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"
51 #ifdef HAVE_SYS_TYPES_H
52 #include <sys/types.h>
66 // Turn off inlining when debugging - it obfuscates things
69 # define STATIC_INLINE static
73 #define USED_WHEN_THREADED_RTS
74 #define USED_WHEN_NON_THREADED_RTS STG_UNUSED
76 #define USED_WHEN_THREADED_RTS STG_UNUSED
77 #define USED_WHEN_NON_THREADED_RTS
83 #define USED_WHEN_SMP STG_UNUSED
86 /* -----------------------------------------------------------------------------
88 * -------------------------------------------------------------------------- */
92 StgTSO* ActiveTSO = NULL; /* for assigning system costs; GranSim-Light only */
93 /* rtsTime TimeOfNextEvent, EndOfTimeSlice; now in GranSim.c */
96 In GranSim we have a runnable and a blocked queue for each processor.
97 In order to minimise code changes new arrays run_queue_hds/tls
98 are created. run_queue_hd is then a short cut (macro) for
99 run_queue_hds[CurrentProc] (see GranSim.h).
102 StgTSO *run_queue_hds[MAX_PROC], *run_queue_tls[MAX_PROC];
103 StgTSO *blocked_queue_hds[MAX_PROC], *blocked_queue_tls[MAX_PROC];
104 StgTSO *ccalling_threadss[MAX_PROC];
105 /* We use the same global list of threads (all_threads) in GranSim as in
106 the std RTS (i.e. we are cheating). However, we don't use this list in
107 the GranSim specific code at the moment (so we are only potentially
112 #if !defined(THREADED_RTS)
113 // Blocked/sleeping thrads
114 StgTSO *blocked_queue_hd = NULL;
115 StgTSO *blocked_queue_tl = NULL;
116 StgTSO *sleeping_queue = NULL; // perhaps replace with a hash table?
119 /* Threads blocked on blackholes.
120 * LOCK: sched_mutex+capability, or all capabilities
122 StgTSO *blackhole_queue = NULL;
125 /* The blackhole_queue should be checked for threads to wake up. See
126 * Schedule.h for more thorough comment.
127 * LOCK: none (doesn't matter if we miss an update)
129 rtsBool blackholes_need_checking = rtsFalse;
131 /* Linked list of all threads.
132 * Used for detecting garbage collected threads.
133 * LOCK: sched_mutex+capability, or all capabilities
135 StgTSO *all_threads = NULL;
137 /* flag set by signal handler to precipitate a context switch
138 * LOCK: none (just an advisory flag)
140 int context_switch = 0;
142 /* flag that tracks whether we have done any execution in this time slice.
143 * LOCK: currently none, perhaps we should lock (but needs to be
144 * updated in the fast path of the scheduler).
146 nat recent_activity = ACTIVITY_YES;
148 /* if this flag is set as well, give up execution
149 * LOCK: none (changes once, from false->true)
151 rtsBool interrupted = rtsFalse;
153 /* Next thread ID to allocate.
156 static StgThreadID next_thread_id = 1;
158 /* The smallest stack size that makes any sense is:
159 * RESERVED_STACK_WORDS (so we can get back from the stack overflow)
160 * + sizeofW(StgStopFrame) (the stg_stop_thread_info frame)
161 * + 1 (the closure to enter)
163 * + 1 (spare slot req'd by stg_ap_v_ret)
165 * A thread with this stack will bomb immediately with a stack
166 * overflow, which will increase its stack size.
168 #define MIN_STACK_WORDS (RESERVED_STACK_WORDS + sizeofW(StgStopFrame) + 3)
174 /* This is used in `TSO.h' and gcc 2.96 insists that this variable actually
175 * exists - earlier gccs apparently didn't.
181 * Set to TRUE when entering a shutdown state (via shutdownHaskellAndExit()) --
182 * in an MT setting, needed to signal that a worker thread shouldn't hang around
183 * in the scheduler when it is out of work.
185 rtsBool shutting_down_scheduler = rtsFalse;
188 * This mutex protects most of the global scheduler data in
189 * the THREADED_RTS and (inc. SMP) runtime.
191 #if defined(THREADED_RTS)
192 Mutex sched_mutex = INIT_MUTEX_VAR;
195 #if defined(PARALLEL_HASKELL)
197 rtsTime TimeOfLastYield;
198 rtsBool emitSchedule = rtsTrue;
201 /* -----------------------------------------------------------------------------
202 * static function prototypes
203 * -------------------------------------------------------------------------- */
205 static Capability *schedule (Capability *initialCapability, Task *task);
208 // These function all encapsulate parts of the scheduler loop, and are
209 // abstracted only to make the structure and control flow of the
210 // scheduler clearer.
212 static void schedulePreLoop (void);
214 static void schedulePushWork(Capability *cap, Task *task);
216 static void scheduleStartSignalHandlers (void);
217 static void scheduleCheckBlockedThreads (Capability *cap);
218 static void scheduleCheckBlackHoles (Capability *cap);
219 static void scheduleDetectDeadlock (Capability *cap, Task *task);
221 static StgTSO *scheduleProcessEvent(rtsEvent *event);
223 #if defined(PARALLEL_HASKELL)
224 static StgTSO *scheduleSendPendingMessages(void);
225 static void scheduleActivateSpark(void);
226 static rtsBool scheduleGetRemoteWork(rtsBool *receivedFinish);
228 #if defined(PAR) || defined(GRAN)
229 static void scheduleGranParReport(void);
231 static void schedulePostRunThread(void);
232 static rtsBool scheduleHandleHeapOverflow( Capability *cap, StgTSO *t );
233 static void scheduleHandleStackOverflow( Capability *cap, Task *task,
235 static rtsBool scheduleHandleYield( Capability *cap, StgTSO *t,
236 nat prev_what_next );
237 static void scheduleHandleThreadBlocked( StgTSO *t );
238 static rtsBool scheduleHandleThreadFinished( Capability *cap, Task *task,
240 static rtsBool scheduleDoHeapProfile(rtsBool ready_to_gc);
241 static void scheduleDoGC(Capability *cap, Task *task, rtsBool force_major);
243 static void unblockThread(Capability *cap, StgTSO *tso);
244 static rtsBool checkBlackHoles(Capability *cap);
245 static void AllRoots(evac_fn evac);
247 static StgTSO *threadStackOverflow(Capability *cap, StgTSO *tso);
249 static void raiseAsync_(Capability *cap, StgTSO *tso, StgClosure *exception,
250 rtsBool stop_at_atomically);
252 static void deleteThread (Capability *cap, StgTSO *tso);
253 static void deleteRunQueue (Capability *cap);
256 static void printThreadBlockage(StgTSO *tso);
257 static void printThreadStatus(StgTSO *tso);
258 void printThreadQueue(StgTSO *tso);
261 #if defined(PARALLEL_HASKELL)
262 StgTSO * createSparkThread(rtsSpark spark);
263 StgTSO * activateSpark (rtsSpark spark);
267 static char *whatNext_strs[] = {
277 /* -----------------------------------------------------------------------------
278 * Putting a thread on the run queue: different scheduling policies
279 * -------------------------------------------------------------------------- */
282 addToRunQueue( Capability *cap, StgTSO *t )
284 #if defined(PARALLEL_HASKELL)
285 if (RtsFlags.ParFlags.doFairScheduling) {
286 // this does round-robin scheduling; good for concurrency
287 appendToRunQueue(cap,t);
289 // this does unfair scheduling; good for parallelism
290 pushOnRunQueue(cap,t);
293 // this does round-robin scheduling; good for concurrency
294 appendToRunQueue(cap,t);
298 /* ---------------------------------------------------------------------------
299 Main scheduling loop.
301 We use round-robin scheduling, each thread returning to the
302 scheduler loop when one of these conditions is detected:
305 * timer expires (thread yields)
311 In a GranSim setup this loop iterates over the global event queue.
312 This revolves around the global event queue, which determines what
313 to do next. Therefore, it's more complicated than either the
314 concurrent or the parallel (GUM) setup.
317 GUM iterates over incoming messages.
318 It starts with nothing to do (thus CurrentTSO == END_TSO_QUEUE),
319 and sends out a fish whenever it has nothing to do; in-between
320 doing the actual reductions (shared code below) it processes the
321 incoming messages and deals with delayed operations
322 (see PendingFetches).
323 This is not the ugliest code you could imagine, but it's bloody close.
325 ------------------------------------------------------------------------ */
328 schedule (Capability *initialCapability, Task *task)
332 StgThreadReturnCode ret;
335 #elif defined(PARALLEL_HASKELL)
338 rtsBool receivedFinish = rtsFalse;
340 nat tp_size, sp_size; // stats only
345 #if defined(THREADED_RTS)
346 rtsBool first = rtsTrue;
349 cap = initialCapability;
351 // Pre-condition: this task owns initialCapability.
352 // The sched_mutex is *NOT* held
353 // NB. on return, we still hold a capability.
356 sched_belch("### NEW SCHEDULER LOOP (task: %p, cap: %p)",
357 task, initialCapability);
362 // -----------------------------------------------------------
363 // Scheduler loop starts here:
365 #if defined(PARALLEL_HASKELL)
366 #define TERMINATION_CONDITION (!receivedFinish)
368 #define TERMINATION_CONDITION ((event = get_next_event()) != (rtsEvent*)NULL)
370 #define TERMINATION_CONDITION rtsTrue
373 while (TERMINATION_CONDITION) {
376 /* Choose the processor with the next event */
377 CurrentProc = event->proc;
378 CurrentTSO = event->tso;
381 #if defined(THREADED_RTS)
383 // don't yield the first time, we want a chance to run this
384 // thread for a bit, even if there are others banging at the
387 ASSERT_CAPABILITY_INVARIANTS(cap,task);
389 // Yield the capability to higher-priority tasks if necessary.
390 yieldCapability(&cap, task);
395 schedulePushWork(cap,task);
398 // Check whether we have re-entered the RTS from Haskell without
399 // going via suspendThread()/resumeThread (i.e. a 'safe' foreign
401 if (cap->in_haskell) {
402 errorBelch("schedule: re-entered unsafely.\n"
403 " Perhaps a 'foreign import unsafe' should be 'safe'?");
404 stg_exit(EXIT_FAILURE);
408 // Test for interruption. If interrupted==rtsTrue, then either
409 // we received a keyboard interrupt (^C), or the scheduler is
410 // trying to shut down all the tasks (shutting_down_scheduler) in
415 if (shutting_down_scheduler) {
416 IF_DEBUG(scheduler, sched_belch("shutting down"));
417 // If we are a worker, just exit. If we're a bound thread
418 // then we will exit below when we've removed our TSO from
420 if (task->tso == NULL && emptyRunQueue(cap)) {
424 IF_DEBUG(scheduler, sched_belch("interrupted"));
428 #if defined(not_yet) && defined(SMP)
430 // Top up the run queue from our spark pool. We try to make the
431 // number of threads in the run queue equal to the number of
432 // free capabilities.
436 if (emptyRunQueue()) {
437 spark = findSpark(rtsFalse);
439 break; /* no more sparks in the pool */
441 createSparkThread(spark);
443 sched_belch("==^^ turning spark of closure %p into a thread",
444 (StgClosure *)spark));
450 scheduleStartSignalHandlers();
452 // Only check the black holes here if we've nothing else to do.
453 // During normal execution, the black hole list only gets checked
454 // at GC time, to avoid repeatedly traversing this possibly long
455 // list each time around the scheduler.
456 if (emptyRunQueue(cap)) { scheduleCheckBlackHoles(cap); }
458 scheduleCheckBlockedThreads(cap);
460 scheduleDetectDeadlock(cap,task);
462 // Normally, the only way we can get here with no threads to
463 // run is if a keyboard interrupt received during
464 // scheduleCheckBlockedThreads() or scheduleDetectDeadlock().
465 // Additionally, it is not fatal for the
466 // threaded RTS to reach here with no threads to run.
468 // win32: might be here due to awaitEvent() being abandoned
469 // as a result of a console event having been delivered.
470 if ( emptyRunQueue(cap) ) {
471 #if !defined(THREADED_RTS) && !defined(mingw32_HOST_OS)
474 continue; // nothing to do
477 #if defined(PARALLEL_HASKELL)
478 scheduleSendPendingMessages();
479 if (emptyRunQueue(cap) && scheduleActivateSpark())
483 ASSERT(next_fish_to_send_at==0); // i.e. no delayed fishes left!
486 /* If we still have no work we need to send a FISH to get a spark
488 if (emptyRunQueue(cap)) {
489 if (!scheduleGetRemoteWork(&receivedFinish)) continue;
490 ASSERT(rtsFalse); // should not happen at the moment
492 // from here: non-empty run queue.
493 // TODO: merge above case with this, only one call processMessages() !
494 if (PacketsWaiting()) { /* process incoming messages, if
495 any pending... only in else
496 because getRemoteWork waits for
498 receivedFinish = processMessages();
503 scheduleProcessEvent(event);
507 // Get a thread to run
509 t = popRunQueue(cap);
511 #if defined(GRAN) || defined(PAR)
512 scheduleGranParReport(); // some kind of debuging output
514 // Sanity check the thread we're about to run. This can be
515 // expensive if there is lots of thread switching going on...
516 IF_DEBUG(sanity,checkTSO(t));
519 #if defined(THREADED_RTS)
520 // Check whether we can run this thread in the current task.
521 // If not, we have to pass our capability to the right task.
523 Task *bound = t->bound;
528 sched_belch("### Running thread %d in bound thread",
530 // yes, the Haskell thread is bound to the current native thread
533 sched_belch("### thread %d bound to another OS thread",
535 // no, bound to a different Haskell thread: pass to that thread
536 pushOnRunQueue(cap,t);
540 // The thread we want to run is unbound.
543 sched_belch("### this OS thread cannot run thread %d", t->id));
544 // no, the current native thread is bound to a different
545 // Haskell thread, so pass it to any worker thread
546 pushOnRunQueue(cap,t);
553 cap->r.rCurrentTSO = t;
555 /* context switches are initiated by the timer signal, unless
556 * the user specified "context switch as often as possible", with
559 if (RtsFlags.ConcFlags.ctxtSwitchTicks == 0
560 && !emptyThreadQueues(cap)) {
566 IF_DEBUG(scheduler, sched_belch("-->> running thread %ld %s ...",
567 (long)t->id, whatNext_strs[t->what_next]));
569 #if defined(PROFILING)
570 startHeapProfTimer();
573 // ----------------------------------------------------------------------
574 // Run the current thread
576 prev_what_next = t->what_next;
578 errno = t->saved_errno;
579 cap->in_haskell = rtsTrue;
581 recent_activity = ACTIVITY_YES;
583 switch (prev_what_next) {
587 /* Thread already finished, return to scheduler. */
588 ret = ThreadFinished;
594 r = StgRun((StgFunPtr) stg_returnToStackTop, &cap->r);
595 cap = regTableToCapability(r);
600 case ThreadInterpret:
601 cap = interpretBCO(cap);
606 barf("schedule: invalid what_next field");
609 cap->in_haskell = rtsFalse;
611 // The TSO might have moved, eg. if it re-entered the RTS and a GC
612 // happened. So find the new location:
613 t = cap->r.rCurrentTSO;
616 // If ret is ThreadBlocked, and this Task is bound to the TSO that
617 // blocked, we are in limbo - the TSO is now owned by whatever it
618 // is blocked on, and may in fact already have been woken up,
619 // perhaps even on a different Capability. It may be the case
620 // that task->cap != cap. We better yield this Capability
621 // immediately and return to normaility.
622 if (ret == ThreadBlocked) {
624 debugBelch("--<< thread %d (%s) stopped: blocked\n",
625 t->id, whatNext_strs[t->what_next]));
630 ASSERT_CAPABILITY_INVARIANTS(cap,task);
632 // And save the current errno in this thread.
633 t->saved_errno = errno;
635 // ----------------------------------------------------------------------
637 // Costs for the scheduler are assigned to CCS_SYSTEM
638 #if defined(PROFILING)
643 // We have run some Haskell code: there might be blackhole-blocked
644 // threads to wake up now.
645 // Lock-free test here should be ok, we're just setting a flag.
