1 /* ---------------------------------------------------------------------------
3 * (c) The GHC Team, 1998-2004
7 * Different GHC ways use this scheduler quite differently (see comments below)
8 * Here is the global picture:
10 * WAY Name CPP flag What's it for
11 * --------------------------------------
12 * mp GUM PARALLEL_HASKELL Parallel execution on a distrib. memory machine
13 * s SMP SMP Parallel execution on a shared memory machine
14 * mg GranSim GRAN Simulation of parallel execution
15 * md GUM/GdH DIST Distributed execution (based on GUM)
17 * --------------------------------------------------------------------------*/
20 * Version with support for distributed memory parallelism aka GUM (WAY=mp):
22 The main scheduling loop in GUM iterates until a finish message is received.
23 In that case a global flag @receivedFinish@ is set and this instance of
24 the RTS shuts down. See ghc/rts/parallel/HLComms.c:processMessages()
25 for the handling of incoming messages, such as PP_FINISH.
26 Note that in the parallel case we have a system manager that coordinates
27 different PEs, each of which are running one instance of the RTS.
28 See ghc/rts/parallel/SysMan.c for the main routine of the parallel program.
29 From this routine processes executing ghc/rts/Main.c are spawned. -- HWL
31 * Version with support for simulating parallel execution aka GranSim (WAY=mg):
33 The main scheduling code in GranSim is quite different from that in std
34 (concurrent) Haskell: while concurrent Haskell just iterates over the
35 threads in the runnable queue, GranSim is event driven, i.e. it iterates
36 over the events in the global event queue. -- HWL
39 #include "PosixSource.h"
44 #include "BlockAlloc.h"
45 #include "OSThreads.h"
49 #define COMPILING_SCHEDULER
51 #include "StgMiscClosures.h"
52 #include "Interpreter.h"
53 #include "Exception.h"
61 #include "ThreadLabels.h"
62 #include "LdvProfile.h"
65 #include "Proftimer.h"
68 #if defined(GRAN) || defined(PARALLEL_HASKELL)
69 # include "GranSimRts.h"
71 # include "ParallelRts.h"
72 # include "Parallel.h"
73 # include "ParallelDebug.h"
78 #include "Capability.h"
81 #ifdef HAVE_SYS_TYPES_H
82 #include <sys/types.h>
96 // Turn off inlining when debugging - it obfuscates things
99 # define STATIC_INLINE static
103 #define USED_IN_THREADED_RTS
105 #define USED_IN_THREADED_RTS STG_UNUSED
108 #ifdef RTS_SUPPORTS_THREADS
109 #define USED_WHEN_RTS_SUPPORTS_THREADS
111 #define USED_WHEN_RTS_SUPPORTS_THREADS STG_UNUSED
114 /* Main thread queue.
115 * Locks required: sched_mutex.
117 StgMainThread *main_threads = NULL;
121 StgTSO* ActiveTSO = NULL; /* for assigning system costs; GranSim-Light only */
122 /* rtsTime TimeOfNextEvent, EndOfTimeSlice; now in GranSim.c */
125 In GranSim we have a runnable and a blocked queue for each processor.
126 In order to minimise code changes new arrays run_queue_hds/tls
127 are created. run_queue_hd is then a short cut (macro) for
128 run_queue_hds[CurrentProc] (see GranSim.h).
131 StgTSO *run_queue_hds[MAX_PROC], *run_queue_tls[MAX_PROC];
132 StgTSO *blocked_queue_hds[MAX_PROC], *blocked_queue_tls[MAX_PROC];
133 StgTSO *ccalling_threadss[MAX_PROC];
134 /* We use the same global list of threads (all_threads) in GranSim as in
135 the std RTS (i.e. we are cheating). However, we don't use this list in
136 the GranSim specific code at the moment (so we are only potentially
142 * Locks required: sched_mutex.
144 StgTSO *run_queue_hd = NULL;
145 StgTSO *run_queue_tl = NULL;
146 StgTSO *blocked_queue_hd = NULL;
147 StgTSO *blocked_queue_tl = NULL;
148 StgTSO *blackhole_queue = NULL;
149 StgTSO *sleeping_queue = NULL; /* perhaps replace with a hash table? */
153 /* The blackhole_queue should be checked for threads to wake up. See
154 * Schedule.h for more thorough comment.
156 rtsBool blackholes_need_checking = rtsFalse;
158 /* Linked list of all threads.
159 * Used for detecting garbage collected threads.
161 StgTSO *all_threads = NULL;
163 /* When a thread performs a safe C call (_ccall_GC, using old
164 * terminology), it gets put on the suspended_ccalling_threads
165 * list. Used by the garbage collector.
167 static StgTSO *suspended_ccalling_threads;
169 /* KH: The following two flags are shared memory locations. There is no need
170 to lock them, since they are only unset at the end of a scheduler
174 /* flag set by signal handler to precipitate a context switch */
175 int context_switch = 0;
177 /* if this flag is set as well, give up execution */
178 rtsBool interrupted = rtsFalse;
180 /* If this flag is set, we are running Haskell code. Used to detect
181 * uses of 'foreign import unsafe' that should be 'safe'.
183 static rtsBool in_haskell = rtsFalse;
185 /* Next thread ID to allocate.
186 * Locks required: thread_id_mutex
188 static StgThreadID next_thread_id = 1;
191 * Pointers to the state of the current thread.
192 * Rule of thumb: if CurrentTSO != NULL, then we're running a Haskell
193 * thread. If CurrentTSO == NULL, then we're at the scheduler level.
196 /* The smallest stack size that makes any sense is:
197 * RESERVED_STACK_WORDS (so we can get back from the stack overflow)
198 * + sizeofW(StgStopFrame) (the stg_stop_thread_info frame)
199 * + 1 (the closure to enter)
201 * + 1 (spare slot req'd by stg_ap_v_ret)
203 * A thread with this stack will bomb immediately with a stack
204 * overflow, which will increase its stack size.
207 #define MIN_STACK_WORDS (RESERVED_STACK_WORDS + sizeofW(StgStopFrame) + 3)
214 /* This is used in `TSO.h' and gcc 2.96 insists that this variable actually
215 * exists - earlier gccs apparently didn't.
221 static Condition gc_pending_cond = INIT_COND_VAR;
224 static rtsBool ready_to_gc;
227 * Set to TRUE when entering a shutdown state (via shutdownHaskellAndExit()) --
228 * in an MT setting, needed to signal that a worker thread shouldn't hang around
229 * in the scheduler when it is out of work.
231 static rtsBool shutting_down_scheduler = rtsFalse;
233 #if defined(RTS_SUPPORTS_THREADS)
234 /* ToDo: carefully document the invariants that go together
235 * with these synchronisation objects.
237 Mutex sched_mutex = INIT_MUTEX_VAR;
238 Mutex term_mutex = INIT_MUTEX_VAR;
240 #endif /* RTS_SUPPORTS_THREADS */
242 #if defined(PARALLEL_HASKELL)
244 rtsTime TimeOfLastYield;
245 rtsBool emitSchedule = rtsTrue;
249 static char *whatNext_strs[] = {
259 /* -----------------------------------------------------------------------------
260 * static function prototypes
261 * -------------------------------------------------------------------------- */
263 #if defined(RTS_SUPPORTS_THREADS)
264 static void taskStart(void);
267 static void schedule( StgMainThread *mainThread USED_WHEN_RTS_SUPPORTS_THREADS,
268 Capability *initialCapability );
271 // These function all encapsulate parts of the scheduler loop, and are
272 // abstracted only to make the structure and control flow of the
273 // scheduler clearer.
275 static void schedulePreLoop(void);
276 static void scheduleStartSignalHandlers(void);
277 static void scheduleCheckBlockedThreads(void);
278 static void scheduleCheckBlackHoles(void);
279 static void scheduleDetectDeadlock(void);
281 static StgTSO *scheduleProcessEvent(rtsEvent *event);
283 #if defined(PARALLEL_HASKELL)
284 static StgTSO *scheduleSendPendingMessages(void);
285 static void scheduleActivateSpark(void);
286 static rtsBool scheduleGetRemoteWork(rtsBool *receivedFinish);
288 #if defined(PAR) || defined(GRAN)
289 static void scheduleGranParReport(void);
291 static void schedulePostRunThread(void);
292 static rtsBool scheduleHandleHeapOverflow( Capability *cap, StgTSO *t );
293 static void scheduleHandleStackOverflow( StgTSO *t);
294 static rtsBool scheduleHandleYield( StgTSO *t, nat prev_what_next );
295 static void scheduleHandleThreadBlocked( StgTSO *t );
296 static rtsBool scheduleHandleThreadFinished( StgMainThread *mainThread,
297 Capability *cap, StgTSO *t );
298 static void scheduleDoHeapProfile(void);
299 static void scheduleDoGC(void);
301 static void unblockThread(StgTSO *tso);
302 static rtsBool checkBlackHoles(void);
303 static SchedulerStatus waitThread_(/*out*/StgMainThread* m,
304 Capability *initialCapability
306 static void scheduleThread_ (StgTSO* tso);
307 static void AllRoots(evac_fn evac);
309 static StgTSO *threadStackOverflow(StgTSO *tso);
311 static void raiseAsync_(StgTSO *tso, StgClosure *exception,
312 rtsBool stop_at_atomically);
314 static void printThreadBlockage(StgTSO *tso);
315 static void printThreadStatus(StgTSO *tso);
317 #if defined(PARALLEL_HASKELL)
318 StgTSO * createSparkThread(rtsSpark spark);
319 StgTSO * activateSpark (rtsSpark spark);
322 /* ----------------------------------------------------------------------------
324 * ------------------------------------------------------------------------- */
326 #if defined(RTS_SUPPORTS_THREADS)
327 static rtsBool startingWorkerThread = rtsFalse;
332 ACQUIRE_LOCK(&sched_mutex);
333 startingWorkerThread = rtsFalse;
336 RELEASE_LOCK(&sched_mutex);
340 startSchedulerTaskIfNecessary(void)
342 if ( !EMPTY_RUN_QUEUE()
343 && !shutting_down_scheduler // not if we're shutting down
344 && !startingWorkerThread )
346 // we don't want to start another worker thread
347 // just because the last one hasn't yet reached the
348 // "waiting for capability" state
349 startingWorkerThread = rtsTrue;
350 if (!maybeStartNewWorker(taskStart)) {
351 startingWorkerThread = rtsFalse;
357 /* -----------------------------------------------------------------------------
358 * Putting a thread on the run queue: different scheduling policies
359 * -------------------------------------------------------------------------- */
362 addToRunQueue( StgTSO *t )
364 #if defined(PARALLEL_HASKELL)
365 if (RtsFlags.ParFlags.doFairScheduling) {
366 // this does round-robin scheduling; good for concurrency
367 APPEND_TO_RUN_QUEUE(t);
369 // this does unfair scheduling; good for parallelism
370 PUSH_ON_RUN_QUEUE(t);
373 // this does round-robin scheduling; good for concurrency
374 APPEND_TO_RUN_QUEUE(t);
378 /* ---------------------------------------------------------------------------
379 Main scheduling loop.
381 We use round-robin scheduling, each thread returning to the
382 scheduler loop when one of these conditions is detected:
385 * timer expires (thread yields)
390 Locking notes: we acquire the scheduler lock once at the beginning
391 of the scheduler loop, and release it when
393 * running a thread, or
394 * waiting for work, or
395 * waiting for a GC to complete.
398 In a GranSim setup this loop iterates over the global event queue.
399 This revolves around the global event queue, which determines what
400 to do next. Therefore, it's more complicated than either the
401 concurrent or the parallel (GUM) setup.
404 GUM iterates over incoming messages.
405 It starts with nothing to do (thus CurrentTSO == END_TSO_QUEUE),
406 and sends out a fish whenever it has nothing to do; in-between
407 doing the actual reductions (shared code below) it processes the
408 incoming messages and deals with delayed operations
409 (see PendingFetches).
410 This is not the ugliest code you could imagine, but it's bloody close.
412 ------------------------------------------------------------------------ */
415 schedule( StgMainThread *mainThread USED_WHEN_RTS_SUPPORTS_THREADS,
416 Capability *initialCapability )
420 StgThreadReturnCode ret;
423 #elif defined(PARALLEL_HASKELL)
426 rtsBool receivedFinish = rtsFalse;
428 nat tp_size, sp_size; // stats only
433 // Pre-condition: sched_mutex is held.
434 // We might have a capability, passed in as initialCapability.
435 cap = initialCapability;
437 #if !defined(RTS_SUPPORTS_THREADS)
438 // simply initialise it in the non-threaded case
439 grabCapability(&cap);
443 sched_belch("### NEW SCHEDULER LOOP (main thr: %p, cap: %p)",
444 mainThread, initialCapability);
449 // -----------------------------------------------------------
450 // Scheduler loop starts here:
452 #if defined(PARALLEL_HASKELL)
453 #define TERMINATION_CONDITION (!receivedFinish)
455 #define TERMINATION_CONDITION ((event = get_next_event()) != (rtsEvent*)NULL)
457 #define TERMINATION_CONDITION rtsTrue
460 while (TERMINATION_CONDITION) {
463 /* Choose the processor with the next event */
464 CurrentProc = event->proc;
465 CurrentTSO = event->tso;
468 IF_DEBUG(scheduler, printAllThreads());
472 // Wait until GC has completed, if necessary.
476 releaseCapability(cap);
477 IF_DEBUG(scheduler,sched_belch("waiting for GC"));
478 waitCondition( &gc_pending_cond, &sched_mutex );
483 #if defined(RTS_SUPPORTS_THREADS)
484 // Yield the capability to higher-priority tasks if necessary.
487 yieldCapability(&cap);
490 // If we do not currently hold a capability, we wait for one
493 waitForCapability(&sched_mutex, &cap,
494 mainThread ? &mainThread->bound_thread_cond : NULL);
497 // We now have a capability...
500 // Check whether we have re-entered the RTS from Haskell without
501 // going via suspendThread()/resumeThread (i.e. a 'safe' foreign
504 errorBelch("schedule: re-entered unsafely.\n"
505 " Perhaps a 'foreign import unsafe' should be 'safe'?");
510 // Test for interruption. If interrupted==rtsTrue, then either
511 // we received a keyboard interrupt (^C), or the scheduler is
512 // trying to shut down all the tasks (shutting_down_scheduler) in
516 if (shutting_down_scheduler) {
517 IF_DEBUG(scheduler, sched_belch("shutting down"));
518 releaseCapability(cap);
520 mainThread->stat = Interrupted;
521 mainThread->ret = NULL;
525 IF_DEBUG(scheduler, sched_belch("interrupted"));
530 #if defined(not_yet) && defined(SMP)
532 // Top up the run queue from our spark pool. We try to make the
533 // number of threads in the run queue equal to the number of
534 // free capabilities.
538 if (EMPTY_RUN_QUEUE()) {
539 spark = findSpark(rtsFalse);
541 break; /* no more sparks in the pool */
543 createSparkThread(spark);
545 sched_belch("==^^ turning spark of closure %p into a thread",
546 (StgClosure *)spark));
552 scheduleStartSignalHandlers();
554 // Only check the black holes here if we've nothing else to do.
555 // During normal execution, the black hole list only gets checked
556 // at GC time, to avoid repeatedly traversing this possibly long
557 // list each time around the scheduler.
558 if (EMPTY_RUN_QUEUE()) { scheduleCheckBlackHoles(); }
560 scheduleCheckBlockedThreads();
562 scheduleDetectDeadlock();
564 // Normally, the only way we can get here with no threads to
565 // run is if a keyboard interrupt received during
566 // scheduleCheckBlockedThreads() or scheduleDetectDeadlock().
567 // Additionally, it is not fatal for the
568 // threaded RTS to reach here with no threads to run.
570 // win32: might be here due to awaitEvent() being abandoned
571 // as a result of a console event having been delivered.
572 if ( EMPTY_RUN_QUEUE() ) {
573 #if !defined(RTS_SUPPORTS_THREADS) && !defined(mingw32_HOST_OS)
576 continue; // nothing to do
579 #if defined(PARALLEL_HASKELL)
580 scheduleSendPendingMessages();
581 if (EMPTY_RUN_QUEUE() && scheduleActivateSpark())
585 ASSERT(next_fish_to_send_at==0); // i.e. no delayed fishes left!
588 /* If we still have no work we need to send a FISH to get a spark
590 if (EMPTY_RUN_QUEUE()) {
591 if (!scheduleGetRemoteWork(&receivedFinish)) continue;
592 ASSERT(rtsFalse); // should not happen at the moment
594 // from here: non-empty run queue.
595 // TODO: merge above case with this, only one call processMessages() !
