1 /* ---------------------------------------------------------------------------
3 * (c) The GHC Team, 1998-2004
7 * Different GHC ways use this scheduler quite differently (see comments below)
8 * Here is the global picture:
10 * WAY Name CPP flag What's it for
11 * --------------------------------------
12 * mp GUM PAR Parallel execution on a distrib. memory machine
13 * s SMP SMP Parallel execution on a shared memory machine
14 * mg GranSim GRAN Simulation of parallel execution
15 * md GUM/GdH DIST Distributed execution (based on GUM)
17 * --------------------------------------------------------------------------*/
20 * Version with support for distributed memory parallelism aka GUM (WAY=mp):
22 The main scheduling loop in GUM iterates until a finish message is received.
23 In that case a global flag @receivedFinish@ is set and this instance of
24 the RTS shuts down. See ghc/rts/parallel/HLComms.c:processMessages()
25 for the handling of incoming messages, such as PP_FINISH.
26 Note that in the parallel case we have a system manager that coordinates
27 different PEs, each of which are running one instance of the RTS.
28 See ghc/rts/parallel/SysMan.c for the main routine of the parallel program.
29 From this routine processes executing ghc/rts/Main.c are spawned. -- HWL
31 * Version with support for simulating parallel execution aka GranSim (WAY=mg):
33 The main scheduling code in GranSim is quite different from that in std
34 (concurrent) Haskell: while concurrent Haskell just iterates over the
35 threads in the runnable queue, GranSim is event driven, i.e. it iterates
36 over the events in the global event queue. -- HWL
39 #include "PosixSource.h"
44 #include "BlockAlloc.h"
48 #define COMPILING_SCHEDULER
50 #include "StgMiscClosures.h"
52 #include "Interpreter.h"
53 #include "Exception.h"
61 #include "ThreadLabels.h"
62 #include "LdvProfile.h"
65 #include "Proftimer.h"
68 #if defined(GRAN) || defined(PAR)
69 # include "GranSimRts.h"
71 # include "ParallelRts.h"
72 # include "Parallel.h"
73 # include "ParallelDebug.h"
78 #include "Capability.h"
79 #include "OSThreads.h"
82 #ifdef HAVE_SYS_TYPES_H
83 #include <sys/types.h>
98 #define USED_IN_THREADED_RTS
100 #define USED_IN_THREADED_RTS STG_UNUSED
103 #ifdef RTS_SUPPORTS_THREADS
104 #define USED_WHEN_RTS_SUPPORTS_THREADS
106 #define USED_WHEN_RTS_SUPPORTS_THREADS STG_UNUSED
109 /* Main thread queue.
110 * Locks required: sched_mutex.
112 StgMainThread *main_threads = NULL;
115 * Locks required: sched_mutex.
119 StgTSO* ActiveTSO = NULL; /* for assigning system costs; GranSim-Light only */
120 /* rtsTime TimeOfNextEvent, EndOfTimeSlice; now in GranSim.c */
123 In GranSim we have a runnable and a blocked queue for each processor.
124 In order to minimise code changes new arrays run_queue_hds/tls
125 are created. run_queue_hd is then a short cut (macro) for
126 run_queue_hds[CurrentProc] (see GranSim.h).
129 StgTSO *run_queue_hds[MAX_PROC], *run_queue_tls[MAX_PROC];
130 StgTSO *blocked_queue_hds[MAX_PROC], *blocked_queue_tls[MAX_PROC];
131 StgTSO *ccalling_threadss[MAX_PROC];
132 /* We use the same global list of threads (all_threads) in GranSim as in
133 the std RTS (i.e. we are cheating). However, we don't use this list in
134 the GranSim specific code at the moment (so we are only potentially
139 StgTSO *run_queue_hd = NULL;
140 StgTSO *run_queue_tl = NULL;
141 StgTSO *blocked_queue_hd = NULL;
142 StgTSO *blocked_queue_tl = NULL;
143 StgTSO *sleeping_queue = NULL; /* perhaps replace with a hash table? */
147 /* Linked list of all threads.
148 * Used for detecting garbage collected threads.
150 StgTSO *all_threads = NULL;
152 /* When a thread performs a safe C call (_ccall_GC, using old
153 * terminology), it gets put on the suspended_ccalling_threads
154 * list. Used by the garbage collector.
156 static StgTSO *suspended_ccalling_threads;
158 static StgTSO *threadStackOverflow(StgTSO *tso);
160 /* KH: The following two flags are shared memory locations. There is no need
161 to lock them, since they are only unset at the end of a scheduler
165 /* flag set by signal handler to precipitate a context switch */
166 int context_switch = 0;
168 /* if this flag is set as well, give up execution */
169 rtsBool interrupted = rtsFalse;
171 /* Next thread ID to allocate.
172 * Locks required: thread_id_mutex
174 static StgThreadID next_thread_id = 1;
177 * Pointers to the state of the current thread.
178 * Rule of thumb: if CurrentTSO != NULL, then we're running a Haskell
179 * thread. If CurrentTSO == NULL, then we're at the scheduler level.
182 /* The smallest stack size that makes any sense is:
183 * RESERVED_STACK_WORDS (so we can get back from the stack overflow)
184 * + sizeofW(StgStopFrame) (the stg_stop_thread_info frame)
185 * + 1 (the closure to enter)
187 * + 1 (spare slot req'd by stg_ap_v_ret)
189 * A thread with this stack will bomb immediately with a stack
190 * overflow, which will increase its stack size.
193 #define MIN_STACK_WORDS (RESERVED_STACK_WORDS + sizeofW(StgStopFrame) + 3)
200 /* This is used in `TSO.h' and gcc 2.96 insists that this variable actually
201 * exists - earlier gccs apparently didn't.
206 static rtsBool ready_to_gc;
209 * Set to TRUE when entering a shutdown state (via shutdownHaskellAndExit()) --
210 * in an MT setting, needed to signal that a worker thread shouldn't hang around
211 * in the scheduler when it is out of work.
213 static rtsBool shutting_down_scheduler = rtsFalse;
215 void addToBlockedQueue ( StgTSO *tso );
217 static void schedule ( StgMainThread *mainThread, Capability *initialCapability );
218 void interruptStgRts ( void );
220 #if !defined(PAR) && !defined(RTS_SUPPORTS_THREADS)
221 static void detectBlackHoles ( void );
224 static void raiseAsync_(StgTSO *tso, StgClosure *exception, rtsBool stop_at_atomically);
226 #if defined(RTS_SUPPORTS_THREADS)
227 /* ToDo: carefully document the invariants that go together
228 * with these synchronisation objects.
230 Mutex sched_mutex = INIT_MUTEX_VAR;
231 Mutex term_mutex = INIT_MUTEX_VAR;
233 #endif /* RTS_SUPPORTS_THREADS */
237 rtsTime TimeOfLastYield;
238 rtsBool emitSchedule = rtsTrue;
242 static char *whatNext_strs[] = {
253 StgTSO * createSparkThread(rtsSpark spark);
254 StgTSO * activateSpark (rtsSpark spark);
257 /* ----------------------------------------------------------------------------
259 * ------------------------------------------------------------------------- */
261 #if defined(RTS_SUPPORTS_THREADS)
262 static rtsBool startingWorkerThread = rtsFalse;
264 static void taskStart(void);
268 ACQUIRE_LOCK(&sched_mutex);
269 startingWorkerThread = rtsFalse;
271 RELEASE_LOCK(&sched_mutex);
275 startSchedulerTaskIfNecessary(void)
277 if(run_queue_hd != END_TSO_QUEUE
278 || blocked_queue_hd != END_TSO_QUEUE
279 || sleeping_queue != END_TSO_QUEUE)
281 if(!startingWorkerThread)
282 { // we don't want to start another worker thread
283 // just because the last one hasn't yet reached the
284 // "waiting for capability" state
285 startingWorkerThread = rtsTrue;
286 if(!startTask(taskStart))
288 startingWorkerThread = rtsFalse;
295 /* ---------------------------------------------------------------------------
296 Main scheduling loop.
298 We use round-robin scheduling, each thread returning to the
299 scheduler loop when one of these conditions is detected:
302 * timer expires (thread yields)
307 Locking notes: we acquire the scheduler lock once at the beginning
308 of the scheduler loop, and release it when
310 * running a thread, or
311 * waiting for work, or
312 * waiting for a GC to complete.
315 In a GranSim setup this loop iterates over the global event queue.
316 This revolves around the global event queue, which determines what
317 to do next. Therefore, it's more complicated than either the
318 concurrent or the parallel (GUM) setup.
321 GUM iterates over incoming messages.
322 It starts with nothing to do (thus CurrentTSO == END_TSO_QUEUE),
323 and sends out a fish whenever it has nothing to do; in-between
324 doing the actual reductions (shared code below) it processes the
325 incoming messages and deals with delayed operations
326 (see PendingFetches).
327 This is not the ugliest code you could imagine, but it's bloody close.
329 ------------------------------------------------------------------------ */
331 schedule( StgMainThread *mainThread USED_WHEN_RTS_SUPPORTS_THREADS,
332 Capability *initialCapability )
336 StgThreadReturnCode ret;
344 rtsBool receivedFinish = rtsFalse;
346 nat tp_size, sp_size; // stats only
349 rtsBool was_interrupted = rtsFalse;
352 // Pre-condition: sched_mutex is held.
353 // We might have a capability, passed in as initialCapability.
354 cap = initialCapability;
356 #if defined(RTS_SUPPORTS_THREADS)
358 // in the threaded case, the capability is either passed in via the
359 // initialCapability parameter, or initialized inside the scheduler
363 sched_belch("### NEW SCHEDULER LOOP (main thr: %p, cap: %p)",
364 mainThread, initialCapability);
367 // simply initialise it in the non-threaded case
368 grabCapability(&cap);
372 /* set up first event to get things going */
373 /* ToDo: assign costs for system setup and init MainTSO ! */
374 new_event(CurrentProc, CurrentProc, CurrentTime[CurrentProc],
376 CurrentTSO, (StgClosure*)NULL, (rtsSpark*)NULL);
379 debugBelch("GRAN: Init CurrentTSO (in schedule) = %p\n", CurrentTSO);
380 G_TSO(CurrentTSO, 5));
382 if (RtsFlags.GranFlags.Light) {
383 /* Save current time; GranSim Light only */
384 CurrentTSO->gran.clock = CurrentTime[CurrentProc];
387 event = get_next_event();
389 while (event!=(rtsEvent*)NULL) {
390 /* Choose the processor with the next event */
391 CurrentProc = event->proc;
392 CurrentTSO = event->tso;
396 while (!receivedFinish) { /* set by processMessages */
397 /* when receiving PP_FINISH message */
399 #else // everything except GRAN and PAR
405 IF_DEBUG(scheduler, printAllThreads());
407 #if defined(RTS_SUPPORTS_THREADS)
408 // Yield the capability to higher-priority tasks if necessary.
411 yieldCapability(&cap);
414 // If we do not currently hold a capability, we wait for one
417 waitForCapability(&sched_mutex, &cap,
418 mainThread ? &mainThread->bound_thread_cond : NULL);
421 // We now have a capability...
425 // If we're interrupted (the user pressed ^C, or some other
426 // termination condition occurred), kill all the currently running
430 IF_DEBUG(scheduler, sched_belch("interrupted"));
431 interrupted = rtsFalse;
432 was_interrupted = rtsTrue;
433 #if defined(RTS_SUPPORTS_THREADS)
434 // In the threaded RTS, deadlock detection doesn't work,
435 // so just exit right away.
436 errorBelch("interrupted");
437 releaseCapability(cap);
438 RELEASE_LOCK(&sched_mutex);
439 shutdownHaskellAndExit(EXIT_SUCCESS);
445 #if defined(RTS_USER_SIGNALS)
446 // check for signals each time around the scheduler
447 if (signals_pending()) {
448 RELEASE_LOCK(&sched_mutex); /* ToDo: kill */
449 startSignalHandlers();
450 ACQUIRE_LOCK(&sched_mutex);
455 // Check whether any waiting threads need to be woken up. If the
456 // run queue is empty, and there are no other tasks running, we
457 // can wait indefinitely for something to happen.
459 if ( !EMPTY_QUEUE(blocked_queue_hd) || !EMPTY_QUEUE(sleeping_queue) )
461 #if defined(RTS_SUPPORTS_THREADS)
462 // We shouldn't be here...
463 barf("schedule: awaitEvent() in threaded RTS");
465 awaitEvent( EMPTY_RUN_QUEUE() );
467 // we can be interrupted while waiting for I/O...
468 if (interrupted) continue;
471 * Detect deadlock: when we have no threads to run, there are no
472 * threads waiting on I/O or sleeping, and all the other tasks are
473 * waiting for work, we must have a deadlock of some description.
475 * We first try to find threads blocked on themselves (ie. black
476 * holes), and generate NonTermination exceptions where necessary.
478 * If no threads are black holed, we have a deadlock situation, so
479 * inform all the main threads.
481 #if !defined(PAR) && !defined(RTS_SUPPORTS_THREADS)
482 if ( EMPTY_THREAD_QUEUES() )
484 IF_DEBUG(scheduler, sched_belch("deadlocked, forcing major GC..."));
486 // Garbage collection can release some new threads due to
487 // either (a) finalizers or (b) threads resurrected because
488 // they are unreachable and will therefore be sent an
489 // exception. Any threads thus released will be immediately
491 GarbageCollect(GetRoots,rtsTrue);
492 if ( !EMPTY_RUN_QUEUE() ) { goto not_deadlocked; }
494 #if defined(RTS_USER_SIGNALS)
495 /* If we have user-installed signal handlers, then wait
496 * for signals to arrive rather then bombing out with a
499 if ( anyUserHandlers() ) {
501 sched_belch("still deadlocked, waiting for signals..."));
505 // we might be interrupted...
