
2981 if (emptyRunQueue())
2982 emitSchedule = rtsTrue;
2984 switch (get_itbl(node)->type) {
2986 ((StgTSO *)bqe)->par.fetchtime += CURRENT_TIME-((StgTSO *)bqe)->par.blockedat;
2991 ((StgTSO *)bqe)->par.blocktime += CURRENT_TIME-((StgTSO *)bqe)->par.blockedat;
2998 barf("{unblockOne}Daq Qagh: unexpected closure in blocking queue");
3005 StgBlockingQueueElement *
3006 unblockOne(StgBlockingQueueElement *bqe, StgClosure *node)
3009 PEs node_loc, tso_loc;
3011 node_loc = where_is(node); // should be lifted out of loop
3012 tso = (StgTSO *)bqe; // wastes an assignment to get the type right
3013 tso_loc = where_is((StgClosure *)tso);
3014 if (IS_LOCAL_TO(PROCS(node),tso_loc)) { // TSO is local
3015 /* !fake_fetch => TSO is on CurrentProc is same as IS_LOCAL_TO */
3016 ASSERT(CurrentProc!=node_loc || tso_loc==CurrentProc);
3017 CurrentTime[CurrentProc] += RtsFlags.GranFlags.Costs.lunblocktime;
3018 // insertThread(tso, node_loc);
3019 new_event(tso_loc, tso_loc, CurrentTime[CurrentProc],
3021 tso, node, (rtsSpark*)NULL);
3022 tso->link = END_TSO_QUEUE; // overwrite link just to be sure
3025 } else { // TSO is remote (actually should be FMBQ)
3026 CurrentTime[CurrentProc] += RtsFlags.GranFlags.Costs.mpacktime +
3027 RtsFlags.GranFlags.Costs.gunblocktime +
3028 RtsFlags.GranFlags.Costs.latency;
3029 new_event(tso_loc, CurrentProc, CurrentTime[CurrentProc],
3031 tso, node, (rtsSpark*)NULL);
3032 tso->link = END_TSO_QUEUE; // overwrite link just to be sure
3035 /* the thread-queue-overhead is accounted for in either Resume or UnblockThread */
3037 debugBelch(" %s TSO %d (%p) [PE %d] (block_info.closure=%p) (next=%p) ,",
3038 (node_loc==tso_loc ? "Local" : "Global"),
3039 tso->id, tso, CurrentProc, tso->block_info.closure, tso->link));
3040 tso->block_info.closure = NULL;
3041 IF_DEBUG(scheduler,debugBelch("-- Waking up thread %ld (%p)\n",
3044 #elif defined(PARALLEL_HASKELL)
3045 StgBlockingQueueElement *
3046 unblockOne(StgBlockingQueueElement *bqe, StgClosure *node)
3048 StgBlockingQueueElement *next;
3050 switch (get_itbl(bqe)->type) {
3052 ASSERT(((StgTSO *)bqe)->why_blocked != NotBlocked);
3053 /* if it's a TSO just push it onto the run_queue */
3055 ((StgTSO *)bqe)->link = END_TSO_QUEUE; // debugging?
3056 APPEND_TO_RUN_QUEUE((StgTSO *)bqe);
3058 unblockCount(bqe, node);
3059 /* reset blocking status after dumping event */
3060 ((StgTSO *)bqe)->why_blocked = NotBlocked;
3064 /* if it's a BLOCKED_FETCH put it on the PendingFetches list */
3066 bqe->link = (StgBlockingQueueElement *)PendingFetches;
3067 PendingFetches = (StgBlockedFetch *)bqe;
3071 /* can ignore this case in a non-debugging setup;
3072 see comments on RBHSave closures above */
3074 /* check that the closure is an RBHSave closure */
3075 ASSERT(get_itbl((StgClosure *)bqe) == &stg_RBH_Save_0_info ||
3076 get_itbl((StgClosure *)bqe) == &stg_RBH_Save_1_info ||
3077 get_itbl((StgClosure *)bqe) == &stg_RBH_Save_2_info);
3081 barf("{unblockOne}Daq Qagh: Unexpected IP (%#lx; %s) in blocking queue at %#lx\n",
3082 get_itbl((StgClosure *)bqe), info_type((StgClosure *)bqe),
3086 IF_PAR_DEBUG(bq, debugBelch(", %p (%s)\n", bqe, info_type((StgClosure*)bqe)));
3092 unblockOne(Capability *cap, StgTSO *tso)
3096 ASSERT(get_itbl(tso)->type == TSO);
3097 ASSERT(tso->why_blocked != NotBlocked);
3098 tso->why_blocked = NotBlocked;
3100 tso->link = END_TSO_QUEUE;
3102 // We might have just migrated this TSO to our Capability:
3104 tso->bound->cap = cap;
3107 appendToRunQueue(cap,tso);
3109 // we're holding a newly woken thread, make sure we context switch
3110 // quickly so we can migrate it if necessary.
