1 /* -----------------------------------------------------------------------------
3 * (c) The GHC Team 1998-2008
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"
25 #include "BlockAlloc.h"
29 #include "RtsSignals.h"
31 #if defined(RTS_GTK_FRONTPANEL)
32 #include "FrontPanel.h"
35 #include "RetainerProfile.h"
36 #include "LdvProfile.h"
37 #include "RaiseAsync.h"
47 #include "MarkStack.h"
52 #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 g0_pcnt_kept = 30; // percentage of g0 live at last minor GC
114 /* Thread-local data for each GC thread
116 gc_thread **gc_threads = NULL;
118 #if !defined(THREADED_RTS)
119 StgWord8 the_gc_thread[sizeof(gc_thread) + 64 * sizeof(gen_workspace)];
122 // Number of threads running in *this* GC. Affects how many
123 // step->todos[] lists we have to look in to find work.
127 long copied; // *words* copied & scavenged during this GC
129 rtsBool work_stealing;
133 /* -----------------------------------------------------------------------------
134 Static function declarations
135 -------------------------------------------------------------------------- */
137 static void mark_root (void *user, StgClosure **root);
138 static void zero_static_object_list (StgClosure* first_static);
139 static nat initialise_N (rtsBool force_major_gc);
140 static void prepare_collected_gen (generation *gen);
141 static void prepare_uncollected_gen (generation *gen);
142 static void init_gc_thread (gc_thread *t);
143 static void resize_generations (void);
144 static void resize_nursery (void);
145 static void start_gc_threads (void);
146 static void scavenge_until_all_done (void);
147 static StgWord inc_running (void);
148 static StgWord dec_running (void);
149 static void wakeup_gc_threads (nat n_threads, nat me);
150 static void shutdown_gc_threads (nat n_threads, nat me);
151 static void collect_gct_blocks (void);
153 #if 0 && defined(DEBUG)
154 static void gcCAFs (void);
157 /* -----------------------------------------------------------------------------
159 -------------------------------------------------------------------------- */
161 bdescr *mark_stack_top_bd; // topmost block in the mark stack
162 bdescr *mark_stack_bd; // current block in the mark stack
163 StgPtr mark_sp; // pointer to the next unallocated mark stack entry
165 /* -----------------------------------------------------------------------------
166 GarbageCollect: the main entry point to the garbage collector.
168 Locks held: all capabilities are held throughout GarbageCollect().
169 -------------------------------------------------------------------------- */
172 GarbageCollect (rtsBool force_major_gc,
173 nat gc_type USED_IF_THREADS,
178 lnat live_blocks, live_words, allocated, max_copied, avg_copied;
179 gc_thread *saved_gct;
182 // necessary if we stole a callee-saves register for gct:
186 CostCentreStack *prev_CCS;
191 #if defined(RTS_USER_SIGNALS)
192 if (RtsFlags.MiscFlags.install_signal_handlers) {
198 ASSERT(sizeof(gen_workspace) == 16 * sizeof(StgWord));
199 // otherwise adjust the padding in gen_workspace.
201 // tell the stats department that we've started a GC
204 // tell the STM to discard any cached closures it's hoping to re-use
207 // lock the StablePtr table
216 // attribute any costs to CCS_GC
222 /* Approximate how much we allocated.
223 * Todo: only when generating stats?
225 allocated = calcAllocated(rtsFalse/* don't count the nursery yet */);
227 /* Figure out which generation to collect
229 n = initialise_N(force_major_gc);
231 #if defined(THREADED_RTS)
232 work_stealing = RtsFlags.ParFlags.parGcLoadBalancingEnabled &&
233 N >= RtsFlags.ParFlags.parGcLoadBalancingGen;
234 // It's not always a good idea to do load balancing in parallel
235 // GC. In particular, for a parallel program we don't want to
236 // lose locality by moving cached data into another CPU's cache
237 // (this effect can be quite significant).
239 // We could have a more complex way to deterimine whether to do
240 // work stealing or not, e.g. it might be a good idea to do it
241 // if the heap is big. For now, we just turn it on or off with
245 /* Start threads, so they can be spinning up while we finish initialisation.
249 #if defined(THREADED_RTS)
250 /* How many threads will be participating in this GC?
251 * We don't try to parallelise minor GCs (unless the user asks for
252 * it with +RTS -gn0), or mark/compact/sweep GC.
254 if (gc_type == PENDING_GC_PAR) {
255 n_gc_threads = RtsFlags.ParFlags.nNodes;
263 debugTrace(DEBUG_gc, "GC (gen %d): %d KB to collect, %ld MB in use, using %d thread(s)",
264 N, n * (BLOCK_SIZE / 1024), mblocks_allocated, n_gc_threads);
266 #ifdef RTS_GTK_FRONTPANEL
267 if (RtsFlags.GcFlags.frontpanel) {
268 updateFrontPanelBeforeGC(N);
273 // check for memory leaks if DEBUG is on
274 memInventory(DEBUG_gc);
277 // check sanity *before* GC
278 IF_DEBUG(sanity, checkSanity(rtsFalse /* before GC */, major_gc));
280 // Initialise all our gc_thread structures
281 for (t = 0; t < n_gc_threads; t++) {
282 init_gc_thread(gc_threads[t]);
285 // Initialise all the generations/steps that we're collecting.
286 for (g = 0; g <= N; g++) {
287 prepare_collected_gen(&generations[g]);
289 // Initialise all the generations/steps that we're *not* collecting.