506 if (interrupted) { continue; }
508 if (signals_pending()) {
509 RELEASE_LOCK(&sched_mutex);
510 startSignalHandlers();
511 ACQUIRE_LOCK(&sched_mutex);
513 ASSERT(!EMPTY_RUN_QUEUE());
518 /* Probably a real deadlock. Send the current main thread the
519 * Deadlock exception (or in the SMP build, send *all* main
520 * threads the deadlock exception, since none of them can make
526 switch (m->tso->why_blocked) {
527 case BlockedOnBlackHole:
528 case BlockedOnException:
530 raiseAsync(m->tso, (StgClosure *)NonTermination_closure);
533 barf("deadlock: main thread blocked in a strange way");
539 #elif defined(RTS_SUPPORTS_THREADS)
540 // ToDo: add deadlock detection in threaded RTS
542 // ToDo: add deadlock detection in GUM (similar to SMP) -- HWL
545 #if defined(RTS_SUPPORTS_THREADS) || defined(mingw32_HOST_OS)
546 /* win32: might be back here due to awaitEvent() being abandoned
547 * as a result of a console event having been delivered.
549 if ( EMPTY_RUN_QUEUE() ) {
550 continue; // nothing to do
555 if (RtsFlags.GranFlags.Light)
556 GranSimLight_enter_system(event, &ActiveTSO); // adjust ActiveTSO etc
558 /* adjust time based on time-stamp */
559 if (event->time > CurrentTime[CurrentProc] &&
560 event->evttype != ContinueThread)
561 CurrentTime[CurrentProc] = event->time;
563 /* Deal with the idle PEs (may issue FindWork or MoveSpark events) */
564 if (!RtsFlags.GranFlags.Light)
567 IF_DEBUG(gran, debugBelch("GRAN: switch by event-type\n"));
569 /* main event dispatcher in GranSim */
570 switch (event->evttype) {
571 /* Should just be continuing execution */
573 IF_DEBUG(gran, debugBelch("GRAN: doing ContinueThread\n"));
574 /* ToDo: check assertion
575 ASSERT(run_queue_hd != (StgTSO*)NULL &&
576 run_queue_hd != END_TSO_QUEUE);
578 /* Ignore ContinueThreads for fetching threads (if synchr comm) */
579 if (!RtsFlags.GranFlags.DoAsyncFetch &&
580 procStatus[CurrentProc]==Fetching) {
581 debugBelch("ghuH: Spurious ContinueThread while Fetching ignored; TSO %d (%p) [PE %d]\n",
582 CurrentTSO->id, CurrentTSO, CurrentProc);
585 /* Ignore ContinueThreads for completed threads */
586 if (CurrentTSO->what_next == ThreadComplete) {
587 debugBelch("ghuH: found a ContinueThread event for completed thread %d (%p) [PE %d] (ignoring ContinueThread)\n",
588 CurrentTSO->id, CurrentTSO, CurrentProc);
591 /* Ignore ContinueThreads for threads that are being migrated */
592 if (PROCS(CurrentTSO)==Nowhere) {
593 debugBelch("ghuH: trying to run the migrating TSO %d (%p) [PE %d] (ignoring ContinueThread)\n",
594 CurrentTSO->id, CurrentTSO, CurrentProc);
597 /* The thread should be at the beginning of the run queue */
598 if (CurrentTSO!=run_queue_hds[CurrentProc]) {
599 debugBelch("ghuH: TSO %d (%p) [PE %d] is not at the start of the run_queue when doing a ContinueThread\n",
600 CurrentTSO->id, CurrentTSO, CurrentProc);
601 break; // run the thread anyway
604 new_event(proc, proc, CurrentTime[proc],
606 (StgTSO*)NULL, (StgClosure*)NULL, (rtsSpark*)NULL);
608 */ /* Catches superfluous CONTINUEs -- should be unnecessary */
609 break; // now actually run the thread; DaH Qu'vam yImuHbej
612 do_the_fetchnode(event);
613 goto next_thread; /* handle next event in event queue */
616 do_the_globalblock(event);
617 goto next_thread; /* handle next event in event queue */
620 do_the_fetchreply(event);
621 goto next_thread; /* handle next event in event queue */
623 case UnblockThread: /* Move from the blocked queue to the tail of */
624 do_the_unblock(event);
625 goto next_thread; /* handle next event in event queue */
627 case ResumeThread: /* Move from the blocked queue to the tail of */
628 /* the runnable queue ( i.e. Qu' SImqa'lu') */
629 event->tso->gran.blocktime +=
630 CurrentTime[CurrentProc] - event->tso->gran.blockedat;
631 do_the_startthread(event);
632 goto next_thread; /* handle next event in event queue */
635 do_the_startthread(event);
636 goto next_thread; /* handle next event in event queue */
639 do_the_movethread(event);
640 goto next_thread; /* handle next event in event queue */
643 do_the_movespark(event);
644 goto next_thread; /* handle next event in event queue */
647 do_the_findwork(event);
648 goto next_thread; /* handle next event in event queue */
651 barf("Illegal event type %u\n", event->evttype);
654 /* This point was scheduler_loop in the old RTS */
656 IF_DEBUG(gran, debugBelch("GRAN: after main switch\n"));
658 TimeOfLastEvent = CurrentTime[CurrentProc];
659 TimeOfNextEvent = get_time_of_next_event();
660 IgnoreEvents=(TimeOfNextEvent==0); // HWL HACK
661 // CurrentTSO = ThreadQueueHd;
663 IF_DEBUG(gran, debugBelch("GRAN: time of next event is: %ld\n",
666 if (RtsFlags.GranFlags.Light)
667 GranSimLight_leave_system(event, &ActiveTSO);
669 EndOfTimeSlice = CurrentTime[CurrentProc]+RtsFlags.GranFlags.time_slice;
672 debugBelch("GRAN: end of time-slice is %#lx\n", EndOfTimeSlice));
674 /* in a GranSim setup the TSO stays on the run queue */
676 /* Take a thread from the run queue. */
677 POP_RUN_QUEUE(t); // take_off_run_queue(t);
680 debugBelch("GRAN: About to run current thread, which is\n");
683 context_switch = 0; // turned on via GranYield, checking events and time slice
686 DumpGranEvent(GR_SCHEDULE, t));
688 procStatus[CurrentProc] = Busy;
691 if (PendingFetches != END_BF_QUEUE) {
695 /* ToDo: phps merge with spark activation above */
696 /* check whether we have local work and send requests if we have none */
697 if (EMPTY_RUN_QUEUE()) { /* no runnable threads */
698 /* :-[ no local threads => look out for local sparks */
699 /* the spark pool for the current PE */
700 pool = &(MainRegTable.rSparks); // generalise to cap = &MainRegTable
701 if (advisory_thread_count < RtsFlags.ParFlags.maxThreads &&
702 pool->hd < pool->tl) {
704 * ToDo: add GC code check that we really have enough heap afterwards!!
706 * If we're here (no runnable threads) and we have pending
707 * sparks, we must have a space problem. Get enough space
708 * to turn one of those pending sparks into a
712 spark = findSpark(rtsFalse); /* get a spark */
713 if (spark != (rtsSpark) NULL) {
714 tso = activateSpark(spark); /* turn the spark into a thread */
715 IF_PAR_DEBUG(schedule,
716 debugBelch("==== schedule: Created TSO %d (%p); %d threads active\n",
717 tso->id, tso, advisory_thread_count));
719 if (tso==END_TSO_QUEUE) { /* failed to activate spark->back to loop */
720 debugBelch("==^^ failed to activate spark\n");
722 } /* otherwise fall through & pick-up new tso */
724 IF_PAR_DEBUG(verbose,
725 debugBelch("==^^ no local sparks (spark pool contains only NFs: %d)\n",
726 spark_queue_len(pool)));
731 /* If we still have no work we need to send a FISH to get a spark
734 if (EMPTY_RUN_QUEUE()) {
735 /* =8-[ no local sparks => look for work on other PEs */
737 * We really have absolutely no work. Send out a fish
738 * (there may be some out there already), and wait for
739 * something to arrive. We clearly can't run any threads
740 * until a SCHEDULE or RESUME arrives, and so that's what
741 * we're hoping to see. (Of course, we still have to
742 * respond to other types of messages.)
744 TIME now = msTime() /*CURRENT_TIME*/;
745 IF_PAR_DEBUG(verbose,
746 debugBelch("-- now=%ld\n", now));
747 IF_PAR_DEBUG(verbose,
748 if (outstandingFishes < RtsFlags.ParFlags.maxFishes &&
749 (last_fish_arrived_at!=0 &&
750 last_fish_arrived_at+RtsFlags.ParFlags.fishDelay > now)) {
751 debugBelch("--$$ delaying FISH until %ld (last fish %ld, delay %ld, now %ld)\n",
752 last_fish_arrived_at+RtsFlags.ParFlags.fishDelay,
753 last_fish_arrived_at,
754 RtsFlags.ParFlags.fishDelay, now);
757 if (outstandingFishes < RtsFlags.ParFlags.maxFishes &&
758 (last_fish_arrived_at==0 ||
759 (last_fish_arrived_at+RtsFlags.ParFlags.fishDelay <= now))) {
760 /* outstandingFishes is set in sendFish, processFish;
761 avoid flooding system with fishes via delay */
763 sendFish(pe, mytid, NEW_FISH_AGE, NEW_FISH_HISTORY,
766 // Global statistics: count no. of fishes
767 if (RtsFlags.ParFlags.ParStats.Global &&
768 RtsFlags.GcFlags.giveStats > NO_GC_STATS) {
769 globalParStats.tot_fish_mess++;
773 receivedFinish = processMessages();
776 } else if (PacketsWaiting()) { /* Look for incoming messages */
777 receivedFinish = processMessages();
780 /* Now we are sure that we have some work available */
781 ASSERT(run_queue_hd != END_TSO_QUEUE);
783 /* Take a thread from the run queue, if we have work */
784 POP_RUN_QUEUE(t); // take_off_run_queue(END_TSO_QUEUE);
785 IF_DEBUG(sanity,checkTSO(t));
787 /* ToDo: write something to the log-file
788 if (RTSflags.ParFlags.granSimStats && !sameThread)
789 DumpGranEvent(GR_SCHEDULE, RunnableThreadsHd);
793 /* the spark pool for the current PE */
794 pool = &(MainRegTable.rSparks); // generalise to cap = &MainRegTable
797 debugBelch("--=^ %d threads, %d sparks on [%#x]\n",
798 run_queue_len(), spark_queue_len(pool), CURRENT_PROC));
801 if (0 && RtsFlags.ParFlags.ParStats.Full &&
802 t && LastTSO && t->id != LastTSO->id &&
803 LastTSO->why_blocked == NotBlocked &&
804 LastTSO->what_next != ThreadComplete) {
805 // if previously scheduled TSO not blocked we have to record the context switch
806 DumpVeryRawGranEvent(TimeOfLastYield, CURRENT_PROC, CURRENT_PROC,
807 GR_DESCHEDULE, LastTSO, (StgClosure *)NULL, 0, 0);
810 if (RtsFlags.ParFlags.ParStats.Full &&
811 (emitSchedule /* forced emit */ ||
812 (t && LastTSO && t->id != LastTSO->id))) {
814 we are running a different TSO, so write a schedule event to log file
815 NB: If we use fair scheduling we also have to write a deschedule
816 event for LastTSO; with unfair scheduling we know that the
817 previous tso has blocked whenever we switch to another tso, so
818 we don't need it in GUM for now
820 DumpRawGranEvent(CURRENT_PROC, CURRENT_PROC,
821 GR_SCHEDULE, t, (StgClosure *)NULL, 0, 0);
822 emitSchedule = rtsFalse;
826 #else /* !GRAN && !PAR */
828 // grab a thread from the run queue
829 ASSERT(run_queue_hd != END_TSO_QUEUE);
832 // Sanity check the thread we're about to run. This can be
833 // expensive if there is lots of thread switching going on...
834 IF_DEBUG(sanity,checkTSO(t));
839 StgMainThread *m = t->main;
846 sched_belch("### Running thread %d in bound thread", t->id));
847 // yes, the Haskell thread is bound to the current native thread
852 sched_belch("### thread %d bound to another OS thread", t->id));
853 // no, bound to a different Haskell thread: pass to that thread
854 PUSH_ON_RUN_QUEUE(t);
855 passCapability(&m->bound_thread_cond);
861 if(mainThread != NULL)
862 // The thread we want to run is bound.
865 sched_belch("### this OS thread cannot run thread %d", t->id));
866 // no, the current native thread is bound to a different
867 // Haskell thread, so pass it to any worker thread
868 PUSH_ON_RUN_QUEUE(t);
869 passCapabilityToWorker();
876 cap->r.rCurrentTSO = t;
878 /* context switches are now initiated by the timer signal, unless
879 * the user specified "context switch as often as possible", with
882 if ((RtsFlags.ConcFlags.ctxtSwitchTicks == 0
883 && (run_queue_hd != END_TSO_QUEUE
884 || blocked_queue_hd != END_TSO_QUEUE
885 || sleeping_queue != END_TSO_QUEUE)))
890 RELEASE_LOCK(&sched_mutex);
892 IF_DEBUG(scheduler, sched_belch("-->> running thread %ld %s ...",
893 (long)t->id, whatNext_strs[t->what_next]));
896 startHeapProfTimer();
899 /* +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ */
900 /* Run the current thread
902 prev_what_next = t->what_next;
904 errno = t->saved_errno;
906 switch (prev_what_next) {
910 /* Thread already finished, return to scheduler. */
911 ret = ThreadFinished;
915 ret = StgRun((StgFunPtr) stg_returnToStackTop, &cap->r);
918 case ThreadInterpret:
919 ret = interpretBCO(cap);
923 barf("schedule: invalid what_next field");
926 // The TSO might have moved, so find the new location:
927 t = cap->r.rCurrentTSO;
929 // And save the current errno in this thread.
930 t->saved_errno = errno;
932 /* +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ */
934 /* Costs for the scheduler are assigned to CCS_SYSTEM */
940 ACQUIRE_LOCK(&sched_mutex);
942 #ifdef RTS_SUPPORTS_THREADS
943 IF_DEBUG(scheduler,debugBelch("sched (task %p): ", osThreadId()););
944 #elif !defined(GRAN) && !defined(PAR)
945 IF_DEBUG(scheduler,debugBelch("sched: "););
949 /* HACK 675: if the last thread didn't yield, make sure to print a
950 SCHEDULE event to the log file when StgRunning the next thread, even
951 if it is the same one as before */
953 TimeOfLastYield = CURRENT_TIME;
959 IF_DEBUG(gran, DumpGranEvent(GR_DESCHEDULE, t));
960 globalGranStats.tot_heapover++;
962 globalParStats.tot_heapover++;
965 // did the task ask for a large block?
