1.1 --- /dev/null Thu Jan 01 00:00:00 1970 +0000
1.2 +++ b/src/share/vm/gc_implementation/g1/concurrentMark.inline.hpp Wed Apr 27 01:25:04 2016 +0800
1.3 @@ -0,0 +1,434 @@
1.4 +/*
1.5 + * Copyright (c) 2001, 2014, Oracle and/or its affiliates. All rights reserved.
1.6 + * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
1.7 + *
1.8 + * This code is free software; you can redistribute it and/or modify it
1.9 + * under the terms of the GNU General Public License version 2 only, as
1.10 + * published by the Free Software Foundation.
1.11 + *
1.12 + * This code is distributed in the hope that it will be useful, but WITHOUT
1.13 + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
1.14 + * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
1.15 + * version 2 for more details (a copy is included in the LICENSE file that
1.16 + * accompanied this code).
1.17 + *
1.18 + * You should have received a copy of the GNU General Public License version
1.19 + * 2 along with this work; if not, write to the Free Software Foundation,
1.20 + * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
1.21 + *
1.22 + * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
1.23 + * or visit www.oracle.com if you need additional information or have any
1.24 + * questions.
1.25 + *
1.26 + */
1.27 +
1.28 +#ifndef SHARE_VM_GC_IMPLEMENTATION_G1_CONCURRENTMARK_INLINE_HPP
1.29 +#define SHARE_VM_GC_IMPLEMENTATION_G1_CONCURRENTMARK_INLINE_HPP
1.30 +
1.31 +#include "gc_implementation/g1/concurrentMark.hpp"
1.32 +#include "gc_implementation/g1/g1CollectedHeap.inline.hpp"
1.33 +
1.34 +// Utility routine to set an exclusive range of cards on the given
1.35 +// card liveness bitmap
1.36 +inline void ConcurrentMark::set_card_bitmap_range(BitMap* card_bm,
1.37 + BitMap::idx_t start_idx,
1.38 + BitMap::idx_t end_idx,
1.39 + bool is_par) {
1.40 +
1.41 + // Set the exclusive bit range [start_idx, end_idx).
1.42 + assert((end_idx - start_idx) > 0, "at least one card");
1.43 + assert(end_idx <= card_bm->size(), "sanity");
1.44 +
1.45 + // Silently clip the end index
1.46 + end_idx = MIN2(end_idx, card_bm->size());
1.47 +
1.48 + // For small ranges use a simple loop; otherwise use set_range or
1.49 + // use par_at_put_range (if parallel). The range is made up of the
1.50 + // cards that are spanned by an object/mem region so 8 cards will
1.51 + // allow up to object sizes up to 4K to be handled using the loop.
1.52 + if ((end_idx - start_idx) <= 8) {
1.53 + for (BitMap::idx_t i = start_idx; i < end_idx; i += 1) {
1.54 + if (is_par) {
1.55 + card_bm->par_set_bit(i);
1.56 + } else {
1.57 + card_bm->set_bit(i);
1.58 + }
1.59 + }
1.60 + } else {
1.61 + // Note BitMap::par_at_put_range() and BitMap::set_range() are exclusive.
1.62 + if (is_par) {
1.63 + card_bm->par_at_put_range(start_idx, end_idx, true);
1.64 + } else {
1.65 + card_bm->set_range(start_idx, end_idx);
1.66 + }
1.67 + }
1.68 +}
1.69 +
1.70 +// Returns the index in the liveness accounting card bitmap
1.71 +// for the given address
1.72 +inline BitMap::idx_t ConcurrentMark::card_bitmap_index_for(HeapWord* addr) {
1.73 + // Below, the term "card num" means the result of shifting an address
1.74 + // by the card shift -- address 0 corresponds to card number 0. One
1.75 + // must subtract the card num of the bottom of the heap to obtain a
1.76 + // card table index.
1.77 + intptr_t card_num = intptr_t(uintptr_t(addr) >> CardTableModRefBS::card_shift);
1.78 + return card_num - heap_bottom_card_num();
1.79 +}
1.80 +
1.81 +// Counts the given memory region in the given task/worker
1.82 +// counting data structures.
1.83 +inline void ConcurrentMark::count_region(MemRegion mr, HeapRegion* hr,
1.84 + size_t* marked_bytes_array,
1.85 + BitMap* task_card_bm) {
1.86 + G1CollectedHeap* g1h = _g1h;
1.87 + CardTableModRefBS* ct_bs = g1h->g1_barrier_set();
1.88 +
1.89 + HeapWord* start = mr.start();
1.90 + HeapWord* end = mr.end();
1.91 + size_t region_size_bytes = mr.byte_size();
1.92 + uint index = hr->hrs_index();
1.93 +
1.94 + assert(!hr->continuesHumongous(), "should not be HC region");
1.95 + assert(hr == g1h->heap_region_containing(start), "sanity");
1.96 + assert(hr == g1h->heap_region_containing(mr.last()), "sanity");
1.97 + assert(marked_bytes_array != NULL, "pre-condition");
1.98 + assert(task_card_bm != NULL, "pre-condition");
1.99 +
1.100 + // Add to the task local marked bytes for this region.