3112 IF_DEBUG(scheduler,sched_belch("waking up thread %ld", (long)tso->id));
3119 awakenBlockedQueue(StgBlockingQueueElement *q, StgClosure *node)
3121 StgBlockingQueueElement *bqe;
3126 debugBelch("##-_ AwBQ for node %p on PE %d @ %ld by TSO %d (%p): \n", \
3127 node, CurrentProc, CurrentTime[CurrentProc],
3128 CurrentTSO->id, CurrentTSO));
3130 node_loc = where_is(node);
3132 ASSERT(q == END_BQ_QUEUE ||
3133 get_itbl(q)->type == TSO || // q is either a TSO or an RBHSave
3134 get_itbl(q)->type == CONSTR); // closure (type constructor)
3135 ASSERT(is_unique(node));
3137 /* FAKE FETCH: magically copy the node to the tso's proc;
3138 no Fetch necessary because in reality the node should not have been
3139 moved to the other PE in the first place
3141 if (CurrentProc!=node_loc) {
3143 debugBelch("## node %p is on PE %d but CurrentProc is %d (TSO %d); assuming fake fetch and adjusting bitmask (old: %#x)\n",
3144 node, node_loc, CurrentProc, CurrentTSO->id,
3145 // CurrentTSO, where_is(CurrentTSO),
3146 node->header.gran.procs));
3147 node->header.gran.procs = (node->header.gran.procs) | PE_NUMBER(CurrentProc);
3149 debugBelch("## new bitmask of node %p is %#x\n",
3150 node, node->header.gran.procs));
3151 if (RtsFlags.GranFlags.GranSimStats.Global) {
3152 globalGranStats.tot_fake_fetches++;
3157 // ToDo: check: ASSERT(CurrentProc==node_loc);
3158 while (get_itbl(bqe)->type==TSO) { // q != END_TSO_QUEUE) {
3161 bqe points to the current element in the queue
3162 next points to the next element in the queue
3164 //tso = (StgTSO *)bqe; // wastes an assignment to get the type right
3165 //tso_loc = where_is(tso);
3167 bqe = unblockOne(bqe, node);
3170 /* if this is the BQ of an RBH, we have to put back the info ripped out of
3171 the closure to make room for the anchor of the BQ */
3172 if (bqe!=END_BQ_QUEUE) {
3173 ASSERT(get_itbl(node)->type == RBH && get_itbl(bqe)->type == CONSTR);
3175 ASSERT((info_ptr==&RBH_Save_0_info) ||
3176 (info_ptr==&RBH_Save_1_info) ||
3177 (info_ptr==&RBH_Save_2_info));
3179 /* cf. convertToRBH in RBH.c for writing the RBHSave closure */
3180 ((StgRBH *)node)->blocking_queue = (StgBlockingQueueElement *)((StgRBHSave *)bqe)->payload[0];
3181 ((StgRBH *)node)->mut_link = (StgMutClosure *)((StgRBHSave *)bqe)->payload[1];
3184 debugBelch("## Filled in RBH_Save for %p (%s) at end of AwBQ\n",
3185 node, info_type(node)));
3188 /* statistics gathering */
3189 if (RtsFlags.GranFlags.GranSimStats.Global) {
3190 // globalGranStats.tot_bq_processing_time += bq_processing_time;
3191 globalGranStats.tot_bq_len += len; // total length of all bqs awakened
3192 // globalGranStats.tot_bq_len_local += len_local; // same for local TSOs only
3193 globalGranStats.tot_awbq++; // total no. of bqs awakened
3196 debugBelch("## BQ Stats of %p: [%d entries] %s\n",
3197 node, len, (bqe!=END_BQ_QUEUE) ? "RBH" : ""));
3199 #elif defined(PARALLEL_HASKELL)
3201 awakenBlockedQueue(StgBlockingQueueElement *q, StgClosure *node)
3203 StgBlockingQueueElement *bqe;
3205 IF_PAR_DEBUG(verbose,
3206 debugBelch("##-_ AwBQ for node %p on [%x]: \n",
3210 if(get_itbl(q)->type == CONSTR || q==END_BQ_QUEUE) {
3211 IF_PAR_DEBUG(verbose, debugBelch("## ... nothing to unblock so lets just return. RFP (BUG?)\n"));
3216 ASSERT(q == END_BQ_QUEUE ||
3217 get_itbl(q)->type == TSO ||
3218 get_itbl(q)->type == BLOCKED_FETCH ||
3219 get_itbl(q)->type == CONSTR);
3222 while (get_itbl(bqe)->type==TSO ||
3223 get_itbl(bqe)->type==BLOCKED_FETCH) {
3224 bqe = unblockOne(bqe, node);
3228 #else /* !GRAN && !PARALLEL_HASKELL */
3231 awakenBlockedQueue(Capability *cap, StgTSO *tso)
3233 if (tso == NULL) return; // hack; see bug #1235728, and comments in
3235 while (tso != END_TSO_QUEUE) {
3236 tso = unblockOne(cap,tso);
3241 /* ---------------------------------------------------------------------------
3243 - usually called inside a signal handler so it mustn't do anything fancy.
3244 ------------------------------------------------------------------------ */
3247 interruptStgRts(void)
3251 #if defined(THREADED_RTS)
3252 prodAllCapabilities();
3256 /* -----------------------------------------------------------------------------
3259 This is for use when we raise an exception in another thread, which
3261 This has nothing to do with the UnblockThread event in GranSim. -- HWL
3262 -------------------------------------------------------------------------- */
3264 #if defined(GRAN) || defined(PARALLEL_HASKELL)
3266 NB: only the type of the blocking queue is different in GranSim and GUM
3267 the operations on the queue-elements are the same
3268 long live polymorphism!
3270 Locks: sched_mutex is held upon entry and exit.
3274 unblockThread(Capability *cap, StgTSO *tso)
3276 StgBlockingQueueElement *t, **last;
3278 switch (tso->why_blocked) {
3281 return; /* not blocked */
3284 // Be careful: nothing to do here! We tell the scheduler that the thread
3285 // is runnable and we leave it to the stack-walking code to abort the
3286 // transaction while unwinding the stack. We should perhaps have a debugging
3287 // test to make sure that this really happens and that the 'zombie' transaction
3288 // does not get committed.