966 if (cap->r.rHpAlloc > BLOCK_SIZE) {
967 // if so, get one and push it on the front of the nursery.
971 blocks = (nat)BLOCK_ROUND_UP(cap->r.rHpAlloc) / BLOCK_SIZE;
973 IF_DEBUG(scheduler,debugBelch("--<< thread %ld (%s) stopped: requesting a large block (size %d)\n",
974 (long)t->id, whatNext_strs[t->what_next], blocks));
976 // don't do this if it would push us over the
977 // alloc_blocks_lim limit; we'll GC first.
978 if (alloc_blocks + blocks < alloc_blocks_lim) {
980 alloc_blocks += blocks;
981 bd = allocGroup( blocks );
983 // link the new group into the list
984 bd->link = cap->r.rCurrentNursery;
985 bd->u.back = cap->r.rCurrentNursery->u.back;
986 if (cap->r.rCurrentNursery->u.back != NULL) {
987 cap->r.rCurrentNursery->u.back->link = bd;
989 ASSERT(g0s0->blocks == cap->r.rCurrentNursery &&
990 g0s0->blocks == cap->r.rNursery);
991 cap->r.rNursery = g0s0->blocks = bd;
993 cap->r.rCurrentNursery->u.back = bd;
995 // initialise it as a nursery block. We initialise the
996 // step, gen_no, and flags field of *every* sub-block in
997 // this large block, because this is easier than making
998 // sure that we always find the block head of a large
999 // block whenever we call Bdescr() (eg. evacuate() and
1000 // isAlive() in the GC would both have to do this, at
1004 for (x = bd; x < bd + blocks; x++) {
1011 // don't forget to update the block count in g0s0.
1012 g0s0->n_blocks += blocks;
1013 // This assert can be a killer if the app is doing lots
1014 // of large block allocations.
1015 ASSERT(countBlocks(g0s0->blocks) == g0s0->n_blocks);
1017 // now update the nursery to point to the new block
1018 cap->r.rCurrentNursery = bd;
1020 // we might be unlucky and have another thread get on the
1021 // run queue before us and steal the large block, but in that
1022 // case the thread will just end up requesting another large
1024 PUSH_ON_RUN_QUEUE(t);
1029 /* make all the running tasks block on a condition variable,
1030 * maybe set context_switch and wait till they all pile in,
1031 * then have them wait on a GC condition variable.
1033 IF_DEBUG(scheduler,debugBelch("--<< thread %ld (%s) stopped: HeapOverflow\n",
1034 (long)t->id, whatNext_strs[t->what_next]));
1037 ASSERT(!is_on_queue(t,CurrentProc));
1039 /* Currently we emit a DESCHEDULE event before GC in GUM.
1040 ToDo: either add separate event to distinguish SYSTEM time from rest
1041 or just nuke this DESCHEDULE (and the following SCHEDULE) */
1042 if (0 && RtsFlags.ParFlags.ParStats.Full) {
1043 DumpRawGranEvent(CURRENT_PROC, CURRENT_PROC,
1044 GR_DESCHEDULE, t, (StgClosure *)NULL, 0, 0);
1045 emitSchedule = rtsTrue;
1049 ready_to_gc = rtsTrue;
1050 context_switch = 1; /* stop other threads ASAP */
1051 PUSH_ON_RUN_QUEUE(t);
1052 /* actual GC is done at the end of the while loop */
1058 DumpGranEvent(GR_DESCHEDULE, t));
1059 globalGranStats.tot_stackover++;
1062 // DumpGranEvent(GR_DESCHEDULE, t);
1063 globalParStats.tot_stackover++;
1065 IF_DEBUG(scheduler,debugBelch("--<< thread %ld (%s) stopped, StackOverflow\n",
1066 (long)t->id, whatNext_strs[t->what_next]));
1067 /* just adjust the stack for this thread, then pop it back
1072 /* enlarge the stack */
1073 StgTSO *new_t = threadStackOverflow(t);
1075 /* This TSO has moved, so update any pointers to it from the
1076 * main thread stack. It better not be on any other queues...
1077 * (it shouldn't be).
1079 if (t->main != NULL) {
1080 t->main->tso = new_t;
1082 PUSH_ON_RUN_QUEUE(new_t);
1086 case ThreadYielding:
1087 // Reset the context switch flag. We don't do this just before
1088 // running the thread, because that would mean we would lose ticks
1089 // during GC, which can lead to unfair scheduling (a thread hogs
1090 // the CPU because the tick always arrives during GC). This way
1091 // penalises threads that do a lot of allocation, but that seems
1092 // better than the alternative.
1097 DumpGranEvent(GR_DESCHEDULE, t));
1098 globalGranStats.tot_yields++;
1101 // DumpGranEvent(GR_DESCHEDULE, t);
1102 globalParStats.tot_yields++;
1104 /* put the thread back on the run queue. Then, if we're ready to
1105 * GC, check whether this is the last task to stop. If so, wake
1106 * up the GC thread. getThread will block during a GC until the
1110 if (t->what_next != prev_what_next) {
1111 debugBelch("--<< thread %ld (%s) stopped to switch evaluators\n",
1112 (long)t->id, whatNext_strs[t->what_next]);
1114 debugBelch("--<< thread %ld (%s) stopped, yielding\n",
1115 (long)t->id, whatNext_strs[t->what_next]);
1120 //debugBelch("&& Doing sanity check on yielding TSO %ld.", t->id);
1122 ASSERT(t->link == END_TSO_QUEUE);
1124 // Shortcut if we're just switching evaluators: don't bother
1125 // doing stack squeezing (which can be expensive), just run the
1127 if (t->what_next != prev_what_next) {
1134 ASSERT(!is_on_queue(t,CurrentProc));
1137 //debugBelch("&& Doing sanity check on all ThreadQueues (and their TSOs).");
1138 checkThreadQsSanity(rtsTrue));
1142 if (RtsFlags.ParFlags.doFairScheduling) {
1143 /* this does round-robin scheduling; good for concurrency */
1144 APPEND_TO_RUN_QUEUE(t);
1146 /* this does unfair scheduling; good for parallelism */
1147 PUSH_ON_RUN_QUEUE(t);
1150 // this does round-robin scheduling; good for concurrency
1151 APPEND_TO_RUN_QUEUE(t);
1155 /* add a ContinueThread event to actually process the thread */
1156 new_event(CurrentProc, CurrentProc, CurrentTime[CurrentProc],
1158 t, (StgClosure*)NULL, (rtsSpark*)NULL);
1160 debugBelch("GRAN: eventq and runnableq after adding yielded thread to queue again:\n");
1169 debugBelch("--<< thread %ld (%p; %s) stopped, blocking on node %p [PE %d] with BQ: \n",
1170 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)));
1171 if (t->block_info.closure!=(StgClosure*)NULL) print_bq(t->block_info.closure));
1173 // ??? needed; should emit block before
1175 DumpGranEvent(GR_DESCHEDULE, t));
1176 prune_eventq(t, (StgClosure *)NULL); // prune ContinueThreads for t
1179 ASSERT(procStatus[CurrentProc]==Busy ||
1180 ((procStatus[CurrentProc]==Fetching) &&
1181 (t->block_info.closure!=(StgClosure*)NULL)));
1182 if (run_queue_hds[CurrentProc] == END_TSO_QUEUE &&
1183 !(!RtsFlags.GranFlags.DoAsyncFetch &&
1184 procStatus[CurrentProc]==Fetching))
1185 procStatus[CurrentProc] = Idle;
1189 debugBelch("--<< thread %ld (%p; %s) stopped, blocking on node %p with BQ: \n",
1190 t->id, t, whatNext_strs[t->what_next], t->block_info.closure));
1193 if (t->block_info.closure!=(StgClosure*)NULL)
1194 print_bq(t->block_info.closure));
1196 /* Send a fetch (if BlockedOnGA) and dump event to log file */
1199 /* whatever we schedule next, we must log that schedule */
1200 emitSchedule = rtsTrue;
1203 /* don't need to do anything. Either the thread is blocked on
1204 * I/O, in which case we'll have called addToBlockedQueue
1205 * previously, or it's blocked on an MVar or Blackhole, in which
1206 * case it'll be on the relevant queue already.
1208 ASSERT(t->why_blocked != NotBlocked);
1210 debugBelch("--<< thread %d (%s) stopped: ",
1211 t->id, whatNext_strs[t->what_next]);
1212 printThreadBlockage(t);
1215 /* Only for dumping event to log file
1216 ToDo: do I need this in GranSim, too?
1223 case ThreadFinished:
1224 /* Need to check whether this was a main thread, and if so, signal
1225 * the task that started it with the return value. If we have no
1226 * more main threads, we probably need to stop all the tasks until
1229 /* We also end up here if the thread kills itself with an
1230 * uncaught exception, see Exception.hc.
1232 IF_DEBUG(scheduler,debugBelch("--++ thread %d (%s) finished\n",
1233 t->id, whatNext_strs[t->what_next]));
1235 endThread(t, CurrentProc); // clean-up the thread
1237 /* For now all are advisory -- HWL */
1238 //if(t->priority==AdvisoryPriority) ??
1239 advisory_thread_count--;
1242 if(t->dist.priority==RevalPriority)
1246 if (RtsFlags.ParFlags.ParStats.Full &&
1247 !RtsFlags.ParFlags.ParStats.Suppressed)
1248 DumpEndEvent(CURRENT_PROC, t, rtsFalse /* not mandatory */);
1252 // Check whether the thread that just completed was a main
1253 // thread, and if so return with the result.
1255 // There is an assumption here that all thread completion goes
1256 // through this point; we need to make sure that if a thread
1257 // ends up in the ThreadKilled state, that it stays on the run
1258 // queue so it can be dealt with here.
1261 #if defined(RTS_SUPPORTS_THREADS)
1264 mainThread->tso == t
1268 // We are a bound thread: this must be our thread that just
1270 ASSERT(mainThread->tso == t);
1272 if (t->what_next == ThreadComplete) {
1273 if (mainThread->ret) {
1274 // NOTE: return val is tso->sp[1] (see StgStartup.hc)
1275 *(mainThread->ret) = (StgClosure *)mainThread->tso->sp[1];
1277 mainThread->stat = Success;
1279 if (mainThread->ret) {
1280 *(mainThread->ret) = NULL;
1282 if (was_interrupted) {
1283 mainThread->stat = Interrupted;
1285 mainThread->stat = Killed;
1289 removeThreadLabel((StgWord)mainThread->tso->id);
1291 if (mainThread->prev == NULL) {
1292 main_threads = mainThread->link;
1294 mainThread->prev->link = mainThread->link;
1296 if (mainThread->link != NULL) {
1297 mainThread->link->prev = NULL;
1299 releaseCapability(cap);
1303 #ifdef RTS_SUPPORTS_THREADS
1304 ASSERT(t->main == NULL);
1306 if (t->main != NULL) {
1307 // Must be a main thread that is not the topmost one. Leave
1308 // it on the run queue until the stack has unwound to the
1309 // point where we can deal with this. Leaving it on the run
1310 // queue also ensures that the garbage collector knows about
1311 // this thread and its return value (it gets dropped from the
1312 // all_threads list so there's no other way to find it).
1313 APPEND_TO_RUN_QUEUE(t);
1319 barf("schedule: invalid thread return code %d", (int)ret);
1323 // When we have +RTS -i0 and we're heap profiling, do a census at
1324 // every GC. This lets us get repeatable runs for debugging.
1325 if (performHeapProfile ||
1326 (RtsFlags.ProfFlags.profileInterval==0 &&
1327 RtsFlags.ProfFlags.doHeapProfile && ready_to_gc)) {
1328 GarbageCollect(GetRoots, rtsTrue);
1330 performHeapProfile = rtsFalse;
1331 ready_to_gc = rtsFalse; // we already GC'd
1336 /* Kick any transactions which are invalid back to their atomically frames.
1337 * When next scheduled they will try to commit, this commit will fail and
1338 * they will retry. */
1339 for (t = all_threads; t != END_TSO_QUEUE; t = t -> link) {
1340 if (t -> what_next != ThreadRelocated && t -> trec != NO_TREC && t -> why_blocked == NotBlocked) {
1341 if (!stmValidateTransaction (t -> trec)) {
1342 IF_DEBUG(stm, sched_belch("trec %p found wasting its time", t));
1344 // strip the stack back to the ATOMICALLY_FRAME, aborting
1345 // the (nested) transaction, and saving the stack of any
1346 // partially-evaluated thunks on the heap.
1347 raiseAsync_(t, NULL, rtsTrue);
1350 ASSERT(get_itbl((StgClosure *)t->sp)->type == ATOMICALLY_FRAME);
1356 /* everybody back, start the GC.
1357 * Could do it in this thread, or signal a condition var
1358 * to do it in another thread. Either way, we need to
1359 * broadcast on gc_pending_cond afterward.
1361 #if defined(RTS_SUPPORTS_THREADS)
1362 IF_DEBUG(scheduler,sched_belch("doing GC"));
1364 GarbageCollect(GetRoots,rtsFalse);
1365 ready_to_gc = rtsFalse;
1367 /* add a ContinueThread event to continue execution of current thread */
1368 new_event(CurrentProc, CurrentProc, CurrentTime[CurrentProc],
1370 t, (StgClosure*)NULL, (rtsSpark*)NULL);
1372 debugBelch("GRAN: eventq and runnableq after Garbage collection:\n\n");
1380 IF_GRAN_DEBUG(unused,
1381 print_eventq(EventHd));
1383 event = get_next_event();
1386 /* ToDo: wait for next message to arrive rather than busy wait */
1389 } /* end of while(1) */
1391 IF_PAR_DEBUG(verbose,
1392 debugBelch("== Leaving schedule() after having received Finish\n"));
1395 /* ---------------------------------------------------------------------------
1396 * rtsSupportsBoundThreads(): is the RTS built to support bound threads?
1397 * used by Control.Concurrent for error checking.
1398 * ------------------------------------------------------------------------- */
1401 rtsSupportsBoundThreads(void)
1410 /* ---------------------------------------------------------------------------
1411 * isThreadBound(tso): check whether tso is bound to an OS thread.
1412 * ------------------------------------------------------------------------- */
1415 isThreadBound(StgTSO* tso USED_IN_THREADED_RTS)
1418 return (tso->main != NULL);
1423 /* ---------------------------------------------------------------------------
1424 * Singleton fork(). Do not copy any running threads.