646 if ( blackhole_queue != END_TSO_QUEUE ) {
647 blackholes_need_checking = rtsTrue;
650 #if defined(THREADED_RTS)
651 IF_DEBUG(scheduler,debugBelch("sched (task %p): ", (void *)(unsigned long)(unsigned int)osThreadId()););
652 #elif !defined(GRAN) && !defined(PARALLEL_HASKELL)
653 IF_DEBUG(scheduler,debugBelch("sched: "););
656 schedulePostRunThread();
658 ready_to_gc = rtsFalse;
662 ready_to_gc = scheduleHandleHeapOverflow(cap,t);
666 scheduleHandleStackOverflow(cap,task,t);
670 if (scheduleHandleYield(cap, t, prev_what_next)) {
671 // shortcut for switching between compiler/interpreter:
677 scheduleHandleThreadBlocked(t);
681 if (scheduleHandleThreadFinished(cap, task, t)) return cap;
682 ASSERT_CAPABILITY_INVARIANTS(cap,task);
686 barf("schedule: invalid thread return code %d", (int)ret);
689 if (scheduleDoHeapProfile(ready_to_gc)) { ready_to_gc = rtsFalse; }
690 if (ready_to_gc) { scheduleDoGC(cap,task,rtsFalse); }
691 } /* end of while() */
693 IF_PAR_DEBUG(verbose,
694 debugBelch("== Leaving schedule() after having received Finish\n"));
697 /* ----------------------------------------------------------------------------
698 * Setting up the scheduler loop
699 * ------------------------------------------------------------------------- */
702 schedulePreLoop(void)
705 /* set up first event to get things going */
706 /* ToDo: assign costs for system setup and init MainTSO ! */
707 new_event(CurrentProc, CurrentProc, CurrentTime[CurrentProc],
709 CurrentTSO, (StgClosure*)NULL, (rtsSpark*)NULL);
712 debugBelch("GRAN: Init CurrentTSO (in schedule) = %p\n",
714 G_TSO(CurrentTSO, 5));
716 if (RtsFlags.GranFlags.Light) {
717 /* Save current time; GranSim Light only */
718 CurrentTSO->gran.clock = CurrentTime[CurrentProc];
723 /* -----------------------------------------------------------------------------
726 * Push work to other Capabilities if we have some.
727 * -------------------------------------------------------------------------- */
731 schedulePushWork(Capability *cap USED_WHEN_SMP,
732 Task *task USED_WHEN_SMP)
734 Capability *free_caps[n_capabilities], *cap0;
737 // Check whether we have more threads on our run queue that we
738 // could hand to another Capability.
739 if (emptyRunQueue(cap) || cap->run_queue_hd->link == END_TSO_QUEUE) {
743 // First grab as many free Capabilities as we can.
744 for (i=0, n_free_caps=0; i < n_capabilities; i++) {
745 cap0 = &capabilities[i];
746 if (cap != cap0 && tryGrabCapability(cap0,task)) {
747 if (!emptyRunQueue(cap0) || cap->returning_tasks_hd != NULL) {
748 // it already has some work, we just grabbed it at
749 // the wrong moment. Or maybe it's deadlocked!
750 releaseCapability(cap0);
752 free_caps[n_free_caps++] = cap0;
757 // we now have n_free_caps free capabilities stashed in
758 // free_caps[]. Share our run queue equally with them. This is
759 // probably the simplest thing we could do; improvements we might
760 // want to do include:
762 // - giving high priority to moving relatively new threads, on
763 // the gournds that they haven't had time to build up a
764 // working set in the cache on this CPU/Capability.
766 // - giving low priority to moving long-lived threads
768 if (n_free_caps > 0) {
769 StgTSO *prev, *t, *next;
770 IF_DEBUG(scheduler, sched_belch("excess threads on run queue and %d free capabilities, sharing...", n_free_caps));
772 prev = cap->run_queue_hd;
774 prev->link = END_TSO_QUEUE;
776 for (; t != END_TSO_QUEUE; t = next) {
778 t->link = END_TSO_QUEUE;
779 if (t->what_next == ThreadRelocated) {
782 } else if (i == n_free_caps) {
788 appendToRunQueue(free_caps[i],t);
789 if (t->bound) { t->bound->cap = free_caps[i]; }
793 cap->run_queue_tl = prev;
795 // release the capabilities
796 for (i = 0; i < n_free_caps; i++) {
797 task->cap = free_caps[i];
798 releaseCapability(free_caps[i]);
801 task->cap = cap; // reset to point to our Capability.
805 /* ----------------------------------------------------------------------------
806 * Start any pending signal handlers
807 * ------------------------------------------------------------------------- */
810 scheduleStartSignalHandlers(void)
812 #if defined(RTS_USER_SIGNALS) && !defined(THREADED_RTS)
813 if (signals_pending()) { // safe outside the lock
814 startSignalHandlers();
819 /* ----------------------------------------------------------------------------
820 * Check for blocked threads that can be woken up.
821 * ------------------------------------------------------------------------- */
824 scheduleCheckBlockedThreads(Capability *cap USED_WHEN_NON_THREADED_RTS)
826 #if !defined(THREADED_RTS)
828 // Check whether any waiting threads need to be woken up. If the
829 // run queue is empty, and there are no other tasks running, we
830 // can wait indefinitely for something to happen.
832 if ( !emptyQueue(blocked_queue_hd) || !emptyQueue(sleeping_queue) )
834 awaitEvent( emptyRunQueue(cap) && !blackholes_need_checking );
840 /* ----------------------------------------------------------------------------
841 * Check for threads blocked on BLACKHOLEs that can be woken up
842 * ------------------------------------------------------------------------- */
844 scheduleCheckBlackHoles (Capability *cap)
846 if ( blackholes_need_checking ) // check without the lock first
848 ACQUIRE_LOCK(&sched_mutex);
849 if ( blackholes_need_checking ) {
850 checkBlackHoles(cap);
851 blackholes_need_checking = rtsFalse;
853 RELEASE_LOCK(&sched_mutex);
857 /* ----------------------------------------------------------------------------
858 * Detect deadlock conditions and attempt to resolve them.
859 * ------------------------------------------------------------------------- */
862 scheduleDetectDeadlock (Capability *cap, Task *task)
865 #if defined(PARALLEL_HASKELL)
866 // ToDo: add deadlock detection in GUM (similar to SMP) -- HWL
871 * Detect deadlock: when we have no threads to run, there are no
872 * threads blocked, waiting for I/O, or sleeping, and all the
873 * other tasks are waiting for work, we must have a deadlock of
876 if ( emptyThreadQueues(cap) )
878 #if defined(THREADED_RTS)
880 * In the threaded RTS, we only check for deadlock if there
881 * has been no activity in a complete timeslice. This means
882 * we won't eagerly start a full GC just because we don't have
883 * any threads to run currently.
885 if (recent_activity != ACTIVITY_INACTIVE) return;
888 IF_DEBUG(scheduler, sched_belch("deadlocked, forcing major GC..."));
890 // Garbage collection can release some new threads due to
891 // either (a) finalizers or (b) threads resurrected because
892 // they are unreachable and will therefore be sent an
893 // exception. Any threads thus released will be immediately
895 scheduleDoGC( cap, task, rtsTrue/*force major GC*/ );
896 recent_activity = ACTIVITY_DONE_GC;
898 if ( !emptyRunQueue(cap) ) return;
900 #if defined(RTS_USER_SIGNALS) && !defined(THREADED_RTS)
901 /* If we have user-installed signal handlers, then wait
902 * for signals to arrive rather then bombing out with a
905 if ( anyUserHandlers() ) {
907 sched_belch("still deadlocked, waiting for signals..."));
911 if (signals_pending()) {
912 startSignalHandlers();
915 // either we have threads to run, or we were interrupted:
916 ASSERT(!emptyRunQueue(cap) || interrupted);
920 #if !defined(THREADED_RTS)
921 /* Probably a real deadlock. Send the current main thread the
922 * Deadlock exception.
925 switch (task->tso->why_blocked) {
927 case BlockedOnBlackHole:
928 case BlockedOnException:
930 raiseAsync(cap, task->tso, (StgClosure *)NonTermination_closure);
933 barf("deadlock: main thread blocked in a strange way");
941 /* ----------------------------------------------------------------------------
942 * Process an event (GRAN only)
943 * ------------------------------------------------------------------------- */
947 scheduleProcessEvent(rtsEvent *event)
951 if (RtsFlags.GranFlags.Light)
952 GranSimLight_enter_system(event, &ActiveTSO); // adjust ActiveTSO etc
954 /* adjust time based on time-stamp */
955 if (event->time > CurrentTime[CurrentProc] &&
956 event->evttype != ContinueThread)
957 CurrentTime[CurrentProc] = event->time;
959 /* Deal with the idle PEs (may issue FindWork or MoveSpark events) */
960 if (!RtsFlags.GranFlags.Light)
963 IF_DEBUG(gran, debugBelch("GRAN: switch by event-type\n"));
965 /* main event dispatcher in GranSim */
966 switch (event->evttype) {
967 /* Should just be continuing execution */
969 IF_DEBUG(gran, debugBelch("GRAN: doing ContinueThread\n"));
970 /* ToDo: check assertion
971 ASSERT(run_queue_hd != (StgTSO*)NULL &&
972 run_queue_hd != END_TSO_QUEUE);
974 /* Ignore ContinueThreads for fetching threads (if synchr comm) */
975 if (!RtsFlags.GranFlags.DoAsyncFetch &&
976 procStatus[CurrentProc]==Fetching) {
977 debugBelch("ghuH: Spurious ContinueThread while Fetching ignored; TSO %d (%p) [PE %d]\n",
978 CurrentTSO->id, CurrentTSO, CurrentProc);
981 /* Ignore ContinueThreads for completed threads */
982 if (CurrentTSO->what_next == ThreadComplete) {
983 debugBelch("ghuH: found a ContinueThread event for completed thread %d (%p) [PE %d] (ignoring ContinueThread)\n",
984 CurrentTSO->id, CurrentTSO, CurrentProc);
987 /* Ignore ContinueThreads for threads that are being migrated */
988 if (PROCS(CurrentTSO)==Nowhere) {
989 debugBelch("ghuH: trying to run the migrating TSO %d (%p) [PE %d] (ignoring ContinueThread)\n",
990 CurrentTSO->id, CurrentTSO, CurrentProc);
993 /* The thread should be at the beginning of the run queue */
994 if (CurrentTSO!=run_queue_hds[CurrentProc]) {
995 debugBelch("ghuH: TSO %d (%p) [PE %d] is not at the start of the run_queue when doing a ContinueThread\n",
996 CurrentTSO->id, CurrentTSO, CurrentProc);
997 break; // run the thread anyway
1000 new_event(proc, proc, CurrentTime[proc],
1002 (StgTSO*)NULL, (StgClosure*)NULL, (rtsSpark*)NULL);
1004 */ /* Catches superfluous CONTINUEs -- should be unnecessary */
1005 break; // now actually run the thread; DaH Qu'vam yImuHbej
1008 do_the_fetchnode(event);
1009 goto next_thread; /* handle next event in event queue */
1012 do_the_globalblock(event);
1013 goto next_thread; /* handle next event in event queue */
1016 do_the_fetchreply(event);
1017 goto next_thread; /* handle next event in event queue */
1019 case UnblockThread: /* Move from the blocked queue to the tail of */
1020 do_the_unblock(event);
1021 goto next_thread; /* handle next event in event queue */
1023 case ResumeThread: /* Move from the blocked queue to the tail of */
1024 /* the runnable queue ( i.e. Qu' SImqa'lu') */
1025 event->tso->gran.blocktime +=
1026 CurrentTime[CurrentProc] - event->tso->gran.blockedat;
1027 do_the_startthread(event);
1028 goto next_thread; /* handle next event in event queue */
1031 do_the_startthread(event);
1032 goto next_thread; /* handle next event in event queue */
1035 do_the_movethread(event);
1036 goto next_thread; /* handle next event in event queue */
1039 do_the_movespark(event);
1040 goto next_thread; /* handle next event in event queue */
1043 do_the_findwork(event);
1044 goto next_thread; /* handle next event in event queue */
1047 barf("Illegal event type %u\n", event->evttype);
1050 /* This point was scheduler_loop in the old RTS */
1052 IF_DEBUG(gran, debugBelch("GRAN: after main switch\n"));
1054 TimeOfLastEvent = CurrentTime[CurrentProc];
1055 TimeOfNextEvent = get_time_of_next_event();
1056 IgnoreEvents=(TimeOfNextEvent==0); // HWL HACK
1057 // CurrentTSO = ThreadQueueHd;
1059 IF_DEBUG(gran, debugBelch("GRAN: time of next event is: %ld\n",
1062 if (RtsFlags.GranFlags.Light)
1063 GranSimLight_leave_system(event, &ActiveTSO);
1065 EndOfTimeSlice = CurrentTime[CurrentProc]+RtsFlags.GranFlags.time_slice;
1068 debugBelch("GRAN: end of time-slice is %#lx\n", EndOfTimeSlice));
1070 /* in a GranSim setup the TSO stays on the run queue */
1072 /* Take a thread from the run queue. */
1073 POP_RUN_QUEUE(t); // take_off_run_queue(t);
1076 debugBelch("GRAN: About to run current thread, which is\n");
1079 context_switch = 0; // turned on via GranYield, checking events and time slice
1082 DumpGranEvent(GR_SCHEDULE, t));
1084 procStatus[CurrentProc] = Busy;
1088 /* ----------------------------------------------------------------------------
1089 * Send pending messages (PARALLEL_HASKELL only)
1090 * ------------------------------------------------------------------------- */
1092 #if defined(PARALLEL_HASKELL)
1094 scheduleSendPendingMessages(void)
1100 # if defined(PAR) // global Mem.Mgmt., omit for now
1101 if (PendingFetches != END_BF_QUEUE) {
1106 if (RtsFlags.ParFlags.BufferTime) {
1107 // if we use message buffering, we must send away all message
1108 // packets which have become too old...
1114 /* ----------------------------------------------------------------------------
1115 * Activate spark threads (PARALLEL_HASKELL only)
1116 * ------------------------------------------------------------------------- */
1118 #if defined(PARALLEL_HASKELL)
1120 scheduleActivateSpark(void)
1123 ASSERT(emptyRunQueue());
1124 /* We get here if the run queue is empty and want some work.
1125 We try to turn a spark into a thread, and add it to the run queue,
1126 from where it will be picked up in the next iteration of the scheduler
1130 /* :-[ no local threads => look out for local sparks */
1131 /* the spark pool for the current PE */
1132 pool = &(cap.r.rSparks); // JB: cap = (old) MainCap
1133 if (advisory_thread_count < RtsFlags.ParFlags.maxThreads &&
1134 pool->hd < pool->tl) {
1136 * ToDo: add GC code check that we really have enough heap afterwards!!
1138 * If we're here (no runnable threads) and we have pending
1139 * sparks, we must have a space problem. Get enough space
1140 * to turn one of those pending sparks into a
1144 spark = findSpark(rtsFalse); /* get a spark */
1145 if (spark != (rtsSpark) NULL) {
1146 tso = createThreadFromSpark(spark); /* turn the spark into a thread */
1147 IF_PAR_DEBUG(fish, // schedule,
1148 debugBelch("==== schedule: Created TSO %d (%p); %d threads active\n",
1149 tso->id, tso, advisory_thread_count));
1151 if (tso==END_TSO_QUEUE) { /* failed to activate spark->back to loop */
1152 IF_PAR_DEBUG(fish, // schedule,
1153 debugBelch("==^^ failed to create thread from spark @ %lx\n",
1155 return rtsFalse; /* failed to generate a thread */
1156 } /* otherwise fall through & pick-up new tso */
1158 IF_PAR_DEBUG(fish, // schedule,
1159 debugBelch("==^^ no local sparks (spark pool contains only NFs: %d)\n",
1160 spark_queue_len(pool)));
1161 return rtsFalse; /* failed to generate a thread */
1163 return rtsTrue; /* success in generating a thread */
1164 } else { /* no more threads permitted or pool empty */
1165 return rtsFalse; /* failed to generateThread */
1168 tso = NULL; // avoid compiler warning only
1169 return rtsFalse; /* dummy in non-PAR setup */
1172 #endif // PARALLEL_HASKELL
1174 /* ----------------------------------------------------------------------------
1175 * Get work from a remote node (PARALLEL_HASKELL only)
1176 * ------------------------------------------------------------------------- */
1178 #if defined(PARALLEL_HASKELL)
1180 scheduleGetRemoteWork(rtsBool *receivedFinish)
1182 ASSERT(emptyRunQueue());
1184 if (RtsFlags.ParFlags.BufferTime) {
1185 IF_PAR_DEBUG(verbose,
1186 debugBelch("...send all pending data,"));
1189 for (i=1; i<=nPEs; i++)
1190 sendImmediately(i); // send all messages away immediately
1194 //++EDEN++ idle() , i.e. send all buffers, wait for work
1195 // suppress fishing in EDEN... just look for incoming messages
1196 // (blocking receive)
1197 IF_PAR_DEBUG(verbose,
1198 debugBelch("...wait for incoming messages...\n"));
1199 *receivedFinish = processMessages(); // blocking receive...