290 for (g = N+1; g < RtsFlags.GcFlags.generations; g++) {
291 prepare_uncollected_gen(&generations[g]);
294 /* Allocate a mark stack if we're doing a major collection.
296 if (major_gc && oldest_gen->mark) {
297 mark_stack_bd = allocBlock();
298 mark_stack_top_bd = mark_stack_bd;
299 mark_stack_bd->link = NULL;
300 mark_stack_bd->u.back = NULL;
301 mark_sp = mark_stack_bd->start;
303 mark_stack_bd = NULL;
304 mark_stack_top_bd = NULL;
308 // this is the main thread
310 if (n_gc_threads == 1) {
311 SET_GCT(gc_threads[0]);
313 SET_GCT(gc_threads[cap->no]);
316 SET_GCT(gc_threads[0]);
319 /* -----------------------------------------------------------------------
320 * follow all the roots that we know about:
323 // the main thread is running: this prevents any other threads from
324 // exiting prematurely, so we can start them now.
325 // NB. do this after the mutable lists have been saved above, otherwise
326 // the other GC threads will be writing into the old mutable lists.
328 wakeup_gc_threads(n_gc_threads, gct->thread_index);
330 // scavenge the capability-private mutable lists. This isn't part
331 // of markSomeCapabilities() because markSomeCapabilities() can only
332 // call back into the GC via mark_root() (due to the gct register
334 if (n_gc_threads == 1) {
335 for (n = 0; n < n_capabilities; n++) {
336 #if defined(THREADED_RTS)
337 scavenge_capability_mut_Lists1(&capabilities[n]);
339 scavenge_capability_mut_lists(&capabilities[n]);
343 scavenge_capability_mut_lists(&capabilities[gct->thread_index]);
346 // follow roots from the CAF list (used by GHCi)
347 gct->evac_gen_no = 0;
348 markCAFs(mark_root, gct);
350 // follow all the roots that the application knows about.
351 gct->evac_gen_no = 0;
352 markSomeCapabilities(mark_root, gct, gct->thread_index, n_gc_threads,
353 rtsTrue/*prune sparks*/);
355 #if defined(RTS_USER_SIGNALS)
356 // mark the signal handlers (signals should be already blocked)
357 markSignalHandlers(mark_root, gct);
360 // Mark the weak pointer list, and prepare to detect dead weak pointers.
364 // Mark the stable pointer table.
365 markStablePtrTable(mark_root, gct);
367 /* -------------------------------------------------------------------------
368 * Repeatedly scavenge all the areas we know about until there's no
369 * more scavenging to be done.
373 scavenge_until_all_done();
374 // The other threads are now stopped. We might recurse back to
375 // here, but from now on this is the only thread.
377 // must be last... invariant is that everything is fully
378 // scavenged at this point.
379 if (traverseWeakPtrList()) { // returns rtsTrue if evaced something
384 // If we get to here, there's really nothing left to do.
388 shutdown_gc_threads(n_gc_threads, gct->thread_index);
390 // Now see which stable names are still alive.
394 if (n_gc_threads == 1) {
395 for (n = 0; n < n_capabilities; n++) {
396 pruneSparkQueue(&capabilities[n]);
399 pruneSparkQueue(&capabilities[gct->thread_index]);
404 // We call processHeapClosureForDead() on every closure destroyed during
405 // the current garbage collection, so we invoke LdvCensusForDead().
406 if (RtsFlags.ProfFlags.doHeapProfile == HEAP_BY_LDV
407 || RtsFlags.ProfFlags.bioSelector != NULL)
411 // NO MORE EVACUATION AFTER THIS POINT!
413 // Two-space collector: free the old to-space.
414 // g0->old_blocks is the old nursery
415 // g0->blocks is to-space from the previous GC
416 if (RtsFlags.GcFlags.generations == 1) {
417 if (g0->blocks != NULL) {
418 freeChain(g0->blocks);
423 // Finally: compact or sweep the oldest generation.
424 if (major_gc && oldest_gen->mark) {
425 if (oldest_gen->compact)
426 compact(gct->scavenged_static_objects);
436 for (i=0; i < n_gc_threads; i++) {
437 if (n_gc_threads > 1) {
438 debugTrace(DEBUG_gc,"thread %d:", i);
439 debugTrace(DEBUG_gc," copied %ld", gc_threads[i]->copied * sizeof(W_));
440 debugTrace(DEBUG_gc," scanned %ld", gc_threads[i]->scanned * sizeof(W_));
441 debugTrace(DEBUG_gc," any_work %ld", gc_threads[i]->any_work);
442 debugTrace(DEBUG_gc," no_work %ld", gc_threads[i]->no_work);
443 debugTrace(DEBUG_gc," scav_find_work %ld", gc_threads[i]->scav_find_work);
445 copied += gc_threads[i]->copied;
446 max_copied = stg_max(gc_threads[i]->copied, max_copied);
448 if (n_gc_threads == 1) {
456 // Run through all the generations/steps and tidy up.
458 // - count the amount of "live" data (live_words, live_blocks)
459 // - count the amount of "copied" data in this GC (copied)
461 // - make to-space the new from-space (set BF_EVACUATED on all blocks)
466 for (g = 0; g < RtsFlags.GcFlags.generations; g++) {
469 generations[g].collections++; // for stats
470 if (n_gc_threads > 1) generations[g].par_collections++;
473 // Count the mutable list as bytes "copied" for the purposes of
474 // stats. Every mutable list is copied during every GC.
