1 /* ---------------------------------------------------------------------------
2 * $Id: Schedule.c,v 1.146 2002/06/26 08:18:42 stolz Exp $
4 * (c) The GHC Team, 1998-2000
8 * Different GHC ways use this scheduler quite differently (see comments below)
9 * Here is the global picture:
11 * WAY Name CPP flag What's it for
12 * --------------------------------------
13 * mp GUM PAR Parallel execution on a distributed memory machine
14 * s SMP SMP Parallel execution on a shared memory machine
15 * mg GranSim GRAN Simulation of parallel execution
16 * md GUM/GdH DIST Distributed execution (based on GUM)
18 * --------------------------------------------------------------------------*/
20 //@node Main scheduling code, , ,
21 //@section Main scheduling code
24 * Version with scheduler monitor support for SMPs (WAY=s):
26 This design provides a high-level API to create and schedule threads etc.
27 as documented in the SMP design document.
29 It uses a monitor design controlled by a single mutex to exercise control
30 over accesses to shared data structures, and builds on the Posix threads
33 The majority of state is shared. In order to keep essential per-task state,
34 there is a Capability structure, which contains all the information
35 needed to run a thread: its STG registers, a pointer to its TSO, a
36 nursery etc. During STG execution, a pointer to the capability is
37 kept in a register (BaseReg).
39 In a non-SMP build, there is one global capability, namely MainRegTable.
43 * Version with support for distributed memory parallelism aka GUM (WAY=mp):
45 The main scheduling loop in GUM iterates until a finish message is received.
46 In that case a global flag @receivedFinish@ is set and this instance of
47 the RTS shuts down. See ghc/rts/parallel/HLComms.c:processMessages()
48 for the handling of incoming messages, such as PP_FINISH.
49 Note that in the parallel case we have a system manager that coordinates
50 different PEs, each of which are running one instance of the RTS.
51 See ghc/rts/parallel/SysMan.c for the main routine of the parallel program.
52 From this routine processes executing ghc/rts/Main.c are spawned. -- HWL
54 * Version with support for simulating parallel execution aka GranSim (WAY=mg):
56 The main scheduling code in GranSim is quite different from that in std
57 (concurrent) Haskell: while concurrent Haskell just iterates over the
58 threads in the runnable queue, GranSim is event driven, i.e. it iterates
59 over the events in the global event queue. -- HWL
64 //* Variables and Data structures::
65 //* Main scheduling loop::
66 //* Suspend and Resume::
68 //* Garbage Collextion Routines::
69 //* Blocking Queue Routines::
70 //* Exception Handling Routines::
71 //* Debugging Routines::
75 //@node Includes, Variables and Data structures, Main scheduling code, Main scheduling code
76 //@subsection Includes
78 #include "PosixSource.h"
85 #include "StgStartup.h"
88 #include "StgMiscClosures.h"
90 #include "Interpreter.h"
91 #include "Exception.h"
99 #include "ThreadLabels.h"
101 #include "Proftimer.h"
102 #include "ProfHeap.h"
104 #if defined(GRAN) || defined(PAR)
105 # include "GranSimRts.h"
106 # include "GranSim.h"
107 # include "ParallelRts.h"
108 # include "Parallel.h"
109 # include "ParallelDebug.h"
110 # include "FetchMe.h"
114 #include "Capability.h"
115 #include "OSThreads.h"
118 #ifdef HAVE_SYS_TYPES_H
119 #include <sys/types.h>
127 //@node Variables and Data structures, Prototypes, Includes, Main scheduling code
128 //@subsection Variables and Data structures
130 /* Main thread queue.
131 * Locks required: sched_mutex.
133 StgMainThread *main_threads;
136 * Locks required: sched_mutex.
140 StgTSO* ActiveTSO = NULL; /* for assigning system costs; GranSim-Light only */
141 /* rtsTime TimeOfNextEvent, EndOfTimeSlice; now in GranSim.c */
144 In GranSim we have a runnable and a blocked queue for each processor.
145 In order to minimise code changes new arrays run_queue_hds/tls
146 are created. run_queue_hd is then a short cut (macro) for
147 run_queue_hds[CurrentProc] (see GranSim.h).
150 StgTSO *run_queue_hds[MAX_PROC], *run_queue_tls[MAX_PROC];
151 StgTSO *blocked_queue_hds[MAX_PROC], *blocked_queue_tls[MAX_PROC];
152 StgTSO *ccalling_threadss[MAX_PROC];
153 /* We use the same global list of threads (all_threads) in GranSim as in
154 the std RTS (i.e. we are cheating). However, we don't use this list in
155 the GranSim specific code at the moment (so we are only potentially
160 StgTSO *run_queue_hd, *run_queue_tl;
161 StgTSO *blocked_queue_hd, *blocked_queue_tl;
162 StgTSO *sleeping_queue; /* perhaps replace with a hash table? */
166 /* Linked list of all threads.
167 * Used for detecting garbage collected threads.
171 /* When a thread performs a safe C call (_ccall_GC, using old
172 * terminology), it gets put on the suspended_ccalling_threads
173 * list. Used by the garbage collector.
175 static StgTSO *suspended_ccalling_threads;
177 static StgTSO *threadStackOverflow(StgTSO *tso);
179 /* KH: The following two flags are shared memory locations. There is no need
180 to lock them, since they are only unset at the end of a scheduler
184 /* flag set by signal handler to precipitate a context switch */
185 //@cindex context_switch
188 /* if this flag is set as well, give up execution */
189 //@cindex interrupted
192 /* Next thread ID to allocate.
193 * Locks required: thread_id_mutex
195 //@cindex next_thread_id
196 StgThreadID next_thread_id = 1;
199 * Pointers to the state of the current thread.
200 * Rule of thumb: if CurrentTSO != NULL, then we're running a Haskell
201 * thread. If CurrentTSO == NULL, then we're at the scheduler level.
204 /* The smallest stack size that makes any sense is:
205 * RESERVED_STACK_WORDS (so we can get back from the stack overflow)
206 * + sizeofW(StgStopFrame) (the stg_stop_thread_info frame)
207 * + 1 (the realworld token for an IO thread)
208 * + 1 (the closure to enter)
210 * A thread with this stack will bomb immediately with a stack
211 * overflow, which will increase its stack size.
214 #define MIN_STACK_WORDS (RESERVED_STACK_WORDS + sizeofW(StgStopFrame) + 2)
221 /* This is used in `TSO.h' and gcc 2.96 insists that this variable actually
222 * exists - earlier gccs apparently didn't.
230 * Set to TRUE when entering a shutdown state (via shutdownHaskellAndExit()) --
231 * in an MT setting, needed to signal that a worker thread shouldn't hang around
232 * in the scheduler when it is out of work.
234 static rtsBool shutting_down_scheduler = rtsFalse;
236 void addToBlockedQueue ( StgTSO *tso );
238 static void schedule ( void );
239 void interruptStgRts ( void );
241 static void detectBlackHoles ( void );
244 static void sched_belch(char *s, ...);
247 #if defined(RTS_SUPPORTS_THREADS)
248 /* ToDo: carefully document the invariants that go together
249 * with these synchronisation objects.
251 Mutex sched_mutex = INIT_MUTEX_VAR;
252 Mutex term_mutex = INIT_MUTEX_VAR;
255 * A heavyweight solution to the problem of protecting
256 * the thread_id from concurrent update.
258 Mutex thread_id_mutex = INIT_MUTEX_VAR;
262 static Condition gc_pending_cond = INIT_COND_VAR;
266 #endif /* RTS_SUPPORTS_THREADS */
270 rtsTime TimeOfLastYield;
271 rtsBool emitSchedule = rtsTrue;
275 char *whatNext_strs[] = {
283 char *threadReturnCode_strs[] = {
284 "HeapOverflow", /* might also be StackOverflow */
293 StgTSO * createSparkThread(rtsSpark spark);
294 StgTSO * activateSpark (rtsSpark spark);
298 * The thread state for the main thread.
299 // ToDo: check whether not needed any more
303 #if defined(PAR) || defined(RTS_SUPPORTS_THREADS)
304 static void taskStart(void);
315 //@node Main scheduling loop, Suspend and Resume, Prototypes, Main scheduling code
316 //@subsection Main scheduling loop
318 /* ---------------------------------------------------------------------------
319 Main scheduling loop.
321 We use round-robin scheduling, each thread returning to the
322 scheduler loop when one of these conditions is detected:
325 * timer expires (thread yields)
330 Locking notes: we acquire the scheduler lock once at the beginning
331 of the scheduler loop, and release it when
333 * running a thread, or
334 * waiting for work, or
335 * waiting for a GC to complete.
338 In a GranSim setup this loop iterates over the global event queue.
339 This revolves around the global event queue, which determines what
340 to do next. Therefore, it's more complicated than either the
341 concurrent or the parallel (GUM) setup.
344 GUM iterates over incoming messages.
345 It starts with nothing to do (thus CurrentTSO == END_TSO_QUEUE),
346 and sends out a fish whenever it has nothing to do; in-between
347 doing the actual reductions (shared code below) it processes the
348 incoming messages and deals with delayed operations
349 (see PendingFetches).
350 This is not the ugliest code you could imagine, but it's bloody close.
352 ------------------------------------------------------------------------ */
359 StgThreadReturnCode ret;
367 rtsBool receivedFinish = rtsFalse;
369 nat tp_size, sp_size; // stats only
372 rtsBool was_interrupted = rtsFalse;
374 ACQUIRE_LOCK(&sched_mutex);
376 #if defined(RTS_SUPPORTS_THREADS)
377 waitForWorkCapability(&sched_mutex, &cap, rtsFalse);
379 /* simply initialise it in the non-threaded case */
380 grabCapability(&cap);
384 /* set up first event to get things going */
385 /* ToDo: assign costs for system setup and init MainTSO ! */
386 new_event(CurrentProc, CurrentProc, CurrentTime[CurrentProc],
388 CurrentTSO, (StgClosure*)NULL, (rtsSpark*)NULL);
391 fprintf(stderr, "GRAN: Init CurrentTSO (in schedule) = %p\n", CurrentTSO);
392 G_TSO(CurrentTSO, 5));
394 if (RtsFlags.GranFlags.Light) {
395 /* Save current time; GranSim Light only */
396 CurrentTSO->gran.clock = CurrentTime[CurrentProc];
399 event = get_next_event();
401 while (event!=(rtsEvent*)NULL) {
402 /* Choose the processor with the next event */
403 CurrentProc = event->proc;
404 CurrentTSO = event->tso;
408 while (!receivedFinish) { /* set by processMessages */
409 /* when receiving PP_FINISH message */
416 IF_DEBUG(scheduler, printAllThreads());
418 #if defined(RTS_SUPPORTS_THREADS)
419 /* Check to see whether there are any worker threads
420 waiting to deposit external call results. If so,
421 yield our capability */
422 yieldToReturningWorker(&sched_mutex, &cap);
425 /* If we're interrupted (the user pressed ^C, or some other
426 * termination condition occurred), kill all the currently running
430 IF_DEBUG(scheduler, sched_belch("interrupted"));
432 interrupted = rtsFalse;
433 was_interrupted = rtsTrue;
436 /* Go through the list of main threads and wake up any
437 * clients whose computations have finished. ToDo: this
438 * should be done more efficiently without a linear scan
439 * of the main threads list, somehow...
441 #if defined(RTS_SUPPORTS_THREADS)
443 StgMainThread *m, **prev;
444 prev = &main_threads;
445 for (m = main_threads; m != NULL; m = m->link) {
446 switch (m->tso->what_next) {
449 *(m->ret) = (StgClosure *)m->tso->sp[0];
453 broadcastCondition(&m->wakeup);
455 removeThreadLabel(m->tso);
459 if (m->ret) *(m->ret) = NULL;
461 if (was_interrupted) {
462 m->stat = Interrupted;
466 broadcastCondition(&m->wakeup);
468 removeThreadLabel(m->tso);
477 #else /* not threaded */
480 /* in GUM do this only on the Main PE */
483 /* If our main thread has finished or been killed, return.
486 StgMainThread *m = main_threads;
487 if (m->tso->what_next == ThreadComplete
488 || m->tso->what_next == ThreadKilled) {
490 removeThreadLabel((StgWord)m->tso);
492 main_threads = main_threads->link;
493 if (m->tso->what_next == ThreadComplete) {
494 /* we finished successfully, fill in the return value */
495 if (m->ret) { *(m->ret) = (StgClosure *)m->tso->sp[0]; };
499 if (m->ret) { *(m->ret) = NULL; };
500 if (was_interrupted) {
501 m->stat = Interrupted;
511 /* Top up the run queue from our spark pool. We try to make the
512 * number of threads in the run queue equal to the number of
515 * Disable spark support in SMP for now, non-essential & requires
516 * a little bit of work to make it compile cleanly. -- sof 1/02.
518 #if 0 /* defined(SMP) */
520 nat n = getFreeCapabilities();
521 StgTSO *tso = run_queue_hd;
523 /* Count the run queue */
524 while (n > 0 && tso != END_TSO_QUEUE) {
531 spark = findSpark(rtsFalse);
533 break; /* no more sparks in the pool */
535 /* I'd prefer this to be done in activateSpark -- HWL */
536 /* tricky - it needs to hold the scheduler lock and
537 * not try to re-acquire it -- SDM */
538 createSparkThread(spark);
540 sched_belch("==^^ turning spark of closure %p into a thread",
541 (StgClosure *)spark));
544 /* We need to wake up the other tasks if we just created some
547 if (getFreeCapabilities() - n > 1) {
548 signalCondition( &thread_ready_cond );
553 /* check for signals each time around the scheduler */
554 #ifndef mingw32_TARGET_OS
555 if (signals_pending()) {
556 RELEASE_LOCK(&sched_mutex); /* ToDo: kill */
557 startSignalHandlers();
558 ACQUIRE_LOCK(&sched_mutex);
562 /* Check whether any waiting threads need to be woken up. If the
563 * run queue is empty, and there are no other tasks running, we
564 * can wait indefinitely for something to happen.
565 * ToDo: what if another client comes along & requests another
568 if ( !EMPTY_QUEUE(blocked_queue_hd) || !EMPTY_QUEUE(sleeping_queue) ) {
569 awaitEvent( EMPTY_RUN_QUEUE()
571 && allFreeCapabilities()
575 /* we can be interrupted while waiting for I/O... */
576 if (interrupted) continue;
579 * Detect deadlock: when we have no threads to run, there are no
580 * threads waiting on I/O or sleeping, and all the other tasks are
581 * waiting for work, we must have a deadlock of some description.
