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 init_collected_gen (nat g, nat threads);
141 static void init_uncollected_gen (nat g, nat threads);
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);
152 #if 0 && defined(DEBUG)
153 static void gcCAFs (void);
156 /* -----------------------------------------------------------------------------
158 -------------------------------------------------------------------------- */
160 bdescr *mark_stack_top_bd; // topmost block in the mark stack
161 bdescr *mark_stack_bd; // current block in the mark stack
162 StgPtr mark_sp; // pointer to the next unallocated mark stack entry
164 /* -----------------------------------------------------------------------------
165 GarbageCollect: the main entry point to the garbage collector.
167 Locks held: all capabilities are held throughout GarbageCollect().
168 -------------------------------------------------------------------------- */
171 GarbageCollect (rtsBool force_major_gc,
172 nat gc_type USED_IF_THREADS,
177 lnat live, allocated, max_copied, avg_copied, slop;
178 gc_thread *saved_gct;
181 // necessary if we stole a callee-saves register for gct:
185 CostCentreStack *prev_CCS;
190 #if defined(RTS_USER_SIGNALS)
191 if (RtsFlags.MiscFlags.install_signal_handlers) {
197 ASSERT(sizeof(gen_workspace) == 16 * sizeof(StgWord));
198 // otherwise adjust the padding in gen_workspace.
200 // tell the stats department that we've started a GC
203 // tell the STM to discard any cached closures it's hoping to re-use
206 // lock the StablePtr table
215 // attribute any costs to CCS_GC
221 /* Approximate how much we allocated.
222 * Todo: only when generating stats?
224 allocated = calcAllocated();
226 /* Figure out which generation to collect
228 n = initialise_N(force_major_gc);
230 #if defined(THREADED_RTS)
231 work_stealing = RtsFlags.ParFlags.parGcLoadBalancingEnabled &&
232 N >= RtsFlags.ParFlags.parGcLoadBalancingGen;
233 // It's not always a good idea to do load balancing in parallel
234 // GC. In particular, for a parallel program we don't want to
235 // lose locality by moving cached data into another CPU's cache
236 // (this effect can be quite significant).
238 // We could have a more complex way to deterimine whether to do
239 // work stealing or not, e.g. it might be a good idea to do it
240 // if the heap is big. For now, we just turn it on or off with
244 /* Start threads, so they can be spinning up while we finish initialisation.
248 #if defined(THREADED_RTS)
249 /* How many threads will be participating in this GC?
250 * We don't try to parallelise minor GCs (unless the user asks for
251 * it with +RTS -gn0), or mark/compact/sweep GC.
253 if (gc_type == PENDING_GC_PAR) {
254 n_gc_threads = RtsFlags.ParFlags.nNodes;
262 debugTrace(DEBUG_gc, "GC (gen %d): %d KB to collect, %ld MB in use, using %d thread(s)",
263 N, n * (BLOCK_SIZE / 1024), mblocks_allocated, n_gc_threads);
265 #ifdef RTS_GTK_FRONTPANEL
266 if (RtsFlags.GcFlags.frontpanel) {
267 updateFrontPanelBeforeGC(N);
272 // check for memory leaks if DEBUG is on
273 memInventory(DEBUG_gc);
276 // check sanity *before* GC
277 IF_DEBUG(sanity, checkSanity(rtsTrue));
279 // Initialise all our gc_thread structures
280 for (t = 0; t < n_gc_threads; t++) {
281 init_gc_thread(gc_threads[t]);
284 // Initialise all the generations/steps that we're collecting.
285 for (g = 0; g <= N; g++) {
286 init_collected_gen(g,n_gc_threads);
289 // Initialise all the generations/steps that we're *not* collecting.
290 for (g = N+1; g < RtsFlags.GcFlags.generations; g++) {
291 init_uncollected_gen(g,n_gc_threads);
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 // Mutable lists from each generation > N
331 // we want to *scavenge* these roots, not evacuate them: they're not
332 // going to move in this GC.
333 // Also do them in reverse generation order, for the usual reason:
334 // namely to reduce the likelihood of spurious old->new pointers.
336 for (g = RtsFlags.GcFlags.generations-1; g > N; g--) {
337 #if defined(THREADED_RTS)
338 if (n_gc_threads > 1) {
339 scavenge_mutable_list(generations[g].saved_mut_list, &generations[g]);
341 scavenge_mutable_list1(generations[g].saved_mut_list, &generations[g]);
344 scavenge_mutable_list(generations[g].saved_mut_list, &generations[g]);
346 freeChain_sync(generations[g].saved_mut_list);
347 generations[g].saved_mut_list = NULL;
351 // scavenge the capability-private mutable lists. This isn't part
352 // of markSomeCapabilities() because markSomeCapabilities() can only
353 // call back into the GC via mark_root() (due to the gct register
355 if (n_gc_threads == 1) {
356 for (n = 0; n < n_capabilities; n++) {
357 #if defined(THREADED_RTS)
358 scavenge_capability_mut_Lists1(&capabilities[n]);
360 scavenge_capability_mut_lists(&capabilities[n]);
364 scavenge_capability_mut_lists(&capabilities[gct->thread_index]);
367 // follow roots from the CAF list (used by GHCi)
369 markCAFs(mark_root, gct);
371 // follow all the roots that the application knows about.