476 nat mut_list_size = 0;
477 for (n = 0; n < n_capabilities; n++) {
478 mut_list_size += countOccupied(capabilities[n].mut_lists[g]);
480 copied += mut_list_size;
483 "mut_list_size: %lu (%d vars, %d arrays, %d MVARs, %d others)",
484 (unsigned long)(mut_list_size * sizeof(W_)),
485 mutlist_MUTVARS, mutlist_MUTARRS, mutlist_MVARS, mutlist_OTHERS);
489 gen = &generations[g];
491 // for generations we collected...
494 /* free old memory and shift to-space into from-space for all
495 * the collected steps (except the allocation area). These
496 * freed blocks will probaby be quickly recycled.
500 // tack the new blocks on the end of the existing blocks
501 if (gen->old_blocks != NULL) {
504 for (bd = gen->old_blocks; bd != NULL; bd = next) {
508 if (!(bd->flags & BF_MARKED))
511 gen->old_blocks = next;
520 gen->n_words += bd->free - bd->start;
522 // NB. this step might not be compacted next
523 // time, so reset the BF_MARKED flags.
524 // They are set before GC if we're going to
525 // compact. (search for BF_MARKED above).
526 bd->flags &= ~BF_MARKED;
528 // between GCs, all blocks in the heap except
529 // for the nursery have the BF_EVACUATED flag set.
530 bd->flags |= BF_EVACUATED;
537 prev->link = gen->blocks;
538 gen->blocks = gen->old_blocks;
541 // add the new blocks to the block tally
542 gen->n_blocks += gen->n_old_blocks;
543 ASSERT(countBlocks(gen->blocks) == gen->n_blocks);
544 ASSERT(countOccupied(gen->blocks) == gen->n_words);
548 freeChain(gen->old_blocks);
551 gen->old_blocks = NULL;
552 gen->n_old_blocks = 0;
554 /* LARGE OBJECTS. The current live large objects are chained on
555 * scavenged_large, having been moved during garbage
556 * collection from large_objects. Any objects left on the
557 * large_objects list are therefore dead, so we free them here.
559 freeChain(gen->large_objects);
560 gen->large_objects = gen->scavenged_large_objects;
561 gen->n_large_blocks = gen->n_scavenged_large_blocks;
562 gen->n_new_large_words = 0;
564 else // for generations > N
566 /* For older generations, we need to append the
567 * scavenged_large_object list (i.e. large objects that have been
568 * promoted during this GC) to the large_object list for that step.
570 for (bd = gen->scavenged_large_objects; bd; bd = next) {
572 dbl_link_onto(bd, &gen->large_objects);
575 // add the new blocks we promoted during this GC
576 gen->n_large_blocks += gen->n_scavenged_large_blocks;
579 ASSERT(countBlocks(gen->large_objects) == gen->n_large_blocks);
581 gen->scavenged_large_objects = NULL;
582 gen->n_scavenged_large_blocks = 0;
585 live_words += genLiveWords(gen);
586 live_blocks += genLiveBlocks(gen);
588 // add in the partial blocks in the gen_workspaces, but ignore gen 0
589 // if this is a local GC (we can't count another capability's part_list)
592 for (i = 0; i < n_capabilities; i++) {
593 live_words += gcThreadLiveWords(i, gen->no);
594 live_blocks += gcThreadLiveBlocks(i, gen->no);
597 } // for all generations
599 // update the max size of older generations after a major GC
600 resize_generations();
602 // Start a new pinned_object_block
603 for (n = 0; n < n_capabilities; n++) {
604 capabilities[n].pinned_object_block = NULL;
607 // Free the mark stack.
608 if (mark_stack_top_bd != NULL) {
609 debugTrace(DEBUG_gc, "mark stack: %d blocks",
610 countBlocks(mark_stack_top_bd));
611 freeChain(mark_stack_top_bd);
615 for (g = 0; g <= N; g++) {
616 gen = &generations[g];
617 if (gen->bitmap != NULL) {
618 freeGroup(gen->bitmap);
623 // Reset the nursery: make the blocks empty
624 allocated += clearNurseries();
630 // mark the garbage collected CAFs as dead
631 #if 0 && defined(DEBUG) // doesn't work at the moment
632 if (major_gc) { gcCAFs(); }
636 // resetStaticObjectForRetainerProfiling() must be called before
638 if (n_gc_threads > 1) {
639 barf("profiling is currently broken with multi-threaded GC");
640 // ToDo: fix the gct->scavenged_static_objects below
642 resetStaticObjectForRetainerProfiling(gct->scavenged_static_objects);
645 // zero the scavenged static object list
648 for (i = 0; i < n_gc_threads; i++) {
649 zero_static_object_list(gc_threads[i]->scavenged_static_objects);
653 // Update the stable pointer hash table.
654 updateStablePtrTable(major_gc);
656 // unlock the StablePtr table. Must be before scheduleFinalizers(),
657 // because a finalizer may call hs_free_fun_ptr() or
658 // hs_free_stable_ptr(), both of which access the StablePtr table.
661 // Start any pending finalizers. Must be after
662 // updateStablePtrTable() and stablePtrPostGC() (see #4221).