1425 * ------------------------------------------------------------------------- */
1427 #ifndef mingw32_HOST_OS
1428 #define FORKPROCESS_PRIMOP_SUPPORTED
1431 #ifdef FORKPROCESS_PRIMOP_SUPPORTED
1433 deleteThreadImmediately(StgTSO *tso);
1436 forkProcess(HsStablePtr *entry
1437 #ifndef FORKPROCESS_PRIMOP_SUPPORTED
1442 #ifdef FORKPROCESS_PRIMOP_SUPPORTED
1448 IF_DEBUG(scheduler,sched_belch("forking!"));
1449 rts_lock(); // This not only acquires sched_mutex, it also
1450 // makes sure that no other threads are running
1454 if (pid) { /* parent */
1456 /* just return the pid */
1460 } else { /* child */
1463 // delete all threads
1464 run_queue_hd = run_queue_tl = END_TSO_QUEUE;
1466 for (t = all_threads; t != END_TSO_QUEUE; t = next) {
1469 // don't allow threads to catch the ThreadKilled exception
1470 deleteThreadImmediately(t);
1473 // wipe the main thread list
1474 while((m = main_threads) != NULL) {
1475 main_threads = m->link;
1476 # ifdef THREADED_RTS
1477 closeCondition(&m->bound_thread_cond);
1482 rc = rts_evalStableIO(entry, NULL); // run the action
1483 rts_checkSchedStatus("forkProcess",rc);
1487 hs_exit(); // clean up and exit
1490 #else /* !FORKPROCESS_PRIMOP_SUPPORTED */
1491 barf("forkProcess#: primop not supported, sorry!\n");
1496 /* ---------------------------------------------------------------------------
1497 * deleteAllThreads(): kill all the live threads.
1499 * This is used when we catch a user interrupt (^C), before performing
1500 * any necessary cleanups and running finalizers.
1502 * Locks: sched_mutex held.
1503 * ------------------------------------------------------------------------- */
1506 deleteAllThreads ( void )
1509 IF_DEBUG(scheduler,sched_belch("deleting all threads"));
1510 for (t = all_threads; t != END_TSO_QUEUE; t = next) {
1511 next = t->global_link;
1515 // The run queue now contains a bunch of ThreadKilled threads. We
1516 // must not throw these away: the main thread(s) will be in there
1517 // somewhere, and the main scheduler loop has to deal with it.
1518 // Also, the run queue is the only thing keeping these threads from
1519 // being GC'd, and we don't want the "main thread has been GC'd" panic.
1521 ASSERT(blocked_queue_hd == END_TSO_QUEUE);
1522 ASSERT(sleeping_queue == END_TSO_QUEUE);
1525 /* startThread and insertThread are now in GranSim.c -- HWL */
1528 /* ---------------------------------------------------------------------------
1529 * Suspending & resuming Haskell threads.
1531 * When making a "safe" call to C (aka _ccall_GC), the task gives back
1532 * its capability before calling the C function. This allows another
1533 * task to pick up the capability and carry on running Haskell
1534 * threads. It also means that if the C call blocks, it won't lock
1537 * The Haskell thread making the C call is put to sleep for the
1538 * duration of the call, on the susepended_ccalling_threads queue. We
1539 * give out a token to the task, which it can use to resume the thread
1540 * on return from the C function.
1541 * ------------------------------------------------------------------------- */
1544 suspendThread( StgRegTable *reg )
1548 int saved_errno = errno;
1550 /* assume that *reg is a pointer to the StgRegTable part
1553 cap = (Capability *)((void *)((unsigned char*)reg - sizeof(StgFunTable)));
1555 ACQUIRE_LOCK(&sched_mutex);
1558 sched_belch("thread %d did a _ccall_gc", cap->r.rCurrentTSO->id));
1560 // XXX this might not be necessary --SDM
1561 cap->r.rCurrentTSO->what_next = ThreadRunGHC;
1563 threadPaused(cap->r.rCurrentTSO);
1564 cap->r.rCurrentTSO->link = suspended_ccalling_threads;
1565 suspended_ccalling_threads = cap->r.rCurrentTSO;
1567 if(cap->r.rCurrentTSO->blocked_exceptions == NULL) {
1568 cap->r.rCurrentTSO->why_blocked = BlockedOnCCall;
1569 cap->r.rCurrentTSO->blocked_exceptions = END_TSO_QUEUE;
1571 cap->r.rCurrentTSO->why_blocked = BlockedOnCCall_NoUnblockExc;
1574 /* Use the thread ID as the token; it should be unique */
1575 tok = cap->r.rCurrentTSO->id;
1577 /* Hand back capability */
1578 releaseCapability(cap);
1580 #if defined(RTS_SUPPORTS_THREADS)
1581 /* Preparing to leave the RTS, so ensure there's a native thread/task
1582 waiting to take over.
1584 IF_DEBUG(scheduler, sched_belch("worker (token %d): leaving RTS", tok));
1587 RELEASE_LOCK(&sched_mutex);
1589 errno = saved_errno;
1594 resumeThread( StgInt tok )
1596 StgTSO *tso, **prev;
1598 int saved_errno = errno;
1600 #if defined(RTS_SUPPORTS_THREADS)
1601 /* Wait for permission to re-enter the RTS with the result. */
1602 ACQUIRE_LOCK(&sched_mutex);
1603 waitForReturnCapability(&sched_mutex, &cap);
1605 IF_DEBUG(scheduler, sched_belch("worker (token %d): re-entering RTS", tok));
1607 grabCapability(&cap);
1610 /* Remove the thread off of the suspended list */
1611 prev = &suspended_ccalling_threads;
1612 for (tso = suspended_ccalling_threads;
1613 tso != END_TSO_QUEUE;
1614 prev = &tso->link, tso = tso->link) {
1615 if (tso->id == (StgThreadID)tok) {
1620 if (tso == END_TSO_QUEUE) {
1621 barf("resumeThread: thread not found");
1623 tso->link = END_TSO_QUEUE;
1625 if(tso->why_blocked == BlockedOnCCall) {
1626 awakenBlockedQueueNoLock(tso->blocked_exceptions);
1627 tso->blocked_exceptions = NULL;
1630 /* Reset blocking status */
1631 tso->why_blocked = NotBlocked;
1633 cap->r.rCurrentTSO = tso;
1634 RELEASE_LOCK(&sched_mutex);
1635 errno = saved_errno;
1640 /* ---------------------------------------------------------------------------
1642 * ------------------------------------------------------------------------ */
1643 static void unblockThread(StgTSO *tso);
1645 /* ---------------------------------------------------------------------------
1646 * Comparing Thread ids.
1648 * This is used from STG land in the implementation of the
1649 * instances of Eq/Ord for ThreadIds.
1650 * ------------------------------------------------------------------------ */
1653 cmp_thread(StgPtr tso1, StgPtr tso2)
1655 StgThreadID id1 = ((StgTSO *)tso1)->id;
1656 StgThreadID id2 = ((StgTSO *)tso2)->id;
1658 if (id1 < id2) return (-1);
1659 if (id1 > id2) return 1;
1663 /* ---------------------------------------------------------------------------
1664 * Fetching the ThreadID from an StgTSO.
1666 * This is used in the implementation of Show for ThreadIds.
1667 * ------------------------------------------------------------------------ */
1669 rts_getThreadId(StgPtr tso)
1671 return ((StgTSO *)tso)->id;
1676 labelThread(StgPtr tso, char *label)
1681 /* Caveat: Once set, you can only set the thread name to "" */
1682 len = strlen(label)+1;
1683 buf = stgMallocBytes(len * sizeof(char), "Schedule.c:labelThread()");
1684 strncpy(buf,label,len);
1685 /* Update will free the old memory for us */
1686 updateThreadLabel(((StgTSO *)tso)->id,buf);
1690 /* ---------------------------------------------------------------------------
1691 Create a new thread.
1693 The new thread starts with the given stack size. Before the
1694 scheduler can run, however, this thread needs to have a closure
1695 (and possibly some arguments) pushed on its stack. See
1696 pushClosure() in Schedule.h.
1698 createGenThread() and createIOThread() (in SchedAPI.h) are
1699 convenient packaged versions of this function.
1701 currently pri (priority) is only used in a GRAN setup -- HWL
1702 ------------------------------------------------------------------------ */
1704 /* currently pri (priority) is only used in a GRAN setup -- HWL */
1706 createThread(nat size, StgInt pri)
1709 createThread(nat size)
1716 /* First check whether we should create a thread at all */
1718 /* check that no more than RtsFlags.ParFlags.maxThreads threads are created */
1719 if (advisory_thread_count >= RtsFlags.ParFlags.maxThreads) {
1721 debugBelch("{createThread}Daq ghuH: refusing to create another thread; no more than %d threads allowed (currently %d)\n",
1722 RtsFlags.ParFlags.maxThreads, advisory_thread_count);
1723 return END_TSO_QUEUE;
1729 ASSERT(!RtsFlags.GranFlags.Light || CurrentProc==0);
1732 // ToDo: check whether size = stack_size - TSO_STRUCT_SIZEW
1734 /* catch ridiculously small stack sizes */
1735 if (size < MIN_STACK_WORDS + TSO_STRUCT_SIZEW) {
1736 size = MIN_STACK_WORDS + TSO_STRUCT_SIZEW;
1739 stack_size = size - TSO_STRUCT_SIZEW;
1741 tso = (StgTSO *)allocate(size);
1742 TICK_ALLOC_TSO(stack_size, 0);
1744 SET_HDR(tso, &stg_TSO_info, CCS_SYSTEM);
1746 SET_GRAN_HDR(tso, ThisPE);
1749 // Always start with the compiled code evaluator
1750 tso->what_next = ThreadRunGHC;
1752 tso->id = next_thread_id++;
1753 tso->why_blocked = NotBlocked;
1754 tso->blocked_exceptions = NULL;
1756 tso->saved_errno = 0;
1759 tso->stack_size = stack_size;
1760 tso->max_stack_size = round_to_mblocks(RtsFlags.GcFlags.maxStkSize)
1762 tso->sp = (P_)&(tso->stack) + stack_size;
1764 tso->trec = NO_TREC;
1767 tso->prof.CCCS = CCS_MAIN;
1770 /* put a stop frame on the stack */
1771 tso->sp -= sizeofW(StgStopFrame);
1772 SET_HDR((StgClosure*)tso->sp,(StgInfoTable *)&stg_stop_thread_info,CCS_SYSTEM);
1773 tso->link = END_TSO_QUEUE;
1777 /* uses more flexible routine in GranSim */
1778 insertThread(tso, CurrentProc);
1780 /* In a non-GranSim setup the pushing of a TSO onto the runq is separated
1786 if (RtsFlags.GranFlags.GranSimStats.Full)
1787 DumpGranEvent(GR_START,tso);
1789 if (RtsFlags.ParFlags.ParStats.Full)
1790 DumpGranEvent(GR_STARTQ,tso);
1791 /* HACk to avoid SCHEDULE
1795 /* Link the new thread on the global thread list.
1797 tso->global_link = all_threads;
1801 tso->dist.priority = MandatoryPriority; //by default that is...
1805 tso->gran.pri = pri;
1807 tso->gran.magic = TSO_MAGIC; // debugging only
1809 tso->gran.sparkname = 0;
1810 tso->gran.startedat = CURRENT_TIME;
1811 tso->gran.exported = 0;
1812 tso->gran.basicblocks = 0;
1813 tso->gran.allocs = 0;
1814 tso->gran.exectime = 0;
1815 tso->gran.fetchtime = 0;
1816 tso->gran.fetchcount = 0;
1817 tso->gran.blocktime = 0;
1818 tso->gran.blockcount = 0;
1819 tso->gran.blockedat = 0;
1820 tso->gran.globalsparks = 0;
1821 tso->gran.localsparks = 0;
1822 if (RtsFlags.GranFlags.Light)
1823 tso->gran.clock = Now; /* local clock */
1825 tso->gran.clock = 0;
1827 IF_DEBUG(gran,printTSO(tso));
1830 tso->par.magic = TSO_MAGIC; // debugging only
1832 tso->par.sparkname = 0;
1833 tso->par.startedat = CURRENT_TIME;
1834 tso->par.exported = 0;
1835 tso->par.basicblocks = 0;
1836 tso->par.allocs = 0;
1837 tso->par.exectime = 0;
1838 tso->par.fetchtime = 0;
1839 tso->par.fetchcount = 0;
1840 tso->par.blocktime = 0;
1841 tso->par.blockcount = 0;
1842 tso->par.blockedat = 0;
1843 tso->par.globalsparks = 0;
1844 tso->par.localsparks = 0;
1848 globalGranStats.tot_threads_created++;
1849 globalGranStats.threads_created_on_PE[CurrentProc]++;
1850 globalGranStats.tot_sq_len += spark_queue_len(CurrentProc);
1851 globalGranStats.tot_sq_probes++;
1853 // collect parallel global statistics (currently done together with GC stats)
1854 if (RtsFlags.ParFlags.ParStats.Global &&
1855 RtsFlags.GcFlags.giveStats > NO_GC_STATS) {
1856 //debugBelch("Creating thread %d @ %11.2f\n", tso->id, usertime());
1857 globalParStats.tot_threads_created++;
1863 sched_belch("==__ schedule: Created TSO %d (%p);",
1864 CurrentProc, tso, tso->id));
1866 IF_PAR_DEBUG(verbose,
1867 sched_belch("==__ schedule: Created TSO %d (%p); %d threads active",
1868 (long)tso->id, tso, advisory_thread_count));
1870 IF_DEBUG(scheduler,sched_belch("created thread %ld, stack size = %lx words",
1871 (long)tso->id, (long)tso->stack_size));
1878 all parallel thread creation calls should fall through the following routine.