1.101 + marked_bytes_array[index] += region_size_bytes;
1.102 +
1.103 + BitMap::idx_t start_idx = card_bitmap_index_for(start);
1.104 + BitMap::idx_t end_idx = card_bitmap_index_for(end);
1.105 +
1.106 + // Note: if we're looking at the last region in heap - end
1.107 + // could be actually just beyond the end of the heap; end_idx
1.108 + // will then correspond to a (non-existent) card that is also
1.109 + // just beyond the heap.
1.110 + if (g1h->is_in_g1_reserved(end) && !ct_bs->is_card_aligned(end)) {
1.111 + // end of region is not card aligned - incremement to cover
1.112 + // all the cards spanned by the region.
1.113 + end_idx += 1;
1.114 + }
1.115 + // The card bitmap is task/worker specific => no need to use
1.116 + // the 'par' BitMap routines.
1.117 + // Set bits in the exclusive bit range [start_idx, end_idx).
1.118 + set_card_bitmap_range(task_card_bm, start_idx, end_idx, false /* is_par */);
1.119 +}
1.120 +
1.121 +// Counts the given memory region in the task/worker counting
1.122 +// data structures for the given worker id.
1.123 +inline void ConcurrentMark::count_region(MemRegion mr,
1.124 + HeapRegion* hr,
1.125 + uint worker_id) {
1.126 + size_t* marked_bytes_array = count_marked_bytes_array_for(worker_id);
1.127 + BitMap* task_card_bm = count_card_bitmap_for(worker_id);
1.128 + count_region(mr, hr, marked_bytes_array, task_card_bm);
1.129 +}
1.130 +
1.131 +// Counts the given memory region, which may be a single object, in the
1.132 +// task/worker counting data structures for the given worker id.
1.133 +inline void ConcurrentMark::count_region(MemRegion mr, uint worker_id) {
1.134 + HeapWord* addr = mr.start();
1.135 + HeapRegion* hr = _g1h->heap_region_containing_raw(addr);
1.136 + count_region(mr, hr, worker_id);
1.137 +}
1.138 +
1.139 +// Counts the given object in the given task/worker counting data structures.
1.140 +inline void ConcurrentMark::count_object(oop obj,
1.141 + HeapRegion* hr,
1.142 + size_t* marked_bytes_array,
1.143 + BitMap* task_card_bm) {
1.144 + MemRegion mr((HeapWord*)obj, obj->size());
1.145 + count_region(mr, hr, marked_bytes_array, task_card_bm);
1.146 +}
1.147 +
1.148 +// Counts the given object in the task/worker counting data
1.149 +// structures for the given worker id.
1.150 +inline void ConcurrentMark::count_object(oop obj,
1.151 + HeapRegion* hr,
1.152 + uint worker_id) {
1.153 + size_t* marked_bytes_array = count_marked_bytes_array_for(worker_id);
1.154 + BitMap* task_card_bm = count_card_bitmap_for(worker_id);
1.155 + HeapWord* addr = (HeapWord*) obj;
1.156 + count_object(obj, hr, marked_bytes_array, task_card_bm);
1.157 +}
1.158 +
1.159 +// Attempts to mark the given object and, if successful, counts
1.160 +// the object in the given task/worker counting structures.
1.161 +inline bool ConcurrentMark::par_mark_and_count(oop obj,
1.162 + HeapRegion* hr,
1.163 + size_t* marked_bytes_array,
1.164 + BitMap* task_card_bm) {
1.165 + HeapWord* addr = (HeapWord*)obj;
1.166 + if (_nextMarkBitMap->parMark(addr)) {
1.167 + // Update the task specific count data for the object.
1.168 + count_object(obj, hr, marked_bytes_array, task_card_bm);
1.169 + return true;
1.170 + }
1.171 + return false;
1.172 +}
1.173 +
1.174 +// Attempts to mark the given object and, if successful, counts
1.175 +// the object in the task/worker counting structures for the
1.176 +// given worker id.
