1 /* ---------------------------------------------------------------------------
2 * $Id: Schedule.c,v 1.47 2000/02/29 14:38:19 simonmar Exp $
4 * (c) The GHC Team, 1998-1999
8 * The main scheduling code in GranSim is quite different from that in std
9 * (concurrent) Haskell: while concurrent Haskell just iterates over the
10 * threads in the runnable queue, GranSim is event driven, i.e. it iterates
11 * over the events in the global event queue. -- HWL
12 * --------------------------------------------------------------------------*/
14 //@node Main scheduling code, , ,
15 //@section Main scheduling code
17 /* Version with scheduler monitor support for SMPs.
19 This design provides a high-level API to create and schedule threads etc.
20 as documented in the SMP design document.
22 It uses a monitor design controlled by a single mutex to exercise control
23 over accesses to shared data structures, and builds on the Posix threads
26 The majority of state is shared. In order to keep essential per-task state,
27 there is a Capability structure, which contains all the information
28 needed to run a thread: its STG registers, a pointer to its TSO, a
29 nursery etc. During STG execution, a pointer to the capability is
30 kept in a register (BaseReg).
32 In a non-SMP build, there is one global capability, namely MainRegTable.
39 //* Variables and Data structures::
41 //* Main scheduling loop::
42 //* Suspend and Resume::
44 //* Garbage Collextion Routines::
45 //* Blocking Queue Routines::
46 //* Exception Handling Routines::
47 //* Debugging Routines::
51 //@node Includes, Variables and Data structures, Main scheduling code, Main scheduling code
52 //@subsection Includes
60 #include "StgStartup.h"
64 #include "StgMiscClosures.h"
66 #include "Evaluator.h"
67 #include "Exception.h"
71 #include "Profiling.h"
75 #if defined(GRAN) || defined(PAR)
76 # include "GranSimRts.h"
78 # include "ParallelRts.h"
79 # include "Parallel.h"
80 # include "ParallelDebug.h"
87 //@node Variables and Data structures, Prototypes, Includes, Main scheduling code
88 //@subsection Variables and Data structures
92 * These are the threads which clients have requested that we run.
94 * In an SMP build, we might have several concurrent clients all
95 * waiting for results, and each one will wait on a condition variable
96 * until the result is available.
98 * In non-SMP, clients are strictly nested: the first client calls
99 * into the RTS, which might call out again to C with a _ccall_GC, and
100 * eventually re-enter the RTS.
102 * Main threads information is kept in a linked list:
104 //@cindex StgMainThread
105 typedef struct StgMainThread_ {
107 SchedulerStatus stat;
110 pthread_cond_t wakeup;
112 struct StgMainThread_ *link;
115 /* Main thread queue.
116 * Locks required: sched_mutex.
118 static StgMainThread *main_threads;
121 * Locks required: sched_mutex.
126 StgTSO* ActiveTSO = NULL; /* for assigning system costs; GranSim-Light only */
127 /* rtsTime TimeOfNextEvent, EndOfTimeSlice; now in GranSim.c */
130 In GranSim we have a runable and a blocked queue for each processor.
131 In order to minimise code changes new arrays run_queue_hds/tls
132 are created. run_queue_hd is then a short cut (macro) for
133 run_queue_hds[CurrentProc] (see GranSim.h).
136 StgTSO *run_queue_hds[MAX_PROC], *run_queue_tls[MAX_PROC];
137 StgTSO *blocked_queue_hds[MAX_PROC], *blocked_queue_tls[MAX_PROC];
138 StgTSO *ccalling_threadss[MAX_PROC];
142 //@cindex run_queue_hd
143 //@cindex run_queue_tl
144 //@cindex blocked_queue_hd
145 //@cindex blocked_queue_tl
146 StgTSO *run_queue_hd, *run_queue_tl;
147 StgTSO *blocked_queue_hd, *blocked_queue_tl;
149 /* Threads suspended in _ccall_GC.
150 * Locks required: sched_mutex.
152 static StgTSO *suspended_ccalling_threads;
154 static void GetRoots(void);
155 static StgTSO *threadStackOverflow(StgTSO *tso);
158 /* KH: The following two flags are shared memory locations. There is no need
159 to lock them, since they are only unset at the end of a scheduler
163 /* flag set by signal handler to precipitate a context switch */
164 //@cindex context_switch
167 /* if this flag is set as well, give up execution */
168 //@cindex interrupted
171 /* Next thread ID to allocate.
172 * Locks required: sched_mutex
174 //@cindex next_thread_id
175 StgThreadID next_thread_id = 1;
178 * Pointers to the state of the current thread.
179 * Rule of thumb: if CurrentTSO != NULL, then we're running a Haskell
180 * thread. If CurrentTSO == NULL, then we're at the scheduler level.
183 /* The smallest stack size that makes any sense is:
184 * RESERVED_STACK_WORDS (so we can get back from the stack overflow)
185 * + sizeofW(StgStopFrame) (the stg_stop_thread_info frame)
186 * + 1 (the realworld token for an IO thread)
187 * + 1 (the closure to enter)
189 * A thread with this stack will bomb immediately with a stack
190 * overflow, which will increase its stack size.
193 #define MIN_STACK_WORDS (RESERVED_STACK_WORDS + sizeofW(StgStopFrame) + 2)
195 /* Free capability list.
196 * Locks required: sched_mutex.
199 //@cindex free_capabilities
200 //@cindex n_free_capabilities
201 Capability *free_capabilities; /* Available capabilities for running threads */
202 nat n_free_capabilities; /* total number of available capabilities */
204 //@cindex MainRegTable
205 Capability MainRegTable; /* for non-SMP, we have one global capability */
209 StgTSO *CurrentTSOs[MAX_PROC];
216 /* All our current task ids, saved in case we need to kill them later.
223 void addToBlockedQueue ( StgTSO *tso );
225 static void schedule ( void );
226 void interruptStgRts ( void );
227 static StgTSO * createThread_ ( nat size, rtsBool have_lock );
230 static void sched_belch(char *s, ...);
234 //@cindex sched_mutex
236 //@cindex thread_ready_cond
237 //@cindex gc_pending_cond
238 pthread_mutex_t sched_mutex = PTHREAD_MUTEX_INITIALIZER;
239 pthread_mutex_t term_mutex = PTHREAD_MUTEX_INITIALIZER;
240 pthread_cond_t thread_ready_cond = PTHREAD_COND_INITIALIZER;
241 pthread_cond_t gc_pending_cond = PTHREAD_COND_INITIALIZER;
248 rtsTime TimeOfLastYield;
252 * The thread state for the main thread.
253 // ToDo: check whether not needed any more
258 //@node Prototypes, Main scheduling loop, Variables and Data structures, Main scheduling code
259 //@subsection Prototypes
261 //@node Main scheduling loop, Suspend and Resume, Prototypes, Main scheduling code
262 //@subsection Main scheduling loop
264 /* ---------------------------------------------------------------------------
265 Main scheduling loop.
267 We use round-robin scheduling, each thread returning to the
268 scheduler loop when one of these conditions is detected:
271 * timer expires (thread yields)
276 Locking notes: we acquire the scheduler lock once at the beginning
277 of the scheduler loop, and release it when
279 * running a thread, or
280 * waiting for work, or
281 * waiting for a GC to complete.
283 ------------------------------------------------------------------------ */
290 StgThreadReturnCode ret;
299 ACQUIRE_LOCK(&sched_mutex);
302 # error ToDo: implement GranSim scheduler
304 while (!GlobalStopPending) { /* GlobalStopPending set in par_exit */
306 if (PendingFetches != END_BF_QUEUE) {
313 /* If we're interrupted (the user pressed ^C, or some other
314 * termination condition occurred), kill all the currently running
318 IF_DEBUG(scheduler, sched_belch("interrupted"));
319 for (t = run_queue_hd; t != END_TSO_QUEUE; t = t->link) {
322 for (t = blocked_queue_hd; t != END_TSO_QUEUE; t = t->link) {
325 run_queue_hd = run_queue_tl = END_TSO_QUEUE;
326 blocked_queue_hd = blocked_queue_tl = END_TSO_QUEUE;
329 /* Go through the list of main threads and wake up any
330 * clients whose computations have finished. ToDo: this
331 * should be done more efficiently without a linear scan
332 * of the main threads list, somehow...
336 StgMainThread *m, **prev;
337 prev = &main_threads;
338 for (m = main_threads; m != NULL; m = m->link) {
339 switch (m->tso->whatNext) {
342 *(m->ret) = (StgClosure *)m->tso->sp[0];
346 pthread_cond_broadcast(&m->wakeup);
351 m->stat = Interrupted;
355 pthread_cond_broadcast(&m->wakeup);
363 /* If our main thread has finished or been killed, return.
366 StgMainThread *m = main_threads;
367 if (m->tso->whatNext == ThreadComplete
368 || m->tso->whatNext == ThreadKilled) {
369 main_threads = main_threads->link;
370 if (m->tso->whatNext == ThreadComplete) {
371 /* we finished successfully, fill in the return value */
372 if (m->ret) { *(m->ret) = (StgClosure *)m->tso->sp[0]; };
377 m->stat = Interrupted;
387 /* Top up the run queue from our spark pool. We try to make the
388 * number of threads in the run queue equal to the number of
393 nat n = n_free_capabilities;
394 StgTSO *tso = run_queue_hd;
396 /* Count the run queue */
397 while (n > 0 && tso != END_TSO_QUEUE) {
406 break; /* no more sparks in the pool */
408 /* I'd prefer this to be done in activateSpark -- HWL */
409 /* tricky - it needs to hold the scheduler lock and
410 * not try to re-acquire it -- SDM */
412 tso = createThread_(RtsFlags.GcFlags.initialStkSize, rtsTrue);
413 pushClosure(tso,spark);
414 PUSH_ON_RUN_QUEUE(tso);
416 advisory_thread_count++;
420 sched_belch("turning spark of closure %p into a thread",
421 (StgClosure *)spark));
424 /* We need to wake up the other tasks if we just created some
427 if (n_free_capabilities - n > 1) {
428 pthread_cond_signal(&thread_ready_cond);
433 /* Check whether any waiting threads need to be woken up. If the
434 * run queue is empty, and there are no other tasks running, we
435 * can wait indefinitely for something to happen.