664 scheduleFinalizers(cap, old_weak_ptr_list);
667 // check sanity after GC
668 // before resurrectThreads(), because that might overwrite some
669 // closures, which will cause problems with THREADED where we don't
671 IF_DEBUG(sanity, checkSanity(rtsTrue /* after GC */, major_gc));
673 // send exceptions to any threads which were about to die
675 resurrectThreads(resurrected_threads);
680 need = BLOCKS_TO_MBLOCKS(n_alloc_blocks);
681 got = mblocks_allocated;
682 /* If the amount of data remains constant, next major GC we'll
683 require (F+1)*need. We leave (F+2)*need in order to reduce
684 repeated deallocation and reallocation. */
685 need = (RtsFlags.GcFlags.oldGenFactor + 2) * need;
687 returnMemoryToOS(got - need);
691 // extra GC trace info
692 IF_DEBUG(gc, statDescribeGens());
695 // symbol-table based profiling
696 /* heapCensus(to_blocks); */ /* ToDo */
699 // restore enclosing cost centre
705 // check for memory leaks if DEBUG is on
706 memInventory(DEBUG_gc);
709 #ifdef RTS_GTK_FRONTPANEL
710 if (RtsFlags.GcFlags.frontpanel) {
711 updateFrontPanelAfterGC( N, live );
715 // ok, GC over: tell the stats department what happened.
716 stat_endGC(allocated, live_words, copied, N, max_copied, avg_copied,
717 live_blocks * BLOCK_SIZE_W - live_words /* slop */);
719 // Guess which generation we'll collect *next* time
720 initialise_N(force_major_gc);
722 #if defined(RTS_USER_SIGNALS)
723 if (RtsFlags.MiscFlags.install_signal_handlers) {
724 // unblock signals again
725 unblockUserSignals();
734 /* -----------------------------------------------------------------------------
735 Figure out which generation to collect, initialise N and major_gc.
737 Also returns the total number of blocks in generations that will be
739 -------------------------------------------------------------------------- */
742 initialise_N (rtsBool force_major_gc)
745 nat blocks, blocks_total;
750 if (force_major_gc) {
751 N = RtsFlags.GcFlags.generations - 1;
756 for (g = RtsFlags.GcFlags.generations - 1; g >= 0; g--) {
758 blocks = generations[g].n_words / BLOCK_SIZE_W
759 + generations[g].n_large_blocks;
761 if (blocks >= generations[g].max_blocks) {
765 blocks_total += blocks;
769 blocks_total += countNurseryBlocks();
771 major_gc = (N == RtsFlags.GcFlags.generations-1);
775 /* -----------------------------------------------------------------------------
776 Initialise the gc_thread structures.
777 -------------------------------------------------------------------------- */
779 #define GC_THREAD_INACTIVE 0
780 #define GC_THREAD_STANDING_BY 1
781 #define GC_THREAD_RUNNING 2
782 #define GC_THREAD_WAITING_TO_CONTINUE 3
785 new_gc_thread (nat n, gc_thread *t)
792 initSpinLock(&t->gc_spin);
793 initSpinLock(&t->mut_spin);
794 ACQUIRE_SPIN_LOCK(&t->gc_spin);
795 t->wakeup = GC_THREAD_INACTIVE; // starts true, so we can wait for the
796 // thread to start up, see wakeup_gc_threads
800 t->free_blocks = NULL;
809 for (g = 0; g < RtsFlags.GcFlags.generations; g++)
812 ws->gen = &generations[g];
813 ASSERT(g == ws->gen->no);
817 // alloc_todo_block(ws,0);
818 // but can't, because it uses gct which isn't set up at this point.
819 // Hence, allocate a block for todo_bd manually:
821 bdescr *bd = allocBlock(); // no lock, locks aren't initialised yet
822 initBdescr(bd, ws->gen, ws->gen->to);
823 bd->flags = BF_EVACUATED;
824 bd->u.scan = bd->free = bd->start;
827 ws->todo_free = bd->free;
828 ws->todo_lim = bd->start + BLOCK_SIZE_W;
831 ws->todo_q = newWSDeque(128);
832 ws->todo_overflow = NULL;
833 ws->n_todo_overflow = 0;
835 ws->part_list = NULL;
836 ws->n_part_blocks = 0;
838 ws->scavd_list = NULL;
839 ws->n_scavd_blocks = 0;
847 if (gc_threads == NULL) {
848 #if defined(THREADED_RTS)
850 gc_threads = stgMallocBytes (RtsFlags.ParFlags.nNodes *
854 for (i = 0; i < RtsFlags.ParFlags.nNodes; i++) {
856 stgMallocBytes(sizeof(gc_thread) +
857 RtsFlags.GcFlags.generations * sizeof(gen_workspace),
860 new_gc_thread(i, gc_threads[i]);
863 gc_threads = stgMallocBytes (sizeof(gc_thread*),"alloc_gc_threads");
865 new_gc_thread(0,gc_threads[0]);
874 if (gc_threads != NULL) {
875 #if defined(THREADED_RTS)
877 for (i = 0; i < n_capabilities; i++) {
878 for (g = 0; g < RtsFlags.GcFlags.generations; g++)
880 freeWSDeque(gc_threads[i]->gens[g].todo_q);
882 stgFree (gc_threads[i]);
884 stgFree (gc_threads);
886 for (g = 0; g < RtsFlags.GcFlags.generations; g++)
888 freeWSDeque(gc_threads[0]->gens[g].todo_q);
890 stgFree (gc_threads);
896 /* ----------------------------------------------------------------------------
898 ------------------------------------------------------------------------- */
900 static volatile StgWord gc_running_threads;
906 new = atomic_inc(&gc_running_threads);
907 ASSERT(new <= n_gc_threads);
914 ASSERT(gc_running_threads != 0);
915 return atomic_dec(&gc_running_threads);
928 // scavenge objects in compacted generation
929 if (mark_stack_bd != NULL && !mark_stack_empty()) {
933 // Check for global work in any step. We don't need to check for
934 // local work, because we have already exited scavenge_loop(),
935 // which means there is no local work for this thread.