373 markSomeCapabilities(mark_root, gct, gct->thread_index, n_gc_threads,
374 rtsTrue/*prune sparks*/);
376 #if defined(RTS_USER_SIGNALS)
377 // mark the signal handlers (signals should be already blocked)
378 markSignalHandlers(mark_root, gct);
381 // Mark the weak pointer list, and prepare to detect dead weak pointers.
385 // Mark the stable pointer table.
386 markStablePtrTable(mark_root, gct);
388 /* -------------------------------------------------------------------------
389 * Repeatedly scavenge all the areas we know about until there's no
390 * more scavenging to be done.
394 scavenge_until_all_done();
395 // The other threads are now stopped. We might recurse back to
396 // here, but from now on this is the only thread.
398 // must be last... invariant is that everything is fully
399 // scavenged at this point.
400 if (traverseWeakPtrList()) { // returns rtsTrue if evaced something
405 // If we get to here, there's really nothing left to do.
409 shutdown_gc_threads(n_gc_threads, gct->thread_index);
411 // Now see which stable names are still alive.
415 // We call processHeapClosureForDead() on every closure destroyed during
416 // the current garbage collection, so we invoke LdvCensusForDead().
417 if (RtsFlags.ProfFlags.doHeapProfile == HEAP_BY_LDV
418 || RtsFlags.ProfFlags.bioSelector != NULL)
422 // NO MORE EVACUATION AFTER THIS POINT!
424 // Two-space collector: free the old to-space.
425 // g0->old_blocks is the old nursery
426 // g0->blocks is to-space from the previous GC
427 if (RtsFlags.GcFlags.generations == 1) {
428 if (g0->blocks != NULL) {
429 freeChain(g0->blocks);
434 // For each workspace, in each thread, move the copied blocks to the step
440 for (t = 0; t < n_gc_threads; t++) {
443 for (g = 0; g < RtsFlags.GcFlags.generations; g++) {
446 // Push the final block
448 push_scanned_block(ws->todo_bd, ws);
451 ASSERT(gct->scan_bd == NULL);
452 ASSERT(countBlocks(ws->scavd_list) == ws->n_scavd_blocks);
455 for (bd = ws->scavd_list; bd != NULL; bd = bd->link) {
456 ws->gen->n_words += bd->free - bd->start;
460 prev->link = ws->gen->blocks;
461 ws->gen->blocks = ws->scavd_list;
463 ws->gen->n_blocks += ws->n_scavd_blocks;
467 // Add all the partial blocks *after* we've added all the full
468 // blocks. This is so that we can grab the partial blocks back
469 // again and try to fill them up in the next GC.
470 for (t = 0; t < n_gc_threads; t++) {
473 for (g = 0; g < RtsFlags.GcFlags.generations; g++) {
477 for (bd = ws->part_list; bd != NULL; bd = next) {
479 if (bd->free == bd->start) {
481 ws->part_list = next;
488 ws->gen->n_words += bd->free - bd->start;
493 prev->link = ws->gen->blocks;
494 ws->gen->blocks = ws->part_list;
496 ws->gen->n_blocks += ws->n_part_blocks;
498 ASSERT(countBlocks(ws->gen->blocks) == ws->gen->n_blocks);
499 ASSERT(countOccupied(ws->gen->blocks) == ws->gen->n_words);
504 // Finally: compact or sweep the oldest generation.
505 if (major_gc && oldest_gen->mark) {
506 if (oldest_gen->compact)
507 compact(gct->scavenged_static_objects);
512 /* run through all the generations/steps and tidy up
519 for (i=0; i < n_gc_threads; i++) {
520 if (n_gc_threads > 1) {
521 debugTrace(DEBUG_gc,"thread %d:", i);
522 debugTrace(DEBUG_gc," copied %ld", gc_threads[i]->copied * sizeof(W_));
523 debugTrace(DEBUG_gc," scanned %ld", gc_threads[i]->scanned * sizeof(W_));
524 debugTrace(DEBUG_gc," any_work %ld", gc_threads[i]->any_work);
525 debugTrace(DEBUG_gc," no_work %ld", gc_threads[i]->no_work);
526 debugTrace(DEBUG_gc," scav_find_work %ld", gc_threads[i]->scav_find_work);
528 copied += gc_threads[i]->copied;
529 max_copied = stg_max(gc_threads[i]->copied, max_copied);
531 if (n_gc_threads == 1) {
539 for (g = 0; g < RtsFlags.GcFlags.generations; g++) {
542 generations[g].collections++; // for stats
543 if (n_gc_threads > 1) generations[g].par_collections++;
546 // Count the mutable list as bytes "copied" for the purposes of
547 // stats. Every mutable list is copied during every GC.
549 nat mut_list_size = 0;
550 for (bd = generations[g].mut_list; bd != NULL; bd = bd->link) {
551 mut_list_size += bd->free - bd->start;
553 for (n = 0; n < n_capabilities; n++) {
554 for (bd = capabilities[n].mut_lists[g];
555 bd != NULL; bd = bd->link) {
556 mut_list_size += bd->free - bd->start;
559 copied += mut_list_size;
562 "mut_list_size: %lu (%d vars, %d arrays, %d MVARs, %d others)",
563 (unsigned long)(mut_list_size * sizeof(W_)),
564 mutlist_MUTVARS, mutlist_MUTARRS, mutlist_MVARS, mutlist_OTHERS);
568 gen = &generations[g];
570 // for generations we collected...
