1 /* -----------------------------------------------------------------------------
3 * (c) The GHC Team 1998-2006
5 * Generational garbage collector
7 * Documentation on the architecture of the Garbage Collector can be
8 * found in the online commentary:
10 * http://hackage.haskell.org/trac/ghc/wiki/Commentary/Rts/Storage/GC
12 * ---------------------------------------------------------------------------*/
14 #include "PosixSource.h"
19 #include "OSThreads.h"
20 #include "LdvProfile.h"
25 #include "BlockAlloc.h"
31 #include "ParTicky.h" // ToDo: move into Rts.h
32 #include "RtsSignals.h"
36 #if defined(RTS_GTK_FRONTPANEL)
37 #include "FrontPanel.h"
40 #include "RetainerProfile.h"
41 #include "RaiseAsync.h"
53 #include <string.h> // for memset()
55 /* -----------------------------------------------------------------------------
57 -------------------------------------------------------------------------- */
59 /* STATIC OBJECT LIST.
62 * We maintain a linked list of static objects that are still live.
63 * The requirements for this list are:
65 * - we need to scan the list while adding to it, in order to
66 * scavenge all the static objects (in the same way that
67 * breadth-first scavenging works for dynamic objects).
69 * - we need to be able to tell whether an object is already on
70 * the list, to break loops.
72 * Each static object has a "static link field", which we use for
73 * linking objects on to the list. We use a stack-type list, consing
74 * objects on the front as they are added (this means that the
75 * scavenge phase is depth-first, not breadth-first, but that
78 * A separate list is kept for objects that have been scavenged
79 * already - this is so that we can zero all the marks afterwards.
81 * An object is on the list if its static link field is non-zero; this
82 * means that we have to mark the end of the list with '1', not NULL.
84 * Extra notes for generational GC:
86 * Each generation has a static object list associated with it. When
87 * collecting generations up to N, we treat the static object lists
88 * from generations > N as roots.
90 * We build up a static object list while collecting generations 0..N,
91 * which is then appended to the static object list of generation N+1.
94 /* N is the oldest generation being collected, where the generations
95 * are numbered starting at 0. A major GC (indicated by the major_gc
96 * flag) is when we're collecting all generations. We only attempt to
97 * deal with static objects and GC CAFs when doing a major GC.
102 /* Data used for allocation area sizing.
104 static lnat g0s0_pcnt_kept = 30; // percentage of g0s0 live at last minor GC
114 /* Thread-local data for each GC thread
116 gc_thread **gc_threads = NULL;
117 // gc_thread *gct = NULL; // this thread's gct TODO: make thread-local
119 // Number of threads running in *this* GC. Affects how many
120 // step->todos[] lists we have to look in to find work.
124 long copied; // *words* copied & scavenged during this GC
127 SpinLock recordMutableGen_sync;
130 /* -----------------------------------------------------------------------------
131 Static function declarations
132 -------------------------------------------------------------------------- */
134 static void mark_root (StgClosure **root);
135 static void zero_static_object_list (StgClosure* first_static);
136 static void initialise_N (rtsBool force_major_gc);
137 static void alloc_gc_threads (void);
138 static void init_collected_gen (nat g, nat threads);
139 static void init_uncollected_gen (nat g, nat threads);
140 static void init_gc_thread (gc_thread *t);
141 static void update_task_list (void);
142 static void resize_generations (void);
143 static void resize_nursery (void);
144 static void start_gc_threads (void);
145 static void gc_thread_work (void);
146 static nat inc_running (void);
147 static nat dec_running (void);
148 static void wakeup_gc_threads (nat n_threads);
150 #if 0 && defined(DEBUG)
151 static void gcCAFs (void);
154 /* -----------------------------------------------------------------------------
155 The mark bitmap & stack.
156 -------------------------------------------------------------------------- */
158 #define MARK_STACK_BLOCKS 4
160 bdescr *mark_stack_bdescr;
165 // Flag and pointers used for falling back to a linear scan when the
166 // mark stack overflows.
167 rtsBool mark_stack_overflowed;
168 bdescr *oldgen_scan_bd;
171 /* -----------------------------------------------------------------------------
172 GarbageCollect: the main entry point to the garbage collector.
174 Locks held: all capabilities are held throughout GarbageCollect().
175 -------------------------------------------------------------------------- */
178 GarbageCollect ( rtsBool force_major_gc )
182 lnat live, allocated;
183 lnat oldgen_saved_blocks = 0;
184 gc_thread *saved_gct;
187 // necessary if we stole a callee-saves register for gct:
191 CostCentreStack *prev_CCS;
196 debugTrace(DEBUG_gc, "starting GC");
198 #if defined(RTS_USER_SIGNALS)
199 if (RtsFlags.MiscFlags.install_signal_handlers) {
205 // tell the stats department that we've started a GC
208 // tell the STM to discard any cached closures it's hoping to re-use
217 // attribute any costs to CCS_GC
223 /* Approximate how much we allocated.
224 * Todo: only when generating stats?
226 allocated = calcAllocated();
228 /* Figure out which generation to collect
230 initialise_N(force_major_gc);
232 /* Allocate + initialise the gc_thread structures.
236 /* Start threads, so they can be spinning up while we finish initialisation.
240 /* How many threads will be participating in this GC?
241 * We don't try to parallelise minor GC.
243 #if defined(THREADED_RTS)
247 n_gc_threads = RtsFlags.ParFlags.gcThreads;
253 #ifdef RTS_GTK_FRONTPANEL
254 if (RtsFlags.GcFlags.frontpanel) {
255 updateFrontPanelBeforeGC(N);
260 // check for memory leaks if DEBUG is on
261 memInventory(traceClass(DEBUG_gc));
264 // check stack sanity *before* GC (ToDo: check all threads)
265 IF_DEBUG(sanity, checkFreeListSanity());
267 // Initialise all the generations/steps that we're collecting.