936 for (g = 0; g < (int)RtsFlags.GcFlags.generations; g++) {
938 if (ws->todo_large_objects) return rtsTrue;
939 if (!looksEmptyWSDeque(ws->todo_q)) return rtsTrue;
940 if (ws->todo_overflow) return rtsTrue;
943 #if defined(THREADED_RTS)
946 // look for work to steal
947 for (n = 0; n < n_gc_threads; n++) {
948 if (n == gct->thread_index) continue;
949 for (g = RtsFlags.GcFlags.generations-1; g >= 0; g--) {
950 ws = &gc_threads[n]->gens[g];
951 if (!looksEmptyWSDeque(ws->todo_q)) return rtsTrue;
958 #if defined(THREADED_RTS)
966 scavenge_until_all_done (void)
972 traceEventGcWork(&capabilities[gct->thread_index]);
974 #if defined(THREADED_RTS)
975 if (n_gc_threads > 1) {
984 collect_gct_blocks();
986 // scavenge_loop() only exits when there's no work to do
989 traceEventGcIdle(&capabilities[gct->thread_index]);
991 debugTrace(DEBUG_gc, "%d GC threads still running", r);
993 while (gc_running_threads != 0) {
999 // any_work() does not remove the work from the queue, it
1000 // just checks for the presence of work. If we find any,
1001 // then we increment gc_running_threads and go back to
1002 // scavenge_loop() to perform any pending work.
1005 traceEventGcDone(&capabilities[gct->thread_index]);
1008 #if defined(THREADED_RTS)
1011 gcWorkerThread (Capability *cap)
1013 gc_thread *saved_gct;
1015 // necessary if we stole a callee-saves register for gct:
1018 gct = gc_threads[cap->no];
1019 gct->id = osThreadId();
1021 // Wait until we're told to wake up
1022 RELEASE_SPIN_LOCK(&gct->mut_spin);
1023 gct->wakeup = GC_THREAD_STANDING_BY;
1024 debugTrace(DEBUG_gc, "GC thread %d standing by...", gct->thread_index);
1025 ACQUIRE_SPIN_LOCK(&gct->gc_spin);
1028 // start performance counters in this thread...
1029 if (gct->papi_events == -1) {
1030 papi_init_eventset(&gct->papi_events);
1032 papi_thread_start_gc1_count(gct->papi_events);
1035 // Every thread evacuates some roots.
1036 gct->evac_gen_no = 0;
1037 markSomeCapabilities(mark_root, gct, gct->thread_index, n_gc_threads,
1038 rtsTrue/*prune sparks*/);
1039 scavenge_capability_mut_lists(&capabilities[gct->thread_index]);
1041 scavenge_until_all_done();
1044 // Now that the whole heap is marked, we discard any sparks that
1045 // were found to be unreachable. The main GC thread is currently
1046 // marking heap reachable via weak pointers, so it is
1047 // non-deterministic whether a spark will be retained if it is
1048 // only reachable via weak pointers. To fix this problem would
1049 // require another GC barrier, which is too high a price.
1050 pruneSparkQueue(cap);
1054 // count events in this thread towards the GC totals
1055 papi_thread_stop_gc1_count(gct->papi_events);
1058 // Wait until we're told to continue
1059 RELEASE_SPIN_LOCK(&gct->gc_spin);
1060 gct->wakeup = GC_THREAD_WAITING_TO_CONTINUE;
1061 debugTrace(DEBUG_gc, "GC thread %d waiting to continue...",
1063 ACQUIRE_SPIN_LOCK(&gct->mut_spin);
1064 debugTrace(DEBUG_gc, "GC thread %d on my way...", gct->thread_index);
1071 #if defined(THREADED_RTS)
1074 waitForGcThreads (Capability *cap USED_IF_THREADS)
1076 const nat n_threads = RtsFlags.ParFlags.nNodes;
1077 const nat me = cap->no;
1079 rtsBool retry = rtsTrue;
1082 for (i=0; i < n_threads; i++) {
1083 if (i == me) continue;
1084 if (gc_threads[i]->wakeup != GC_THREAD_STANDING_BY) {
1085 prodCapability(&capabilities[i], cap->running_task);
1088 for (j=0; j < 10; j++) {
1090 for (i=0; i < n_threads; i++) {
1091 if (i == me) continue;
1093 setContextSwitches();
1094 if (gc_threads[i]->wakeup != GC_THREAD_STANDING_BY) {
1104 #endif // THREADED_RTS
1107 start_gc_threads (void)
1109 #if defined(THREADED_RTS)
1110 gc_running_threads = 0;
1115 wakeup_gc_threads (nat n_threads USED_IF_THREADS, nat me USED_IF_THREADS)
1117 #if defined(THREADED_RTS)
1119 for (i=0; i < n_threads; i++) {
1120 if (i == me) continue;
1122 debugTrace(DEBUG_gc, "waking up gc thread %d", i);
1123 if (gc_threads[i]->wakeup != GC_THREAD_STANDING_BY) barf("wakeup_gc_threads");
1125 gc_threads[i]->wakeup = GC_THREAD_RUNNING;
1126 ACQUIRE_SPIN_LOCK(&gc_threads[i]->mut_spin);
1127 RELEASE_SPIN_LOCK(&gc_threads[i]->gc_spin);
1132 // After GC is complete, we must wait for all GC threads to enter the
1133 // standby state, otherwise they may still be executing inside
1134 // any_work(), and may even remain awake until the next GC starts.