573 /* free old memory and shift to-space into from-space for all
574 * the collected steps (except the allocation area). These
575 * freed blocks will probaby be quickly recycled.
579 // tack the new blocks on the end of the existing blocks
580 if (gen->old_blocks != NULL) {
583 for (bd = gen->old_blocks; bd != NULL; bd = next) {
587 if (!(bd->flags & BF_MARKED))
590 gen->old_blocks = next;
599 gen->n_words += bd->free - bd->start;
601 // NB. this step might not be compacted next
602 // time, so reset the BF_MARKED flags.
603 // They are set before GC if we're going to
604 // compact. (search for BF_MARKED above).
605 bd->flags &= ~BF_MARKED;
607 // between GCs, all blocks in the heap except
608 // for the nursery have the BF_EVACUATED flag set.
609 bd->flags |= BF_EVACUATED;
616 prev->link = gen->blocks;
617 gen->blocks = gen->old_blocks;
620 // add the new blocks to the block tally
621 gen->n_blocks += gen->n_old_blocks;
622 ASSERT(countBlocks(gen->blocks) == gen->n_blocks);
623 ASSERT(countOccupied(gen->blocks) == gen->n_words);
627 freeChain(gen->old_blocks);
630 gen->old_blocks = NULL;
631 gen->n_old_blocks = 0;
633 /* LARGE OBJECTS. The current live large objects are chained on
634 * scavenged_large, having been moved during garbage
635 * collection from large_objects. Any objects left on the
636 * large_objects list are therefore dead, so we free them here.
638 freeChain(gen->large_objects);
639 gen->large_objects = gen->scavenged_large_objects;
640 gen->n_large_blocks = gen->n_scavenged_large_blocks;
641 gen->n_new_large_blocks = 0;
642 ASSERT(countBlocks(gen->large_objects) == gen->n_large_blocks);
644 else // for generations > N
646 /* For older generations, we need to append the
647 * scavenged_large_object list (i.e. large objects that have been
648 * promoted during this GC) to the large_object list for that step.
650 for (bd = gen->scavenged_large_objects; bd; bd = next) {
652 dbl_link_onto(bd, &gen->large_objects);
655 // add the new blocks we promoted during this GC
656 gen->n_large_blocks += gen->n_scavenged_large_blocks;
657 ASSERT(countBlocks(gen->large_objects) == gen->n_large_blocks);
659 } // for all generations
661 // update the max size of older generations after a major GC
662 resize_generations();
664 // Calculate the amount of live data for stats.
665 live = calcLiveWords();
667 // Free the small objects allocated via allocate(), since this will
668 // all have been copied into G0S1 now.
669 alloc_blocks_lim = RtsFlags.GcFlags.minAllocAreaSize;
671 // Start a new pinned_object_block
672 for (n = 0; n < n_capabilities; n++) {
673 capabilities[n].pinned_object_block = NULL;
676 // Free the mark stack.
677 if (mark_stack_top_bd != NULL) {
678 debugTrace(DEBUG_gc, "mark stack: %d blocks",
679 countBlocks(mark_stack_top_bd));
680 freeChain(mark_stack_top_bd);
684 for (g = 0; g <= N; g++) {
685 gen = &generations[g];
686 if (gen->bitmap != NULL) {
687 freeGroup(gen->bitmap);
694 // mark the garbage collected CAFs as dead
695 #if 0 && defined(DEBUG) // doesn't work at the moment
696 if (major_gc) { gcCAFs(); }
700 // resetStaticObjectForRetainerProfiling() must be called before
702 if (n_gc_threads > 1) {
703 barf("profiling is currently broken with multi-threaded GC");
704 // ToDo: fix the gct->scavenged_static_objects below
706 resetStaticObjectForRetainerProfiling(gct->scavenged_static_objects);
709 // zero the scavenged static object list
712 for (i = 0; i < n_gc_threads; i++) {
713 zero_static_object_list(gc_threads[i]->scavenged_static_objects);
720 // start any pending finalizers
722 scheduleFinalizers(cap, old_weak_ptr_list);
725 // send exceptions to any threads which were about to die
727 resurrectThreads(resurrected_threads);
730 // Update the stable pointer hash table.
731 updateStablePtrTable(major_gc);
733 // check sanity after GC
734 IF_DEBUG(sanity, checkSanity(rtsTrue));
736 // extra GC trace info
737 IF_DEBUG(gc, statDescribeGens());
740 // symbol-table based profiling
741 /* heapCensus(to_blocks); */ /* ToDo */
744 // restore enclosing cost centre
750 // check for memory leaks if DEBUG is on
751 memInventory(DEBUG_gc);
754 #ifdef RTS_GTK_FRONTPANEL
755 if (RtsFlags.GcFlags.frontpanel) {
756 updateFrontPanelAfterGC( N, live );
760 // ok, GC over: tell the stats department what happened.