268 for (g = 0; g <= N; g++) {
269 init_collected_gen(g,n_gc_threads);
272 // Initialise all the generations/steps that we're *not* collecting.
273 for (g = N+1; g < RtsFlags.GcFlags.generations; g++) {
274 init_uncollected_gen(g,n_gc_threads);
277 /* Allocate a mark stack if we're doing a major collection.
280 mark_stack_bdescr = allocGroup(MARK_STACK_BLOCKS);
281 mark_stack = (StgPtr *)mark_stack_bdescr->start;
282 mark_sp = mark_stack;
283 mark_splim = mark_stack + (MARK_STACK_BLOCKS * BLOCK_SIZE_W);
285 mark_stack_bdescr = NULL;
288 // Initialise all our gc_thread structures
289 for (t = 0; t < n_gc_threads; t++) {
290 init_gc_thread(gc_threads[t]);
293 // the main thread is running: this prevents any other threads from
294 // exiting prematurely, so we can start them now.
296 wakeup_gc_threads(n_gc_threads);
298 // this is the main thread
301 /* -----------------------------------------------------------------------
302 * follow all the roots that we know about:
303 * - mutable lists from each generation > N
304 * we want to *scavenge* these roots, not evacuate them: they're not
305 * going to move in this GC.
306 * Also do them in reverse generation order, for the usual reason:
307 * namely to reduce the likelihood of spurious old->new pointers.
310 for (g = RtsFlags.GcFlags.generations-1; g > N; g--) {
311 generations[g].saved_mut_list = generations[g].mut_list;
312 generations[g].mut_list = allocBlock();
313 // mut_list always has at least one block.
315 for (g = RtsFlags.GcFlags.generations-1; g > N; g--) {
316 scavenge_mutable_list(&generations[g]);
320 // follow roots from the CAF list (used by GHCi)
324 // follow all the roots that the application knows about.
328 #if defined(RTS_USER_SIGNALS)
329 // mark the signal handlers (signals should be already blocked)
330 markSignalHandlers(mark_root);
333 // Mark the weak pointer list, and prepare to detect dead weak pointers.
337 // Mark the stable pointer table.
338 markStablePtrTable(mark_root);
340 /* -------------------------------------------------------------------------
341 * Repeatedly scavenge all the areas we know about until there's no
342 * more scavenging to be done.
347 // The other threads are now stopped. We might recurse back to
348 // here, but from now on this is the only thread.
350 // if any blackholes are alive, make the threads that wait on
352 if (traverseBlackholeQueue()) {
357 // must be last... invariant is that everything is fully
358 // scavenged at this point.
359 if (traverseWeakPtrList()) { // returns rtsTrue if evaced something
364 // If we get to here, there's really nothing left to do.
368 // Update pointers from the Task list
371 // Now see which stable names are still alive.
375 // We call processHeapClosureForDead() on every closure destroyed during
376 // the current garbage collection, so we invoke LdvCensusForDead().
377 if (RtsFlags.ProfFlags.doHeapProfile == HEAP_BY_LDV
378 || RtsFlags.ProfFlags.bioSelector != NULL)
382 // NO MORE EVACUATION AFTER THIS POINT!
383 // Finally: compaction of the oldest generation.
384 if (major_gc && oldest_gen->steps[0].is_compacted) {
385 // save number of blocks for stats
386 oldgen_saved_blocks = oldest_gen->steps[0].n_old_blocks;
390 IF_DEBUG(sanity, checkGlobalTSOList(rtsFalse));
392 // Two-space collector: free the old to-space.
393 // g0s0->old_blocks is the old nursery
394 // g0s0->blocks is to-space from the previous GC
395 if (RtsFlags.GcFlags.generations == 1) {
396 if (g0s0->blocks != NULL) {
397 freeChain(g0s0->blocks);
402 // For each workspace, in each thread:
403 // * clear the BF_EVACUATED flag from each copied block
404 // * move the copied blocks to the step
410 for (t = 0; t < n_gc_threads; t++) {
414 for (s = 1; s < total_steps; s++) {
417 // ASSERT( ws->scan_bd == ws->todo_bd );
418 ASSERT( ws->scan_bd ? ws->scan == ws->scan_bd->free : 1 );
420 // Push the final block
421 if (ws->scan_bd) { push_scan_block(ws->scan_bd, ws); }
423 ASSERT(countBlocks(ws->scavd_list) == ws->n_scavd_blocks);
425 prev = ws->scavd_list;
426 for (bd = ws->scavd_list; bd != NULL; bd = bd->link) {
427 bd->flags &= ~BF_EVACUATED; // now from-space
430 prev->link = ws->stp->blocks;
431 ws->stp->blocks = ws->scavd_list;
432 ws->stp->n_blocks += ws->n_scavd_blocks;
433 ASSERT(countBlocks(ws->stp->blocks) == ws->stp->n_blocks);
438 // Two-space collector: swap the semi-spaces around.
439 // Currently: g0s0->old_blocks is the old nursery
440 // g0s0->blocks is to-space from this GC
441 // We want these the other way around.