583 * We first try to find threads blocked on themselves (ie. black
584 * holes), and generate NonTermination exceptions where necessary.
586 * If no threads are black holed, we have a deadlock situation, so
587 * inform all the main threads.
590 if ( EMPTY_THREAD_QUEUES()
591 #if defined(RTS_SUPPORTS_THREADS)
592 && EMPTY_QUEUE(suspended_ccalling_threads)
595 && allFreeCapabilities()
599 IF_DEBUG(scheduler, sched_belch("deadlocked, forcing major GC..."));
600 #if defined(THREADED_RTS)
601 /* and SMP mode ..? */
602 releaseCapability(cap);
604 // Garbage collection can release some new threads due to
605 // either (a) finalizers or (b) threads resurrected because
606 // they are about to be send BlockedOnDeadMVar. Any threads
607 // thus released will be immediately runnable.
608 GarbageCollect(GetRoots,rtsTrue);
610 if ( !EMPTY_RUN_QUEUE() ) { goto not_deadlocked; }
613 sched_belch("still deadlocked, checking for black holes..."));
616 if ( !EMPTY_RUN_QUEUE() ) { goto not_deadlocked; }
618 #ifndef mingw32_TARGET_OS
619 /* If we have user-installed signal handlers, then wait
620 * for signals to arrive rather then bombing out with a
623 #if defined(RTS_SUPPORTS_THREADS)
624 if ( 0 ) { /* hmm..what to do? Simply stop waiting for
625 a signal with no runnable threads (or I/O
626 suspended ones) leads nowhere quick.
627 For now, simply shut down when we reach this
630 ToDo: define precisely under what conditions
631 the Scheduler should shut down in an MT setting.
634 if ( anyUserHandlers() ) {
637 sched_belch("still deadlocked, waiting for signals..."));
641 // we might be interrupted...
642 if (interrupted) { continue; }
644 if (signals_pending()) {
645 RELEASE_LOCK(&sched_mutex);
646 startSignalHandlers();
647 ACQUIRE_LOCK(&sched_mutex);
649 ASSERT(!EMPTY_RUN_QUEUE());
654 /* Probably a real deadlock. Send the current main thread the
655 * Deadlock exception (or in the SMP build, send *all* main
656 * threads the deadlock exception, since none of them can make
661 #if defined(RTS_SUPPORTS_THREADS)
662 for (m = main_threads; m != NULL; m = m->link) {
663 switch (m->tso->why_blocked) {
664 case BlockedOnBlackHole:
665 raiseAsync(m->tso, (StgClosure *)NonTermination_closure);
667 case BlockedOnException:
669 raiseAsync(m->tso, (StgClosure *)Deadlock_closure);
672 barf("deadlock: main thread blocked in a strange way");
677 switch (m->tso->why_blocked) {
678 case BlockedOnBlackHole:
679 raiseAsync(m->tso, (StgClosure *)NonTermination_closure);
681 case BlockedOnException:
683 raiseAsync(m->tso, (StgClosure *)Deadlock_closure);
686 barf("deadlock: main thread blocked in a strange way");
691 #if defined(RTS_SUPPORTS_THREADS)
692 /* ToDo: revisit conditions (and mechanism) for shutting
693 down a multi-threaded world */
694 IF_DEBUG(scheduler, sched_belch("all done, i think...shutting down."));
695 RELEASE_LOCK(&sched_mutex);
703 /* ToDo: add deadlock detection in GUM (similar to SMP) -- HWL */
707 /* If there's a GC pending, don't do anything until it has
711 IF_DEBUG(scheduler,sched_belch("waiting for GC"));
712 waitCondition( &gc_pending_cond, &sched_mutex );
716 #if defined(RTS_SUPPORTS_THREADS)
717 /* block until we've got a thread on the run queue and a free
721 if ( EMPTY_RUN_QUEUE() ) {
722 /* Give up our capability */
723 releaseCapability(cap);
725 /* If we're in the process of shutting down (& running the
726 * a batch of finalisers), don't wait around.
728 if ( shutting_down_scheduler ) {
729 RELEASE_LOCK(&sched_mutex);
732 IF_DEBUG(scheduler, sched_belch("thread %d: waiting for work", osThreadId()));
733 waitForWorkCapability(&sched_mutex, &cap, rtsTrue);
734 IF_DEBUG(scheduler, sched_belch("thread %d: work now available", osThreadId()));
739 if (RtsFlags.GranFlags.Light)
740 GranSimLight_enter_system(event, &ActiveTSO); // adjust ActiveTSO etc
742 /* adjust time based on time-stamp */
743 if (event->time > CurrentTime[CurrentProc] &&
744 event->evttype != ContinueThread)
745 CurrentTime[CurrentProc] = event->time;
747 /* Deal with the idle PEs (may issue FindWork or MoveSpark events) */
748 if (!RtsFlags.GranFlags.Light)
751 IF_DEBUG(gran, fprintf(stderr, "GRAN: switch by event-type\n"));
753 /* main event dispatcher in GranSim */
754 switch (event->evttype) {
755 /* Should just be continuing execution */
757 IF_DEBUG(gran, fprintf(stderr, "GRAN: doing ContinueThread\n"));
758 /* ToDo: check assertion
759 ASSERT(run_queue_hd != (StgTSO*)NULL &&
760 run_queue_hd != END_TSO_QUEUE);
762 /* Ignore ContinueThreads for fetching threads (if synchr comm) */
763 if (!RtsFlags.GranFlags.DoAsyncFetch &&
764 procStatus[CurrentProc]==Fetching) {
765 belch("ghuH: Spurious ContinueThread while Fetching ignored; TSO %d (%p) [PE %d]",
766 CurrentTSO->id, CurrentTSO, CurrentProc);
769 /* Ignore ContinueThreads for completed threads */
770 if (CurrentTSO->what_next == ThreadComplete) {
771 belch("ghuH: found a ContinueThread event for completed thread %d (%p) [PE %d] (ignoring ContinueThread)",
772 CurrentTSO->id, CurrentTSO, CurrentProc);
775 /* Ignore ContinueThreads for threads that are being migrated */
776 if (PROCS(CurrentTSO)==Nowhere) {
777 belch("ghuH: trying to run the migrating TSO %d (%p) [PE %d] (ignoring ContinueThread)",
778 CurrentTSO->id, CurrentTSO, CurrentProc);
781 /* The thread should be at the beginning of the run queue */
782 if (CurrentTSO!=run_queue_hds[CurrentProc]) {
783 belch("ghuH: TSO %d (%p) [PE %d] is not at the start of the run_queue when doing a ContinueThread",
784 CurrentTSO->id, CurrentTSO, CurrentProc);
785 break; // run the thread anyway
788 new_event(proc, proc, CurrentTime[proc],
790 (StgTSO*)NULL, (StgClosure*)NULL, (rtsSpark*)NULL);
792 */ /* Catches superfluous CONTINUEs -- should be unnecessary */
793 break; // now actually run the thread; DaH Qu'vam yImuHbej
796 do_the_fetchnode(event);
797 goto next_thread; /* handle next event in event queue */
800 do_the_globalblock(event);
801 goto next_thread; /* handle next event in event queue */
804 do_the_fetchreply(event);
805 goto next_thread; /* handle next event in event queue */
807 case UnblockThread: /* Move from the blocked queue to the tail of */
808 do_the_unblock(event);
809 goto next_thread; /* handle next event in event queue */
811 case ResumeThread: /* Move from the blocked queue to the tail of */
812 /* the runnable queue ( i.e. Qu' SImqa'lu') */
813 event->tso->gran.blocktime +=
814 CurrentTime[CurrentProc] - event->tso->gran.blockedat;
815 do_the_startthread(event);
816 goto next_thread; /* handle next event in event queue */
819 do_the_startthread(event);
820 goto next_thread; /* handle next event in event queue */
823 do_the_movethread(event);
824 goto next_thread; /* handle next event in event queue */
827 do_the_movespark(event);
828 goto next_thread; /* handle next event in event queue */
831 do_the_findwork(event);
832 goto next_thread; /* handle next event in event queue */
835 barf("Illegal event type %u\n", event->evttype);
838 /* This point was scheduler_loop in the old RTS */
840 IF_DEBUG(gran, belch("GRAN: after main switch"));
842 TimeOfLastEvent = CurrentTime[CurrentProc];
843 TimeOfNextEvent = get_time_of_next_event();
844 IgnoreEvents=(TimeOfNextEvent==0); // HWL HACK
845 // CurrentTSO = ThreadQueueHd;
847 IF_DEBUG(gran, belch("GRAN: time of next event is: %ld",
850 if (RtsFlags.GranFlags.Light)
851 GranSimLight_leave_system(event, &ActiveTSO);
853 EndOfTimeSlice = CurrentTime[CurrentProc]+RtsFlags.GranFlags.time_slice;
856 belch("GRAN: end of time-slice is %#lx", EndOfTimeSlice));
858 /* in a GranSim setup the TSO stays on the run queue */
860 /* Take a thread from the run queue. */
861 t = POP_RUN_QUEUE(); // take_off_run_queue(t);
864 fprintf(stderr, "GRAN: About to run current thread, which is\n");
867 context_switch = 0; // turned on via GranYield, checking events and time slice
870 DumpGranEvent(GR_SCHEDULE, t));
872 procStatus[CurrentProc] = Busy;
875 if (PendingFetches != END_BF_QUEUE) {
879 /* ToDo: phps merge with spark activation above */
880 /* check whether we have local work and send requests if we have none */
881 if (EMPTY_RUN_QUEUE()) { /* no runnable threads */
882 /* :-[ no local threads => look out for local sparks */
883 /* the spark pool for the current PE */
884 pool = &(MainRegTable.rSparks); // generalise to cap = &MainRegTable
885 if (advisory_thread_count < RtsFlags.ParFlags.maxThreads &&
886 pool->hd < pool->tl) {
888 * ToDo: add GC code check that we really have enough heap afterwards!!
890 * If we're here (no runnable threads) and we have pending
891 * sparks, we must have a space problem. Get enough space
892 * to turn one of those pending sparks into a
896 spark = findSpark(rtsFalse); /* get a spark */
897 if (spark != (rtsSpark) NULL) {
898 tso = activateSpark(spark); /* turn the spark into a thread */
899 IF_PAR_DEBUG(schedule,
900 belch("==== schedule: Created TSO %d (%p); %d threads active",
901 tso->id, tso, advisory_thread_count));
903 if (tso==END_TSO_QUEUE) { /* failed to activate spark->back to loop */
904 belch("==^^ failed to activate spark");
906 } /* otherwise fall through & pick-up new tso */
908 IF_PAR_DEBUG(verbose,
909 belch("==^^ no local sparks (spark pool contains only NFs: %d)",
910 spark_queue_len(pool)));
915 /* If we still have no work we need to send a FISH to get a spark
918 if (EMPTY_RUN_QUEUE()) {
919 /* =8-[ no local sparks => look for work on other PEs */
921 * We really have absolutely no work. Send out a fish
922 * (there may be some out there already), and wait for
923 * something to arrive. We clearly can't run any threads
924 * until a SCHEDULE or RESUME arrives, and so that's what
925 * we're hoping to see. (Of course, we still have to
926 * respond to other types of messages.)
928 TIME now = msTime() /*CURRENT_TIME*/;
929 IF_PAR_DEBUG(verbose,
930 belch("-- now=%ld", now));
931 IF_PAR_DEBUG(verbose,
932 if (outstandingFishes < RtsFlags.ParFlags.maxFishes &&
933 (last_fish_arrived_at!=0 &&
934 last_fish_arrived_at+RtsFlags.ParFlags.fishDelay > now)) {
935 belch("--$$ delaying FISH until %ld (last fish %ld, delay %ld, now %ld)",
936 last_fish_arrived_at+RtsFlags.ParFlags.fishDelay,
937 last_fish_arrived_at,
938 RtsFlags.ParFlags.fishDelay, now);
941 if (outstandingFishes < RtsFlags.ParFlags.maxFishes &&
942 (last_fish_arrived_at==0 ||
943 (last_fish_arrived_at+RtsFlags.ParFlags.fishDelay <= now))) {
944 /* outstandingFishes is set in sendFish, processFish;
945 avoid flooding system with fishes via delay */
947 sendFish(pe, mytid, NEW_FISH_AGE, NEW_FISH_HISTORY,
950 // Global statistics: count no. of fishes
951 if (RtsFlags.ParFlags.ParStats.Global &&
952 RtsFlags.GcFlags.giveStats > NO_GC_STATS) {
953 globalParStats.tot_fish_mess++;
957 receivedFinish = processMessages();
960 } else if (PacketsWaiting()) { /* Look for incoming messages */
961 receivedFinish = processMessages();
964 /* Now we are sure that we have some work available */
965 ASSERT(run_queue_hd != END_TSO_QUEUE);
967 /* Take a thread from the run queue, if we have work */
968 t = POP_RUN_QUEUE(); // take_off_run_queue(END_TSO_QUEUE);
969 IF_DEBUG(sanity,checkTSO(t));
971 /* ToDo: write something to the log-file
972 if (RTSflags.ParFlags.granSimStats && !sameThread)
973 DumpGranEvent(GR_SCHEDULE, RunnableThreadsHd);
977 /* the spark pool for the current PE */
978 pool = &(MainRegTable.rSparks); // generalise to cap = &MainRegTable
981 belch("--=^ %d threads, %d sparks on [%#x]",
982 run_queue_len(), spark_queue_len(pool), CURRENT_PROC));
985 if (0 && RtsFlags.ParFlags.ParStats.Full &&
986 t && LastTSO && t->id != LastTSO->id &&
987 LastTSO->why_blocked == NotBlocked &&
988 LastTSO->what_next != ThreadComplete) {
989 // if previously scheduled TSO not blocked we have to record the context switch
990 DumpVeryRawGranEvent(TimeOfLastYield, CURRENT_PROC, CURRENT_PROC,
991 GR_DESCHEDULE, LastTSO, (StgClosure *)NULL, 0, 0);
994 if (RtsFlags.ParFlags.ParStats.Full &&
995 (emitSchedule /* forced emit */ ||
996 (t && LastTSO && t->id != LastTSO->id))) {
998 we are running a different TSO, so write a schedule event to log file
999 NB: If we use fair scheduling we also have to write a deschedule
1000 event for LastTSO; with unfair scheduling we know that the
1001 previous tso has blocked whenever we switch to another tso, so
1002 we don't need it in GUM for now
1004 DumpRawGranEvent(CURRENT_PROC, CURRENT_PROC,
1005 GR_SCHEDULE, t, (StgClosure *)NULL, 0, 0);
1006 emitSchedule = rtsFalse;
1010 #else /* !GRAN && !PAR */
1012 /* grab a thread from the run queue */
1013 ASSERT(run_queue_hd != END_TSO_QUEUE);
1014 t = POP_RUN_QUEUE();
1015 // Sanity check the thread we're about to run. This can be
1016 // expensive if there is lots of thread switching going on...