436 * ToDo: what if another client comes along & requests another
439 if (blocked_queue_hd != END_TSO_QUEUE) {
441 (run_queue_hd == END_TSO_QUEUE)
443 && (n_free_capabilities == RtsFlags.ParFlags.nNodes)
448 /* check for signals each time around the scheduler */
450 if (signals_pending()) {
451 start_signal_handlers();
455 /* Detect deadlock: when we have no threads to run, there are
456 * no threads waiting on I/O or sleeping, and all the other
457 * tasks are waiting for work, we must have a deadlock. Inform
458 * all the main threads.
461 if (blocked_queue_hd == END_TSO_QUEUE
462 && run_queue_hd == END_TSO_QUEUE
463 && (n_free_capabilities == RtsFlags.ParFlags.nNodes)
466 for (m = main_threads; m != NULL; m = m->link) {
469 pthread_cond_broadcast(&m->wakeup);
474 if (blocked_queue_hd == END_TSO_QUEUE
475 && run_queue_hd == END_TSO_QUEUE) {
476 StgMainThread *m = main_threads;
479 main_threads = m->link;
485 /* If there's a GC pending, don't do anything until it has
489 IF_DEBUG(scheduler,sched_belch("waiting for GC"));
490 pthread_cond_wait(&gc_pending_cond, &sched_mutex);
493 /* block until we've got a thread on the run queue and a free
496 while (run_queue_hd == END_TSO_QUEUE || free_capabilities == NULL) {
497 IF_DEBUG(scheduler, sched_belch("waiting for work"));
498 pthread_cond_wait(&thread_ready_cond, &sched_mutex);
499 IF_DEBUG(scheduler, sched_belch("work now available"));
504 # error ToDo: implement GranSim scheduler
506 /* ToDo: phps merge with spark activation above */
507 /* check whether we have local work and send requests if we have none */
508 if (run_queue_hd == END_TSO_QUEUE) { /* no runnable threads */
509 /* :-[ no local threads => look out for local sparks */
510 if (advisory_thread_count < RtsFlags.ParFlags.maxThreads &&
511 (pending_sparks_hd[REQUIRED_POOL] < pending_sparks_tl[REQUIRED_POOL] ||
512 pending_sparks_hd[ADVISORY_POOL] < pending_sparks_tl[ADVISORY_POOL])) {
514 * ToDo: add GC code check that we really have enough heap afterwards!!
516 * If we're here (no runnable threads) and we have pending
517 * sparks, we must have a space problem. Get enough space
518 * to turn one of those pending sparks into a
522 spark = findSpark(); /* get a spark */
523 if (spark != (rtsSpark) NULL) {
524 tso = activateSpark(spark); /* turn the spark into a thread */
525 IF_PAR_DEBUG(verbose,
526 belch("== [%x] schedule: Created TSO %p (%d); %d threads active",
527 mytid, tso, tso->id, advisory_thread_count));
529 if (tso==END_TSO_QUEUE) { /* failed to activate spark->back to loop */
530 belch("^^ failed to activate spark");
532 } /* otherwise fall through & pick-up new tso */
534 IF_PAR_DEBUG(verbose,
535 belch("^^ no local sparks (spark pool contains only NFs: %d)",
536 spark_queue_len(ADVISORY_POOL)));
540 /* =8-[ no local sparks => look for work on other PEs */
543 * We really have absolutely no work. Send out a fish
544 * (there may be some out there already), and wait for
545 * something to arrive. We clearly can't run any threads
546 * until a SCHEDULE or RESUME arrives, and so that's what
547 * we're hoping to see. (Of course, we still have to
548 * respond to other types of messages.)
551 outstandingFishes < RtsFlags.ParFlags.maxFishes ) { // &&
552 // (last_fish_arrived_at+FISH_DELAY < CURRENT_TIME)) {
553 /* fishing set in sendFish, processFish;
554 avoid flooding system with fishes via delay */
556 sendFish(pe, mytid, NEW_FISH_AGE, NEW_FISH_HISTORY,
564 } else if (PacketsWaiting()) { /* Look for incoming messages */
568 /* Now we are sure that we have some work available */
569 ASSERT(run_queue_hd != END_TSO_QUEUE);
570 /* Take a thread from the run queue, if we have work */
571 t = take_off_run_queue(END_TSO_QUEUE);
573 /* ToDo: write something to the log-file
574 if (RTSflags.ParFlags.granSimStats && !sameThread)
575 DumpGranEvent(GR_SCHEDULE, RunnableThreadsHd);
580 IF_DEBUG(scheduler, belch("--^^ %d sparks on [%#x] (hd=%x; tl=%x; lim=%x)",
581 spark_queue_len(ADVISORY_POOL), CURRENT_PROC,
582 pending_sparks_hd[ADVISORY_POOL],
583 pending_sparks_tl[ADVISORY_POOL],
584 pending_sparks_lim[ADVISORY_POOL]));
586 IF_DEBUG(scheduler, belch("--== %d threads on [%#x] (hd=%x; tl=%x)",
587 run_queue_len(), CURRENT_PROC,
588 run_queue_hd, run_queue_tl));
592 we are running a different TSO, so write a schedule event to log file
593 NB: If we use fair scheduling we also have to write a deschedule
594 event for LastTSO; with unfair scheduling we know that the
595 previous tso has blocked whenever we switch to another tso, so
596 we don't need it in GUM for now
598 DumpRawGranEvent(CURRENT_PROC, CURRENT_PROC,
599 GR_SCHEDULE, t, (StgClosure *)NULL, 0, 0);
603 #else /* !GRAN && !PAR */
605 /* grab a thread from the run queue
608 IF_DEBUG(sanity,checkTSO(t));
615 cap = free_capabilities;
616 free_capabilities = cap->link;
617 n_free_capabilities--;
622 cap->rCurrentTSO = t;
624 /* set the context_switch flag
626 if (run_queue_hd == END_TSO_QUEUE)
631 RELEASE_LOCK(&sched_mutex);
633 IF_DEBUG(scheduler,sched_belch("running thread %d", t->id));
635 /* +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ */
636 /* Run the current thread
638 switch (cap->rCurrentTSO->whatNext) {
641 /* Thread already finished, return to scheduler. */
642 ret = ThreadFinished;
645 ret = StgRun((StgFunPtr) stg_enterStackTop, cap);
648 ret = StgRun((StgFunPtr) stg_returnToStackTop, cap);
650 case ThreadEnterHugs:
654 IF_DEBUG(scheduler,sched_belch("entering Hugs"));
655 c = (StgClosure *)(cap->rCurrentTSO->sp[0]);
656 cap->rCurrentTSO->sp += 1;
661 barf("Panic: entered a BCO but no bytecode interpreter in this build");
664 barf("schedule: invalid whatNext field");
666 /* +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ */
668 /* Costs for the scheduler are assigned to CCS_SYSTEM */
673 ACQUIRE_LOCK(&sched_mutex);
676 IF_DEBUG(scheduler,fprintf(stderr,"scheduler (task %ld): ", pthread_self()););
678 IF_DEBUG(scheduler,fprintf(stderr,"scheduler: "););
680 t = cap->rCurrentTSO;
684 /* make all the running tasks block on a condition variable,
685 * maybe set context_switch and wait till they all pile in,
686 * then have them wait on a GC condition variable.
688 IF_DEBUG(scheduler,belch("thread %ld stopped: HeapOverflow", t->id));
691 ready_to_gc = rtsTrue;
692 context_switch = 1; /* stop other threads ASAP */
693 PUSH_ON_RUN_QUEUE(t);
697 /* just adjust the stack for this thread, then pop it back
700 IF_DEBUG(scheduler,belch("thread %ld stopped, StackOverflow", t->id));
704 /* enlarge the stack */
705 StgTSO *new_t = threadStackOverflow(t);
707 /* This TSO has moved, so update any pointers to it from the
708 * main thread stack. It better not be on any other queues...
711 for (m = main_threads; m != NULL; m = m->link) {
717 ready_to_gc = rtsTrue;
719 PUSH_ON_RUN_QUEUE(new_t);
726 DumpGranEvent(GR_DESCHEDULE, t));
727 globalGranStats.tot_yields++;
730 DumpGranEvent(GR_DESCHEDULE, t));
732 /* put the thread back on the run queue. Then, if we're ready to
733 * GC, check whether this is the last task to stop. If so, wake
734 * up the GC thread. getThread will block during a GC until the
738 if (t->whatNext == ThreadEnterHugs) {
739 /* ToDo: or maybe a timer expired when we were in Hugs?
740 * or maybe someone hit ctrl-C
742 belch("thread %ld stopped to switch to Hugs", t->id);
744 belch("thread %ld stopped, yielding", t->id);
748 APPEND_TO_RUN_QUEUE(t);
753 # error ToDo: implement GranSim scheduler
756 DumpGranEvent(GR_DESCHEDULE, t));
759 /* don't need to do anything. Either the thread is blocked on
760 * I/O, in which case we'll have called addToBlockedQueue
761 * previously, or it's blocked on an MVar or Blackhole, in which
762 * case it'll be on the relevant queue already.