1201 // and reenter scheduling loop after having received something
1202 // (return rtsFalse below)
1204 # else /* activate SPARKS machinery */
1205 /* We get here, if we have no work, tried to activate a local spark, but still
1206 have no work. We try to get a remote spark, by sending a FISH message.
1207 Thread migration should be added here, and triggered when a sequence of
1208 fishes returns without work. */
1209 delay = (RtsFlags.ParFlags.fishDelay!=0ll ? RtsFlags.ParFlags.fishDelay : 0ll);
1211 /* =8-[ no local sparks => look for work on other PEs */
1213 * We really have absolutely no work. Send out a fish
1214 * (there may be some out there already), and wait for
1215 * something to arrive. We clearly can't run any threads
1216 * until a SCHEDULE or RESUME arrives, and so that's what
1217 * we're hoping to see. (Of course, we still have to
1218 * respond to other types of messages.)
1220 rtsTime now = msTime() /*CURRENT_TIME*/;
1221 IF_PAR_DEBUG(verbose,
1222 debugBelch("-- now=%ld\n", now));
1223 IF_PAR_DEBUG(fish, // verbose,
1224 if (outstandingFishes < RtsFlags.ParFlags.maxFishes &&
1225 (last_fish_arrived_at!=0 &&
1226 last_fish_arrived_at+delay > now)) {
1227 debugBelch("--$$ <%llu> delaying FISH until %llu (last fish %llu, delay %llu)\n",
1228 now, last_fish_arrived_at+delay,
1229 last_fish_arrived_at,
1233 if (outstandingFishes < RtsFlags.ParFlags.maxFishes &&
1234 advisory_thread_count < RtsFlags.ParFlags.maxThreads) { // send a FISH, but when?
1235 if (last_fish_arrived_at==0 ||
1236 (last_fish_arrived_at+delay <= now)) { // send FISH now!
1237 /* outstandingFishes is set in sendFish, processFish;
1238 avoid flooding system with fishes via delay */
1239 next_fish_to_send_at = 0;
1241 /* ToDo: this should be done in the main scheduling loop to avoid the
1242 busy wait here; not so bad if fish delay is very small */
1243 int iq = 0; // DEBUGGING -- HWL
1244 next_fish_to_send_at = last_fish_arrived_at+delay; // remember when to send
1245 /* send a fish when ready, but process messages that arrive in the meantime */
1247 if (PacketsWaiting()) {
1249 *receivedFinish = processMessages();
1252 } while (!*receivedFinish || now<next_fish_to_send_at);
1253 // JB: This means the fish could become obsolete, if we receive
1254 // work. Better check for work again?
1255 // last line: while (!receivedFinish || !haveWork || now<...)
1256 // next line: if (receivedFinish || haveWork )
1258 if (*receivedFinish) // no need to send a FISH if we are finishing anyway
1259 return rtsFalse; // NB: this will leave scheduler loop
1260 // immediately after return!
1262 IF_PAR_DEBUG(fish, // verbose,
1263 debugBelch("--$$ <%llu> sent delayed fish (%d processMessages); active/total threads=%d/%d\n",now,iq,run_queue_len(),advisory_thread_count));
1267 // JB: IMHO, this should all be hidden inside sendFish(...)
1269 sendFish(pe, thisPE, NEW_FISH_AGE, NEW_FISH_HISTORY,
1272 // Global statistics: count no. of fishes
1273 if (RtsFlags.ParFlags.ParStats.Global &&
1274 RtsFlags.GcFlags.giveStats > NO_GC_STATS) {
1275 globalParStats.tot_fish_mess++;
1279 /* delayed fishes must have been sent by now! */
1280 next_fish_to_send_at = 0;
1283 *receivedFinish = processMessages();
1284 # endif /* SPARKS */
1287 /* NB: this function always returns rtsFalse, meaning the scheduler
1288 loop continues with the next iteration;
1290 return code means success in finding work; we enter this function
1291 if there is no local work, thus have to send a fish which takes
1292 time until it arrives with work; in the meantime we should process
1293 messages in the main loop;
1296 #endif // PARALLEL_HASKELL
1298 /* ----------------------------------------------------------------------------
1299 * PAR/GRAN: Report stats & debugging info(?)
1300 * ------------------------------------------------------------------------- */
1302 #if defined(PAR) || defined(GRAN)
1304 scheduleGranParReport(void)
1306 ASSERT(run_queue_hd != END_TSO_QUEUE);
1308 /* Take a thread from the run queue, if we have work */
1309 POP_RUN_QUEUE(t); // take_off_run_queue(END_TSO_QUEUE);
1311 /* If this TSO has got its outport closed in the meantime,
1312 * it mustn't be run. Instead, we have to clean it up as if it was finished.
1313 * It has to be marked as TH_DEAD for this purpose.
1314 * If it is TH_TERM instead, it is supposed to have finished in the normal way.
1316 JB: TODO: investigate wether state change field could be nuked
1317 entirely and replaced by the normal tso state (whatnext
1318 field). All we want to do is to kill tsos from outside.
1321 /* ToDo: write something to the log-file
1322 if (RTSflags.ParFlags.granSimStats && !sameThread)
1323 DumpGranEvent(GR_SCHEDULE, RunnableThreadsHd);
1327 /* the spark pool for the current PE */
1328 pool = &(cap.r.rSparks); // cap = (old) MainCap
1331 debugBelch("--=^ %d threads, %d sparks on [%#x]\n",
1332 run_queue_len(), spark_queue_len(pool), CURRENT_PROC));
1335 debugBelch("--=^ %d threads, %d sparks on [%#x]\n",
1336 run_queue_len(), spark_queue_len(pool), CURRENT_PROC));
1338 if (RtsFlags.ParFlags.ParStats.Full &&
1339 (t->par.sparkname != (StgInt)0) && // only log spark generated threads
1340 (emitSchedule || // forced emit
1341 (t && LastTSO && t->id != LastTSO->id))) {
1343 we are running a different TSO, so write a schedule event to log file
1344 NB: If we use fair scheduling we also have to write a deschedule
1345 event for LastTSO; with unfair scheduling we know that the
1346 previous tso has blocked whenever we switch to another tso, so
1347 we don't need it in GUM for now
1349 IF_PAR_DEBUG(fish, // schedule,
1350 debugBelch("____ scheduling spark generated thread %d (%lx) (%lx) via a forced emit\n",t->id,t,t->par.sparkname));
1352 DumpRawGranEvent(CURRENT_PROC, CURRENT_PROC,
1353 GR_SCHEDULE, t, (StgClosure *)NULL, 0, 0);
1354 emitSchedule = rtsFalse;
1359 /* ----------------------------------------------------------------------------
1360 * After running a thread...
1361 * ------------------------------------------------------------------------- */
1364 schedulePostRunThread(void)
1367 /* HACK 675: if the last thread didn't yield, make sure to print a
1368 SCHEDULE event to the log file when StgRunning the next thread, even
1369 if it is the same one as before */
1371 TimeOfLastYield = CURRENT_TIME;
1374 /* some statistics gathering in the parallel case */
1376 #if defined(GRAN) || defined(PAR) || defined(EDEN)
1380 IF_DEBUG(gran, DumpGranEvent(GR_DESCHEDULE, t));
1381 globalGranStats.tot_heapover++;
1383 globalParStats.tot_heapover++;
1390 DumpGranEvent(GR_DESCHEDULE, t));
1391 globalGranStats.tot_stackover++;
1394 // DumpGranEvent(GR_DESCHEDULE, t);
1395 globalParStats.tot_stackover++;
1399 case ThreadYielding:
1402 DumpGranEvent(GR_DESCHEDULE, t));
1403 globalGranStats.tot_yields++;
1406 // DumpGranEvent(GR_DESCHEDULE, t);
1407 globalParStats.tot_yields++;
1414 debugBelch("--<< thread %ld (%p; %s) stopped, blocking on node %p [PE %d] with BQ: ",
1415 t->id, t, whatNext_strs[t->what_next], t->block_info.closure,
1416 (t->block_info.closure==(StgClosure*)NULL ? 99 : where_is(t->block_info.closure)));
1417 if (t->block_info.closure!=(StgClosure*)NULL)
1418 print_bq(t->block_info.closure);
1421 // ??? needed; should emit block before
1423 DumpGranEvent(GR_DESCHEDULE, t));
1424 prune_eventq(t, (StgClosure *)NULL); // prune ContinueThreads for t
1427 ASSERT(procStatus[CurrentProc]==Busy ||
1428 ((procStatus[CurrentProc]==Fetching) &&
1429 (t->block_info.closure!=(StgClosure*)NULL)));
1430 if (run_queue_hds[CurrentProc] == END_TSO_QUEUE &&
1431 !(!RtsFlags.GranFlags.DoAsyncFetch &&
1432 procStatus[CurrentProc]==Fetching))
1433 procStatus[CurrentProc] = Idle;
1436 //++PAR++ blockThread() writes the event (change?)
1440 case ThreadFinished:
1444 barf("parGlobalStats: unknown return code");
1450 /* -----------------------------------------------------------------------------
1451 * Handle a thread that returned to the scheduler with ThreadHeepOverflow
1452 * -------------------------------------------------------------------------- */
1455 scheduleHandleHeapOverflow( Capability *cap, StgTSO *t )
1457 // did the task ask for a large block?
1458 if (cap->r.rHpAlloc > BLOCK_SIZE) {
1459 // if so, get one and push it on the front of the nursery.
1463 blocks = (lnat)BLOCK_ROUND_UP(cap->r.rHpAlloc) / BLOCK_SIZE;
1466 debugBelch("--<< thread %ld (%s) stopped: requesting a large block (size %ld)\n",
1467 (long)t->id, whatNext_strs[t->what_next], blocks));
1469 // don't do this if the nursery is (nearly) full, we'll GC first.
1470 if (cap->r.rCurrentNursery->link != NULL ||
1471 cap->r.rNursery->n_blocks == 1) { // paranoia to prevent infinite loop
1472 // if the nursery has only one block.
1475 bd = allocGroup( blocks );
1477 cap->r.rNursery->n_blocks += blocks;
1479 // link the new group into the list
1480 bd->link = cap->r.rCurrentNursery;
1481 bd->u.back = cap->r.rCurrentNursery->u.back;
1482 if (cap->r.rCurrentNursery->u.back != NULL) {
1483 cap->r.rCurrentNursery->u.back->link = bd;
1486 ASSERT(g0s0->blocks == cap->r.rCurrentNursery &&
1487 g0s0 == cap->r.rNursery);
1489 cap->r.rNursery->blocks = bd;
1491 cap->r.rCurrentNursery->u.back = bd;
1493 // initialise it as a nursery block. We initialise the
1494 // step, gen_no, and flags field of *every* sub-block in
1495 // this large block, because this is easier than making
1496 // sure that we always find the block head of a large
1497 // block whenever we call Bdescr() (eg. evacuate() and
1498 // isAlive() in the GC would both have to do this, at
1502 for (x = bd; x < bd + blocks; x++) {
1503 x->step = cap->r.rNursery;
1509 // This assert can be a killer if the app is doing lots
1510 // of large block allocations.
1511 IF_DEBUG(sanity, checkNurserySanity(cap->r.rNursery));
1513 // now update the nursery to point to the new block
1514 cap->r.rCurrentNursery = bd;
1516 // we might be unlucky and have another thread get on the
1517 // run queue before us and steal the large block, but in that
1518 // case the thread will just end up requesting another large
1520 pushOnRunQueue(cap,t);
1521 return rtsFalse; /* not actually GC'ing */
1526 debugBelch("--<< thread %ld (%s) stopped: HeapOverflow\n",
1527 (long)t->id, whatNext_strs[t->what_next]));
1529 ASSERT(!is_on_queue(t,CurrentProc));
1530 #elif defined(PARALLEL_HASKELL)
1531 /* Currently we emit a DESCHEDULE event before GC in GUM.
1532 ToDo: either add separate event to distinguish SYSTEM time from rest
1533 or just nuke this DESCHEDULE (and the following SCHEDULE) */
1534 if (0 && RtsFlags.ParFlags.ParStats.Full) {
1535 DumpRawGranEvent(CURRENT_PROC, CURRENT_PROC,
1536 GR_DESCHEDULE, t, (StgClosure *)NULL, 0, 0);
1537 emitSchedule = rtsTrue;
1541 pushOnRunQueue(cap,t);
1543 /* actual GC is done at the end of the while loop in schedule() */
1546 /* -----------------------------------------------------------------------------
1547 * Handle a thread that returned to the scheduler with ThreadStackOverflow
1548 * -------------------------------------------------------------------------- */
1551 scheduleHandleStackOverflow (Capability *cap, Task *task, StgTSO *t)
1553 IF_DEBUG(scheduler,debugBelch("--<< thread %ld (%s) stopped, StackOverflow\n",
1554 (long)t->id, whatNext_strs[t->what_next]));
1555 /* just adjust the stack for this thread, then pop it back
1559 /* enlarge the stack */
1560 StgTSO *new_t = threadStackOverflow(cap, t);
1562 /* The TSO attached to this Task may have moved, so update the
1565 if (task->tso == t) {
1568 pushOnRunQueue(cap,new_t);
1572 /* -----------------------------------------------------------------------------
1573 * Handle a thread that returned to the scheduler with ThreadYielding
1574 * -------------------------------------------------------------------------- */
1577 scheduleHandleYield( Capability *cap, StgTSO *t, nat prev_what_next )
1579 // Reset the context switch flag. We don't do this just before
1580 // running the thread, because that would mean we would lose ticks
1581 // during GC, which can lead to unfair scheduling (a thread hogs
1582 // the CPU because the tick always arrives during GC). This way
1583 // penalises threads that do a lot of allocation, but that seems
1584 // better than the alternative.
1587 /* put the thread back on the run queue. Then, if we're ready to
1588 * GC, check whether this is the last task to stop. If so, wake
1589 * up the GC thread. getThread will block during a GC until the
1593 if (t->what_next != prev_what_next) {
1594 debugBelch("--<< thread %ld (%s) stopped to switch evaluators\n",
1595 (long)t->id, whatNext_strs[t->what_next]);
1597 debugBelch("--<< thread %ld (%s) stopped, yielding\n",
1598 (long)t->id, whatNext_strs[t->what_next]);
1603 //debugBelch("&& Doing sanity check on yielding TSO %ld.", t->id);
1605 ASSERT(t->link == END_TSO_QUEUE);
1607 // Shortcut if we're just switching evaluators: don't bother
1608 // doing stack squeezing (which can be expensive), just run the
1610 if (t->what_next != prev_what_next) {
1615 ASSERT(!is_on_queue(t,CurrentProc));
1618 //debugBelch("&& Doing sanity check on all ThreadQueues (and their TSOs).");
1619 checkThreadQsSanity(rtsTrue));
1623 addToRunQueue(cap,t);
1626 /* add a ContinueThread event to actually process the thread */
1627 new_event(CurrentProc, CurrentProc, CurrentTime[CurrentProc],
1629 t, (StgClosure*)NULL, (rtsSpark*)NULL);
1631 debugBelch("GRAN: eventq and runnableq after adding yielded thread to queue again:\n");
1638 /* -----------------------------------------------------------------------------
1639 * Handle a thread that returned to the scheduler with ThreadBlocked
1640 * -------------------------------------------------------------------------- */
1643 scheduleHandleThreadBlocked( StgTSO *t
1644 #if !defined(GRAN) && !defined(DEBUG)
1651 debugBelch("--<< thread %ld (%p; %s) stopped, blocking on node %p [PE %d] with BQ: \n",
1652 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)));
1653 if (t->block_info.closure!=(StgClosure*)NULL) print_bq(t->block_info.closure));
1655 // ??? needed; should emit block before
1657 DumpGranEvent(GR_DESCHEDULE, t));
1658 prune_eventq(t, (StgClosure *)NULL); // prune ContinueThreads for t
1661 ASSERT(procStatus[CurrentProc]==Busy ||
1662 ((procStatus[CurrentProc]==Fetching) &&
1663 (t->block_info.closure!=(StgClosure*)NULL)));
1664 if (run_queue_hds[CurrentProc] == END_TSO_QUEUE &&
1665 !(!RtsFlags.GranFlags.DoAsyncFetch &&
1666 procStatus[CurrentProc]==Fetching))
1667 procStatus[CurrentProc] = Idle;
1671 debugBelch("--<< thread %ld (%p; %s) stopped, blocking on node %p with BQ: \n",
1672 t->id, t, whatNext_strs[t->what_next], t->block_info.closure));
1675 if (t->block_info.closure!=(StgClosure*)NULL)
1676 print_bq(t->block_info.closure));
1678 /* Send a fetch (if BlockedOnGA) and dump event to log file */
1681 /* whatever we schedule next, we must log that schedule */
1682 emitSchedule = rtsTrue;
1686 // We don't need to do anything. The thread is blocked, and it
1687 // has tidied up its stack and placed itself on whatever queue
1688 // it needs to be on.