1017 IF_DEBUG(sanity,checkTSO(t));
1020 cap->r.rCurrentTSO = t;
1022 /* context switches are now initiated by the timer signal, unless
1023 * the user specified "context switch as often as possible", with
1028 RtsFlags.ProfFlags.profileInterval == 0 ||
1030 (RtsFlags.ConcFlags.ctxtSwitchTicks == 0
1031 && (run_queue_hd != END_TSO_QUEUE
1032 || blocked_queue_hd != END_TSO_QUEUE
1033 || sleeping_queue != END_TSO_QUEUE)))
1038 RELEASE_LOCK(&sched_mutex);
1040 IF_DEBUG(scheduler, sched_belch("-->> Running TSO %ld (%p) %s ...",
1041 t->id, t, whatNext_strs[t->what_next]));
1044 startHeapProfTimer();
1047 /* +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ */
1048 /* Run the current thread
1050 switch (cap->r.rCurrentTSO->what_next) {
1052 case ThreadComplete:
1053 /* Thread already finished, return to scheduler. */
1054 ret = ThreadFinished;
1056 case ThreadEnterGHC:
1057 ret = StgRun((StgFunPtr) stg_enterStackTop, &cap->r);
1060 ret = StgRun((StgFunPtr) stg_returnToStackTop, &cap->r);
1062 case ThreadEnterInterp:
1063 ret = interpretBCO(cap);
1066 barf("schedule: invalid what_next field");
1068 /* +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ */
1070 /* Costs for the scheduler are assigned to CCS_SYSTEM */
1072 stopHeapProfTimer();
1076 ACQUIRE_LOCK(&sched_mutex);
1079 IF_DEBUG(scheduler,fprintf(stderr,"scheduler (task %ld): ", osThreadId()););
1080 #elif !defined(GRAN) && !defined(PAR)
1081 IF_DEBUG(scheduler,fprintf(stderr,"scheduler: "););
1083 t = cap->r.rCurrentTSO;
1086 /* HACK 675: if the last thread didn't yield, make sure to print a
1087 SCHEDULE event to the log file when StgRunning the next thread, even
1088 if it is the same one as before */
1090 TimeOfLastYield = CURRENT_TIME;
1096 IF_DEBUG(gran, DumpGranEvent(GR_DESCHEDULE, t));
1097 globalGranStats.tot_heapover++;
1099 globalParStats.tot_heapover++;
1102 // did the task ask for a large block?
1103 if (cap->r.rHpAlloc > BLOCK_SIZE_W) {
1104 // if so, get one and push it on the front of the nursery.
1108 blocks = (nat)BLOCK_ROUND_UP(cap->r.rHpAlloc * sizeof(W_)) / BLOCK_SIZE;
1110 IF_DEBUG(scheduler,belch("--<< thread %ld (%p; %s) stopped: requesting a large block (size %d)",
1112 whatNext_strs[t->what_next], blocks));
1114 // don't do this if it would push us over the
1115 // alloc_blocks_lim limit; we'll GC first.
1116 if (alloc_blocks + blocks < alloc_blocks_lim) {
1118 alloc_blocks += blocks;
1119 bd = allocGroup( blocks );
1121 // link the new group into the list
1122 bd->link = cap->r.rCurrentNursery;
1123 bd->u.back = cap->r.rCurrentNursery->u.back;
1124 if (cap->r.rCurrentNursery->u.back != NULL) {
1125 cap->r.rCurrentNursery->u.back->link = bd;
1127 ASSERT(g0s0->blocks == cap->r.rCurrentNursery &&
1128 g0s0->blocks == cap->r.rNursery);
1129 cap->r.rNursery = g0s0->blocks = bd;
1131 cap->r.rCurrentNursery->u.back = bd;
1133 // initialise it as a nursery block
1137 bd->free = bd->start;
1139 // don't forget to update the block count in g0s0.
1140 g0s0->n_blocks += blocks;
1141 ASSERT(countBlocks(g0s0->blocks) == g0s0->n_blocks);
1143 // now update the nursery to point to the new block
1144 cap->r.rCurrentNursery = bd;
1146 // we might be unlucky and have another thread get on the
1147 // run queue before us and steal the large block, but in that
1148 // case the thread will just end up requesting another large
1150 PUSH_ON_RUN_QUEUE(t);
1155 /* make all the running tasks block on a condition variable,
1156 * maybe set context_switch and wait till they all pile in,
1157 * then have them wait on a GC condition variable.
1159 IF_DEBUG(scheduler,belch("--<< thread %ld (%p; %s) stopped: HeapOverflow",
1160 t->id, t, whatNext_strs[t->what_next]));
1163 ASSERT(!is_on_queue(t,CurrentProc));
1165 /* Currently we emit a DESCHEDULE event before GC in GUM.
1166 ToDo: either add separate event to distinguish SYSTEM time from rest
1167 or just nuke this DESCHEDULE (and the following SCHEDULE) */
1168 if (0 && RtsFlags.ParFlags.ParStats.Full) {
1169 DumpRawGranEvent(CURRENT_PROC, CURRENT_PROC,
1170 GR_DESCHEDULE, t, (StgClosure *)NULL, 0, 0);
1171 emitSchedule = rtsTrue;
1175 ready_to_gc = rtsTrue;
1176 context_switch = 1; /* stop other threads ASAP */
1177 PUSH_ON_RUN_QUEUE(t);
1178 /* actual GC is done at the end of the while loop */
1184 DumpGranEvent(GR_DESCHEDULE, t));
1185 globalGranStats.tot_stackover++;
1188 // DumpGranEvent(GR_DESCHEDULE, t);
1189 globalParStats.tot_stackover++;
1191 IF_DEBUG(scheduler,belch("--<< thread %ld (%p; %s) stopped, StackOverflow",
1192 t->id, t, whatNext_strs[t->what_next]));
1193 /* just adjust the stack for this thread, then pop it back
1199 /* enlarge the stack */
1200 StgTSO *new_t = threadStackOverflow(t);
1202 /* This TSO has moved, so update any pointers to it from the
1203 * main thread stack. It better not be on any other queues...
1204 * (it shouldn't be).
1206 for (m = main_threads; m != NULL; m = m->link) {
1211 threadPaused(new_t);
1212 PUSH_ON_RUN_QUEUE(new_t);
1216 case ThreadYielding:
1219 DumpGranEvent(GR_DESCHEDULE, t));
1220 globalGranStats.tot_yields++;
1223 // DumpGranEvent(GR_DESCHEDULE, t);
1224 globalParStats.tot_yields++;
1226 /* put the thread back on the run queue. Then, if we're ready to
1227 * GC, check whether this is the last task to stop. If so, wake
1228 * up the GC thread. getThread will block during a GC until the
1232 if (t->what_next == ThreadEnterInterp) {
1233 /* ToDo: or maybe a timer expired when we were in Hugs?
1234 * or maybe someone hit ctrl-C
1236 belch("--<< thread %ld (%p; %s) stopped to switch to Hugs",
1237 t->id, t, whatNext_strs[t->what_next]);
1239 belch("--<< thread %ld (%p; %s) stopped, yielding",
1240 t->id, t, whatNext_strs[t->what_next]);
1247 //belch("&& Doing sanity check on yielding TSO %ld.", t->id);
1249 ASSERT(t->link == END_TSO_QUEUE);
1251 ASSERT(!is_on_queue(t,CurrentProc));
1254 //belch("&& Doing sanity check on all ThreadQueues (and their TSOs).");
1255 checkThreadQsSanity(rtsTrue));
1258 if (RtsFlags.ParFlags.doFairScheduling) {
1259 /* this does round-robin scheduling; good for concurrency */
1260 APPEND_TO_RUN_QUEUE(t);
1262 /* this does unfair scheduling; good for parallelism */
1263 PUSH_ON_RUN_QUEUE(t);
1266 /* this does round-robin scheduling; good for concurrency */
1267 APPEND_TO_RUN_QUEUE(t);
1270 /* add a ContinueThread event to actually process the thread */
1271 new_event(CurrentProc, CurrentProc, CurrentTime[CurrentProc],
1273 t, (StgClosure*)NULL, (rtsSpark*)NULL);
1275 belch("GRAN: eventq and runnableq after adding yielded thread to queue again:");
1284 belch("--<< thread %ld (%p; %s) stopped, blocking on node %p [PE %d] with BQ: ",
1285 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)));
1286 if (t->block_info.closure!=(StgClosure*)NULL) print_bq(t->block_info.closure));
1288 // ??? needed; should emit block before
1290 DumpGranEvent(GR_DESCHEDULE, t));
1291 prune_eventq(t, (StgClosure *)NULL); // prune ContinueThreads for t
1294 ASSERT(procStatus[CurrentProc]==Busy ||
1295 ((procStatus[CurrentProc]==Fetching) &&
1296 (t->block_info.closure!=(StgClosure*)NULL)));
1297 if (run_queue_hds[CurrentProc] == END_TSO_QUEUE &&
1298 !(!RtsFlags.GranFlags.DoAsyncFetch &&
1299 procStatus[CurrentProc]==Fetching))
1300 procStatus[CurrentProc] = Idle;
1304 belch("--<< thread %ld (%p; %s) stopped, blocking on node %p with BQ: ",
1305 t->id, t, whatNext_strs[t->what_next], t->block_info.closure));
1308 if (t->block_info.closure!=(StgClosure*)NULL)
1309 print_bq(t->block_info.closure));
1311 /* Send a fetch (if BlockedOnGA) and dump event to log file */
1314 /* whatever we schedule next, we must log that schedule */
1315 emitSchedule = rtsTrue;
1318 /* don't need to do anything. Either the thread is blocked on
1319 * I/O, in which case we'll have called addToBlockedQueue
1320 * previously, or it's blocked on an MVar or Blackhole, in which
1321 * case it'll be on the relevant queue already.
1324 fprintf(stderr, "--<< thread %d (%p) stopped: ", t->id, t);
1325 printThreadBlockage(t);
1326 fprintf(stderr, "\n"));
1328 /* Only for dumping event to log file
1329 ToDo: do I need this in GranSim, too?
1336 case ThreadFinished:
1337 /* Need to check whether this was a main thread, and if so, signal
1338 * the task that started it with the return value. If we have no
1339 * more main threads, we probably need to stop all the tasks until
1342 /* We also end up here if the thread kills itself with an
1343 * uncaught exception, see Exception.hc.
1345 IF_DEBUG(scheduler,belch("--++ thread %d (%p) finished", t->id, t));
1347 endThread(t, CurrentProc); // clean-up the thread
1349 /* For now all are advisory -- HWL */
1350 //if(t->priority==AdvisoryPriority) ??
1351 advisory_thread_count--;
1354 if(t->dist.priority==RevalPriority)
1358 if (RtsFlags.ParFlags.ParStats.Full &&
1359 !RtsFlags.ParFlags.ParStats.Suppressed)
1360 DumpEndEvent(CURRENT_PROC, t, rtsFalse /* not mandatory */);
1365 barf("schedule: invalid thread return code %d", (int)ret);
1369 if (RtsFlags.ProfFlags.profileInterval==0 || performHeapProfile) {
1370 GarbageCollect(GetRoots, rtsTrue);
1372 performHeapProfile = rtsFalse;
1373 ready_to_gc = rtsFalse; // we already GC'd
1379 && allFreeCapabilities()
1382 /* everybody back, start the GC.
1383 * Could do it in this thread, or signal a condition var
1384 * to do it in another thread. Either way, we need to
1385 * broadcast on gc_pending_cond afterward.
1387 #if defined(RTS_SUPPORTS_THREADS)
1388 IF_DEBUG(scheduler,sched_belch("doing GC"));
1390 GarbageCollect(GetRoots,rtsFalse);
1391 ready_to_gc = rtsFalse;
1393 broadcastCondition(&gc_pending_cond);
1396 /* add a ContinueThread event to continue execution of current thread */
1397 new_event(CurrentProc, CurrentProc, CurrentTime[CurrentProc],
1399 t, (StgClosure*)NULL, (rtsSpark*)NULL);
1401 fprintf(stderr, "GRAN: eventq and runnableq after Garbage collection:\n");
1409 IF_GRAN_DEBUG(unused,
1410 print_eventq(EventHd));
1412 event = get_next_event();
1415 /* ToDo: wait for next message to arrive rather than busy wait */
1418 } /* end of while(1) */
1420 IF_PAR_DEBUG(verbose,
1421 belch("== Leaving schedule() after having received Finish"));
1424 /* ---------------------------------------------------------------------------
1425 * Singleton fork(). Do not copy any running threads.
1426 * ------------------------------------------------------------------------- */
1428 StgInt forkProcess(StgTSO* tso) {
1430 #ifndef mingw32_TARGET_OS
1434 IF_DEBUG(scheduler,sched_belch("forking!"));
1437 if (pid) { /* parent */
1439 /* just return the pid */
1441 } else { /* child */
1442 /* wipe all other threads */
1444 tso->link = END_TSO_QUEUE;
1446 /* DO NOT TOUCH THE QUEUES directly because most of the code around
1447 us is picky about finding the threat still in its queue when
1448 handling the deleteThread() */
1450 for (t = all_threads; t != END_TSO_QUEUE; t = next) {
1452 if (t->id != tso->id) {
1459 barf("forkProcess#: primop not implemented for mingw32, sorry! (%u)\n", tso->id);
1460 /* pointlessly printing out the TSOs 'id' to avoid CC unused warning. */
1462 #endif /* mingw32 */
1465 /* ---------------------------------------------------------------------------
1466 * deleteAllThreads(): kill all the live threads.
1468 * This is used when we catch a user interrupt (^C), before performing
1469 * any necessary cleanups and running finalizers.
1471 * Locks: sched_mutex held.
1472 * ------------------------------------------------------------------------- */
1474 void deleteAllThreads ( void )
1477 IF_DEBUG(scheduler,sched_belch("deleting all threads"));
1478 for (t = all_threads; t != END_TSO_QUEUE; t = next) {
1479 next = t->global_link;
1482 run_queue_hd = run_queue_tl = END_TSO_QUEUE;
1483 blocked_queue_hd = blocked_queue_tl = END_TSO_QUEUE;
1484 sleeping_queue = END_TSO_QUEUE;
1487 /* startThread and insertThread are now in GranSim.c -- HWL */
1490 //@node Suspend and Resume, Run queue code, Main scheduling loop, Main scheduling code
1491 //@subsection Suspend and Resume
1493 /* ---------------------------------------------------------------------------
1494 * Suspending & resuming Haskell threads.