3292 ASSERT(get_itbl(tso->block_info.closure)->type == MVAR);
3294 StgBlockingQueueElement *last_tso = END_BQ_QUEUE;
3295 StgMVar *mvar = (StgMVar *)(tso->block_info.closure);
3297 last = (StgBlockingQueueElement **)&mvar->head;
3298 for (t = (StgBlockingQueueElement *)mvar->head;
3300 last = &t->link, last_tso = t, t = t->link) {
3301 if (t == (StgBlockingQueueElement *)tso) {
3302 *last = (StgBlockingQueueElement *)tso->link;
3303 if (mvar->tail == tso) {
3304 mvar->tail = (StgTSO *)last_tso;
3309 barf("unblockThread (MVAR): TSO not found");
3312 case BlockedOnBlackHole:
3313 ASSERT(get_itbl(tso->block_info.closure)->type == BLACKHOLE_BQ);
3315 StgBlockingQueue *bq = (StgBlockingQueue *)(tso->block_info.closure);
3317 last = &bq->blocking_queue;
3318 for (t = bq->blocking_queue;
3320 last = &t->link, t = t->link) {
3321 if (t == (StgBlockingQueueElement *)tso) {
3322 *last = (StgBlockingQueueElement *)tso->link;
3326 barf("unblockThread (BLACKHOLE): TSO not found");
3329 case BlockedOnException:
3331 StgTSO *target = tso->block_info.tso;
3333 ASSERT(get_itbl(target)->type == TSO);
3335 if (target->what_next == ThreadRelocated) {
3336 target = target->link;
3337 ASSERT(get_itbl(target)->type == TSO);
3340 ASSERT(target->blocked_exceptions != NULL);
3342 last = (StgBlockingQueueElement **)&target->blocked_exceptions;
3343 for (t = (StgBlockingQueueElement *)target->blocked_exceptions;
3345 last = &t->link, t = t->link) {
3346 ASSERT(get_itbl(t)->type == TSO);
3347 if (t == (StgBlockingQueueElement *)tso) {
3348 *last = (StgBlockingQueueElement *)tso->link;
3352 barf("unblockThread (Exception): TSO not found");
3356 case BlockedOnWrite:
3357 #if defined(mingw32_HOST_OS)
3358 case BlockedOnDoProc:
3361 /* take TSO off blocked_queue */
3362 StgBlockingQueueElement *prev = NULL;
3363 for (t = (StgBlockingQueueElement *)blocked_queue_hd; t != END_BQ_QUEUE;
3364 prev = t, t = t->link) {
3365 if (t == (StgBlockingQueueElement *)tso) {
3367 blocked_queue_hd = (StgTSO *)t->link;
3368 if ((StgBlockingQueueElement *)blocked_queue_tl == t) {
3369 blocked_queue_tl = END_TSO_QUEUE;
3372 prev->link = t->link;
3373 if ((StgBlockingQueueElement *)blocked_queue_tl == t) {
3374 blocked_queue_tl = (StgTSO *)prev;
3377 #if defined(mingw32_HOST_OS)
3378 /* (Cooperatively) signal that the worker thread should abort
3381 abandonWorkRequest(tso->block_info.async_result->reqID);
3386 barf("unblockThread (I/O): TSO not found");
3389 case BlockedOnDelay:
3391 /* take TSO off sleeping_queue */
3392 StgBlockingQueueElement *prev = NULL;
3393 for (t = (StgBlockingQueueElement *)sleeping_queue; t != END_BQ_QUEUE;
3394 prev = t, t = t->link) {
3395 if (t == (StgBlockingQueueElement *)tso) {
3397 sleeping_queue = (StgTSO *)t->link;
3399 prev->link = t->link;
3404 barf("unblockThread (delay): TSO not found");
3408 barf("unblockThread");
3412 tso->link = END_TSO_QUEUE;
3413 tso->why_blocked = NotBlocked;
3414 tso->block_info.closure = NULL;
3415 pushOnRunQueue(cap,tso);
3419 unblockThread(Capability *cap, StgTSO *tso)
3423 /* To avoid locking unnecessarily. */
3424 if (tso->why_blocked == NotBlocked) {
3428 switch (tso->why_blocked) {
3431 // Be careful: nothing to do here! We tell the scheduler that the thread
3432 // is runnable and we leave it to the stack-walking code to abort the
3433 // transaction while unwinding the stack. We should perhaps have a debugging
3434 // test to make sure that this really happens and that the 'zombie' transaction
3435 // does not get committed.