442 if (RtsFlags.GcFlags.generations == 1) {
443 bdescr *nursery_blocks = g0s0->old_blocks;
444 nat n_nursery_blocks = g0s0->n_old_blocks;
445 g0s0->old_blocks = g0s0->blocks;
446 g0s0->n_old_blocks = g0s0->n_blocks;
447 g0s0->blocks = nursery_blocks;
448 g0s0->n_blocks = n_nursery_blocks;
451 /* run through all the generations/steps and tidy up
456 for (i=0; i < n_gc_threads; i++) {
458 trace(TRACE_gc,"thread %d:", i);
459 trace(TRACE_gc," copied %ld", gc_threads[i]->copied * sizeof(W_));
460 trace(TRACE_gc," any_work %ld", gc_threads[i]->any_work);
461 trace(TRACE_gc," no_work %ld", gc_threads[i]->no_work);
462 trace(TRACE_gc," scav_global_work %ld", gc_threads[i]->scav_global_work);
463 trace(TRACE_gc," scav_local_work %ld", gc_threads[i]->scav_local_work);
465 copied += gc_threads[i]->copied;
469 for (g = 0; g < RtsFlags.GcFlags.generations; g++) {
472 generations[g].collections++; // for stats
475 // Count the mutable list as bytes "copied" for the purposes of
476 // stats. Every mutable list is copied during every GC.
478 nat mut_list_size = 0;
479 for (bd = generations[g].mut_list; bd != NULL; bd = bd->link) {
480 mut_list_size += bd->free - bd->start;
482 copied += mut_list_size;
485 "mut_list_size: %lu (%d vars, %d arrays, %d MVARs, %d others)",
486 (unsigned long)(mut_list_size * sizeof(W_)),
487 mutlist_MUTVARS, mutlist_MUTARRS, mutlist_MVARS, mutlist_OTHERS);
490 for (s = 0; s < generations[g].n_steps; s++) {
492 stp = &generations[g].steps[s];
494 // for generations we collected...
497 /* free old memory and shift to-space into from-space for all
498 * the collected steps (except the allocation area). These
499 * freed blocks will probaby be quickly recycled.
501 if (!(g == 0 && s == 0 && RtsFlags.GcFlags.generations > 1)) {
502 if (stp->is_compacted)
504 // for a compacted step, just shift the new to-space
505 // onto the front of the now-compacted existing blocks.
506 for (bd = stp->blocks; bd != NULL; bd = bd->link) {
507 bd->flags &= ~BF_EVACUATED; // now from-space
509 // tack the new blocks on the end of the existing blocks
510 if (stp->old_blocks != NULL) {
511 for (bd = stp->old_blocks; bd != NULL; bd = next) {
512 // NB. this step might not be compacted next
513 // time, so reset the BF_COMPACTED flags.
514 // They are set before GC if we're going to
515 // compact. (search for BF_COMPACTED above).
516 bd->flags &= ~BF_COMPACTED;
519 bd->link = stp->blocks;
522 stp->blocks = stp->old_blocks;
524 // add the new blocks to the block tally
525 stp->n_blocks += stp->n_old_blocks;
526 ASSERT(countBlocks(stp->blocks) == stp->n_blocks);
530 freeChain(stp->old_blocks);
532 stp->old_blocks = NULL;
533 stp->n_old_blocks = 0;
536 /* LARGE OBJECTS. The current live large objects are chained on
537 * scavenged_large, having been moved during garbage
538 * collection from large_objects. Any objects left on
539 * large_objects list are therefore dead, so we free them here.
541 for (bd = stp->large_objects; bd != NULL; bd = next) {
547 // update the count of blocks used by large objects
548 for (bd = stp->scavenged_large_objects; bd != NULL; bd = bd->link) {
549 bd->flags &= ~BF_EVACUATED;
551 stp->large_objects = stp->scavenged_large_objects;
552 stp->n_large_blocks = stp->n_scavenged_large_blocks;
555 else // for older generations...
557 /* For older generations, we need to append the
558 * scavenged_large_object list (i.e. large objects that have been
559 * promoted during this GC) to the large_object list for that step.
561 for (bd = stp->scavenged_large_objects; bd; bd = next) {
563 bd->flags &= ~BF_EVACUATED;
564 dbl_link_onto(bd, &stp->large_objects);
567 // add the new blocks we promoted during this GC
568 stp->n_large_blocks += stp->n_scavenged_large_blocks;
573 // update the max size of older generations after a major GC
574 resize_generations();
576 // Guess the amount of live data for stats.
577 live = calcLiveBlocks() * BLOCK_SIZE_W;
578 debugTrace(DEBUG_gc, "Slop: %ldKB",
579 (live - calcLiveWords()) / (1024/sizeof(W_)));
581 // Free the small objects allocated via allocate(), since this will
582 // all have been copied into G0S1 now.
583 if (RtsFlags.GcFlags.generations > 1) {
584 if (g0s0->blocks != NULL) {
585 freeChain(g0s0->blocks);
591 alloc_blocks_lim = RtsFlags.GcFlags.minAllocAreaSize;
593 // Start a new pinned_object_block
594 pinned_object_block = NULL;
596 // Free the mark stack.
597 if (mark_stack_bdescr != NULL) {
598 freeGroup(mark_stack_bdescr);
602 for (g = 0; g <= N; g++) {
603 for (s = 0; s < generations[g].n_steps; s++) {
604 stp = &generations[g].steps[s];
605 if (stp->bitmap != NULL) {
606 freeGroup(stp->bitmap);
614 // mark the garbage collected CAFs as dead
615 #if 0 && defined(DEBUG) // doesn't work at the moment
616 if (major_gc) { gcCAFs(); }
620 // resetStaticObjectForRetainerProfiling() must be called before
622 if (n_gc_threads > 1) {
623 barf("profiling is currently broken with multi-threaded GC");
624 // ToDo: fix the gct->scavenged_static_objects below
626 resetStaticObjectForRetainerProfiling(gct->scavenged_static_objects);
629 // zero the scavenged static object list
632 for (i = 0; i < n_gc_threads; i++) {
633 zero_static_object_list(gc_threads[i]->scavenged_static_objects);
640 // start any pending finalizers
642 scheduleFinalizers(last_free_capability, old_weak_ptr_list);
645 // send exceptions to any threads which were about to die
647 resurrectThreads(resurrected_threads);
650 // Update the stable pointer hash table.