1496 * When making a "safe" call to C (aka _ccall_GC), the task gives back
1497 * its capability before calling the C function. This allows another
1498 * task to pick up the capability and carry on running Haskell
1499 * threads. It also means that if the C call blocks, it won't lock
1502 * The Haskell thread making the C call is put to sleep for the
1503 * duration of the call, on the susepended_ccalling_threads queue. We
1504 * give out a token to the task, which it can use to resume the thread
1505 * on return from the C function.
1506 * ------------------------------------------------------------------------- */
1509 suspendThread( StgRegTable *reg,
1511 #if !defined(RTS_SUPPORTS_THREADS) && !defined(DEBUG)
1519 /* assume that *reg is a pointer to the StgRegTable part
1522 cap = (Capability *)((void *)reg - sizeof(StgFunTable));
1524 ACQUIRE_LOCK(&sched_mutex);
1527 sched_belch("thread %d did a _ccall_gc (is_concurrent: %d)", cap->r.rCurrentTSO->id,concCall));
1529 threadPaused(cap->r.rCurrentTSO);
1530 cap->r.rCurrentTSO->link = suspended_ccalling_threads;
1531 suspended_ccalling_threads = cap->r.rCurrentTSO;
1533 #if defined(RTS_SUPPORTS_THREADS)
1534 cap->r.rCurrentTSO->why_blocked = BlockedOnCCall;
1537 /* Use the thread ID as the token; it should be unique */
1538 tok = cap->r.rCurrentTSO->id;
1540 /* Hand back capability */
1541 releaseCapability(cap);
1543 #if defined(RTS_SUPPORTS_THREADS)
1544 /* Preparing to leave the RTS, so ensure there's a native thread/task
1545 waiting to take over.
1547 ToDo: optimise this and only create a new task if there's a need
1548 for one (i.e., if there's only one Concurrent Haskell thread alive,
1549 there's no need to create a new task).
1551 IF_DEBUG(scheduler, sched_belch("worker thread (%d): leaving RTS", tok));
1553 startTask(taskStart);
1557 /* Other threads _might_ be available for execution; signal this */
1559 RELEASE_LOCK(&sched_mutex);
1564 resumeThread( StgInt tok,
1566 #if !defined(RTS_SUPPORTS_THREADS)
1571 StgTSO *tso, **prev;
1574 #if defined(RTS_SUPPORTS_THREADS)
1575 /* Wait for permission to re-enter the RTS with the result. */
1577 ACQUIRE_LOCK(&sched_mutex);
1578 grabReturnCapability(&sched_mutex, &cap);
1580 grabCapability(&cap);
1583 grabCapability(&cap);
1586 /* Remove the thread off of the suspended list */
1587 prev = &suspended_ccalling_threads;
1588 for (tso = suspended_ccalling_threads;
1589 tso != END_TSO_QUEUE;
1590 prev = &tso->link, tso = tso->link) {
1591 if (tso->id == (StgThreadID)tok) {
1596 if (tso == END_TSO_QUEUE) {
1597 barf("resumeThread: thread not found");
1599 tso->link = END_TSO_QUEUE;
1600 /* Reset blocking status */
1601 tso->why_blocked = NotBlocked;
1603 cap->r.rCurrentTSO = tso;
1604 RELEASE_LOCK(&sched_mutex);
1609 /* ---------------------------------------------------------------------------
1611 * ------------------------------------------------------------------------ */
1612 static void unblockThread(StgTSO *tso);
1614 /* ---------------------------------------------------------------------------
1615 * Comparing Thread ids.
1617 * This is used from STG land in the implementation of the
1618 * instances of Eq/Ord for ThreadIds.
1619 * ------------------------------------------------------------------------ */
1621 int cmp_thread(const StgTSO *tso1, const StgTSO *tso2)
1623 StgThreadID id1 = tso1->id;
1624 StgThreadID id2 = tso2->id;
1626 if (id1 < id2) return (-1);
1627 if (id1 > id2) return 1;
1631 /* ---------------------------------------------------------------------------
1632 * Fetching the ThreadID from an StgTSO.
1634 * This is used in the implementation of Show for ThreadIds.
1635 * ------------------------------------------------------------------------ */
1636 int rts_getThreadId(const StgTSO *tso)
1642 void labelThread(StgTSO *tso, char *label)
1647 /* Caveat: Once set, you can only set the thread name to "" */
1648 len = strlen(label)+1;
1651 fprintf(stderr,"insufficient memory for labelThread!\n");
1653 strncpy(buf,label,len);
1654 /* Update will free the old memory for us */
1655 updateThreadLabel((StgWord)tso,buf);
1659 /* ---------------------------------------------------------------------------
1660 Create a new thread.
1662 The new thread starts with the given stack size. Before the
1663 scheduler can run, however, this thread needs to have a closure
1664 (and possibly some arguments) pushed on its stack. See
1665 pushClosure() in Schedule.h.
1667 createGenThread() and createIOThread() (in SchedAPI.h) are
1668 convenient packaged versions of this function.
1670 currently pri (priority) is only used in a GRAN setup -- HWL
1671 ------------------------------------------------------------------------ */
1672 //@cindex createThread
1674 /* currently pri (priority) is only used in a GRAN setup -- HWL */
1676 createThread(nat size, StgInt pri)
1679 createThread(nat size)
1686 /* First check whether we should create a thread at all */
1688 /* check that no more than RtsFlags.ParFlags.maxThreads threads are created */
1689 if (advisory_thread_count >= RtsFlags.ParFlags.maxThreads) {
1691 belch("{createThread}Daq ghuH: refusing to create another thread; no more than %d threads allowed (currently %d)",
1692 RtsFlags.ParFlags.maxThreads, advisory_thread_count);
1693 return END_TSO_QUEUE;
1699 ASSERT(!RtsFlags.GranFlags.Light || CurrentProc==0);
1702 // ToDo: check whether size = stack_size - TSO_STRUCT_SIZEW
1704 /* catch ridiculously small stack sizes */
1705 if (size < MIN_STACK_WORDS + TSO_STRUCT_SIZEW) {
1706 size = MIN_STACK_WORDS + TSO_STRUCT_SIZEW;
1709 stack_size = size - TSO_STRUCT_SIZEW;
1711 tso = (StgTSO *)allocate(size);
1712 TICK_ALLOC_TSO(stack_size, 0);
1714 SET_HDR(tso, &stg_TSO_info, CCS_SYSTEM);
1716 SET_GRAN_HDR(tso, ThisPE);
1718 tso->what_next = ThreadEnterGHC;
1720 /* tso->id needs to be unique. For now we use a heavyweight mutex to
1721 * protect the increment operation on next_thread_id.
1722 * In future, we could use an atomic increment instead.
1724 ACQUIRE_LOCK(&thread_id_mutex);
1725 tso->id = next_thread_id++;
1726 RELEASE_LOCK(&thread_id_mutex);
1728 tso->why_blocked = NotBlocked;
1729 tso->blocked_exceptions = NULL;
1731 tso->stack_size = stack_size;
1732 tso->max_stack_size = round_to_mblocks(RtsFlags.GcFlags.maxStkSize)
1734 tso->sp = (P_)&(tso->stack) + stack_size;
1737 tso->prof.CCCS = CCS_MAIN;
1740 /* put a stop frame on the stack */
1741 tso->sp -= sizeofW(StgStopFrame);
1742 SET_HDR((StgClosure*)tso->sp,(StgInfoTable *)&stg_stop_thread_info,CCS_SYSTEM);
1743 tso->su = (StgUpdateFrame*)tso->sp;
1747 tso->link = END_TSO_QUEUE;
1748 /* uses more flexible routine in GranSim */
1749 insertThread(tso, CurrentProc);
1751 /* In a non-GranSim setup the pushing of a TSO onto the runq is separated
1757 if (RtsFlags.GranFlags.GranSimStats.Full)
1758 DumpGranEvent(GR_START,tso);
1760 if (RtsFlags.ParFlags.ParStats.Full)
1761 DumpGranEvent(GR_STARTQ,tso);
1762 /* HACk to avoid SCHEDULE
1766 /* Link the new thread on the global thread list.
1768 tso->global_link = all_threads;
1772 tso->dist.priority = MandatoryPriority; //by default that is...
1776 tso->gran.pri = pri;
1778 tso->gran.magic = TSO_MAGIC; // debugging only
1780 tso->gran.sparkname = 0;
1781 tso->gran.startedat = CURRENT_TIME;
1782 tso->gran.exported = 0;
1783 tso->gran.basicblocks = 0;
1784 tso->gran.allocs = 0;
1785 tso->gran.exectime = 0;
1786 tso->gran.fetchtime = 0;
1787 tso->gran.fetchcount = 0;
1788 tso->gran.blocktime = 0;
1789 tso->gran.blockcount = 0;
1790 tso->gran.blockedat = 0;
1791 tso->gran.globalsparks = 0;
1792 tso->gran.localsparks = 0;
1793 if (RtsFlags.GranFlags.Light)
1794 tso->gran.clock = Now; /* local clock */
1796 tso->gran.clock = 0;
1798 IF_DEBUG(gran,printTSO(tso));
1801 tso->par.magic = TSO_MAGIC; // debugging only
1803 tso->par.sparkname = 0;
1804 tso->par.startedat = CURRENT_TIME;
1805 tso->par.exported = 0;
1806 tso->par.basicblocks = 0;
1807 tso->par.allocs = 0;
1808 tso->par.exectime = 0;
1809 tso->par.fetchtime = 0;
1810 tso->par.fetchcount = 0;
1811 tso->par.blocktime = 0;
1812 tso->par.blockcount = 0;
1813 tso->par.blockedat = 0;
1814 tso->par.globalsparks = 0;
1815 tso->par.localsparks = 0;
1819 globalGranStats.tot_threads_created++;
1820 globalGranStats.threads_created_on_PE[CurrentProc]++;
1821 globalGranStats.tot_sq_len += spark_queue_len(CurrentProc);
1822 globalGranStats.tot_sq_probes++;
1824 // collect parallel global statistics (currently done together with GC stats)
1825 if (RtsFlags.ParFlags.ParStats.Global &&
1826 RtsFlags.GcFlags.giveStats > NO_GC_STATS) {
1827 //fprintf(stderr, "Creating thread %d @ %11.2f\n", tso->id, usertime());
1828 globalParStats.tot_threads_created++;
1834 belch("==__ schedule: Created TSO %d (%p);",
1835 CurrentProc, tso, tso->id));
1837 IF_PAR_DEBUG(verbose,
1838 belch("==__ schedule: Created TSO %d (%p); %d threads active",
1839 tso->id, tso, advisory_thread_count));
1841 IF_DEBUG(scheduler,sched_belch("created thread %ld, stack size = %lx words",
1842 tso->id, tso->stack_size));
1849 all parallel thread creation calls should fall through the following routine.
1852 createSparkThread(rtsSpark spark)
1854 ASSERT(spark != (rtsSpark)NULL);
1855 if (advisory_thread_count >= RtsFlags.ParFlags.maxThreads)
1857 barf("{createSparkThread}Daq ghuH: refusing to create another thread; no more than %d threads allowed (currently %d)",
1858 RtsFlags.ParFlags.maxThreads, advisory_thread_count);
1859 return END_TSO_QUEUE;
1863 tso = createThread(RtsFlags.GcFlags.initialStkSize);
1864 if (tso==END_TSO_QUEUE)
1865 barf("createSparkThread: Cannot create TSO");
1867 tso->priority = AdvisoryPriority;
1869 pushClosure(tso,spark);
1870 PUSH_ON_RUN_QUEUE(tso);
1871 advisory_thread_count++;
1878 Turn a spark into a thread.
1879 ToDo: fix for SMP (needs to acquire SCHED_MUTEX!)
1882 //@cindex activateSpark
1884 activateSpark (rtsSpark spark)
1888 tso = createSparkThread(spark);
1889 if (RtsFlags.ParFlags.ParStats.Full) {
1890 //ASSERT(run_queue_hd == END_TSO_QUEUE); // I think ...
1891 IF_PAR_DEBUG(verbose,
1892 belch("==^^ activateSpark: turning spark of closure %p (%s) into a thread",
1893 (StgClosure *)spark, info_type((StgClosure *)spark)));
1895 // ToDo: fwd info on local/global spark to thread -- HWL
1896 // tso->gran.exported = spark->exported;
1897 // tso->gran.locked = !spark->global;
1898 // tso->gran.sparkname = spark->name;
1904 static SchedulerStatus waitThread_(/*out*/StgMainThread* m
1905 #if defined(THREADED_RTS)
1906 , rtsBool blockWaiting
1911 /* ---------------------------------------------------------------------------
1914 * scheduleThread puts a thread on the head of the runnable queue.
1915 * This will usually be done immediately after a thread is created.
1916 * The caller of scheduleThread must create the thread using e.g.
1917 * createThread and push an appropriate closure
1918 * on this thread's stack before the scheduler is invoked.
1919 * ------------------------------------------------------------------------ */
1921 static void scheduleThread_ (StgTSO* tso, rtsBool createTask);
1924 scheduleThread_(StgTSO *tso
1925 , rtsBool createTask
1926 #if !defined(THREADED_RTS)
1931 ACQUIRE_LOCK(&sched_mutex);
1933 /* Put the new thread on the head of the runnable queue. The caller
1934 * better push an appropriate closure on this thread's stack
1935 * beforehand. In the SMP case, the thread may start running as
1936 * soon as we release the scheduler lock below.
1938 PUSH_ON_RUN_QUEUE(tso);
1939 #if defined(THREADED_RTS)
1940 /* If main() is scheduling a thread, don't bother creating a
1944 startTask(taskStart);
1950 IF_DEBUG(scheduler,printTSO(tso));
1952 RELEASE_LOCK(&sched_mutex);
1955 void scheduleThread(StgTSO* tso)
1957 scheduleThread_(tso, rtsFalse);
1961 scheduleWaitThread(StgTSO* tso, /*[out]*/HaskellObj* ret)
1965 m = stgMallocBytes(sizeof(StgMainThread), "waitThread");
1969 #if defined(RTS_SUPPORTS_THREADS)
1970 initCondition(&m->wakeup);
1973 /* Put the thread on the main-threads list prior to scheduling the TSO.
