Mercurial > hg > graal-jvmci-8
view src/share/vm/memory/space.inline.hpp @ 20304:a22acf6d7598
8048112: G1 Full GC needs to support the case when the very first region is not available
Summary: Refactor preparation for compaction during Full GC so that it lazily initializes the first compaction point. This also avoids problems later when the first region may not be committed. Also reviewed by K. Barrett.
Reviewed-by: brutisso
author | tschatzl |
---|---|
date | Mon, 21 Jul 2014 10:00:31 +0200 |
parents | c49dcaf78a65 |
children |
line wrap: on
line source
/* * Copyright (c) 2000, 2012, Oracle and/or its affiliates. All rights reserved. * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. * * This code is free software; you can redistribute it and/or modify it * under the terms of the GNU General Public License version 2 only, as * published by the Free Software Foundation. * * This code is distributed in the hope that it will be useful, but WITHOUT * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License * version 2 for more details (a copy is included in the LICENSE file that * accompanied this code). * * You should have received a copy of the GNU General Public License version * 2 along with this work; if not, write to the Free Software Foundation, * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. * * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA * or visit www.oracle.com if you need additional information or have any * questions. * */ #ifndef SHARE_VM_MEMORY_SPACE_INLINE_HPP #define SHARE_VM_MEMORY_SPACE_INLINE_HPP #include "gc_interface/collectedHeap.hpp" #include "memory/space.hpp" #include "memory/universe.hpp" #include "runtime/prefetch.inline.hpp" #include "runtime/safepoint.hpp" inline HeapWord* Space::block_start(const void* p) { return block_start_const(p); } #define SCAN_AND_FORWARD(cp,scan_limit,block_is_obj,block_size) { \ /* Compute the new addresses for the live objects and store it in the mark \ * Used by universe::mark_sweep_phase2() \ */ \ HeapWord* compact_top; /* This is where we are currently compacting to. */ \ \ /* We're sure to be here before any objects are compacted into this \ * space, so this is a good time to initialize this: \ */ \ set_compaction_top(bottom()); \ \ if (cp->space == NULL) { \ assert(cp->gen != NULL, "need a generation"); \ assert(cp->threshold == NULL, "just checking"); \ assert(cp->gen->first_compaction_space() == this, "just checking"); \ cp->space = cp->gen->first_compaction_space(); \ compact_top = cp->space->bottom(); \ cp->space->set_compaction_top(compact_top); \ cp->threshold = cp->space->initialize_threshold(); \ } else { \ compact_top = cp->space->compaction_top(); \ } \ \ /* We allow some amount of garbage towards the bottom of the space, so \ * we don't start compacting before there is a significant gain to be made.\ * Occasionally, we want to ensure a full compaction, which is determined \ * by the MarkSweepAlwaysCompactCount parameter. \ */ \ uint invocations = MarkSweep::total_invocations(); \ bool skip_dead = ((invocations % MarkSweepAlwaysCompactCount) != 0); \ \ size_t allowed_deadspace = 0; \ if (skip_dead) { \ const size_t ratio = allowed_dead_ratio(); \ allowed_deadspace = (capacity() * ratio / 100) / HeapWordSize; \ } \ \ HeapWord* q = bottom(); \ HeapWord* t = scan_limit(); \ \ HeapWord* end_of_live= q; /* One byte beyond the last byte of the last \ live object. */ \ HeapWord* first_dead = end();/* The first dead object. */ \ LiveRange* liveRange = NULL; /* The current live range, recorded in the \ first header of preceding free area. */ \ _first_dead = first_dead; \ \ const intx interval = PrefetchScanIntervalInBytes; \ \ while (q < t) { \ assert(!block_is_obj(q) || \ oop(q)->mark()->is_marked() || oop(q)->mark()->is_unlocked() || \ oop(q)->mark()->has_bias_pattern(), \ "these are the only valid states during a mark sweep"); \ if (block_is_obj(q) && oop(q)->is_gc_marked()) { \ /* prefetch beyond q */ \ Prefetch::write(q, interval); \ size_t size = block_size(q); \ compact_top = cp->space->forward(oop(q), size, cp, compact_top); \ q += size; \ end_of_live = q; \ } else { \ /* run over all the contiguous dead objects */ \ HeapWord* end = q; \ do { \ /* prefetch beyond end */ \ Prefetch::write(end, interval); \ end += block_size(end); \ } while (end < t && (!block_is_obj(end) || !oop(end)->is_gc_marked()));\ \ /* see if we might want to pretend this object is alive so that \ * we don't have to compact quite as often. \ */ \ if (allowed_deadspace > 0 && q == compact_top) { \ size_t sz = pointer_delta(end, q); \ if (insert_deadspace(allowed_deadspace, q, sz)) { \ compact_top = cp->space->forward(oop(q), sz, cp, compact_top); \ q = end; \ end_of_live = end; \ continue; \ } \ } \ \ /* otherwise, it really is a free region. */ \ \ /* for the previous LiveRange, record the end of the live objects. */ \ if (liveRange) { \ liveRange->set_end(q); \ } \ \ /* record the current LiveRange object. \ * liveRange->start() is overlaid on the mark word. \ */ \ liveRange = (LiveRange*)q; \ liveRange->set_start(end); \ liveRange->set_end(end); \ \ /* see if this is the first dead region. */ \ if (q < first_dead) { \ first_dead = q; \ } \ \ /* move on to the next object */ \ q = end; \ } \ } \ \ assert(q == t, "just checking"); \ if (liveRange != NULL) { \ liveRange->set_end(q); \ } \ _end_of_live = end_of_live; \ if (end_of_live < first_dead) { \ first_dead = end_of_live; \ } \ _first_dead = first_dead; \ \ /* save the compaction_top of the compaction space. */ \ cp->space->set_compaction_top(compact_top); \ } #define SCAN_AND_ADJUST_POINTERS(adjust_obj_size) { \ /* adjust all the interior pointers to point at the new locations of objects \ * Used by MarkSweep::mark_sweep_phase3() */ \ \ HeapWord* q = bottom(); \ HeapWord* t = _end_of_live; /* Established by "prepare_for_compaction". */ \ \ assert(_first_dead <= _end_of_live, "Stands to reason, no?"); \ \ if (q < t && _first_dead > q && \ !oop(q)->is_gc_marked()) { \ /* we have a chunk of the space which hasn't moved and we've \ * reinitialized the mark word during the previous pass, so we can't \ * use is_gc_marked for the traversal. */ \ HeapWord* end = _first_dead; \ \ while (q < end) { \ /* I originally tried to conjoin "block_start(q) == q" to the \ * assertion below, but that doesn't work, because you can't \ * accurately traverse previous objects to get to the current one \ * after their pointers have been \ * updated, until the actual compaction is done. dld, 4/00 */ \ assert(block_is_obj(q), \ "should be at block boundaries, and should be looking at objs"); \ \ /* point all the oops to the new location */ \ size_t size = oop(q)->adjust_pointers(); \ size = adjust_obj_size(size); \ \ q += size; \ } \ \ if (_first_dead == t) { \ q = t; \ } else { \ /* $$$ This is funky. Using this to read the previously written \ * LiveRange. See also use below. */ \ q = (HeapWord*)oop(_first_dead)->mark()->decode_pointer(); \ } \ } \ \ const intx interval = PrefetchScanIntervalInBytes; \ \ debug_only(HeapWord* prev_q = NULL); \ while (q < t) { \ /* prefetch beyond q */ \ Prefetch::write(q, interval); \ if (oop(q)->is_gc_marked()) { \ /* q is alive */ \ /* point all the oops to the new location */ \ size_t size = oop(q)->adjust_pointers(); \ size = adjust_obj_size(size); \ debug_only(prev_q = q); \ q += size; \ } else { \ /* q is not a live object, so its mark should point at the next \ * live object */ \ debug_only(prev_q = q); \ q = (HeapWord*) oop(q)->mark()->decode_pointer(); \ assert(q > prev_q, "we should be moving forward through memory"); \ } \ } \ \ assert(q == t, "just checking"); \ } #define SCAN_AND_COMPACT(obj_size) { \ /* Copy all live objects to their new location \ * Used by MarkSweep::mark_sweep_phase4() */ \ \ HeapWord* q = bottom(); \ HeapWord* const t = _end_of_live; \ debug_only(HeapWord* prev_q = NULL); \ \ if (q < t && _first_dead > q && \ !oop(q)->is_gc_marked()) { \ debug_only( \ /* we have a chunk of the space which hasn't moved and we've reinitialized \ * the mark word during the previous pass, so we can't use is_gc_marked for \ * the traversal. */ \ HeapWord* const end = _first_dead; \ \ while (q < end) { \ size_t size = obj_size(q); \ assert(!oop(q)->is_gc_marked(), \ "should be unmarked (special dense prefix handling)"); \ debug_only(prev_q = q); \ q += size; \ } \ ) /* debug_only */ \ \ if (_first_dead == t) { \ q = t; \ } else { \ /* $$$ Funky */ \ q = (HeapWord*) oop(_first_dead)->mark()->decode_pointer(); \ } \ } \ \ const intx scan_interval = PrefetchScanIntervalInBytes; \ const intx copy_interval = PrefetchCopyIntervalInBytes; \ while (q < t) { \ if (!oop(q)->is_gc_marked()) { \ /* mark is pointer to next marked oop */ \ debug_only(prev_q = q); \ q = (HeapWord*) oop(q)->mark()->decode_pointer(); \ assert(q > prev_q, "we should be moving forward through memory"); \ } else { \ /* prefetch beyond q */ \ Prefetch::read(q, scan_interval); \ \ /* size and destination */ \ size_t size = obj_size(q); \ HeapWord* compaction_top = (HeapWord*)oop(q)->forwardee(); \ \ /* prefetch beyond compaction_top */ \ Prefetch::write(compaction_top, copy_interval); \ \ /* copy object and reinit its mark */ \ assert(q != compaction_top, "everything in this pass should be moving"); \ Copy::aligned_conjoint_words(q, compaction_top, size); \ oop(compaction_top)->init_mark(); \ assert(oop(compaction_top)->klass() != NULL, "should have a class"); \ \ debug_only(prev_q = q); \ q += size; \ } \ } \ \ /* Let's remember if we were empty before we did the compaction. */ \ bool was_empty = used_region().is_empty(); \ /* Reset space after compaction is complete */ \ reset_after_compaction(); \ /* We do this clear, below, since it has overloaded meanings for some */ \ /* space subtypes. For example, OffsetTableContigSpace's that were */ \ /* compacted into will have had their offset table thresholds updated */ \ /* continuously, but those that weren't need to have their thresholds */ \ /* re-initialized. Also mangles unused area for debugging. */ \ if (used_region().is_empty()) { \ if (!was_empty) clear(SpaceDecorator::Mangle); \ } else { \ if (ZapUnusedHeapArea) mangle_unused_area(); \ } \ } inline HeapWord* OffsetTableContigSpace::allocate(size_t size) { HeapWord* res = ContiguousSpace::allocate(size); if (res != NULL) { _offsets.alloc_block(res, size); } return res; } // Because of the requirement of keeping "_offsets" up to date with the // allocations, we sequentialize these with a lock. Therefore, best if // this is used for larger LAB allocations only. inline HeapWord* OffsetTableContigSpace::par_allocate(size_t size) { MutexLocker x(&_par_alloc_lock); // This ought to be just "allocate", because of the lock above, but that // ContiguousSpace::allocate asserts that either the allocating thread // holds the heap lock or it is the VM thread and we're at a safepoint. // The best I (dld) could figure was to put a field in ContiguousSpace // meaning "locking at safepoint taken care of", and set/reset that // here. But this will do for now, especially in light of the comment // above. Perhaps in the future some lock-free manner of keeping the // coordination. HeapWord* res = ContiguousSpace::par_allocate(size); if (res != NULL) { _offsets.alloc_block(res, size); } return res; } inline HeapWord* OffsetTableContigSpace::block_start_const(const void* p) const { return _offsets.block_start(p); } #endif // SHARE_VM_MEMORY_SPACE_INLINE_HPP