651 updateStablePtrTable(major_gc);
653 // check sanity after GC
654 IF_DEBUG(sanity, checkSanity());
656 // extra GC trace info
657 if (traceClass(TRACE_gc)) statDescribeGens();
660 // symbol-table based profiling
661 /* heapCensus(to_blocks); */ /* ToDo */
664 // restore enclosing cost centre
670 // check for memory leaks if DEBUG is on
671 memInventory(traceClass(DEBUG_gc));
674 #ifdef RTS_GTK_FRONTPANEL
675 if (RtsFlags.GcFlags.frontpanel) {
676 updateFrontPanelAfterGC( N, live );
680 // ok, GC over: tell the stats department what happened.
681 stat_endGC(allocated, live, copied, N);
683 #if defined(RTS_USER_SIGNALS)
684 if (RtsFlags.MiscFlags.install_signal_handlers) {
685 // unblock signals again
686 unblockUserSignals();
695 /* -----------------------------------------------------------------------------
696 * Mark all nodes pointed to by sparks in the spark queues (for GC) Does an
697 * implicit slide i.e. after marking all sparks are at the beginning of the
698 * spark pool and the spark pool only contains sparkable closures
699 * -------------------------------------------------------------------------- */
703 markSparkQueue (evac_fn evac, Capability *cap)
705 StgClosure **sparkp, **to_sparkp;
706 nat n, pruned_sparks; // stats only
709 PAR_TICKY_MARK_SPARK_QUEUE_START();
714 pool = &(cap->r.rSparks);
716 ASSERT_SPARK_POOL_INVARIANTS(pool);
718 #if defined(PARALLEL_HASKELL)
725 to_sparkp = pool->hd;
726 while (sparkp != pool->tl) {
727 ASSERT(*sparkp!=NULL);
728 ASSERT(LOOKS_LIKE_CLOSURE_PTR(((StgClosure *)*sparkp)));
729 // ToDo?: statistics gathering here (also for GUM!)
730 if (closure_SHOULD_SPARK(*sparkp)) {
732 *to_sparkp++ = *sparkp;
733 if (to_sparkp == pool->lim) {
734 to_sparkp = pool->base;
741 if (sparkp == pool->lim) {
745 pool->tl = to_sparkp;
747 PAR_TICKY_MARK_SPARK_QUEUE_END(n);
749 #if defined(PARALLEL_HASKELL)
750 debugTrace(DEBUG_sched,
751 "marked %d sparks and pruned %d sparks on [%x]",
752 n, pruned_sparks, mytid);
754 debugTrace(DEBUG_sched,
755 "marked %d sparks and pruned %d sparks",
759 debugTrace(DEBUG_sched,
760 "new spark queue len=%d; (hd=%p; tl=%p)\n",
761 sparkPoolSize(pool), pool->hd, pool->tl);
765 /* ---------------------------------------------------------------------------
766 Where are the roots that we know about?
768 - all the threads on the runnable queue
769 - all the threads on the blocked queue
770 - all the threads on the sleeping queue
771 - all the thread currently executing a _ccall_GC
772 - all the "main threads"
774 ------------------------------------------------------------------------ */
777 GetRoots( evac_fn evac )
783 // Each GC thread is responsible for following roots from the
784 // Capability of the same number. There will usually be the same
785 // or fewer Capabilities as GC threads, but just in case there
786 // are more, we mark every Capability whose number is the GC
787 // thread's index plus a multiple of the number of GC threads.
788 for (i = gct->thread_index; i < n_capabilities; i += n_gc_threads) {
789 cap = &capabilities[i];
790 evac((StgClosure **)(void *)&cap->run_queue_hd);
791 evac((StgClosure **)(void *)&cap->run_queue_tl);
792 #if defined(THREADED_RTS)
793 evac((StgClosure **)(void *)&cap->wakeup_queue_hd);
794 evac((StgClosure **)(void *)&cap->wakeup_queue_tl);
796 for (task = cap->suspended_ccalling_tasks; task != NULL;
798 debugTrace(DEBUG_sched,
799 "evac'ing suspended TSO %lu", (unsigned long)task->suspended_tso->id);
800 evac((StgClosure **)(void *)&task->suspended_tso);
803 #if defined(THREADED_RTS)
804 markSparkQueue(evac,cap);
808 #if !defined(THREADED_RTS)
809 evac((StgClosure **)(void *)&blocked_queue_hd);
810 evac((StgClosure **)(void *)&blocked_queue_tl);
811 evac((StgClosure **)(void *)&sleeping_queue);
815 /* -----------------------------------------------------------------------------
816 isAlive determines whether the given closure is still alive (after
817 a garbage collection) or not. It returns the new address of the
818 closure if it is alive, or NULL otherwise.
820 NOTE: Use it before compaction only!
821 It untags and (if needed) retags pointers to closures.
822 -------------------------------------------------------------------------- */
826 isAlive(StgClosure *p)
828 const StgInfoTable *info;
834 /* The tag and the pointer are split, to be merged later when needed. */
835 tag = GET_CLOSURE_TAG(p);
836 q = UNTAG_CLOSURE(p);
838 ASSERT(LOOKS_LIKE_CLOSURE_PTR(q));
841 // ignore static closures
843 // ToDo: for static closures, check the static link field.
844 // Problem here is that we sometimes don't set the link field, eg.
845 // for static closures with an empty SRT or CONSTR_STATIC_NOCAFs.