1974 Failure to do so introduces a race condition in the MT case (as
1975 identified by Wolfgang Thaller), whereby the new task/OS thread
1976 created by scheduleThread_() would complete prior to the thread
1977 that spawned it managed to put 'itself' on the main-threads list.
1978 The upshot of it all being that the worker thread wouldn't get to
1979 signal the completion of the its work item for the main thread to
1980 see (==> it got stuck waiting.) -- sof 6/02.
1982 ACQUIRE_LOCK(&sched_mutex);
1983 IF_DEBUG(scheduler, sched_belch("== scheduler: waiting for thread (%d)\n", tso->id));
1985 m->link = main_threads;
1988 /* Inefficient (scheduleThread_() acquires it again right away),
1989 * but obviously correct.
1991 RELEASE_LOCK(&sched_mutex);
1993 scheduleThread_(tso, rtsTrue);
1994 #if defined(THREADED_RTS)
1995 return waitThread_(m, rtsTrue);
1997 return waitThread_(m);
2001 /* ---------------------------------------------------------------------------
2004 * Initialise the scheduler. This resets all the queues - if the
2005 * queues contained any threads, they'll be garbage collected at the
2008 * ------------------------------------------------------------------------ */
2012 term_handler(int sig STG_UNUSED)
2015 ACQUIRE_LOCK(&term_mutex);
2017 RELEASE_LOCK(&term_mutex);
2028 for (i=0; i<=MAX_PROC; i++) {
2029 run_queue_hds[i] = END_TSO_QUEUE;
2030 run_queue_tls[i] = END_TSO_QUEUE;
2031 blocked_queue_hds[i] = END_TSO_QUEUE;
2032 blocked_queue_tls[i] = END_TSO_QUEUE;
2033 ccalling_threadss[i] = END_TSO_QUEUE;
2034 sleeping_queue = END_TSO_QUEUE;
2037 run_queue_hd = END_TSO_QUEUE;
2038 run_queue_tl = END_TSO_QUEUE;
2039 blocked_queue_hd = END_TSO_QUEUE;
2040 blocked_queue_tl = END_TSO_QUEUE;
2041 sleeping_queue = END_TSO_QUEUE;
2044 suspended_ccalling_threads = END_TSO_QUEUE;
2046 main_threads = NULL;
2047 all_threads = END_TSO_QUEUE;
2052 RtsFlags.ConcFlags.ctxtSwitchTicks =
2053 RtsFlags.ConcFlags.ctxtSwitchTime / TICK_MILLISECS;
2055 #if defined(RTS_SUPPORTS_THREADS)
2056 /* Initialise the mutex and condition variables used by
2058 initMutex(&sched_mutex);
2059 initMutex(&term_mutex);
2060 initMutex(&thread_id_mutex);
2062 initCondition(&thread_ready_cond);
2066 initCondition(&gc_pending_cond);
2069 #if defined(RTS_SUPPORTS_THREADS)
2070 ACQUIRE_LOCK(&sched_mutex);
2073 /* Install the SIGHUP handler */
2076 struct sigaction action,oact;
2078 action.sa_handler = term_handler;
2079 sigemptyset(&action.sa_mask);
2080 action.sa_flags = 0;
2081 if (sigaction(SIGTERM, &action, &oact) != 0) {
2082 barf("can't install TERM handler");
2087 /* A capability holds the state a native thread needs in
2088 * order to execute STG code. At least one capability is
2089 * floating around (only SMP builds have more than one).
2093 #if defined(RTS_SUPPORTS_THREADS)
2094 /* start our haskell execution tasks */
2096 startTaskManager(RtsFlags.ParFlags.nNodes, taskStart);
2098 startTaskManager(0,taskStart);
2102 #if /* defined(SMP) ||*/ defined(PAR)
2106 #if defined(RTS_SUPPORTS_THREADS)
2107 RELEASE_LOCK(&sched_mutex);
2113 exitScheduler( void )
2115 #if defined(RTS_SUPPORTS_THREADS)
2118 shutting_down_scheduler = rtsTrue;
2121 /* -----------------------------------------------------------------------------
2122 Managing the per-task allocation areas.
2124 Each capability comes with an allocation area. These are
2125 fixed-length block lists into which allocation can be done.
2127 ToDo: no support for two-space collection at the moment???
2128 -------------------------------------------------------------------------- */
2130 /* -----------------------------------------------------------------------------
2131 * waitThread is the external interface for running a new computation
2132 * and waiting for the result.
2134 * In the non-SMP case, we create a new main thread, push it on the
2135 * main-thread stack, and invoke the scheduler to run it. The
2136 * scheduler will return when the top main thread on the stack has
2137 * completed or died, and fill in the necessary fields of the
2138 * main_thread structure.
2140 * In the SMP case, we create a main thread as before, but we then
2141 * create a new condition variable and sleep on it. When our new
2142 * main thread has completed, we'll be woken up and the status/result
2143 * will be in the main_thread struct.
2144 * -------------------------------------------------------------------------- */
2147 howManyThreadsAvail ( void )
2151 for (q = run_queue_hd; q != END_TSO_QUEUE; q = q->link)
2153 for (q = blocked_queue_hd; q != END_TSO_QUEUE; q = q->link)
2155 for (q = sleeping_queue; q != END_TSO_QUEUE; q = q->link)
2161 finishAllThreads ( void )
2164 while (run_queue_hd != END_TSO_QUEUE) {
2165 waitThread ( run_queue_hd, NULL);
2167 while (blocked_queue_hd != END_TSO_QUEUE) {
2168 waitThread ( blocked_queue_hd, NULL);
2170 while (sleeping_queue != END_TSO_QUEUE) {
2171 waitThread ( blocked_queue_hd, NULL);
2174 (blocked_queue_hd != END_TSO_QUEUE ||
2175 run_queue_hd != END_TSO_QUEUE ||
2176 sleeping_queue != END_TSO_QUEUE);
2180 waitThread(StgTSO *tso, /*out*/StgClosure **ret)
2184 m = stgMallocBytes(sizeof(StgMainThread), "waitThread");
2188 #if defined(RTS_SUPPORTS_THREADS)
2189 initCondition(&m->wakeup);
2192 /* see scheduleWaitThread() comment */
2193 ACQUIRE_LOCK(&sched_mutex);
2194 IF_DEBUG(scheduler, sched_belch("== scheduler: waiting for thread (%d)\n", tso->id));
2195 m->link = main_threads;
2197 RELEASE_LOCK(&sched_mutex);
2199 IF_DEBUG(scheduler, sched_belch("== scheduler: waiting for thread (%d)\n", tso->id));
2200 #if defined(THREADED_RTS)
2201 return waitThread_(m, rtsFalse);
2203 return waitThread_(m);
2209 waitThread_(StgMainThread* m
2210 #if defined(THREADED_RTS)
2211 , rtsBool blockWaiting
2215 SchedulerStatus stat;
2217 IF_DEBUG(scheduler, sched_belch("== scheduler: new main thread (%d)\n", m->tso->id));
2219 #if defined(RTS_SUPPORTS_THREADS)
2221 # if defined(THREADED_RTS)
2222 if (!blockWaiting) {
2223 /* In the threaded case, the OS thread that called main()
2224 * gets to enter the RTS directly without going via another
2228 ASSERT(m->stat != NoStatus);
2232 ACQUIRE_LOCK(&sched_mutex);
2234 waitCondition(&m->wakeup, &sched_mutex);
2235 } while (m->stat == NoStatus);
2238 /* GranSim specific init */
2239 CurrentTSO = m->tso; // the TSO to run
2240 procStatus[MainProc] = Busy; // status of main PE
2241 CurrentProc = MainProc; // PE to run it on
2245 RELEASE_LOCK(&sched_mutex);
2247 ASSERT(m->stat != NoStatus);
2252 #if defined(RTS_SUPPORTS_THREADS)
2253 closeCondition(&m->wakeup);
2256 IF_DEBUG(scheduler, fprintf(stderr, "== scheduler: main thread (%d) finished\n",
2260 #if defined(THREADED_RTS)
2263 RELEASE_LOCK(&sched_mutex);
2268 //@node Run queue code, Garbage Collextion Routines, Suspend and Resume, Main scheduling code
2269 //@subsection Run queue code
2273 NB: In GranSim we have many run queues; run_queue_hd is actually a macro
2274 unfolding to run_queue_hds[CurrentProc], thus CurrentProc is an
2275 implicit global variable that has to be correct when calling these
2279 /* Put the new thread on the head of the runnable queue.
2280 * The caller of createThread better push an appropriate closure
2281 * on this thread's stack before the scheduler is invoked.
2283 static /* inline */ void
2284 add_to_run_queue(tso)
2287 ASSERT(tso!=run_queue_hd && tso!=run_queue_tl);
2288 tso->link = run_queue_hd;
2290 if (run_queue_tl == END_TSO_QUEUE) {
2295 /* Put the new thread at the end of the runnable queue. */
2296 static /* inline */ void
2297 push_on_run_queue(tso)
2300 ASSERT(get_itbl((StgClosure *)tso)->type == TSO);
2301 ASSERT(run_queue_hd!=NULL && run_queue_tl!=NULL);
2302 ASSERT(tso!=run_queue_hd && tso!=run_queue_tl);
2303 if (run_queue_hd == END_TSO_QUEUE) {
2306 run_queue_tl->link = tso;
2312 Should be inlined because it's used very often in schedule. The tso
2313 argument is actually only needed in GranSim, where we want to have the
2314 possibility to schedule *any* TSO on the run queue, irrespective of the
2315 actual ordering. Therefore, if tso is not the nil TSO then we traverse
2316 the run queue and dequeue the tso, adjusting the links in the queue.
2318 //@cindex take_off_run_queue
2319 static /* inline */ StgTSO*
2320 take_off_run_queue(StgTSO *tso) {
2324 qetlaHbogh Qu' ngaSbogh ghomDaQ {tso} yIteq!
2326 if tso is specified, unlink that tso from the run_queue (doesn't have
2327 to be at the beginning of the queue); GranSim only
2329 if (tso!=END_TSO_QUEUE) {
2330 /* find tso in queue */
2331 for (t=run_queue_hd, prev=END_TSO_QUEUE;
2332 t!=END_TSO_QUEUE && t!=tso;
2336 /* now actually dequeue the tso */
2337 if (prev!=END_TSO_QUEUE) {
2338 ASSERT(run_queue_hd!=t);
2339 prev->link = t->link;
2341 /* t is at beginning of thread queue */
2342 ASSERT(run_queue_hd==t);
2343 run_queue_hd = t->link;
2345 /* t is at end of thread queue */
2346 if (t->link==END_TSO_QUEUE) {
2347 ASSERT(t==run_queue_tl);
2348 run_queue_tl = prev;
2350 ASSERT(run_queue_tl!=t);
2352 t->link = END_TSO_QUEUE;
2354 /* take tso from the beginning of the queue; std concurrent code */
2356 if (t != END_TSO_QUEUE) {
2357 run_queue_hd = t->link;
2358 t->link = END_TSO_QUEUE;
2359 if (run_queue_hd == END_TSO_QUEUE) {
2360 run_queue_tl = END_TSO_QUEUE;
2369 //@node Garbage Collextion Routines, Blocking Queue Routines, Run queue code, Main scheduling code
2370 //@subsection Garbage Collextion Routines
2372 /* ---------------------------------------------------------------------------
2373 Where are the roots that we know about?
2375 - all the threads on the runnable queue
2376 - all the threads on the blocked queue
2377 - all the threads on the sleeping queue
2378 - all the thread currently executing a _ccall_GC
2379 - all the "main threads"
2381 ------------------------------------------------------------------------ */
2383 /* This has to be protected either by the scheduler monitor, or by the
2384 garbage collection monitor (probably the latter).
2389 GetRoots(evac_fn evac)
2394 for (i=0; i<=RtsFlags.GranFlags.proc; i++) {
2395 if ((run_queue_hds[i] != END_TSO_QUEUE) && ((run_queue_hds[i] != NULL)))
2396 evac((StgClosure **)&run_queue_hds[i]);
2397 if ((run_queue_tls[i] != END_TSO_QUEUE) && ((run_queue_tls[i] != NULL)))
2398 evac((StgClosure **)&run_queue_tls[i]);
2400 if ((blocked_queue_hds[i] != END_TSO_QUEUE) && ((blocked_queue_hds[i] != NULL)))
2401 evac((StgClosure **)&blocked_queue_hds[i]);
2402 if ((blocked_queue_tls[i] != END_TSO_QUEUE) && ((blocked_queue_tls[i] != NULL)))
2403 evac((StgClosure **)&blocked_queue_tls[i]);
2404 if ((ccalling_threadss[i] != END_TSO_QUEUE) && ((ccalling_threadss[i] != NULL)))
2405 evac((StgClosure **)&ccalling_threads[i]);
2412 if (run_queue_hd != END_TSO_QUEUE) {
2413 ASSERT(run_queue_tl != END_TSO_QUEUE);
2414 evac((StgClosure **)&run_queue_hd);
2415 evac((StgClosure **)&run_queue_tl);
2418 if (blocked_queue_hd != END_TSO_QUEUE) {
2419 ASSERT(blocked_queue_tl != END_TSO_QUEUE);
2420 evac((StgClosure **)&blocked_queue_hd);
2421 evac((StgClosure **)&blocked_queue_tl);
2424 if (sleeping_queue != END_TSO_QUEUE) {
2425 evac((StgClosure **)&sleeping_queue);
2429 if (suspended_ccalling_threads != END_TSO_QUEUE) {
2430 evac((StgClosure **)&suspended_ccalling_threads);
2433 #if defined(PAR) || defined(GRAN)
2434 markSparkQueue(evac);
2438 /* -----------------------------------------------------------------------------
2441 This is the interface to the garbage collector from Haskell land.
2442 We provide this so that external C code can allocate and garbage
2443 collect when called from Haskell via _ccall_GC.
2445 It might be useful to provide an interface whereby the programmer
2446 can specify more roots (ToDo).