596 if (PacketsWaiting()) { /* process incoming messages, if
597 any pending... only in else
598 because getRemoteWork waits for
600 receivedFinish = processMessages();
605 scheduleProcessEvent(event);
609 // Get a thread to run
611 ASSERT(run_queue_hd != END_TSO_QUEUE);
614 #if defined(GRAN) || defined(PAR)
615 scheduleGranParReport(); // some kind of debuging output
617 // Sanity check the thread we're about to run. This can be
618 // expensive if there is lots of thread switching going on...
619 IF_DEBUG(sanity,checkTSO(t));
622 #if defined(RTS_SUPPORTS_THREADS)
623 // Check whether we can run this thread in the current task.
624 // If not, we have to pass our capability to the right task.
626 StgMainThread *m = t->main;
633 sched_belch("### Running thread %d in bound thread", t->id));
634 // yes, the Haskell thread is bound to the current native thread
639 sched_belch("### thread %d bound to another OS thread", t->id));
640 // no, bound to a different Haskell thread: pass to that thread
641 PUSH_ON_RUN_QUEUE(t);
642 passCapability(&m->bound_thread_cond);
648 if(mainThread != NULL)
649 // The thread we want to run is bound.
652 sched_belch("### this OS thread cannot run thread %d", t->id));
653 // no, the current native thread is bound to a different
654 // Haskell thread, so pass it to any worker thread
655 PUSH_ON_RUN_QUEUE(t);
656 passCapabilityToWorker();
663 cap->r.rCurrentTSO = t;
665 /* context switches are now initiated by the timer signal, unless
666 * the user specified "context switch as often as possible", with
669 if ((RtsFlags.ConcFlags.ctxtSwitchTicks == 0
670 && (run_queue_hd != END_TSO_QUEUE
671 || blocked_queue_hd != END_TSO_QUEUE
672 || sleeping_queue != END_TSO_QUEUE)))
677 RELEASE_LOCK(&sched_mutex);
679 IF_DEBUG(scheduler, sched_belch("-->> running thread %ld %s ...",
680 (long)t->id, whatNext_strs[t->what_next]));
682 #if defined(PROFILING)
683 startHeapProfTimer();
686 // ----------------------------------------------------------------------
687 // Run the current thread
689 prev_what_next = t->what_next;
691 errno = t->saved_errno;
692 in_haskell = rtsTrue;
694 switch (prev_what_next) {
698 /* Thread already finished, return to scheduler. */
699 ret = ThreadFinished;
703 ret = StgRun((StgFunPtr) stg_returnToStackTop, &cap->r);
706 case ThreadInterpret:
707 ret = interpretBCO(cap);
711 barf("schedule: invalid what_next field");
714 // We have run some Haskell code: there might be blackhole-blocked
715 // threads to wake up now.
716 if ( blackhole_queue != END_TSO_QUEUE ) {
717 blackholes_need_checking = rtsTrue;
720 in_haskell = rtsFalse;
722 // The TSO might have moved, eg. if it re-entered the RTS and a GC
723 // happened. So find the new location:
724 t = cap->r.rCurrentTSO;
726 // And save the current errno in this thread.
727 t->saved_errno = errno;
729 // ----------------------------------------------------------------------
731 /* Costs for the scheduler are assigned to CCS_SYSTEM */
732 #if defined(PROFILING)
737 ACQUIRE_LOCK(&sched_mutex);
739 #if defined(RTS_SUPPORTS_THREADS)
740 IF_DEBUG(scheduler,debugBelch("sched (task %p): ", osThreadId()););
741 #elif !defined(GRAN) && !defined(PARALLEL_HASKELL)
742 IF_DEBUG(scheduler,debugBelch("sched: "););
745 schedulePostRunThread();
749 ready_to_gc = scheduleHandleHeapOverflow(cap,t);
753 scheduleHandleStackOverflow(t);
757 if (scheduleHandleYield(t, prev_what_next)) {
758 // shortcut for switching between compiler/interpreter:
764 scheduleHandleThreadBlocked(t);
769 if (scheduleHandleThreadFinished(mainThread, cap, t)) return;;
773 barf("schedule: invalid thread return code %d", (int)ret);
776 scheduleDoHeapProfile();
778 } /* end of while() */
780 IF_PAR_DEBUG(verbose,
781 debugBelch("== Leaving schedule() after having received Finish\n"));
784 /* ----------------------------------------------------------------------------
785 * Setting up the scheduler loop
786 * ASSUMES: sched_mutex
787 * ------------------------------------------------------------------------- */
790 schedulePreLoop(void)
793 /* set up first event to get things going */
794 /* ToDo: assign costs for system setup and init MainTSO ! */
795 new_event(CurrentProc, CurrentProc, CurrentTime[CurrentProc],
797 CurrentTSO, (StgClosure*)NULL, (rtsSpark*)NULL);
800 debugBelch("GRAN: Init CurrentTSO (in schedule) = %p\n",
802 G_TSO(CurrentTSO, 5));
804 if (RtsFlags.GranFlags.Light) {
805 /* Save current time; GranSim Light only */
806 CurrentTSO->gran.clock = CurrentTime[CurrentProc];
811 /* ----------------------------------------------------------------------------
812 * Start any pending signal handlers
813 * ASSUMES: sched_mutex
814 * ------------------------------------------------------------------------- */
817 scheduleStartSignalHandlers(void)
819 #if defined(RTS_USER_SIGNALS) && !defined(RTS_SUPPORTS_THREADS)
820 if (signals_pending()) {
821 RELEASE_LOCK(&sched_mutex); /* ToDo: kill */
822 startSignalHandlers();
823 ACQUIRE_LOCK(&sched_mutex);
828 /* ----------------------------------------------------------------------------
829 * Check for blocked threads that can be woken up.
830 * ASSUMES: sched_mutex
831 * ------------------------------------------------------------------------- */
834 scheduleCheckBlockedThreads(void)
837 // Check whether any waiting threads need to be woken up. If the
838 // run queue is empty, and there are no other tasks running, we
839 // can wait indefinitely for something to happen.
841 if ( !EMPTY_QUEUE(blocked_queue_hd) || !EMPTY_QUEUE(sleeping_queue) )
843 #if defined(RTS_SUPPORTS_THREADS)
844 // We shouldn't be here...
845 barf("schedule: awaitEvent() in threaded RTS");
847 awaitEvent( EMPTY_RUN_QUEUE() && !blackholes_need_checking );
852 /* ----------------------------------------------------------------------------
853 * Check for threads blocked on BLACKHOLEs that can be woken up
854 * ASSUMES: sched_mutex
855 * ------------------------------------------------------------------------- */
857 scheduleCheckBlackHoles( void )
859 if ( blackholes_need_checking )
862 blackholes_need_checking = rtsFalse;
866 /* ----------------------------------------------------------------------------
867 * Detect deadlock conditions and attempt to resolve them.
868 * ASSUMES: sched_mutex
869 * ------------------------------------------------------------------------- */
872 scheduleDetectDeadlock(void)
875 * Detect deadlock: when we have no threads to run, there are no
876 * threads blocked, waiting for I/O, or sleeping, and all the
877 * other tasks are waiting for work, we must have a deadlock of
880 if ( EMPTY_THREAD_QUEUES() )
882 #if !defined(PARALLEL_HASKELL) && !defined(RTS_SUPPORTS_THREADS)
883 IF_DEBUG(scheduler, sched_belch("deadlocked, forcing major GC..."));
885 // Garbage collection can release some new threads due to
886 // either (a) finalizers or (b) threads resurrected because
887 // they are unreachable and will therefore be sent an
888 // exception. Any threads thus released will be immediately
890 GarbageCollect(GetRoots,rtsTrue);
891 if ( !EMPTY_RUN_QUEUE() ) return;
893 #if defined(RTS_USER_SIGNALS)
894 /* If we have user-installed signal handlers, then wait
895 * for signals to arrive rather then bombing out with a
898 if ( anyUserHandlers() ) {
900 sched_belch("still deadlocked, waiting for signals..."));
904 #if !defined(RTS_SUPPORTS_THREADS)
905 if (signals_pending()) {
906 RELEASE_LOCK(&sched_mutex);
907 startSignalHandlers();
908 ACQUIRE_LOCK(&sched_mutex);
912 // either we have threads to run, or we were interrupted:
913 ASSERT(!EMPTY_RUN_QUEUE() || interrupted);
917 /* Probably a real deadlock. Send the current main thread the
918 * Deadlock exception (or in the SMP build, send *all* main
919 * threads the deadlock exception, since none of them can make
925 switch (m->tso->why_blocked) {
926 case BlockedOnBlackHole:
927 case BlockedOnException:
929 raiseAsync(m->tso, (StgClosure *)NonTermination_closure);
932 barf("deadlock: main thread blocked in a strange way");
936 #elif defined(RTS_SUPPORTS_THREADS)
937 // ToDo: add deadlock detection in threaded RTS
938 #elif defined(PARALLEL_HASKELL)
939 // ToDo: add deadlock detection in GUM (similar to SMP) -- HWL
944 /* ----------------------------------------------------------------------------
945 * Process an event (GRAN only)
946 * ------------------------------------------------------------------------- */
950 scheduleProcessEvent(rtsEvent *event)
954 if (RtsFlags.GranFlags.Light)
955 GranSimLight_enter_system(event, &ActiveTSO); // adjust ActiveTSO etc
957 /* adjust time based on time-stamp */
958 if (event->time > CurrentTime[CurrentProc] &&
959 event->evttype != ContinueThread)
960 CurrentTime[CurrentProc] = event->time;
962 /* Deal with the idle PEs (may issue FindWork or MoveSpark events) */
963 if (!RtsFlags.GranFlags.Light)
966 IF_DEBUG(gran, debugBelch("GRAN: switch by event-type\n"));
968 /* main event dispatcher in GranSim */
969 switch (event->evttype) {
970 /* Should just be continuing execution */
972 IF_DEBUG(gran, debugBelch("GRAN: doing ContinueThread\n"));
973 /* ToDo: check assertion
974 ASSERT(run_queue_hd != (StgTSO*)NULL &&
975 run_queue_hd != END_TSO_QUEUE);
977 /* Ignore ContinueThreads for fetching threads (if synchr comm) */
978 if (!RtsFlags.GranFlags.DoAsyncFetch &&
979 procStatus[CurrentProc]==Fetching) {
980 debugBelch("ghuH: Spurious ContinueThread while Fetching ignored; TSO %d (%p) [PE %d]\n",
981 CurrentTSO->id, CurrentTSO, CurrentProc);
984 /* Ignore ContinueThreads for completed threads */
985 if (CurrentTSO->what_next == ThreadComplete) {
986 debugBelch("ghuH: found a ContinueThread event for completed thread %d (%p) [PE %d] (ignoring ContinueThread)\n",
987 CurrentTSO->id, CurrentTSO, CurrentProc);
990 /* Ignore ContinueThreads for threads that are being migrated */
991 if (PROCS(CurrentTSO)==Nowhere) {
992 debugBelch("ghuH: trying to run the migrating TSO %d (%p) [PE %d] (ignoring ContinueThread)\n",
993 CurrentTSO->id, CurrentTSO, CurrentProc);
996 /* The thread should be at the beginning of the run queue */
997 if (CurrentTSO!=run_queue_hds[CurrentProc]) {
998 debugBelch("ghuH: TSO %d (%p) [PE %d] is not at the start of the run_queue when doing a ContinueThread\n",
999 CurrentTSO->id, CurrentTSO, CurrentProc);
1000 break; // run the thread anyway
1003 new_event(proc, proc, CurrentTime[proc],
1005 (StgTSO*)NULL, (StgClosure*)NULL, (rtsSpark*)NULL);
1007 */ /* Catches superfluous CONTINUEs -- should be unnecessary */
1008 break; // now actually run the thread; DaH Qu'vam yImuHbej
1011 do_the_fetchnode(event);
1012 goto next_thread; /* handle next event in event queue */
1015 do_the_globalblock(event);
1016 goto next_thread; /* handle next event in event queue */
1019 do_the_fetchreply(event);
1020 goto next_thread; /* handle next event in event queue */
1022 case UnblockThread: /* Move from the blocked queue to the tail of */
1023 do_the_unblock(event);
1024 goto next_thread; /* handle next event in event queue */
1026 case ResumeThread: /* Move from the blocked queue to the tail of */
1027 /* the runnable queue ( i.e. Qu' SImqa'lu') */
1028 event->tso->gran.blocktime +=
1029 CurrentTime[CurrentProc] - event->tso->gran.blockedat;
1030 do_the_startthread(event);
1031 goto next_thread; /* handle next event in event queue */
1034 do_the_startthread(event);
1035 goto next_thread; /* handle next event in event queue */
1038 do_the_movethread(event);
1039 goto next_thread; /* handle next event in event queue */
1042 do_the_movespark(event);
1043 goto next_thread; /* handle next event in event queue */
1046 do_the_findwork(event);
1047 goto next_thread; /* handle next event in event queue */
1050 barf("Illegal event type %u\n", event->evttype);
1053 /* This point was scheduler_loop in the old RTS */
1055 IF_DEBUG(gran, debugBelch("GRAN: after main switch\n"));
1057 TimeOfLastEvent = CurrentTime[CurrentProc];
1058 TimeOfNextEvent = get_time_of_next_event();
1059 IgnoreEvents=(TimeOfNextEvent==0); // HWL HACK
1060 // CurrentTSO = ThreadQueueHd;
1062 IF_DEBUG(gran, debugBelch("GRAN: time of next event is: %ld\n",
1065 if (RtsFlags.GranFlags.Light)
1066 GranSimLight_leave_system(event, &ActiveTSO);
1068 EndOfTimeSlice = CurrentTime[CurrentProc]+RtsFlags.GranFlags.time_slice;
1071 debugBelch("GRAN: end of time-slice is %#lx\n", EndOfTimeSlice));
1073 /* in a GranSim setup the TSO stays on the run queue */
1075 /* Take a thread from the run queue. */
1076 POP_RUN_QUEUE(t); // take_off_run_queue(t);
1079 debugBelch("GRAN: About to run current thread, which is\n");
1082 context_switch = 0; // turned on via GranYield, checking events and time slice
1085 DumpGranEvent(GR_SCHEDULE, t));
1087 procStatus[CurrentProc] = Busy;
1091 /* ----------------------------------------------------------------------------
1092 * Send pending messages (PARALLEL_HASKELL only)
1093 * ------------------------------------------------------------------------- */
1095 #if defined(PARALLEL_HASKELL)
1097 scheduleSendPendingMessages(void)
1103 # if defined(PAR) // global Mem.Mgmt., omit for now
1104 if (PendingFetches != END_BF_QUEUE) {
1109 if (RtsFlags.ParFlags.BufferTime) {
1110 // if we use message buffering, we must send away all message
1111 // packets which have become too old...
1117 /* ----------------------------------------------------------------------------
1118 * Activate spark threads (PARALLEL_HASKELL only)
1119 * ------------------------------------------------------------------------- */
1121 #if defined(PARALLEL_HASKELL)
1123 scheduleActivateSpark(void)
1126 ASSERT(EMPTY_RUN_QUEUE());
1127 /* We get here if the run queue is empty and want some work.
1128 We try to turn a spark into a thread, and add it to the run queue,
1129 from where it will be picked up in the next iteration of the scheduler
1133 /* :-[ no local threads => look out for local sparks */
1134 /* the spark pool for the current PE */
1135 pool = &(cap.r.rSparks); // JB: cap = (old) MainCap
1136 if (advisory_thread_count < RtsFlags.ParFlags.maxThreads &&
1137 pool->hd < pool->tl) {
1139 * ToDo: add GC code check that we really have enough heap afterwards!!