847 if (!HEAP_ALLOCED(q)) {
851 // ignore closures in generations that we're not collecting.
853 if (bd->gen_no > N) {
857 // if it's a pointer into to-space, then we're done
858 if (bd->flags & BF_EVACUATED) {
862 // large objects use the evacuated flag
863 if (bd->flags & BF_LARGE) {
867 // check the mark bit for compacted steps
868 if ((bd->flags & BF_COMPACTED) && is_marked((P_)q,bd)) {
872 switch (info->type) {
877 case IND_OLDGEN: // rely on compatible layout with StgInd
878 case IND_OLDGEN_PERM:
879 // follow indirections
880 p = ((StgInd *)q)->indirectee;
885 return ((StgEvacuated *)q)->evacuee;
888 if (((StgTSO *)q)->what_next == ThreadRelocated) {
889 p = (StgClosure *)((StgTSO *)q)->link;
901 /* -----------------------------------------------------------------------------
902 Figure out which generation to collect, initialise N and major_gc.
903 -------------------------------------------------------------------------- */
906 initialise_N (rtsBool force_major_gc)
910 if (force_major_gc) {
911 N = RtsFlags.GcFlags.generations - 1;
915 for (g = 0; g < RtsFlags.GcFlags.generations; g++) {
916 if (generations[g].steps[0].n_blocks +
917 generations[g].steps[0].n_large_blocks
918 >= generations[g].max_blocks) {
922 major_gc = (N == RtsFlags.GcFlags.generations-1);
926 /* -----------------------------------------------------------------------------
927 Initialise the gc_thread structures.
928 -------------------------------------------------------------------------- */
931 alloc_gc_thread (int n)
937 t = stgMallocBytes(sizeof(gc_thread) + total_steps * sizeof(step_workspace),
942 initCondition(&t->wake_cond);
943 initMutex(&t->wake_mutex);
944 t->wakeup = rtsFalse;
949 t->free_blocks = NULL;
958 for (s = 0; s < total_steps; s++)
961 ws->stp = &all_steps[s];
962 ASSERT(s == ws->stp->abs_no);
969 ws->buffer_todo_bd = NULL;
971 ws->scavd_list = NULL;
972 ws->n_scavd_blocks = 0;
980 alloc_gc_threads (void)
982 if (gc_threads == NULL) {
983 #if defined(THREADED_RTS)
985 gc_threads = stgMallocBytes (RtsFlags.ParFlags.gcThreads *
989 for (i = 0; i < RtsFlags.ParFlags.gcThreads; i++) {
990 gc_threads[i] = alloc_gc_thread(i);
993 gc_threads = stgMallocBytes (sizeof(gc_thread*),
996 gc_threads[0] = alloc_gc_thread(0);
1001 /* ----------------------------------------------------------------------------
1003 ------------------------------------------------------------------------- */
1005 static nat gc_running_threads;
1007 #if defined(THREADED_RTS)
1008 static Mutex gc_running_mutex;
1015 ACQUIRE_LOCK(&gc_running_mutex);
1016 n_running = ++gc_running_threads;
1017 RELEASE_LOCK(&gc_running_mutex);
1025 ACQUIRE_LOCK(&gc_running_mutex);
1026 n_running = --gc_running_threads;
1027 RELEASE_LOCK(&gc_running_mutex);
1032 // gc_thread_work(): Scavenge until there's no work left to do and all
1033 // the running threads are idle.
1036 gc_thread_work (void)
1040 debugTrace(DEBUG_gc, "GC thread %d working", gct->thread_index);
1042 // gc_running_threads has already been incremented for us; either
1043 // this is the main thread and we incremented it inside
1044 // GarbageCollect(), or this is a worker thread and the main
1045 // thread bumped gc_running_threads before waking us up.
1047 // Every thread evacuates some roots.
1049 GetRoots(mark_root);
1053 // scavenge_loop() only exits when there's no work to do
1056 debugTrace(DEBUG_gc, "GC thread %d idle (%d still running)",
1057 gct->thread_index, r);
1059 while (gc_running_threads != 0) {
1064 // any_work() does not remove the work from the queue, it
1065 // just checks for the presence of work. If we find any,
1066 // then we increment gc_running_threads and go back to
1067 // scavenge_loop() to perform any pending work.
1070 // All threads are now stopped
1071 debugTrace(DEBUG_gc, "GC thread %d finished.", gct->thread_index);
1075 #if defined(THREADED_RTS)
1077 gc_thread_mainloop (void)
1079 while (!gct->exit) {
1081 // Wait until we're told to wake up
1082 ACQUIRE_LOCK(&gct->wake_mutex);
1083 while (!gct->wakeup) {
1084 debugTrace(DEBUG_gc, "GC thread %d standing by...",
1086 waitCondition(&gct->wake_cond, &gct->wake_mutex);
1088 RELEASE_LOCK(&gct->wake_mutex);
1089 gct->wakeup = rtsFalse;
1090 if (gct->exit) break;
1093 // start performance counters in this thread...