765 fprintf(stderr, "thread %d stopped, ", t->id);
766 printThreadBlockage(t);
767 fprintf(stderr, "\n"));
772 /* Need to check whether this was a main thread, and if so, signal
773 * the task that started it with the return value. If we have no
774 * more main threads, we probably need to stop all the tasks until
777 IF_DEBUG(scheduler,belch("thread %ld finished", t->id));
778 t->whatNext = ThreadComplete;
780 // ToDo: endThread(t, CurrentProc); // clean-up the thread
782 advisory_thread_count--;
783 if (RtsFlags.ParFlags.ParStats.Full)
784 DumpEndEvent(CURRENT_PROC, t, rtsFalse /* not mandatory */);
789 barf("doneThread: invalid thread return code");
793 cap->link = free_capabilities;
794 free_capabilities = cap;
795 n_free_capabilities++;
799 if (ready_to_gc && n_free_capabilities == RtsFlags.ParFlags.nNodes)
804 /* everybody back, start the GC.
805 * Could do it in this thread, or signal a condition var
806 * to do it in another thread. Either way, we need to
807 * broadcast on gc_pending_cond afterward.
810 IF_DEBUG(scheduler,sched_belch("doing GC"));
812 GarbageCollect(GetRoots);
813 ready_to_gc = rtsFalse;
815 pthread_cond_broadcast(&gc_pending_cond);
820 IF_GRAN_DEBUG(unused,
821 print_eventq(EventHd));
823 event = get_next_event();
827 /* ToDo: wait for next message to arrive rather than busy wait */
832 t = take_off_run_queue(END_TSO_QUEUE);
835 } /* end of while(1) */
838 /* A hack for Hugs concurrency support. Needs sanitisation (?) */
839 void deleteAllThreads ( void )
842 IF_DEBUG(scheduler,sched_belch("deleteAllThreads()"));
843 for (t = run_queue_hd; t != END_TSO_QUEUE; t = t->link) {
846 for (t = blocked_queue_hd; t != END_TSO_QUEUE; t = t->link) {
849 run_queue_hd = run_queue_tl = END_TSO_QUEUE;
850 blocked_queue_hd = blocked_queue_tl = END_TSO_QUEUE;
853 /* startThread and insertThread are now in GranSim.c -- HWL */
855 //@node Suspend and Resume, Run queue code, Main scheduling loop, Main scheduling code
856 //@subsection Suspend and Resume
858 /* ---------------------------------------------------------------------------
859 * Suspending & resuming Haskell threads.
861 * When making a "safe" call to C (aka _ccall_GC), the task gives back
862 * its capability before calling the C function. This allows another
863 * task to pick up the capability and carry on running Haskell
864 * threads. It also means that if the C call blocks, it won't lock
867 * The Haskell thread making the C call is put to sleep for the
868 * duration of the call, on the susepended_ccalling_threads queue. We
869 * give out a token to the task, which it can use to resume the thread
870 * on return from the C function.
871 * ------------------------------------------------------------------------- */
874 suspendThread( Capability *cap )
878 ACQUIRE_LOCK(&sched_mutex);
881 sched_belch("thread %d did a _ccall_gc\n", cap->rCurrentTSO->id));
883 threadPaused(cap->rCurrentTSO);
884 cap->rCurrentTSO->link = suspended_ccalling_threads;
885 suspended_ccalling_threads = cap->rCurrentTSO;
887 /* Use the thread ID as the token; it should be unique */
888 tok = cap->rCurrentTSO->id;
891 cap->link = free_capabilities;
892 free_capabilities = cap;
893 n_free_capabilities++;
896 RELEASE_LOCK(&sched_mutex);
901 resumeThread( StgInt tok )
906 ACQUIRE_LOCK(&sched_mutex);
908 prev = &suspended_ccalling_threads;
909 for (tso = suspended_ccalling_threads;
910 tso != END_TSO_QUEUE;
911 prev = &tso->link, tso = tso->link) {
912 if (tso->id == (StgThreadID)tok) {
917 if (tso == END_TSO_QUEUE) {
918 barf("resumeThread: thread not found");
922 while (free_capabilities == NULL) {
923 IF_DEBUG(scheduler, sched_belch("waiting to resume"));
924 pthread_cond_wait(&thread_ready_cond, &sched_mutex);
925 IF_DEBUG(scheduler, sched_belch("resuming thread %d", tso->id));
927 cap = free_capabilities;
928 free_capabilities = cap->link;
929 n_free_capabilities--;
934 cap->rCurrentTSO = tso;
936 RELEASE_LOCK(&sched_mutex);
941 /* ---------------------------------------------------------------------------
943 * ------------------------------------------------------------------------ */
944 static void unblockThread(StgTSO *tso);
946 /* ---------------------------------------------------------------------------
947 * Comparing Thread ids.
949 * This is used from STG land in the implementation of the
950 * instances of Eq/Ord for ThreadIds.
951 * ------------------------------------------------------------------------ */
953 int cmp_thread(const StgTSO *tso1, const StgTSO *tso2)
955 StgThreadID id1 = tso1->id;
956 StgThreadID id2 = tso2->id;
958 if (id1 < id2) return (-1);
959 if (id1 > id2) return 1;
963 /* ---------------------------------------------------------------------------
966 The new thread starts with the given stack size. Before the
967 scheduler can run, however, this thread needs to have a closure
968 (and possibly some arguments) pushed on its stack. See
969 pushClosure() in Schedule.h.
971 createGenThread() and createIOThread() (in SchedAPI.h) are
972 convenient packaged versions of this function.
973 ------------------------------------------------------------------------ */
974 //@cindex createThread
976 /* currently pri (priority) is only used in a GRAN setup -- HWL */
978 createThread(nat stack_size, StgInt pri)
980 return createThread_(stack_size, rtsFalse, pri);
984 createThread_(nat size, rtsBool have_lock, StgInt pri)
988 createThread(nat stack_size)
990 return createThread_(stack_size, rtsFalse);
994 createThread_(nat size, rtsBool have_lock)
1000 /* First check whether we should create a thread at all */
1002 /* check that no more than RtsFlags.ParFlags.maxThreads threads are created */
1003 if (advisory_thread_count >= RtsFlags.ParFlags.maxThreads) {
1005 belch("{createThread}Daq ghuH: refusing to create another thread; no more than %d threads allowed (currently %d)",
1006 RtsFlags.ParFlags.maxThreads, advisory_thread_count);
1007 return END_TSO_QUEUE;
1013 ASSERT(!RtsFlags.GranFlags.Light || CurrentProc==0);
1016 // ToDo: check whether size = stack_size - TSO_STRUCT_SIZEW
1018 /* catch ridiculously small stack sizes */
1019 if (size < MIN_STACK_WORDS + TSO_STRUCT_SIZEW) {
1020 size = MIN_STACK_WORDS + TSO_STRUCT_SIZEW;
1023 tso = (StgTSO *)allocate(size);
1024 TICK_ALLOC_TSO(size-sizeofW(StgTSO),0);
1026 stack_size = size - TSO_STRUCT_SIZEW;
1028 // Hmm, this CCS_MAIN is not protected by a PROFILING cpp var;
1029 SET_HDR(tso, &TSO_info, CCS_MAIN);
1031 SET_GRAN_HDR(tso, ThisPE);
1033 tso->whatNext = ThreadEnterGHC;
1035 /* tso->id needs to be unique. For now we use a heavyweight mutex to
1036 protect the increment operation on next_thread_id.
1037 In future, we could use an atomic increment instead.
1040 if (!have_lock) { ACQUIRE_LOCK(&sched_mutex); }
1041 tso->id = next_thread_id++;
1042 if (!have_lock) { RELEASE_LOCK(&sched_mutex); }
1044 tso->why_blocked = NotBlocked;
1045 tso->blocked_exceptions = NULL;
1047 tso->splim = (P_)&(tso->stack) + RESERVED_STACK_WORDS;
1048 tso->stack_size = stack_size;
1049 tso->max_stack_size = round_to_mblocks(RtsFlags.GcFlags.maxStkSize)
1051 tso->sp = (P_)&(tso->stack) + stack_size;
1054 tso->prof.CCCS = CCS_MAIN;
1057 /* put a stop frame on the stack */
1058 tso->sp -= sizeofW(StgStopFrame);
1059 SET_HDR((StgClosure*)tso->sp,(StgInfoTable *)&stg_stop_thread_info,CCS_MAIN);
1060 tso->su = (StgUpdateFrame*)tso->sp;
1062 IF_DEBUG(scheduler,belch("---- Initialised TSO %ld (%p), stack size = %lx words",
1063 tso->id, tso, tso->stack_size));
1067 tso->link = END_TSO_QUEUE;
1068 /* uses more flexible routine in GranSim */
1069 insertThread(tso, CurrentProc);
1071 /* In a non-GranSim setup the pushing of a TSO onto the runq is separated
1077 tso->gran.pri = pri;
1078 tso->gran.magic = TSO_MAGIC; // debugging only
1079 tso->gran.sparkname = 0;
1080 tso->gran.startedat = CURRENT_TIME;
1081 tso->gran.exported = 0;
1082 tso->gran.basicblocks = 0;
1083 tso->gran.allocs = 0;
1084 tso->gran.exectime = 0;
1085 tso->gran.fetchtime = 0;
1086 tso->gran.fetchcount = 0;
1087 tso->gran.blocktime = 0;
1088 tso->gran.blockcount = 0;
1089 tso->gran.blockedat = 0;
1090 tso->gran.globalsparks = 0;
1091 tso->gran.localsparks = 0;
1092 if (RtsFlags.GranFlags.Light)
1093 tso->gran.clock = Now; /* local clock */
1095 tso->gran.clock = 0;
1097 IF_DEBUG(gran,printTSO(tso));
1099 tso->par.sparkname = 0;
1100 tso->par.startedat = CURRENT_TIME;
1101 tso->par.exported = 0;
1102 tso->par.basicblocks = 0;
1103 tso->par.allocs = 0;
1104 tso->par.exectime = 0;
1105 tso->par.fetchtime = 0;
1106 tso->par.fetchcount = 0;
1107 tso->par.blocktime = 0;
1108 tso->par.blockcount = 0;
1109 tso->par.blockedat = 0;
1110 tso->par.globalsparks = 0;
1111 tso->par.localsparks = 0;
1115 globalGranStats.tot_threads_created++;
1116 globalGranStats.threads_created_on_PE[CurrentProc]++;
1117 globalGranStats.tot_sq_len += spark_queue_len(CurrentProc);
1118 globalGranStats.tot_sq_probes++;
1121 IF_DEBUG(scheduler,sched_belch("created thread %ld, stack size = %lx words",
1122 tso->id, tso->stack_size));
1126 /* ---------------------------------------------------------------------------
1129 * scheduleThread puts a thread on the head of the runnable queue.