1136 shutdown_gc_threads (nat n_threads USED_IF_THREADS, nat me USED_IF_THREADS)
1138 #if defined(THREADED_RTS)
1140 for (i=0; i < n_threads; i++) {
1141 if (i == me) continue;
1142 while (gc_threads[i]->wakeup != GC_THREAD_WAITING_TO_CONTINUE) { write_barrier(); }
1147 #if defined(THREADED_RTS)
1149 releaseGCThreads (Capability *cap USED_IF_THREADS)
1151 const nat n_threads = RtsFlags.ParFlags.nNodes;
1152 const nat me = cap->no;
1154 for (i=0; i < n_threads; i++) {
1155 if (i == me) continue;
1156 if (gc_threads[i]->wakeup != GC_THREAD_WAITING_TO_CONTINUE)
1157 barf("releaseGCThreads");
1159 gc_threads[i]->wakeup = GC_THREAD_INACTIVE;
1160 ACQUIRE_SPIN_LOCK(&gc_threads[i]->gc_spin);
1161 RELEASE_SPIN_LOCK(&gc_threads[i]->mut_spin);
1166 /* ----------------------------------------------------------------------------
1167 Initialise a generation that is to be collected
1168 ------------------------------------------------------------------------- */
1171 prepare_collected_gen (generation *gen)
1177 // Throw away the current mutable list. Invariant: the mutable
1178 // list always has at least one block; this means we can avoid a
1179 // check for NULL in recordMutable().
1182 for (i = 0; i < n_capabilities; i++) {
1183 freeChain(capabilities[i].mut_lists[g]);
1184 capabilities[i].mut_lists[g] = allocBlock();
1188 gen = &generations[g];
1189 ASSERT(gen->no == g);
1191 // we'll construct a new list of threads in this step
1192 // during GC, throw away the current list.
1193 gen->old_threads = gen->threads;
1194 gen->threads = END_TSO_QUEUE;
1196 // deprecate the existing blocks
1197 gen->old_blocks = gen->blocks;
1198 gen->n_old_blocks = gen->n_blocks;
1202 gen->live_estimate = 0;
1204 // initialise the large object queues.
1205 ASSERT(gen->scavenged_large_objects == NULL);
1206 ASSERT(gen->n_scavenged_large_blocks == 0);
1208 // grab all the partial blocks stashed in the gc_thread workspaces and
1209 // move them to the old_blocks list of this gen.
1210 for (n = 0; n < n_capabilities; n++) {
1211 ws = &gc_threads[n]->gens[gen->no];
1213 for (bd = ws->part_list; bd != NULL; bd = next) {
1215 bd->link = gen->old_blocks;
1216 gen->old_blocks = bd;
1217 gen->n_old_blocks += bd->blocks;
1219 ws->part_list = NULL;
1220 ws->n_part_blocks = 0;
1222 ASSERT(ws->scavd_list == NULL);
1223 ASSERT(ws->n_scavd_blocks == 0);
1225 if (ws->todo_free != ws->todo_bd->start) {
1226 ws->todo_bd->free = ws->todo_free;
1227 ws->todo_bd->link = gen->old_blocks;
1228 gen->old_blocks = ws->todo_bd;
1229 gen->n_old_blocks += ws->todo_bd->blocks;
1230 alloc_todo_block(ws,0); // always has one block.
1234 // mark the small objects as from-space
1235 for (bd = gen->old_blocks; bd; bd = bd->link) {
1236 bd->flags &= ~BF_EVACUATED;
1239 // mark the large objects as from-space
1240 for (bd = gen->large_objects; bd; bd = bd->link) {
1241 bd->flags &= ~BF_EVACUATED;
1244 // for a compacted generation, we need to allocate the bitmap
1246 nat bitmap_size; // in bytes
1247 bdescr *bitmap_bdescr;
1250 bitmap_size = gen->n_old_blocks * BLOCK_SIZE / (sizeof(W_)*BITS_PER_BYTE);
1252 if (bitmap_size > 0) {
1253 bitmap_bdescr = allocGroup((lnat)BLOCK_ROUND_UP(bitmap_size)
1255 gen->bitmap = bitmap_bdescr;
1256 bitmap = bitmap_bdescr->start;
1258 debugTrace(DEBUG_gc, "bitmap_size: %d, bitmap: %p",
1259 bitmap_size, bitmap);
1261 // don't forget to fill it with zeros!
1262 memset(bitmap, 0, bitmap_size);
1264 // For each block in this step, point to its bitmap from the
1265 // block descriptor.
1266 for (bd=gen->old_blocks; bd != NULL; bd = bd->link) {
1267 bd->u.bitmap = bitmap;
1268 bitmap += BLOCK_SIZE_W / (sizeof(W_)*BITS_PER_BYTE);
1270 // Also at this point we set the BF_MARKED flag
1271 // for this block. The invariant is that
1272 // BF_MARKED is always unset, except during GC
1273 // when it is set on those blocks which will be
1275 if (!(bd->flags & BF_FRAGMENTED)) {
1276 bd->flags |= BF_MARKED;
1279 // BF_SWEPT should be marked only for blocks that are being
1280 // collected in sweep()
1281 bd->flags &= ~BF_SWEPT;
1288 /* ----------------------------------------------------------------------------
1289 Save the mutable lists in saved_mut_lists
1290 ------------------------------------------------------------------------- */
1293 stash_mut_list (Capability *cap, nat gen_no)
1295 cap->saved_mut_lists[gen_no] = cap->mut_lists[gen_no];
1296 cap->mut_lists[gen_no] = allocBlock_sync();
1299 /* ----------------------------------------------------------------------------
1300 Initialise a generation that is *not* to be collected
1301 ------------------------------------------------------------------------- */
1304 prepare_uncollected_gen (generation *gen)
1309 ASSERT(gen->no > 0);
1311 // save the current mutable lists for this generation, and
1312 // allocate a fresh block for each one. We'll traverse these
1313 // mutable lists as roots early on in the GC.