1881 createSparkThread(rtsSpark spark)
1883 ASSERT(spark != (rtsSpark)NULL);
1884 if (advisory_thread_count >= RtsFlags.ParFlags.maxThreads)
1886 barf("{createSparkThread}Daq ghuH: refusing to create another thread; no more than %d threads allowed (currently %d)",
1887 RtsFlags.ParFlags.maxThreads, advisory_thread_count);
1888 return END_TSO_QUEUE;
1892 tso = createThread(RtsFlags.GcFlags.initialStkSize);
1893 if (tso==END_TSO_QUEUE)
1894 barf("createSparkThread: Cannot create TSO");
1896 tso->priority = AdvisoryPriority;
1898 pushClosure(tso,spark);
1899 PUSH_ON_RUN_QUEUE(tso);
1900 advisory_thread_count++;
1907 Turn a spark into a thread.
1908 ToDo: fix for SMP (needs to acquire SCHED_MUTEX!)
1912 activateSpark (rtsSpark spark)
1916 tso = createSparkThread(spark);
1917 if (RtsFlags.ParFlags.ParStats.Full) {
1918 //ASSERT(run_queue_hd == END_TSO_QUEUE); // I think ...
1919 IF_PAR_DEBUG(verbose,
1920 debugBelch("==^^ activateSpark: turning spark of closure %p (%s) into a thread\n",
1921 (StgClosure *)spark, info_type((StgClosure *)spark)));
1923 // ToDo: fwd info on local/global spark to thread -- HWL
1924 // tso->gran.exported = spark->exported;
1925 // tso->gran.locked = !spark->global;
1926 // tso->gran.sparkname = spark->name;
1932 static SchedulerStatus waitThread_(/*out*/StgMainThread* m,
1933 Capability *initialCapability
1937 /* ---------------------------------------------------------------------------
1940 * scheduleThread puts a thread on the head of the runnable queue.
1941 * This will usually be done immediately after a thread is created.
1942 * The caller of scheduleThread must create the thread using e.g.
1943 * createThread and push an appropriate closure
1944 * on this thread's stack before the scheduler is invoked.
1945 * ------------------------------------------------------------------------ */
1947 static void scheduleThread_ (StgTSO* tso);
1950 scheduleThread_(StgTSO *tso)
1952 // The thread goes at the *end* of the run-queue, to avoid possible
1953 // starvation of any threads already on the queue.
1954 APPEND_TO_RUN_QUEUE(tso);
1959 scheduleThread(StgTSO* tso)
1961 ACQUIRE_LOCK(&sched_mutex);
1962 scheduleThread_(tso);
1963 RELEASE_LOCK(&sched_mutex);
1966 #if defined(RTS_SUPPORTS_THREADS)
1967 static Condition bound_cond_cache;
1968 static int bound_cond_cache_full = 0;
1973 scheduleWaitThread(StgTSO* tso, /*[out]*/HaskellObj* ret,
1974 Capability *initialCapability)
1976 // Precondition: sched_mutex must be held
1979 m = stgMallocBytes(sizeof(StgMainThread), "waitThread");
1984 m->link = main_threads;
1986 if (main_threads != NULL) {
1987 main_threads->prev = m;
1991 #if defined(RTS_SUPPORTS_THREADS)
1992 // Allocating a new condition for each thread is expensive, so we
1993 // cache one. This is a pretty feeble hack, but it helps speed up
1994 // consecutive call-ins quite a bit.
1995 if (bound_cond_cache_full) {
1996 m->bound_thread_cond = bound_cond_cache;
1997 bound_cond_cache_full = 0;
1999 initCondition(&m->bound_thread_cond);
2003 /* Put the thread on the main-threads list prior to scheduling the TSO.
2004 Failure to do so introduces a race condition in the MT case (as
2005 identified by Wolfgang Thaller), whereby the new task/OS thread
2006 created by scheduleThread_() would complete prior to the thread
2007 that spawned it managed to put 'itself' on the main-threads list.
2008 The upshot of it all being that the worker thread wouldn't get to
2009 signal the completion of the its work item for the main thread to
2010 see (==> it got stuck waiting.) -- sof 6/02.
2012 IF_DEBUG(scheduler, sched_belch("waiting for thread (%d)", tso->id));
2014 APPEND_TO_RUN_QUEUE(tso);
2015 // NB. Don't call threadRunnable() here, because the thread is
2016 // bound and only runnable by *this* OS thread, so waking up other
2017 // workers will just slow things down.
2019 return waitThread_(m, initialCapability);
2022 /* ---------------------------------------------------------------------------
2025 * Initialise the scheduler. This resets all the queues - if the
2026 * queues contained any threads, they'll be garbage collected at the
2029 * ------------------------------------------------------------------------ */
2037 for (i=0; i<=MAX_PROC; i++) {
2038 run_queue_hds[i] = END_TSO_QUEUE;
2039 run_queue_tls[i] = END_TSO_QUEUE;
2040 blocked_queue_hds[i] = END_TSO_QUEUE;
2041 blocked_queue_tls[i] = END_TSO_QUEUE;
2042 ccalling_threadss[i] = END_TSO_QUEUE;
2043 sleeping_queue = END_TSO_QUEUE;
2046 run_queue_hd = END_TSO_QUEUE;
2047 run_queue_tl = END_TSO_QUEUE;
2048 blocked_queue_hd = END_TSO_QUEUE;
2049 blocked_queue_tl = END_TSO_QUEUE;
2050 sleeping_queue = END_TSO_QUEUE;
2053 suspended_ccalling_threads = END_TSO_QUEUE;
2055 main_threads = NULL;
2056 all_threads = END_TSO_QUEUE;
2061 RtsFlags.ConcFlags.ctxtSwitchTicks =
2062 RtsFlags.ConcFlags.ctxtSwitchTime / TICK_MILLISECS;
2064 #if defined(RTS_SUPPORTS_THREADS)
2065 /* Initialise the mutex and condition variables used by
2067 initMutex(&sched_mutex);
2068 initMutex(&term_mutex);
2071 ACQUIRE_LOCK(&sched_mutex);
2073 /* A capability holds the state a native thread needs in
2074 * order to execute STG code. At least one capability is
2075 * floating around (only SMP builds have more than one).
2079 #if defined(RTS_SUPPORTS_THREADS)
2080 /* start our haskell execution tasks */
2081 startTaskManager(0,taskStart);
2084 #if /* defined(SMP) ||*/ defined(PAR)
2088 RELEASE_LOCK(&sched_mutex);
2092 exitScheduler( void )
2094 #if defined(RTS_SUPPORTS_THREADS)
2097 shutting_down_scheduler = rtsTrue;
2100 /* ----------------------------------------------------------------------------
2101 Managing the per-task allocation areas.
2103 Each capability comes with an allocation area. These are
2104 fixed-length block lists into which allocation can be done.
2106 ToDo: no support for two-space collection at the moment???
2107 ------------------------------------------------------------------------- */
2111 waitThread_(StgMainThread* m, Capability *initialCapability)
2113 SchedulerStatus stat;
2115 // Precondition: sched_mutex must be held.
2116 IF_DEBUG(scheduler, sched_belch("new main thread (%d)", m->tso->id));
2119 /* GranSim specific init */
2120 CurrentTSO = m->tso; // the TSO to run
2121 procStatus[MainProc] = Busy; // status of main PE
2122 CurrentProc = MainProc; // PE to run it on
2123 schedule(m,initialCapability);
2125 schedule(m,initialCapability);
2126 ASSERT(m->stat != NoStatus);
2131 #if defined(RTS_SUPPORTS_THREADS)
2132 // Free the condition variable, returning it to the cache if possible.
2133 if (!bound_cond_cache_full) {
2134 bound_cond_cache = m->bound_thread_cond;
2135 bound_cond_cache_full = 1;
2137 closeCondition(&m->bound_thread_cond);
2141 IF_DEBUG(scheduler, sched_belch("main thread (%d) finished", m->tso->id));
2144 // Postcondition: sched_mutex still held
2148 /* ---------------------------------------------------------------------------
2149 Where are the roots that we know about?
2151 - all the threads on the runnable queue
2152 - all the threads on the blocked queue
2153 - all the threads on the sleeping queue
2154 - all the thread currently executing a _ccall_GC
2155 - all the "main threads"
2157 ------------------------------------------------------------------------ */
2159 /* This has to be protected either by the scheduler monitor, or by the
2160 garbage collection monitor (probably the latter).
2165 GetRoots( evac_fn evac )
2170 for (i=0; i<=RtsFlags.GranFlags.proc; i++) {
2171 if ((run_queue_hds[i] != END_TSO_QUEUE) && ((run_queue_hds[i] != NULL)))
2172 evac((StgClosure **)&run_queue_hds[i]);
2173 if ((run_queue_tls[i] != END_TSO_QUEUE) && ((run_queue_tls[i] != NULL)))
2174 evac((StgClosure **)&run_queue_tls[i]);
2176 if ((blocked_queue_hds[i] != END_TSO_QUEUE) && ((blocked_queue_hds[i] != NULL)))
2177 evac((StgClosure **)&blocked_queue_hds[i]);
2178 if ((blocked_queue_tls[i] != END_TSO_QUEUE) && ((blocked_queue_tls[i] != NULL)))
2179 evac((StgClosure **)&blocked_queue_tls[i]);
2180 if ((ccalling_threadss[i] != END_TSO_QUEUE) && ((ccalling_threadss[i] != NULL)))
2181 evac((StgClosure **)&ccalling_threads[i]);
2188 if (run_queue_hd != END_TSO_QUEUE) {
2189 ASSERT(run_queue_tl != END_TSO_QUEUE);
2190 evac((StgClosure **)&run_queue_hd);
2191 evac((StgClosure **)&run_queue_tl);
2194 if (blocked_queue_hd != END_TSO_QUEUE) {
2195 ASSERT(blocked_queue_tl != END_TSO_QUEUE);
2196 evac((StgClosure **)&blocked_queue_hd);
2197 evac((StgClosure **)&blocked_queue_tl);
2200 if (sleeping_queue != END_TSO_QUEUE) {
2201 evac((StgClosure **)&sleeping_queue);
2205 if (suspended_ccalling_threads != END_TSO_QUEUE) {
2206 evac((StgClosure **)&suspended_ccalling_threads);
2209 #if defined(PAR) || defined(GRAN)
2210 markSparkQueue(evac);
2213 #if defined(RTS_USER_SIGNALS)
2214 // mark the signal handlers (signals should be already blocked)
2215 markSignalHandlers(evac);
2219 /* -----------------------------------------------------------------------------
2222 This is the interface to the garbage collector from Haskell land.
2223 We provide this so that external C code can allocate and garbage
2224 collect when called from Haskell via _ccall_GC.
2226 It might be useful to provide an interface whereby the programmer
2227 can specify more roots (ToDo).
2229 This needs to be protected by the GC condition variable above. KH.
2230 -------------------------------------------------------------------------- */
2232 static void (*extra_roots)(evac_fn);
2237 /* Obligated to hold this lock upon entry */
2238 ACQUIRE_LOCK(&sched_mutex);
2239 GarbageCollect(GetRoots,rtsFalse);
2240 RELEASE_LOCK(&sched_mutex);
2244 performMajorGC(void)
2246 ACQUIRE_LOCK(&sched_mutex);
2247 GarbageCollect(GetRoots,rtsTrue);
2248 RELEASE_LOCK(&sched_mutex);
2252 AllRoots(evac_fn evac)
2254 GetRoots(evac); // the scheduler's roots
2255 extra_roots(evac); // the user's roots
2259 performGCWithRoots(void (*get_roots)(evac_fn))
2261 ACQUIRE_LOCK(&sched_mutex);
2262 extra_roots = get_roots;
2263 GarbageCollect(AllRoots,rtsFalse);
2264 RELEASE_LOCK(&sched_mutex);
2267 /* -----------------------------------------------------------------------------
2270 If the thread has reached its maximum stack size, then raise the
2271 StackOverflow exception in the offending thread. Otherwise
2272 relocate the TSO into a larger chunk of memory and adjust its stack
2274 -------------------------------------------------------------------------- */
2277 threadStackOverflow(StgTSO *tso)
2279 nat new_stack_size, new_tso_size, stack_words;
2283 IF_DEBUG(sanity,checkTSO(tso));
2284 if (tso->stack_size >= tso->max_stack_size) {
2287 debugBelch("@@ threadStackOverflow of TSO %ld (%p): stack too large (now %ld; max is %ld)\n",
2288 (long)tso->id, tso, (long)tso->stack_size, (long)tso->max_stack_size);
2289 /* If we're debugging, just print out the top of the stack */
2290 printStackChunk(tso->sp, stg_min(tso->stack+tso->stack_size,
2293 /* Send this thread the StackOverflow exception */
2294 raiseAsync(tso, (StgClosure *)stackOverflow_closure);
2298 /* Try to double the current stack size. If that takes us over the
2299 * maximum stack size for this thread, then use the maximum instead.
2300 * Finally round up so the TSO ends up as a whole number of blocks.
2302 new_stack_size = stg_min(tso->stack_size * 2, tso->max_stack_size);
2303 new_tso_size = (nat)BLOCK_ROUND_UP(new_stack_size * sizeof(W_) +
2304 TSO_STRUCT_SIZE)/sizeof(W_);
2305 new_tso_size = round_to_mblocks(new_tso_size); /* Be MBLOCK-friendly */
2306 new_stack_size = new_tso_size - TSO_STRUCT_SIZEW;
2308 IF_DEBUG(scheduler, debugBelch("== sched: increasing stack size from %d words to %d.\n", tso->stack_size, new_stack_size));
2310 dest = (StgTSO *)allocate(new_tso_size);
2311 TICK_ALLOC_TSO(new_stack_size,0);
2313 /* copy the TSO block and the old stack into the new area */
2314 memcpy(dest,tso,TSO_STRUCT_SIZE);
2315 stack_words = tso->stack + tso->stack_size - tso->sp;
2316 new_sp = (P_)dest + new_tso_size - stack_words;
2317 memcpy(new_sp, tso->sp, stack_words * sizeof(W_));
2319 /* relocate the stack pointers... */
2321 dest->stack_size = new_stack_size;
2323 /* Mark the old TSO as relocated. We have to check for relocated
2324 * TSOs in the garbage collector and any primops that deal with TSOs.