761 slop = calcLiveBlocks() * BLOCK_SIZE_W - live;
762 stat_endGC(allocated, live, copied, N, max_copied, avg_copied, slop);
764 // unlock the StablePtr table
767 // Guess which generation we'll collect *next* time
768 initialise_N(force_major_gc);
770 #if defined(RTS_USER_SIGNALS)
771 if (RtsFlags.MiscFlags.install_signal_handlers) {
772 // unblock signals again
773 unblockUserSignals();
782 /* -----------------------------------------------------------------------------
783 Figure out which generation to collect, initialise N and major_gc.
785 Also returns the total number of blocks in generations that will be
787 -------------------------------------------------------------------------- */
790 initialise_N (rtsBool force_major_gc)
793 nat blocks, blocks_total;
798 if (force_major_gc) {
799 N = RtsFlags.GcFlags.generations - 1;
804 for (g = RtsFlags.GcFlags.generations - 1; g >= 0; g--) {
806 blocks = generations[g].n_words / BLOCK_SIZE_W
807 + generations[g].n_large_blocks;
809 if (blocks >= generations[g].max_blocks) {
813 blocks_total += blocks;
817 blocks_total += countNurseryBlocks();
819 major_gc = (N == RtsFlags.GcFlags.generations-1);
823 /* -----------------------------------------------------------------------------
824 Initialise the gc_thread structures.
825 -------------------------------------------------------------------------- */
827 #define GC_THREAD_INACTIVE 0
828 #define GC_THREAD_STANDING_BY 1
829 #define GC_THREAD_RUNNING 2
830 #define GC_THREAD_WAITING_TO_CONTINUE 3
833 new_gc_thread (nat n, gc_thread *t)
840 initSpinLock(&t->gc_spin);
841 initSpinLock(&t->mut_spin);
842 ACQUIRE_SPIN_LOCK(&t->gc_spin);
843 t->wakeup = GC_THREAD_INACTIVE; // starts true, so we can wait for the
844 // thread to start up, see wakeup_gc_threads
848 t->free_blocks = NULL;
857 for (g = 0; g < RtsFlags.GcFlags.generations; g++)
860 ws->gen = &generations[g];
861 ASSERT(g == ws->gen->no);
865 ws->todo_q = newWSDeque(128);
866 ws->todo_overflow = NULL;
867 ws->n_todo_overflow = 0;
869 ws->part_list = NULL;
870 ws->n_part_blocks = 0;
872 ws->scavd_list = NULL;
873 ws->n_scavd_blocks = 0;
881 if (gc_threads == NULL) {
882 #if defined(THREADED_RTS)
884 gc_threads = stgMallocBytes (RtsFlags.ParFlags.nNodes *
888 for (i = 0; i < RtsFlags.ParFlags.nNodes; i++) {
890 stgMallocBytes(sizeof(gc_thread) +
891 RtsFlags.GcFlags.generations * sizeof(gen_workspace),
894 new_gc_thread(i, gc_threads[i]);
897 gc_threads = stgMallocBytes (sizeof(gc_thread*),"alloc_gc_threads");
899 new_gc_thread(0,gc_threads[0]);
908 if (gc_threads != NULL) {
909 #if defined(THREADED_RTS)
911 for (i = 0; i < n_capabilities; i++) {
912 for (g = 0; g < RtsFlags.GcFlags.generations; g++)
914 freeWSDeque(gc_threads[i]->gens[g].todo_q);
916 stgFree (gc_threads[i]);
918 stgFree (gc_threads);
920 for (g = 0; g < RtsFlags.GcFlags.generations; g++)
922 freeWSDeque(gc_threads[0]->gens[g].todo_q);
924 stgFree (gc_threads);
930 /* ----------------------------------------------------------------------------
932 ------------------------------------------------------------------------- */
934 static volatile StgWord gc_running_threads;
940 new = atomic_inc(&gc_running_threads);
941 ASSERT(new <= n_gc_threads);
948 ASSERT(gc_running_threads != 0);
949 return atomic_dec(&gc_running_threads);
962 // scavenge objects in compacted generation
963 if (mark_stack_bd != NULL && !mark_stack_empty()) {
967 // Check for global work in any step. We don't need to check for
968 // local work, because we have already exited scavenge_loop(),
969 // which means there is no local work for this thread.
970 for (g = 0; g < (int)RtsFlags.GcFlags.generations; g++) {
972 if (ws->todo_large_objects) return rtsTrue;
973 if (!looksEmptyWSDeque(ws->todo_q)) return rtsTrue;
974 if (ws->todo_overflow) return rtsTrue;
977 #if defined(THREADED_RTS)
980 // look for work to steal
981 for (n = 0; n < n_gc_threads; n++) {
982 if (n == gct->thread_index) continue;
983 for (g = RtsFlags.GcFlags.generations-1; g >= 0; g--) {
984 ws = &gc_threads[n]->gens[g];
985 if (!looksEmptyWSDeque(ws->todo_q)) return rtsTrue;
992 #if defined(THREADED_RTS)
1000 scavenge_until_all_done (void)
1006 traceEventGcWork(&capabilities[gct->thread_index]);
1008 #if defined(THREADED_RTS)
1009 if (n_gc_threads > 1) {
1018 // scavenge_loop() only exits when there's no work to do
1021 traceEventGcIdle(&capabilities[gct->thread_index]);
1023 debugTrace(DEBUG_gc, "%d GC threads still running", r);
1025 while (gc_running_threads != 0) {
1031 // any_work() does not remove the work from the queue, it
1032 // just checks for the presence of work. If we find any,
1033 // then we increment gc_running_threads and go back to
1034 // scavenge_loop() to perform any pending work.