1094 if (gct->papi_events == -1) {
1095 papi_init_eventset(&gct->papi_events);
1097 papi_thread_start_gc1_count(gct->papi_events);
1103 // count events in this thread towards the GC totals
1104 papi_thread_stop_gc1_count(gct->papi_events);
1110 #if defined(THREADED_RTS)
1112 gc_thread_entry (gc_thread *my_gct)
1115 debugTrace(DEBUG_gc, "GC thread %d starting...", gct->thread_index);
1116 gct->id = osThreadId();
1117 gc_thread_mainloop();
1122 start_gc_threads (void)
1124 #if defined(THREADED_RTS)
1127 static rtsBool done = rtsFalse;
1129 gc_running_threads = 0;
1130 initMutex(&gc_running_mutex);
1133 // Start from 1: the main thread is 0
1134 for (i = 1; i < RtsFlags.ParFlags.gcThreads; i++) {
1135 createOSThread(&id, (OSThreadProc*)&gc_thread_entry,
1144 wakeup_gc_threads (nat n_threads USED_IF_THREADS)
1146 #if defined(THREADED_RTS)
1148 for (i=1; i < n_threads; i++) {
1150 ACQUIRE_LOCK(&gc_threads[i]->wake_mutex);
1151 gc_threads[i]->wakeup = rtsTrue;
1152 signalCondition(&gc_threads[i]->wake_cond);
1153 RELEASE_LOCK(&gc_threads[i]->wake_mutex);
1158 /* ----------------------------------------------------------------------------
1159 Initialise a generation that is to be collected
1160 ------------------------------------------------------------------------- */
1163 init_collected_gen (nat g, nat n_threads)
1170 // Throw away the current mutable list. Invariant: the mutable
1171 // list always has at least one block; this means we can avoid a
1172 // check for NULL in recordMutable().
1174 freeChain(generations[g].mut_list);
1175 generations[g].mut_list = allocBlock();
1176 for (i = 0; i < n_capabilities; i++) {
1177 freeChain(capabilities[i].mut_lists[g]);
1178 capabilities[i].mut_lists[g] = allocBlock();
1182 for (s = 0; s < generations[g].n_steps; s++) {
1184 // generation 0, step 0 doesn't need to-space
1185 if (g == 0 && s == 0 && RtsFlags.GcFlags.generations > 1) {
1189 stp = &generations[g].steps[s];
1190 ASSERT(stp->gen_no == g);
1192 // deprecate the existing blocks
1193 stp->old_blocks = stp->blocks;
1194 stp->n_old_blocks = stp->n_blocks;
1198 // we don't have any to-be-scavenged blocks yet
1202 // initialise the large object queues.
1203 stp->scavenged_large_objects = NULL;
1204 stp->n_scavenged_large_blocks = 0;
1206 // mark the large objects as not evacuated yet
1207 for (bd = stp->large_objects; bd; bd = bd->link) {
1208 bd->flags &= ~BF_EVACUATED;
1211 // for a compacted step, we need to allocate the bitmap
1212 if (stp->is_compacted) {
1213 nat bitmap_size; // in bytes
1214 bdescr *bitmap_bdescr;
1217 bitmap_size = stp->n_old_blocks * BLOCK_SIZE / (sizeof(W_)*BITS_PER_BYTE);
1219 if (bitmap_size > 0) {
1220 bitmap_bdescr = allocGroup((lnat)BLOCK_ROUND_UP(bitmap_size)
1222 stp->bitmap = bitmap_bdescr;
1223 bitmap = bitmap_bdescr->start;
1225 debugTrace(DEBUG_gc, "bitmap_size: %d, bitmap: %p",
1226 bitmap_size, bitmap);
1228 // don't forget to fill it with zeros!
1229 memset(bitmap, 0, bitmap_size);
1231 // For each block in this step, point to its bitmap from the
1232 // block descriptor.
1233 for (bd=stp->old_blocks; bd != NULL; bd = bd->link) {
1234 bd->u.bitmap = bitmap;
1235 bitmap += BLOCK_SIZE_W / (sizeof(W_)*BITS_PER_BYTE);
1237 // Also at this point we set the BF_COMPACTED flag
1238 // for this block. The invariant is that
1239 // BF_COMPACTED is always unset, except during GC
1240 // when it is set on those blocks which will be
1242 bd->flags |= BF_COMPACTED;
1248 // For each GC thread, for each step, allocate a "todo" block to
1249 // store evacuated objects to be scavenged, and a block to store
1250 // evacuated objects that do not need to be scavenged.
1251 for (t = 0; t < n_threads; t++) {
1252 for (s = 0; s < generations[g].n_steps; s++) {
1254 // we don't copy objects into g0s0, unless -G0
1255 if (g==0 && s==0 && RtsFlags.GcFlags.generations > 1) continue;
1257 ws = &gc_threads[t]->steps[g * RtsFlags.GcFlags.steps + s];
1262 ws->todo_large_objects = NULL;
1264 // allocate the first to-space block; extra blocks will be
1265 // chained on as necessary.
1267 ws->buffer_todo_bd = NULL;
1268 gc_alloc_todo_block(ws);
1270 ws->scavd_list = NULL;
1271 ws->n_scavd_blocks = 0;
1277 /* ----------------------------------------------------------------------------
1278 Initialise a generation that is *not* to be collected
1279 ------------------------------------------------------------------------- */
1282 init_uncollected_gen (nat g, nat threads)
1289 for (s = 0; s < generations[g].n_steps; s++) {
1290 stp = &generations[g].steps[s];
1291 stp->scavenged_large_objects = NULL;
1292 stp->n_scavenged_large_blocks = 0;
1295 for (t = 0; t < threads; t++) {
1296 for (s = 0; s < generations[g].n_steps; s++) {
1298 ws = &gc_threads[t]->steps[g * RtsFlags.GcFlags.steps + s];
1301 ws->buffer_todo_bd = NULL;
1302 ws->todo_large_objects = NULL;
1304 ws->scavd_list = NULL;
1305 ws->n_scavd_blocks = 0;
1307 // If the block at the head of the list in this generation
1308 // is less than 3/4 full, then use it as a todo block.