2448 This needs to be protected by the GC condition variable above. KH.
2449 -------------------------------------------------------------------------- */
2451 void (*extra_roots)(evac_fn);
2456 /* Obligated to hold this lock upon entry */
2457 ACQUIRE_LOCK(&sched_mutex);
2458 GarbageCollect(GetRoots,rtsFalse);
2459 RELEASE_LOCK(&sched_mutex);
2463 performMajorGC(void)
2465 ACQUIRE_LOCK(&sched_mutex);
2466 GarbageCollect(GetRoots,rtsTrue);
2467 RELEASE_LOCK(&sched_mutex);
2471 AllRoots(evac_fn evac)
2473 GetRoots(evac); // the scheduler's roots
2474 extra_roots(evac); // the user's roots
2478 performGCWithRoots(void (*get_roots)(evac_fn))
2480 ACQUIRE_LOCK(&sched_mutex);
2481 extra_roots = get_roots;
2482 GarbageCollect(AllRoots,rtsFalse);
2483 RELEASE_LOCK(&sched_mutex);
2486 /* -----------------------------------------------------------------------------
2489 If the thread has reached its maximum stack size, then raise the
2490 StackOverflow exception in the offending thread. Otherwise
2491 relocate the TSO into a larger chunk of memory and adjust its stack
2493 -------------------------------------------------------------------------- */
2496 threadStackOverflow(StgTSO *tso)
2498 nat new_stack_size, new_tso_size, diff, stack_words;
2502 IF_DEBUG(sanity,checkTSO(tso));
2503 if (tso->stack_size >= tso->max_stack_size) {
2506 belch("@@ threadStackOverflow of TSO %d (%p): stack too large (now %ld; max is %ld",
2507 tso->id, tso, tso->stack_size, tso->max_stack_size);
2508 /* If we're debugging, just print out the top of the stack */
2509 printStackChunk(tso->sp, stg_min(tso->stack+tso->stack_size,
2512 /* Send this thread the StackOverflow exception */
2513 raiseAsync(tso, (StgClosure *)stackOverflow_closure);
2517 /* Try to double the current stack size. If that takes us over the
2518 * maximum stack size for this thread, then use the maximum instead.
2519 * Finally round up so the TSO ends up as a whole number of blocks.
2521 new_stack_size = stg_min(tso->stack_size * 2, tso->max_stack_size);
2522 new_tso_size = (nat)BLOCK_ROUND_UP(new_stack_size * sizeof(W_) +
2523 TSO_STRUCT_SIZE)/sizeof(W_);
2524 new_tso_size = round_to_mblocks(new_tso_size); /* Be MBLOCK-friendly */
2525 new_stack_size = new_tso_size - TSO_STRUCT_SIZEW;
2527 IF_DEBUG(scheduler, fprintf(stderr,"== scheduler: increasing stack size from %d words to %d.\n", tso->stack_size, new_stack_size));
2529 dest = (StgTSO *)allocate(new_tso_size);
2530 TICK_ALLOC_TSO(new_stack_size,0);
2532 /* copy the TSO block and the old stack into the new area */
2533 memcpy(dest,tso,TSO_STRUCT_SIZE);
2534 stack_words = tso->stack + tso->stack_size - tso->sp;
2535 new_sp = (P_)dest + new_tso_size - stack_words;
2536 memcpy(new_sp, tso->sp, stack_words * sizeof(W_));
2538 /* relocate the stack pointers... */
2539 diff = (P_)new_sp - (P_)tso->sp; /* In *words* */
2540 dest->su = (StgUpdateFrame *) ((P_)dest->su + diff);
2542 dest->stack_size = new_stack_size;
2544 /* and relocate the update frame list */
2545 relocate_stack(dest, diff);
2547 /* Mark the old TSO as relocated. We have to check for relocated
2548 * TSOs in the garbage collector and any primops that deal with TSOs.
2550 * It's important to set the sp and su values to just beyond the end
2551 * of the stack, so we don't attempt to scavenge any part of the
2554 tso->what_next = ThreadRelocated;
2556 tso->sp = (P_)&(tso->stack[tso->stack_size]);
2557 tso->su = (StgUpdateFrame *)tso->sp;
2558 tso->why_blocked = NotBlocked;
2559 dest->mut_link = NULL;
2561 IF_PAR_DEBUG(verbose,
2562 belch("@@ threadStackOverflow of TSO %d (now at %p): stack size increased to %ld",
2563 tso->id, tso, tso->stack_size);
2564 /* If we're debugging, just print out the top of the stack */
2565 printStackChunk(tso->sp, stg_min(tso->stack+tso->stack_size,
2568 IF_DEBUG(sanity,checkTSO(tso));
2570 IF_DEBUG(scheduler,printTSO(dest));
2576 //@node Blocking Queue Routines, Exception Handling Routines, Garbage Collextion Routines, Main scheduling code
2577 //@subsection Blocking Queue Routines
2579 /* ---------------------------------------------------------------------------
2580 Wake up a queue that was blocked on some resource.
2581 ------------------------------------------------------------------------ */
2585 unblockCount ( StgBlockingQueueElement *bqe, StgClosure *node )
2590 unblockCount ( StgBlockingQueueElement *bqe, StgClosure *node )
2592 /* write RESUME events to log file and
2593 update blocked and fetch time (depending on type of the orig closure) */
2594 if (RtsFlags.ParFlags.ParStats.Full) {
2595 DumpRawGranEvent(CURRENT_PROC, CURRENT_PROC,
2596 GR_RESUMEQ, ((StgTSO *)bqe), ((StgTSO *)bqe)->block_info.closure,
2597 0, 0 /* spark_queue_len(ADVISORY_POOL) */);
2598 if (EMPTY_RUN_QUEUE())
2599 emitSchedule = rtsTrue;
2601 switch (get_itbl(node)->type) {
2603 ((StgTSO *)bqe)->par.fetchtime += CURRENT_TIME-((StgTSO *)bqe)->par.blockedat;
2608 ((StgTSO *)bqe)->par.blocktime += CURRENT_TIME-((StgTSO *)bqe)->par.blockedat;
2615 barf("{unblockOneLocked}Daq Qagh: unexpected closure in blocking queue");
2622 static StgBlockingQueueElement *
2623 unblockOneLocked(StgBlockingQueueElement *bqe, StgClosure *node)
2626 PEs node_loc, tso_loc;
2628 node_loc = where_is(node); // should be lifted out of loop
2629 tso = (StgTSO *)bqe; // wastes an assignment to get the type right
2630 tso_loc = where_is((StgClosure *)tso);
2631 if (IS_LOCAL_TO(PROCS(node),tso_loc)) { // TSO is local
2632 /* !fake_fetch => TSO is on CurrentProc is same as IS_LOCAL_TO */
2633 ASSERT(CurrentProc!=node_loc || tso_loc==CurrentProc);
2634 CurrentTime[CurrentProc] += RtsFlags.GranFlags.Costs.lunblocktime;
2635 // insertThread(tso, node_loc);
2636 new_event(tso_loc, tso_loc, CurrentTime[CurrentProc],
2638 tso, node, (rtsSpark*)NULL);
2639 tso->link = END_TSO_QUEUE; // overwrite link just to be sure
2642 } else { // TSO is remote (actually should be FMBQ)
2643 CurrentTime[CurrentProc] += RtsFlags.GranFlags.Costs.mpacktime +
2644 RtsFlags.GranFlags.Costs.gunblocktime +
2645 RtsFlags.GranFlags.Costs.latency;
2646 new_event(tso_loc, CurrentProc, CurrentTime[CurrentProc],
2648 tso, node, (rtsSpark*)NULL);
2649 tso->link = END_TSO_QUEUE; // overwrite link just to be sure
2652 /* the thread-queue-overhead is accounted for in either Resume or UnblockThread */
2654 fprintf(stderr," %s TSO %d (%p) [PE %d] (block_info.closure=%p) (next=%p) ,",
2655 (node_loc==tso_loc ? "Local" : "Global"),
2656 tso->id, tso, CurrentProc, tso->block_info.closure, tso->link));
2657 tso->block_info.closure = NULL;
2658 IF_DEBUG(scheduler,belch("-- Waking up thread %ld (%p)",
2662 static StgBlockingQueueElement *
2663 unblockOneLocked(StgBlockingQueueElement *bqe, StgClosure *node)
2665 StgBlockingQueueElement *next;
2667 switch (get_itbl(bqe)->type) {
2669 ASSERT(((StgTSO *)bqe)->why_blocked != NotBlocked);
2670 /* if it's a TSO just push it onto the run_queue */
2672 // ((StgTSO *)bqe)->link = END_TSO_QUEUE; // debugging?
2673 PUSH_ON_RUN_QUEUE((StgTSO *)bqe);
2675 unblockCount(bqe, node);
2676 /* reset blocking status after dumping event */
2677 ((StgTSO *)bqe)->why_blocked = NotBlocked;
2681 /* if it's a BLOCKED_FETCH put it on the PendingFetches list */
2683 bqe->link = (StgBlockingQueueElement *)PendingFetches;
2684 PendingFetches = (StgBlockedFetch *)bqe;
2688 /* can ignore this case in a non-debugging setup;
2689 see comments on RBHSave closures above */
2691 /* check that the closure is an RBHSave closure */
2692 ASSERT(get_itbl((StgClosure *)bqe) == &stg_RBH_Save_0_info ||
2693 get_itbl((StgClosure *)bqe) == &stg_RBH_Save_1_info ||
2694 get_itbl((StgClosure *)bqe) == &stg_RBH_Save_2_info);
2698 barf("{unblockOneLocked}Daq Qagh: Unexpected IP (%#lx; %s) in blocking queue at %#lx\n",
2699 get_itbl((StgClosure *)bqe), info_type((StgClosure *)bqe),
2703 IF_PAR_DEBUG(bq, fprintf(stderr, ", %p (%s)", bqe, info_type((StgClosure*)bqe)));
2707 #else /* !GRAN && !PAR */
2709 unblockOneLocked(StgTSO *tso)
2713 ASSERT(get_itbl(tso)->type == TSO);
2714 ASSERT(tso->why_blocked != NotBlocked);
2715 tso->why_blocked = NotBlocked;
2717 PUSH_ON_RUN_QUEUE(tso);
2719 IF_DEBUG(scheduler,sched_belch("waking up thread %ld", tso->id));
2724 #if defined(GRAN) || defined(PAR)
2725 inline StgBlockingQueueElement *
2726 unblockOne(StgBlockingQueueElement *bqe, StgClosure *node)
2728 ACQUIRE_LOCK(&sched_mutex);
2729 bqe = unblockOneLocked(bqe, node);
2730 RELEASE_LOCK(&sched_mutex);
2735 unblockOne(StgTSO *tso)
2737 ACQUIRE_LOCK(&sched_mutex);
2738 tso = unblockOneLocked(tso);
2739 RELEASE_LOCK(&sched_mutex);
2746 awakenBlockedQueue(StgBlockingQueueElement *q, StgClosure *node)
2748 StgBlockingQueueElement *bqe;
2753 belch("##-_ AwBQ for node %p on PE %d @ %ld by TSO %d (%p): ", \
2754 node, CurrentProc, CurrentTime[CurrentProc],
2755 CurrentTSO->id, CurrentTSO));
2757 node_loc = where_is(node);
2759 ASSERT(q == END_BQ_QUEUE ||
2760 get_itbl(q)->type == TSO || // q is either a TSO or an RBHSave
2761 get_itbl(q)->type == CONSTR); // closure (type constructor)
2762 ASSERT(is_unique(node));
2764 /* FAKE FETCH: magically copy the node to the tso's proc;
2765 no Fetch necessary because in reality the node should not have been
2766 moved to the other PE in the first place
2768 if (CurrentProc!=node_loc) {
2770 belch("## node %p is on PE %d but CurrentProc is %d (TSO %d); assuming fake fetch and adjusting bitmask (old: %#x)",
2771 node, node_loc, CurrentProc, CurrentTSO->id,
2772 // CurrentTSO, where_is(CurrentTSO),
2773 node->header.gran.procs));
2774 node->header.gran.procs = (node->header.gran.procs) | PE_NUMBER(CurrentProc);
2776 belch("## new bitmask of node %p is %#x",
2777 node, node->header.gran.procs));
2778 if (RtsFlags.GranFlags.GranSimStats.Global) {
2779 globalGranStats.tot_fake_fetches++;
2784 // ToDo: check: ASSERT(CurrentProc==node_loc);
2785 while (get_itbl(bqe)->type==TSO) { // q != END_TSO_QUEUE) {
2788 bqe points to the current element in the queue
2789 next points to the next element in the queue
2791 //tso = (StgTSO *)bqe; // wastes an assignment to get the type right
2792 //tso_loc = where_is(tso);
2794 bqe = unblockOneLocked(bqe, node);
2797 /* if this is the BQ of an RBH, we have to put back the info ripped out of
2798 the closure to make room for the anchor of the BQ */
2799 if (bqe!=END_BQ_QUEUE) {
2800 ASSERT(get_itbl(node)->type == RBH && get_itbl(bqe)->type == CONSTR);
2802 ASSERT((info_ptr==&RBH_Save_0_info) ||
2803 (info_ptr==&RBH_Save_1_info) ||
2804 (info_ptr==&RBH_Save_2_info));
2806 /* cf. convertToRBH in RBH.c for writing the RBHSave closure */
2807 ((StgRBH *)node)->blocking_queue = (StgBlockingQueueElement *)((StgRBHSave *)bqe)->payload[0];
2808 ((StgRBH *)node)->mut_link = (StgMutClosure *)((StgRBHSave *)bqe)->payload[1];
2811 belch("## Filled in RBH_Save for %p (%s) at end of AwBQ",
2812 node, info_type(node)));
2815 /* statistics gathering */
2816 if (RtsFlags.GranFlags.GranSimStats.Global) {
2817 // globalGranStats.tot_bq_processing_time += bq_processing_time;
2818 globalGranStats.tot_bq_len += len; // total length of all bqs awakened
2819 // globalGranStats.tot_bq_len_local += len_local; // same for local TSOs only
2820 globalGranStats.tot_awbq++; // total no. of bqs awakened
2823 fprintf(stderr,"## BQ Stats of %p: [%d entries] %s\n",
2824 node, len, (bqe!=END_BQ_QUEUE) ? "RBH" : ""));
2828 awakenBlockedQueue(StgBlockingQueueElement *q, StgClosure *node)
2830 StgBlockingQueueElement *bqe;
2832 ACQUIRE_LOCK(&sched_mutex);
2834 IF_PAR_DEBUG(verbose,
2835 belch("##-_ AwBQ for node %p on [%x]: ",
2839 if(get_itbl(q)->type == CONSTR || q==END_BQ_QUEUE) {
2840 IF_PAR_DEBUG(verbose, belch("## ... nothing to unblock so lets just return. RFP (BUG?)"));
2845 ASSERT(q == END_BQ_QUEUE ||
2846 get_itbl(q)->type == TSO ||
2847 get_itbl(q)->type == BLOCKED_FETCH ||
2848 get_itbl(q)->type == CONSTR);
2851 while (get_itbl(bqe)->type==TSO ||
2852 get_itbl(bqe)->type==BLOCKED_FETCH) {
2853 bqe = unblockOneLocked(bqe, node);
2855 RELEASE_LOCK(&sched_mutex);
2858 #else /* !GRAN && !PAR */
2860 awakenBlockedQueue(StgTSO *tso)
2862 ACQUIRE_LOCK(&sched_mutex);
2863 while (tso != END_TSO_QUEUE) {
2864 tso = unblockOneLocked(tso);
2866 RELEASE_LOCK(&sched_mutex);
2870 //@node Exception Handling Routines, Debugging Routines, Blocking Queue Routines, Main scheduling code
2871 //@subsection Exception Handling Routines
2873 /* ---------------------------------------------------------------------------
2875 - usually called inside a signal handler so it mustn't do anything fancy.