1130 * This will usually be done immediately after a thread is created.
1131 * The caller of scheduleThread must create the thread using e.g.
1132 * createThread and push an appropriate closure
1133 * on this thread's stack before the scheduler is invoked.
1134 * ------------------------------------------------------------------------ */
1137 scheduleThread(StgTSO *tso)
1139 ACQUIRE_LOCK(&sched_mutex);
1141 /* Put the new thread on the head of the runnable queue. The caller
1142 * better push an appropriate closure on this thread's stack
1143 * beforehand. In the SMP case, the thread may start running as
1144 * soon as we release the scheduler lock below.
1146 PUSH_ON_RUN_QUEUE(tso);
1149 IF_DEBUG(scheduler,printTSO(tso));
1150 RELEASE_LOCK(&sched_mutex);
1153 /* ---------------------------------------------------------------------------
1156 * Start up Posix threads to run each of the scheduler tasks.
1157 * I believe the task ids are not needed in the system as defined.
1159 * ------------------------------------------------------------------------ */
1163 taskStart( void *arg STG_UNUSED )
1170 /* ---------------------------------------------------------------------------
1173 * Initialise the scheduler. This resets all the queues - if the
1174 * queues contained any threads, they'll be garbage collected at the
1177 * This now calls startTasks(), so should only be called once! KH @ 25/10/99
1178 * ------------------------------------------------------------------------ */
1182 term_handler(int sig STG_UNUSED)
1185 ACQUIRE_LOCK(&term_mutex);
1187 RELEASE_LOCK(&term_mutex);
1192 //@cindex initScheduler
1199 for (i=0; i<=MAX_PROC; i++) {
1200 run_queue_hds[i] = END_TSO_QUEUE;
1201 run_queue_tls[i] = END_TSO_QUEUE;
1202 blocked_queue_hds[i] = END_TSO_QUEUE;
1203 blocked_queue_tls[i] = END_TSO_QUEUE;
1204 ccalling_threadss[i] = END_TSO_QUEUE;
1207 run_queue_hd = END_TSO_QUEUE;
1208 run_queue_tl = END_TSO_QUEUE;
1209 blocked_queue_hd = END_TSO_QUEUE;
1210 blocked_queue_tl = END_TSO_QUEUE;
1213 suspended_ccalling_threads = END_TSO_QUEUE;
1215 main_threads = NULL;
1220 enteredCAFs = END_CAF_LIST;
1222 /* Install the SIGHUP handler */
1225 struct sigaction action,oact;
1227 action.sa_handler = term_handler;
1228 sigemptyset(&action.sa_mask);
1229 action.sa_flags = 0;
1230 if (sigaction(SIGTERM, &action, &oact) != 0) {
1231 barf("can't install TERM handler");
1237 /* Allocate N Capabilities */
1240 Capability *cap, *prev;
1243 for (i = 0; i < RtsFlags.ParFlags.nNodes; i++) {
1244 cap = stgMallocBytes(sizeof(Capability), "initScheduler:capabilities");
1248 free_capabilities = cap;
1249 n_free_capabilities = RtsFlags.ParFlags.nNodes;
1251 IF_DEBUG(scheduler,fprintf(stderr,"scheduler: Allocated %d capabilities\n",
1252 n_free_capabilities););
1255 #if defined(SMP) || defined(PAR)
1268 /* make some space for saving all the thread ids */
1269 task_ids = stgMallocBytes(RtsFlags.ParFlags.nNodes * sizeof(task_info),
1270 "initScheduler:task_ids");
1272 /* and create all the threads */
1273 for (i = 0; i < RtsFlags.ParFlags.nNodes; i++) {
1274 r = pthread_create(&tid,NULL,taskStart,NULL);
1276 barf("startTasks: Can't create new Posix thread");
1278 task_ids[i].id = tid;
1279 task_ids[i].mut_time = 0.0;
1280 task_ids[i].mut_etime = 0.0;
1281 task_ids[i].gc_time = 0.0;
1282 task_ids[i].gc_etime = 0.0;
1283 task_ids[i].elapsedtimestart = elapsedtime();
1284 IF_DEBUG(scheduler,fprintf(stderr,"scheduler: Started task: %ld\n",tid););
1290 exitScheduler( void )
1295 /* Don't want to use pthread_cancel, since we'd have to install
1296 * these silly exception handlers (pthread_cleanup_{push,pop}) around
1300 /* Cancel all our tasks */
1301 for (i = 0; i < RtsFlags.ParFlags.nNodes; i++) {
1302 pthread_cancel(task_ids[i].id);
1305 /* Wait for all the tasks to terminate */
1306 for (i = 0; i < RtsFlags.ParFlags.nNodes; i++) {
1307 IF_DEBUG(scheduler,fprintf(stderr,"scheduler: waiting for task %ld\n",
1309 pthread_join(task_ids[i].id, NULL);
1313 /* Send 'em all a SIGHUP. That should shut 'em up.
1315 await_death = RtsFlags.ParFlags.nNodes;
1316 for (i = 0; i < RtsFlags.ParFlags.nNodes; i++) {
1317 pthread_kill(task_ids[i].id,SIGTERM);
1319 while (await_death > 0) {
1325 /* -----------------------------------------------------------------------------
1326 Managing the per-task allocation areas.
1328 Each capability comes with an allocation area. These are
1329 fixed-length block lists into which allocation can be done.
1331 ToDo: no support for two-space collection at the moment???
1332 -------------------------------------------------------------------------- */
1334 /* -----------------------------------------------------------------------------
1335 * waitThread is the external interface for running a new computataion
1336 * and waiting for the result.
1338 * In the non-SMP case, we create a new main thread, push it on the
1339 * main-thread stack, and invoke the scheduler to run it. The
1340 * scheduler will return when the top main thread on the stack has
1341 * completed or died, and fill in the necessary fields of the
1342 * main_thread structure.
1344 * In the SMP case, we create a main thread as before, but we then
1345 * create a new condition variable and sleep on it. When our new
1346 * main thread has completed, we'll be woken up and the status/result
1347 * will be in the main_thread struct.
1348 * -------------------------------------------------------------------------- */
1351 waitThread(StgTSO *tso, /*out*/StgClosure **ret)
1354 SchedulerStatus stat;
1356 ACQUIRE_LOCK(&sched_mutex);
1358 m = stgMallocBytes(sizeof(StgMainThread), "waitThread");
1364 pthread_cond_init(&m->wakeup, NULL);
1367 m->link = main_threads;
1370 IF_DEBUG(scheduler, fprintf(stderr, "scheduler: new main thread (%d)\n",
1375 pthread_cond_wait(&m->wakeup, &sched_mutex);
1376 } while (m->stat == NoStatus);
1379 ASSERT(m->stat != NoStatus);
1385 pthread_cond_destroy(&m->wakeup);
1388 IF_DEBUG(scheduler, fprintf(stderr, "scheduler: main thread (%d) finished\n",
1392 RELEASE_LOCK(&sched_mutex);
1397 //@node Run queue code, Garbage Collextion Routines, Suspend and Resume, Main scheduling code
1398 //@subsection Run queue code
1402 NB: In GranSim we have many run queues; run_queue_hd is actually a macro
1403 unfolding to run_queue_hds[CurrentProc], thus CurrentProc is an
1404 implicit global variable that has to be correct when calling these
1408 /* Put the new thread on the head of the runnable queue.
1409 * The caller of createThread better push an appropriate closure
1410 * on this thread's stack before the scheduler is invoked.
1412 static /* inline */ void
1413 add_to_run_queue(tso)
1416 ASSERT(tso!=run_queue_hd && tso!=run_queue_tl);
1417 tso->link = run_queue_hd;
1419 if (run_queue_tl == END_TSO_QUEUE) {
1424 /* Put the new thread at the end of the runnable queue. */
1425 static /* inline */ void
1426 push_on_run_queue(tso)
1429 ASSERT(get_itbl((StgClosure *)tso)->type == TSO);
1430 ASSERT(run_queue_hd!=NULL && run_queue_tl!=NULL);
1431 ASSERT(tso!=run_queue_hd && tso!=run_queue_tl);
1432 if (run_queue_hd == END_TSO_QUEUE) {
1435 run_queue_tl->link = tso;
1441 Should be inlined because it's used very often in schedule. The tso
1442 argument is actually only needed in GranSim, where we want to have the
1443 possibility to schedule *any* TSO on the run queue, irrespective of the
1444 actual ordering. Therefore, if tso is not the nil TSO then we traverse
1445 the run queue and dequeue the tso, adjusting the links in the queue.