1309 if (stp->blocks && isPartiallyFull(stp->blocks))
1311 ws->todo_bd = stp->blocks;
1312 ws->todo_free = ws->todo_bd->free;
1313 ws->todo_lim = ws->todo_bd->start + BLOCK_SIZE_W;
1314 stp->blocks = stp->blocks->link;
1316 ws->todo_bd->link = NULL;
1318 // this block is also the scan block; we must scan
1319 // from the current end point.
1320 ws->scan_bd = ws->todo_bd;
1321 ws->scan = ws->scan_bd->free;
1323 // subtract the contents of this block from the stats,
1324 // because we'll count the whole block later.
1325 copied -= ws->scan_bd->free - ws->scan_bd->start;
1332 gc_alloc_todo_block(ws);
1337 // Move the private mutable lists from each capability onto the
1338 // main mutable list for the generation.
1339 for (i = 0; i < n_capabilities; i++) {
1340 for (bd = capabilities[i].mut_lists[g];
1341 bd->link != NULL; bd = bd->link) {
1344 bd->link = generations[g].mut_list;
1345 generations[g].mut_list = capabilities[i].mut_lists[g];
1346 capabilities[i].mut_lists[g] = allocBlock();
1350 /* -----------------------------------------------------------------------------
1351 Initialise a gc_thread before GC
1352 -------------------------------------------------------------------------- */
1355 init_gc_thread (gc_thread *t)
1357 t->static_objects = END_OF_STATIC_LIST;
1358 t->scavenged_static_objects = END_OF_STATIC_LIST;
1360 t->failed_to_evac = rtsFalse;
1361 t->eager_promotion = rtsTrue;
1362 t->thunk_selector_depth = 0;
1366 t->scav_global_work = 0;
1367 t->scav_local_work = 0;
1371 /* -----------------------------------------------------------------------------
1372 Function we pass to GetRoots to evacuate roots.
1373 -------------------------------------------------------------------------- */
1376 mark_root(StgClosure **root)
1381 /* -----------------------------------------------------------------------------
1382 Initialising the static object & mutable lists
1383 -------------------------------------------------------------------------- */
1386 zero_static_object_list(StgClosure* first_static)
1390 const StgInfoTable *info;
1392 for (p = first_static; p != END_OF_STATIC_LIST; p = link) {
1394 link = *STATIC_LINK(info, p);
1395 *STATIC_LINK(info,p) = NULL;
1399 /* -----------------------------------------------------------------------------
1401 -------------------------------------------------------------------------- */
1408 for (c = (StgIndStatic *)revertible_caf_list; c != NULL;
1409 c = (StgIndStatic *)c->static_link)
1411 SET_INFO(c, c->saved_info);
1412 c->saved_info = NULL;
1413 // could, but not necessary: c->static_link = NULL;
1415 revertible_caf_list = NULL;
1419 markCAFs( evac_fn evac )
1423 for (c = (StgIndStatic *)caf_list; c != NULL;
1424 c = (StgIndStatic *)c->static_link)
1426 evac(&c->indirectee);
1428 for (c = (StgIndStatic *)revertible_caf_list; c != NULL;
1429 c = (StgIndStatic *)c->static_link)
1431 evac(&c->indirectee);
1435 /* ----------------------------------------------------------------------------
1436 Update the pointers from the task list
1438 These are treated as weak pointers because we want to allow a main
1439 thread to get a BlockedOnDeadMVar exception in the same way as any
1440 other thread. Note that the threads should all have been retained
1441 by GC by virtue of being on the all_threads list, we're just
1442 updating pointers here.
1443 ------------------------------------------------------------------------- */
1446 update_task_list (void)
1450 for (task = all_tasks; task != NULL; task = task->all_link) {
1451 if (!task->stopped && task->tso) {
1452 ASSERT(task->tso->bound == task);
1453 tso = (StgTSO *) isAlive((StgClosure *)task->tso);
1455 barf("task %p: main thread %d has been GC'd",
1468 /* ----------------------------------------------------------------------------
1469 Reset the sizes of the older generations when we do a major
1472 CURRENT STRATEGY: make all generations except zero the same size.
1473 We have to stay within the maximum heap size, and leave a certain
1474 percentage of the maximum heap size available to allocate into.
1475 ------------------------------------------------------------------------- */
1478 resize_generations (void)
1482 if (major_gc && RtsFlags.GcFlags.generations > 1) {
1483 nat live, size, min_alloc;
1484 nat max = RtsFlags.GcFlags.maxHeapSize;
1485 nat gens = RtsFlags.GcFlags.generations;
1487 // live in the oldest generations
1488 live = oldest_gen->steps[0].n_blocks +
1489 oldest_gen->steps[0].n_large_blocks;
1491 // default max size for all generations except zero
1492 size = stg_max(live * RtsFlags.GcFlags.oldGenFactor,
1493 RtsFlags.GcFlags.minOldGenSize);
1495 // minimum size for generation zero
1496 min_alloc = stg_max((RtsFlags.GcFlags.pcFreeHeap * max) / 200,
1497 RtsFlags.GcFlags.minAllocAreaSize);
1499 // Auto-enable compaction when the residency reaches a
1500 // certain percentage of the maximum heap size (default: 30%).
1501 if (RtsFlags.GcFlags.generations > 1 &&
1502 (RtsFlags.GcFlags.compact ||
1504 oldest_gen->steps[0].n_blocks >
1505 (RtsFlags.GcFlags.compactThreshold * max) / 100))) {
1506 oldest_gen->steps[0].is_compacted = 1;
1507 // debugBelch("compaction: on\n", live);
1509 oldest_gen->steps[0].is_compacted = 0;
1510 // debugBelch("compaction: off\n", live);
1513 // if we're going to go over the maximum heap size, reduce the
1514 // size of the generations accordingly. The calculation is
1515 // different if compaction is turned on, because we don't need
1516 // to double the space required to collect the old generation.