2876 ------------------------------------------------------------------------ */
2879 interruptStgRts(void)
2885 /* -----------------------------------------------------------------------------
2888 This is for use when we raise an exception in another thread, which
2890 This has nothing to do with the UnblockThread event in GranSim. -- HWL
2891 -------------------------------------------------------------------------- */
2893 #if defined(GRAN) || defined(PAR)
2895 NB: only the type of the blocking queue is different in GranSim and GUM
2896 the operations on the queue-elements are the same
2897 long live polymorphism!
2899 Locks: sched_mutex is held upon entry and exit.
2903 unblockThread(StgTSO *tso)
2905 StgBlockingQueueElement *t, **last;
2907 switch (tso->why_blocked) {
2910 return; /* not blocked */
2913 ASSERT(get_itbl(tso->block_info.closure)->type == MVAR);
2915 StgBlockingQueueElement *last_tso = END_BQ_QUEUE;
2916 StgMVar *mvar = (StgMVar *)(tso->block_info.closure);
2918 last = (StgBlockingQueueElement **)&mvar->head;
2919 for (t = (StgBlockingQueueElement *)mvar->head;
2921 last = &t->link, last_tso = t, t = t->link) {
2922 if (t == (StgBlockingQueueElement *)tso) {
2923 *last = (StgBlockingQueueElement *)tso->link;
2924 if (mvar->tail == tso) {
2925 mvar->tail = (StgTSO *)last_tso;
2930 barf("unblockThread (MVAR): TSO not found");
2933 case BlockedOnBlackHole:
2934 ASSERT(get_itbl(tso->block_info.closure)->type == BLACKHOLE_BQ);
2936 StgBlockingQueue *bq = (StgBlockingQueue *)(tso->block_info.closure);
2938 last = &bq->blocking_queue;
2939 for (t = bq->blocking_queue;
2941 last = &t->link, t = t->link) {
2942 if (t == (StgBlockingQueueElement *)tso) {
2943 *last = (StgBlockingQueueElement *)tso->link;
2947 barf("unblockThread (BLACKHOLE): TSO not found");
2950 case BlockedOnException:
2952 StgTSO *target = tso->block_info.tso;
2954 ASSERT(get_itbl(target)->type == TSO);
2956 if (target->what_next == ThreadRelocated) {
2957 target = target->link;
2958 ASSERT(get_itbl(target)->type == TSO);
2961 ASSERT(target->blocked_exceptions != NULL);
2963 last = (StgBlockingQueueElement **)&target->blocked_exceptions;
2964 for (t = (StgBlockingQueueElement *)target->blocked_exceptions;
2966 last = &t->link, t = t->link) {
2967 ASSERT(get_itbl(t)->type == TSO);
2968 if (t == (StgBlockingQueueElement *)tso) {
2969 *last = (StgBlockingQueueElement *)tso->link;
2973 barf("unblockThread (Exception): TSO not found");
2977 case BlockedOnWrite:
2979 /* take TSO off blocked_queue */
2980 StgBlockingQueueElement *prev = NULL;
2981 for (t = (StgBlockingQueueElement *)blocked_queue_hd; t != END_BQ_QUEUE;
2982 prev = t, t = t->link) {
2983 if (t == (StgBlockingQueueElement *)tso) {
2985 blocked_queue_hd = (StgTSO *)t->link;
2986 if ((StgBlockingQueueElement *)blocked_queue_tl == t) {
2987 blocked_queue_tl = END_TSO_QUEUE;
2990 prev->link = t->link;
2991 if ((StgBlockingQueueElement *)blocked_queue_tl == t) {
2992 blocked_queue_tl = (StgTSO *)prev;
2998 barf("unblockThread (I/O): TSO not found");
3001 case BlockedOnDelay:
3003 /* take TSO off sleeping_queue */
3004 StgBlockingQueueElement *prev = NULL;
3005 for (t = (StgBlockingQueueElement *)sleeping_queue; t != END_BQ_QUEUE;
3006 prev = t, t = t->link) {
3007 if (t == (StgBlockingQueueElement *)tso) {
3009 sleeping_queue = (StgTSO *)t->link;
3011 prev->link = t->link;
3016 barf("unblockThread (I/O): TSO not found");
3020 barf("unblockThread");
3024 tso->link = END_TSO_QUEUE;
3025 tso->why_blocked = NotBlocked;
3026 tso->block_info.closure = NULL;
3027 PUSH_ON_RUN_QUEUE(tso);
3031 unblockThread(StgTSO *tso)
3035 /* To avoid locking unnecessarily. */
3036 if (tso->why_blocked == NotBlocked) {
3040 switch (tso->why_blocked) {
3043 ASSERT(get_itbl(tso->block_info.closure)->type == MVAR);
3045 StgTSO *last_tso = END_TSO_QUEUE;
3046 StgMVar *mvar = (StgMVar *)(tso->block_info.closure);
3049 for (t = mvar->head; t != END_TSO_QUEUE;
3050 last = &t->link, last_tso = t, t = t->link) {
3053 if (mvar->tail == tso) {
3054 mvar->tail = last_tso;
3059 barf("unblockThread (MVAR): TSO not found");
3062 case BlockedOnBlackHole:
3063 ASSERT(get_itbl(tso->block_info.closure)->type == BLACKHOLE_BQ);
3065 StgBlockingQueue *bq = (StgBlockingQueue *)(tso->block_info.closure);
3067 last = &bq->blocking_queue;
3068 for (t = bq->blocking_queue; t != END_TSO_QUEUE;
3069 last = &t->link, t = t->link) {
3075 barf("unblockThread (BLACKHOLE): TSO not found");
3078 case BlockedOnException:
3080 StgTSO *target = tso->block_info.tso;
3082 ASSERT(get_itbl(target)->type == TSO);
3084 while (target->what_next == ThreadRelocated) {
3085 target = target->link;
3086 ASSERT(get_itbl(target)->type == TSO);
3089 ASSERT(target->blocked_exceptions != NULL);
3091 last = &target->blocked_exceptions;
3092 for (t = target->blocked_exceptions; t != END_TSO_QUEUE;
3093 last = &t->link, t = t->link) {
3094 ASSERT(get_itbl(t)->type == TSO);
3100 barf("unblockThread (Exception): TSO not found");
3104 case BlockedOnWrite:
3106 StgTSO *prev = NULL;
3107 for (t = blocked_queue_hd; t != END_TSO_QUEUE;
3108 prev = t, t = t->link) {
3111 blocked_queue_hd = t->link;
3112 if (blocked_queue_tl == t) {
3113 blocked_queue_tl = END_TSO_QUEUE;
3116 prev->link = t->link;
3117 if (blocked_queue_tl == t) {
3118 blocked_queue_tl = prev;
3124 barf("unblockThread (I/O): TSO not found");
3127 case BlockedOnDelay:
3129 StgTSO *prev = NULL;
3130 for (t = sleeping_queue; t != END_TSO_QUEUE;
3131 prev = t, t = t->link) {
3134 sleeping_queue = t->link;
3136 prev->link = t->link;
3141 barf("unblockThread (I/O): TSO not found");
3145 barf("unblockThread");
3149 tso->link = END_TSO_QUEUE;
3150 tso->why_blocked = NotBlocked;
3151 tso->block_info.closure = NULL;
3152 PUSH_ON_RUN_QUEUE(tso);
3156 /* -----------------------------------------------------------------------------
3159 * The following function implements the magic for raising an
3160 * asynchronous exception in an existing thread.
3162 * We first remove the thread from any queue on which it might be
3163 * blocked. The possible blockages are MVARs and BLACKHOLE_BQs.
3165 * We strip the stack down to the innermost CATCH_FRAME, building
3166 * thunks in the heap for all the active computations, so they can
3167 * be restarted if necessary. When we reach a CATCH_FRAME, we build
3168 * an application of the handler to the exception, and push it on
3169 * the top of the stack.
3171 * How exactly do we save all the active computations? We create an
3172 * AP_UPD for every UpdateFrame on the stack. Entering one of these
3173 * AP_UPDs pushes everything from the corresponding update frame
3174 * upwards onto the stack. (Actually, it pushes everything up to the
3175 * next update frame plus a pointer to the next AP_UPD object.
3176 * Entering the next AP_UPD object pushes more onto the stack until we
3177 * reach the last AP_UPD object - at which point the stack should look
3178 * exactly as it did when we killed the TSO and we can continue
3179 * execution by entering the closure on top of the stack.
3181 * We can also kill a thread entirely - this happens if either (a) the
3182 * exception passed to raiseAsync is NULL, or (b) there's no
3183 * CATCH_FRAME on the stack. In either case, we strip the entire
3184 * stack and replace the thread with a zombie.
3186 * Locks: sched_mutex held upon entry nor exit.
3188 * -------------------------------------------------------------------------- */
3191 deleteThread(StgTSO *tso)
3193 raiseAsync(tso,NULL);
3197 raiseAsyncWithLock(StgTSO *tso, StgClosure *exception)
3199 /* When raising async exs from contexts where sched_mutex isn't held;
3200 use raiseAsyncWithLock(). */
3201 ACQUIRE_LOCK(&sched_mutex);
3202 raiseAsync(tso,exception);
3203 RELEASE_LOCK(&sched_mutex);
3207 raiseAsync(StgTSO *tso, StgClosure *exception)
3209 StgUpdateFrame* su = tso->su;
3210 StgPtr sp = tso->sp;
3212 /* Thread already dead? */
3213 if (tso->what_next == ThreadComplete || tso->what_next == ThreadKilled) {
3217 IF_DEBUG(scheduler, sched_belch("raising exception in thread %ld.", tso->id));
3219 /* Remove it from any blocking queues */
3222 IF_DEBUG(scheduler, sched_belch("raising exception in thread %ld.", tso->id));
3223 /* The stack freezing code assumes there's a closure pointer on
3224 * the top of the stack. This isn't always the case with compiled
3225 * code, so we have to push a dummy closure on the top which just
3226 * returns to the next return address on the stack.
3228 if ( LOOKS_LIKE_GHC_INFO((void*)*sp) ) {
3229 *(--sp) = (W_)&stg_dummy_ret_closure;
3233 nat words = ((P_)su - (P_)sp) - 1;
3237 /* If we find a CATCH_FRAME, and we've got an exception to raise,
3238 * then build the THUNK raise(exception), and leave it on
3239 * top of the CATCH_FRAME ready to enter.
3241 if (get_itbl(su)->type == CATCH_FRAME && exception != NULL) {
3243 StgCatchFrame *cf = (StgCatchFrame *)su;
3247 /* we've got an exception to raise, so let's pass it to the
3248 * handler in this frame.
3250 raise = (StgClosure *)allocate(sizeofW(StgClosure)+1);
3251 TICK_ALLOC_SE_THK(1,0);
3252 SET_HDR(raise,&stg_raise_info,cf->header.prof.ccs);
3253 raise->payload[0] = exception;
3255 /* throw away the stack from Sp up to the CATCH_FRAME.
3259 /* Ensure that async excpetions are blocked now, so we don't get
3260 * a surprise exception before we get around to executing the
3263 if (tso->blocked_exceptions == NULL) {
3264 tso->blocked_exceptions = END_TSO_QUEUE;
3267 /* Put the newly-built THUNK on top of the stack, ready to execute
3268 * when the thread restarts.
3273 tso->what_next = ThreadEnterGHC;
3274 IF_DEBUG(sanity, checkTSO(tso));
3278 /* First build an AP_UPD consisting of the stack chunk above the
3279 * current update frame, with the top word on the stack as the
3282 ap = (StgAP_UPD *)allocate(AP_sizeW(words));
3287 ap->fun = (StgClosure *)sp[0];
3289 for(i=0; i < (nat)words; ++i) {
3290 ap->payload[i] = (StgClosure *)*sp++;
3293 switch (get_itbl(su)->type) {
3297 SET_HDR(ap,&stg_AP_UPD_info,su->header.prof.ccs /* ToDo */);
3298 TICK_ALLOC_UP_THK(words+1,0);
3301 fprintf(stderr, "scheduler: Updating ");
3302 printPtr((P_)su->updatee);
3303 fprintf(stderr, " with ");
3304 printObj((StgClosure *)ap);
3307 /* Replace the updatee with an indirection - happily
3308 * this will also wake up any threads currently
3309 * waiting on the result.
3311 * Warning: if we're in a loop, more than one update frame on
3312 * the stack may point to the same object. Be careful not to
3313 * overwrite an IND_OLDGEN in this case, because we'll screw
3314 * up the mutable lists. To be on the safe side, don't
3315 * overwrite any kind of indirection at all. See also
3316 * threadSqueezeStack in GC.c, where we have to make a similar
3319 if (!closure_IND(su->updatee)) {
3320 UPD_IND_NOLOCK(su->updatee,ap); /* revert the black hole */
3323 sp += sizeofW(StgUpdateFrame) -1;
3324 sp[0] = (W_)ap; /* push onto stack */
3330 StgCatchFrame *cf = (StgCatchFrame *)su;
3333 /* We want a PAP, not an AP_UPD. Fortunately, the
3334 * layout's the same.