1141 * If we're here (no runnable threads) and we have pending
1142 * sparks, we must have a space problem. Get enough space
1143 * to turn one of those pending sparks into a
1147 spark = findSpark(rtsFalse); /* get a spark */
1148 if (spark != (rtsSpark) NULL) {
1149 tso = createThreadFromSpark(spark); /* turn the spark into a thread */
1150 IF_PAR_DEBUG(fish, // schedule,
1151 debugBelch("==== schedule: Created TSO %d (%p); %d threads active\n",
1152 tso->id, tso, advisory_thread_count));
1154 if (tso==END_TSO_QUEUE) { /* failed to activate spark->back to loop */
1155 IF_PAR_DEBUG(fish, // schedule,
1156 debugBelch("==^^ failed to create thread from spark @ %lx\n",
1158 return rtsFalse; /* failed to generate a thread */
1159 } /* otherwise fall through & pick-up new tso */
1161 IF_PAR_DEBUG(fish, // schedule,
1162 debugBelch("==^^ no local sparks (spark pool contains only NFs: %d)\n",
1163 spark_queue_len(pool)));
1164 return rtsFalse; /* failed to generate a thread */
1166 return rtsTrue; /* success in generating a thread */
1167 } else { /* no more threads permitted or pool empty */
1168 return rtsFalse; /* failed to generateThread */
1171 tso = NULL; // avoid compiler warning only
1172 return rtsFalse; /* dummy in non-PAR setup */
1175 #endif // PARALLEL_HASKELL
1177 /* ----------------------------------------------------------------------------
1178 * Get work from a remote node (PARALLEL_HASKELL only)
1179 * ------------------------------------------------------------------------- */
1181 #if defined(PARALLEL_HASKELL)
1183 scheduleGetRemoteWork(rtsBool *receivedFinish)
1185 ASSERT(EMPTY_RUN_QUEUE());
1187 if (RtsFlags.ParFlags.BufferTime) {
1188 IF_PAR_DEBUG(verbose,
1189 debugBelch("...send all pending data,"));
1192 for (i=1; i<=nPEs; i++)
1193 sendImmediately(i); // send all messages away immediately
1197 //++EDEN++ idle() , i.e. send all buffers, wait for work
1198 // suppress fishing in EDEN... just look for incoming messages
1199 // (blocking receive)
1200 IF_PAR_DEBUG(verbose,
1201 debugBelch("...wait for incoming messages...\n"));
1202 *receivedFinish = processMessages(); // blocking receive...
1204 // and reenter scheduling loop after having received something
1205 // (return rtsFalse below)
1207 # else /* activate SPARKS machinery */
1208 /* We get here, if we have no work, tried to activate a local spark, but still
1209 have no work. We try to get a remote spark, by sending a FISH message.
1210 Thread migration should be added here, and triggered when a sequence of
1211 fishes returns without work. */
1212 delay = (RtsFlags.ParFlags.fishDelay!=0ll ? RtsFlags.ParFlags.fishDelay : 0ll);
1214 /* =8-[ no local sparks => look for work on other PEs */
1216 * We really have absolutely no work. Send out a fish
1217 * (there may be some out there already), and wait for
1218 * something to arrive. We clearly can't run any threads
1219 * until a SCHEDULE or RESUME arrives, and so that's what
1220 * we're hoping to see. (Of course, we still have to
1221 * respond to other types of messages.)
1223 rtsTime now = msTime() /*CURRENT_TIME*/;
1224 IF_PAR_DEBUG(verbose,
1225 debugBelch("-- now=%ld\n", now));
1226 IF_PAR_DEBUG(fish, // verbose,
1227 if (outstandingFishes < RtsFlags.ParFlags.maxFishes &&
1228 (last_fish_arrived_at!=0 &&
1229 last_fish_arrived_at+delay > now)) {
1230 debugBelch("--$$ <%llu> delaying FISH until %llu (last fish %llu, delay %llu)\n",
1231 now, last_fish_arrived_at+delay,
1232 last_fish_arrived_at,
1236 if (outstandingFishes < RtsFlags.ParFlags.maxFishes &&
1237 advisory_thread_count < RtsFlags.ParFlags.maxThreads) { // send a FISH, but when?
1238 if (last_fish_arrived_at==0 ||
1239 (last_fish_arrived_at+delay <= now)) { // send FISH now!
1240 /* outstandingFishes is set in sendFish, processFish;
1241 avoid flooding system with fishes via delay */
1242 next_fish_to_send_at = 0;
1244 /* ToDo: this should be done in the main scheduling loop to avoid the
1245 busy wait here; not so bad if fish delay is very small */
1246 int iq = 0; // DEBUGGING -- HWL
1247 next_fish_to_send_at = last_fish_arrived_at+delay; // remember when to send
1248 /* send a fish when ready, but process messages that arrive in the meantime */
1250 if (PacketsWaiting()) {
1252 *receivedFinish = processMessages();
1255 } while (!*receivedFinish || now<next_fish_to_send_at);
1256 // JB: This means the fish could become obsolete, if we receive
1257 // work. Better check for work again?
1258 // last line: while (!receivedFinish || !haveWork || now<...)
1259 // next line: if (receivedFinish || haveWork )
1261 if (*receivedFinish) // no need to send a FISH if we are finishing anyway
1262 return rtsFalse; // NB: this will leave scheduler loop
1263 // immediately after return!
1265 IF_PAR_DEBUG(fish, // verbose,
1266 debugBelch("--$$ <%llu> sent delayed fish (%d processMessages); active/total threads=%d/%d\n",now,iq,run_queue_len(),advisory_thread_count));
1270 // JB: IMHO, this should all be hidden inside sendFish(...)
1272 sendFish(pe, thisPE, NEW_FISH_AGE, NEW_FISH_HISTORY,
1275 // Global statistics: count no. of fishes
1276 if (RtsFlags.ParFlags.ParStats.Global &&
1277 RtsFlags.GcFlags.giveStats > NO_GC_STATS) {
1278 globalParStats.tot_fish_mess++;
1282 /* delayed fishes must have been sent by now! */
1283 next_fish_to_send_at = 0;
1286 *receivedFinish = processMessages();
1287 # endif /* SPARKS */
1290 /* NB: this function always returns rtsFalse, meaning the scheduler
1291 loop continues with the next iteration;
1293 return code means success in finding work; we enter this function
1294 if there is no local work, thus have to send a fish which takes
1295 time until it arrives with work; in the meantime we should process
1296 messages in the main loop;
1299 #endif // PARALLEL_HASKELL
1301 /* ----------------------------------------------------------------------------
1302 * PAR/GRAN: Report stats & debugging info(?)
1303 * ------------------------------------------------------------------------- */
1305 #if defined(PAR) || defined(GRAN)
1307 scheduleGranParReport(void)
1309 ASSERT(run_queue_hd != END_TSO_QUEUE);
1311 /* Take a thread from the run queue, if we have work */
1312 POP_RUN_QUEUE(t); // take_off_run_queue(END_TSO_QUEUE);
1314 /* If this TSO has got its outport closed in the meantime,
1315 * it mustn't be run. Instead, we have to clean it up as if it was finished.
1316 * It has to be marked as TH_DEAD for this purpose.
1317 * If it is TH_TERM instead, it is supposed to have finished in the normal way.
1319 JB: TODO: investigate wether state change field could be nuked
1320 entirely and replaced by the normal tso state (whatnext
1321 field). All we want to do is to kill tsos from outside.
1324 /* ToDo: write something to the log-file
1325 if (RTSflags.ParFlags.granSimStats && !sameThread)
1326 DumpGranEvent(GR_SCHEDULE, RunnableThreadsHd);
1330 /* the spark pool for the current PE */
1331 pool = &(cap.r.rSparks); // cap = (old) MainCap
1334 debugBelch("--=^ %d threads, %d sparks on [%#x]\n",
1335 run_queue_len(), spark_queue_len(pool), CURRENT_PROC));
1338 debugBelch("--=^ %d threads, %d sparks on [%#x]\n",
1339 run_queue_len(), spark_queue_len(pool), CURRENT_PROC));
1341 if (RtsFlags.ParFlags.ParStats.Full &&
1342 (t->par.sparkname != (StgInt)0) && // only log spark generated threads
1343 (emitSchedule || // forced emit
1344 (t && LastTSO && t->id != LastTSO->id))) {
1346 we are running a different TSO, so write a schedule event to log file
1347 NB: If we use fair scheduling we also have to write a deschedule
1348 event for LastTSO; with unfair scheduling we know that the
1349 previous tso has blocked whenever we switch to another tso, so
1350 we don't need it in GUM for now
1352 IF_PAR_DEBUG(fish, // schedule,
1353 debugBelch("____ scheduling spark generated thread %d (%lx) (%lx) via a forced emit\n",t->id,t,t->par.sparkname));
1355 DumpRawGranEvent(CURRENT_PROC, CURRENT_PROC,
1356 GR_SCHEDULE, t, (StgClosure *)NULL, 0, 0);
1357 emitSchedule = rtsFalse;
1362 /* ----------------------------------------------------------------------------
1363 * After running a thread...
1364 * ASSUMES: sched_mutex
1365 * ------------------------------------------------------------------------- */
1368 schedulePostRunThread(void)
1371 /* HACK 675: if the last thread didn't yield, make sure to print a
1372 SCHEDULE event to the log file when StgRunning the next thread, even
1373 if it is the same one as before */
1375 TimeOfLastYield = CURRENT_TIME;
1378 /* some statistics gathering in the parallel case */
1380 #if defined(GRAN) || defined(PAR) || defined(EDEN)
1384 IF_DEBUG(gran, DumpGranEvent(GR_DESCHEDULE, t));
1385 globalGranStats.tot_heapover++;
1387 globalParStats.tot_heapover++;
1394 DumpGranEvent(GR_DESCHEDULE, t));
1395 globalGranStats.tot_stackover++;
1398 // DumpGranEvent(GR_DESCHEDULE, t);
1399 globalParStats.tot_stackover++;
1403 case ThreadYielding:
1406 DumpGranEvent(GR_DESCHEDULE, t));
1407 globalGranStats.tot_yields++;
1410 // DumpGranEvent(GR_DESCHEDULE, t);
1411 globalParStats.tot_yields++;
1418 debugBelch("--<< thread %ld (%p; %s) stopped, blocking on node %p [PE %d] with BQ: ",
1419 t->id, t, whatNext_strs[t->what_next], t->block_info.closure,
1420 (t->block_info.closure==(StgClosure*)NULL ? 99 : where_is(t->block_info.closure)));
1421 if (t->block_info.closure!=(StgClosure*)NULL)
1422 print_bq(t->block_info.closure);
1425 // ??? needed; should emit block before
1427 DumpGranEvent(GR_DESCHEDULE, t));
1428 prune_eventq(t, (StgClosure *)NULL); // prune ContinueThreads for t
1431 ASSERT(procStatus[CurrentProc]==Busy ||
1432 ((procStatus[CurrentProc]==Fetching) &&
1433 (t->block_info.closure!=(StgClosure*)NULL)));
1434 if (run_queue_hds[CurrentProc] == END_TSO_QUEUE &&
1435 !(!RtsFlags.GranFlags.DoAsyncFetch &&
1436 procStatus[CurrentProc]==Fetching))
1437 procStatus[CurrentProc] = Idle;
1440 //++PAR++ blockThread() writes the event (change?)
1444 case ThreadFinished:
1448 barf("parGlobalStats: unknown return code");
1454 /* -----------------------------------------------------------------------------
1455 * Handle a thread that returned to the scheduler with ThreadHeepOverflow
1456 * ASSUMES: sched_mutex
1457 * -------------------------------------------------------------------------- */
1460 scheduleHandleHeapOverflow( Capability *cap, StgTSO *t )
1462 // did the task ask for a large block?
1463 if (cap->r.rHpAlloc > BLOCK_SIZE) {
1464 // if so, get one and push it on the front of the nursery.
1468 blocks = (lnat)BLOCK_ROUND_UP(cap->r.rHpAlloc) / BLOCK_SIZE;
1471 debugBelch("--<< thread %ld (%s) stopped: requesting a large block (size %ld)\n",
1472 (long)t->id, whatNext_strs[t->what_next], blocks));
1474 // don't do this if it would push us over the
1475 // alloc_blocks_lim limit; we'll GC first.
1476 if (alloc_blocks + blocks < alloc_blocks_lim) {
1478 alloc_blocks += blocks;
1479 bd = allocGroup( blocks );
1481 // link the new group into the list
1482 bd->link = cap->r.rCurrentNursery;
1483 bd->u.back = cap->r.rCurrentNursery->u.back;
1484 if (cap->r.rCurrentNursery->u.back != NULL) {
1485 cap->r.rCurrentNursery->u.back->link = bd;
1487 ASSERT(g0s0->blocks == cap->r.rCurrentNursery &&
1488 g0s0->blocks == cap->r.rNursery);
1489 cap->r.rNursery = g0s0->blocks = bd;
1491 cap->r.rCurrentNursery->u.back = bd;
1493 // initialise it as a nursery block. We initialise the
1494 // step, gen_no, and flags field of *every* sub-block in
1495 // this large block, because this is easier than making
1496 // sure that we always find the block head of a large
1497 // block whenever we call Bdescr() (eg. evacuate() and
1498 // isAlive() in the GC would both have to do this, at
1502 for (x = bd; x < bd + blocks; x++) {
1509 // don't forget to update the block count in g0s0.
1510 g0s0->n_blocks += blocks;
1511 // This assert can be a killer if the app is doing lots
1512 // of large block allocations.
1513 ASSERT(countBlocks(g0s0->blocks) == g0s0->n_blocks);
1515 // now update the nursery to point to the new block
1516 cap->r.rCurrentNursery = bd;
1518 // we might be unlucky and have another thread get on the
1519 // run queue before us and steal the large block, but in that
1520 // case the thread will just end up requesting another large
1522 PUSH_ON_RUN_QUEUE(t);
1523 return rtsFalse; /* not actually GC'ing */
1527 /* make all the running tasks block on a condition variable,
1528 * maybe set context_switch and wait till they all pile in,
1529 * then have them wait on a GC condition variable.
1532 debugBelch("--<< thread %ld (%s) stopped: HeapOverflow\n",
1533 (long)t->id, whatNext_strs[t->what_next]));
1536 ASSERT(!is_on_queue(t,CurrentProc));
1537 #elif defined(PARALLEL_HASKELL)
1538 /* Currently we emit a DESCHEDULE event before GC in GUM.
1539 ToDo: either add separate event to distinguish SYSTEM time from rest
1540 or just nuke this DESCHEDULE (and the following SCHEDULE) */
1541 if (0 && RtsFlags.ParFlags.ParStats.Full) {
1542 DumpRawGranEvent(CURRENT_PROC, CURRENT_PROC,
1543 GR_DESCHEDULE, t, (StgClosure *)NULL, 0, 0);
1544 emitSchedule = rtsTrue;
1548 PUSH_ON_RUN_QUEUE(t);
1550 /* actual GC is done at the end of the while loop in schedule() */
1553 /* -----------------------------------------------------------------------------
1554 * Handle a thread that returned to the scheduler with ThreadStackOverflow
1555 * ASSUMES: sched_mutex
1556 * -------------------------------------------------------------------------- */
1559 scheduleHandleStackOverflow( StgTSO *t)
1561 IF_DEBUG(scheduler,debugBelch("--<< thread %ld (%s) stopped, StackOverflow\n",
1562 (long)t->id, whatNext_strs[t->what_next]));
1563 /* just adjust the stack for this thread, then pop it back
1568 /* enlarge the stack */
1569 StgTSO *new_t = threadStackOverflow(t);
1571 /* This TSO has moved, so update any pointers to it from the
1572 * main thread stack. It better not be on any other queues...
1573 * (it shouldn't be).
1575 if (t->main != NULL) {
1576 t->main->tso = new_t;
1578 PUSH_ON_RUN_QUEUE(new_t);
1582 /* -----------------------------------------------------------------------------
1583 * Handle a thread that returned to the scheduler with ThreadYielding
1584 * ASSUMES: sched_mutex
1585 * -------------------------------------------------------------------------- */
1588 scheduleHandleYield( StgTSO *t, nat prev_what_next )
1590 // Reset the context switch flag. We don't do this just before
1591 // running the thread, because that would mean we would lose ticks
1592 // during GC, which can lead to unfair scheduling (a thread hogs
1593 // the CPU because the tick always arrives during GC). This way
1594 // penalises threads that do a lot of allocation, but that seems
1595 // better than the alternative.