1037 traceEventGcDone(&capabilities[gct->thread_index]);
1040 #if defined(THREADED_RTS)
1043 gcWorkerThread (Capability *cap)
1045 gc_thread *saved_gct;
1047 // necessary if we stole a callee-saves register for gct:
1050 gct = gc_threads[cap->no];
1051 gct->id = osThreadId();
1053 // Wait until we're told to wake up
1054 RELEASE_SPIN_LOCK(&gct->mut_spin);
1055 gct->wakeup = GC_THREAD_STANDING_BY;
1056 debugTrace(DEBUG_gc, "GC thread %d standing by...", gct->thread_index);
1057 ACQUIRE_SPIN_LOCK(&gct->gc_spin);
1060 // start performance counters in this thread...
1061 if (gct->papi_events == -1) {
1062 papi_init_eventset(&gct->papi_events);
1064 papi_thread_start_gc1_count(gct->papi_events);
1067 // Every thread evacuates some roots.
1069 markSomeCapabilities(mark_root, gct, gct->thread_index, n_gc_threads,
1070 rtsTrue/*prune sparks*/);
1071 scavenge_capability_mut_lists(&capabilities[gct->thread_index]);
1073 scavenge_until_all_done();
1076 // count events in this thread towards the GC totals
1077 papi_thread_stop_gc1_count(gct->papi_events);
1080 // Wait until we're told to continue
1081 RELEASE_SPIN_LOCK(&gct->gc_spin);
1082 gct->wakeup = GC_THREAD_WAITING_TO_CONTINUE;
1083 debugTrace(DEBUG_gc, "GC thread %d waiting to continue...",
1085 ACQUIRE_SPIN_LOCK(&gct->mut_spin);
1086 debugTrace(DEBUG_gc, "GC thread %d on my way...", gct->thread_index);
1093 #if defined(THREADED_RTS)
1096 waitForGcThreads (Capability *cap USED_IF_THREADS)
1098 nat n_threads = RtsFlags.ParFlags.nNodes;
1101 rtsBool retry = rtsTrue;
1104 for (i=0; i < n_threads; i++) {
1105 if (i == me) continue;
1106 if (gc_threads[i]->wakeup != GC_THREAD_STANDING_BY) {
1107 prodCapability(&capabilities[i], cap->running_task);
1110 for (j=0; j < 10; j++) {
1112 for (i=0; i < n_threads; i++) {
1113 if (i == me) continue;
1115 setContextSwitches();
1116 if (gc_threads[i]->wakeup != GC_THREAD_STANDING_BY) {
1126 #endif // THREADED_RTS
1129 start_gc_threads (void)
1131 #if defined(THREADED_RTS)
1132 gc_running_threads = 0;
1137 wakeup_gc_threads (nat n_threads USED_IF_THREADS, nat me USED_IF_THREADS)
1139 #if defined(THREADED_RTS)
1141 for (i=0; i < n_threads; i++) {
1142 if (i == me) continue;
1144 debugTrace(DEBUG_gc, "waking up gc thread %d", i);
1145 if (gc_threads[i]->wakeup != GC_THREAD_STANDING_BY) barf("wakeup_gc_threads");
1147 gc_threads[i]->wakeup = GC_THREAD_RUNNING;
1148 ACQUIRE_SPIN_LOCK(&gc_threads[i]->mut_spin);
1149 RELEASE_SPIN_LOCK(&gc_threads[i]->gc_spin);
1154 // After GC is complete, we must wait for all GC threads to enter the
1155 // standby state, otherwise they may still be executing inside
1156 // any_work(), and may even remain awake until the next GC starts.
1158 shutdown_gc_threads (nat n_threads USED_IF_THREADS, nat me USED_IF_THREADS)
1160 #if defined(THREADED_RTS)
1162 for (i=0; i < n_threads; i++) {
1163 if (i == me) continue;
1164 while (gc_threads[i]->wakeup != GC_THREAD_WAITING_TO_CONTINUE) { write_barrier(); }
1169 #if defined(THREADED_RTS)
1171 releaseGCThreads (Capability *cap USED_IF_THREADS)
1173 nat n_threads = RtsFlags.ParFlags.nNodes;
1176 for (i=0; i < n_threads; i++) {
1177 if (i == me) continue;
1178 if (gc_threads[i]->wakeup != GC_THREAD_WAITING_TO_CONTINUE)
1179 barf("releaseGCThreads");
1181 gc_threads[i]->wakeup = GC_THREAD_INACTIVE;
1182 ACQUIRE_SPIN_LOCK(&gc_threads[i]->gc_spin);
1183 RELEASE_SPIN_LOCK(&gc_threads[i]->mut_spin);
1188 /* ----------------------------------------------------------------------------
1189 Initialise a generation that is to be collected
1190 ------------------------------------------------------------------------- */
1193 init_collected_gen (nat g, nat n_threads)
1200 // Throw away the current mutable list. Invariant: the mutable
1201 // list always has at least one block; this means we can avoid a
1202 // check for NULL in recordMutable().