3439 ASSERT(get_itbl(tso->block_info.closure)->type == MVAR);
3441 StgTSO *last_tso = END_TSO_QUEUE;
3442 StgMVar *mvar = (StgMVar *)(tso->block_info.closure);
3445 for (t = mvar->head; t != END_TSO_QUEUE;
3446 last = &t->link, last_tso = t, t = t->link) {
3449 if (mvar->tail == tso) {
3450 mvar->tail = last_tso;
3455 barf("unblockThread (MVAR): TSO not found");
3458 case BlockedOnBlackHole:
3460 last = &blackhole_queue;
3461 for (t = blackhole_queue; t != END_TSO_QUEUE;
3462 last = &t->link, t = t->link) {
3468 barf("unblockThread (BLACKHOLE): TSO not found");
3471 case BlockedOnException:
3473 StgTSO *target = tso->block_info.tso;
3475 ASSERT(get_itbl(target)->type == TSO);
3477 while (target->what_next == ThreadRelocated) {
3478 target = target->link;
3479 ASSERT(get_itbl(target)->type == TSO);
3482 ASSERT(target->blocked_exceptions != NULL);
3484 last = &target->blocked_exceptions;
3485 for (t = target->blocked_exceptions; t != END_TSO_QUEUE;
3486 last = &t->link, t = t->link) {
3487 ASSERT(get_itbl(t)->type == TSO);
3493 barf("unblockThread (Exception): TSO not found");
3496 #if !defined(THREADED_RTS)
3498 case BlockedOnWrite:
3499 #if defined(mingw32_HOST_OS)
3500 case BlockedOnDoProc:
3503 StgTSO *prev = NULL;
3504 for (t = blocked_queue_hd; t != END_TSO_QUEUE;
3505 prev = t, t = t->link) {
3508 blocked_queue_hd = t->link;
3509 if (blocked_queue_tl == t) {
3510 blocked_queue_tl = END_TSO_QUEUE;
3513 prev->link = t->link;
3514 if (blocked_queue_tl == t) {
3515 blocked_queue_tl = prev;
3518 #if defined(mingw32_HOST_OS)
3519 /* (Cooperatively) signal that the worker thread should abort
3522 abandonWorkRequest(tso->block_info.async_result->reqID);
3527 barf("unblockThread (I/O): TSO not found");
3530 case BlockedOnDelay:
3532 StgTSO *prev = NULL;
3533 for (t = sleeping_queue; t != END_TSO_QUEUE;
3534 prev = t, t = t->link) {
3537 sleeping_queue = t->link;
3539 prev->link = t->link;
3544 barf("unblockThread (delay): TSO not found");
3549 barf("unblockThread");
3553 tso->link = END_TSO_QUEUE;
3554 tso->why_blocked = NotBlocked;
3555 tso->block_info.closure = NULL;
3556 appendToRunQueue(cap,tso);
3560 /* -----------------------------------------------------------------------------
3563 * Check the blackhole_queue for threads that can be woken up. We do
3564 * this periodically: before every GC, and whenever the run queue is
3567 * An elegant solution might be to just wake up all the blocked
3568 * threads with awakenBlockedQueue occasionally: they'll go back to
3569 * sleep again if the object is still a BLACKHOLE. Unfortunately this
3570 * doesn't give us a way to tell whether we've actually managed to
3571 * wake up any threads, so we would be busy-waiting.
3573 * -------------------------------------------------------------------------- */
3576 checkBlackHoles (Capability *cap)
3579 rtsBool any_woke_up = rtsFalse;
3582 // blackhole_queue is global:
3583 ASSERT_LOCK_HELD(&sched_mutex);
3585 IF_DEBUG(scheduler, sched_belch("checking threads blocked on black holes"));
3587 // ASSUMES: sched_mutex
3588 prev = &blackhole_queue;
3589 t = blackhole_queue;
3590 while (t != END_TSO_QUEUE) {
3591 ASSERT(t->why_blocked == BlockedOnBlackHole);
3592 type = get_itbl(t->block_info.closure)->type;
3593 if (type != BLACKHOLE && type != CAF_BLACKHOLE) {
3594 IF_DEBUG(sanity,checkTSO(t));
3595 t = unblockOne(cap, t);
3596 // urk, the threads migrate to the current capability
3597 // here, but we'd like to keep them on the original one.
3599 any_woke_up = rtsTrue;
3609 /* -----------------------------------------------------------------------------
3612 * The following function implements the magic for raising an
3613 * asynchronous exception in an existing thread.
3615 * We first remove the thread from any queue on which it might be
3616 * blocked. The possible blockages are MVARs and BLACKHOLE_BQs.
3618 * We strip the stack down to the innermost CATCH_FRAME, building
3619 * thunks in the heap for all the active computations, so they can
3620 * be restarted if necessary. When we reach a CATCH_FRAME, we build
3621 * an application of the handler to the exception, and push it on
3622 * the top of the stack.
3624 * How exactly do we save all the active computations? We create an
3625 * AP_STACK for every UpdateFrame on the stack. Entering one of these
3626 * AP_STACKs pushes everything from the corresponding update frame
3627 * upwards onto the stack. (Actually, it pushes everything up to the
3628 * next update frame plus a pointer to the next AP_STACK object.
3629 * Entering the next AP_STACK object pushes more onto the stack until we
3630 * reach the last AP_STACK object - at which point the stack should look
3631 * exactly as it did when we killed the TSO and we can continue
3632 * execution by entering the closure on top of the stack.
3634 * We can also kill a thread entirely - this happens if either (a) the
3635 * exception passed to raiseAsync is NULL, or (b) there's no
3636 * CATCH_FRAME on the stack. In either case, we strip the entire
3637 * stack and replace the thread with a zombie.
3639 * ToDo: in THREADED_RTS mode, this function is only safe if either
3640 * (a) we hold all the Capabilities (eg. in GC, or if there is only
3641 * one Capability), or (b) we own the Capability that the TSO is
3642 * currently blocked on or on the run queue of.
3644 * -------------------------------------------------------------------------- */
3647 raiseAsync(Capability *cap, StgTSO *tso, StgClosure *exception)
3649 raiseAsync_(cap, tso, exception, rtsFalse, NULL);
3653 suspendComputation(Capability *cap, StgTSO *tso, StgPtr stop_here)
3655 raiseAsync_(cap, tso, NULL, rtsFalse, stop_here);
3659 raiseAsync_(Capability *cap, StgTSO *tso, StgClosure *exception,
3660 rtsBool stop_at_atomically, StgPtr stop_here)
3662 StgRetInfoTable *info;
3666 // Thread already dead?
3667 if (tso->what_next == ThreadComplete || tso->what_next == ThreadKilled) {
3672 sched_belch("raising exception in thread %ld.", (long)tso->id));
3674 // Remove it from any blocking queues
3675 unblockThread(cap,tso);
3677 // mark it dirty; we're about to change its stack.