1598 /* put the thread back on the run queue. Then, if we're ready to
1599 * GC, check whether this is the last task to stop. If so, wake
1600 * up the GC thread. getThread will block during a GC until the
1604 if (t->what_next != prev_what_next) {
1605 debugBelch("--<< thread %ld (%s) stopped to switch evaluators\n",
1606 (long)t->id, whatNext_strs[t->what_next]);
1608 debugBelch("--<< thread %ld (%s) stopped, yielding\n",
1609 (long)t->id, whatNext_strs[t->what_next]);
1614 //debugBelch("&& Doing sanity check on yielding TSO %ld.", t->id);
1616 ASSERT(t->link == END_TSO_QUEUE);
1618 // Shortcut if we're just switching evaluators: don't bother
1619 // doing stack squeezing (which can be expensive), just run the
1621 if (t->what_next != prev_what_next) {
1628 ASSERT(!is_on_queue(t,CurrentProc));
1631 //debugBelch("&& Doing sanity check on all ThreadQueues (and their TSOs).");
1632 checkThreadQsSanity(rtsTrue));
1639 /* add a ContinueThread event to actually process the thread */
1640 new_event(CurrentProc, CurrentProc, CurrentTime[CurrentProc],
1642 t, (StgClosure*)NULL, (rtsSpark*)NULL);
1644 debugBelch("GRAN: eventq and runnableq after adding yielded thread to queue again:\n");
1651 /* -----------------------------------------------------------------------------
1652 * Handle a thread that returned to the scheduler with ThreadBlocked
1653 * ASSUMES: sched_mutex
1654 * -------------------------------------------------------------------------- */
1657 scheduleHandleThreadBlocked( StgTSO *t
1658 #if !defined(GRAN) && !defined(DEBUG)
1665 debugBelch("--<< thread %ld (%p; %s) stopped, blocking on node %p [PE %d] with BQ: \n",
1666 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)));
1667 if (t->block_info.closure!=(StgClosure*)NULL) print_bq(t->block_info.closure));
1669 // ??? needed; should emit block before
1671 DumpGranEvent(GR_DESCHEDULE, t));
1672 prune_eventq(t, (StgClosure *)NULL); // prune ContinueThreads for t
1675 ASSERT(procStatus[CurrentProc]==Busy ||
1676 ((procStatus[CurrentProc]==Fetching) &&
1677 (t->block_info.closure!=(StgClosure*)NULL)));
1678 if (run_queue_hds[CurrentProc] == END_TSO_QUEUE &&
1679 !(!RtsFlags.GranFlags.DoAsyncFetch &&
1680 procStatus[CurrentProc]==Fetching))
1681 procStatus[CurrentProc] = Idle;
1685 debugBelch("--<< thread %ld (%p; %s) stopped, blocking on node %p with BQ: \n",
1686 t->id, t, whatNext_strs[t->what_next], t->block_info.closure));
1689 if (t->block_info.closure!=(StgClosure*)NULL)
1690 print_bq(t->block_info.closure));
1692 /* Send a fetch (if BlockedOnGA) and dump event to log file */
1695 /* whatever we schedule next, we must log that schedule */
1696 emitSchedule = rtsTrue;
1699 /* don't need to do anything. Either the thread is blocked on
1700 * I/O, in which case we'll have called addToBlockedQueue
1701 * previously, or it's blocked on an MVar or Blackhole, in which
1702 * case it'll be on the relevant queue already.
1704 ASSERT(t->why_blocked != NotBlocked);
1706 debugBelch("--<< thread %d (%s) stopped: ",
1707 t->id, whatNext_strs[t->what_next]);
1708 printThreadBlockage(t);
1711 /* Only for dumping event to log file
1712 ToDo: do I need this in GranSim, too?
1718 /* -----------------------------------------------------------------------------
1719 * Handle a thread that returned to the scheduler with ThreadFinished
1720 * ASSUMES: sched_mutex
1721 * -------------------------------------------------------------------------- */
1724 scheduleHandleThreadFinished( StgMainThread *mainThread
1725 USED_WHEN_RTS_SUPPORTS_THREADS,
1729 /* Need to check whether this was a main thread, and if so,
1730 * return with the return value.
1732 * We also end up here if the thread kills itself with an
1733 * uncaught exception, see Exception.cmm.
1735 IF_DEBUG(scheduler,debugBelch("--++ thread %d (%s) finished\n",
1736 t->id, whatNext_strs[t->what_next]));
1739 endThread(t, CurrentProc); // clean-up the thread
1740 #elif defined(PARALLEL_HASKELL)
1741 /* For now all are advisory -- HWL */
1742 //if(t->priority==AdvisoryPriority) ??
1743 advisory_thread_count--; // JB: Caution with this counter, buggy!
1746 if(t->dist.priority==RevalPriority)
1750 # if defined(EDENOLD)
1751 // the thread could still have an outport... (BUG)
1752 if (t->eden.outport != -1) {
1753 // delete the outport for the tso which has finished...
1754 IF_PAR_DEBUG(eden_ports,
1755 debugBelch("WARNING: Scheduler removes outport %d for TSO %d.\n",
1756 t->eden.outport, t->id));
1759 // thread still in the process (HEAVY BUG! since outport has just been closed...)
1760 if (t->eden.epid != -1) {
1761 IF_PAR_DEBUG(eden_ports,
1762 debugBelch("WARNING: Scheduler removes TSO %d from process %d .\n",
1763 t->id, t->eden.epid));
1764 removeTSOfromProcess(t);
1769 if (RtsFlags.ParFlags.ParStats.Full &&
1770 !RtsFlags.ParFlags.ParStats.Suppressed)
1771 DumpEndEvent(CURRENT_PROC, t, rtsFalse /* not mandatory */);
1773 // t->par only contains statistics: left out for now...
1775 debugBelch("**** end thread: ended sparked thread %d (%lx); sparkname: %lx\n",
1776 t->id,t,t->par.sparkname));
1778 #endif // PARALLEL_HASKELL
1781 // Check whether the thread that just completed was a main
1782 // thread, and if so return with the result.
1784 // There is an assumption here that all thread completion goes
1785 // through this point; we need to make sure that if a thread
1786 // ends up in the ThreadKilled state, that it stays on the run
1787 // queue so it can be dealt with here.
1790 #if defined(RTS_SUPPORTS_THREADS)
1793 mainThread->tso == t
1797 // We are a bound thread: this must be our thread that just
1799 ASSERT(mainThread->tso == t);
1801 if (t->what_next == ThreadComplete) {
1802 if (mainThread->ret) {
1803 // NOTE: return val is tso->sp[1] (see StgStartup.hc)
1804 *(mainThread->ret) = (StgClosure *)mainThread->tso->sp[1];
1806 mainThread->stat = Success;
1808 if (mainThread->ret) {
1809 *(mainThread->ret) = NULL;
1812 mainThread->stat = Interrupted;
1814 mainThread->stat = Killed;
1818 removeThreadLabel((StgWord)mainThread->tso->id);
1820 if (mainThread->prev == NULL) {
1821 main_threads = mainThread->link;
1823 mainThread->prev->link = mainThread->link;
1825 if (mainThread->link != NULL) {
1826 mainThread->link->prev = NULL;
1828 releaseCapability(cap);
1829 return rtsTrue; // tells schedule() to return
1832 #ifdef RTS_SUPPORTS_THREADS
1833 ASSERT(t->main == NULL);
1835 if (t->main != NULL) {
1836 // Must be a main thread that is not the topmost one. Leave
1837 // it on the run queue until the stack has unwound to the
1838 // point where we can deal with this. Leaving it on the run
1839 // queue also ensures that the garbage collector knows about
1840 // this thread and its return value (it gets dropped from the
1841 // all_threads list so there's no other way to find it).
1842 APPEND_TO_RUN_QUEUE(t);
1848 /* -----------------------------------------------------------------------------
1849 * Perform a heap census, if PROFILING
1850 * -------------------------------------------------------------------------- */
1853 scheduleDoHeapProfile(void)
1856 // When we have +RTS -i0 and we're heap profiling, do a census at
1857 // every GC. This lets us get repeatable runs for debugging.
1858 if (performHeapProfile ||
1859 (RtsFlags.ProfFlags.profileInterval==0 &&
1860 RtsFlags.ProfFlags.doHeapProfile && ready_to_gc)) {
1861 GarbageCollect(GetRoots, rtsTrue);
1863 performHeapProfile = rtsFalse;
1864 ready_to_gc = rtsFalse; // we already GC'd
1869 /* -----------------------------------------------------------------------------
1870 * Perform a garbage collection if necessary
1871 * ASSUMES: sched_mutex
1872 * -------------------------------------------------------------------------- */
1880 // The last task to stop actually gets to do the GC. The rest
1881 // of the tasks release their capabilities and wait gc_pending_cond.
1882 if (ready_to_gc && allFreeCapabilities())
1887 /* Kick any transactions which are invalid back to their
1888 * atomically frames. When next scheduled they will try to
1889 * commit, this commit will fail and they will retry.
1891 for (t = all_threads; t != END_TSO_QUEUE; t = t -> link) {
1892 if (t -> what_next != ThreadRelocated && t -> trec != NO_TREC && t -> why_blocked == NotBlocked) {
1893 if (!stmValidateTransaction (t -> trec)) {
1894 IF_DEBUG(stm, sched_belch("trec %p found wasting its time", t));
1896 // strip the stack back to the ATOMICALLY_FRAME, aborting
1897 // the (nested) transaction, and saving the stack of any
1898 // partially-evaluated thunks on the heap.
1899 raiseAsync_(t, NULL, rtsTrue);
1902 ASSERT(get_itbl((StgClosure *)t->sp)->type == ATOMICALLY_FRAME);
1908 // so this happens periodically:
1909 scheduleCheckBlackHoles();
1911 /* everybody back, start the GC.
1912 * Could do it in this thread, or signal a condition var
1913 * to do it in another thread. Either way, we need to
1914 * broadcast on gc_pending_cond afterward.
1916 #if defined(RTS_SUPPORTS_THREADS)
1917 IF_DEBUG(scheduler,sched_belch("doing GC"));
1919 GarbageCollect(GetRoots,rtsFalse);
1920 ready_to_gc = rtsFalse;
1922 broadcastCondition(&gc_pending_cond);
1925 /* add a ContinueThread event to continue execution of current thread */
1926 new_event(CurrentProc, CurrentProc, CurrentTime[CurrentProc],
1928 t, (StgClosure*)NULL, (rtsSpark*)NULL);
1930 debugBelch("GRAN: eventq and runnableq after Garbage collection:\n\n");
1937 /* ---------------------------------------------------------------------------
1938 * rtsSupportsBoundThreads(): is the RTS built to support bound threads?
1939 * used by Control.Concurrent for error checking.
1940 * ------------------------------------------------------------------------- */
1943 rtsSupportsBoundThreads(void)
1952 /* ---------------------------------------------------------------------------
1953 * isThreadBound(tso): check whether tso is bound to an OS thread.
1954 * ------------------------------------------------------------------------- */
1957 isThreadBound(StgTSO* tso USED_IN_THREADED_RTS)
1960 return (tso->main != NULL);
1965 /* ---------------------------------------------------------------------------
1966 * Singleton fork(). Do not copy any running threads.
1967 * ------------------------------------------------------------------------- */
1969 #ifndef mingw32_HOST_OS
1970 #define FORKPROCESS_PRIMOP_SUPPORTED
1973 #ifdef FORKPROCESS_PRIMOP_SUPPORTED
1975 deleteThreadImmediately(StgTSO *tso);
1978 forkProcess(HsStablePtr *entry
1979 #ifndef FORKPROCESS_PRIMOP_SUPPORTED
1984 #ifdef FORKPROCESS_PRIMOP_SUPPORTED
1990 IF_DEBUG(scheduler,sched_belch("forking!"));
1991 rts_lock(); // This not only acquires sched_mutex, it also
1992 // makes sure that no other threads are running
1996 if (pid) { /* parent */
1998 /* just return the pid */
2002 } else { /* child */
2005 // delete all threads
2006 run_queue_hd = run_queue_tl = END_TSO_QUEUE;
2008 for (t = all_threads; t != END_TSO_QUEUE; t = next) {
2011 // don't allow threads to catch the ThreadKilled exception
2012 deleteThreadImmediately(t);
2015 // wipe the main thread list
2016 while((m = main_threads) != NULL) {
2017 main_threads = m->link;
2018 # ifdef THREADED_RTS
2019 closeCondition(&m->bound_thread_cond);
2024 rc = rts_evalStableIO(entry, NULL); // run the action
2025 rts_checkSchedStatus("forkProcess",rc);
2029 hs_exit(); // clean up and exit
2032 #else /* !FORKPROCESS_PRIMOP_SUPPORTED */
2033 barf("forkProcess#: primop not supported, sorry!\n");
2038 /* ---------------------------------------------------------------------------
2039 * deleteAllThreads(): kill all the live threads.
2041 * This is used when we catch a user interrupt (^C), before performing
2042 * any necessary cleanups and running finalizers.
2044 * Locks: sched_mutex held.
2045 * ------------------------------------------------------------------------- */
2048 deleteAllThreads ( void )
2051 IF_DEBUG(scheduler,sched_belch("deleting all threads"));
2052 for (t = all_threads; t != END_TSO_QUEUE; t = next) {
2053 next = t->global_link;
2057 // The run queue now contains a bunch of ThreadKilled threads. We
2058 // must not throw these away: the main thread(s) will be in there
2059 // somewhere, and the main scheduler loop has to deal with it.
2060 // Also, the run queue is the only thing keeping these threads from
2061 // being GC'd, and we don't want the "main thread has been GC'd" panic.
2063 ASSERT(blocked_queue_hd == END_TSO_QUEUE);
2064 ASSERT(blackhole_queue == END_TSO_QUEUE);
2065 ASSERT(sleeping_queue == END_TSO_QUEUE);
2068 /* startThread and insertThread are now in GranSim.c -- HWL */
2071 /* ---------------------------------------------------------------------------
2072 * Suspending & resuming Haskell threads.
2074 * When making a "safe" call to C (aka _ccall_GC), the task gives back
2075 * its capability before calling the C function. This allows another
2076 * task to pick up the capability and carry on running Haskell
2077 * threads. It also means that if the C call blocks, it won't lock
2080 * The Haskell thread making the C call is put to sleep for the
2081 * duration of the call, on the susepended_ccalling_threads queue. We
2082 * give out a token to the task, which it can use to resume the thread
2083 * on return from the C function.
2084 * ------------------------------------------------------------------------- */
2087 suspendThread( StgRegTable *reg )
2091 int saved_errno = errno;
2093 /* assume that *reg is a pointer to the StgRegTable part
2096 cap = (Capability *)((void *)((unsigned char*)reg - sizeof(StgFunTable)));
2098 ACQUIRE_LOCK(&sched_mutex);
2101 sched_belch("thread %d did a _ccall_gc", cap->r.rCurrentTSO->id));
2103 // XXX this might not be necessary --SDM
2104 cap->r.rCurrentTSO->what_next = ThreadRunGHC;
2106 threadPaused(cap->r.rCurrentTSO);
2107 cap->r.rCurrentTSO->link = suspended_ccalling_threads;
2108 suspended_ccalling_threads = cap->r.rCurrentTSO;
2110 if(cap->r.rCurrentTSO->blocked_exceptions == NULL) {
2111 cap->r.rCurrentTSO->why_blocked = BlockedOnCCall;
2112 cap->r.rCurrentTSO->blocked_exceptions = END_TSO_QUEUE;
2114 cap->r.rCurrentTSO->why_blocked = BlockedOnCCall_NoUnblockExc;
2117 /* Use the thread ID as the token; it should be unique */
2118 tok = cap->r.rCurrentTSO->id;
2120 /* Hand back capability */
2121 releaseCapability(cap);
2123 #if defined(RTS_SUPPORTS_THREADS)
2124 /* Preparing to leave the RTS, so ensure there's a native thread/task
2125 waiting to take over.
2127 IF_DEBUG(scheduler, sched_belch("worker (token %d): leaving RTS", tok));
2130 in_haskell = rtsFalse;
2131 RELEASE_LOCK(&sched_mutex);
2133 errno = saved_errno;
2138 resumeThread( StgInt tok )
2140 StgTSO *tso, **prev;
2142 int saved_errno = errno;
2144 #if defined(RTS_SUPPORTS_THREADS)
2145 /* Wait for permission to re-enter the RTS with the result. */
2146 ACQUIRE_LOCK(&sched_mutex);
2147 waitForReturnCapability(&sched_mutex, &cap);
2149 IF_DEBUG(scheduler, sched_belch("worker (token %d): re-entering RTS", tok));
2151 grabCapability(&cap);
2154 /* Remove the thread off of the suspended list */
2155 prev = &suspended_ccalling_threads;
2156 for (tso = suspended_ccalling_threads;
2157 tso != END_TSO_QUEUE;
2158 prev = &tso->link, tso = tso->link) {
2159 if (tso->id == (StgThreadID)tok) {
2164 if (tso == END_TSO_QUEUE) {
2165 barf("resumeThread: thread not found");
2167 tso->link = END_TSO_QUEUE;
2169 if(tso->why_blocked == BlockedOnCCall) {
2170 awakenBlockedQueueNoLock(tso->blocked_exceptions);
2171 tso->blocked_exceptions = NULL;
2174 /* Reset blocking status */
2175 tso->why_blocked = NotBlocked;
2177 cap->r.rCurrentTSO = tso;
2178 in_haskell = rtsTrue;
2179 RELEASE_LOCK(&sched_mutex);
2180 errno = saved_errno;
2184 /* ---------------------------------------------------------------------------
2185 * Comparing Thread ids.