1447 //@cindex take_off_run_queue
1448 static /* inline */ StgTSO*
1449 take_off_run_queue(StgTSO *tso) {
1453 qetlaHbogh Qu' ngaSbogh ghomDaQ {tso} yIteq!
1455 if tso is specified, unlink that tso from the run_queue (doesn't have
1456 to be at the beginning of the queue); GranSim only
1458 if (tso!=END_TSO_QUEUE) {
1459 /* find tso in queue */
1460 for (t=run_queue_hd, prev=END_TSO_QUEUE;
1461 t!=END_TSO_QUEUE && t!=tso;
1465 /* now actually dequeue the tso */
1466 if (prev!=END_TSO_QUEUE) {
1467 ASSERT(run_queue_hd!=t);
1468 prev->link = t->link;
1470 /* t is at beginning of thread queue */
1471 ASSERT(run_queue_hd==t);
1472 run_queue_hd = t->link;
1474 /* t is at end of thread queue */
1475 if (t->link==END_TSO_QUEUE) {
1476 ASSERT(t==run_queue_tl);
1477 run_queue_tl = prev;
1479 ASSERT(run_queue_tl!=t);
1481 t->link = END_TSO_QUEUE;
1483 /* take tso from the beginning of the queue; std concurrent code */
1485 if (t != END_TSO_QUEUE) {
1486 run_queue_hd = t->link;
1487 t->link = END_TSO_QUEUE;
1488 if (run_queue_hd == END_TSO_QUEUE) {
1489 run_queue_tl = END_TSO_QUEUE;
1498 //@node Garbage Collextion Routines, Blocking Queue Routines, Run queue code, Main scheduling code
1499 //@subsection Garbage Collextion Routines
1501 /* ---------------------------------------------------------------------------
1502 Where are the roots that we know about?
1504 - all the threads on the runnable queue
1505 - all the threads on the blocked queue
1506 - all the thread currently executing a _ccall_GC
1507 - all the "main threads"
1509 ------------------------------------------------------------------------ */
1511 /* This has to be protected either by the scheduler monitor, or by the
1512 garbage collection monitor (probably the latter).
1516 static void GetRoots(void)
1523 for (i=0; i<=RtsFlags.GranFlags.proc; i++) {
1524 if ((run_queue_hds[i] != END_TSO_QUEUE) && ((run_queue_hds[i] != NULL)))
1525 run_queue_hds[i] = (StgTSO *)MarkRoot((StgClosure *)run_queue_hds[i]);
1526 if ((run_queue_tls[i] != END_TSO_QUEUE) && ((run_queue_tls[i] != NULL)))
1527 run_queue_tls[i] = (StgTSO *)MarkRoot((StgClosure *)run_queue_tls[i]);
1529 if ((blocked_queue_hds[i] != END_TSO_QUEUE) && ((blocked_queue_hds[i] != NULL)))
1530 blocked_queue_hds[i] = (StgTSO *)MarkRoot((StgClosure *)blocked_queue_hds[i]);
1531 if ((blocked_queue_tls[i] != END_TSO_QUEUE) && ((blocked_queue_tls[i] != NULL)))
1532 blocked_queue_tls[i] = (StgTSO *)MarkRoot((StgClosure *)blocked_queue_tls[i]);
1533 if ((ccalling_threadss[i] != END_TSO_QUEUE) && ((ccalling_threadss[i] != NULL)))
1534 ccalling_threadss[i] = (StgTSO *)MarkRoot((StgClosure *)ccalling_threadss[i]);
1541 run_queue_hd = (StgTSO *)MarkRoot((StgClosure *)run_queue_hd);
1542 run_queue_tl = (StgTSO *)MarkRoot((StgClosure *)run_queue_tl);
1544 blocked_queue_hd = (StgTSO *)MarkRoot((StgClosure *)blocked_queue_hd);
1545 blocked_queue_tl = (StgTSO *)MarkRoot((StgClosure *)blocked_queue_tl);
1548 for (m = main_threads; m != NULL; m = m->link) {
1549 m->tso = (StgTSO *)MarkRoot((StgClosure *)m->tso);
1551 suspended_ccalling_threads =
1552 (StgTSO *)MarkRoot((StgClosure *)suspended_ccalling_threads);
1554 #if defined(SMP) || defined(PAR) || defined(GRAN)
1559 /* -----------------------------------------------------------------------------
1562 This is the interface to the garbage collector from Haskell land.
1563 We provide this so that external C code can allocate and garbage
1564 collect when called from Haskell via _ccall_GC.
1566 It might be useful to provide an interface whereby the programmer
1567 can specify more roots (ToDo).
1569 This needs to be protected by the GC condition variable above. KH.
1570 -------------------------------------------------------------------------- */
1572 void (*extra_roots)(void);
1577 GarbageCollect(GetRoots);
1583 GetRoots(); /* the scheduler's roots */
1584 extra_roots(); /* the user's roots */
1588 performGCWithRoots(void (*get_roots)(void))
1590 extra_roots = get_roots;
1592 GarbageCollect(AllRoots);
1595 /* -----------------------------------------------------------------------------
1598 If the thread has reached its maximum stack size, then raise the
1599 StackOverflow exception in the offending thread. Otherwise
1600 relocate the TSO into a larger chunk of memory and adjust its stack
1602 -------------------------------------------------------------------------- */
1605 threadStackOverflow(StgTSO *tso)
1607 nat new_stack_size, new_tso_size, diff, stack_words;
1611 IF_DEBUG(sanity,checkTSO(tso));
1612 if (tso->stack_size >= tso->max_stack_size) {
1614 /* If we're debugging, just print out the top of the stack */
1615 printStackChunk(tso->sp, stg_min(tso->stack+tso->stack_size,
1619 fprintf(stderr, "fatal: stack overflow in Hugs; aborting\n" );
1622 /* Send this thread the StackOverflow exception */
1623 raiseAsync(tso, (StgClosure *)&stackOverflow_closure);
1628 /* Try to double the current stack size. If that takes us over the
1629 * maximum stack size for this thread, then use the maximum instead.
1630 * Finally round up so the TSO ends up as a whole number of blocks.
1632 new_stack_size = stg_min(tso->stack_size * 2, tso->max_stack_size);
1633 new_tso_size = (nat)BLOCK_ROUND_UP(new_stack_size * sizeof(W_) +
1634 TSO_STRUCT_SIZE)/sizeof(W_);
1635 new_tso_size = round_to_mblocks(new_tso_size); /* Be MBLOCK-friendly */
1636 new_stack_size = new_tso_size - TSO_STRUCT_SIZEW;
1638 IF_DEBUG(scheduler, fprintf(stderr,"scheduler: increasing stack size from %d words to %d.\n", tso->stack_size, new_stack_size));
1640 dest = (StgTSO *)allocate(new_tso_size);
1641 TICK_ALLOC_TSO(new_tso_size-sizeofW(StgTSO),0);
1643 /* copy the TSO block and the old stack into the new area */
1644 memcpy(dest,tso,TSO_STRUCT_SIZE);
1645 stack_words = tso->stack + tso->stack_size - tso->sp;
1646 new_sp = (P_)dest + new_tso_size - stack_words;
1647 memcpy(new_sp, tso->sp, stack_words * sizeof(W_));
1649 /* relocate the stack pointers... */
1650 diff = (P_)new_sp - (P_)tso->sp; /* In *words* */
1651 dest->su = (StgUpdateFrame *) ((P_)dest->su + diff);
1653 dest->splim = (P_)dest->splim + (nat)((P_)dest - (P_)tso);
1654 dest->stack_size = new_stack_size;
1656 /* and relocate the update frame list */
1657 relocate_TSO(tso, dest);
1659 /* Mark the old TSO as relocated. We have to check for relocated
1660 * TSOs in the garbage collector and any primops that deal with TSOs.
1662 * It's important to set the sp and su values to just beyond the end
1663 * of the stack, so we don't attempt to scavenge any part of the
1666 tso->whatNext = ThreadRelocated;
1668 tso->sp = (P_)&(tso->stack[tso->stack_size]);
1669 tso->su = (StgUpdateFrame *)tso->sp;
1670 tso->why_blocked = NotBlocked;
1671 dest->mut_link = NULL;
1673 IF_DEBUG(sanity,checkTSO(tso));
1675 IF_DEBUG(scheduler,printTSO(dest));
1681 //@node Blocking Queue Routines, Exception Handling Routines, Garbage Collextion Routines, Main scheduling code
1682 //@subsection Blocking Queue Routines
1684 /* ---------------------------------------------------------------------------