1204 freeChain(generations[g].mut_list);
1205 generations[g].mut_list = allocBlock();
1206 for (i = 0; i < n_capabilities; i++) {
1207 freeChain(capabilities[i].mut_lists[g]);
1208 capabilities[i].mut_lists[g] = allocBlock();
1212 gen = &generations[g];
1213 ASSERT(gen->no == g);
1215 // we'll construct a new list of threads in this step
1216 // during GC, throw away the current list.
1217 gen->old_threads = gen->threads;
1218 gen->threads = END_TSO_QUEUE;
1220 // deprecate the existing blocks
1221 gen->old_blocks = gen->blocks;
1222 gen->n_old_blocks = gen->n_blocks;
1226 gen->live_estimate = 0;
1228 // initialise the large object queues.
1229 gen->scavenged_large_objects = NULL;
1230 gen->n_scavenged_large_blocks = 0;
1232 // mark the small objects as from-space
1233 for (bd = gen->old_blocks; bd; bd = bd->link) {
1234 bd->flags &= ~BF_EVACUATED;
1237 // mark the large objects as from-space
1238 for (bd = gen->large_objects; bd; bd = bd->link) {
1239 bd->flags &= ~BF_EVACUATED;
1242 // for a compacted generation, we need to allocate the bitmap
1244 nat bitmap_size; // in bytes
1245 bdescr *bitmap_bdescr;
1248 bitmap_size = gen->n_old_blocks * BLOCK_SIZE / (sizeof(W_)*BITS_PER_BYTE);
1250 if (bitmap_size > 0) {
1251 bitmap_bdescr = allocGroup((lnat)BLOCK_ROUND_UP(bitmap_size)
1253 gen->bitmap = bitmap_bdescr;
1254 bitmap = bitmap_bdescr->start;
1256 debugTrace(DEBUG_gc, "bitmap_size: %d, bitmap: %p",
1257 bitmap_size, bitmap);
1259 // don't forget to fill it with zeros!
1260 memset(bitmap, 0, bitmap_size);
1262 // For each block in this step, point to its bitmap from the
1263 // block descriptor.
1264 for (bd=gen->old_blocks; bd != NULL; bd = bd->link) {
1265 bd->u.bitmap = bitmap;
1266 bitmap += BLOCK_SIZE_W / (sizeof(W_)*BITS_PER_BYTE);
1268 // Also at this point we set the BF_MARKED flag
1269 // for this block. The invariant is that
1270 // BF_MARKED is always unset, except during GC
1271 // when it is set on those blocks which will be
1273 if (!(bd->flags & BF_FRAGMENTED)) {
1274 bd->flags |= BF_MARKED;
1280 // For each GC thread, for each step, allocate a "todo" block to
1281 // store evacuated objects to be scavenged, and a block to store
1282 // evacuated objects that do not need to be scavenged.
1283 for (t = 0; t < n_threads; t++) {
1284 ws = &gc_threads[t]->gens[g];
1286 ws->todo_large_objects = NULL;
1288 ws->part_list = NULL;
1289 ws->n_part_blocks = 0;
1291 // allocate the first to-space block; extra blocks will be
1292 // chained on as necessary.
1294 ASSERT(looksEmptyWSDeque(ws->todo_q));
1295 alloc_todo_block(ws,0);
1297 ws->todo_overflow = NULL;
1298 ws->n_todo_overflow = 0;
1300 ws->scavd_list = NULL;
1301 ws->n_scavd_blocks = 0;
1306 /* ----------------------------------------------------------------------------
1307 Initialise a generation that is *not* to be collected
1308 ------------------------------------------------------------------------- */
1311 init_uncollected_gen (nat g, nat threads)
1318 // save the current mutable lists for this generation, and
1319 // allocate a fresh block for each one. We'll traverse these
1320 // mutable lists as roots early on in the GC.
1321 generations[g].saved_mut_list = generations[g].mut_list;
1322 generations[g].mut_list = allocBlock();
1323 for (n = 0; n < n_capabilities; n++) {
1324 capabilities[n].saved_mut_lists[g] = capabilities[n].mut_lists[g];
1325 capabilities[n].mut_lists[g] = allocBlock();
1328 gen = &generations[g];
1330 gen->scavenged_large_objects = NULL;
1331 gen->n_scavenged_large_blocks = 0;
1333 for (t = 0; t < threads; t++) {
1334 ws = &gc_threads[t]->gens[g];
1336 ASSERT(looksEmptyWSDeque(ws->todo_q));
1337 ws->todo_large_objects = NULL;
1339 ws->part_list = NULL;
1340 ws->n_part_blocks = 0;
1342 ws->scavd_list = NULL;
1343 ws->n_scavd_blocks = 0;
1345 // If the block at the head of the list in this generation
1346 // is less than 3/4 full, then use it as a todo block.