2187 * This is used from STG land in the implementation of the
2188 * instances of Eq/Ord for ThreadIds.
2189 * ------------------------------------------------------------------------ */
2192 cmp_thread(StgPtr tso1, StgPtr tso2)
2194 StgThreadID id1 = ((StgTSO *)tso1)->id;
2195 StgThreadID id2 = ((StgTSO *)tso2)->id;
2197 if (id1 < id2) return (-1);
2198 if (id1 > id2) return 1;
2202 /* ---------------------------------------------------------------------------
2203 * Fetching the ThreadID from an StgTSO.
2205 * This is used in the implementation of Show for ThreadIds.
2206 * ------------------------------------------------------------------------ */
2208 rts_getThreadId(StgPtr tso)
2210 return ((StgTSO *)tso)->id;
2215 labelThread(StgPtr tso, char *label)
2220 /* Caveat: Once set, you can only set the thread name to "" */
2221 len = strlen(label)+1;
2222 buf = stgMallocBytes(len * sizeof(char), "Schedule.c:labelThread()");
2223 strncpy(buf,label,len);
2224 /* Update will free the old memory for us */
2225 updateThreadLabel(((StgTSO *)tso)->id,buf);
2229 /* ---------------------------------------------------------------------------
2230 Create a new thread.
2232 The new thread starts with the given stack size. Before the
2233 scheduler can run, however, this thread needs to have a closure
2234 (and possibly some arguments) pushed on its stack. See
2235 pushClosure() in Schedule.h.
2237 createGenThread() and createIOThread() (in SchedAPI.h) are
2238 convenient packaged versions of this function.
2240 currently pri (priority) is only used in a GRAN setup -- HWL
2241 ------------------------------------------------------------------------ */
2243 /* currently pri (priority) is only used in a GRAN setup -- HWL */
2245 createThread(nat size, StgInt pri)
2248 createThread(nat size)
2255 /* First check whether we should create a thread at all */
2256 #if defined(PARALLEL_HASKELL)
2257 /* check that no more than RtsFlags.ParFlags.maxThreads threads are created */
2258 if (advisory_thread_count >= RtsFlags.ParFlags.maxThreads) {
2260 debugBelch("{createThread}Daq ghuH: refusing to create another thread; no more than %d threads allowed (currently %d)\n",
2261 RtsFlags.ParFlags.maxThreads, advisory_thread_count);
2262 return END_TSO_QUEUE;
2268 ASSERT(!RtsFlags.GranFlags.Light || CurrentProc==0);
2271 // ToDo: check whether size = stack_size - TSO_STRUCT_SIZEW
2273 /* catch ridiculously small stack sizes */
2274 if (size < MIN_STACK_WORDS + TSO_STRUCT_SIZEW) {
2275 size = MIN_STACK_WORDS + TSO_STRUCT_SIZEW;
2278 stack_size = size - TSO_STRUCT_SIZEW;
2280 tso = (StgTSO *)allocate(size);
2281 TICK_ALLOC_TSO(stack_size, 0);
2283 SET_HDR(tso, &stg_TSO_info, CCS_SYSTEM);
2285 SET_GRAN_HDR(tso, ThisPE);
2288 // Always start with the compiled code evaluator
2289 tso->what_next = ThreadRunGHC;
2291 tso->id = next_thread_id++;
2292 tso->why_blocked = NotBlocked;
2293 tso->blocked_exceptions = NULL;
2295 tso->saved_errno = 0;
2298 tso->stack_size = stack_size;
2299 tso->max_stack_size = round_to_mblocks(RtsFlags.GcFlags.maxStkSize)
2301 tso->sp = (P_)&(tso->stack) + stack_size;
2303 tso->trec = NO_TREC;
2306 tso->prof.CCCS = CCS_MAIN;
2309 /* put a stop frame on the stack */
2310 tso->sp -= sizeofW(StgStopFrame);
2311 SET_HDR((StgClosure*)tso->sp,(StgInfoTable *)&stg_stop_thread_info,CCS_SYSTEM);
2312 tso->link = END_TSO_QUEUE;
2316 /* uses more flexible routine in GranSim */
2317 insertThread(tso, CurrentProc);
2319 /* In a non-GranSim setup the pushing of a TSO onto the runq is separated
2325 if (RtsFlags.GranFlags.GranSimStats.Full)
2326 DumpGranEvent(GR_START,tso);
2327 #elif defined(PARALLEL_HASKELL)
2328 if (RtsFlags.ParFlags.ParStats.Full)
2329 DumpGranEvent(GR_STARTQ,tso);
2330 /* HACk to avoid SCHEDULE
2334 /* Link the new thread on the global thread list.
2336 tso->global_link = all_threads;
2340 tso->dist.priority = MandatoryPriority; //by default that is...
2344 tso->gran.pri = pri;
2346 tso->gran.magic = TSO_MAGIC; // debugging only
2348 tso->gran.sparkname = 0;
2349 tso->gran.startedat = CURRENT_TIME;
2350 tso->gran.exported = 0;
2351 tso->gran.basicblocks = 0;
2352 tso->gran.allocs = 0;
2353 tso->gran.exectime = 0;
2354 tso->gran.fetchtime = 0;
2355 tso->gran.fetchcount = 0;
2356 tso->gran.blocktime = 0;
2357 tso->gran.blockcount = 0;
2358 tso->gran.blockedat = 0;
2359 tso->gran.globalsparks = 0;
2360 tso->gran.localsparks = 0;
2361 if (RtsFlags.GranFlags.Light)
2362 tso->gran.clock = Now; /* local clock */
2364 tso->gran.clock = 0;
2366 IF_DEBUG(gran,printTSO(tso));
2367 #elif defined(PARALLEL_HASKELL)
2369 tso->par.magic = TSO_MAGIC; // debugging only
2371 tso->par.sparkname = 0;
2372 tso->par.startedat = CURRENT_TIME;
2373 tso->par.exported = 0;
2374 tso->par.basicblocks = 0;
2375 tso->par.allocs = 0;
2376 tso->par.exectime = 0;
2377 tso->par.fetchtime = 0;
2378 tso->par.fetchcount = 0;
2379 tso->par.blocktime = 0;
2380 tso->par.blockcount = 0;
2381 tso->par.blockedat = 0;
2382 tso->par.globalsparks = 0;
2383 tso->par.localsparks = 0;
2387 globalGranStats.tot_threads_created++;
2388 globalGranStats.threads_created_on_PE[CurrentProc]++;
2389 globalGranStats.tot_sq_len += spark_queue_len(CurrentProc);
2390 globalGranStats.tot_sq_probes++;
2391 #elif defined(PARALLEL_HASKELL)
2392 // collect parallel global statistics (currently done together with GC stats)
2393 if (RtsFlags.ParFlags.ParStats.Global &&
2394 RtsFlags.GcFlags.giveStats > NO_GC_STATS) {
2395 //debugBelch("Creating thread %d @ %11.2f\n", tso->id, usertime());
2396 globalParStats.tot_threads_created++;
2402 sched_belch("==__ schedule: Created TSO %d (%p);",
2403 CurrentProc, tso, tso->id));
2404 #elif defined(PARALLEL_HASKELL)
2405 IF_PAR_DEBUG(verbose,
2406 sched_belch("==__ schedule: Created TSO %d (%p); %d threads active",
2407 (long)tso->id, tso, advisory_thread_count));
2409 IF_DEBUG(scheduler,sched_belch("created thread %ld, stack size = %lx words",
2410 (long)tso->id, (long)tso->stack_size));
2417 all parallel thread creation calls should fall through the following routine.
2420 createThreadFromSpark(rtsSpark spark)
2422 ASSERT(spark != (rtsSpark)NULL);
2423 // JB: TAKE CARE OF THIS COUNTER! BUGGY
2424 if (advisory_thread_count >= RtsFlags.ParFlags.maxThreads)
2426 barf("{createSparkThread}Daq ghuH: refusing to create another thread; no more than %d threads allowed (currently %d)",
2427 RtsFlags.ParFlags.maxThreads, advisory_thread_count);
2428 return END_TSO_QUEUE;
2432 tso = createThread(RtsFlags.GcFlags.initialStkSize);
2433 if (tso==END_TSO_QUEUE)
2434 barf("createSparkThread: Cannot create TSO");
2436 tso->priority = AdvisoryPriority;
2438 pushClosure(tso,spark);
2440 advisory_thread_count++; // JB: TAKE CARE OF THIS COUNTER! BUGGY
2447 Turn a spark into a thread.
2448 ToDo: fix for SMP (needs to acquire SCHED_MUTEX!)
2452 activateSpark (rtsSpark spark)
2456 tso = createSparkThread(spark);
2457 if (RtsFlags.ParFlags.ParStats.Full) {
2458 //ASSERT(run_queue_hd == END_TSO_QUEUE); // I think ...
2459 IF_PAR_DEBUG(verbose,
2460 debugBelch("==^^ activateSpark: turning spark of closure %p (%s) into a thread\n",
2461 (StgClosure *)spark, info_type((StgClosure *)spark)));
2463 // ToDo: fwd info on local/global spark to thread -- HWL
2464 // tso->gran.exported = spark->exported;
2465 // tso->gran.locked = !spark->global;
2466 // tso->gran.sparkname = spark->name;
2472 /* ---------------------------------------------------------------------------
2475 * scheduleThread puts a thread on the head of the runnable queue.
2476 * This will usually be done immediately after a thread is created.
2477 * The caller of scheduleThread must create the thread using e.g.
2478 * createThread and push an appropriate closure
2479 * on this thread's stack before the scheduler is invoked.
2480 * ------------------------------------------------------------------------ */
2483 scheduleThread_(StgTSO *tso)
2485 // The thread goes at the *end* of the run-queue, to avoid possible
2486 // starvation of any threads already on the queue.
2487 APPEND_TO_RUN_QUEUE(tso);
2492 scheduleThread(StgTSO* tso)
2494 ACQUIRE_LOCK(&sched_mutex);
2495 scheduleThread_(tso);
2496 RELEASE_LOCK(&sched_mutex);
2499 #if defined(RTS_SUPPORTS_THREADS)
2500 static Condition bound_cond_cache;
2501 static int bound_cond_cache_full = 0;
2506 scheduleWaitThread(StgTSO* tso, /*[out]*/HaskellObj* ret,
2507 Capability *initialCapability)
2509 // Precondition: sched_mutex must be held
2512 m = stgMallocBytes(sizeof(StgMainThread), "waitThread");
2517 m->link = main_threads;
2519 if (main_threads != NULL) {
2520 main_threads->prev = m;
2524 #if defined(RTS_SUPPORTS_THREADS)
2525 // Allocating a new condition for each thread is expensive, so we
2526 // cache one. This is a pretty feeble hack, but it helps speed up
2527 // consecutive call-ins quite a bit.
2528 if (bound_cond_cache_full) {
2529 m->bound_thread_cond = bound_cond_cache;
2530 bound_cond_cache_full = 0;
2532 initCondition(&m->bound_thread_cond);
2536 /* Put the thread on the main-threads list prior to scheduling the TSO.
2537 Failure to do so introduces a race condition in the MT case (as
2538 identified by Wolfgang Thaller), whereby the new task/OS thread
2539 created by scheduleThread_() would complete prior to the thread
2540 that spawned it managed to put 'itself' on the main-threads list.
2541 The upshot of it all being that the worker thread wouldn't get to
2542 signal the completion of the its work item for the main thread to
2543 see (==> it got stuck waiting.) -- sof 6/02.
2545 IF_DEBUG(scheduler, sched_belch("waiting for thread (%d)", tso->id));
2547 APPEND_TO_RUN_QUEUE(tso);
2548 // NB. Don't call threadRunnable() here, because the thread is
2549 // bound and only runnable by *this* OS thread, so waking up other
2550 // workers will just slow things down.
2552 return waitThread_(m, initialCapability);
2555 /* ---------------------------------------------------------------------------
2558 * Initialise the scheduler. This resets all the queues - if the
2559 * queues contained any threads, they'll be garbage collected at the
2562 * ------------------------------------------------------------------------ */
2570 for (i=0; i<=MAX_PROC; i++) {
2571 run_queue_hds[i] = END_TSO_QUEUE;
2572 run_queue_tls[i] = END_TSO_QUEUE;
2573 blocked_queue_hds[i] = END_TSO_QUEUE;
2574 blocked_queue_tls[i] = END_TSO_QUEUE;
2575 ccalling_threadss[i] = END_TSO_QUEUE;
2576 blackhole_queue[i] = END_TSO_QUEUE;
2577 sleeping_queue = END_TSO_QUEUE;
2580 run_queue_hd = END_TSO_QUEUE;
2581 run_queue_tl = END_TSO_QUEUE;
2582 blocked_queue_hd = END_TSO_QUEUE;
2583 blocked_queue_tl = END_TSO_QUEUE;
2584 blackhole_queue = END_TSO_QUEUE;
2585 sleeping_queue = END_TSO_QUEUE;
2588 suspended_ccalling_threads = END_TSO_QUEUE;
2590 main_threads = NULL;
2591 all_threads = END_TSO_QUEUE;
2596 RtsFlags.ConcFlags.ctxtSwitchTicks =
2597 RtsFlags.ConcFlags.ctxtSwitchTime / TICK_MILLISECS;
2599 #if defined(RTS_SUPPORTS_THREADS)
2600 /* Initialise the mutex and condition variables used by
2602 initMutex(&sched_mutex);
2603 initMutex(&term_mutex);
2606 ACQUIRE_LOCK(&sched_mutex);
2608 /* A capability holds the state a native thread needs in
2609 * order to execute STG code. At least one capability is
2610 * floating around (only SMP builds have more than one).
2614 #if defined(RTS_SUPPORTS_THREADS)
2619 /* eagerly start some extra workers */
2620 startTasks(RtsFlags.ParFlags.nNodes, taskStart);
2623 #if /* defined(SMP) ||*/ defined(PARALLEL_HASKELL)
2627 RELEASE_LOCK(&sched_mutex);
2631 exitScheduler( void )
2633 interrupted = rtsTrue;
2634 shutting_down_scheduler = rtsTrue;
2635 #if defined(RTS_SUPPORTS_THREADS)
2636 if (threadIsTask(osThreadId())) { taskStop(); }
2641 /* ----------------------------------------------------------------------------
2642 Managing the per-task allocation areas.
2644 Each capability comes with an allocation area. These are
2645 fixed-length block lists into which allocation can be done.
2647 ToDo: no support for two-space collection at the moment???
2648 ------------------------------------------------------------------------- */
2650 static SchedulerStatus
2651 waitThread_(StgMainThread* m, Capability *initialCapability)
2653 SchedulerStatus stat;
2655 // Precondition: sched_mutex must be held.
2656 IF_DEBUG(scheduler, sched_belch("new main thread (%d)", m->tso->id));
2659 /* GranSim specific init */
2660 CurrentTSO = m->tso; // the TSO to run
2661 procStatus[MainProc] = Busy; // status of main PE
2662 CurrentProc = MainProc; // PE to run it on
2663 schedule(m,initialCapability);
2665 schedule(m,initialCapability);
2666 ASSERT(m->stat != NoStatus);
2671 #if defined(RTS_SUPPORTS_THREADS)
2672 // Free the condition variable, returning it to the cache if possible.
2673 if (!bound_cond_cache_full) {
2674 bound_cond_cache = m->bound_thread_cond;
2675 bound_cond_cache_full = 1;
2677 closeCondition(&m->bound_thread_cond);
2681 IF_DEBUG(scheduler, sched_belch("main thread (%d) finished", m->tso->id));
2684 // Postcondition: sched_mutex still held
2688 /* ---------------------------------------------------------------------------
2689 Where are the roots that we know about?
2691 - all the threads on the runnable queue
2692 - all the threads on the blocked queue
2693 - all the threads on the sleeping queue
2694 - all the thread currently executing a _ccall_GC
2695 - all the "main threads"
2697 ------------------------------------------------------------------------ */
2699 /* This has to be protected either by the scheduler monitor, or by the
2700 garbage collection monitor (probably the latter).