3682 // The stack freezing code assumes there's a closure pointer on
3683 // the top of the stack, so we have to arrange that this is the case...
3685 if (sp[0] == (W_)&stg_enter_info) {
3689 sp[0] = (W_)&stg_dummy_ret_closure;
3693 while (stop_here == NULL || frame < stop_here) {
3695 // 1. Let the top of the stack be the "current closure"
3697 // 2. Walk up the stack until we find either an UPDATE_FRAME or a
3700 // 3. If it's an UPDATE_FRAME, then make an AP_STACK containing the
3701 // current closure applied to the chunk of stack up to (but not
3702 // including) the update frame. This closure becomes the "current
3703 // closure". Go back to step 2.
3705 // 4. If it's a CATCH_FRAME, then leave the exception handler on
3706 // top of the stack applied to the exception.
3708 // 5. If it's a STOP_FRAME, then kill the thread.
3710 // NB: if we pass an ATOMICALLY_FRAME then abort the associated
3713 info = get_ret_itbl((StgClosure *)frame);
3715 switch (info->i.type) {
3722 // First build an AP_STACK consisting of the stack chunk above the
3723 // current update frame, with the top word on the stack as the
3726 words = frame - sp - 1;
3727 ap = (StgAP_STACK *)allocateLocal(cap,AP_STACK_sizeW(words));
3730 ap->fun = (StgClosure *)sp[0];
3732 for(i=0; i < (nat)words; ++i) {
3733 ap->payload[i] = (StgClosure *)*sp++;
3736 SET_HDR(ap,&stg_AP_STACK_info,
3737 ((StgClosure *)frame)->header.prof.ccs /* ToDo */);
3738 TICK_ALLOC_UP_THK(words+1,0);
3741 debugBelch("sched: Updating ");
3742 printPtr((P_)((StgUpdateFrame *)frame)->updatee);
3743 debugBelch(" with ");
3744 printObj((StgClosure *)ap);
3747 // Replace the updatee with an indirection
3749 // Warning: if we're in a loop, more than one update frame on
3750 // the stack may point to the same object. Be careful not to
3751 // overwrite an IND_OLDGEN in this case, because we'll screw
3752 // up the mutable lists. To be on the safe side, don't
3753 // overwrite any kind of indirection at all. See also
3754 // threadSqueezeStack in GC.c, where we have to make a similar
3757 if (!closure_IND(((StgUpdateFrame *)frame)->updatee)) {
3758 // revert the black hole
3759 UPD_IND_NOLOCK(((StgUpdateFrame *)frame)->updatee,
3762 sp += sizeofW(StgUpdateFrame) - 1;
3763 sp[0] = (W_)ap; // push onto stack
3765 continue; //no need to bump frame
3769 // We've stripped the entire stack, the thread is now dead.
3770 tso->what_next = ThreadKilled;
3771 tso->sp = frame + sizeofW(StgStopFrame);
3775 // If we find a CATCH_FRAME, and we've got an exception to raise,
3776 // then build the THUNK raise(exception), and leave it on
3777 // top of the CATCH_FRAME ready to enter.
3781 StgCatchFrame *cf = (StgCatchFrame *)frame;
3785 if (exception == NULL) break;
3787 // we've got an exception to raise, so let's pass it to the
3788 // handler in this frame.
3790 raise = (StgThunk *)allocateLocal(cap,sizeofW(StgThunk)+1);
3791 TICK_ALLOC_SE_THK(1,0);
3792 SET_HDR(raise,&stg_raise_info,cf->header.prof.ccs);
3793 raise->payload[0] = exception;
3795 // throw away the stack from Sp up to the CATCH_FRAME.
3799 /* Ensure that async excpetions are blocked now, so we don't get
3800 * a surprise exception before we get around to executing the
3803 if (tso->blocked_exceptions == NULL) {
3804 tso->blocked_exceptions = END_TSO_QUEUE;
3807 /* Put the newly-built THUNK on top of the stack, ready to execute
3808 * when the thread restarts.
3811 sp[-1] = (W_)&stg_enter_info;
3813 tso->what_next = ThreadRunGHC;
3814 IF_DEBUG(sanity, checkTSO(tso));
3818 case ATOMICALLY_FRAME:
3819 if (stop_at_atomically) {
3820 ASSERT(stmGetEnclosingTRec(tso->trec) == NO_TREC);
3821 stmCondemnTransaction(cap, tso -> trec);
3825 // R1 is not a register: the return convention for IO in
3826 // this case puts the return value on the stack, so we
3827 // need to set up the stack to return to the atomically
3828 // frame properly...
3829 tso->sp = frame - 2;
3830 tso->sp[1] = (StgWord) &stg_NO_FINALIZER_closure; // why not?
3831 tso->sp[0] = (StgWord) &stg_ut_1_0_unreg_info;
3833 tso->what_next = ThreadRunGHC;
3836 // Not stop_at_atomically... fall through and abort the
3839 case CATCH_RETRY_FRAME:
3840 // IF we find an ATOMICALLY_FRAME then we abort the
3841 // current transaction and propagate the exception. In
3842 // this case (unlike ordinary exceptions) we do not care
3843 // whether the transaction is valid or not because its
3844 // possible validity cannot have caused the exception
3845 // and will not be visible after the abort.