1519 // this test is necessary to ensure that the calculations
1520 // below don't have any negative results - we're working
1521 // with unsigned values here.
1522 if (max < min_alloc) {
1526 if (oldest_gen->steps[0].is_compacted) {
1527 if ( (size + (size - 1) * (gens - 2) * 2) + min_alloc > max ) {
1528 size = (max - min_alloc) / ((gens - 1) * 2 - 1);
1531 if ( (size * (gens - 1) * 2) + min_alloc > max ) {
1532 size = (max - min_alloc) / ((gens - 1) * 2);
1542 debugBelch("live: %d, min_alloc: %d, size : %d, max = %d\n", live,
1543 min_alloc, size, max);
1546 for (g = 0; g < gens; g++) {
1547 generations[g].max_blocks = size;
1552 /* -----------------------------------------------------------------------------
1553 Calculate the new size of the nursery, and resize it.
1554 -------------------------------------------------------------------------- */
1557 resize_nursery (void)
1559 if (RtsFlags.GcFlags.generations == 1)
1560 { // Two-space collector:
1563 /* set up a new nursery. Allocate a nursery size based on a
1564 * function of the amount of live data (by default a factor of 2)
1565 * Use the blocks from the old nursery if possible, freeing up any
1568 * If we get near the maximum heap size, then adjust our nursery
1569 * size accordingly. If the nursery is the same size as the live
1570 * data (L), then we need 3L bytes. We can reduce the size of the
1571 * nursery to bring the required memory down near 2L bytes.
1573 * A normal 2-space collector would need 4L bytes to give the same
1574 * performance we get from 3L bytes, reducing to the same
1575 * performance at 2L bytes.
1577 blocks = g0s0->n_old_blocks;
1579 if ( RtsFlags.GcFlags.maxHeapSize != 0 &&
1580 blocks * RtsFlags.GcFlags.oldGenFactor * 2 >
1581 RtsFlags.GcFlags.maxHeapSize )
1583 long adjusted_blocks; // signed on purpose
1586 adjusted_blocks = (RtsFlags.GcFlags.maxHeapSize - 2 * blocks);
1588 debugTrace(DEBUG_gc, "near maximum heap size of 0x%x blocks, blocks = %d, adjusted to %ld",
1589 RtsFlags.GcFlags.maxHeapSize, blocks, adjusted_blocks);
1591 pc_free = adjusted_blocks * 100 / RtsFlags.GcFlags.maxHeapSize;
1592 if (pc_free < RtsFlags.GcFlags.pcFreeHeap) /* might even * be < 0 */
1596 blocks = adjusted_blocks;
1600 blocks *= RtsFlags.GcFlags.oldGenFactor;
1601 if (blocks < RtsFlags.GcFlags.minAllocAreaSize)
1603 blocks = RtsFlags.GcFlags.minAllocAreaSize;
1606 resizeNurseries(blocks);
1608 else // Generational collector
1611 * If the user has given us a suggested heap size, adjust our
1612 * allocation area to make best use of the memory available.
1614 if (RtsFlags.GcFlags.heapSizeSuggestion)
1617 nat needed = calcNeeded(); // approx blocks needed at next GC
1619 /* Guess how much will be live in generation 0 step 0 next time.
1620 * A good approximation is obtained by finding the
1621 * percentage of g0s0 that was live at the last minor GC.
1623 * We have an accurate figure for the amount of copied data in
1624 * 'copied', but we must convert this to a number of blocks, with
1625 * a small adjustment for estimated slop at the end of a block
1630 g0s0_pcnt_kept = ((copied / (BLOCK_SIZE_W - 10)) * 100)
1631 / countNurseryBlocks();
1634 /* Estimate a size for the allocation area based on the
1635 * information available. We might end up going slightly under
1636 * or over the suggested heap size, but we should be pretty
1639 * Formula: suggested - needed
1640 * ----------------------------
1641 * 1 + g0s0_pcnt_kept/100
1643 * where 'needed' is the amount of memory needed at the next
1644 * collection for collecting all steps except g0s0.
1647 (((long)RtsFlags.GcFlags.heapSizeSuggestion - (long)needed) * 100) /
1648 (100 + (long)g0s0_pcnt_kept);
1650 if (blocks < (long)RtsFlags.GcFlags.minAllocAreaSize) {
1651 blocks = RtsFlags.GcFlags.minAllocAreaSize;
1654 resizeNurseries((nat)blocks);
1658 // we might have added extra large blocks to the nursery, so
1659 // resize back to minAllocAreaSize again.
1660 resizeNurseriesFixed(RtsFlags.GcFlags.minAllocAreaSize);
1665 /* -----------------------------------------------------------------------------
1666 Sanity code for CAF garbage collection.
1668 With DEBUG turned on, we manage a CAF list in addition to the SRT
1669 mechanism. After GC, we run down the CAF list and blackhole any
1670 CAFs which have been garbage collected. This means we get an error
1671 whenever the program tries to enter a garbage collected CAF.
1673 Any garbage collected CAFs are taken off the CAF list at the same
1675 -------------------------------------------------------------------------- */
1677 #if 0 && defined(DEBUG)
1684 const StgInfoTable *info;
1695 ASSERT(info->type == IND_STATIC);
1697 if (STATIC_LINK(info,p) == NULL) {
1698 debugTrace(DEBUG_gccafs, "CAF gc'd at 0x%04lx", (long)p);
1700 SET_INFO(p,&stg_BLACKHOLE_info);
1701 p = STATIC_LINK2(info,p);
1705 pp = &STATIC_LINK2(info,p);
1712 debugTrace(DEBUG_gccafs, "%d CAFs live", i);