2326 * It's important to set the sp value to just beyond the end
2327 * of the stack, so we don't attempt to scavenge any part of the
2330 tso->what_next = ThreadRelocated;
2332 tso->sp = (P_)&(tso->stack[tso->stack_size]);
2333 tso->why_blocked = NotBlocked;
2334 dest->mut_link = NULL;
2336 IF_PAR_DEBUG(verbose,
2337 debugBelch("@@ threadStackOverflow of TSO %d (now at %p): stack size increased to %ld\n",
2338 tso->id, tso, tso->stack_size);
2339 /* If we're debugging, just print out the top of the stack */
2340 printStackChunk(tso->sp, stg_min(tso->stack+tso->stack_size,
2343 IF_DEBUG(sanity,checkTSO(tso));
2345 IF_DEBUG(scheduler,printTSO(dest));
2351 /* ---------------------------------------------------------------------------
2352 Wake up a queue that was blocked on some resource.
2353 ------------------------------------------------------------------------ */
2357 unblockCount ( StgBlockingQueueElement *bqe, StgClosure *node )
2362 unblockCount ( StgBlockingQueueElement *bqe, StgClosure *node )
2364 /* write RESUME events to log file and
2365 update blocked and fetch time (depending on type of the orig closure) */
2366 if (RtsFlags.ParFlags.ParStats.Full) {
2367 DumpRawGranEvent(CURRENT_PROC, CURRENT_PROC,
2368 GR_RESUMEQ, ((StgTSO *)bqe), ((StgTSO *)bqe)->block_info.closure,
2369 0, 0 /* spark_queue_len(ADVISORY_POOL) */);
2370 if (EMPTY_RUN_QUEUE())
2371 emitSchedule = rtsTrue;
2373 switch (get_itbl(node)->type) {
2375 ((StgTSO *)bqe)->par.fetchtime += CURRENT_TIME-((StgTSO *)bqe)->par.blockedat;
2380 ((StgTSO *)bqe)->par.blocktime += CURRENT_TIME-((StgTSO *)bqe)->par.blockedat;
2387 barf("{unblockOneLocked}Daq Qagh: unexpected closure in blocking queue");
2394 static StgBlockingQueueElement *
2395 unblockOneLocked(StgBlockingQueueElement *bqe, StgClosure *node)
2398 PEs node_loc, tso_loc;
2400 node_loc = where_is(node); // should be lifted out of loop
2401 tso = (StgTSO *)bqe; // wastes an assignment to get the type right
2402 tso_loc = where_is((StgClosure *)tso);
2403 if (IS_LOCAL_TO(PROCS(node),tso_loc)) { // TSO is local
2404 /* !fake_fetch => TSO is on CurrentProc is same as IS_LOCAL_TO */
2405 ASSERT(CurrentProc!=node_loc || tso_loc==CurrentProc);
2406 CurrentTime[CurrentProc] += RtsFlags.GranFlags.Costs.lunblocktime;
2407 // insertThread(tso, node_loc);
2408 new_event(tso_loc, tso_loc, CurrentTime[CurrentProc],
2410 tso, node, (rtsSpark*)NULL);
2411 tso->link = END_TSO_QUEUE; // overwrite link just to be sure
2414 } else { // TSO is remote (actually should be FMBQ)
2415 CurrentTime[CurrentProc] += RtsFlags.GranFlags.Costs.mpacktime +
2416 RtsFlags.GranFlags.Costs.gunblocktime +
2417 RtsFlags.GranFlags.Costs.latency;
2418 new_event(tso_loc, CurrentProc, CurrentTime[CurrentProc],
2420 tso, node, (rtsSpark*)NULL);
2421 tso->link = END_TSO_QUEUE; // overwrite link just to be sure
2424 /* the thread-queue-overhead is accounted for in either Resume or UnblockThread */
2426 debugBelch(" %s TSO %d (%p) [PE %d] (block_info.closure=%p) (next=%p) ,",
2427 (node_loc==tso_loc ? "Local" : "Global"),
2428 tso->id, tso, CurrentProc, tso->block_info.closure, tso->link));
2429 tso->block_info.closure = NULL;
2430 IF_DEBUG(scheduler,debugBelch("-- Waking up thread %ld (%p)\n",
2434 static StgBlockingQueueElement *
2435 unblockOneLocked(StgBlockingQueueElement *bqe, StgClosure *node)
2437 StgBlockingQueueElement *next;
2439 switch (get_itbl(bqe)->type) {
2441 ASSERT(((StgTSO *)bqe)->why_blocked != NotBlocked);
2442 /* if it's a TSO just push it onto the run_queue */
2444 ((StgTSO *)bqe)->link = END_TSO_QUEUE; // debugging?
2445 APPEND_TO_RUN_QUEUE((StgTSO *)bqe);
2447 unblockCount(bqe, node);
2448 /* reset blocking status after dumping event */
2449 ((StgTSO *)bqe)->why_blocked = NotBlocked;
2453 /* if it's a BLOCKED_FETCH put it on the PendingFetches list */
2455 bqe->link = (StgBlockingQueueElement *)PendingFetches;
2456 PendingFetches = (StgBlockedFetch *)bqe;
2460 /* can ignore this case in a non-debugging setup;
2461 see comments on RBHSave closures above */
2463 /* check that the closure is an RBHSave closure */
2464 ASSERT(get_itbl((StgClosure *)bqe) == &stg_RBH_Save_0_info ||
2465 get_itbl((StgClosure *)bqe) == &stg_RBH_Save_1_info ||
2466 get_itbl((StgClosure *)bqe) == &stg_RBH_Save_2_info);
2470 barf("{unblockOneLocked}Daq Qagh: Unexpected IP (%#lx; %s) in blocking queue at %#lx\n",
2471 get_itbl((StgClosure *)bqe), info_type((StgClosure *)bqe),
2475 IF_PAR_DEBUG(bq, debugBelch(", %p (%s)\n", bqe, info_type((StgClosure*)bqe)));
2479 #else /* !GRAN && !PAR */
2481 unblockOneLocked(StgTSO *tso)
2485 ASSERT(get_itbl(tso)->type == TSO);
2486 ASSERT(tso->why_blocked != NotBlocked);
2487 tso->why_blocked = NotBlocked;
2489 tso->link = END_TSO_QUEUE;
2490 APPEND_TO_RUN_QUEUE(tso);
2492 IF_DEBUG(scheduler,sched_belch("waking up thread %ld", (long)tso->id));
2497 #if defined(GRAN) || defined(PAR)
2498 INLINE_ME StgBlockingQueueElement *
2499 unblockOne(StgBlockingQueueElement *bqe, StgClosure *node)
2501 ACQUIRE_LOCK(&sched_mutex);
2502 bqe = unblockOneLocked(bqe, node);
2503 RELEASE_LOCK(&sched_mutex);
2508 unblockOne(StgTSO *tso)
2510 ACQUIRE_LOCK(&sched_mutex);
2511 tso = unblockOneLocked(tso);
2512 RELEASE_LOCK(&sched_mutex);
2519 awakenBlockedQueue(StgBlockingQueueElement *q, StgClosure *node)
2521 StgBlockingQueueElement *bqe;
2526 debugBelch("##-_ AwBQ for node %p on PE %d @ %ld by TSO %d (%p): \n", \
2527 node, CurrentProc, CurrentTime[CurrentProc],
2528 CurrentTSO->id, CurrentTSO));
2530 node_loc = where_is(node);
2532 ASSERT(q == END_BQ_QUEUE ||
2533 get_itbl(q)->type == TSO || // q is either a TSO or an RBHSave
2534 get_itbl(q)->type == CONSTR); // closure (type constructor)
2535 ASSERT(is_unique(node));
2537 /* FAKE FETCH: magically copy the node to the tso's proc;
2538 no Fetch necessary because in reality the node should not have been
2539 moved to the other PE in the first place
2541 if (CurrentProc!=node_loc) {
2543 debugBelch("## node %p is on PE %d but CurrentProc is %d (TSO %d); assuming fake fetch and adjusting bitmask (old: %#x)\n",
2544 node, node_loc, CurrentProc, CurrentTSO->id,
2545 // CurrentTSO, where_is(CurrentTSO),
2546 node->header.gran.procs));
2547 node->header.gran.procs = (node->header.gran.procs) | PE_NUMBER(CurrentProc);
2549 debugBelch("## new bitmask of node %p is %#x\n",
2550 node, node->header.gran.procs));
2551 if (RtsFlags.GranFlags.GranSimStats.Global) {
2552 globalGranStats.tot_fake_fetches++;
2557 // ToDo: check: ASSERT(CurrentProc==node_loc);
2558 while (get_itbl(bqe)->type==TSO) { // q != END_TSO_QUEUE) {
2561 bqe points to the current element in the queue
2562 next points to the next element in the queue
2564 //tso = (StgTSO *)bqe; // wastes an assignment to get the type right
2565 //tso_loc = where_is(tso);
2567 bqe = unblockOneLocked(bqe, node);
2570 /* if this is the BQ of an RBH, we have to put back the info ripped out of
2571 the closure to make room for the anchor of the BQ */
2572 if (bqe!=END_BQ_QUEUE) {
2573 ASSERT(get_itbl(node)->type == RBH && get_itbl(bqe)->type == CONSTR);
2575 ASSERT((info_ptr==&RBH_Save_0_info) ||
2576 (info_ptr==&RBH_Save_1_info) ||
2577 (info_ptr==&RBH_Save_2_info));
2579 /* cf. convertToRBH in RBH.c for writing the RBHSave closure */
2580 ((StgRBH *)node)->blocking_queue = (StgBlockingQueueElement *)((StgRBHSave *)bqe)->payload[0];
2581 ((StgRBH *)node)->mut_link = (StgMutClosure *)((StgRBHSave *)bqe)->payload[1];
2584 debugBelch("## Filled in RBH_Save for %p (%s) at end of AwBQ\n",
2585 node, info_type(node)));
2588 /* statistics gathering */
2589 if (RtsFlags.GranFlags.GranSimStats.Global) {
2590 // globalGranStats.tot_bq_processing_time += bq_processing_time;
2591 globalGranStats.tot_bq_len += len; // total length of all bqs awakened
2592 // globalGranStats.tot_bq_len_local += len_local; // same for local TSOs only
2593 globalGranStats.tot_awbq++; // total no. of bqs awakened
2596 debugBelch("## BQ Stats of %p: [%d entries] %s\n",
2597 node, len, (bqe!=END_BQ_QUEUE) ? "RBH" : ""));
2601 awakenBlockedQueue(StgBlockingQueueElement *q, StgClosure *node)
2603 StgBlockingQueueElement *bqe;
2605 ACQUIRE_LOCK(&sched_mutex);
2607 IF_PAR_DEBUG(verbose,
2608 debugBelch("##-_ AwBQ for node %p on [%x]: \n",
2612 if(get_itbl(q)->type == CONSTR || q==END_BQ_QUEUE) {
2613 IF_PAR_DEBUG(verbose, debugBelch("## ... nothing to unblock so lets just return. RFP (BUG?)\n"));
2618 ASSERT(q == END_BQ_QUEUE ||
2619 get_itbl(q)->type == TSO ||
2620 get_itbl(q)->type == BLOCKED_FETCH ||
2621 get_itbl(q)->type == CONSTR);
2624 while (get_itbl(bqe)->type==TSO ||
2625 get_itbl(bqe)->type==BLOCKED_FETCH) {
2626 bqe = unblockOneLocked(bqe, node);
2628 RELEASE_LOCK(&sched_mutex);
2631 #else /* !GRAN && !PAR */
2634 awakenBlockedQueueNoLock(StgTSO *tso)
2636 while (tso != END_TSO_QUEUE) {
2637 tso = unblockOneLocked(tso);
2642 awakenBlockedQueue(StgTSO *tso)
2644 ACQUIRE_LOCK(&sched_mutex);
2645 while (tso != END_TSO_QUEUE) {
2646 tso = unblockOneLocked(tso);
2648 RELEASE_LOCK(&sched_mutex);
2652 /* ---------------------------------------------------------------------------
2654 - usually called inside a signal handler so it mustn't do anything fancy.
2655 ------------------------------------------------------------------------ */
2658 interruptStgRts(void)
2664 /* -----------------------------------------------------------------------------
2667 This is for use when we raise an exception in another thread, which
2669 This has nothing to do with the UnblockThread event in GranSim. -- HWL
2670 -------------------------------------------------------------------------- */
2672 #if defined(GRAN) || defined(PAR)
2674 NB: only the type of the blocking queue is different in GranSim and GUM
2675 the operations on the queue-elements are the same
2676 long live polymorphism!
2678 Locks: sched_mutex is held upon entry and exit.
2682 unblockThread(StgTSO *tso)
2684 StgBlockingQueueElement *t, **last;
2686 switch (tso->why_blocked) {
2689 return; /* not blocked */
2692 // Be careful: nothing to do here! We tell the scheduler that the thread
2693 // is runnable and we leave it to the stack-walking code to abort the
2694 // transaction while unwinding the stack. We should perhaps have a debugging
2695 // test to make sure that this really happens and that the 'zombie' transaction
2696 // does not get committed.