3847 debugBelch("Found atomically block delivering async exception\n"));
3848 StgTRecHeader *trec = tso -> trec;
3849 StgTRecHeader *outer = stmGetEnclosingTRec(trec);
3850 stmAbortTransaction(cap, trec);
3851 tso -> trec = outer;
3858 // move on to the next stack frame
3859 frame += stack_frame_sizeW((StgClosure *)frame);
3862 // if we got here, then we stopped at stop_here
3863 ASSERT(stop_here != NULL);
3866 /* -----------------------------------------------------------------------------
3869 This is used for interruption (^C) and forking, and corresponds to
3870 raising an exception but without letting the thread catch the
3872 -------------------------------------------------------------------------- */
3875 deleteThread (Capability *cap, StgTSO *tso)
3877 if (tso->why_blocked != BlockedOnCCall &&
3878 tso->why_blocked != BlockedOnCCall_NoUnblockExc) {
3879 raiseAsync(cap,tso,NULL);
3883 #ifdef FORKPROCESS_PRIMOP_SUPPORTED
3885 deleteThreadImmediately(Capability *cap, StgTSO *tso)
3886 { // for forkProcess only:
3887 // delete thread without giving it a chance to catch the KillThread exception
3889 if (tso->what_next == ThreadComplete || tso->what_next == ThreadKilled) {
3893 if (tso->why_blocked != BlockedOnCCall &&
3894 tso->why_blocked != BlockedOnCCall_NoUnblockExc) {
3895 unblockThread(cap,tso);
3898 tso->what_next = ThreadKilled;
3902 /* -----------------------------------------------------------------------------
3903 raiseExceptionHelper
3905 This function is called by the raise# primitve, just so that we can
3906 move some of the tricky bits of raising an exception from C-- into
3907 C. Who knows, it might be a useful re-useable thing here too.
3908 -------------------------------------------------------------------------- */
3911 raiseExceptionHelper (StgRegTable *reg, StgTSO *tso, StgClosure *exception)
3913 Capability *cap = regTableToCapability(reg);
3914 StgThunk *raise_closure = NULL;
3916 StgRetInfoTable *info;
3918 // This closure represents the expression 'raise# E' where E
3919 // is the exception raise. It is used to overwrite all the
3920 // thunks which are currently under evaluataion.
3923 // OLD COMMENT (we don't have MIN_UPD_SIZE now):
3924 // LDV profiling: stg_raise_info has THUNK as its closure
3925 // type. Since a THUNK takes at least MIN_UPD_SIZE words in its
3926 // payload, MIN_UPD_SIZE is more approprate than 1. It seems that
3927 // 1 does not cause any problem unless profiling is performed.
3928 // However, when LDV profiling goes on, we need to linearly scan
3929 // small object pool, where raise_closure is stored, so we should
3930 // use MIN_UPD_SIZE.
3932 // raise_closure = (StgClosure *)RET_STGCALL1(P_,allocate,
3933 // sizeofW(StgClosure)+1);
3937 // Walk up the stack, looking for the catch frame. On the way,
3938 // we update any closures pointed to from update frames with the
3939 // raise closure that we just built.
3943 info = get_ret_itbl((StgClosure *)p);
3944 next = p + stack_frame_sizeW((StgClosure *)p);
3945 switch (info->i.type) {
3948 // Only create raise_closure if we need to.
3949 if (raise_closure == NULL) {
3951 (StgThunk *)allocateLocal(cap,sizeofW(StgThunk)+1);
3952 SET_HDR(raise_closure, &stg_raise_info, CCCS);
3953 raise_closure->payload[0] = exception;
3955 UPD_IND(((StgUpdateFrame *)p)->updatee,(StgClosure *)raise_closure);
3959 case ATOMICALLY_FRAME:
3960 IF_DEBUG(stm, debugBelch("Found ATOMICALLY_FRAME at %p\n", p));
3962 return ATOMICALLY_FRAME;
3968 case CATCH_STM_FRAME:
3969 IF_DEBUG(stm, debugBelch("Found CATCH_STM_FRAME at %p\n", p));
3971 return CATCH_STM_FRAME;
3977 case CATCH_RETRY_FRAME:
3986 /* -----------------------------------------------------------------------------
3987 findRetryFrameHelper
3989 This function is called by the retry# primitive. It traverses the stack
3990 leaving tso->sp referring to the frame which should handle the retry.
3992 This should either be a CATCH_RETRY_FRAME (if the retry# is within an orElse#)
3993 or should be a ATOMICALLY_FRAME (if the retry# reaches the top level).
3995 We skip CATCH_STM_FRAMEs because retries are not considered to be exceptions,
3996 despite the similar implementation.
3998 We should not expect to see CATCH_FRAME or STOP_FRAME because those should
3999 not be created within memory transactions.
4000 -------------------------------------------------------------------------- */
4003 findRetryFrameHelper (StgTSO *tso)
4006 StgRetInfoTable *info;
4010 info = get_ret_itbl((StgClosure *)p);
4011 next = p + stack_frame_sizeW((StgClosure *)p);
4012 switch (info->i.type) {
4014 case ATOMICALLY_FRAME:
4015 IF_DEBUG(stm, debugBelch("Found ATOMICALLY_FRAME at %p during retrry\n", p));
4017 return ATOMICALLY_FRAME;
4019 case CATCH_RETRY_FRAME:
4020 IF_DEBUG(stm, debugBelch("Found CATCH_RETRY_FRAME at %p during retrry\n", p));
4022 return CATCH_RETRY_FRAME;
4024 case CATCH_STM_FRAME:
4026 ASSERT(info->i.type != CATCH_FRAME);
4027 ASSERT(info->i.type != STOP_FRAME);
4034 /* -----------------------------------------------------------------------------
4035 resurrectThreads is called after garbage collection on the list of
4036 threads found to be garbage. Each of these threads will be woken
4037 up and sent a signal: BlockedOnDeadMVar if the thread was blocked
4038 on an MVar, or NonTermination if the thread was blocked on a Black
4041 Locks: assumes we hold *all* the capabilities.
4042 -------------------------------------------------------------------------- */
4045 resurrectThreads (StgTSO *threads)
4050 for (tso = threads; tso != END_TSO_QUEUE; tso = next) {
4051 next = tso->global_link;
4052 tso->global_link = all_threads;
4054 IF_DEBUG(scheduler, sched_belch("resurrecting thread %d", tso->id));
4056 // Wake up the thread on the Capability it was last on for a
4057 // bound thread, or last_free_capability otherwise.