1685 Wake up a queue that was blocked on some resource.
1686 ------------------------------------------------------------------------ */
1688 /* ToDo: check push_on_run_queue vs. PUSH_ON_RUN_QUEUE */
1692 unblockCount ( StgBlockingQueueElement *bqe, StgClosure *node )
1697 unblockCount ( StgBlockingQueueElement *bqe, StgClosure *node )
1699 /* write RESUME events to log file and
1700 update blocked and fetch time (depending on type of the orig closure) */
1701 if (RtsFlags.ParFlags.ParStats.Full) {
1702 DumpRawGranEvent(CURRENT_PROC, CURRENT_PROC,
1703 GR_RESUMEQ, ((StgTSO *)bqe), ((StgTSO *)bqe)->block_info.closure,
1704 0, 0 /* spark_queue_len(ADVISORY_POOL) */);
1706 switch (get_itbl(node)->type) {
1708 ((StgTSO *)bqe)->par.fetchtime += CURRENT_TIME-((StgTSO *)bqe)->par.blockedat;
1713 ((StgTSO *)bqe)->par.blocktime += CURRENT_TIME-((StgTSO *)bqe)->par.blockedat;
1716 barf("{unblockOneLocked}Daq Qagh: unexpected closure in blocking queue");
1723 static StgBlockingQueueElement *
1724 unblockOneLocked(StgBlockingQueueElement *bqe, StgClosure *node)
1726 StgBlockingQueueElement *next;
1727 PEs node_loc, tso_loc;
1729 node_loc = where_is(node); // should be lifted out of loop
1730 tso = (StgTSO *)bqe; // wastes an assignment to get the type right
1731 tso_loc = where_is(tso);
1732 if (IS_LOCAL_TO(PROCS(node),tso_loc)) { // TSO is local
1733 /* !fake_fetch => TSO is on CurrentProc is same as IS_LOCAL_TO */
1734 ASSERT(CurrentProc!=node_loc || tso_loc==CurrentProc);
1735 bq_processing_time += RtsFlags.GranFlags.Costs.lunblocktime;
1736 // insertThread(tso, node_loc);
1737 new_event(tso_loc, tso_loc,
1738 CurrentTime[CurrentProc]+bq_processing_time,
1740 tso, node, (rtsSpark*)NULL);
1741 tso->link = END_TSO_QUEUE; // overwrite link just to be sure
1744 } else { // TSO is remote (actually should be FMBQ)
1745 bq_processing_time += RtsFlags.GranFlags.Costs.mpacktime;
1746 bq_processing_time += RtsFlags.GranFlags.Costs.gunblocktime;
1747 new_event(tso_loc, CurrentProc,
1748 CurrentTime[CurrentProc]+bq_processing_time+
1749 RtsFlags.GranFlags.Costs.latency,
1751 tso, node, (rtsSpark*)NULL);
1752 tso->link = END_TSO_QUEUE; // overwrite link just to be sure
1753 bq_processing_time += RtsFlags.GranFlags.Costs.mtidytime;
1756 /* the thread-queue-overhead is accounted for in either Resume or UnblockThread */
1758 fprintf(stderr," %s TSO %d (%p) [PE %d] (blocked_on=%p) (next=%p) ,",
1759 (node_loc==tso_loc ? "Local" : "Global"),
1760 tso->id, tso, CurrentProc, tso->blocked_on, tso->link))
1761 tso->blocked_on = NULL;
1762 IF_DEBUG(scheduler,belch("-- Waking up thread %ld (%p)",
1766 /* if this is the BQ of an RBH, we have to put back the info ripped out of
1767 the closure to make room for the anchor of the BQ */
1768 if (next!=END_BQ_QUEUE) {
1769 ASSERT(get_itbl(node)->type == RBH && get_itbl(next)->type == CONSTR);
1771 ASSERT((info_ptr==&RBH_Save_0_info) ||
1772 (info_ptr==&RBH_Save_1_info) ||
1773 (info_ptr==&RBH_Save_2_info));
1775 /* cf. convertToRBH in RBH.c for writing the RBHSave closure */
1776 ((StgRBH *)node)->blocking_queue = ((StgRBHSave *)next)->payload[0];
1777 ((StgRBH *)node)->mut_link = ((StgRBHSave *)next)->payload[1];
1780 belch("## Filled in RBH_Save for %p (%s) at end of AwBQ",
1781 node, info_type(node)));
1785 static StgBlockingQueueElement *
1786 unblockOneLocked(StgBlockingQueueElement *bqe, StgClosure *node)
1788 StgBlockingQueueElement *next;
1790 switch (get_itbl(bqe)->type) {
1792 ASSERT(((StgTSO *)bqe)->why_blocked != NotBlocked);
1793 /* if it's a TSO just push it onto the run_queue */
1795 // ((StgTSO *)bqe)->link = END_TSO_QUEUE; // debugging?
1796 PUSH_ON_RUN_QUEUE((StgTSO *)bqe);
1798 unblockCount(bqe, node);
1799 /* reset blocking status after dumping event */
1800 ((StgTSO *)bqe)->why_blocked = NotBlocked;
1804 /* if it's a BLOCKED_FETCH put it on the PendingFetches list */
1806 bqe->link = PendingFetches;
1807 PendingFetches = bqe;
1811 /* can ignore this case in a non-debugging setup;
1812 see comments on RBHSave closures above */
1814 /* check that the closure is an RBHSave closure */
1815 ASSERT(get_itbl((StgClosure *)bqe) == &RBH_Save_0_info ||
1816 get_itbl((StgClosure *)bqe) == &RBH_Save_1_info ||
1817 get_itbl((StgClosure *)bqe) == &RBH_Save_2_info);
1821 barf("{unblockOneLocked}Daq Qagh: Unexpected IP (%#lx; %s) in blocking queue at %#lx\n",
1822 get_itbl((StgClosure *)bqe), info_type((StgClosure *)bqe),
1826 // IF_DEBUG(scheduler,sched_belch("waking up thread %ld", tso->id));
1830 #else /* !GRAN && !PAR */
1832 unblockOneLocked(StgTSO *tso)
1836 ASSERT(get_itbl(tso)->type == TSO);
1837 ASSERT(tso->why_blocked != NotBlocked);
1838 tso->why_blocked = NotBlocked;
1840 PUSH_ON_RUN_QUEUE(tso);
1842 IF_DEBUG(scheduler,sched_belch("waking up thread %ld", tso->id));
1849 unblockOne(StgTSO *tso, StgClosure *node)
1851 ACQUIRE_LOCK(&sched_mutex);
1852 tso = unblockOneLocked(tso, node);
1853 RELEASE_LOCK(&sched_mutex);
1858 unblockOne(StgTSO *tso, StgClosure *node)
1860 ACQUIRE_LOCK(&sched_mutex);
1861 tso = unblockOneLocked(tso, node);
1862 RELEASE_LOCK(&sched_mutex);
1867 unblockOne(StgTSO *tso)
1869 ACQUIRE_LOCK(&sched_mutex);
1870 tso = unblockOneLocked(tso);
1871 RELEASE_LOCK(&sched_mutex);
1878 awakenBlockedQueue(StgBlockingQueueElement *q, StgClosure *node)
1880 StgBlockingQueueElement *bqe, *next;
1882 PEs node_loc, tso_loc;
1883 rtsTime bq_processing_time = 0;
1884 nat len = 0, len_local = 0;
1887 belch("## AwBQ for node %p on PE %d @ %ld by TSO %d (%p): ", \
1888 node, CurrentProc, CurrentTime[CurrentProc],
1889 CurrentTSO->id, CurrentTSO));
1891 node_loc = where_is(node);
1893 ASSERT(get_itbl(q)->type == TSO || // q is either a TSO or an RBHSave
1894 get_itbl(q)->type == CONSTR); // closure (type constructor)
1895 ASSERT(is_unique(node));
1897 /* FAKE FETCH: magically copy the node to the tso's proc;
1898 no Fetch necessary because in reality the node should not have been
1899 moved to the other PE in the first place
1901 if (CurrentProc!=node_loc) {
1903 belch("## node %p is on PE %d but CurrentProc is %d (TSO %d); assuming fake fetch and adjusting bitmask (old: %#x)",
1904 node, node_loc, CurrentProc, CurrentTSO->id,
1905 // CurrentTSO, where_is(CurrentTSO),
1906 node->header.gran.procs));
1907 node->header.gran.procs = (node->header.gran.procs) | PE_NUMBER(CurrentProc);
1909 belch("## new bitmask of node %p is %#x",
1910 node, node->header.gran.procs));
1911 if (RtsFlags.GranFlags.GranSimStats.Global) {
1912 globalGranStats.tot_fake_fetches++;
1917 // ToDo: check: ASSERT(CurrentProc==node_loc);
1918 while (get_itbl(bqe)->type==TSO) { // q != END_TSO_QUEUE) {
1921 bqe points to the current element in the queue
1922 next points to the next element in the queue
1924 //tso = (StgTSO *)bqe; // wastes an assignment to get the type right
1925 //tso_loc = where_is(tso);
1926 bqe = unblockOneLocked(bqe, node);
1929 /* statistics gathering */
1930 /* ToDo: fix counters
1931 if (RtsFlags.GranFlags.GranSimStats.Global) {
1932 globalGranStats.tot_bq_processing_time += bq_processing_time;
1933 globalGranStats.tot_bq_len += len; // total length of all bqs awakened
1934 globalGranStats.tot_bq_len_local += len_local; // same for local TSOs only
1935 globalGranStats.tot_awbq++; // total no. of bqs awakened
1938 fprintf(stderr,"## BQ Stats of %p: [%d entries, %d local] %s\n",
1939 node, len, len_local, (next!=END_TSO_QUEUE) ? "RBH" : ""));
1944 awakenBlockedQueue(StgBlockingQueueElement *q, StgClosure *node)
1946 StgBlockingQueueElement *bqe, *next;
1948 ACQUIRE_LOCK(&sched_mutex);
1950 IF_PAR_DEBUG(verbose,
1951 belch("## AwBQ for node %p on [%x]: ",
1954 ASSERT(get_itbl(q)->type == TSO ||
1955 get_itbl(q)->type == BLOCKED_FETCH ||
1956 get_itbl(q)->type == CONSTR);
1959 while (get_itbl(bqe)->type==TSO ||
1960 get_itbl(bqe)->type==BLOCKED_FETCH) {
1961 bqe = unblockOneLocked(bqe, node);
1963 RELEASE_LOCK(&sched_mutex);
1966 #else /* !GRAN && !PAR */
1968 awakenBlockedQueue(StgTSO *tso)
1970 ACQUIRE_LOCK(&sched_mutex);
1971 while (tso != END_TSO_QUEUE) {
1972 tso = unblockOneLocked(tso);
1974 RELEASE_LOCK(&sched_mutex);
1978 //@node Exception Handling Routines, Debugging Routines, Blocking Queue Routines, Main scheduling code
1979 //@subsection Exception Handling Routines
1981 /* ---------------------------------------------------------------------------
1983 - usually called inside a signal handler so it mustn't do anything fancy.