1691 ASSERT(t->why_blocked != NotBlocked);
1692 // This might not be true under SMP: we don't have
1693 // exclusive access to this TSO, so someone might have
1694 // woken it up by now. This actually happens: try
1695 // conc023 +RTS -N2.
1699 debugBelch("--<< thread %d (%s) stopped: ",
1700 t->id, whatNext_strs[t->what_next]);
1701 printThreadBlockage(t);
1704 /* Only for dumping event to log file
1705 ToDo: do I need this in GranSim, too?
1711 /* -----------------------------------------------------------------------------
1712 * Handle a thread that returned to the scheduler with ThreadFinished
1713 * -------------------------------------------------------------------------- */
1716 scheduleHandleThreadFinished (Capability *cap STG_UNUSED, Task *task, StgTSO *t)
1718 /* Need to check whether this was a main thread, and if so,
1719 * return with the return value.
1721 * We also end up here if the thread kills itself with an
1722 * uncaught exception, see Exception.cmm.
1724 IF_DEBUG(scheduler,debugBelch("--++ thread %d (%s) finished\n",
1725 t->id, whatNext_strs[t->what_next]));
1728 endThread(t, CurrentProc); // clean-up the thread
1729 #elif defined(PARALLEL_HASKELL)
1730 /* For now all are advisory -- HWL */
1731 //if(t->priority==AdvisoryPriority) ??
1732 advisory_thread_count--; // JB: Caution with this counter, buggy!
1735 if(t->dist.priority==RevalPriority)
1739 # if defined(EDENOLD)
1740 // the thread could still have an outport... (BUG)
1741 if (t->eden.outport != -1) {
1742 // delete the outport for the tso which has finished...
1743 IF_PAR_DEBUG(eden_ports,
1744 debugBelch("WARNING: Scheduler removes outport %d for TSO %d.\n",
1745 t->eden.outport, t->id));
1748 // thread still in the process (HEAVY BUG! since outport has just been closed...)
1749 if (t->eden.epid != -1) {
1750 IF_PAR_DEBUG(eden_ports,
1751 debugBelch("WARNING: Scheduler removes TSO %d from process %d .\n",
1752 t->id, t->eden.epid));
1753 removeTSOfromProcess(t);
1758 if (RtsFlags.ParFlags.ParStats.Full &&
1759 !RtsFlags.ParFlags.ParStats.Suppressed)
1760 DumpEndEvent(CURRENT_PROC, t, rtsFalse /* not mandatory */);
1762 // t->par only contains statistics: left out for now...
1764 debugBelch("**** end thread: ended sparked thread %d (%lx); sparkname: %lx\n",
1765 t->id,t,t->par.sparkname));
1767 #endif // PARALLEL_HASKELL
1770 // Check whether the thread that just completed was a bound
1771 // thread, and if so return with the result.
1773 // There is an assumption here that all thread completion goes
1774 // through this point; we need to make sure that if a thread
1775 // ends up in the ThreadKilled state, that it stays on the run
1776 // queue so it can be dealt with here.
1781 if (t->bound != task) {
1782 #if !defined(THREADED_RTS)
1783 // Must be a bound thread that is not the topmost one. Leave
1784 // it on the run queue until the stack has unwound to the
1785 // point where we can deal with this. Leaving it on the run
1786 // queue also ensures that the garbage collector knows about
1787 // this thread and its return value (it gets dropped from the
1788 // all_threads list so there's no other way to find it).
1789 appendToRunQueue(cap,t);
1792 // this cannot happen in the threaded RTS, because a
1793 // bound thread can only be run by the appropriate Task.
1794 barf("finished bound thread that isn't mine");
1798 ASSERT(task->tso == t);
1800 if (t->what_next == ThreadComplete) {
1802 // NOTE: return val is tso->sp[1] (see StgStartup.hc)
1803 *(task->ret) = (StgClosure *)task->tso->sp[1];
1805 task->stat = Success;
1808 *(task->ret) = NULL;
1811 task->stat = Interrupted;
1813 task->stat = Killed;
1817 removeThreadLabel((StgWord)task->tso->id);
1819 return rtsTrue; // tells schedule() to return
1825 /* -----------------------------------------------------------------------------
1826 * Perform a heap census, if PROFILING
1827 * -------------------------------------------------------------------------- */
1830 scheduleDoHeapProfile( rtsBool ready_to_gc STG_UNUSED )
1832 #if defined(PROFILING)
1833 // When we have +RTS -i0 and we're heap profiling, do a census at
1834 // every GC. This lets us get repeatable runs for debugging.
1835 if (performHeapProfile ||
1836 (RtsFlags.ProfFlags.profileInterval==0 &&
1837 RtsFlags.ProfFlags.doHeapProfile && ready_to_gc)) {
1838 GarbageCollect(GetRoots, rtsTrue);
1840 performHeapProfile = rtsFalse;
1841 return rtsTrue; // true <=> we already GC'd
1847 /* -----------------------------------------------------------------------------
1848 * Perform a garbage collection if necessary
1849 * -------------------------------------------------------------------------- */
1852 scheduleDoGC( Capability *cap, Task *task USED_WHEN_SMP, rtsBool force_major )
1856 static volatile StgWord waiting_for_gc;
1857 rtsBool was_waiting;
1862 // In order to GC, there must be no threads running Haskell code.
1863 // Therefore, the GC thread needs to hold *all* the capabilities,
1864 // and release them after the GC has completed.
1866 // This seems to be the simplest way: previous attempts involved
1867 // making all the threads with capabilities give up their
1868 // capabilities and sleep except for the *last* one, which
1869 // actually did the GC. But it's quite hard to arrange for all
1870 // the other tasks to sleep and stay asleep.
1873 was_waiting = cas(&waiting_for_gc, 0, 1);
1876 IF_DEBUG(scheduler, sched_belch("someone else is trying to GC..."));
1877 yieldCapability(&cap,task);
1878 } while (waiting_for_gc);
1882 for (i=0; i < n_capabilities; i++) {
1883 IF_DEBUG(scheduler, sched_belch("ready_to_gc, grabbing all the capabilies (%d/%d)", i, n_capabilities));
1884 if (cap != &capabilities[i]) {
1885 Capability *pcap = &capabilities[i];
1886 // we better hope this task doesn't get migrated to
1887 // another Capability while we're waiting for this one.
1888 // It won't, because load balancing happens while we have
1889 // all the Capabilities, but even so it's a slightly
1890 // unsavoury invariant.
1893 waitForReturnCapability(&pcap, task);
1894 if (pcap != &capabilities[i]) {
1895 barf("scheduleDoGC: got the wrong capability");
1900 waiting_for_gc = rtsFalse;
1903 /* Kick any transactions which are invalid back to their
1904 * atomically frames. When next scheduled they will try to
1905 * commit, this commit will fail and they will retry.
1910 for (t = all_threads; t != END_TSO_QUEUE; t = next) {
1911 if (t->what_next == ThreadRelocated) {
1914 next = t->global_link;
1915 if (t -> trec != NO_TREC && t -> why_blocked == NotBlocked) {
1916 if (!stmValidateNestOfTransactions (t -> trec)) {
1917 IF_DEBUG(stm, sched_belch("trec %p found wasting its time", t));
1919 // strip the stack back to the
1920 // ATOMICALLY_FRAME, aborting the (nested)
1921 // transaction, and saving the stack of any
1922 // partially-evaluated thunks on the heap.
1923 raiseAsync_(cap, t, NULL, rtsTrue);
1926 ASSERT(get_itbl((StgClosure *)t->sp)->type == ATOMICALLY_FRAME);
1934 // so this happens periodically:
1935 scheduleCheckBlackHoles(cap);
1937 IF_DEBUG(scheduler, printAllThreads());
1939 /* everybody back, start the GC.
1940 * Could do it in this thread, or signal a condition var
1941 * to do it in another thread. Either way, we need to
1942 * broadcast on gc_pending_cond afterward.
1944 #if defined(THREADED_RTS)
1945 IF_DEBUG(scheduler,sched_belch("doing GC"));
1947 GarbageCollect(GetRoots, force_major);
1950 // release our stash of capabilities.
1951 for (i = 0; i < n_capabilities; i++) {
1952 if (cap != &capabilities[i]) {
1953 task->cap = &capabilities[i];
1954 releaseCapability(&capabilities[i]);
1961 /* add a ContinueThread event to continue execution of current thread */
1962 new_event(CurrentProc, CurrentProc, CurrentTime[CurrentProc],
1964 t, (StgClosure*)NULL, (rtsSpark*)NULL);
1966 debugBelch("GRAN: eventq and runnableq after Garbage collection:\n\n");
1972 /* ---------------------------------------------------------------------------
1973 * rtsSupportsBoundThreads(): is the RTS built to support bound threads?
1974 * used by Control.Concurrent for error checking.
1975 * ------------------------------------------------------------------------- */
1978 rtsSupportsBoundThreads(void)
1980 #if defined(THREADED_RTS)
1987 /* ---------------------------------------------------------------------------
1988 * isThreadBound(tso): check whether tso is bound to an OS thread.
1989 * ------------------------------------------------------------------------- */
1992 isThreadBound(StgTSO* tso USED_WHEN_THREADED_RTS)
1994 #if defined(THREADED_RTS)
1995 return (tso->bound != NULL);
2000 /* ---------------------------------------------------------------------------
2001 * Singleton fork(). Do not copy any running threads.
2002 * ------------------------------------------------------------------------- */
2004 #if !defined(mingw32_HOST_OS) && !defined(SMP)
2005 #define FORKPROCESS_PRIMOP_SUPPORTED
2008 #ifdef FORKPROCESS_PRIMOP_SUPPORTED
2010 deleteThreadImmediately(Capability *cap, StgTSO *tso);
2013 forkProcess(HsStablePtr *entry
2014 #ifndef FORKPROCESS_PRIMOP_SUPPORTED
2019 #ifdef FORKPROCESS_PRIMOP_SUPPORTED
2025 IF_DEBUG(scheduler,sched_belch("forking!"));
2027 // ToDo: for SMP, we should probably acquire *all* the capabilities
2032 if (pid) { // parent
2034 // just return the pid
2040 // delete all threads
2041 cap->run_queue_hd = END_TSO_QUEUE;
2042 cap->run_queue_tl = END_TSO_QUEUE;
2044 for (t = all_threads; t != END_TSO_QUEUE; t = next) {
2047 // don't allow threads to catch the ThreadKilled exception
2048 deleteThreadImmediately(cap,t);
2051 // wipe the main thread list
2052 while ((task = all_tasks) != NULL) {
2053 all_tasks = task->all_link;
2057 cap = rts_evalStableIO(cap, entry, NULL); // run the action
2058 rts_checkSchedStatus("forkProcess",cap);
2061 hs_exit(); // clean up and exit
2062 stg_exit(EXIT_SUCCESS);
2064 #else /* !FORKPROCESS_PRIMOP_SUPPORTED */
2065 barf("forkProcess#: primop not supported on this platform, sorry!\n");
2070 /* ---------------------------------------------------------------------------
2071 * Delete the threads on the run queue of the current capability.
2072 * ------------------------------------------------------------------------- */
2075 deleteRunQueue (Capability *cap)
2078 for (t = cap->run_queue_hd; t != END_TSO_QUEUE; t = next) {
2079 ASSERT(t->what_next != ThreadRelocated);
2081 deleteThread(cap, t);
2085 /* startThread and insertThread are now in GranSim.c -- HWL */
2088 /* -----------------------------------------------------------------------------
2089 Managing the suspended_ccalling_tasks list.
2090 Locks required: sched_mutex
2091 -------------------------------------------------------------------------- */
2094 suspendTask (Capability *cap, Task *task)
2096 ASSERT(task->next == NULL && task->prev == NULL);
2097 task->next = cap->suspended_ccalling_tasks;
2099 if (cap->suspended_ccalling_tasks) {
2100 cap->suspended_ccalling_tasks->prev = task;
2102 cap->suspended_ccalling_tasks = task;
2106 recoverSuspendedTask (Capability *cap, Task *task)
2109 task->prev->next = task->next;
2111 ASSERT(cap->suspended_ccalling_tasks == task);
2112 cap->suspended_ccalling_tasks = task->next;
2115 task->next->prev = task->prev;
2117 task->next = task->prev = NULL;
2120 /* ---------------------------------------------------------------------------
2121 * Suspending & resuming Haskell threads.
2123 * When making a "safe" call to C (aka _ccall_GC), the task gives back
2124 * its capability before calling the C function. This allows another
2125 * task to pick up the capability and carry on running Haskell
2126 * threads. It also means that if the C call blocks, it won't lock
2129 * The Haskell thread making the C call is put to sleep for the
2130 * duration of the call, on the susepended_ccalling_threads queue. We
2131 * give out a token to the task, which it can use to resume the thread
2132 * on return from the C function.
2133 * ------------------------------------------------------------------------- */
2136 suspendThread (StgRegTable *reg)
2139 int saved_errno = errno;
2143 /* assume that *reg is a pointer to the StgRegTable part of a Capability.
2145 cap = regTableToCapability(reg);
2147 task = cap->running_task;
2148 tso = cap->r.rCurrentTSO;
2151 sched_belch("thread %d did a safe foreign call", cap->r.rCurrentTSO->id));
2153 // XXX this might not be necessary --SDM
2154 tso->what_next = ThreadRunGHC;
2158 if(tso->blocked_exceptions == NULL) {
2159 tso->why_blocked = BlockedOnCCall;
2160 tso->blocked_exceptions = END_TSO_QUEUE;
2162 tso->why_blocked = BlockedOnCCall_NoUnblockExc;
2165 // Hand back capability
2166 task->suspended_tso = tso;
2168 ACQUIRE_LOCK(&cap->lock);
2170 suspendTask(cap,task);
2171 cap->in_haskell = rtsFalse;
2172 releaseCapability_(cap);
2174 RELEASE_LOCK(&cap->lock);
2176 #if defined(THREADED_RTS)
2177 /* Preparing to leave the RTS, so ensure there's a native thread/task
2178 waiting to take over.
2180 IF_DEBUG(scheduler, sched_belch("thread %d: leaving RTS", tso->id));
2183 errno = saved_errno;
2188 resumeThread (void *task_)
2192 int saved_errno = errno;
2196 // Wait for permission to re-enter the RTS with the result.
2197 waitForReturnCapability(&cap,task);
2198 // we might be on a different capability now... but if so, our
2199 // entry on the suspended_ccalling_tasks list will also have been
2202 // Remove the thread from the suspended list
2203 recoverSuspendedTask(cap,task);
2205 tso = task->suspended_tso;
2206 task->suspended_tso = NULL;
2207 tso->link = END_TSO_QUEUE;
2208 IF_DEBUG(scheduler, sched_belch("thread %d: re-entering RTS", tso->id));
2210 if (tso->why_blocked == BlockedOnCCall) {
2211 awakenBlockedQueue(cap,tso->blocked_exceptions);
2212 tso->blocked_exceptions = NULL;
2215 /* Reset blocking status */
2216 tso->why_blocked = NotBlocked;
2218 cap->r.rCurrentTSO = tso;
2219 cap->in_haskell = rtsTrue;
2220 errno = saved_errno;
2225 /* ---------------------------------------------------------------------------
2226 * Comparing Thread ids.
2228 * This is used from STG land in the implementation of the
2229 * instances of Eq/Ord for ThreadIds.