2700 ASSERT(get_itbl(tso->block_info.closure)->type == MVAR);
2702 StgBlockingQueueElement *last_tso = END_BQ_QUEUE;
2703 StgMVar *mvar = (StgMVar *)(tso->block_info.closure);
2705 last = (StgBlockingQueueElement **)&mvar->head;
2706 for (t = (StgBlockingQueueElement *)mvar->head;
2708 last = &t->link, last_tso = t, t = t->link) {
2709 if (t == (StgBlockingQueueElement *)tso) {
2710 *last = (StgBlockingQueueElement *)tso->link;
2711 if (mvar->tail == tso) {
2712 mvar->tail = (StgTSO *)last_tso;
2717 barf("unblockThread (MVAR): TSO not found");
2720 case BlockedOnBlackHole:
2721 ASSERT(get_itbl(tso->block_info.closure)->type == BLACKHOLE_BQ);
2723 StgBlockingQueue *bq = (StgBlockingQueue *)(tso->block_info.closure);
2725 last = &bq->blocking_queue;
2726 for (t = bq->blocking_queue;
2728 last = &t->link, t = t->link) {
2729 if (t == (StgBlockingQueueElement *)tso) {
2730 *last = (StgBlockingQueueElement *)tso->link;
2734 barf("unblockThread (BLACKHOLE): TSO not found");
2737 case BlockedOnException:
2739 StgTSO *target = tso->block_info.tso;
2741 ASSERT(get_itbl(target)->type == TSO);
2743 if (target->what_next == ThreadRelocated) {
2744 target = target->link;
2745 ASSERT(get_itbl(target)->type == TSO);
2748 ASSERT(target->blocked_exceptions != NULL);
2750 last = (StgBlockingQueueElement **)&target->blocked_exceptions;
2751 for (t = (StgBlockingQueueElement *)target->blocked_exceptions;
2753 last = &t->link, t = t->link) {
2754 ASSERT(get_itbl(t)->type == TSO);
2755 if (t == (StgBlockingQueueElement *)tso) {
2756 *last = (StgBlockingQueueElement *)tso->link;
2760 barf("unblockThread (Exception): TSO not found");
2764 case BlockedOnWrite:
2765 #if defined(mingw32_HOST_OS)
2766 case BlockedOnDoProc:
2769 /* take TSO off blocked_queue */
2770 StgBlockingQueueElement *prev = NULL;
2771 for (t = (StgBlockingQueueElement *)blocked_queue_hd; t != END_BQ_QUEUE;
2772 prev = t, t = t->link) {
2773 if (t == (StgBlockingQueueElement *)tso) {
2775 blocked_queue_hd = (StgTSO *)t->link;
2776 if ((StgBlockingQueueElement *)blocked_queue_tl == t) {
2777 blocked_queue_tl = END_TSO_QUEUE;
2780 prev->link = t->link;
2781 if ((StgBlockingQueueElement *)blocked_queue_tl == t) {
2782 blocked_queue_tl = (StgTSO *)prev;
2788 barf("unblockThread (I/O): TSO not found");
2791 case BlockedOnDelay:
2793 /* take TSO off sleeping_queue */
2794 StgBlockingQueueElement *prev = NULL;
2795 for (t = (StgBlockingQueueElement *)sleeping_queue; t != END_BQ_QUEUE;
2796 prev = t, t = t->link) {
2797 if (t == (StgBlockingQueueElement *)tso) {
2799 sleeping_queue = (StgTSO *)t->link;
2801 prev->link = t->link;
2806 barf("unblockThread (delay): TSO not found");
2810 barf("unblockThread");
2814 tso->link = END_TSO_QUEUE;
2815 tso->why_blocked = NotBlocked;
2816 tso->block_info.closure = NULL;
2817 PUSH_ON_RUN_QUEUE(tso);
2821 unblockThread(StgTSO *tso)
2825 /* To avoid locking unnecessarily. */
2826 if (tso->why_blocked == NotBlocked) {
2830 switch (tso->why_blocked) {
2833 // Be careful: nothing to do here! We tell the scheduler that the thread
2834 // is runnable and we leave it to the stack-walking code to abort the
2835 // transaction while unwinding the stack. We should perhaps have a debugging
2836 // test to make sure that this really happens and that the 'zombie' transaction
2837 // does not get committed.
2841 ASSERT(get_itbl(tso->block_info.closure)->type == MVAR);
2843 StgTSO *last_tso = END_TSO_QUEUE;
2844 StgMVar *mvar = (StgMVar *)(tso->block_info.closure);
2847 for (t = mvar->head; t != END_TSO_QUEUE;
2848 last = &t->link, last_tso = t, t = t->link) {
2851 if (mvar->tail == tso) {
2852 mvar->tail = last_tso;
2857 barf("unblockThread (MVAR): TSO not found");
2860 case BlockedOnBlackHole:
2861 ASSERT(get_itbl(tso->block_info.closure)->type == BLACKHOLE_BQ);
2863 StgBlockingQueue *bq = (StgBlockingQueue *)(tso->block_info.closure);
2865 last = &bq->blocking_queue;
2866 for (t = bq->blocking_queue; t != END_TSO_QUEUE;
2867 last = &t->link, t = t->link) {
2873 barf("unblockThread (BLACKHOLE): TSO not found");
2876 case BlockedOnException:
2878 StgTSO *target = tso->block_info.tso;
2880 ASSERT(get_itbl(target)->type == TSO);
2882 while (target->what_next == ThreadRelocated) {
2883 target = target->link;
2884 ASSERT(get_itbl(target)->type == TSO);
2887 ASSERT(target->blocked_exceptions != NULL);
2889 last = &target->blocked_exceptions;
2890 for (t = target->blocked_exceptions; t != END_TSO_QUEUE;
2891 last = &t->link, t = t->link) {
2892 ASSERT(get_itbl(t)->type == TSO);
2898 barf("unblockThread (Exception): TSO not found");
2902 case BlockedOnWrite:
2903 #if defined(mingw32_HOST_OS)
2904 case BlockedOnDoProc:
2907 StgTSO *prev = NULL;
2908 for (t = blocked_queue_hd; t != END_TSO_QUEUE;
2909 prev = t, t = t->link) {
2912 blocked_queue_hd = t->link;
2913 if (blocked_queue_tl == t) {
2914 blocked_queue_tl = END_TSO_QUEUE;
2917 prev->link = t->link;
2918 if (blocked_queue_tl == t) {
2919 blocked_queue_tl = prev;
2925 barf("unblockThread (I/O): TSO not found");
2928 case BlockedOnDelay:
2930 StgTSO *prev = NULL;
2931 for (t = sleeping_queue; t != END_TSO_QUEUE;
2932 prev = t, t = t->link) {
2935 sleeping_queue = t->link;
2937 prev->link = t->link;
2942 barf("unblockThread (delay): TSO not found");
2946 barf("unblockThread");
2950 tso->link = END_TSO_QUEUE;
2951 tso->why_blocked = NotBlocked;
2952 tso->block_info.closure = NULL;
2953 APPEND_TO_RUN_QUEUE(tso);
2957 /* -----------------------------------------------------------------------------
2960 * The following function implements the magic for raising an
2961 * asynchronous exception in an existing thread.
2963 * We first remove the thread from any queue on which it might be
2964 * blocked. The possible blockages are MVARs and BLACKHOLE_BQs.
2966 * We strip the stack down to the innermost CATCH_FRAME, building
2967 * thunks in the heap for all the active computations, so they can
2968 * be restarted if necessary. When we reach a CATCH_FRAME, we build
2969 * an application of the handler to the exception, and push it on
2970 * the top of the stack.
2972 * How exactly do we save all the active computations? We create an
2973 * AP_STACK for every UpdateFrame on the stack. Entering one of these
2974 * AP_STACKs pushes everything from the corresponding update frame
2975 * upwards onto the stack. (Actually, it pushes everything up to the
2976 * next update frame plus a pointer to the next AP_STACK object.
2977 * Entering the next AP_STACK object pushes more onto the stack until we
2978 * reach the last AP_STACK object - at which point the stack should look
2979 * exactly as it did when we killed the TSO and we can continue
2980 * execution by entering the closure on top of the stack.
2982 * We can also kill a thread entirely - this happens if either (a) the
2983 * exception passed to raiseAsync is NULL, or (b) there's no
2984 * CATCH_FRAME on the stack. In either case, we strip the entire
2985 * stack and replace the thread with a zombie.
2987 * Locks: sched_mutex held upon entry nor exit.
2989 * -------------------------------------------------------------------------- */
2992 deleteThread(StgTSO *tso)
2994 if (tso->why_blocked != BlockedOnCCall &&
2995 tso->why_blocked != BlockedOnCCall_NoUnblockExc) {
2996 raiseAsync(tso,NULL);
3000 #ifdef FORKPROCESS_PRIMOP_SUPPORTED
3002 deleteThreadImmediately(StgTSO *tso)
3003 { // for forkProcess only:
3004 // delete thread without giving it a chance to catch the KillThread exception
3006 if (tso->what_next == ThreadComplete || tso->what_next == ThreadKilled) {
3010 if (tso->why_blocked != BlockedOnCCall &&
3011 tso->why_blocked != BlockedOnCCall_NoUnblockExc) {
3015 tso->what_next = ThreadKilled;
3020 raiseAsyncWithLock(StgTSO *tso, StgClosure *exception)
3022 /* When raising async exs from contexts where sched_mutex isn't held;
3023 use raiseAsyncWithLock(). */
3024 ACQUIRE_LOCK(&sched_mutex);
3025 raiseAsync(tso,exception);
3026 RELEASE_LOCK(&sched_mutex);
3030 raiseAsync(StgTSO *tso, StgClosure *exception)
3032 raiseAsync_(tso, exception, rtsFalse);
3036 raiseAsync_(StgTSO *tso, StgClosure *exception, rtsBool stop_at_atomically)
3038 StgRetInfoTable *info;
3041 // Thread already dead?
3042 if (tso->what_next == ThreadComplete || tso->what_next == ThreadKilled) {
3047 sched_belch("raising exception in thread %ld.", (long)tso->id));
3049 // Remove it from any blocking queues
3054 // The stack freezing code assumes there's a closure pointer on
3055 // the top of the stack, so we have to arrange that this is the case...
3057 if (sp[0] == (W_)&stg_enter_info) {
3061 sp[0] = (W_)&stg_dummy_ret_closure;
3067 // 1. Let the top of the stack be the "current closure"
3069 // 2. Walk up the stack until we find either an UPDATE_FRAME or a
3072 // 3. If it's an UPDATE_FRAME, then make an AP_STACK containing the
3073 // current closure applied to the chunk of stack up to (but not
3074 // including) the update frame. This closure becomes the "current
3075 // closure". Go back to step 2.
3077 // 4. If it's a CATCH_FRAME, then leave the exception handler on
3078 // top of the stack applied to the exception.
3080 // 5. If it's a STOP_FRAME, then kill the thread.
3082 // NB: if we pass an ATOMICALLY_FRAME then abort the associated
3089 info = get_ret_itbl((StgClosure *)frame);
3091 while (info->i.type != UPDATE_FRAME
3092 && (info->i.type != CATCH_FRAME || exception == NULL)
3093 && info->i.type != STOP_FRAME
3094 && (info->i.type != ATOMICALLY_FRAME || stop_at_atomically == rtsFalse))
3096 if (info->i.type == CATCH_RETRY_FRAME || info->i.type == ATOMICALLY_FRAME) {
3097 // IF we find an ATOMICALLY_FRAME then we abort the
3098 // current transaction and propagate the exception. In
3099 // this case (unlike ordinary exceptions) we do not care
3100 // whether the transaction is valid or not because its
3101 // possible validity cannot have caused the exception
3102 // and will not be visible after the abort.
3104 debugBelch("Found atomically block delivering async exception\n"));
3105 stmAbortTransaction(tso -> trec);
3106 tso -> trec = stmGetEnclosingTRec(tso -> trec);
3108 frame += stack_frame_sizeW((StgClosure *)frame);
3109 info = get_ret_itbl((StgClosure *)frame);
3112 switch (info->i.type) {
3114 case ATOMICALLY_FRAME:
3115 ASSERT(stop_at_atomically);
3116 ASSERT(stmGetEnclosingTRec(tso->trec) == NO_TREC);
3117 stmCondemnTransaction(tso -> trec);
3121 // R1 is not a register: the return convention for IO in
3122 // this case puts the return value on the stack, so we
3123 // need to set up the stack to return to the atomically
3124 // frame properly...
3125 tso->sp = frame - 2;
3126 tso->sp[1] = (StgWord) &stg_NO_FINALIZER_closure; // why not?
3127 tso->sp[0] = (StgWord) &stg_ut_1_0_unreg_info;
3129 tso->what_next = ThreadRunGHC;
3133 // If we find a CATCH_FRAME, and we've got an exception to raise,
3134 // then build the THUNK raise(exception), and leave it on
3135 // top of the CATCH_FRAME ready to enter.
3139 StgCatchFrame *cf = (StgCatchFrame *)frame;
3143 // we've got an exception to raise, so let's pass it to the
3144 // handler in this frame.
3146 raise = (StgClosure *)allocate(sizeofW(StgClosure)+1);
3147 TICK_ALLOC_SE_THK(1,0);
3148 SET_HDR(raise,&stg_raise_info,cf->header.prof.ccs);
3149 raise->payload[0] = exception;
3151 // throw away the stack from Sp up to the CATCH_FRAME.
3155 /* Ensure that async excpetions are blocked now, so we don't get
3156 * a surprise exception before we get around to executing the
3159 if (tso->blocked_exceptions == NULL) {
3160 tso->blocked_exceptions = END_TSO_QUEUE;
3163 /* Put the newly-built THUNK on top of the stack, ready to execute
3164 * when the thread restarts.
3167 sp[-1] = (W_)&stg_enter_info;
3169 tso->what_next = ThreadRunGHC;
3170 IF_DEBUG(sanity, checkTSO(tso));
3179 // First build an AP_STACK consisting of the stack chunk above the
3180 // current update frame, with the top word on the stack as the
3183 words = frame - sp - 1;
3184 ap = (StgAP_STACK *)allocate(PAP_sizeW(words));
3187 ap->fun = (StgClosure *)sp[0];
3189 for(i=0; i < (nat)words; ++i) {
3190 ap->payload[i] = (StgClosure *)*sp++;
3193 SET_HDR(ap,&stg_AP_STACK_info,
3194 ((StgClosure *)frame)->header.prof.ccs /* ToDo */);
3195 TICK_ALLOC_UP_THK(words+1,0);
3198 debugBelch("sched: Updating ");
3199 printPtr((P_)((StgUpdateFrame *)frame)->updatee);
3200 debugBelch(" with ");
3201 printObj((StgClosure *)ap);
3204 // Replace the updatee with an indirection - happily
3205 // this will also wake up any threads currently
3206 // waiting on the result.
3208 // Warning: if we're in a loop, more than one update frame on
3209 // the stack may point to the same object. Be careful not to
3210 // overwrite an IND_OLDGEN in this case, because we'll screw
3211 // up the mutable lists. To be on the safe side, don't
3212 // overwrite any kind of indirection at all. See also
3213 // threadSqueezeStack in GC.c, where we have to make a similar
3216 if (!closure_IND(((StgUpdateFrame *)frame)->updatee)) {
3217 // revert the black hole
3218 UPD_IND_NOLOCK(((StgUpdateFrame *)frame)->updatee,
3221 sp += sizeofW(StgUpdateFrame) - 1;
3222 sp[0] = (W_)ap; // push onto stack
3227 // We've stripped the entire stack, the thread is now dead.
3228 sp += sizeofW(StgStopFrame);
3229 tso->what_next = ThreadKilled;
3240 /* -----------------------------------------------------------------------------
3241 raiseExceptionHelper
3243 This function is called by the raise# primitve, just so that we can
3244 move some of the tricky bits of raising an exception from C-- into
3245 C. Who knows, it might be a useful re-useable thing here too.
3246 -------------------------------------------------------------------------- */
3249 raiseExceptionHelper (StgTSO *tso, StgClosure *exception)
3251 StgClosure *raise_closure = NULL;
3253 StgRetInfoTable *info;
3255 // This closure represents the expression 'raise# E' where E
3256 // is the exception raise. It is used to overwrite all the
3257 // thunks which are currently under evaluataion.