4059 cap = tso->bound->cap;
4061 cap = last_free_capability;
4064 switch (tso->why_blocked) {
4066 case BlockedOnException:
4067 /* Called by GC - sched_mutex lock is currently held. */
4068 raiseAsync(cap, tso,(StgClosure *)BlockedOnDeadMVar_closure);
4070 case BlockedOnBlackHole:
4071 raiseAsync(cap, tso,(StgClosure *)NonTermination_closure);
4074 raiseAsync(cap, tso,(StgClosure *)BlockedIndefinitely_closure);
4077 /* This might happen if the thread was blocked on a black hole
4078 * belonging to a thread that we've just woken up (raiseAsync
4079 * can wake up threads, remember...).
4083 barf("resurrectThreads: thread blocked in a strange way");
4088 /* ----------------------------------------------------------------------------
4089 * Debugging: why is a thread blocked
4090 * [Also provides useful information when debugging threaded programs
4091 * at the Haskell source code level, so enable outside of DEBUG. --sof 7/02]
4092 ------------------------------------------------------------------------- */
4096 printThreadBlockage(StgTSO *tso)
4098 switch (tso->why_blocked) {
4100 debugBelch("is blocked on read from fd %d", (int)(tso->block_info.fd));
4102 case BlockedOnWrite:
4103 debugBelch("is blocked on write to fd %d", (int)(tso->block_info.fd));
4105 #if defined(mingw32_HOST_OS)
4106 case BlockedOnDoProc:
4107 debugBelch("is blocked on proc (request: %ld)", tso->block_info.async_result->reqID);
4110 case BlockedOnDelay:
4111 debugBelch("is blocked until %ld", (long)(tso->block_info.target));
4114 debugBelch("is blocked on an MVar @ %p", tso->block_info.closure);
4116 case BlockedOnException:
4117 debugBelch("is blocked on delivering an exception to thread %d",
4118 tso->block_info.tso->id);
4120 case BlockedOnBlackHole:
4121 debugBelch("is blocked on a black hole");
4124 debugBelch("is not blocked");
4126 #if defined(PARALLEL_HASKELL)
4128 debugBelch("is blocked on global address; local FM_BQ is %p (%s)",
4129 tso->block_info.closure, info_type(tso->block_info.closure));
4131 case BlockedOnGA_NoSend:
4132 debugBelch("is blocked on global address (no send); local FM_BQ is %p (%s)",
4133 tso->block_info.closure, info_type(tso->block_info.closure));
4136 case BlockedOnCCall:
4137 debugBelch("is blocked on an external call");
4139 case BlockedOnCCall_NoUnblockExc:
4140 debugBelch("is blocked on an external call (exceptions were already blocked)");
4143 debugBelch("is blocked on an STM operation");
4146 barf("printThreadBlockage: strange tso->why_blocked: %d for TSO %d (%d)",
4147 tso->why_blocked, tso->id, tso);
4152 printThreadStatus(StgTSO *t)
4154 debugBelch("\tthread %4d @ %p ", t->id, (void *)t);
4156 void *label = lookupThreadLabel(t->id);
4157 if (label) debugBelch("[\"%s\"] ",(char *)label);
4159 if (t->what_next == ThreadRelocated) {
4160 debugBelch("has been relocated...\n");
4162 switch (t->what_next) {
4164 debugBelch("has been killed");
4166 case ThreadComplete:
4167 debugBelch("has completed");
4170 printThreadBlockage(t);
4177 printAllThreads(void)
4184 char time_string[TIME_STR_LEN], node_str[NODE_STR_LEN];
4185 ullong_format_string(TIME_ON_PROC(CurrentProc),
4186 time_string, rtsFalse/*no commas!*/);
4188 debugBelch("all threads at [%s]:\n", time_string);
4189 # elif defined(PARALLEL_HASKELL)
4190 char time_string[TIME_STR_LEN], node_str[NODE_STR_LEN];
4191 ullong_format_string(CURRENT_TIME,
4192 time_string, rtsFalse/*no commas!*/);
4194 debugBelch("all threads at [%s]:\n", time_string);
4196 debugBelch("all threads:\n");
4199 for (i = 0; i < n_capabilities; i++) {
4200 cap = &capabilities[i];
4201 debugBelch("threads on capability %d:\n", cap->no);
4202 for (t = cap->run_queue_hd; t != END_TSO_QUEUE; t = t->link) {
4203 printThreadStatus(t);
4207 debugBelch("other threads:\n");
4208 for (t = all_threads; t != END_TSO_QUEUE; t = next) {
4209 if (t->why_blocked != NotBlocked) {
4210 printThreadStatus(t);
4212 if (t->what_next == ThreadRelocated) {
4215 next = t->global_link;
4222 printThreadQueue(StgTSO *t)
4225 for (; t != END_TSO_QUEUE; t = t->link) {
4226 printThreadStatus(t);
4229 debugBelch("%d threads on queue\n", i);