2230 * ------------------------------------------------------------------------ */
2233 cmp_thread(StgPtr tso1, StgPtr tso2)
2235 StgThreadID id1 = ((StgTSO *)tso1)->id;
2236 StgThreadID id2 = ((StgTSO *)tso2)->id;
2238 if (id1 < id2) return (-1);
2239 if (id1 > id2) return 1;
2243 /* ---------------------------------------------------------------------------
2244 * Fetching the ThreadID from an StgTSO.
2246 * This is used in the implementation of Show for ThreadIds.
2247 * ------------------------------------------------------------------------ */
2249 rts_getThreadId(StgPtr tso)
2251 return ((StgTSO *)tso)->id;
2256 labelThread(StgPtr tso, char *label)
2261 /* Caveat: Once set, you can only set the thread name to "" */
2262 len = strlen(label)+1;
2263 buf = stgMallocBytes(len * sizeof(char), "Schedule.c:labelThread()");
2264 strncpy(buf,label,len);
2265 /* Update will free the old memory for us */
2266 updateThreadLabel(((StgTSO *)tso)->id,buf);
2270 /* ---------------------------------------------------------------------------
2271 Create a new thread.
2273 The new thread starts with the given stack size. Before the
2274 scheduler can run, however, this thread needs to have a closure
2275 (and possibly some arguments) pushed on its stack. See
2276 pushClosure() in Schedule.h.
2278 createGenThread() and createIOThread() (in SchedAPI.h) are
2279 convenient packaged versions of this function.
2281 currently pri (priority) is only used in a GRAN setup -- HWL
2282 ------------------------------------------------------------------------ */
2284 /* currently pri (priority) is only used in a GRAN setup -- HWL */
2286 createThread(nat size, StgInt pri)
2289 createThread(Capability *cap, nat size)
2295 /* sched_mutex is *not* required */
2297 /* First check whether we should create a thread at all */
2298 #if defined(PARALLEL_HASKELL)
2299 /* check that no more than RtsFlags.ParFlags.maxThreads threads are created */
2300 if (advisory_thread_count >= RtsFlags.ParFlags.maxThreads) {
2302 debugBelch("{createThread}Daq ghuH: refusing to create another thread; no more than %d threads allowed (currently %d)\n",
2303 RtsFlags.ParFlags.maxThreads, advisory_thread_count);
2304 return END_TSO_QUEUE;
2310 ASSERT(!RtsFlags.GranFlags.Light || CurrentProc==0);
2313 // ToDo: check whether size = stack_size - TSO_STRUCT_SIZEW
2315 /* catch ridiculously small stack sizes */
2316 if (size < MIN_STACK_WORDS + TSO_STRUCT_SIZEW) {
2317 size = MIN_STACK_WORDS + TSO_STRUCT_SIZEW;
2320 stack_size = size - TSO_STRUCT_SIZEW;
2322 tso = (StgTSO *)allocateLocal(cap, size);
2323 TICK_ALLOC_TSO(stack_size, 0);
2325 SET_HDR(tso, &stg_TSO_info, CCS_SYSTEM);
2327 SET_GRAN_HDR(tso, ThisPE);
2330 // Always start with the compiled code evaluator
2331 tso->what_next = ThreadRunGHC;
2333 tso->why_blocked = NotBlocked;
2334 tso->blocked_exceptions = NULL;
2336 tso->saved_errno = 0;
2339 tso->stack_size = stack_size;
2340 tso->max_stack_size = round_to_mblocks(RtsFlags.GcFlags.maxStkSize)
2342 tso->sp = (P_)&(tso->stack) + stack_size;
2344 tso->trec = NO_TREC;
2347 tso->prof.CCCS = CCS_MAIN;
2350 /* put a stop frame on the stack */
2351 tso->sp -= sizeofW(StgStopFrame);
2352 SET_HDR((StgClosure*)tso->sp,(StgInfoTable *)&stg_stop_thread_info,CCS_SYSTEM);
2353 tso->link = END_TSO_QUEUE;
2357 /* uses more flexible routine in GranSim */
2358 insertThread(tso, CurrentProc);
2360 /* In a non-GranSim setup the pushing of a TSO onto the runq is separated
2366 if (RtsFlags.GranFlags.GranSimStats.Full)
2367 DumpGranEvent(GR_START,tso);
2368 #elif defined(PARALLEL_HASKELL)
2369 if (RtsFlags.ParFlags.ParStats.Full)
2370 DumpGranEvent(GR_STARTQ,tso);
2371 /* HACk to avoid SCHEDULE
2375 /* Link the new thread on the global thread list.
2377 ACQUIRE_LOCK(&sched_mutex);
2378 tso->id = next_thread_id++; // while we have the mutex
2379 tso->global_link = all_threads;
2381 RELEASE_LOCK(&sched_mutex);
2384 tso->dist.priority = MandatoryPriority; //by default that is...
2388 tso->gran.pri = pri;
2390 tso->gran.magic = TSO_MAGIC; // debugging only
2392 tso->gran.sparkname = 0;
2393 tso->gran.startedat = CURRENT_TIME;
2394 tso->gran.exported = 0;
2395 tso->gran.basicblocks = 0;
2396 tso->gran.allocs = 0;
2397 tso->gran.exectime = 0;
2398 tso->gran.fetchtime = 0;
2399 tso->gran.fetchcount = 0;
2400 tso->gran.blocktime = 0;
2401 tso->gran.blockcount = 0;
2402 tso->gran.blockedat = 0;
2403 tso->gran.globalsparks = 0;
2404 tso->gran.localsparks = 0;
2405 if (RtsFlags.GranFlags.Light)
2406 tso->gran.clock = Now; /* local clock */
2408 tso->gran.clock = 0;
2410 IF_DEBUG(gran,printTSO(tso));
2411 #elif defined(PARALLEL_HASKELL)
2413 tso->par.magic = TSO_MAGIC; // debugging only
2415 tso->par.sparkname = 0;
2416 tso->par.startedat = CURRENT_TIME;
2417 tso->par.exported = 0;
2418 tso->par.basicblocks = 0;
2419 tso->par.allocs = 0;
2420 tso->par.exectime = 0;
2421 tso->par.fetchtime = 0;
2422 tso->par.fetchcount = 0;
2423 tso->par.blocktime = 0;
2424 tso->par.blockcount = 0;
2425 tso->par.blockedat = 0;
2426 tso->par.globalsparks = 0;
2427 tso->par.localsparks = 0;
2431 globalGranStats.tot_threads_created++;
2432 globalGranStats.threads_created_on_PE[CurrentProc]++;
2433 globalGranStats.tot_sq_len += spark_queue_len(CurrentProc);
2434 globalGranStats.tot_sq_probes++;
2435 #elif defined(PARALLEL_HASKELL)
2436 // collect parallel global statistics (currently done together with GC stats)
2437 if (RtsFlags.ParFlags.ParStats.Global &&
2438 RtsFlags.GcFlags.giveStats > NO_GC_STATS) {
2439 //debugBelch("Creating thread %d @ %11.2f\n", tso->id, usertime());
2440 globalParStats.tot_threads_created++;
2446 sched_belch("==__ schedule: Created TSO %d (%p);",
2447 CurrentProc, tso, tso->id));
2448 #elif defined(PARALLEL_HASKELL)
2449 IF_PAR_DEBUG(verbose,
2450 sched_belch("==__ schedule: Created TSO %d (%p); %d threads active",
2451 (long)tso->id, tso, advisory_thread_count));
2453 IF_DEBUG(scheduler,sched_belch("created thread %ld, stack size = %lx words",
2454 (long)tso->id, (long)tso->stack_size));
2461 all parallel thread creation calls should fall through the following routine.
2464 createThreadFromSpark(rtsSpark spark)
2466 ASSERT(spark != (rtsSpark)NULL);
2467 // JB: TAKE CARE OF THIS COUNTER! BUGGY
2468 if (advisory_thread_count >= RtsFlags.ParFlags.maxThreads)
2470 barf("{createSparkThread}Daq ghuH: refusing to create another thread; no more than %d threads allowed (currently %d)",
2471 RtsFlags.ParFlags.maxThreads, advisory_thread_count);
2472 return END_TSO_QUEUE;
2476 tso = createThread(RtsFlags.GcFlags.initialStkSize);
2477 if (tso==END_TSO_QUEUE)
2478 barf("createSparkThread: Cannot create TSO");
2480 tso->priority = AdvisoryPriority;
2482 pushClosure(tso,spark);
2484 advisory_thread_count++; // JB: TAKE CARE OF THIS COUNTER! BUGGY
2491 Turn a spark into a thread.
2492 ToDo: fix for SMP (needs to acquire SCHED_MUTEX!)
2496 activateSpark (rtsSpark spark)
2500 tso = createSparkThread(spark);
2501 if (RtsFlags.ParFlags.ParStats.Full) {
2502 //ASSERT(run_queue_hd == END_TSO_QUEUE); // I think ...
2503 IF_PAR_DEBUG(verbose,
2504 debugBelch("==^^ activateSpark: turning spark of closure %p (%s) into a thread\n",
2505 (StgClosure *)spark, info_type((StgClosure *)spark)));
2507 // ToDo: fwd info on local/global spark to thread -- HWL
2508 // tso->gran.exported = spark->exported;
2509 // tso->gran.locked = !spark->global;
2510 // tso->gran.sparkname = spark->name;
2516 /* ---------------------------------------------------------------------------
2519 * scheduleThread puts a thread on the end of the runnable queue.
2520 * This will usually be done immediately after a thread is created.
2521 * The caller of scheduleThread must create the thread using e.g.
2522 * createThread and push an appropriate closure
2523 * on this thread's stack before the scheduler is invoked.
2524 * ------------------------------------------------------------------------ */
2527 scheduleThread(Capability *cap, StgTSO *tso)
2529 // The thread goes at the *end* of the run-queue, to avoid possible
2530 // starvation of any threads already on the queue.
2531 appendToRunQueue(cap,tso);
2535 scheduleWaitThread (StgTSO* tso, /*[out]*/HaskellObj* ret, Capability *cap)
2539 // We already created/initialised the Task
2540 task = cap->running_task;
2542 // This TSO is now a bound thread; make the Task and TSO
2543 // point to each other.
2548 task->stat = NoStatus;
2550 appendToRunQueue(cap,tso);
2552 IF_DEBUG(scheduler, sched_belch("new bound thread (%d)", tso->id));
2555 /* GranSim specific init */
2556 CurrentTSO = m->tso; // the TSO to run
2557 procStatus[MainProc] = Busy; // status of main PE
2558 CurrentProc = MainProc; // PE to run it on
2561 cap = schedule(cap,task);
2563 ASSERT(task->stat != NoStatus);
2564 ASSERT_CAPABILITY_INVARIANTS(cap,task);
2566 IF_DEBUG(scheduler, sched_belch("bound thread (%d) finished", task->tso->id));
2570 /* ----------------------------------------------------------------------------
2572 * ------------------------------------------------------------------------- */
2574 #if defined(THREADED_RTS)
2576 workerStart(Task *task)
2580 // See startWorkerTask().
2581 ACQUIRE_LOCK(&task->lock);
2583 RELEASE_LOCK(&task->lock);
2585 // set the thread-local pointer to the Task:
2588 // schedule() runs without a lock.
2589 cap = schedule(cap,task);
2591 // On exit from schedule(), we have a Capability.
2592 releaseCapability(cap);
2597 /* ---------------------------------------------------------------------------
2600 * Initialise the scheduler. This resets all the queues - if the
2601 * queues contained any threads, they'll be garbage collected at the
2604 * ------------------------------------------------------------------------ */
2611 for (i=0; i<=MAX_PROC; i++) {
2612 run_queue_hds[i] = END_TSO_QUEUE;
2613 run_queue_tls[i] = END_TSO_QUEUE;
2614 blocked_queue_hds[i] = END_TSO_QUEUE;
2615 blocked_queue_tls[i] = END_TSO_QUEUE;
2616 ccalling_threadss[i] = END_TSO_QUEUE;
2617 blackhole_queue[i] = END_TSO_QUEUE;
2618 sleeping_queue = END_TSO_QUEUE;
2620 #elif !defined(THREADED_RTS)
2621 blocked_queue_hd = END_TSO_QUEUE;
2622 blocked_queue_tl = END_TSO_QUEUE;
2623 sleeping_queue = END_TSO_QUEUE;
2626 blackhole_queue = END_TSO_QUEUE;
2627 all_threads = END_TSO_QUEUE;
2632 RtsFlags.ConcFlags.ctxtSwitchTicks =
2633 RtsFlags.ConcFlags.ctxtSwitchTime / TICK_MILLISECS;
2635 #if defined(THREADED_RTS)
2636 /* Initialise the mutex and condition variables used by
2638 initMutex(&sched_mutex);
2641 ACQUIRE_LOCK(&sched_mutex);
2643 /* A capability holds the state a native thread needs in
2644 * order to execute STG code. At least one capability is
2645 * floating around (only SMP builds have more than one).
2653 * Eagerly start one worker to run each Capability, except for
2654 * Capability 0. The idea is that we're probably going to start a
2655 * bound thread on Capability 0 pretty soon, so we don't want a
2656 * worker task hogging it.
2661 for (i = 1; i < n_capabilities; i++) {
2662 cap = &capabilities[i];
2663 ACQUIRE_LOCK(&cap->lock);
2664 startWorkerTask(cap, workerStart);
2665 RELEASE_LOCK(&cap->lock);
2670 #if /* defined(SMP) ||*/ defined(PARALLEL_HASKELL)
2674 RELEASE_LOCK(&sched_mutex);
2678 exitScheduler( void )
2680 interrupted = rtsTrue;
2681 shutting_down_scheduler = rtsTrue;
2683 #if defined(THREADED_RTS)
2688 ACQUIRE_LOCK(&sched_mutex);
2689 task = newBoundTask();
2690 RELEASE_LOCK(&sched_mutex);
2692 for (i = 0; i < n_capabilities; i++) {
2693 shutdownCapability(&capabilities[i], task);
2695 boundTaskExiting(task);
2701 /* ---------------------------------------------------------------------------
2702 Where are the roots that we know about?
2704 - all the threads on the runnable queue
2705 - all the threads on the blocked queue
2706 - all the threads on the sleeping queue
2707 - all the thread currently executing a _ccall_GC
2708 - all the "main threads"
2710 ------------------------------------------------------------------------ */
2712 /* This has to be protected either by the scheduler monitor, or by the
2713 garbage collection monitor (probably the latter).
2718 GetRoots( evac_fn evac )
2725 for (i=0; i<=RtsFlags.GranFlags.proc; i++) {
2726 if ((run_queue_hds[i] != END_TSO_QUEUE) && ((run_queue_hds[i] != NULL)))
2727 evac((StgClosure **)&run_queue_hds[i]);
2728 if ((run_queue_tls[i] != END_TSO_QUEUE) && ((run_queue_tls[i] != NULL)))
2729 evac((StgClosure **)&run_queue_tls[i]);
2731 if ((blocked_queue_hds[i] != END_TSO_QUEUE) && ((blocked_queue_hds[i] != NULL)))
2732 evac((StgClosure **)&blocked_queue_hds[i]);
2733 if ((blocked_queue_tls[i] != END_TSO_QUEUE) && ((blocked_queue_tls[i] != NULL)))
2734 evac((StgClosure **)&blocked_queue_tls[i]);
2735 if ((ccalling_threadss[i] != END_TSO_QUEUE) && ((ccalling_threadss[i] != NULL)))
2736 evac((StgClosure **)&ccalling_threads[i]);
2743 for (i = 0; i < n_capabilities; i++) {
2744 cap = &capabilities[i];
2745 evac((StgClosure **)&cap->run_queue_hd);
2746 evac((StgClosure **)&cap->run_queue_tl);
2748 for (task = cap->suspended_ccalling_tasks; task != NULL;
2750 evac((StgClosure **)&task->suspended_tso);
2754 #if !defined(THREADED_RTS)
2755 evac((StgClosure **)&blocked_queue_hd);
2756 evac((StgClosure **)&blocked_queue_tl);
2757 evac((StgClosure **)&sleeping_queue);
2761 evac((StgClosure **)&blackhole_queue);
2763 #if defined(PARALLEL_HASKELL) || defined(GRAN)
2764 markSparkQueue(evac);
2767 #if defined(RTS_USER_SIGNALS)
2768 // mark the signal handlers (signals should be already blocked)
2769 markSignalHandlers(evac);
2773 /* -----------------------------------------------------------------------------
2776 This is the interface to the garbage collector from Haskell land.
2777 We provide this so that external C code can allocate and garbage
2778 collect when called from Haskell via _ccall_GC.