2705 GetRoots( evac_fn evac )
2710 for (i=0; i<=RtsFlags.GranFlags.proc; i++) {
2711 if ((run_queue_hds[i] != END_TSO_QUEUE) && ((run_queue_hds[i] != NULL)))
2712 evac((StgClosure **)&run_queue_hds[i]);
2713 if ((run_queue_tls[i] != END_TSO_QUEUE) && ((run_queue_tls[i] != NULL)))
2714 evac((StgClosure **)&run_queue_tls[i]);
2716 if ((blocked_queue_hds[i] != END_TSO_QUEUE) && ((blocked_queue_hds[i] != NULL)))
2717 evac((StgClosure **)&blocked_queue_hds[i]);
2718 if ((blocked_queue_tls[i] != END_TSO_QUEUE) && ((blocked_queue_tls[i] != NULL)))
2719 evac((StgClosure **)&blocked_queue_tls[i]);
2720 if ((ccalling_threadss[i] != END_TSO_QUEUE) && ((ccalling_threadss[i] != NULL)))
2721 evac((StgClosure **)&ccalling_threads[i]);
2728 if (run_queue_hd != END_TSO_QUEUE) {
2729 ASSERT(run_queue_tl != END_TSO_QUEUE);
2730 evac((StgClosure **)&run_queue_hd);
2731 evac((StgClosure **)&run_queue_tl);
2734 if (blocked_queue_hd != END_TSO_QUEUE) {
2735 ASSERT(blocked_queue_tl != END_TSO_QUEUE);
2736 evac((StgClosure **)&blocked_queue_hd);
2737 evac((StgClosure **)&blocked_queue_tl);
2740 if (sleeping_queue != END_TSO_QUEUE) {
2741 evac((StgClosure **)&sleeping_queue);
2745 if (blackhole_queue != END_TSO_QUEUE) {
2746 evac((StgClosure **)&blackhole_queue);
2749 if (suspended_ccalling_threads != END_TSO_QUEUE) {
2750 evac((StgClosure **)&suspended_ccalling_threads);
2753 #if defined(PARALLEL_HASKELL) || defined(GRAN)
2754 markSparkQueue(evac);
2757 #if defined(RTS_USER_SIGNALS)
2758 // mark the signal handlers (signals should be already blocked)
2759 markSignalHandlers(evac);
2763 /* -----------------------------------------------------------------------------
2766 This is the interface to the garbage collector from Haskell land.
2767 We provide this so that external C code can allocate and garbage
2768 collect when called from Haskell via _ccall_GC.
2770 It might be useful to provide an interface whereby the programmer
2771 can specify more roots (ToDo).
2773 This needs to be protected by the GC condition variable above. KH.
2774 -------------------------------------------------------------------------- */
2776 static void (*extra_roots)(evac_fn);
2781 /* Obligated to hold this lock upon entry */
2782 ACQUIRE_LOCK(&sched_mutex);
2783 GarbageCollect(GetRoots,rtsFalse);
2784 RELEASE_LOCK(&sched_mutex);
2788 performMajorGC(void)
2790 ACQUIRE_LOCK(&sched_mutex);
2791 GarbageCollect(GetRoots,rtsTrue);
2792 RELEASE_LOCK(&sched_mutex);
2796 AllRoots(evac_fn evac)
2798 GetRoots(evac); // the scheduler's roots
2799 extra_roots(evac); // the user's roots
2803 performGCWithRoots(void (*get_roots)(evac_fn))
2805 ACQUIRE_LOCK(&sched_mutex);
2806 extra_roots = get_roots;
2807 GarbageCollect(AllRoots,rtsFalse);
2808 RELEASE_LOCK(&sched_mutex);
2811 /* -----------------------------------------------------------------------------
2814 If the thread has reached its maximum stack size, then raise the
2815 StackOverflow exception in the offending thread. Otherwise
2816 relocate the TSO into a larger chunk of memory and adjust its stack
2818 -------------------------------------------------------------------------- */
2821 threadStackOverflow(StgTSO *tso)
2823 nat new_stack_size, stack_words;
2828 IF_DEBUG(sanity,checkTSO(tso));
2829 if (tso->stack_size >= tso->max_stack_size) {
2832 debugBelch("@@ threadStackOverflow of TSO %ld (%p): stack too large (now %ld; max is %ld)\n",
2833 (long)tso->id, tso, (long)tso->stack_size, (long)tso->max_stack_size);
2834 /* If we're debugging, just print out the top of the stack */
2835 printStackChunk(tso->sp, stg_min(tso->stack+tso->stack_size,
2838 /* Send this thread the StackOverflow exception */
2839 raiseAsync(tso, (StgClosure *)stackOverflow_closure);
2843 /* Try to double the current stack size. If that takes us over the
2844 * maximum stack size for this thread, then use the maximum instead.
2845 * Finally round up so the TSO ends up as a whole number of blocks.
2847 new_stack_size = stg_min(tso->stack_size * 2, tso->max_stack_size);
2848 new_tso_size = (lnat)BLOCK_ROUND_UP(new_stack_size * sizeof(W_) +
2849 TSO_STRUCT_SIZE)/sizeof(W_);
2850 new_tso_size = round_to_mblocks(new_tso_size); /* Be MBLOCK-friendly */
2851 new_stack_size = new_tso_size - TSO_STRUCT_SIZEW;
2853 IF_DEBUG(scheduler, debugBelch("== sched: increasing stack size from %d words to %d.\n", tso->stack_size, new_stack_size));
2855 dest = (StgTSO *)allocate(new_tso_size);
2856 TICK_ALLOC_TSO(new_stack_size,0);
2858 /* copy the TSO block and the old stack into the new area */
2859 memcpy(dest,tso,TSO_STRUCT_SIZE);
2860 stack_words = tso->stack + tso->stack_size - tso->sp;
2861 new_sp = (P_)dest + new_tso_size - stack_words;
2862 memcpy(new_sp, tso->sp, stack_words * sizeof(W_));
2864 /* relocate the stack pointers... */
2866 dest->stack_size = new_stack_size;
2868 /* Mark the old TSO as relocated. We have to check for relocated
2869 * TSOs in the garbage collector and any primops that deal with TSOs.
2871 * It's important to set the sp value to just beyond the end
2872 * of the stack, so we don't attempt to scavenge any part of the
2875 tso->what_next = ThreadRelocated;
2877 tso->sp = (P_)&(tso->stack[tso->stack_size]);
2878 tso->why_blocked = NotBlocked;
2880 IF_PAR_DEBUG(verbose,
2881 debugBelch("@@ threadStackOverflow of TSO %d (now at %p): stack size increased to %ld\n",
2882 tso->id, tso, tso->stack_size);
2883 /* If we're debugging, just print out the top of the stack */
2884 printStackChunk(tso->sp, stg_min(tso->stack+tso->stack_size,
2887 IF_DEBUG(sanity,checkTSO(tso));
2889 IF_DEBUG(scheduler,printTSO(dest));
2895 /* ---------------------------------------------------------------------------
2896 Wake up a queue that was blocked on some resource.
2897 ------------------------------------------------------------------------ */
2901 unblockCount ( StgBlockingQueueElement *bqe, StgClosure *node )
2904 #elif defined(PARALLEL_HASKELL)
2906 unblockCount ( StgBlockingQueueElement *bqe, StgClosure *node )
2908 /* write RESUME events to log file and
2909 update blocked and fetch time (depending on type of the orig closure) */
2910 if (RtsFlags.ParFlags.ParStats.Full) {
2911 DumpRawGranEvent(CURRENT_PROC, CURRENT_PROC,
2912 GR_RESUMEQ, ((StgTSO *)bqe), ((StgTSO *)bqe)->block_info.closure,
2913 0, 0 /* spark_queue_len(ADVISORY_POOL) */);
2914 if (EMPTY_RUN_QUEUE())
2915 emitSchedule = rtsTrue;
2917 switch (get_itbl(node)->type) {
2919 ((StgTSO *)bqe)->par.fetchtime += CURRENT_TIME-((StgTSO *)bqe)->par.blockedat;
2924 ((StgTSO *)bqe)->par.blocktime += CURRENT_TIME-((StgTSO *)bqe)->par.blockedat;
2931 barf("{unblockOneLocked}Daq Qagh: unexpected closure in blocking queue");
2938 static StgBlockingQueueElement *
2939 unblockOneLocked(StgBlockingQueueElement *bqe, StgClosure *node)
2942 PEs node_loc, tso_loc;
2944 node_loc = where_is(node); // should be lifted out of loop
2945 tso = (StgTSO *)bqe; // wastes an assignment to get the type right
2946 tso_loc = where_is((StgClosure *)tso);
2947 if (IS_LOCAL_TO(PROCS(node),tso_loc)) { // TSO is local
2948 /* !fake_fetch => TSO is on CurrentProc is same as IS_LOCAL_TO */
2949 ASSERT(CurrentProc!=node_loc || tso_loc==CurrentProc);
2950 CurrentTime[CurrentProc] += RtsFlags.GranFlags.Costs.lunblocktime;
2951 // insertThread(tso, node_loc);
2952 new_event(tso_loc, tso_loc, CurrentTime[CurrentProc],
2954 tso, node, (rtsSpark*)NULL);
2955 tso->link = END_TSO_QUEUE; // overwrite link just to be sure
2958 } else { // TSO is remote (actually should be FMBQ)
2959 CurrentTime[CurrentProc] += RtsFlags.GranFlags.Costs.mpacktime +
2960 RtsFlags.GranFlags.Costs.gunblocktime +
2961 RtsFlags.GranFlags.Costs.latency;
2962 new_event(tso_loc, CurrentProc, CurrentTime[CurrentProc],
2964 tso, node, (rtsSpark*)NULL);
2965 tso->link = END_TSO_QUEUE; // overwrite link just to be sure
2968 /* the thread-queue-overhead is accounted for in either Resume or UnblockThread */
2970 debugBelch(" %s TSO %d (%p) [PE %d] (block_info.closure=%p) (next=%p) ,",
2971 (node_loc==tso_loc ? "Local" : "Global"),
2972 tso->id, tso, CurrentProc, tso->block_info.closure, tso->link));
2973 tso->block_info.closure = NULL;
2974 IF_DEBUG(scheduler,debugBelch("-- Waking up thread %ld (%p)\n",
2977 #elif defined(PARALLEL_HASKELL)
2978 static StgBlockingQueueElement *
2979 unblockOneLocked(StgBlockingQueueElement *bqe, StgClosure *node)
2981 StgBlockingQueueElement *next;
2983 switch (get_itbl(bqe)->type) {
2985 ASSERT(((StgTSO *)bqe)->why_blocked != NotBlocked);
2986 /* if it's a TSO just push it onto the run_queue */
2988 ((StgTSO *)bqe)->link = END_TSO_QUEUE; // debugging?
2989 APPEND_TO_RUN_QUEUE((StgTSO *)bqe);
2991 unblockCount(bqe, node);
2992 /* reset blocking status after dumping event */
2993 ((StgTSO *)bqe)->why_blocked = NotBlocked;
2997 /* if it's a BLOCKED_FETCH put it on the PendingFetches list */
2999 bqe->link = (StgBlockingQueueElement *)PendingFetches;
3000 PendingFetches = (StgBlockedFetch *)bqe;
3004 /* can ignore this case in a non-debugging setup;
3005 see comments on RBHSave closures above */
3007 /* check that the closure is an RBHSave closure */
3008 ASSERT(get_itbl((StgClosure *)bqe) == &stg_RBH_Save_0_info ||
3009 get_itbl((StgClosure *)bqe) == &stg_RBH_Save_1_info ||
3010 get_itbl((StgClosure *)bqe) == &stg_RBH_Save_2_info);
3014 barf("{unblockOneLocked}Daq Qagh: Unexpected IP (%#lx; %s) in blocking queue at %#lx\n",
3015 get_itbl((StgClosure *)bqe), info_type((StgClosure *)bqe),
3019 IF_PAR_DEBUG(bq, debugBelch(", %p (%s)\n", bqe, info_type((StgClosure*)bqe)));
3023 #else /* !GRAN && !PARALLEL_HASKELL */
3025 unblockOneLocked(StgTSO *tso)
3029 ASSERT(get_itbl(tso)->type == TSO);
3030 ASSERT(tso->why_blocked != NotBlocked);
3031 tso->why_blocked = NotBlocked;
3033 tso->link = END_TSO_QUEUE;
3034 APPEND_TO_RUN_QUEUE(tso);
3036 IF_DEBUG(scheduler,sched_belch("waking up thread %ld", (long)tso->id));
3041 #if defined(GRAN) || defined(PARALLEL_HASKELL)
3042 INLINE_ME StgBlockingQueueElement *
3043 unblockOne(StgBlockingQueueElement *bqe, StgClosure *node)
3045 ACQUIRE_LOCK(&sched_mutex);
3046 bqe = unblockOneLocked(bqe, node);
3047 RELEASE_LOCK(&sched_mutex);
3052 unblockOne(StgTSO *tso)
3054 ACQUIRE_LOCK(&sched_mutex);
3055 tso = unblockOneLocked(tso);
3056 RELEASE_LOCK(&sched_mutex);
3063 awakenBlockedQueue(StgBlockingQueueElement *q, StgClosure *node)
3065 StgBlockingQueueElement *bqe;
3070 debugBelch("##-_ AwBQ for node %p on PE %d @ %ld by TSO %d (%p): \n", \
3071 node, CurrentProc, CurrentTime[CurrentProc],
3072 CurrentTSO->id, CurrentTSO));
3074 node_loc = where_is(node);
3076 ASSERT(q == END_BQ_QUEUE ||
3077 get_itbl(q)->type == TSO || // q is either a TSO or an RBHSave
3078 get_itbl(q)->type == CONSTR); // closure (type constructor)
3079 ASSERT(is_unique(node));
3081 /* FAKE FETCH: magically copy the node to the tso's proc;
3082 no Fetch necessary because in reality the node should not have been
3083 moved to the other PE in the first place
3085 if (CurrentProc!=node_loc) {
3087 debugBelch("## node %p is on PE %d but CurrentProc is %d (TSO %d); assuming fake fetch and adjusting bitmask (old: %#x)\n",
3088 node, node_loc, CurrentProc, CurrentTSO->id,
3089 // CurrentTSO, where_is(CurrentTSO),
3090 node->header.gran.procs));
3091 node->header.gran.procs = (node->header.gran.procs) | PE_NUMBER(CurrentProc);
3093 debugBelch("## new bitmask of node %p is %#x\n",
3094 node, node->header.gran.procs));
3095 if (RtsFlags.GranFlags.GranSimStats.Global) {
3096 globalGranStats.tot_fake_fetches++;
3101 // ToDo: check: ASSERT(CurrentProc==node_loc);
3102 while (get_itbl(bqe)->type==TSO) { // q != END_TSO_QUEUE) {
3105 bqe points to the current element in the queue
3106 next points to the next element in the queue
3108 //tso = (StgTSO *)bqe; // wastes an assignment to get the type right
3109 //tso_loc = where_is(tso);
3111 bqe = unblockOneLocked(bqe, node);
3114 /* if this is the BQ of an RBH, we have to put back the info ripped out of
3115 the closure to make room for the anchor of the BQ */
3116 if (bqe!=END_BQ_QUEUE) {
3117 ASSERT(get_itbl(node)->type == RBH && get_itbl(bqe)->type == CONSTR);
3119 ASSERT((info_ptr==&RBH_Save_0_info) ||
3120 (info_ptr==&RBH_Save_1_info) ||
3121 (info_ptr==&RBH_Save_2_info));
3123 /* cf. convertToRBH in RBH.c for writing the RBHSave closure */
3124 ((StgRBH *)node)->blocking_queue = (StgBlockingQueueElement *)((StgRBHSave *)bqe)->payload[0];
3125 ((StgRBH *)node)->mut_link = (StgMutClosure *)((StgRBHSave *)bqe)->payload[1];
3128 debugBelch("## Filled in RBH_Save for %p (%s) at end of AwBQ\n",
3129 node, info_type(node)));
3132 /* statistics gathering */
3133 if (RtsFlags.GranFlags.GranSimStats.Global) {
3134 // globalGranStats.tot_bq_processing_time += bq_processing_time;
3135 globalGranStats.tot_bq_len += len; // total length of all bqs awakened
3136 // globalGranStats.tot_bq_len_local += len_local; // same for local TSOs only
3137 globalGranStats.tot_awbq++; // total no. of bqs awakened
3140 debugBelch("## BQ Stats of %p: [%d entries] %s\n",
3141 node, len, (bqe!=END_BQ_QUEUE) ? "RBH" : ""));
3143 #elif defined(PARALLEL_HASKELL)
3145 awakenBlockedQueue(StgBlockingQueueElement *q, StgClosure *node)
3147 StgBlockingQueueElement *bqe;
3149 ACQUIRE_LOCK(&sched_mutex);
3151 IF_PAR_DEBUG(verbose,
3152 debugBelch("##-_ AwBQ for node %p on [%x]: \n",
3156 if(get_itbl(q)->type == CONSTR || q==END_BQ_QUEUE) {
3157 IF_PAR_DEBUG(verbose, debugBelch("## ... nothing to unblock so lets just return. RFP (BUG?)\n"));
3162 ASSERT(q == END_BQ_QUEUE ||
3163 get_itbl(q)->type == TSO ||
3164 get_itbl(q)->type == BLOCKED_FETCH ||
3165 get_itbl(q)->type == CONSTR);
3168 while (get_itbl(bqe)->type==TSO ||
3169 get_itbl(bqe)->type==BLOCKED_FETCH) {
3170 bqe = unblockOneLocked(bqe, node);
3172 RELEASE_LOCK(&sched_mutex);
3175 #else /* !GRAN && !PARALLEL_HASKELL */
3178 awakenBlockedQueueNoLock(StgTSO *tso)
3180 while (tso != END_TSO_QUEUE) {
3181 tso = unblockOneLocked(tso);
3186 awakenBlockedQueue(StgTSO *tso)
3188 ACQUIRE_LOCK(&sched_mutex);
3189 while (tso != END_TSO_QUEUE) {
3190 tso = unblockOneLocked(tso);
3192 RELEASE_LOCK(&sched_mutex);
3196 /* ---------------------------------------------------------------------------
3198 - usually called inside a signal handler so it mustn't do anything fancy.