1347 if (gen->blocks && isPartiallyFull(gen->blocks))
1349 ws->todo_bd = gen->blocks;
1350 ws->todo_free = ws->todo_bd->free;
1351 ws->todo_lim = ws->todo_bd->start + BLOCK_SIZE_W;
1352 gen->blocks = gen->blocks->link;
1354 gen->n_words -= ws->todo_bd->free - ws->todo_bd->start;
1355 ws->todo_bd->link = NULL;
1356 // we must scan from the current end point.
1357 ws->todo_bd->u.scan = ws->todo_bd->free;
1362 alloc_todo_block(ws,0);
1366 // deal out any more partial blocks to the threads' part_lists
1368 while (gen->blocks && isPartiallyFull(gen->blocks))
1371 gen->blocks = bd->link;
1372 ws = &gc_threads[t]->gens[g];
1373 bd->link = ws->part_list;
1375 ws->n_part_blocks += 1;
1376 bd->u.scan = bd->free;
1378 gen->n_words -= bd->free - bd->start;
1380 if (t == n_gc_threads) t = 0;
1384 /* -----------------------------------------------------------------------------
1385 Initialise a gc_thread before GC
1386 -------------------------------------------------------------------------- */
1389 init_gc_thread (gc_thread *t)
1391 t->static_objects = END_OF_STATIC_LIST;
1392 t->scavenged_static_objects = END_OF_STATIC_LIST;
1394 t->mut_lists = capabilities[t->thread_index].mut_lists;
1396 t->failed_to_evac = rtsFalse;
1397 t->eager_promotion = rtsTrue;
1398 t->thunk_selector_depth = 0;
1403 t->scav_find_work = 0;
1406 /* -----------------------------------------------------------------------------
1407 Function we pass to evacuate roots.
1408 -------------------------------------------------------------------------- */
1411 mark_root(void *user USED_IF_THREADS, StgClosure **root)
1413 // we stole a register for gct, but this function is called from
1414 // *outside* the GC where the register variable is not in effect,
1415 // so we need to save and restore it here. NB. only call
1416 // mark_root() from the main GC thread, otherwise gct will be
1418 gc_thread *saved_gct;
1427 /* -----------------------------------------------------------------------------
1428 Initialising the static object & mutable lists
1429 -------------------------------------------------------------------------- */
1432 zero_static_object_list(StgClosure* first_static)
1436 const StgInfoTable *info;
1438 for (p = first_static; p != END_OF_STATIC_LIST; p = link) {
1440 link = *STATIC_LINK(info, p);
1441 *STATIC_LINK(info,p) = NULL;
1445 /* ----------------------------------------------------------------------------
1446 Reset the sizes of the older generations when we do a major
1449 CURRENT STRATEGY: make all generations except zero the same size.
1450 We have to stay within the maximum heap size, and leave a certain
1451 percentage of the maximum heap size available to allocate into.
1452 ------------------------------------------------------------------------- */
1455 resize_generations (void)
1459 if (major_gc && RtsFlags.GcFlags.generations > 1) {
1460 nat live, size, min_alloc, words;
1461 nat max = RtsFlags.GcFlags.maxHeapSize;
1462 nat gens = RtsFlags.GcFlags.generations;
1464 // live in the oldest generations
1465 if (oldest_gen->live_estimate != 0) {
1466 words = oldest_gen->live_estimate;
1468 words = oldest_gen->n_words;
1470 live = (words + BLOCK_SIZE_W - 1) / BLOCK_SIZE_W +
1471 oldest_gen->n_large_blocks;
1473 // default max size for all generations except zero
1474 size = stg_max(live * RtsFlags.GcFlags.oldGenFactor,
1475 RtsFlags.GcFlags.minOldGenSize);
1477 if (RtsFlags.GcFlags.heapSizeSuggestionAuto) {
1478 RtsFlags.GcFlags.heapSizeSuggestion = size;
1481 // minimum size for generation zero
1482 min_alloc = stg_max((RtsFlags.GcFlags.pcFreeHeap * max) / 200,
1483 RtsFlags.GcFlags.minAllocAreaSize);
1485 // Auto-enable compaction when the residency reaches a
1486 // certain percentage of the maximum heap size (default: 30%).
1487 if (RtsFlags.GcFlags.generations > 1 &&
1488 (RtsFlags.GcFlags.compact ||
1490 oldest_gen->n_blocks >
1491 (RtsFlags.GcFlags.compactThreshold * max) / 100))) {
1492 oldest_gen->mark = 1;
1493 oldest_gen->compact = 1;
1494 // debugBelch("compaction: on\n", live);
1496 oldest_gen->mark = 0;
1497 oldest_gen->compact = 0;
1498 // debugBelch("compaction: off\n", live);
1501 if (RtsFlags.GcFlags.sweep) {
1502 oldest_gen->mark = 1;
1505 // if we're going to go over the maximum heap size, reduce the
1506 // size of the generations accordingly. The calculation is
1507 // different if compaction is turned on, because we don't need
1508 // to double the space required to collect the old generation.
1511 // this test is necessary to ensure that the calculations
1512 // below don't have any negative results - we're working
1513 // with unsigned values here.