1314 for (i = 0; i < n_capabilities; i++) {
1315 stash_mut_list(&capabilities[i], gen->no);
1318 ASSERT(gen->scavenged_large_objects == NULL);
1319 ASSERT(gen->n_scavenged_large_blocks == 0);
1322 /* -----------------------------------------------------------------------------
1323 Collect the completed blocks from a GC thread and attach them to
1325 -------------------------------------------------------------------------- */
1328 collect_gct_blocks (void)
1334 for (g = 0; g < RtsFlags.GcFlags.generations; g++) {
1337 // there may still be a block attached to ws->todo_bd;
1338 // leave it there to use next time.
1340 if (ws->scavd_list != NULL) {
1341 ACQUIRE_SPIN_LOCK(&ws->gen->sync);
1343 ASSERT(gct->scan_bd == NULL);
1344 ASSERT(countBlocks(ws->scavd_list) == ws->n_scavd_blocks);
1347 for (bd = ws->scavd_list; bd != NULL; bd = bd->link) {
1348 ws->gen->n_words += bd->free - bd->start;
1352 prev->link = ws->gen->blocks;
1353 ws->gen->blocks = ws->scavd_list;
1355 ws->gen->n_blocks += ws->n_scavd_blocks;
1357 ws->scavd_list = NULL;
1358 ws->n_scavd_blocks = 0;
1360 RELEASE_SPIN_LOCK(&ws->gen->sync);
1365 /* -----------------------------------------------------------------------------
1366 Initialise a gc_thread before GC
1367 -------------------------------------------------------------------------- */
1370 init_gc_thread (gc_thread *t)
1372 t->static_objects = END_OF_STATIC_LIST;
1373 t->scavenged_static_objects = END_OF_STATIC_LIST;
1375 t->mut_lists = capabilities[t->thread_index].mut_lists;
1377 t->failed_to_evac = rtsFalse;
1378 t->eager_promotion = rtsTrue;
1379 t->thunk_selector_depth = 0;
1384 t->scav_find_work = 0;
1387 /* -----------------------------------------------------------------------------
1388 Function we pass to evacuate roots.
1389 -------------------------------------------------------------------------- */
1392 mark_root(void *user USED_IF_THREADS, StgClosure **root)
1394 // we stole a register for gct, but this function is called from
1395 // *outside* the GC where the register variable is not in effect,
1396 // so we need to save and restore it here. NB. only call
1397 // mark_root() from the main GC thread, otherwise gct will be
1399 gc_thread *saved_gct;
1408 /* -----------------------------------------------------------------------------
1409 Initialising the static object & mutable lists
1410 -------------------------------------------------------------------------- */
1413 zero_static_object_list(StgClosure* first_static)
1417 const StgInfoTable *info;
1419 for (p = first_static; p != END_OF_STATIC_LIST; p = link) {
1421 link = *STATIC_LINK(info, p);
1422 *STATIC_LINK(info,p) = NULL;
1426 /* ----------------------------------------------------------------------------
1427 Reset the sizes of the older generations when we do a major
1430 CURRENT STRATEGY: make all generations except zero the same size.
1431 We have to stay within the maximum heap size, and leave a certain
1432 percentage of the maximum heap size available to allocate into.
1433 ------------------------------------------------------------------------- */
1436 resize_generations (void)
1440 if (major_gc && RtsFlags.GcFlags.generations > 1) {
1441 nat live, size, min_alloc, words;
1442 const nat max = RtsFlags.GcFlags.maxHeapSize;
1443 const nat gens = RtsFlags.GcFlags.generations;
1445 // live in the oldest generations
1446 if (oldest_gen->live_estimate != 0) {
1447 words = oldest_gen->live_estimate;
1449 words = oldest_gen->n_words;
1451 live = (words + BLOCK_SIZE_W - 1) / BLOCK_SIZE_W +
1452 oldest_gen->n_large_blocks;
1454 // default max size for all generations except zero
1455 size = stg_max(live * RtsFlags.GcFlags.oldGenFactor,
1456 RtsFlags.GcFlags.minOldGenSize);
1458 if (RtsFlags.GcFlags.heapSizeSuggestionAuto) {
1459 RtsFlags.GcFlags.heapSizeSuggestion = size;
1462 // minimum size for generation zero
1463 min_alloc = stg_max((RtsFlags.GcFlags.pcFreeHeap * max) / 200,
1464 RtsFlags.GcFlags.minAllocAreaSize);
1466 // Auto-enable compaction when the residency reaches a
1467 // certain percentage of the maximum heap size (default: 30%).
1468 if (RtsFlags.GcFlags.compact ||
1470 oldest_gen->n_blocks >
1471 (RtsFlags.GcFlags.compactThreshold * max) / 100)) {
1472 oldest_gen->mark = 1;
1473 oldest_gen->compact = 1;
1474 // debugBelch("compaction: on\n", live);
1476 oldest_gen->mark = 0;
1477 oldest_gen->compact = 0;
1478 // debugBelch("compaction: off\n", live);
1481 if (RtsFlags.GcFlags.sweep) {
1482 oldest_gen->mark = 1;
1485 // if we're going to go over the maximum heap size, reduce the
1486 // size of the generations accordingly. The calculation is
1487 // different if compaction is turned on, because we don't need
1488 // to double the space required to collect the old generation.