1984 ------------------------------------------------------------------------ */
1987 interruptStgRts(void)
1993 /* -----------------------------------------------------------------------------
1996 This is for use when we raise an exception in another thread, which
1998 This has nothing to do with the UnblockThread event in GranSim. -- HWL
1999 -------------------------------------------------------------------------- */
2002 unblockThread(StgTSO *tso)
2006 ACQUIRE_LOCK(&sched_mutex);
2007 switch (tso->why_blocked) {
2010 return; /* not blocked */
2013 ASSERT(get_itbl(tso->block_info.closure)->type == MVAR);
2015 StgTSO *last_tso = END_TSO_QUEUE;
2016 StgMVar *mvar = (StgMVar *)(tso->block_info.closure);
2019 for (t = mvar->head; t != END_TSO_QUEUE;
2020 last = &t->link, last_tso = t, t = t->link) {
2023 if (mvar->tail == tso) {
2024 mvar->tail = last_tso;
2029 barf("unblockThread (MVAR): TSO not found");
2032 case BlockedOnBlackHole:
2033 ASSERT(get_itbl(tso->block_info.closure)->type == BLACKHOLE_BQ);
2035 StgBlockingQueue *bq = (StgBlockingQueue *)(tso->block_info.closure);
2037 last = &bq->blocking_queue;
2038 for (t = bq->blocking_queue; t != END_TSO_QUEUE;
2039 last = &t->link, t = t->link) {
2045 barf("unblockThread (BLACKHOLE): TSO not found");
2048 case BlockedOnException:
2050 StgTSO *target = tso->block_info.tso;
2052 ASSERT(get_itbl(target)->type == TSO);
2053 ASSERT(target->blocked_exceptions != NULL);
2055 last = &target->blocked_exceptions;
2056 for (t = target->blocked_exceptions; t != END_TSO_QUEUE;
2057 last = &t->link, t = t->link) {
2058 ASSERT(get_itbl(t)->type == TSO);
2064 barf("unblockThread (Exception): TSO not found");
2067 case BlockedOnDelay:
2069 case BlockedOnWrite:
2071 StgTSO *prev = NULL;
2072 for (t = blocked_queue_hd; t != END_TSO_QUEUE;
2073 prev = t, t = t->link) {
2076 blocked_queue_hd = t->link;
2077 if (blocked_queue_tl == t) {
2078 blocked_queue_tl = END_TSO_QUEUE;
2081 prev->link = t->link;
2082 if (blocked_queue_tl == t) {
2083 blocked_queue_tl = prev;
2089 barf("unblockThread (I/O): TSO not found");
2093 barf("unblockThread");
2097 tso->link = END_TSO_QUEUE;
2098 tso->why_blocked = NotBlocked;
2099 tso->block_info.closure = NULL;
2100 PUSH_ON_RUN_QUEUE(tso);
2101 RELEASE_LOCK(&sched_mutex);
2104 /* -----------------------------------------------------------------------------
2107 * The following function implements the magic for raising an
2108 * asynchronous exception in an existing thread.
2110 * We first remove the thread from any queue on which it might be
2111 * blocked. The possible blockages are MVARs and BLACKHOLE_BQs.
2113 * We strip the stack down to the innermost CATCH_FRAME, building
2114 * thunks in the heap for all the active computations, so they can
2115 * be restarted if necessary. When we reach a CATCH_FRAME, we build
2116 * an application of the handler to the exception, and push it on
2117 * the top of the stack.
2119 * How exactly do we save all the active computations? We create an
2120 * AP_UPD for every UpdateFrame on the stack. Entering one of these
2121 * AP_UPDs pushes everything from the corresponding update frame
2122 * upwards onto the stack. (Actually, it pushes everything up to the
2123 * next update frame plus a pointer to the next AP_UPD object.
2124 * Entering the next AP_UPD object pushes more onto the stack until we
2125 * reach the last AP_UPD object - at which point the stack should look
2126 * exactly as it did when we killed the TSO and we can continue
2127 * execution by entering the closure on top of the stack.
2129 * We can also kill a thread entirely - this happens if either (a) the
2130 * exception passed to raiseAsync is NULL, or (b) there's no
2131 * CATCH_FRAME on the stack. In either case, we strip the entire
2132 * stack and replace the thread with a zombie.
2134 * -------------------------------------------------------------------------- */
2137 deleteThread(StgTSO *tso)
2139 raiseAsync(tso,NULL);
2143 raiseAsync(StgTSO *tso, StgClosure *exception)
2145 StgUpdateFrame* su = tso->su;
2146 StgPtr sp = tso->sp;
2148 /* Thread already dead? */
2149 if (tso->whatNext == ThreadComplete || tso->whatNext == ThreadKilled) {
2153 IF_DEBUG(scheduler, sched_belch("raising exception in thread %ld.", tso->id));
2155 /* Remove it from any blocking queues */
2158 /* The stack freezing code assumes there's a closure pointer on
2159 * the top of the stack. This isn't always the case with compiled
2160 * code, so we have to push a dummy closure on the top which just
2161 * returns to the next return address on the stack.
2163 if ( LOOKS_LIKE_GHC_INFO((void*)*sp) ) {
2164 *(--sp) = (W_)&dummy_ret_closure;
2168 int words = ((P_)su - (P_)sp) - 1;
2172 /* If we find a CATCH_FRAME, and we've got an exception to raise,
2173 * then build PAP(handler,exception,realworld#), and leave it on
2174 * top of the stack ready to enter.
2176 if (get_itbl(su)->type == CATCH_FRAME && exception != NULL) {
2177 StgCatchFrame *cf = (StgCatchFrame *)su;
2178 /* we've got an exception to raise, so let's pass it to the
2179 * handler in this frame.
2181 ap = (StgAP_UPD *)allocate(sizeofW(StgPAP) + 2);
2182 TICK_ALLOC_UPD_PAP(3,0);
2183 SET_HDR(ap,&PAP_info,cf->header.prof.ccs);
2186 ap->fun = cf->handler; /* :: Exception -> IO a */
2187 ap->payload[0] = (P_)exception;
2188 ap->payload[1] = ARG_TAG(0); /* realworld token */
2190 /* throw away the stack from Sp up to and including the
2193 sp = (P_)su + sizeofW(StgCatchFrame) - 1;
2196 /* Restore the blocked/unblocked state for asynchronous exceptions
2197 * at the CATCH_FRAME.
2199 * If exceptions were unblocked at the catch, arrange that they
2200 * are unblocked again after executing the handler by pushing an
2201 * unblockAsyncExceptions_ret stack frame.
2203 if (!cf->exceptions_blocked) {
2204 *(sp--) = (W_)&unblockAsyncExceptionszh_ret_info;
2207 /* Ensure that async exceptions are blocked when running the handler.
2209 if (tso->blocked_exceptions == NULL) {
2210 tso->blocked_exceptions = END_TSO_QUEUE;
2213 /* Put the newly-built PAP on top of the stack, ready to execute
2214 * when the thread restarts.
2218 tso->whatNext = ThreadEnterGHC;
2222 /* First build an AP_UPD consisting of the stack chunk above the
2223 * current update frame, with the top word on the stack as the
2226 ap = (StgAP_UPD *)allocate(AP_sizeW(words));
2231 ap->fun = (StgClosure *)sp[0];
2233 for(i=0; i < (nat)words; ++i) {
2234 ap->payload[i] = (P_)*sp++;
2237 switch (get_itbl(su)->type) {
2241 SET_HDR(ap,&AP_UPD_info,su->header.prof.ccs /* ToDo */);
2242 TICK_ALLOC_UP_THK(words+1,0);
2245 fprintf(stderr, "scheduler: Updating ");
2246 printPtr((P_)su->updatee);
2247 fprintf(stderr, " with ");
2248 printObj((StgClosure *)ap);
2251 /* Replace the updatee with an indirection - happily
2252 * this will also wake up any threads currently
2253 * waiting on the result.
2255 UPD_IND_NOLOCK(su->updatee,ap); /* revert the black hole */
2257 sp += sizeofW(StgUpdateFrame) -1;
2258 sp[0] = (W_)ap; /* push onto stack */
2264 StgCatchFrame *cf = (StgCatchFrame *)su;
2267 /* We want a PAP, not an AP_UPD. Fortunately, the
2268 * layout's the same.