3199 ------------------------------------------------------------------------ */
3202 interruptStgRts(void)
3208 /* -----------------------------------------------------------------------------
3211 This is for use when we raise an exception in another thread, which
3213 This has nothing to do with the UnblockThread event in GranSim. -- HWL
3214 -------------------------------------------------------------------------- */
3216 #if defined(GRAN) || defined(PARALLEL_HASKELL)
3218 NB: only the type of the blocking queue is different in GranSim and GUM
3219 the operations on the queue-elements are the same
3220 long live polymorphism!
3222 Locks: sched_mutex is held upon entry and exit.
3226 unblockThread(StgTSO *tso)
3228 StgBlockingQueueElement *t, **last;
3230 switch (tso->why_blocked) {
3233 return; /* not blocked */
3236 // Be careful: nothing to do here! We tell the scheduler that the thread
3237 // is runnable and we leave it to the stack-walking code to abort the
3238 // transaction while unwinding the stack. We should perhaps have a debugging
3239 // test to make sure that this really happens and that the 'zombie' transaction
3240 // does not get committed.
3244 ASSERT(get_itbl(tso->block_info.closure)->type == MVAR);
3246 StgBlockingQueueElement *last_tso = END_BQ_QUEUE;
3247 StgMVar *mvar = (StgMVar *)(tso->block_info.closure);
3249 last = (StgBlockingQueueElement **)&mvar->head;
3250 for (t = (StgBlockingQueueElement *)mvar->head;
3252 last = &t->link, last_tso = t, t = t->link) {
3253 if (t == (StgBlockingQueueElement *)tso) {
3254 *last = (StgBlockingQueueElement *)tso->link;
3255 if (mvar->tail == tso) {
3256 mvar->tail = (StgTSO *)last_tso;
3261 barf("unblockThread (MVAR): TSO not found");
3264 case BlockedOnBlackHole:
3265 ASSERT(get_itbl(tso->block_info.closure)->type == BLACKHOLE_BQ);
3267 StgBlockingQueue *bq = (StgBlockingQueue *)(tso->block_info.closure);
3269 last = &bq->blocking_queue;
3270 for (t = bq->blocking_queue;
3272 last = &t->link, t = t->link) {
3273 if (t == (StgBlockingQueueElement *)tso) {
3274 *last = (StgBlockingQueueElement *)tso->link;
3278 barf("unblockThread (BLACKHOLE): TSO not found");
3281 case BlockedOnException:
3283 StgTSO *target = tso->block_info.tso;
3285 ASSERT(get_itbl(target)->type == TSO);
3287 if (target->what_next == ThreadRelocated) {
3288 target = target->link;
3289 ASSERT(get_itbl(target)->type == TSO);
3292 ASSERT(target->blocked_exceptions != NULL);
3294 last = (StgBlockingQueueElement **)&target->blocked_exceptions;
3295 for (t = (StgBlockingQueueElement *)target->blocked_exceptions;
3297 last = &t->link, t = t->link) {
3298 ASSERT(get_itbl(t)->type == TSO);
3299 if (t == (StgBlockingQueueElement *)tso) {
3300 *last = (StgBlockingQueueElement *)tso->link;
3304 barf("unblockThread (Exception): TSO not found");
3308 case BlockedOnWrite:
3309 #if defined(mingw32_HOST_OS)
3310 case BlockedOnDoProc:
3313 /* take TSO off blocked_queue */
3314 StgBlockingQueueElement *prev = NULL;
3315 for (t = (StgBlockingQueueElement *)blocked_queue_hd; t != END_BQ_QUEUE;
3316 prev = t, t = t->link) {
3317 if (t == (StgBlockingQueueElement *)tso) {
3319 blocked_queue_hd = (StgTSO *)t->link;
3320 if ((StgBlockingQueueElement *)blocked_queue_tl == t) {
3321 blocked_queue_tl = END_TSO_QUEUE;
3324 prev->link = t->link;
3325 if ((StgBlockingQueueElement *)blocked_queue_tl == t) {
3326 blocked_queue_tl = (StgTSO *)prev;
3332 barf("unblockThread (I/O): TSO not found");
3335 case BlockedOnDelay:
3337 /* take TSO off sleeping_queue */
3338 StgBlockingQueueElement *prev = NULL;
3339 for (t = (StgBlockingQueueElement *)sleeping_queue; t != END_BQ_QUEUE;
3340 prev = t, t = t->link) {
3341 if (t == (StgBlockingQueueElement *)tso) {
3343 sleeping_queue = (StgTSO *)t->link;
3345 prev->link = t->link;
3350 barf("unblockThread (delay): TSO not found");
3354 barf("unblockThread");
3358 tso->link = END_TSO_QUEUE;
3359 tso->why_blocked = NotBlocked;
3360 tso->block_info.closure = NULL;
3361 PUSH_ON_RUN_QUEUE(tso);
3365 unblockThread(StgTSO *tso)
3369 /* To avoid locking unnecessarily. */
3370 if (tso->why_blocked == NotBlocked) {
3374 switch (tso->why_blocked) {
3377 // Be careful: nothing to do here! We tell the scheduler that the thread
3378 // is runnable and we leave it to the stack-walking code to abort the
3379 // transaction while unwinding the stack. We should perhaps have a debugging
3380 // test to make sure that this really happens and that the 'zombie' transaction
3381 // does not get committed.
3385 ASSERT(get_itbl(tso->block_info.closure)->type == MVAR);
3387 StgTSO *last_tso = END_TSO_QUEUE;
3388 StgMVar *mvar = (StgMVar *)(tso->block_info.closure);
3391 for (t = mvar->head; t != END_TSO_QUEUE;
3392 last = &t->link, last_tso = t, t = t->link) {
3395 if (mvar->tail == tso) {
3396 mvar->tail = last_tso;
3401 barf("unblockThread (MVAR): TSO not found");
3404 case BlockedOnBlackHole:
3406 last = &blackhole_queue;
3407 for (t = blackhole_queue; t != END_TSO_QUEUE;
3408 last = &t->link, t = t->link) {
3414 barf("unblockThread (BLACKHOLE): TSO not found");
3417 case BlockedOnException:
3419 StgTSO *target = tso->block_info.tso;
3421 ASSERT(get_itbl(target)->type == TSO);
3423 while (target->what_next == ThreadRelocated) {
3424 target = target->link;
3425 ASSERT(get_itbl(target)->type == TSO);
3428 ASSERT(target->blocked_exceptions != NULL);
3430 last = &target->blocked_exceptions;
3431 for (t = target->blocked_exceptions; t != END_TSO_QUEUE;
3432 last = &t->link, t = t->link) {
3433 ASSERT(get_itbl(t)->type == TSO);
3439 barf("unblockThread (Exception): TSO not found");
3443 case BlockedOnWrite:
3444 #if defined(mingw32_HOST_OS)
3445 case BlockedOnDoProc:
3448 StgTSO *prev = NULL;
3449 for (t = blocked_queue_hd; t != END_TSO_QUEUE;
3450 prev = t, t = t->link) {
3453 blocked_queue_hd = t->link;
3454 if (blocked_queue_tl == t) {
3455 blocked_queue_tl = END_TSO_QUEUE;
3458 prev->link = t->link;
3459 if (blocked_queue_tl == t) {
3460 blocked_queue_tl = prev;
3466 barf("unblockThread (I/O): TSO not found");
3469 case BlockedOnDelay:
3471 StgTSO *prev = NULL;
3472 for (t = sleeping_queue; t != END_TSO_QUEUE;
3473 prev = t, t = t->link) {
3476 sleeping_queue = t->link;
3478 prev->link = t->link;
3483 barf("unblockThread (delay): TSO not found");
3487 barf("unblockThread");
3491 tso->link = END_TSO_QUEUE;
3492 tso->why_blocked = NotBlocked;
3493 tso->block_info.closure = NULL;
3494 APPEND_TO_RUN_QUEUE(tso);
3498 /* -----------------------------------------------------------------------------
3501 * Check the blackhole_queue for threads that can be woken up. We do
3502 * this periodically: before every GC, and whenever the run queue is
3505 * An elegant solution might be to just wake up all the blocked
3506 * threads with awakenBlockedQueue occasionally: they'll go back to
3507 * sleep again if the object is still a BLACKHOLE. Unfortunately this
3508 * doesn't give us a way to tell whether we've actually managed to
3509 * wake up any threads, so we would be busy-waiting.
3511 * -------------------------------------------------------------------------- */
3514 checkBlackHoles( void )
3517 rtsBool any_woke_up = rtsFalse;
3520 IF_DEBUG(scheduler, sched_belch("checking threads blocked on black holes"));
3522 // ASSUMES: sched_mutex
3523 prev = &blackhole_queue;
3524 t = blackhole_queue;
3525 while (t != END_TSO_QUEUE) {
3526 ASSERT(t->why_blocked == BlockedOnBlackHole);
3527 type = get_itbl(t->block_info.closure)->type;
3528 if (type != BLACKHOLE && type != CAF_BLACKHOLE) {
3529 t = unblockOneLocked(t);
3531 any_woke_up = rtsTrue;
3541 /* -----------------------------------------------------------------------------
3544 * The following function implements the magic for raising an
3545 * asynchronous exception in an existing thread.
3547 * We first remove the thread from any queue on which it might be
3548 * blocked. The possible blockages are MVARs and BLACKHOLE_BQs.
3550 * We strip the stack down to the innermost CATCH_FRAME, building
3551 * thunks in the heap for all the active computations, so they can
3552 * be restarted if necessary. When we reach a CATCH_FRAME, we build
3553 * an application of the handler to the exception, and push it on
3554 * the top of the stack.
3556 * How exactly do we save all the active computations? We create an
3557 * AP_STACK for every UpdateFrame on the stack. Entering one of these
3558 * AP_STACKs pushes everything from the corresponding update frame
3559 * upwards onto the stack. (Actually, it pushes everything up to the
3560 * next update frame plus a pointer to the next AP_STACK object.
3561 * Entering the next AP_STACK object pushes more onto the stack until we
3562 * reach the last AP_STACK object - at which point the stack should look
3563 * exactly as it did when we killed the TSO and we can continue
3564 * execution by entering the closure on top of the stack.
3566 * We can also kill a thread entirely - this happens if either (a) the
3567 * exception passed to raiseAsync is NULL, or (b) there's no
3568 * CATCH_FRAME on the stack. In either case, we strip the entire
3569 * stack and replace the thread with a zombie.
3571 * Locks: sched_mutex held upon entry nor exit.
3573 * -------------------------------------------------------------------------- */
3576 deleteThread(StgTSO *tso)
3578 if (tso->why_blocked != BlockedOnCCall &&
3579 tso->why_blocked != BlockedOnCCall_NoUnblockExc) {
3580 raiseAsync(tso,NULL);
3584 #ifdef FORKPROCESS_PRIMOP_SUPPORTED
3586 deleteThreadImmediately(StgTSO *tso)
3587 { // for forkProcess only:
3588 // delete thread without giving it a chance to catch the KillThread exception
3590 if (tso->what_next == ThreadComplete || tso->what_next == ThreadKilled) {
3594 if (tso->why_blocked != BlockedOnCCall &&
3595 tso->why_blocked != BlockedOnCCall_NoUnblockExc) {
3599 tso->what_next = ThreadKilled;
3604 raiseAsyncWithLock(StgTSO *tso, StgClosure *exception)
3606 /* When raising async exs from contexts where sched_mutex isn't held;
3607 use raiseAsyncWithLock(). */
3608 ACQUIRE_LOCK(&sched_mutex);
3609 raiseAsync(tso,exception);
3610 RELEASE_LOCK(&sched_mutex);
3614 raiseAsync(StgTSO *tso, StgClosure *exception)
3616 raiseAsync_(tso, exception, rtsFalse);
3620 raiseAsync_(StgTSO *tso, StgClosure *exception, rtsBool stop_at_atomically)
3622 StgRetInfoTable *info;
3625 // Thread already dead?
3626 if (tso->what_next == ThreadComplete || tso->what_next == ThreadKilled) {
3631 sched_belch("raising exception in thread %ld.", (long)tso->id));
3633 // Remove it from any blocking queues
3638 // The stack freezing code assumes there's a closure pointer on
3639 // the top of the stack, so we have to arrange that this is the case...
3641 if (sp[0] == (W_)&stg_enter_info) {
3645 sp[0] = (W_)&stg_dummy_ret_closure;
3651 // 1. Let the top of the stack be the "current closure"
3653 // 2. Walk up the stack until we find either an UPDATE_FRAME or a
3656 // 3. If it's an UPDATE_FRAME, then make an AP_STACK containing the
3657 // current closure applied to the chunk of stack up to (but not
3658 // including) the update frame. This closure becomes the "current
3659 // closure". Go back to step 2.
3661 // 4. If it's a CATCH_FRAME, then leave the exception handler on
3662 // top of the stack applied to the exception.
3664 // 5. If it's a STOP_FRAME, then kill the thread.
3666 // NB: if we pass an ATOMICALLY_FRAME then abort the associated
3673 info = get_ret_itbl((StgClosure *)frame);
3675 while (info->i.type != UPDATE_FRAME
3676 && (info->i.type != CATCH_FRAME || exception == NULL)
3677 && info->i.type != STOP_FRAME
3678 && (info->i.type != ATOMICALLY_FRAME || stop_at_atomically == rtsFalse))
3680 if (info->i.type == CATCH_RETRY_FRAME || info->i.type == ATOMICALLY_FRAME) {
3681 // IF we find an ATOMICALLY_FRAME then we abort the
3682 // current transaction and propagate the exception. In
3683 // this case (unlike ordinary exceptions) we do not care
3684 // whether the transaction is valid or not because its
3685 // possible validity cannot have caused the exception
3686 // and will not be visible after the abort.
3688 debugBelch("Found atomically block delivering async exception\n"));
3689 stmAbortTransaction(tso -> trec);
3690 tso -> trec = stmGetEnclosingTRec(tso -> trec);
3692 frame += stack_frame_sizeW((StgClosure *)frame);
3693 info = get_ret_itbl((StgClosure *)frame);
3696 switch (info->i.type) {
3698 case ATOMICALLY_FRAME:
3699 ASSERT(stop_at_atomically);
3700 ASSERT(stmGetEnclosingTRec(tso->trec) == NO_TREC);
3701 stmCondemnTransaction(tso -> trec);
3705 // R1 is not a register: the return convention for IO in
3706 // this case puts the return value on the stack, so we
3707 // need to set up the stack to return to the atomically
3708 // frame properly...
3709 tso->sp = frame - 2;
3710 tso->sp[1] = (StgWord) &stg_NO_FINALIZER_closure; // why not?
3711 tso->sp[0] = (StgWord) &stg_ut_1_0_unreg_info;
3713 tso->what_next = ThreadRunGHC;
3717 // If we find a CATCH_FRAME, and we've got an exception to raise,
3718 // then build the THUNK raise(exception), and leave it on
3719 // top of the CATCH_FRAME ready to enter.
3723 StgCatchFrame *cf = (StgCatchFrame *)frame;
3727 // we've got an exception to raise, so let's pass it to the
3728 // handler in this frame.
3730 raise = (StgClosure *)allocate(sizeofW(StgClosure)+1);
3731 TICK_ALLOC_SE_THK(1,0);
3732 SET_HDR(raise,&stg_raise_info,cf->header.prof.ccs);
3733 raise->payload[0] = exception;
3735 // throw away the stack from Sp up to the CATCH_FRAME.