4233 Print a whole blocking queue attached to node (debugging only).
4235 # if defined(PARALLEL_HASKELL)
4237 print_bq (StgClosure *node)
4239 StgBlockingQueueElement *bqe;
4243 debugBelch("## BQ of closure %p (%s): ",
4244 node, info_type(node));
4246 /* should cover all closures that may have a blocking queue */
4247 ASSERT(get_itbl(node)->type == BLACKHOLE_BQ ||
4248 get_itbl(node)->type == FETCH_ME_BQ ||
4249 get_itbl(node)->type == RBH ||
4250 get_itbl(node)->type == MVAR);
4252 ASSERT(node!=(StgClosure*)NULL); // sanity check
4254 print_bqe(((StgBlockingQueue*)node)->blocking_queue);
4258 Print a whole blocking queue starting with the element bqe.
4261 print_bqe (StgBlockingQueueElement *bqe)
4266 NB: In a parallel setup a BQ of an RBH must end with an RBH_Save closure;
4268 for (end = (bqe==END_BQ_QUEUE);
4269 !end; // iterate until bqe points to a CONSTR
4270 end = (get_itbl(bqe)->type == CONSTR) || (bqe->link==END_BQ_QUEUE),
4271 bqe = end ? END_BQ_QUEUE : bqe->link) {
4272 ASSERT(bqe != END_BQ_QUEUE); // sanity check
4273 ASSERT(bqe != (StgBlockingQueueElement *)NULL); // sanity check
4274 /* types of closures that may appear in a blocking queue */
4275 ASSERT(get_itbl(bqe)->type == TSO ||
4276 get_itbl(bqe)->type == BLOCKED_FETCH ||
4277 get_itbl(bqe)->type == CONSTR);
4278 /* only BQs of an RBH end with an RBH_Save closure */
4279 //ASSERT(get_itbl(bqe)->type != CONSTR || get_itbl(node)->type == RBH);
4281 switch (get_itbl(bqe)->type) {
4283 debugBelch(" TSO %u (%x),",
4284 ((StgTSO *)bqe)->id, ((StgTSO *)bqe));
4287 debugBelch(" BF (node=%p, ga=((%x, %d, %x)),",
4288 ((StgBlockedFetch *)bqe)->node,
4289 ((StgBlockedFetch *)bqe)->ga.payload.gc.gtid,
4290 ((StgBlockedFetch *)bqe)->ga.payload.gc.slot,
4291 ((StgBlockedFetch *)bqe)->ga.weight);
4294 debugBelch(" %s (IP %p),",
4295 (get_itbl(bqe) == &stg_RBH_Save_0_info ? "RBH_Save_0" :
4296 get_itbl(bqe) == &stg_RBH_Save_1_info ? "RBH_Save_1" :
4297 get_itbl(bqe) == &stg_RBH_Save_2_info ? "RBH_Save_2" :
4298 "RBH_Save_?"), get_itbl(bqe));
4301 barf("Unexpected closure type %s in blocking queue", // of %p (%s)",
4302 info_type((StgClosure *)bqe)); // , node, info_type(node));
4308 # elif defined(GRAN)
4310 print_bq (StgClosure *node)
4312 StgBlockingQueueElement *bqe;
4313 PEs node_loc, tso_loc;
4316 /* should cover all closures that may have a blocking queue */
4317 ASSERT(get_itbl(node)->type == BLACKHOLE_BQ ||
4318 get_itbl(node)->type == FETCH_ME_BQ ||
4319 get_itbl(node)->type == RBH);
4321 ASSERT(node!=(StgClosure*)NULL); // sanity check
4322 node_loc = where_is(node);
4324 debugBelch("## BQ of closure %p (%s) on [PE %d]: ",
4325 node, info_type(node), node_loc);
4328 NB: In a parallel setup a BQ of an RBH must end with an RBH_Save closure;
4330 for (bqe = ((StgBlockingQueue*)node)->blocking_queue, end = (bqe==END_BQ_QUEUE);
4331 !end; // iterate until bqe points to a CONSTR
4332 end = (get_itbl(bqe)->type == CONSTR) || (bqe->link==END_BQ_QUEUE), bqe = end ? END_BQ_QUEUE : bqe->link) {
4333 ASSERT(bqe != END_BQ_QUEUE); // sanity check
4334 ASSERT(bqe != (StgBlockingQueueElement *)NULL); // sanity check
4335 /* types of closures that may appear in a blocking queue */
4336 ASSERT(get_itbl(bqe)->type == TSO ||
4337 get_itbl(bqe)->type == CONSTR);
4338 /* only BQs of an RBH end with an RBH_Save closure */
4339 ASSERT(get_itbl(bqe)->type != CONSTR || get_itbl(node)->type == RBH);
4341 tso_loc = where_is((StgClosure *)bqe);
4342 switch (get_itbl(bqe)->type) {
4344 debugBelch(" TSO %d (%p) on [PE %d],",
4345 ((StgTSO *)bqe)->id, (StgTSO *)bqe, tso_loc);
4348 debugBelch(" %s (IP %p),",
4349 (get_itbl(bqe) == &stg_RBH_Save_0_info ? "RBH_Save_0" :
4350 get_itbl(bqe) == &stg_RBH_Save_1_info ? "RBH_Save_1" :
4351 get_itbl(bqe) == &stg_RBH_Save_2_info ? "RBH_Save_2" :
4352 "RBH_Save_?"), get_itbl(bqe));
4355 barf("Unexpected closure type %s in blocking queue of %p (%s)",
4356 info_type((StgClosure *)bqe), node, info_type(node));
4364 #if defined(PARALLEL_HASKELL)
4371 for (i=0, tso=run_queue_hd;
4372 tso != END_TSO_QUEUE;
4373 i++, tso=tso->link) {
4382 sched_belch(char *s, ...)
4387 debugBelch("sched (task %p): ", (void *)(unsigned long)(unsigned int)osThreadId());
4388 #elif defined(PARALLEL_HASKELL)
4391 debugBelch("sched: ");