2780 It might be useful to provide an interface whereby the programmer
2781 can specify more roots (ToDo).
2783 This needs to be protected by the GC condition variable above. KH.
2784 -------------------------------------------------------------------------- */
2786 static void (*extra_roots)(evac_fn);
2792 // ToDo: we have to grab all the capabilities here.
2793 errorBelch("performGC not supported in threaded RTS (yet)");
2794 stg_exit(EXIT_FAILURE);
2796 /* Obligated to hold this lock upon entry */
2797 GarbageCollect(GetRoots,rtsFalse);
2801 performMajorGC(void)
2804 errorBelch("performMayjorGC not supported in threaded RTS (yet)");
2805 stg_exit(EXIT_FAILURE);
2807 GarbageCollect(GetRoots,rtsTrue);
2811 AllRoots(evac_fn evac)
2813 GetRoots(evac); // the scheduler's roots
2814 extra_roots(evac); // the user's roots
2818 performGCWithRoots(void (*get_roots)(evac_fn))
2821 errorBelch("performGCWithRoots not supported in threaded RTS (yet)");
2822 stg_exit(EXIT_FAILURE);
2824 extra_roots = get_roots;
2825 GarbageCollect(AllRoots,rtsFalse);
2828 /* -----------------------------------------------------------------------------
2831 If the thread has reached its maximum stack size, then raise the
2832 StackOverflow exception in the offending thread. Otherwise
2833 relocate the TSO into a larger chunk of memory and adjust its stack
2835 -------------------------------------------------------------------------- */
2838 threadStackOverflow(Capability *cap, StgTSO *tso)
2840 nat new_stack_size, stack_words;
2845 IF_DEBUG(sanity,checkTSO(tso));
2846 if (tso->stack_size >= tso->max_stack_size) {
2849 debugBelch("@@ threadStackOverflow of TSO %ld (%p): stack too large (now %ld; max is %ld)\n",
2850 (long)tso->id, tso, (long)tso->stack_size, (long)tso->max_stack_size);
2851 /* If we're debugging, just print out the top of the stack */
2852 printStackChunk(tso->sp, stg_min(tso->stack+tso->stack_size,
2855 /* Send this thread the StackOverflow exception */
2856 raiseAsync(cap, tso, (StgClosure *)stackOverflow_closure);
2860 /* Try to double the current stack size. If that takes us over the
2861 * maximum stack size for this thread, then use the maximum instead.
2862 * Finally round up so the TSO ends up as a whole number of blocks.
2864 new_stack_size = stg_min(tso->stack_size * 2, tso->max_stack_size);
2865 new_tso_size = (lnat)BLOCK_ROUND_UP(new_stack_size * sizeof(W_) +
2866 TSO_STRUCT_SIZE)/sizeof(W_);
2867 new_tso_size = round_to_mblocks(new_tso_size); /* Be MBLOCK-friendly */
2868 new_stack_size = new_tso_size - TSO_STRUCT_SIZEW;
2870 IF_DEBUG(scheduler, sched_belch("increasing stack size from %ld words to %d.\n", (long)tso->stack_size, new_stack_size));
2872 dest = (StgTSO *)allocate(new_tso_size);
2873 TICK_ALLOC_TSO(new_stack_size,0);
2875 /* copy the TSO block and the old stack into the new area */
2876 memcpy(dest,tso,TSO_STRUCT_SIZE);
2877 stack_words = tso->stack + tso->stack_size - tso->sp;
2878 new_sp = (P_)dest + new_tso_size - stack_words;
2879 memcpy(new_sp, tso->sp, stack_words * sizeof(W_));
2881 /* relocate the stack pointers... */
2883 dest->stack_size = new_stack_size;
2885 /* Mark the old TSO as relocated. We have to check for relocated
2886 * TSOs in the garbage collector and any primops that deal with TSOs.
2888 * It's important to set the sp value to just beyond the end
2889 * of the stack, so we don't attempt to scavenge any part of the
2892 tso->what_next = ThreadRelocated;
2894 tso->sp = (P_)&(tso->stack[tso->stack_size]);
2895 tso->why_blocked = NotBlocked;
2897 IF_PAR_DEBUG(verbose,
2898 debugBelch("@@ threadStackOverflow of TSO %d (now at %p): stack size increased to %ld\n",
2899 tso->id, tso, tso->stack_size);
2900 /* If we're debugging, just print out the top of the stack */
2901 printStackChunk(tso->sp, stg_min(tso->stack+tso->stack_size,
2904 IF_DEBUG(sanity,checkTSO(tso));
2906 IF_DEBUG(scheduler,printTSO(dest));
2912 /* ---------------------------------------------------------------------------
2913 Wake up a queue that was blocked on some resource.
2914 ------------------------------------------------------------------------ */
2918 unblockCount ( StgBlockingQueueElement *bqe, StgClosure *node )
2921 #elif defined(PARALLEL_HASKELL)
2923 unblockCount ( StgBlockingQueueElement *bqe, StgClosure *node )
2925 /* write RESUME events to log file and
2926 update blocked and fetch time (depending on type of the orig closure) */
2927 if (RtsFlags.ParFlags.ParStats.Full) {
2928 DumpRawGranEvent(CURRENT_PROC, CURRENT_PROC,
2929 GR_RESUMEQ, ((StgTSO *)bqe), ((StgTSO *)bqe)->block_info.closure,
2930 0, 0 /* spark_queue_len(ADVISORY_POOL) */);
2931 if (emptyRunQueue())
2932 emitSchedule = rtsTrue;
2934 switch (get_itbl(node)->type) {
2936 ((StgTSO *)bqe)->par.fetchtime += CURRENT_TIME-((StgTSO *)bqe)->par.blockedat;
2941 ((StgTSO *)bqe)->par.blocktime += CURRENT_TIME-((StgTSO *)bqe)->par.blockedat;
2948 barf("{unblockOne}Daq Qagh: unexpected closure in blocking queue");
2955 StgBlockingQueueElement *
2956 unblockOne(StgBlockingQueueElement *bqe, StgClosure *node)
2959 PEs node_loc, tso_loc;
2961 node_loc = where_is(node); // should be lifted out of loop
2962 tso = (StgTSO *)bqe; // wastes an assignment to get the type right
2963 tso_loc = where_is((StgClosure *)tso);
2964 if (IS_LOCAL_TO(PROCS(node),tso_loc)) { // TSO is local
2965 /* !fake_fetch => TSO is on CurrentProc is same as IS_LOCAL_TO */
2966 ASSERT(CurrentProc!=node_loc || tso_loc==CurrentProc);
2967 CurrentTime[CurrentProc] += RtsFlags.GranFlags.Costs.lunblocktime;
2968 // insertThread(tso, node_loc);
2969 new_event(tso_loc, tso_loc, CurrentTime[CurrentProc],
2971 tso, node, (rtsSpark*)NULL);
2972 tso->link = END_TSO_QUEUE; // overwrite link just to be sure
2975 } else { // TSO is remote (actually should be FMBQ)
2976 CurrentTime[CurrentProc] += RtsFlags.GranFlags.Costs.mpacktime +
2977 RtsFlags.GranFlags.Costs.gunblocktime +
2978 RtsFlags.GranFlags.Costs.latency;
2979 new_event(tso_loc, CurrentProc, CurrentTime[CurrentProc],
2981 tso, node, (rtsSpark*)NULL);
2982 tso->link = END_TSO_QUEUE; // overwrite link just to be sure
2985 /* the thread-queue-overhead is accounted for in either Resume or UnblockThread */
2987 debugBelch(" %s TSO %d (%p) [PE %d] (block_info.closure=%p) (next=%p) ,",
2988 (node_loc==tso_loc ? "Local" : "Global"),
2989 tso->id, tso, CurrentProc, tso->block_info.closure, tso->link));
2990 tso->block_info.closure = NULL;
2991 IF_DEBUG(scheduler,debugBelch("-- Waking up thread %ld (%p)\n",
2994 #elif defined(PARALLEL_HASKELL)
2995 StgBlockingQueueElement *
2996 unblockOne(StgBlockingQueueElement *bqe, StgClosure *node)
2998 StgBlockingQueueElement *next;
3000 switch (get_itbl(bqe)->type) {
3002 ASSERT(((StgTSO *)bqe)->why_blocked != NotBlocked);
3003 /* if it's a TSO just push it onto the run_queue */
3005 ((StgTSO *)bqe)->link = END_TSO_QUEUE; // debugging?
3006 APPEND_TO_RUN_QUEUE((StgTSO *)bqe);
3008 unblockCount(bqe, node);
3009 /* reset blocking status after dumping event */
3010 ((StgTSO *)bqe)->why_blocked = NotBlocked;
3014 /* if it's a BLOCKED_FETCH put it on the PendingFetches list */
3016 bqe->link = (StgBlockingQueueElement *)PendingFetches;
3017 PendingFetches = (StgBlockedFetch *)bqe;
3021 /* can ignore this case in a non-debugging setup;
3022 see comments on RBHSave closures above */
3024 /* check that the closure is an RBHSave closure */
3025 ASSERT(get_itbl((StgClosure *)bqe) == &stg_RBH_Save_0_info ||
3026 get_itbl((StgClosure *)bqe) == &stg_RBH_Save_1_info ||
3027 get_itbl((StgClosure *)bqe) == &stg_RBH_Save_2_info);
3031 barf("{unblockOne}Daq Qagh: Unexpected IP (%#lx; %s) in blocking queue at %#lx\n",
3032 get_itbl((StgClosure *)bqe), info_type((StgClosure *)bqe),
3036 IF_PAR_DEBUG(bq, debugBelch(", %p (%s)\n", bqe, info_type((StgClosure*)bqe)));
3042 unblockOne(Capability *cap, StgTSO *tso)
3046 ASSERT(get_itbl(tso)->type == TSO);
3047 ASSERT(tso->why_blocked != NotBlocked);
3048 tso->why_blocked = NotBlocked;
3050 tso->link = END_TSO_QUEUE;
3052 // We might have just migrated this TSO to our Capability:
3054 tso->bound->cap = cap;
3057 appendToRunQueue(cap,tso);
3059 // we're holding a newly woken thread, make sure we context switch
3060 // quickly so we can migrate it if necessary.
3062 IF_DEBUG(scheduler,sched_belch("waking up thread %ld", (long)tso->id));
3069 awakenBlockedQueue(StgBlockingQueueElement *q, StgClosure *node)
3071 StgBlockingQueueElement *bqe;
3076 debugBelch("##-_ AwBQ for node %p on PE %d @ %ld by TSO %d (%p): \n", \
3077 node, CurrentProc, CurrentTime[CurrentProc],
3078 CurrentTSO->id, CurrentTSO));
3080 node_loc = where_is(node);
3082 ASSERT(q == END_BQ_QUEUE ||
3083 get_itbl(q)->type == TSO || // q is either a TSO or an RBHSave
3084 get_itbl(q)->type == CONSTR); // closure (type constructor)
3085 ASSERT(is_unique(node));
3087 /* FAKE FETCH: magically copy the node to the tso's proc;
3088 no Fetch necessary because in reality the node should not have been
3089 moved to the other PE in the first place
3091 if (CurrentProc!=node_loc) {
3093 debugBelch("## node %p is on PE %d but CurrentProc is %d (TSO %d); assuming fake fetch and adjusting bitmask (old: %#x)\n",
3094 node, node_loc, CurrentProc, CurrentTSO->id,
3095 // CurrentTSO, where_is(CurrentTSO),
3096 node->header.gran.procs));
3097 node->header.gran.procs = (node->header.gran.procs) | PE_NUMBER(CurrentProc);
3099 debugBelch("## new bitmask of node %p is %#x\n",
3100 node, node->header.gran.procs));
3101 if (RtsFlags.GranFlags.GranSimStats.Global) {
3102 globalGranStats.tot_fake_fetches++;
3107 // ToDo: check: ASSERT(CurrentProc==node_loc);
3108 while (get_itbl(bqe)->type==TSO) { // q != END_TSO_QUEUE) {
3111 bqe points to the current element in the queue
3112 next points to the next element in the queue
3114 //tso = (StgTSO *)bqe; // wastes an assignment to get the type right
3115 //tso_loc = where_is(tso);
3117 bqe = unblockOne(bqe, node);
3120 /* if this is the BQ of an RBH, we have to put back the info ripped out of
3121 the closure to make room for the anchor of the BQ */
3122 if (bqe!=END_BQ_QUEUE) {
3123 ASSERT(get_itbl(node)->type == RBH && get_itbl(bqe)->type == CONSTR);
3125 ASSERT((info_ptr==&RBH_Save_0_info) ||
3126 (info_ptr==&RBH_Save_1_info) ||
3127 (info_ptr==&RBH_Save_2_info));
3129 /* cf. convertToRBH in RBH.c for writing the RBHSave closure */
3130 ((StgRBH *)node)->blocking_queue = (StgBlockingQueueElement *)((StgRBHSave *)bqe)->payload[0];
3131 ((StgRBH *)node)->mut_link = (StgMutClosure *)((StgRBHSave *)bqe)->payload[1];
3134 debugBelch("## Filled in RBH_Save for %p (%s) at end of AwBQ\n",
3135 node, info_type(node)));
3138 /* statistics gathering */
3139 if (RtsFlags.GranFlags.GranSimStats.Global) {
3140 // globalGranStats.tot_bq_processing_time += bq_processing_time;
3141 globalGranStats.tot_bq_len += len; // total length of all bqs awakened
3142 // globalGranStats.tot_bq_len_local += len_local; // same for local TSOs only
3143 globalGranStats.tot_awbq++; // total no. of bqs awakened
3146 debugBelch("## BQ Stats of %p: [%d entries] %s\n",
3147 node, len, (bqe!=END_BQ_QUEUE) ? "RBH" : ""));
3149 #elif defined(PARALLEL_HASKELL)
3151 awakenBlockedQueue(StgBlockingQueueElement *q, StgClosure *node)
3153 StgBlockingQueueElement *bqe;
3155 IF_PAR_DEBUG(verbose,
3156 debugBelch("##-_ AwBQ for node %p on [%x]: \n",
3160 if(get_itbl(q)->type == CONSTR || q==END_BQ_QUEUE) {
3161 IF_PAR_DEBUG(verbose, debugBelch("## ... nothing to unblock so lets just return. RFP (BUG?)\n"));
3166 ASSERT(q == END_BQ_QUEUE ||
3167 get_itbl(q)->type == TSO ||
3168 get_itbl(q)->type == BLOCKED_FETCH ||
3169 get_itbl(q)->type == CONSTR);
3172 while (get_itbl(bqe)->type==TSO ||
3173 get_itbl(bqe)->type==BLOCKED_FETCH) {
3174 bqe = unblockOne(bqe, node);
3178 #else /* !GRAN && !PARALLEL_HASKELL */
3181 awakenBlockedQueue(Capability *cap, StgTSO *tso)
3183 if (tso == NULL) return; // hack; see bug #1235728, and comments in
3185 while (tso != END_TSO_QUEUE) {
3186 tso = unblockOne(cap,tso);
3191 /* ---------------------------------------------------------------------------
3193 - usually called inside a signal handler so it mustn't do anything fancy.
3194 ------------------------------------------------------------------------ */
3197 interruptStgRts(void)
3201 #if defined(THREADED_RTS)
3202 prodAllCapabilities();
3206 /* -----------------------------------------------------------------------------
3209 This is for use when we raise an exception in another thread, which
3211 This has nothing to do with the UnblockThread event in GranSim. -- HWL
3212 -------------------------------------------------------------------------- */
3214 #if defined(GRAN) || defined(PARALLEL_HASKELL)
3216 NB: only the type of the blocking queue is different in GranSim and GUM
3217 the operations on the queue-elements are the same
3218 long live polymorphism!
3220 Locks: sched_mutex is held upon entry and exit.
3224 unblockThread(Capability *cap, StgTSO *tso)
3226 StgBlockingQueueElement *t, **last;
3228 switch (tso->why_blocked) {
3231 return; /* not blocked */
3234 // Be careful: nothing to do here! We tell the scheduler that the thread
3235 // is runnable and we leave it to the stack-walking code to abort the
3236 // transaction while unwinding the stack. We should perhaps have a debugging
3237 // test to make sure that this really happens and that the 'zombie' transaction
3238 // does not get committed.