3336 SET_HDR(ap,&stg_PAP_info,su->header.prof.ccs /* ToDo */);
3337 TICK_ALLOC_UPD_PAP(words+1,0);
3339 /* now build o = FUN(catch,ap,handler) */
3340 o = (StgClosure *)allocate(sizeofW(StgClosure)+2);
3341 TICK_ALLOC_FUN(2,0);
3342 SET_HDR(o,&stg_catch_info,su->header.prof.ccs /* ToDo */);
3343 o->payload[0] = (StgClosure *)ap;
3344 o->payload[1] = cf->handler;
3347 fprintf(stderr, "scheduler: Built ");
3348 printObj((StgClosure *)o);
3351 /* pop the old handler and put o on the stack */
3353 sp += sizeofW(StgCatchFrame) - 1;
3360 StgSeqFrame *sf = (StgSeqFrame *)su;
3363 SET_HDR(ap,&stg_PAP_info,su->header.prof.ccs /* ToDo */);
3364 TICK_ALLOC_UPD_PAP(words+1,0);
3366 /* now build o = FUN(seq,ap) */
3367 o = (StgClosure *)allocate(sizeofW(StgClosure)+1);
3368 TICK_ALLOC_SE_THK(1,0);
3369 SET_HDR(o,&stg_seq_info,su->header.prof.ccs /* ToDo */);
3370 o->payload[0] = (StgClosure *)ap;
3373 fprintf(stderr, "scheduler: Built ");
3374 printObj((StgClosure *)o);
3377 /* pop the old handler and put o on the stack */
3379 sp += sizeofW(StgSeqFrame) - 1;
3385 /* We've stripped the entire stack, the thread is now dead. */
3386 sp += sizeofW(StgStopFrame) - 1;
3387 sp[0] = (W_)exception; /* save the exception */
3388 tso->what_next = ThreadKilled;
3389 tso->su = (StgUpdateFrame *)(sp+1);
3400 /* -----------------------------------------------------------------------------
3401 resurrectThreads is called after garbage collection on the list of
3402 threads found to be garbage. Each of these threads will be woken
3403 up and sent a signal: BlockedOnDeadMVar if the thread was blocked
3404 on an MVar, or NonTermination if the thread was blocked on a Black
3407 Locks: sched_mutex isn't held upon entry nor exit.
3408 -------------------------------------------------------------------------- */
3411 resurrectThreads( StgTSO *threads )
3415 for (tso = threads; tso != END_TSO_QUEUE; tso = next) {
3416 next = tso->global_link;
3417 tso->global_link = all_threads;
3419 IF_DEBUG(scheduler, sched_belch("resurrecting thread %d", tso->id));
3421 switch (tso->why_blocked) {
3423 case BlockedOnException:
3424 /* Called by GC - sched_mutex lock is currently held. */
3425 raiseAsync(tso,(StgClosure *)BlockedOnDeadMVar_closure);
3427 case BlockedOnBlackHole:
3428 raiseAsync(tso,(StgClosure *)NonTermination_closure);
3431 /* This might happen if the thread was blocked on a black hole
3432 * belonging to a thread that we've just woken up (raiseAsync
3433 * can wake up threads, remember...).
3437 barf("resurrectThreads: thread blocked in a strange way");
3442 /* -----------------------------------------------------------------------------
3443 * Blackhole detection: if we reach a deadlock, test whether any
3444 * threads are blocked on themselves. Any threads which are found to
3445 * be self-blocked get sent a NonTermination exception.
3447 * This is only done in a deadlock situation in order to avoid
3448 * performance overhead in the normal case.
3450 * Locks: sched_mutex is held upon entry and exit.
3451 * -------------------------------------------------------------------------- */
3454 detectBlackHoles( void )
3456 StgTSO *t = all_threads;
3457 StgUpdateFrame *frame;
3458 StgClosure *blocked_on;
3460 for (t = all_threads; t != END_TSO_QUEUE; t = t->global_link) {
3462 while (t->what_next == ThreadRelocated) {
3464 ASSERT(get_itbl(t)->type == TSO);
3467 if (t->why_blocked != BlockedOnBlackHole) {
3471 blocked_on = t->block_info.closure;
3473 for (frame = t->su; ; frame = frame->link) {
3474 switch (get_itbl(frame)->type) {
3477 if (frame->updatee == blocked_on) {
3478 /* We are blocking on one of our own computations, so
3479 * send this thread the NonTermination exception.
3482 sched_belch("thread %d is blocked on itself", t->id));
3483 raiseAsync(t, (StgClosure *)NonTermination_closure);
3504 //@node Debugging Routines, Index, Exception Handling Routines, Main scheduling code
3505 //@subsection Debugging Routines
3507 /* -----------------------------------------------------------------------------
3508 Debugging: why is a thread blocked
3509 -------------------------------------------------------------------------- */
3514 printThreadBlockage(StgTSO *tso)
3516 switch (tso->why_blocked) {
3518 fprintf(stderr,"is blocked on read from fd %d", tso->block_info.fd);
3520 case BlockedOnWrite:
3521 fprintf(stderr,"is blocked on write to fd %d", tso->block_info.fd);
3523 case BlockedOnDelay:
3524 fprintf(stderr,"is blocked until %d", tso->block_info.target);
3527 fprintf(stderr,"is blocked on an MVar");
3529 case BlockedOnException:
3530 fprintf(stderr,"is blocked on delivering an exception to thread %d",
3531 tso->block_info.tso->id);
3533 case BlockedOnBlackHole:
3534 fprintf(stderr,"is blocked on a black hole");
3537 fprintf(stderr,"is not blocked");
3541 fprintf(stderr,"is blocked on global address; local FM_BQ is %p (%s)",
3542 tso->block_info.closure, info_type(tso->block_info.closure));
3544 case BlockedOnGA_NoSend:
3545 fprintf(stderr,"is blocked on global address (no send); local FM_BQ is %p (%s)",
3546 tso->block_info.closure, info_type(tso->block_info.closure));
3549 #if defined(RTS_SUPPORTS_THREADS)
3550 case BlockedOnCCall:
3551 fprintf(stderr,"is blocked on an external call");
3555 barf("printThreadBlockage: strange tso->why_blocked: %d for TSO %d (%d)",
3556 tso->why_blocked, tso->id, tso);
3561 printThreadStatus(StgTSO *tso)
3563 switch (tso->what_next) {
3565 fprintf(stderr,"has been killed");
3567 case ThreadComplete:
3568 fprintf(stderr,"has completed");
3571 printThreadBlockage(tso);
3576 printAllThreads(void)
3582 char time_string[TIME_STR_LEN], node_str[NODE_STR_LEN];
3583 ullong_format_string(TIME_ON_PROC(CurrentProc),
3584 time_string, rtsFalse/*no commas!*/);
3586 sched_belch("all threads at [%s]:", time_string);
3588 char time_string[TIME_STR_LEN], node_str[NODE_STR_LEN];
3589 ullong_format_string(CURRENT_TIME,
3590 time_string, rtsFalse/*no commas!*/);
3592 sched_belch("all threads at [%s]:", time_string);
3594 sched_belch("all threads:");
3597 for (t = all_threads; t != END_TSO_QUEUE; t = t->global_link) {
3598 fprintf(stderr, "\tthread %d @ %p ", t->id, (void *)t);
3599 label = lookupThreadLabel((StgWord)t);
3600 if (label) fprintf(stderr,"[\"%s\"] ",(char *)label);
3601 printThreadStatus(t);
3602 fprintf(stderr,"\n");
3607 Print a whole blocking queue attached to node (debugging only).
3612 print_bq (StgClosure *node)
3614 StgBlockingQueueElement *bqe;
3618 fprintf(stderr,"## BQ of closure %p (%s): ",
3619 node, info_type(node));
3621 /* should cover all closures that may have a blocking queue */
3622 ASSERT(get_itbl(node)->type == BLACKHOLE_BQ ||
3623 get_itbl(node)->type == FETCH_ME_BQ ||
3624 get_itbl(node)->type == RBH ||
3625 get_itbl(node)->type == MVAR);
3627 ASSERT(node!=(StgClosure*)NULL); // sanity check
3629 print_bqe(((StgBlockingQueue*)node)->blocking_queue);
3633 Print a whole blocking queue starting with the element bqe.
3636 print_bqe (StgBlockingQueueElement *bqe)
3641 NB: In a parallel setup a BQ of an RBH must end with an RBH_Save closure;
3643 for (end = (bqe==END_BQ_QUEUE);
3644 !end; // iterate until bqe points to a CONSTR
3645 end = (get_itbl(bqe)->type == CONSTR) || (bqe->link==END_BQ_QUEUE),
3646 bqe = end ? END_BQ_QUEUE : bqe->link) {
3647 ASSERT(bqe != END_BQ_QUEUE); // sanity check
3648 ASSERT(bqe != (StgBlockingQueueElement *)NULL); // sanity check
3649 /* types of closures that may appear in a blocking queue */
3650 ASSERT(get_itbl(bqe)->type == TSO ||
3651 get_itbl(bqe)->type == BLOCKED_FETCH ||
3652 get_itbl(bqe)->type == CONSTR);
3653 /* only BQs of an RBH end with an RBH_Save closure */
3654 //ASSERT(get_itbl(bqe)->type != CONSTR || get_itbl(node)->type == RBH);
3656 switch (get_itbl(bqe)->type) {
3658 fprintf(stderr," TSO %u (%x),",
3659 ((StgTSO *)bqe)->id, ((StgTSO *)bqe));
3662 fprintf(stderr," BF (node=%p, ga=((%x, %d, %x)),",
3663 ((StgBlockedFetch *)bqe)->node,
3664 ((StgBlockedFetch *)bqe)->ga.payload.gc.gtid,
3665 ((StgBlockedFetch *)bqe)->ga.payload.gc.slot,
3666 ((StgBlockedFetch *)bqe)->ga.weight);
3669 fprintf(stderr," %s (IP %p),",
3670 (get_itbl(bqe) == &stg_RBH_Save_0_info ? "RBH_Save_0" :
3671 get_itbl(bqe) == &stg_RBH_Save_1_info ? "RBH_Save_1" :
3672 get_itbl(bqe) == &stg_RBH_Save_2_info ? "RBH_Save_2" :
3673 "RBH_Save_?"), get_itbl(bqe));
3676 barf("Unexpected closure type %s in blocking queue", // of %p (%s)",
3677 info_type((StgClosure *)bqe)); // , node, info_type(node));
3681 fputc('\n', stderr);
3683 # elif defined(GRAN)
3685 print_bq (StgClosure *node)
3687 StgBlockingQueueElement *bqe;
3688 PEs node_loc, tso_loc;
3691 /* should cover all closures that may have a blocking queue */
3692 ASSERT(get_itbl(node)->type == BLACKHOLE_BQ ||
3693 get_itbl(node)->type == FETCH_ME_BQ ||
3694 get_itbl(node)->type == RBH);
3696 ASSERT(node!=(StgClosure*)NULL); // sanity check
3697 node_loc = where_is(node);
3699 fprintf(stderr,"## BQ of closure %p (%s) on [PE %d]: ",
3700 node, info_type(node), node_loc);
3703 NB: In a parallel setup a BQ of an RBH must end with an RBH_Save closure;
3705 for (bqe = ((StgBlockingQueue*)node)->blocking_queue, end = (bqe==END_BQ_QUEUE);
3706 !end; // iterate until bqe points to a CONSTR
3707 end = (get_itbl(bqe)->type == CONSTR) || (bqe->link==END_BQ_QUEUE), bqe = end ? END_BQ_QUEUE : bqe->link) {
3708 ASSERT(bqe != END_BQ_QUEUE); // sanity check
3709 ASSERT(bqe != (StgBlockingQueueElement *)NULL); // sanity check
3710 /* types of closures that may appear in a blocking queue */
3711 ASSERT(get_itbl(bqe)->type == TSO ||
3712 get_itbl(bqe)->type == CONSTR);
3713 /* only BQs of an RBH end with an RBH_Save closure */
3714 ASSERT(get_itbl(bqe)->type != CONSTR || get_itbl(node)->type == RBH);
3716 tso_loc = where_is((StgClosure *)bqe);
3717 switch (get_itbl(bqe)->type) {
3719 fprintf(stderr," TSO %d (%p) on [PE %d],",
3720 ((StgTSO *)bqe)->id, (StgTSO *)bqe, tso_loc);
3723 fprintf(stderr," %s (IP %p),",
3724 (get_itbl(bqe) == &stg_RBH_Save_0_info ? "RBH_Save_0" :
3725 get_itbl(bqe) == &stg_RBH_Save_1_info ? "RBH_Save_1" :
3726 get_itbl(bqe) == &stg_RBH_Save_2_info ? "RBH_Save_2" :
3727 "RBH_Save_?"), get_itbl(bqe));
3730 barf("Unexpected closure type %s in blocking queue of %p (%s)",
3731 info_type((StgClosure *)bqe), node, info_type(node));
3735 fputc('\n', stderr);
3739 Nice and easy: only TSOs on the blocking queue
3742 print_bq (StgClosure *node)
3746 ASSERT(node!=(StgClosure*)NULL); // sanity check
3747 for (tso = ((StgBlockingQueue*)node)->blocking_queue;
3748 tso != END_TSO_QUEUE;
3750 ASSERT(tso!=NULL && tso!=END_TSO_QUEUE); // sanity check
3751 ASSERT(get_itbl(tso)->type == TSO); // guess what, sanity check
3752 fprintf(stderr," TSO %d (%p),", tso->id, tso);
3754 fputc('\n', stderr);
3765 for (i=0, tso=run_queue_hd;
3766 tso != END_TSO_QUEUE;
3775 sched_belch(char *s, ...)
3780 fprintf(stderr, "scheduler (task %ld): ", osThreadId());
3782 fprintf(stderr, "== ");
3784 fprintf(stderr, "scheduler: ");
3786 vfprintf(stderr, s, ap);
3787 fprintf(stderr, "\n");
3794 //@node Index, , Debugging Routines, Main scheduling code
3798 //* StgMainThread:: @cindex\s-+StgMainThread
3799 //* awaken_blocked_queue:: @cindex\s-+awaken_blocked_queue
3800 //* blocked_queue_hd:: @cindex\s-+blocked_queue_hd
3801 //* blocked_queue_tl:: @cindex\s-+blocked_queue_tl
3802 //* context_switch:: @cindex\s-+context_switch
3803 //* createThread:: @cindex\s-+createThread
3804 //* gc_pending_cond:: @cindex\s-+gc_pending_cond
3805 //* initScheduler:: @cindex\s-+initScheduler
3806 //* interrupted:: @cindex\s-+interrupted
3807 //* next_thread_id:: @cindex\s-+next_thread_id
3808 //* print_bq:: @cindex\s-+print_bq
3809 //* run_queue_hd:: @cindex\s-+run_queue_hd
3810 //* run_queue_tl:: @cindex\s-+run_queue_tl
3811 //* sched_mutex:: @cindex\s-+sched_mutex
3812 //* schedule:: @cindex\s-+schedule
3813 //* take_off_run_queue:: @cindex\s-+take_off_run_queue
3814 //* term_mutex:: @cindex\s-+term_mutex