3261 // LDV profiling: stg_raise_info has THUNK as its closure
3262 // type. Since a THUNK takes at least MIN_UPD_SIZE words in its
3263 // payload, MIN_UPD_SIZE is more approprate than 1. It seems that
3264 // 1 does not cause any problem unless profiling is performed.
3265 // However, when LDV profiling goes on, we need to linearly scan
3266 // small object pool, where raise_closure is stored, so we should
3267 // use MIN_UPD_SIZE.
3269 // raise_closure = (StgClosure *)RET_STGCALL1(P_,allocate,
3270 // sizeofW(StgClosure)+1);
3274 // Walk up the stack, looking for the catch frame. On the way,
3275 // we update any closures pointed to from update frames with the
3276 // raise closure that we just built.
3280 info = get_ret_itbl((StgClosure *)p);
3281 next = p + stack_frame_sizeW((StgClosure *)p);
3282 switch (info->i.type) {
3285 // Only create raise_closure if we need to.
3286 if (raise_closure == NULL) {
3288 (StgClosure *)allocate(sizeofW(StgClosure)+MIN_UPD_SIZE);
3289 SET_HDR(raise_closure, &stg_raise_info, CCCS);
3290 raise_closure->payload[0] = exception;
3292 UPD_IND(((StgUpdateFrame *)p)->updatee,raise_closure);
3296 case ATOMICALLY_FRAME:
3297 IF_DEBUG(stm, debugBelch("Found ATOMICALLY_FRAME at %p\n", p));
3299 return ATOMICALLY_FRAME;
3305 case CATCH_STM_FRAME:
3306 IF_DEBUG(stm, debugBelch("Found CATCH_STM_FRAME at %p\n", p));
3308 return CATCH_STM_FRAME;
3314 case CATCH_RETRY_FRAME:
3323 /* -----------------------------------------------------------------------------
3324 findRetryFrameHelper
3326 This function is called by the retry# primitive. It traverses the stack
3327 leaving tso->sp referring to the frame which should handle the retry.
3329 This should either be a CATCH_RETRY_FRAME (if the retry# is within an orElse#)
3330 or should be a ATOMICALLY_FRAME (if the retry# reaches the top level).
3332 We skip CATCH_STM_FRAMEs because retries are not considered to be exceptions,
3333 despite the similar implementation.
3335 We should not expect to see CATCH_FRAME or STOP_FRAME because those should
3336 not be created within memory transactions.
3337 -------------------------------------------------------------------------- */
3340 findRetryFrameHelper (StgTSO *tso)
3343 StgRetInfoTable *info;
3347 info = get_ret_itbl((StgClosure *)p);
3348 next = p + stack_frame_sizeW((StgClosure *)p);
3349 switch (info->i.type) {
3351 case ATOMICALLY_FRAME:
3352 IF_DEBUG(stm, debugBelch("Found ATOMICALLY_FRAME at %p during retrry\n", p));
3354 return ATOMICALLY_FRAME;
3356 case CATCH_RETRY_FRAME:
3357 IF_DEBUG(stm, debugBelch("Found CATCH_RETRY_FRAME at %p during retrry\n", p));
3359 return CATCH_RETRY_FRAME;
3361 case CATCH_STM_FRAME:
3363 ASSERT(info->i.type != CATCH_FRAME);
3364 ASSERT(info->i.type != STOP_FRAME);
3371 /* -----------------------------------------------------------------------------
3372 resurrectThreads is called after garbage collection on the list of
3373 threads found to be garbage. Each of these threads will be woken
3374 up and sent a signal: BlockedOnDeadMVar if the thread was blocked
3375 on an MVar, or NonTermination if the thread was blocked on a Black
3378 Locks: sched_mutex isn't held upon entry nor exit.
3379 -------------------------------------------------------------------------- */
3382 resurrectThreads( StgTSO *threads )
3386 for (tso = threads; tso != END_TSO_QUEUE; tso = next) {
3387 next = tso->global_link;
3388 tso->global_link = all_threads;
3390 IF_DEBUG(scheduler, sched_belch("resurrecting thread %d", tso->id));
3392 switch (tso->why_blocked) {
3394 case BlockedOnException:
3395 /* Called by GC - sched_mutex lock is currently held. */
3396 raiseAsync(tso,(StgClosure *)BlockedOnDeadMVar_closure);
3398 case BlockedOnBlackHole:
3399 raiseAsync(tso,(StgClosure *)NonTermination_closure);
3402 raiseAsync(tso,(StgClosure *)BlockedIndefinitely_closure);
3405 /* This might happen if the thread was blocked on a black hole
3406 * belonging to a thread that we've just woken up (raiseAsync
3407 * can wake up threads, remember...).
3411 barf("resurrectThreads: thread blocked in a strange way");
3416 /* ----------------------------------------------------------------------------
3417 * Debugging: why is a thread blocked
3418 * [Also provides useful information when debugging threaded programs
3419 * at the Haskell source code level, so enable outside of DEBUG. --sof 7/02]
3420 ------------------------------------------------------------------------- */
3424 printThreadBlockage(StgTSO *tso)
3426 switch (tso->why_blocked) {
3428 debugBelch("is blocked on read from fd %d", tso->block_info.fd);
3430 case BlockedOnWrite:
3431 debugBelch("is blocked on write to fd %d", tso->block_info.fd);
3433 #if defined(mingw32_HOST_OS)
3434 case BlockedOnDoProc:
3435 debugBelch("is blocked on proc (request: %d)", tso->block_info.async_result->reqID);
3438 case BlockedOnDelay:
3439 debugBelch("is blocked until %d", tso->block_info.target);
3442 debugBelch("is blocked on an MVar");
3444 case BlockedOnException:
3445 debugBelch("is blocked on delivering an exception to thread %d",
3446 tso->block_info.tso->id);
3448 case BlockedOnBlackHole:
3449 debugBelch("is blocked on a black hole");
3452 debugBelch("is not blocked");
3456 debugBelch("is blocked on global address; local FM_BQ is %p (%s)",
3457 tso->block_info.closure, info_type(tso->block_info.closure));
3459 case BlockedOnGA_NoSend:
3460 debugBelch("is blocked on global address (no send); local FM_BQ is %p (%s)",
3461 tso->block_info.closure, info_type(tso->block_info.closure));
3464 case BlockedOnCCall:
3465 debugBelch("is blocked on an external call");
3467 case BlockedOnCCall_NoUnblockExc:
3468 debugBelch("is blocked on an external call (exceptions were already blocked)");
3471 debugBelch("is blocked on an STM operation");
3474 barf("printThreadBlockage: strange tso->why_blocked: %d for TSO %d (%d)",
3475 tso->why_blocked, tso->id, tso);
3481 printThreadStatus(StgTSO *tso)
3483 switch (tso->what_next) {
3485 debugBelch("has been killed");
3487 case ThreadComplete:
3488 debugBelch("has completed");
3491 printThreadBlockage(tso);
3496 printAllThreads(void)
3501 char time_string[TIME_STR_LEN], node_str[NODE_STR_LEN];
3502 ullong_format_string(TIME_ON_PROC(CurrentProc),
3503 time_string, rtsFalse/*no commas!*/);
3505 debugBelch("all threads at [%s]:\n", time_string);
3507 char time_string[TIME_STR_LEN], node_str[NODE_STR_LEN];
3508 ullong_format_string(CURRENT_TIME,
3509 time_string, rtsFalse/*no commas!*/);
3511 debugBelch("all threads at [%s]:\n", time_string);
3513 debugBelch("all threads:\n");
3516 for (t = all_threads; t != END_TSO_QUEUE; t = t->global_link) {
3517 debugBelch("\tthread %d @ %p ", t->id, (void *)t);
3520 void *label = lookupThreadLabel(t->id);
3521 if (label) debugBelch("[\"%s\"] ",(char *)label);
3524 printThreadStatus(t);
3532 Print a whole blocking queue attached to node (debugging only).
3536 print_bq (StgClosure *node)
3538 StgBlockingQueueElement *bqe;
3542 debugBelch("## BQ of closure %p (%s): ",
3543 node, info_type(node));
3545 /* should cover all closures that may have a blocking queue */
3546 ASSERT(get_itbl(node)->type == BLACKHOLE_BQ ||
3547 get_itbl(node)->type == FETCH_ME_BQ ||
3548 get_itbl(node)->type == RBH ||
3549 get_itbl(node)->type == MVAR);
3551 ASSERT(node!=(StgClosure*)NULL); // sanity check
3553 print_bqe(((StgBlockingQueue*)node)->blocking_queue);
3557 Print a whole blocking queue starting with the element bqe.
3560 print_bqe (StgBlockingQueueElement *bqe)
3565 NB: In a parallel setup a BQ of an RBH must end with an RBH_Save closure;
3567 for (end = (bqe==END_BQ_QUEUE);
3568 !end; // iterate until bqe points to a CONSTR
3569 end = (get_itbl(bqe)->type == CONSTR) || (bqe->link==END_BQ_QUEUE),
3570 bqe = end ? END_BQ_QUEUE : bqe->link) {
3571 ASSERT(bqe != END_BQ_QUEUE); // sanity check
3572 ASSERT(bqe != (StgBlockingQueueElement *)NULL); // sanity check
3573 /* types of closures that may appear in a blocking queue */
3574 ASSERT(get_itbl(bqe)->type == TSO ||
3575 get_itbl(bqe)->type == BLOCKED_FETCH ||
3576 get_itbl(bqe)->type == CONSTR);
3577 /* only BQs of an RBH end with an RBH_Save closure */
3578 //ASSERT(get_itbl(bqe)->type != CONSTR || get_itbl(node)->type == RBH);
3580 switch (get_itbl(bqe)->type) {
3582 debugBelch(" TSO %u (%x),",
3583 ((StgTSO *)bqe)->id, ((StgTSO *)bqe));
3586 debugBelch(" BF (node=%p, ga=((%x, %d, %x)),",
3587 ((StgBlockedFetch *)bqe)->node,
3588 ((StgBlockedFetch *)bqe)->ga.payload.gc.gtid,
3589 ((StgBlockedFetch *)bqe)->ga.payload.gc.slot,
3590 ((StgBlockedFetch *)bqe)->ga.weight);
3593 debugBelch(" %s (IP %p),",
3594 (get_itbl(bqe) == &stg_RBH_Save_0_info ? "RBH_Save_0" :
3595 get_itbl(bqe) == &stg_RBH_Save_1_info ? "RBH_Save_1" :
3596 get_itbl(bqe) == &stg_RBH_Save_2_info ? "RBH_Save_2" :
3597 "RBH_Save_?"), get_itbl(bqe));
3600 barf("Unexpected closure type %s in blocking queue", // of %p (%s)",
3601 info_type((StgClosure *)bqe)); // , node, info_type(node));
3607 # elif defined(GRAN)
3609 print_bq (StgClosure *node)
3611 StgBlockingQueueElement *bqe;
3612 PEs node_loc, tso_loc;
3615 /* should cover all closures that may have a blocking queue */
3616 ASSERT(get_itbl(node)->type == BLACKHOLE_BQ ||
3617 get_itbl(node)->type == FETCH_ME_BQ ||
3618 get_itbl(node)->type == RBH);
3620 ASSERT(node!=(StgClosure*)NULL); // sanity check
3621 node_loc = where_is(node);
3623 debugBelch("## BQ of closure %p (%s) on [PE %d]: ",
3624 node, info_type(node), node_loc);
3627 NB: In a parallel setup a BQ of an RBH must end with an RBH_Save closure;
3629 for (bqe = ((StgBlockingQueue*)node)->blocking_queue, end = (bqe==END_BQ_QUEUE);
3630 !end; // iterate until bqe points to a CONSTR
3631 end = (get_itbl(bqe)->type == CONSTR) || (bqe->link==END_BQ_QUEUE), bqe = end ? END_BQ_QUEUE : bqe->link) {
3632 ASSERT(bqe != END_BQ_QUEUE); // sanity check
3633 ASSERT(bqe != (StgBlockingQueueElement *)NULL); // sanity check
3634 /* types of closures that may appear in a blocking queue */
3635 ASSERT(get_itbl(bqe)->type == TSO ||
3636 get_itbl(bqe)->type == CONSTR);
3637 /* only BQs of an RBH end with an RBH_Save closure */
3638 ASSERT(get_itbl(bqe)->type != CONSTR || get_itbl(node)->type == RBH);
3640 tso_loc = where_is((StgClosure *)bqe);
3641 switch (get_itbl(bqe)->type) {
3643 debugBelch(" TSO %d (%p) on [PE %d],",
3644 ((StgTSO *)bqe)->id, (StgTSO *)bqe, tso_loc);
3647 debugBelch(" %s (IP %p),",
3648 (get_itbl(bqe) == &stg_RBH_Save_0_info ? "RBH_Save_0" :
3649 get_itbl(bqe) == &stg_RBH_Save_1_info ? "RBH_Save_1" :
3650 get_itbl(bqe) == &stg_RBH_Save_2_info ? "RBH_Save_2" :
3651 "RBH_Save_?"), get_itbl(bqe));
3654 barf("Unexpected closure type %s in blocking queue of %p (%s)",
3655 info_type((StgClosure *)bqe), node, info_type(node));
3663 Nice and easy: only TSOs on the blocking queue
3666 print_bq (StgClosure *node)
3670 ASSERT(node!=(StgClosure*)NULL); // sanity check
3671 for (tso = ((StgBlockingQueue*)node)->blocking_queue;
3672 tso != END_TSO_QUEUE;
3674 ASSERT(tso!=NULL && tso!=END_TSO_QUEUE); // sanity check
3675 ASSERT(get_itbl(tso)->type == TSO); // guess what, sanity check
3676 debugBelch(" TSO %d (%p),", tso->id, tso);
3689 for (i=0, tso=run_queue_hd;
3690 tso != END_TSO_QUEUE;
3699 sched_belch(char *s, ...)
3703 #ifdef RTS_SUPPORTS_THREADS
3704 debugBelch("sched (task %p): ", osThreadId());
3708 debugBelch("sched: ");