3242 ASSERT(get_itbl(tso->block_info.closure)->type == MVAR);
3244 StgBlockingQueueElement *last_tso = END_BQ_QUEUE;
3245 StgMVar *mvar = (StgMVar *)(tso->block_info.closure);
3247 last = (StgBlockingQueueElement **)&mvar->head;
3248 for (t = (StgBlockingQueueElement *)mvar->head;
3250 last = &t->link, last_tso = t, t = t->link) {
3251 if (t == (StgBlockingQueueElement *)tso) {
3252 *last = (StgBlockingQueueElement *)tso->link;
3253 if (mvar->tail == tso) {
3254 mvar->tail = (StgTSO *)last_tso;
3259 barf("unblockThread (MVAR): TSO not found");
3262 case BlockedOnBlackHole:
3263 ASSERT(get_itbl(tso->block_info.closure)->type == BLACKHOLE_BQ);
3265 StgBlockingQueue *bq = (StgBlockingQueue *)(tso->block_info.closure);
3267 last = &bq->blocking_queue;
3268 for (t = bq->blocking_queue;
3270 last = &t->link, t = t->link) {
3271 if (t == (StgBlockingQueueElement *)tso) {
3272 *last = (StgBlockingQueueElement *)tso->link;
3276 barf("unblockThread (BLACKHOLE): TSO not found");
3279 case BlockedOnException:
3281 StgTSO *target = tso->block_info.tso;
3283 ASSERT(get_itbl(target)->type == TSO);
3285 if (target->what_next == ThreadRelocated) {
3286 target = target->link;
3287 ASSERT(get_itbl(target)->type == TSO);
3290 ASSERT(target->blocked_exceptions != NULL);
3292 last = (StgBlockingQueueElement **)&target->blocked_exceptions;
3293 for (t = (StgBlockingQueueElement *)target->blocked_exceptions;
3295 last = &t->link, t = t->link) {
3296 ASSERT(get_itbl(t)->type == TSO);
3297 if (t == (StgBlockingQueueElement *)tso) {
3298 *last = (StgBlockingQueueElement *)tso->link;
3302 barf("unblockThread (Exception): TSO not found");
3306 case BlockedOnWrite:
3307 #if defined(mingw32_HOST_OS)
3308 case BlockedOnDoProc:
3311 /* take TSO off blocked_queue */
3312 StgBlockingQueueElement *prev = NULL;
3313 for (t = (StgBlockingQueueElement *)blocked_queue_hd; t != END_BQ_QUEUE;
3314 prev = t, t = t->link) {
3315 if (t == (StgBlockingQueueElement *)tso) {
3317 blocked_queue_hd = (StgTSO *)t->link;
3318 if ((StgBlockingQueueElement *)blocked_queue_tl == t) {
3319 blocked_queue_tl = END_TSO_QUEUE;
3322 prev->link = t->link;
3323 if ((StgBlockingQueueElement *)blocked_queue_tl == t) {
3324 blocked_queue_tl = (StgTSO *)prev;
3327 #if defined(mingw32_HOST_OS)
3328 /* (Cooperatively) signal that the worker thread should abort
3331 abandonWorkRequest(tso->block_info.async_result->reqID);
3336 barf("unblockThread (I/O): TSO not found");
3339 case BlockedOnDelay:
3341 /* take TSO off sleeping_queue */
3342 StgBlockingQueueElement *prev = NULL;
3343 for (t = (StgBlockingQueueElement *)sleeping_queue; t != END_BQ_QUEUE;
3344 prev = t, t = t->link) {
3345 if (t == (StgBlockingQueueElement *)tso) {
3347 sleeping_queue = (StgTSO *)t->link;
3349 prev->link = t->link;
3354 barf("unblockThread (delay): TSO not found");
3358 barf("unblockThread");
3362 tso->link = END_TSO_QUEUE;
3363 tso->why_blocked = NotBlocked;
3364 tso->block_info.closure = NULL;
3365 pushOnRunQueue(cap,tso);
3369 unblockThread(Capability *cap, StgTSO *tso)
3373 /* To avoid locking unnecessarily. */
3374 if (tso->why_blocked == NotBlocked) {
3378 switch (tso->why_blocked) {
3381 // Be careful: nothing to do here! We tell the scheduler that the thread
3382 // is runnable and we leave it to the stack-walking code to abort the
3383 // transaction while unwinding the stack. We should perhaps have a debugging
3384 // test to make sure that this really happens and that the 'zombie' transaction
3385 // does not get committed.
3389 ASSERT(get_itbl(tso->block_info.closure)->type == MVAR);
3391 StgTSO *last_tso = END_TSO_QUEUE;
3392 StgMVar *mvar = (StgMVar *)(tso->block_info.closure);
3395 for (t = mvar->head; t != END_TSO_QUEUE;
3396 last = &t->link, last_tso = t, t = t->link) {
3399 if (mvar->tail == tso) {
3400 mvar->tail = last_tso;
3405 barf("unblockThread (MVAR): TSO not found");
3408 case BlockedOnBlackHole:
3410 last = &blackhole_queue;
3411 for (t = blackhole_queue; t != END_TSO_QUEUE;
3412 last = &t->link, t = t->link) {
3418 barf("unblockThread (BLACKHOLE): TSO not found");
3421 case BlockedOnException:
3423 StgTSO *target = tso->block_info.tso;
3425 ASSERT(get_itbl(target)->type == TSO);
3427 while (target->what_next == ThreadRelocated) {
3428 target = target->link;
3429 ASSERT(get_itbl(target)->type == TSO);
3432 ASSERT(target->blocked_exceptions != NULL);
3434 last = &target->blocked_exceptions;
3435 for (t = target->blocked_exceptions; t != END_TSO_QUEUE;
3436 last = &t->link, t = t->link) {
3437 ASSERT(get_itbl(t)->type == TSO);
3443 barf("unblockThread (Exception): TSO not found");
3446 #if !defined(THREADED_RTS)
3448 case BlockedOnWrite:
3449 #if defined(mingw32_HOST_OS)
3450 case BlockedOnDoProc:
3453 StgTSO *prev = NULL;
3454 for (t = blocked_queue_hd; t != END_TSO_QUEUE;
3455 prev = t, t = t->link) {
3458 blocked_queue_hd = t->link;
3459 if (blocked_queue_tl == t) {
3460 blocked_queue_tl = END_TSO_QUEUE;
3463 prev->link = t->link;
3464 if (blocked_queue_tl == t) {
3465 blocked_queue_tl = prev;
3468 #if defined(mingw32_HOST_OS)
3469 /* (Cooperatively) signal that the worker thread should abort
3472 abandonWorkRequest(tso->block_info.async_result->reqID);
3477 barf("unblockThread (I/O): TSO not found");
3480 case BlockedOnDelay:
3482 StgTSO *prev = NULL;
3483 for (t = sleeping_queue; t != END_TSO_QUEUE;
3484 prev = t, t = t->link) {
3487 sleeping_queue = t->link;
3489 prev->link = t->link;
3494 barf("unblockThread (delay): TSO not found");
3499 barf("unblockThread");
3503 tso->link = END_TSO_QUEUE;
3504 tso->why_blocked = NotBlocked;
3505 tso->block_info.closure = NULL;
3506 appendToRunQueue(cap,tso);
3510 /* -----------------------------------------------------------------------------
3513 * Check the blackhole_queue for threads that can be woken up. We do
3514 * this periodically: before every GC, and whenever the run queue is
3517 * An elegant solution might be to just wake up all the blocked
3518 * threads with awakenBlockedQueue occasionally: they'll go back to
3519 * sleep again if the object is still a BLACKHOLE. Unfortunately this
3520 * doesn't give us a way to tell whether we've actually managed to
3521 * wake up any threads, so we would be busy-waiting.
3523 * -------------------------------------------------------------------------- */
3526 checkBlackHoles (Capability *cap)
3529 rtsBool any_woke_up = rtsFalse;
3532 // blackhole_queue is global:
3533 ASSERT_LOCK_HELD(&sched_mutex);
3535 IF_DEBUG(scheduler, sched_belch("checking threads blocked on black holes"));
3537 // ASSUMES: sched_mutex
3538 prev = &blackhole_queue;
3539 t = blackhole_queue;
3540 while (t != END_TSO_QUEUE) {
3541 ASSERT(t->why_blocked == BlockedOnBlackHole);
3542 type = get_itbl(t->block_info.closure)->type;
3543 if (type != BLACKHOLE && type != CAF_BLACKHOLE) {
3544 IF_DEBUG(sanity,checkTSO(t));
3545 t = unblockOne(cap, t);
3546 // urk, the threads migrate to the current capability
3547 // here, but we'd like to keep them on the original one.
3549 any_woke_up = rtsTrue;
3559 /* -----------------------------------------------------------------------------
3562 * The following function implements the magic for raising an
3563 * asynchronous exception in an existing thread.
3565 * We first remove the thread from any queue on which it might be
3566 * blocked. The possible blockages are MVARs and BLACKHOLE_BQs.
3568 * We strip the stack down to the innermost CATCH_FRAME, building
3569 * thunks in the heap for all the active computations, so they can
3570 * be restarted if necessary. When we reach a CATCH_FRAME, we build
3571 * an application of the handler to the exception, and push it on
3572 * the top of the stack.
3574 * How exactly do we save all the active computations? We create an
3575 * AP_STACK for every UpdateFrame on the stack. Entering one of these
3576 * AP_STACKs pushes everything from the corresponding update frame
3577 * upwards onto the stack. (Actually, it pushes everything up to the
3578 * next update frame plus a pointer to the next AP_STACK object.
3579 * Entering the next AP_STACK object pushes more onto the stack until we
3580 * reach the last AP_STACK object - at which point the stack should look
3581 * exactly as it did when we killed the TSO and we can continue
3582 * execution by entering the closure on top of the stack.
3584 * We can also kill a thread entirely - this happens if either (a) the
3585 * exception passed to raiseAsync is NULL, or (b) there's no
3586 * CATCH_FRAME on the stack. In either case, we strip the entire
3587 * stack and replace the thread with a zombie.
3589 * ToDo: in SMP mode, this function is only safe if either (a) we hold
3590 * all the Capabilities (eg. in GC), or (b) we own the Capability that
3591 * the TSO is currently blocked on or on the run queue of.
3593 * -------------------------------------------------------------------------- */
3596 raiseAsync(Capability *cap, StgTSO *tso, StgClosure *exception)
3598 raiseAsync_(cap, tso, exception, rtsFalse);
3602 raiseAsync_(Capability *cap, StgTSO *tso, StgClosure *exception,
3603 rtsBool stop_at_atomically)
3605 StgRetInfoTable *info;
3608 // Thread already dead?
3609 if (tso->what_next == ThreadComplete || tso->what_next == ThreadKilled) {
3614 sched_belch("raising exception in thread %ld.", (long)tso->id));
3616 // Remove it from any blocking queues
3617 unblockThread(cap,tso);
3621 // The stack freezing code assumes there's a closure pointer on
3622 // the top of the stack, so we have to arrange that this is the case...
3624 if (sp[0] == (W_)&stg_enter_info) {
3628 sp[0] = (W_)&stg_dummy_ret_closure;
3634 // 1. Let the top of the stack be the "current closure"
3636 // 2. Walk up the stack until we find either an UPDATE_FRAME or a
3639 // 3. If it's an UPDATE_FRAME, then make an AP_STACK containing the
3640 // current closure applied to the chunk of stack up to (but not
3641 // including) the update frame. This closure becomes the "current
3642 // closure". Go back to step 2.
3644 // 4. If it's a CATCH_FRAME, then leave the exception handler on
3645 // top of the stack applied to the exception.
3647 // 5. If it's a STOP_FRAME, then kill the thread.
3649 // NB: if we pass an ATOMICALLY_FRAME then abort the associated
3656 info = get_ret_itbl((StgClosure *)frame);
3658 while (info->i.type != UPDATE_FRAME
3659 && (info->i.type != CATCH_FRAME || exception == NULL)
3660 && info->i.type != STOP_FRAME
3661 && (info->i.type != ATOMICALLY_FRAME || stop_at_atomically == rtsFalse))
3663 if (info->i.type == CATCH_RETRY_FRAME || info->i.type == ATOMICALLY_FRAME) {
3664 // IF we find an ATOMICALLY_FRAME then we abort the
3665 // current transaction and propagate the exception. In
3666 // this case (unlike ordinary exceptions) we do not care
3667 // whether the transaction is valid or not because its
3668 // possible validity cannot have caused the exception
3669 // and will not be visible after the abort.
3671 debugBelch("Found atomically block delivering async exception\n"));
3672 stmAbortTransaction(tso -> trec);
3673 tso -> trec = stmGetEnclosingTRec(tso -> trec);
3675 frame += stack_frame_sizeW((StgClosure *)frame);
3676 info = get_ret_itbl((StgClosure *)frame);
3679 switch (info->i.type) {
3681 case ATOMICALLY_FRAME:
3682 ASSERT(stop_at_atomically);
3683 ASSERT(stmGetEnclosingTRec(tso->trec) == NO_TREC);
3684 stmCondemnTransaction(tso -> trec);
3688 // R1 is not a register: the return convention for IO in
3689 // this case puts the return value on the stack, so we
3690 // need to set up the stack to return to the atomically
3691 // frame properly...
3692 tso->sp = frame - 2;
3693 tso->sp[1] = (StgWord) &stg_NO_FINALIZER_closure; // why not?
3694 tso->sp[0] = (StgWord) &stg_ut_1_0_unreg_info;
3696 tso->what_next = ThreadRunGHC;
3700 // If we find a CATCH_FRAME, and we've got an exception to raise,
3701 // then build the THUNK raise(exception), and leave it on
3702 // top of the CATCH_FRAME ready to enter.
3706 StgCatchFrame *cf = (StgCatchFrame *)frame;
3710 // we've got an exception to raise, so let's pass it to the
3711 // handler in this frame.
3713 raise = (StgThunk *)allocateLocal(cap,sizeofW(StgThunk)+MIN_UPD_SIZE);
3714 TICK_ALLOC_SE_THK(1,0);
3715 SET_HDR(raise,&stg_raise_info,cf->header.prof.ccs);
3716 raise->payload[0] = exception;
3718 // throw away the stack from Sp up to the CATCH_FRAME.
3722 /* Ensure that async excpetions are blocked now, so we don't get
3723 * a surprise exception before we get around to executing the
3726 if (tso->blocked_exceptions == NULL) {
3727 tso->blocked_exceptions = END_TSO_QUEUE;
3730 /* Put the newly-built THUNK on top of the stack, ready to execute
3731 * when the thread restarts.
3734 sp[-1] = (W_)&stg_enter_info;
3736 tso->what_next = ThreadRunGHC;
3737 IF_DEBUG(sanity, checkTSO(tso));
3746 // First build an AP_STACK consisting of the stack chunk above the
3747 // current update frame, with the top word on the stack as the
3750 words = frame - sp - 1;
3751 ap = (StgAP_STACK *)allocateLocal(cap,AP_STACK_sizeW(words));
3754 ap->fun = (StgClosure *)sp[0];
3756 for(i=0; i < (nat)words; ++i) {
3757 ap->payload[i] = (StgClosure *)*sp++;
3760 SET_HDR(ap,&stg_AP_STACK_info,
3761 ((StgClosure *)frame)->header.prof.ccs /* ToDo */);
3762 TICK_ALLOC_UP_THK(words+1,0);
3765 debugBelch("sched: Updating ");
3766 printPtr((P_)((StgUpdateFrame *)frame)->updatee);
3767 debugBelch(" with ");
3768 printObj((StgClosure *)ap);
3771 // Replace the updatee with an indirection - happily
3772 // this will also wake up any threads currently
3773 // waiting on the result.
3775 // Warning: if we're in a loop, more than one update frame on
3776 // the stack may point to the same object. Be careful not to
3777 // overwrite an IND_OLDGEN in this case, because we'll screw
3778 // up the mutable lists. To be on the safe side, don't
3779 // overwrite any kind of indirection at all. See also
3780 // threadSqueezeStack in GC.c, where we have to make a similar
3783 if (!closure_IND(((StgUpdateFrame *)frame)->updatee)) {
3784 // revert the black hole
3785 UPD_IND_NOLOCK(((StgUpdateFrame *)frame)->updatee,
3788 sp += sizeofW(StgUpdateFrame) - 1;
3789 sp[0] = (W_)ap; // push onto stack
3794 // We've stripped the entire stack, the thread is now dead.