3739 /* Ensure that async excpetions are blocked now, so we don't get
3740 * a surprise exception before we get around to executing the
3743 if (tso->blocked_exceptions == NULL) {
3744 tso->blocked_exceptions = END_TSO_QUEUE;
3747 /* Put the newly-built THUNK on top of the stack, ready to execute
3748 * when the thread restarts.
3751 sp[-1] = (W_)&stg_enter_info;
3753 tso->what_next = ThreadRunGHC;
3754 IF_DEBUG(sanity, checkTSO(tso));
3763 // First build an AP_STACK consisting of the stack chunk above the
3764 // current update frame, with the top word on the stack as the
3767 words = frame - sp - 1;
3768 ap = (StgAP_STACK *)allocate(PAP_sizeW(words));
3771 ap->fun = (StgClosure *)sp[0];
3773 for(i=0; i < (nat)words; ++i) {
3774 ap->payload[i] = (StgClosure *)*sp++;
3777 SET_HDR(ap,&stg_AP_STACK_info,
3778 ((StgClosure *)frame)->header.prof.ccs /* ToDo */);
3779 TICK_ALLOC_UP_THK(words+1,0);
3782 debugBelch("sched: Updating ");
3783 printPtr((P_)((StgUpdateFrame *)frame)->updatee);
3784 debugBelch(" with ");
3785 printObj((StgClosure *)ap);
3788 // Replace the updatee with an indirection - happily
3789 // this will also wake up any threads currently
3790 // waiting on the result.
3792 // Warning: if we're in a loop, more than one update frame on
3793 // the stack may point to the same object. Be careful not to
3794 // overwrite an IND_OLDGEN in this case, because we'll screw
3795 // up the mutable lists. To be on the safe side, don't
3796 // overwrite any kind of indirection at all. See also
3797 // threadSqueezeStack in GC.c, where we have to make a similar
3800 if (!closure_IND(((StgUpdateFrame *)frame)->updatee)) {
3801 // revert the black hole
3802 UPD_IND_NOLOCK(((StgUpdateFrame *)frame)->updatee,
3805 sp += sizeofW(StgUpdateFrame) - 1;
3806 sp[0] = (W_)ap; // push onto stack
3811 // We've stripped the entire stack, the thread is now dead.
3812 sp += sizeofW(StgStopFrame);
3813 tso->what_next = ThreadKilled;
3824 /* -----------------------------------------------------------------------------
3825 raiseExceptionHelper
3827 This function is called by the raise# primitve, just so that we can
3828 move some of the tricky bits of raising an exception from C-- into
3829 C. Who knows, it might be a useful re-useable thing here too.
3830 -------------------------------------------------------------------------- */
3833 raiseExceptionHelper (StgTSO *tso, StgClosure *exception)
3835 StgClosure *raise_closure = NULL;
3837 StgRetInfoTable *info;
3839 // This closure represents the expression 'raise# E' where E
3840 // is the exception raise. It is used to overwrite all the
3841 // thunks which are currently under evaluataion.
3845 // LDV profiling: stg_raise_info has THUNK as its closure
3846 // type. Since a THUNK takes at least MIN_UPD_SIZE words in its
3847 // payload, MIN_UPD_SIZE is more approprate than 1. It seems that
3848 // 1 does not cause any problem unless profiling is performed.
3849 // However, when LDV profiling goes on, we need to linearly scan
3850 // small object pool, where raise_closure is stored, so we should
3851 // use MIN_UPD_SIZE.
3853 // raise_closure = (StgClosure *)RET_STGCALL1(P_,allocate,
3854 // sizeofW(StgClosure)+1);
3858 // Walk up the stack, looking for the catch frame. On the way,
3859 // we update any closures pointed to from update frames with the
3860 // raise closure that we just built.
3864 info = get_ret_itbl((StgClosure *)p);
3865 next = p + stack_frame_sizeW((StgClosure *)p);
3866 switch (info->i.type) {
3869 // Only create raise_closure if we need to.
3870 if (raise_closure == NULL) {
3872 (StgClosure *)allocate(sizeofW(StgClosure)+MIN_UPD_SIZE);
3873 SET_HDR(raise_closure, &stg_raise_info, CCCS);
3874 raise_closure->payload[0] = exception;
3876 UPD_IND(((StgUpdateFrame *)p)->updatee,raise_closure);
3880 case ATOMICALLY_FRAME:
3881 IF_DEBUG(stm, debugBelch("Found ATOMICALLY_FRAME at %p\n", p));
3883 return ATOMICALLY_FRAME;
3889 case CATCH_STM_FRAME:
3890 IF_DEBUG(stm, debugBelch("Found CATCH_STM_FRAME at %p\n", p));
3892 return CATCH_STM_FRAME;
3898 case CATCH_RETRY_FRAME:
3907 /* -----------------------------------------------------------------------------
3908 findRetryFrameHelper
3910 This function is called by the retry# primitive. It traverses the stack
3911 leaving tso->sp referring to the frame which should handle the retry.
3913 This should either be a CATCH_RETRY_FRAME (if the retry# is within an orElse#)
3914 or should be a ATOMICALLY_FRAME (if the retry# reaches the top level).
3916 We skip CATCH_STM_FRAMEs because retries are not considered to be exceptions,
3917 despite the similar implementation.
3919 We should not expect to see CATCH_FRAME or STOP_FRAME because those should
3920 not be created within memory transactions.
3921 -------------------------------------------------------------------------- */
3924 findRetryFrameHelper (StgTSO *tso)
3927 StgRetInfoTable *info;
3931 info = get_ret_itbl((StgClosure *)p);
3932 next = p + stack_frame_sizeW((StgClosure *)p);
3933 switch (info->i.type) {
3935 case ATOMICALLY_FRAME:
3936 IF_DEBUG(stm, debugBelch("Found ATOMICALLY_FRAME at %p during retrry\n", p));
3938 return ATOMICALLY_FRAME;
3940 case CATCH_RETRY_FRAME:
3941 IF_DEBUG(stm, debugBelch("Found CATCH_RETRY_FRAME at %p during retrry\n", p));
3943 return CATCH_RETRY_FRAME;
3945 case CATCH_STM_FRAME:
3947 ASSERT(info->i.type != CATCH_FRAME);
3948 ASSERT(info->i.type != STOP_FRAME);
3955 /* -----------------------------------------------------------------------------
3956 resurrectThreads is called after garbage collection on the list of
3957 threads found to be garbage. Each of these threads will be woken
3958 up and sent a signal: BlockedOnDeadMVar if the thread was blocked
3959 on an MVar, or NonTermination if the thread was blocked on a Black
3962 Locks: sched_mutex isn't held upon entry nor exit.
3963 -------------------------------------------------------------------------- */
3966 resurrectThreads( StgTSO *threads )
3970 for (tso = threads; tso != END_TSO_QUEUE; tso = next) {
3971 next = tso->global_link;
3972 tso->global_link = all_threads;
3974 IF_DEBUG(scheduler, sched_belch("resurrecting thread %d", tso->id));
3976 switch (tso->why_blocked) {
3978 case BlockedOnException:
3979 /* Called by GC - sched_mutex lock is currently held. */
3980 raiseAsync(tso,(StgClosure *)BlockedOnDeadMVar_closure);
3982 case BlockedOnBlackHole:
3983 raiseAsync(tso,(StgClosure *)NonTermination_closure);
3986 raiseAsync(tso,(StgClosure *)BlockedIndefinitely_closure);
3989 /* This might happen if the thread was blocked on a black hole
3990 * belonging to a thread that we've just woken up (raiseAsync
3991 * can wake up threads, remember...).
3995 barf("resurrectThreads: thread blocked in a strange way");
4000 /* ----------------------------------------------------------------------------
4001 * Debugging: why is a thread blocked
4002 * [Also provides useful information when debugging threaded programs
4003 * at the Haskell source code level, so enable outside of DEBUG. --sof 7/02]
4004 ------------------------------------------------------------------------- */
4007 printThreadBlockage(StgTSO *tso)
4009 switch (tso->why_blocked) {
4011 debugBelch("is blocked on read from fd %ld", tso->block_info.fd);
4013 case BlockedOnWrite:
4014 debugBelch("is blocked on write to fd %ld", tso->block_info.fd);
4016 #if defined(mingw32_HOST_OS)
4017 case BlockedOnDoProc:
4018 debugBelch("is blocked on proc (request: %ld)", tso->block_info.async_result->reqID);
4021 case BlockedOnDelay:
4022 debugBelch("is blocked until %ld", tso->block_info.target);
4025 debugBelch("is blocked on an MVar");
4027 case BlockedOnException:
4028 debugBelch("is blocked on delivering an exception to thread %d",
4029 tso->block_info.tso->id);
4031 case BlockedOnBlackHole:
4032 debugBelch("is blocked on a black hole");
4035 debugBelch("is not blocked");
4037 #if defined(PARALLEL_HASKELL)
4039 debugBelch("is blocked on global address; local FM_BQ is %p (%s)",
4040 tso->block_info.closure, info_type(tso->block_info.closure));
4042 case BlockedOnGA_NoSend:
4043 debugBelch("is blocked on global address (no send); local FM_BQ is %p (%s)",
4044 tso->block_info.closure, info_type(tso->block_info.closure));
4047 case BlockedOnCCall:
4048 debugBelch("is blocked on an external call");
4050 case BlockedOnCCall_NoUnblockExc:
4051 debugBelch("is blocked on an external call (exceptions were already blocked)");
4054 debugBelch("is blocked on an STM operation");
4057 barf("printThreadBlockage: strange tso->why_blocked: %d for TSO %d (%d)",
4058 tso->why_blocked, tso->id, tso);
4063 printThreadStatus(StgTSO *tso)
4065 switch (tso->what_next) {
4067 debugBelch("has been killed");
4069 case ThreadComplete:
4070 debugBelch("has completed");
4073 printThreadBlockage(tso);
4078 printAllThreads(void)
4083 char time_string[TIME_STR_LEN], node_str[NODE_STR_LEN];
4084 ullong_format_string(TIME_ON_PROC(CurrentProc),
4085 time_string, rtsFalse/*no commas!*/);
4087 debugBelch("all threads at [%s]:\n", time_string);
4088 # elif defined(PARALLEL_HASKELL)
4089 char time_string[TIME_STR_LEN], node_str[NODE_STR_LEN];
4090 ullong_format_string(CURRENT_TIME,
4091 time_string, rtsFalse/*no commas!*/);
4093 debugBelch("all threads at [%s]:\n", time_string);
4095 debugBelch("all threads:\n");
4098 for (t = all_threads; t != END_TSO_QUEUE; t = t->global_link) {
4099 debugBelch("\tthread %d @ %p ", t->id, (void *)t);
4102 void *label = lookupThreadLabel(t->id);
4103 if (label) debugBelch("[\"%s\"] ",(char *)label);
4106 printThreadStatus(t);
4114 Print a whole blocking queue attached to node (debugging only).
4116 # if defined(PARALLEL_HASKELL)
4118 print_bq (StgClosure *node)
4120 StgBlockingQueueElement *bqe;
4124 debugBelch("## BQ of closure %p (%s): ",
4125 node, info_type(node));
4127 /* should cover all closures that may have a blocking queue */
4128 ASSERT(get_itbl(node)->type == BLACKHOLE_BQ ||
4129 get_itbl(node)->type == FETCH_ME_BQ ||
4130 get_itbl(node)->type == RBH ||
4131 get_itbl(node)->type == MVAR);
4133 ASSERT(node!=(StgClosure*)NULL); // sanity check
4135 print_bqe(((StgBlockingQueue*)node)->blocking_queue);
4139 Print a whole blocking queue starting with the element bqe.
4142 print_bqe (StgBlockingQueueElement *bqe)
4147 NB: In a parallel setup a BQ of an RBH must end with an RBH_Save closure;
4149 for (end = (bqe==END_BQ_QUEUE);
4150 !end; // iterate until bqe points to a CONSTR
4151 end = (get_itbl(bqe)->type == CONSTR) || (bqe->link==END_BQ_QUEUE),
4152 bqe = end ? END_BQ_QUEUE : bqe->link) {
4153 ASSERT(bqe != END_BQ_QUEUE); // sanity check
4154 ASSERT(bqe != (StgBlockingQueueElement *)NULL); // sanity check
4155 /* types of closures that may appear in a blocking queue */
4156 ASSERT(get_itbl(bqe)->type == TSO ||
4157 get_itbl(bqe)->type == BLOCKED_FETCH ||
4158 get_itbl(bqe)->type == CONSTR);
4159 /* only BQs of an RBH end with an RBH_Save closure */
4160 //ASSERT(get_itbl(bqe)->type != CONSTR || get_itbl(node)->type == RBH);
4162 switch (get_itbl(bqe)->type) {
4164 debugBelch(" TSO %u (%x),",
4165 ((StgTSO *)bqe)->id, ((StgTSO *)bqe));
4168 debugBelch(" BF (node=%p, ga=((%x, %d, %x)),",
4169 ((StgBlockedFetch *)bqe)->node,
4170 ((StgBlockedFetch *)bqe)->ga.payload.gc.gtid,
4171 ((StgBlockedFetch *)bqe)->ga.payload.gc.slot,
4172 ((StgBlockedFetch *)bqe)->ga.weight);
4175 debugBelch(" %s (IP %p),",
4176 (get_itbl(bqe) == &stg_RBH_Save_0_info ? "RBH_Save_0" :
4177 get_itbl(bqe) == &stg_RBH_Save_1_info ? "RBH_Save_1" :
4178 get_itbl(bqe) == &stg_RBH_Save_2_info ? "RBH_Save_2" :
4179 "RBH_Save_?"), get_itbl(bqe));
4182 barf("Unexpected closure type %s in blocking queue", // of %p (%s)",
4183 info_type((StgClosure *)bqe)); // , node, info_type(node));
4189 # elif defined(GRAN)
4191 print_bq (StgClosure *node)
4193 StgBlockingQueueElement *bqe;
4194 PEs node_loc, tso_loc;
4197 /* should cover all closures that may have a blocking queue */
4198 ASSERT(get_itbl(node)->type == BLACKHOLE_BQ ||
4199 get_itbl(node)->type == FETCH_ME_BQ ||
4200 get_itbl(node)->type == RBH);
4202 ASSERT(node!=(StgClosure*)NULL); // sanity check
4203 node_loc = where_is(node);
4205 debugBelch("## BQ of closure %p (%s) on [PE %d]: ",
4206 node, info_type(node), node_loc);
4209 NB: In a parallel setup a BQ of an RBH must end with an RBH_Save closure;
4211 for (bqe = ((StgBlockingQueue*)node)->blocking_queue, end = (bqe==END_BQ_QUEUE);
4212 !end; // iterate until bqe points to a CONSTR
4213 end = (get_itbl(bqe)->type == CONSTR) || (bqe->link==END_BQ_QUEUE), bqe = end ? END_BQ_QUEUE : bqe->link) {
4214 ASSERT(bqe != END_BQ_QUEUE); // sanity check
4215 ASSERT(bqe != (StgBlockingQueueElement *)NULL); // sanity check
4216 /* types of closures that may appear in a blocking queue */
4217 ASSERT(get_itbl(bqe)->type == TSO ||
4218 get_itbl(bqe)->type == CONSTR);
4219 /* only BQs of an RBH end with an RBH_Save closure */
4220 ASSERT(get_itbl(bqe)->type != CONSTR || get_itbl(node)->type == RBH);
4222 tso_loc = where_is((StgClosure *)bqe);
4223 switch (get_itbl(bqe)->type) {
4225 debugBelch(" TSO %d (%p) on [PE %d],",
4226 ((StgTSO *)bqe)->id, (StgTSO *)bqe, tso_loc);
4229 debugBelch(" %s (IP %p),",
4230 (get_itbl(bqe) == &stg_RBH_Save_0_info ? "RBH_Save_0" :
4231 get_itbl(bqe) == &stg_RBH_Save_1_info ? "RBH_Save_1" :
4232 get_itbl(bqe) == &stg_RBH_Save_2_info ? "RBH_Save_2" :
4233 "RBH_Save_?"), get_itbl(bqe));
4236 barf("Unexpected closure type %s in blocking queue of %p (%s)",
4237 info_type((StgClosure *)bqe), node, info_type(node));
4245 #if defined(PARALLEL_HASKELL)
4252 for (i=0, tso=run_queue_hd;
4253 tso != END_TSO_QUEUE;
4262 sched_belch(char *s, ...)
4266 #ifdef RTS_SUPPORTS_THREADS
4267 debugBelch("sched (task %p): ", osThreadId());
4268 #elif defined(PARALLEL_HASKELL)
4271 debugBelch("sched: ");