2270 SET_HDR(ap,&PAP_info,su->header.prof.ccs /* ToDo */);
2271 TICK_ALLOC_UPD_PAP(words+1,0);
2273 /* now build o = FUN(catch,ap,handler) */
2274 o = (StgClosure *)allocate(sizeofW(StgClosure)+2);
2275 TICK_ALLOC_FUN(2,0);
2276 SET_HDR(o,&catch_info,su->header.prof.ccs /* ToDo */);
2277 o->payload[0] = (StgClosure *)ap;
2278 o->payload[1] = cf->handler;
2281 fprintf(stderr, "scheduler: Built ");
2282 printObj((StgClosure *)o);
2285 /* pop the old handler and put o on the stack */
2287 sp += sizeofW(StgCatchFrame) - 1;
2294 StgSeqFrame *sf = (StgSeqFrame *)su;
2297 SET_HDR(ap,&PAP_info,su->header.prof.ccs /* ToDo */);
2298 TICK_ALLOC_UPD_PAP(words+1,0);
2300 /* now build o = FUN(seq,ap) */
2301 o = (StgClosure *)allocate(sizeofW(StgClosure)+1);
2302 TICK_ALLOC_SE_THK(1,0);
2303 SET_HDR(o,&seq_info,su->header.prof.ccs /* ToDo */);
2304 payloadCPtr(o,0) = (StgClosure *)ap;
2307 fprintf(stderr, "scheduler: Built ");
2308 printObj((StgClosure *)o);
2311 /* pop the old handler and put o on the stack */
2313 sp += sizeofW(StgSeqFrame) - 1;
2319 /* We've stripped the entire stack, the thread is now dead. */
2320 sp += sizeofW(StgStopFrame) - 1;
2321 sp[0] = (W_)exception; /* save the exception */
2322 tso->whatNext = ThreadKilled;
2323 tso->su = (StgUpdateFrame *)(sp+1);
2334 //@node Debugging Routines, Index, Exception Handling Routines, Main scheduling code
2335 //@subsection Debugging Routines
2337 /* -----------------------------------------------------------------------------
2338 Debugging: why is a thread blocked
2339 -------------------------------------------------------------------------- */
2343 void printThreadBlockage(StgTSO *tso)
2345 switch (tso->why_blocked) {
2347 fprintf(stderr,"blocked on read from fd %d", tso->block_info.fd);
2349 case BlockedOnWrite:
2350 fprintf(stderr,"blocked on write to fd %d", tso->block_info.fd);
2352 case BlockedOnDelay:
2353 fprintf(stderr,"blocked on delay of %d ms", tso->block_info.delay);
2356 fprintf(stderr,"blocked on an MVar");
2358 case BlockedOnException:
2359 fprintf(stderr,"blocked on delivering an exception to thread %d",
2360 tso->block_info.tso->id);
2362 case BlockedOnBlackHole:
2363 fprintf(stderr,"blocked on a black hole");
2366 fprintf(stderr,"not blocked");
2370 fprintf(stderr,"blocked on global address");
2377 Print a whole blocking queue attached to node (debugging only).
2382 print_bq (StgClosure *node)
2384 StgBlockingQueueElement *bqe;
2388 fprintf(stderr,"## BQ of closure %p (%s): ",
2389 node, info_type(node));
2391 /* should cover all closures that may have a blocking queue */
2392 ASSERT(get_itbl(node)->type == BLACKHOLE_BQ ||
2393 get_itbl(node)->type == FETCH_ME_BQ ||
2394 get_itbl(node)->type == RBH);
2396 ASSERT(node!=(StgClosure*)NULL); // sanity check
2398 NB: In a parallel setup a BQ of an RBH must end with an RBH_Save closure;
2400 for (bqe = ((StgBlockingQueue*)node)->blocking_queue, end = (bqe==END_BQ_QUEUE);
2401 !end; // iterate until bqe points to a CONSTR
2402 end = (get_itbl(bqe)->type == CONSTR) || (bqe->link==END_BQ_QUEUE), bqe = end ? END_BQ_QUEUE : bqe->link) {
2403 ASSERT(bqe != END_BQ_QUEUE); // sanity check
2404 ASSERT(bqe != (StgTSO*)NULL); // sanity check
2405 /* types of closures that may appear in a blocking queue */
2406 ASSERT(get_itbl(bqe)->type == TSO ||
2407 get_itbl(bqe)->type == BLOCKED_FETCH ||
2408 get_itbl(bqe)->type == CONSTR);
2409 /* only BQs of an RBH end with an RBH_Save closure */
2410 ASSERT(get_itbl(bqe)->type != CONSTR || get_itbl(node)->type == RBH);
2412 switch (get_itbl(bqe)->type) {
2414 fprintf(stderr," TSO %d (%x),",
2415 ((StgTSO *)bqe)->id, ((StgTSO *)bqe));
2418 fprintf(stderr," BF (node=%p, ga=((%x, %d, %x)),",
2419 ((StgBlockedFetch *)bqe)->node,
2420 ((StgBlockedFetch *)bqe)->ga.payload.gc.gtid,
2421 ((StgBlockedFetch *)bqe)->ga.payload.gc.slot,
2422 ((StgBlockedFetch *)bqe)->ga.weight);
2425 fprintf(stderr," %s (IP %p),",
2426 (get_itbl(bqe) == &RBH_Save_0_info ? "RBH_Save_0" :
2427 get_itbl(bqe) == &RBH_Save_1_info ? "RBH_Save_1" :
2428 get_itbl(bqe) == &RBH_Save_2_info ? "RBH_Save_2" :
2429 "RBH_Save_?"), get_itbl(bqe));
2432 barf("Unexpected closure type %s in blocking queue of %p (%s)",
2433 info_type(bqe), node, info_type(node));
2437 fputc('\n', stderr);
2439 # elif defined(GRAN)
2441 print_bq (StgClosure *node)
2443 StgBlockingQueueElement *bqe;
2445 PEs node_loc, tso_loc;
2448 /* should cover all closures that may have a blocking queue */
2449 ASSERT(get_itbl(node)->type == BLACKHOLE_BQ ||
2450 get_itbl(node)->type == FETCH_ME_BQ ||
2451 get_itbl(node)->type == RBH);
2453 ASSERT(node!=(StgClosure*)NULL); // sanity check
2454 node_loc = where_is(node);
2456 fprintf(stderr,"## BQ of closure %p (%s) on [PE %d]: ",
2457 node, info_type(node), node_loc);
2460 NB: In a parallel setup a BQ of an RBH must end with an RBH_Save closure;
2462 for (bqe = ((StgBlockingQueue*)node)->blocking_queue, end = (bqe==END_BQ_QUEUE);
2463 !end; // iterate until bqe points to a CONSTR
2464 end = (get_itbl(bqe)->type == CONSTR) || (bqe->link==END_BQ_QUEUE), bqe = end ? END_BQ_QUEUE : bqe->link) {
2465 ASSERT(bqe != END_BQ_QUEUE); // sanity check
2466 ASSERT(bqe != (StgTSO*)NULL); // sanity check
2467 /* types of closures that may appear in a blocking queue */
2468 ASSERT(get_itbl(bqe)->type == TSO ||
2469 get_itbl(bqe)->type == CONSTR);
2470 /* only BQs of an RBH end with an RBH_Save closure */
2471 ASSERT(get_itbl(bqe)->type != CONSTR || get_itbl(node)->type == RBH);
2473 tso_loc = where_is((StgClosure *)bqe);
2474 switch (get_itbl(bqe)->type) {
2476 fprintf(stderr," TSO %d (%x) on [PE %d],",
2477 ((StgTSO *)bqe)->id, ((StgTSO *)bqe), tso_loc);
2480 fprintf(stderr," %s (IP %p),",
2481 (get_itbl(bqe) == &RBH_Save_0_info ? "RBH_Save_0" :
2482 get_itbl(bqe) == &RBH_Save_1_info ? "RBH_Save_1" :
2483 get_itbl(bqe) == &RBH_Save_2_info ? "RBH_Save_2" :
2484 "RBH_Save_?"), get_itbl(bqe));
2487 barf("Unexpected closure type %s in blocking queue of %p (%s)",
2488 info_type(bqe), node, info_type(node));
2492 fputc('\n', stderr);
2496 Nice and easy: only TSOs on the blocking queue
2499 print_bq (StgClosure *node)
2503 ASSERT(node!=(StgClosure*)NULL); // sanity check
2504 for (tso = ((StgBlockingQueue*)node)->blocking_queue;
2505 tso != END_TSO_QUEUE;
2507 ASSERT(tso!=NULL && tso!=END_TSO_QUEUE); // sanity check
2508 ASSERT(get_itbl(tso)->type == TSO); // guess what, sanity check
2509 fprintf(stderr," TSO %d (%p),", tso->id, tso);
2511 fputc('\n', stderr);
2522 for (i=0, tso=run_queue_hd;
2523 tso != END_TSO_QUEUE;
2532 sched_belch(char *s, ...)
2537 fprintf(stderr, "scheduler (task %ld): ", pthread_self());
2539 fprintf(stderr, "scheduler: ");
2541 vfprintf(stderr, s, ap);
2542 fprintf(stderr, "\n");
2547 //@node Index, , Debugging Routines, Main scheduling code
2551 //* MainRegTable:: @cindex\s-+MainRegTable
2552 //* StgMainThread:: @cindex\s-+StgMainThread
2553 //* awaken_blocked_queue:: @cindex\s-+awaken_blocked_queue
2554 //* blocked_queue_hd:: @cindex\s-+blocked_queue_hd
2555 //* blocked_queue_tl:: @cindex\s-+blocked_queue_tl
2556 //* context_switch:: @cindex\s-+context_switch
2557 //* createThread:: @cindex\s-+createThread
2558 //* free_capabilities:: @cindex\s-+free_capabilities
2559 //* gc_pending_cond:: @cindex\s-+gc_pending_cond
2560 //* initScheduler:: @cindex\s-+initScheduler
2561 //* interrupted:: @cindex\s-+interrupted
2562 //* n_free_capabilities:: @cindex\s-+n_free_capabilities
2563 //* next_thread_id:: @cindex\s-+next_thread_id
2564 //* print_bq:: @cindex\s-+print_bq
2565 //* run_queue_hd:: @cindex\s-+run_queue_hd
2566 //* run_queue_tl:: @cindex\s-+run_queue_tl
2567 //* sched_mutex:: @cindex\s-+sched_mutex
2568 //* schedule:: @cindex\s-+schedule
2569 //* take_off_run_queue:: @cindex\s-+take_off_run_queue
2570 //* task_ids:: @cindex\s-+task_ids
2571 //* term_mutex:: @cindex\s-+term_mutex
2572 //* thread_ready_cond:: @cindex\s-+thread_ready_cond