3795 sp += sizeofW(StgStopFrame);
3796 tso->what_next = ThreadKilled;
3807 /* -----------------------------------------------------------------------------
3810 This is used for interruption (^C) and forking, and corresponds to
3811 raising an exception but without letting the thread catch the
3813 -------------------------------------------------------------------------- */
3816 deleteThread (Capability *cap, StgTSO *tso)
3818 if (tso->why_blocked != BlockedOnCCall &&
3819 tso->why_blocked != BlockedOnCCall_NoUnblockExc) {
3820 raiseAsync(cap,tso,NULL);
3824 #ifdef FORKPROCESS_PRIMOP_SUPPORTED
3826 deleteThreadImmediately(Capability *cap, StgTSO *tso)
3827 { // for forkProcess only:
3828 // delete thread without giving it a chance to catch the KillThread exception
3830 if (tso->what_next == ThreadComplete || tso->what_next == ThreadKilled) {
3834 if (tso->why_blocked != BlockedOnCCall &&
3835 tso->why_blocked != BlockedOnCCall_NoUnblockExc) {
3836 unblockThread(cap,tso);
3839 tso->what_next = ThreadKilled;
3843 /* -----------------------------------------------------------------------------
3844 raiseExceptionHelper
3846 This function is called by the raise# primitve, just so that we can
3847 move some of the tricky bits of raising an exception from C-- into
3848 C. Who knows, it might be a useful re-useable thing here too.
3849 -------------------------------------------------------------------------- */
3852 raiseExceptionHelper (StgRegTable *reg, StgTSO *tso, StgClosure *exception)
3854 Capability *cap = regTableToCapability(reg);
3855 StgThunk *raise_closure = NULL;
3857 StgRetInfoTable *info;
3859 // This closure represents the expression 'raise# E' where E
3860 // is the exception raise. It is used to overwrite all the
3861 // thunks which are currently under evaluataion.
3865 // LDV profiling: stg_raise_info has THUNK as its closure
3866 // type. Since a THUNK takes at least MIN_UPD_SIZE words in its
3867 // payload, MIN_UPD_SIZE is more approprate than 1. It seems that
3868 // 1 does not cause any problem unless profiling is performed.
3869 // However, when LDV profiling goes on, we need to linearly scan
3870 // small object pool, where raise_closure is stored, so we should
3871 // use MIN_UPD_SIZE.
3873 // raise_closure = (StgClosure *)RET_STGCALL1(P_,allocate,
3874 // sizeofW(StgClosure)+1);
3878 // Walk up the stack, looking for the catch frame. On the way,
3879 // we update any closures pointed to from update frames with the
3880 // raise closure that we just built.
3884 info = get_ret_itbl((StgClosure *)p);
3885 next = p + stack_frame_sizeW((StgClosure *)p);
3886 switch (info->i.type) {
3889 // Only create raise_closure if we need to.
3890 if (raise_closure == NULL) {
3892 (StgThunk *)allocateLocal(cap,sizeofW(StgThunk)+MIN_UPD_SIZE);
3893 SET_HDR(raise_closure, &stg_raise_info, CCCS);
3894 raise_closure->payload[0] = exception;
3896 UPD_IND(((StgUpdateFrame *)p)->updatee,(StgClosure *)raise_closure);
3900 case ATOMICALLY_FRAME:
3901 IF_DEBUG(stm, debugBelch("Found ATOMICALLY_FRAME at %p\n", p));
3903 return ATOMICALLY_FRAME;
3909 case CATCH_STM_FRAME:
3910 IF_DEBUG(stm, debugBelch("Found CATCH_STM_FRAME at %p\n", p));
3912 return CATCH_STM_FRAME;
3918 case CATCH_RETRY_FRAME:
3927 /* -----------------------------------------------------------------------------
3928 findRetryFrameHelper
3930 This function is called by the retry# primitive. It traverses the stack
3931 leaving tso->sp referring to the frame which should handle the retry.
3933 This should either be a CATCH_RETRY_FRAME (if the retry# is within an orElse#)
3934 or should be a ATOMICALLY_FRAME (if the retry# reaches the top level).
3936 We skip CATCH_STM_FRAMEs because retries are not considered to be exceptions,
3937 despite the similar implementation.
3939 We should not expect to see CATCH_FRAME or STOP_FRAME because those should
3940 not be created within memory transactions.
3941 -------------------------------------------------------------------------- */
3944 findRetryFrameHelper (StgTSO *tso)
3947 StgRetInfoTable *info;
3951 info = get_ret_itbl((StgClosure *)p);
3952 next = p + stack_frame_sizeW((StgClosure *)p);
3953 switch (info->i.type) {
3955 case ATOMICALLY_FRAME:
3956 IF_DEBUG(stm, debugBelch("Found ATOMICALLY_FRAME at %p during retrry\n", p));
3958 return ATOMICALLY_FRAME;
3960 case CATCH_RETRY_FRAME:
3961 IF_DEBUG(stm, debugBelch("Found CATCH_RETRY_FRAME at %p during retrry\n", p));
3963 return CATCH_RETRY_FRAME;
3965 case CATCH_STM_FRAME:
3967 ASSERT(info->i.type != CATCH_FRAME);
3968 ASSERT(info->i.type != STOP_FRAME);
3975 /* -----------------------------------------------------------------------------
3976 resurrectThreads is called after garbage collection on the list of
3977 threads found to be garbage. Each of these threads will be woken
3978 up and sent a signal: BlockedOnDeadMVar if the thread was blocked
3979 on an MVar, or NonTermination if the thread was blocked on a Black
3982 Locks: assumes we hold *all* the capabilities.
3983 -------------------------------------------------------------------------- */
3986 resurrectThreads (StgTSO *threads)
3991 for (tso = threads; tso != END_TSO_QUEUE; tso = next) {
3992 next = tso->global_link;
3993 tso->global_link = all_threads;
3995 IF_DEBUG(scheduler, sched_belch("resurrecting thread %d", tso->id));
3997 // Wake up the thread on the Capability it was last on for a
3998 // bound thread, or last_free_capability otherwise.
4000 cap = tso->bound->cap;
4002 cap = last_free_capability;
4005 switch (tso->why_blocked) {
4007 case BlockedOnException:
4008 /* Called by GC - sched_mutex lock is currently held. */
4009 raiseAsync(cap, tso,(StgClosure *)BlockedOnDeadMVar_closure);
4011 case BlockedOnBlackHole:
4012 raiseAsync(cap, tso,(StgClosure *)NonTermination_closure);
4015 raiseAsync(cap, tso,(StgClosure *)BlockedIndefinitely_closure);
4018 /* This might happen if the thread was blocked on a black hole
4019 * belonging to a thread that we've just woken up (raiseAsync
4020 * can wake up threads, remember...).
4024 barf("resurrectThreads: thread blocked in a strange way");
4029 /* ----------------------------------------------------------------------------
4030 * Debugging: why is a thread blocked
4031 * [Also provides useful information when debugging threaded programs
4032 * at the Haskell source code level, so enable outside of DEBUG. --sof 7/02]
4033 ------------------------------------------------------------------------- */
4037 printThreadBlockage(StgTSO *tso)
4039 switch (tso->why_blocked) {
4041 debugBelch("is blocked on read from fd %d", (int)(tso->block_info.fd));
4043 case BlockedOnWrite:
4044 debugBelch("is blocked on write to fd %d", (int)(tso->block_info.fd));
4046 #if defined(mingw32_HOST_OS)
4047 case BlockedOnDoProc:
4048 debugBelch("is blocked on proc (request: %ld)", tso->block_info.async_result->reqID);
4051 case BlockedOnDelay:
4052 debugBelch("is blocked until %ld", (long)(tso->block_info.target));
4055 debugBelch("is blocked on an MVar @ %p", tso->block_info.closure);
4057 case BlockedOnException:
4058 debugBelch("is blocked on delivering an exception to thread %d",
4059 tso->block_info.tso->id);
4061 case BlockedOnBlackHole:
4062 debugBelch("is blocked on a black hole");
4065 debugBelch("is not blocked");
4067 #if defined(PARALLEL_HASKELL)
4069 debugBelch("is blocked on global address; local FM_BQ is %p (%s)",
4070 tso->block_info.closure, info_type(tso->block_info.closure));
4072 case BlockedOnGA_NoSend:
4073 debugBelch("is blocked on global address (no send); local FM_BQ is %p (%s)",
4074 tso->block_info.closure, info_type(tso->block_info.closure));
4077 case BlockedOnCCall:
4078 debugBelch("is blocked on an external call");
4080 case BlockedOnCCall_NoUnblockExc:
4081 debugBelch("is blocked on an external call (exceptions were already blocked)");
4084 debugBelch("is blocked on an STM operation");
4087 barf("printThreadBlockage: strange tso->why_blocked: %d for TSO %d (%d)",
4088 tso->why_blocked, tso->id, tso);
4093 printThreadStatus(StgTSO *t)
4095 debugBelch("\tthread %4d @ %p ", t->id, (void *)t);
4097 void *label = lookupThreadLabel(t->id);
4098 if (label) debugBelch("[\"%s\"] ",(char *)label);
4100 if (t->what_next == ThreadRelocated) {
4101 debugBelch("has been relocated...\n");
4103 switch (t->what_next) {
4105 debugBelch("has been killed");
4107 case ThreadComplete:
4108 debugBelch("has completed");
4111 printThreadBlockage(t);
4118 printAllThreads(void)
4125 char time_string[TIME_STR_LEN], node_str[NODE_STR_LEN];
4126 ullong_format_string(TIME_ON_PROC(CurrentProc),
4127 time_string, rtsFalse/*no commas!*/);
4129 debugBelch("all threads at [%s]:\n", time_string);
4130 # elif defined(PARALLEL_HASKELL)
4131 char time_string[TIME_STR_LEN], node_str[NODE_STR_LEN];
4132 ullong_format_string(CURRENT_TIME,
4133 time_string, rtsFalse/*no commas!*/);
4135 debugBelch("all threads at [%s]:\n", time_string);
4137 debugBelch("all threads:\n");
4140 for (i = 0; i < n_capabilities; i++) {
4141 cap = &capabilities[i];
4142 debugBelch("threads on capability %d:\n", cap->no);
4143 for (t = cap->run_queue_hd; t != END_TSO_QUEUE; t = t->link) {
4144 printThreadStatus(t);
4148 for (t = all_threads; t != END_TSO_QUEUE; t = next) {
4149 if (t->why_blocked != NotBlocked) {
4150 printThreadStatus(t);
4152 if (t->what_next == ThreadRelocated) {
4155 next = t->global_link;
4162 printThreadQueue(StgTSO *t)
4165 for (; t != END_TSO_QUEUE; t = t->link) {
4166 printThreadStatus(t);
4169 debugBelch("%d threads on queue\n", i);
4173 Print a whole blocking queue attached to node (debugging only).
4175 # if defined(PARALLEL_HASKELL)
4177 print_bq (StgClosure *node)
4179 StgBlockingQueueElement *bqe;
4183 debugBelch("## BQ of closure %p (%s): ",
4184 node, info_type(node));
4186 /* should cover all closures that may have a blocking queue */
4187 ASSERT(get_itbl(node)->type == BLACKHOLE_BQ ||
4188 get_itbl(node)->type == FETCH_ME_BQ ||
4189 get_itbl(node)->type == RBH ||
4190 get_itbl(node)->type == MVAR);
4192 ASSERT(node!=(StgClosure*)NULL); // sanity check
4194 print_bqe(((StgBlockingQueue*)node)->blocking_queue);
4198 Print a whole blocking queue starting with the element bqe.
4201 print_bqe (StgBlockingQueueElement *bqe)
4206 NB: In a parallel setup a BQ of an RBH must end with an RBH_Save closure;
4208 for (end = (bqe==END_BQ_QUEUE);
4209 !end; // iterate until bqe points to a CONSTR
4210 end = (get_itbl(bqe)->type == CONSTR) || (bqe->link==END_BQ_QUEUE),
4211 bqe = end ? END_BQ_QUEUE : bqe->link) {
4212 ASSERT(bqe != END_BQ_QUEUE); // sanity check
4213 ASSERT(bqe != (StgBlockingQueueElement *)NULL); // sanity check
4214 /* types of closures that may appear in a blocking queue */
4215 ASSERT(get_itbl(bqe)->type == TSO ||
4216 get_itbl(bqe)->type == BLOCKED_FETCH ||
4217 get_itbl(bqe)->type == CONSTR);
4218 /* only BQs of an RBH end with an RBH_Save closure */
4219 //ASSERT(get_itbl(bqe)->type != CONSTR || get_itbl(node)->type == RBH);
4221 switch (get_itbl(bqe)->type) {
4223 debugBelch(" TSO %u (%x),",
4224 ((StgTSO *)bqe)->id, ((StgTSO *)bqe));
4227 debugBelch(" BF (node=%p, ga=((%x, %d, %x)),",
4228 ((StgBlockedFetch *)bqe)->node,
4229 ((StgBlockedFetch *)bqe)->ga.payload.gc.gtid,
4230 ((StgBlockedFetch *)bqe)->ga.payload.gc.slot,
4231 ((StgBlockedFetch *)bqe)->ga.weight);
4234 debugBelch(" %s (IP %p),",
4235 (get_itbl(bqe) == &stg_RBH_Save_0_info ? "RBH_Save_0" :
4236 get_itbl(bqe) == &stg_RBH_Save_1_info ? "RBH_Save_1" :
4237 get_itbl(bqe) == &stg_RBH_Save_2_info ? "RBH_Save_2" :
4238 "RBH_Save_?"), get_itbl(bqe));
4241 barf("Unexpected closure type %s in blocking queue", // of %p (%s)",
4242 info_type((StgClosure *)bqe)); // , node, info_type(node));
4248 # elif defined(GRAN)
4250 print_bq (StgClosure *node)
4252 StgBlockingQueueElement *bqe;
4253 PEs node_loc, tso_loc;
4256 /* should cover all closures that may have a blocking queue */
4257 ASSERT(get_itbl(node)->type == BLACKHOLE_BQ ||
4258 get_itbl(node)->type == FETCH_ME_BQ ||
4259 get_itbl(node)->type == RBH);
4261 ASSERT(node!=(StgClosure*)NULL); // sanity check
4262 node_loc = where_is(node);
4264 debugBelch("## BQ of closure %p (%s) on [PE %d]: ",
4265 node, info_type(node), node_loc);
4268 NB: In a parallel setup a BQ of an RBH must end with an RBH_Save closure;
4270 for (bqe = ((StgBlockingQueue*)node)->blocking_queue, end = (bqe==END_BQ_QUEUE);
4271 !end; // iterate until bqe points to a CONSTR
4272 end = (get_itbl(bqe)->type == CONSTR) || (bqe->link==END_BQ_QUEUE), bqe = end ? END_BQ_QUEUE : bqe->link) {
4273 ASSERT(bqe != END_BQ_QUEUE); // sanity check
4274 ASSERT(bqe != (StgBlockingQueueElement *)NULL); // sanity check
4275 /* types of closures that may appear in a blocking queue */
4276 ASSERT(get_itbl(bqe)->type == TSO ||
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 tso_loc = where_is((StgClosure *)bqe);
4282 switch (get_itbl(bqe)->type) {
4284 debugBelch(" TSO %d (%p) on [PE %d],",
4285 ((StgTSO *)bqe)->id, (StgTSO *)bqe, tso_loc);
4288 debugBelch(" %s (IP %p),",
4289 (get_itbl(bqe) == &stg_RBH_Save_0_info ? "RBH_Save_0" :
4290 get_itbl(bqe) == &stg_RBH_Save_1_info ? "RBH_Save_1" :
4291 get_itbl(bqe) == &stg_RBH_Save_2_info ? "RBH_Save_2" :
4292 "RBH_Save_?"), get_itbl(bqe));
4295 barf("Unexpected closure type %s in blocking queue of %p (%s)",
4296 info_type((StgClosure *)bqe), node, info_type(node));
4304 #if defined(PARALLEL_HASKELL)
4311 for (i=0, tso=run_queue_hd;
4312 tso != END_TSO_QUEUE;
4313 i++, tso=tso->link) {
4322 sched_belch(char *s, ...)
4327 debugBelch("sched (task %p): ", (void *)(unsigned long)(unsigned int)osThreadId());
4328 #elif defined(PARALLEL_HASKELL)
4331 debugBelch("sched: ");