1491 // this test is necessary to ensure that the calculations
1492 // below don't have any negative results - we're working
1493 // with unsigned values here.
1494 if (max < min_alloc) {
1498 if (oldest_gen->compact) {
1499 if ( (size + (size - 1) * (gens - 2) * 2) + min_alloc > max ) {
1500 size = (max - min_alloc) / ((gens - 1) * 2 - 1);
1503 if ( (size * (gens - 1) * 2) + min_alloc > max ) {
1504 size = (max - min_alloc) / ((gens - 1) * 2);
1514 debugBelch("live: %d, min_alloc: %d, size : %d, max = %d\n", live,
1515 min_alloc, size, max);
1518 for (g = 0; g < gens; g++) {
1519 generations[g].max_blocks = size;
1524 /* -----------------------------------------------------------------------------
1525 Calculate the new size of the nursery, and resize it.
1526 -------------------------------------------------------------------------- */
1529 resize_nursery (void)
1531 const lnat min_nursery = RtsFlags.GcFlags.minAllocAreaSize * n_capabilities;
1533 if (RtsFlags.GcFlags.generations == 1)
1534 { // Two-space collector:
1537 /* set up a new nursery. Allocate a nursery size based on a
1538 * function of the amount of live data (by default a factor of 2)
1539 * Use the blocks from the old nursery if possible, freeing up any
1542 * If we get near the maximum heap size, then adjust our nursery
1543 * size accordingly. If the nursery is the same size as the live
1544 * data (L), then we need 3L bytes. We can reduce the size of the
1545 * nursery to bring the required memory down near 2L bytes.
1547 * A normal 2-space collector would need 4L bytes to give the same
1548 * performance we get from 3L bytes, reducing to the same
1549 * performance at 2L bytes.
1551 blocks = generations[0].n_blocks;
1553 if ( RtsFlags.GcFlags.maxHeapSize != 0 &&
1554 blocks * RtsFlags.GcFlags.oldGenFactor * 2 >
1555 RtsFlags.GcFlags.maxHeapSize )
1557 long adjusted_blocks; // signed on purpose
1560 adjusted_blocks = (RtsFlags.GcFlags.maxHeapSize - 2 * blocks);
1562 debugTrace(DEBUG_gc, "near maximum heap size of 0x%x blocks, blocks = %d, adjusted to %ld",
1563 RtsFlags.GcFlags.maxHeapSize, blocks, adjusted_blocks);
1565 pc_free = adjusted_blocks * 100 / RtsFlags.GcFlags.maxHeapSize;
1566 if (pc_free < RtsFlags.GcFlags.pcFreeHeap) /* might even * be < 0 */
1570 blocks = adjusted_blocks;
1574 blocks *= RtsFlags.GcFlags.oldGenFactor;
1575 if (blocks < min_nursery)
1577 blocks = min_nursery;
1580 resizeNurseries(blocks);
1582 else // Generational collector
1585 * If the user has given us a suggested heap size, adjust our
1586 * allocation area to make best use of the memory available.
1588 if (RtsFlags.GcFlags.heapSizeSuggestion)
1591 const nat needed = calcNeeded(); // approx blocks needed at next GC
1593 /* Guess how much will be live in generation 0 step 0 next time.
1594 * A good approximation is obtained by finding the
1595 * percentage of g0 that was live at the last minor GC.
1597 * We have an accurate figure for the amount of copied data in
1598 * 'copied', but we must convert this to a number of blocks, with
1599 * a small adjustment for estimated slop at the end of a block
1604 g0_pcnt_kept = ((copied / (BLOCK_SIZE_W - 10)) * 100)
1605 / countNurseryBlocks();
1608 /* Estimate a size for the allocation area based on the
1609 * information available. We might end up going slightly under
1610 * or over the suggested heap size, but we should be pretty
1613 * Formula: suggested - needed
1614 * ----------------------------
1615 * 1 + g0_pcnt_kept/100
1617 * where 'needed' is the amount of memory needed at the next
1618 * collection for collecting all gens except g0.
1621 (((long)RtsFlags.GcFlags.heapSizeSuggestion - (long)needed) * 100) /
1622 (100 + (long)g0_pcnt_kept);
1624 if (blocks < (long)min_nursery) {
1625 blocks = min_nursery;
1628 resizeNurseries((nat)blocks);
1632 // we might have added extra large blocks to the nursery, so
1633 // resize back to minAllocAreaSize again.
1634 resizeNurseriesFixed(RtsFlags.GcFlags.minAllocAreaSize);
1639 /* -----------------------------------------------------------------------------
1640 Sanity code for CAF garbage collection.
1642 With DEBUG turned on, we manage a CAF list in addition to the SRT
1643 mechanism. After GC, we run down the CAF list and blackhole any
1644 CAFs which have been garbage collected. This means we get an error
1645 whenever the program tries to enter a garbage collected CAF.
1647 Any garbage collected CAFs are taken off the CAF list at the same
1649 -------------------------------------------------------------------------- */
1651 #if 0 && defined(DEBUG)
1658 const StgInfoTable *info;
1669 ASSERT(info->type == IND_STATIC);
1671 if (STATIC_LINK(info,p) == NULL) {
1672 debugTrace(DEBUG_gccafs, "CAF gc'd at 0x%04lx", (long)p);
1674 SET_INFO(p,&stg_BLACKHOLE_info);
1675 p = STATIC_LINK2(info,p);
1679 pp = &STATIC_LINK2(info,p);
1686 debugTrace(DEBUG_gccafs, "%d CAFs live", i);