1514 if (max < min_alloc) {
1518 if (oldest_gen->compact) {
1519 if ( (size + (size - 1) * (gens - 2) * 2) + min_alloc > max ) {
1520 size = (max - min_alloc) / ((gens - 1) * 2 - 1);
1523 if ( (size * (gens - 1) * 2) + min_alloc > max ) {
1524 size = (max - min_alloc) / ((gens - 1) * 2);
1534 debugBelch("live: %d, min_alloc: %d, size : %d, max = %d\n", live,
1535 min_alloc, size, max);
1538 for (g = 0; g < gens; g++) {
1539 generations[g].max_blocks = size;
1544 /* -----------------------------------------------------------------------------
1545 Calculate the new size of the nursery, and resize it.
1546 -------------------------------------------------------------------------- */
1549 resize_nursery (void)
1551 lnat min_nursery = RtsFlags.GcFlags.minAllocAreaSize * n_capabilities;
1553 if (RtsFlags.GcFlags.generations == 1)
1554 { // Two-space collector:
1557 /* set up a new nursery. Allocate a nursery size based on a
1558 * function of the amount of live data (by default a factor of 2)
1559 * Use the blocks from the old nursery if possible, freeing up any
1562 * If we get near the maximum heap size, then adjust our nursery
1563 * size accordingly. If the nursery is the same size as the live
1564 * data (L), then we need 3L bytes. We can reduce the size of the
1565 * nursery to bring the required memory down near 2L bytes.
1567 * A normal 2-space collector would need 4L bytes to give the same
1568 * performance we get from 3L bytes, reducing to the same
1569 * performance at 2L bytes.
1571 blocks = generations[0].n_blocks;
1573 if ( RtsFlags.GcFlags.maxHeapSize != 0 &&
1574 blocks * RtsFlags.GcFlags.oldGenFactor * 2 >
1575 RtsFlags.GcFlags.maxHeapSize )
1577 long adjusted_blocks; // signed on purpose
1580 adjusted_blocks = (RtsFlags.GcFlags.maxHeapSize - 2 * blocks);
1582 debugTrace(DEBUG_gc, "near maximum heap size of 0x%x blocks, blocks = %d, adjusted to %ld",
1583 RtsFlags.GcFlags.maxHeapSize, blocks, adjusted_blocks);
1585 pc_free = adjusted_blocks * 100 / RtsFlags.GcFlags.maxHeapSize;
1586 if (pc_free < RtsFlags.GcFlags.pcFreeHeap) /* might even * be < 0 */
1590 blocks = adjusted_blocks;
1594 blocks *= RtsFlags.GcFlags.oldGenFactor;
1595 if (blocks < min_nursery)
1597 blocks = min_nursery;
1600 resizeNurseries(blocks);
1602 else // Generational collector
1605 * If the user has given us a suggested heap size, adjust our
1606 * allocation area to make best use of the memory available.
1608 if (RtsFlags.GcFlags.heapSizeSuggestion)
1611 nat needed = calcNeeded(); // approx blocks needed at next GC
1613 /* Guess how much will be live in generation 0 step 0 next time.
1614 * A good approximation is obtained by finding the
1615 * percentage of g0 that was live at the last minor GC.
1617 * We have an accurate figure for the amount of copied data in
1618 * 'copied', but we must convert this to a number of blocks, with
1619 * a small adjustment for estimated slop at the end of a block
1624 g0_pcnt_kept = ((copied / (BLOCK_SIZE_W - 10)) * 100)
1625 / countNurseryBlocks();
1628 /* Estimate a size for the allocation area based on the
1629 * information available. We might end up going slightly under
1630 * or over the suggested heap size, but we should be pretty
1633 * Formula: suggested - needed
1634 * ----------------------------
1635 * 1 + g0_pcnt_kept/100
1637 * where 'needed' is the amount of memory needed at the next
1638 * collection for collecting all gens except g0.
1641 (((long)RtsFlags.GcFlags.heapSizeSuggestion - (long)needed) * 100) /
1642 (100 + (long)g0_pcnt_kept);
1644 if (blocks < (long)min_nursery) {
1645 blocks = min_nursery;
1648 resizeNurseries((nat)blocks);
1652 // we might have added extra large blocks to the nursery, so
1653 // resize back to minAllocAreaSize again.
1654 resizeNurseriesFixed(RtsFlags.GcFlags.minAllocAreaSize);
1659 /* -----------------------------------------------------------------------------
1660 Sanity code for CAF garbage collection.
1662 With DEBUG turned on, we manage a CAF list in addition to the SRT
1663 mechanism. After GC, we run down the CAF list and blackhole any
1664 CAFs which have been garbage collected. This means we get an error
1665 whenever the program tries to enter a garbage collected CAF.
1667 Any garbage collected CAFs are taken off the CAF list at the same
1669 -------------------------------------------------------------------------- */
1671 #if 0 && defined(DEBUG)
1678 const StgInfoTable *info;
1689 ASSERT(info->type == IND_STATIC);
1691 if (STATIC_LINK(info,p) == NULL) {
1692 debugTrace(DEBUG_gccafs, "CAF gc'd at 0x%04lx", (long)p);
1694 SET_INFO(p,&stg_BLACKHOLE_info);
1695 p = STATIC_LINK2(info,p);
1699 pp = &STATIC_LINK2(info,p);
1706 debugTrace(DEBUG_gccafs, "%d CAFs live", i);