1.177 +inline bool ConcurrentMark::par_mark_and_count(oop obj,
1.178 + size_t word_size,
1.179 + HeapRegion* hr,
1.180 + uint worker_id) {
1.181 + HeapWord* addr = (HeapWord*)obj;
1.182 + if (_nextMarkBitMap->parMark(addr)) {
1.183 + MemRegion mr(addr, word_size);
1.184 + count_region(mr, hr, worker_id);
1.185 + return true;
1.186 + }
1.187 + return false;
1.188 +}
1.189 +
1.190 +// Attempts to mark the given object and, if successful, counts
1.191 +// the object in the task/worker counting structures for the
1.192 +// given worker id.
1.193 +inline bool ConcurrentMark::par_mark_and_count(oop obj,
1.194 + HeapRegion* hr,
1.195 + uint worker_id) {
1.196 + HeapWord* addr = (HeapWord*)obj;
1.197 + if (_nextMarkBitMap->parMark(addr)) {
1.198 + // Update the task specific count data for the object.
1.199 + count_object(obj, hr, worker_id);
1.200 + return true;
1.201 + }
1.202 + return false;
1.203 +}
1.204 +
1.205 +// As above - but we don't know the heap region containing the
1.206 +// object and so have to supply it.
1.207 +inline bool ConcurrentMark::par_mark_and_count(oop obj, uint worker_id) {
1.208 + HeapWord* addr = (HeapWord*)obj;
1.209 + HeapRegion* hr = _g1h->heap_region_containing_raw(addr);
1.210 + return par_mark_and_count(obj, hr, worker_id);
1.211 +}
1.212 +
1.213 +// Similar to the above routine but we already know the size, in words, of
1.214 +// the object that we wish to mark/count
1.215 +inline bool ConcurrentMark::par_mark_and_count(oop obj,
1.216 + size_t word_size,
1.217 + uint worker_id) {
1.218 + HeapWord* addr = (HeapWord*)obj;
1.219 + if (_nextMarkBitMap->parMark(addr)) {
1.220 + // Update the task specific count data for the object.
1.221 + MemRegion mr(addr, word_size);
1.222 + count_region(mr, worker_id);
1.223 + return true;
1.224 + }
1.225 + return false;
1.226 +}
1.227 +
1.228 +// Unconditionally mark the given object, and unconditinally count
1.229 +// the object in the counting structures for worker id 0.
1.230 +// Should *not* be called from parallel code.
1.231 +inline bool ConcurrentMark::mark_and_count(oop obj, HeapRegion* hr) {
1.232 + HeapWord* addr = (HeapWord*)obj;
1.233 + _nextMarkBitMap->mark(addr);
1.234 + // Update the task specific count data for the object.
1.235 + count_object(obj, hr, 0 /* worker_id */);
1.236 + return true;
1.237 +}
1.238 +
1.239 +// As above - but we don't have the heap region containing the
1.240 +// object, so we have to supply it.
1.241 +inline bool ConcurrentMark::mark_and_count(oop obj) {
1.242 + HeapWord* addr = (HeapWord*)obj;
1.243 + HeapRegion* hr = _g1h->heap_region_containing_raw(addr);
1.244 + return mark_and_count(obj, hr);
1.245 +}
1.246 +
1.247 +inline bool CMBitMapRO::iterate(BitMapClosure* cl, MemRegion mr) {
1.248 + HeapWord* start_addr = MAX2(startWord(), mr.start());
1.249 + HeapWord* end_addr = MIN2(endWord(), mr.end());
1.250 +
1.251 + if (end_addr > start_addr) {
1.252 + // Right-open interval [start-offset, end-offset).
1.253 + BitMap::idx_t start_offset = heapWordToOffset(start_addr);
1.254 + BitMap::idx_t end_offset = heapWordToOffset(end_addr);
1.255 +
1.256 + start_offset = _bm.get_next_one_offset(start_offset, end_offset);
1.257 + while (start_offset < end_offset) {
1.258 + if (!cl->do_bit(start_offset)) {
1.259 + return false;
1.260 + }
1.261 + HeapWord* next_addr = MIN2(nextObject(offsetToHeapWord(start_offset)), end_addr);
1.262 + BitMap::idx_t next_offset = heapWordToOffset(next_addr);
1.263 + start_offset = _bm.get_next_one_offset(next_offset, end_offset);
1.264 + }
1.265 + }
1.266 + return true;
1.267 +}
1.268 +
1.269 +inline bool CMBitMapRO::iterate(BitMapClosure* cl) {
1.270 + MemRegion mr(startWord(), sizeInWords());
1.271 + return iterate(cl, mr);
1.272 +}
1.273 +
1.274 +inline void CMTask::push(oop obj) {
1.275 + HeapWord* objAddr = (HeapWord*) obj;
1.276 + assert(_g1h->is_in_g1_reserved(objAddr), "invariant");
1.277 + assert(!_g1h->is_on_master_free_list(
1.278 + _g1h->heap_region_containing((HeapWord*) objAddr)), "invariant");
1.279 + assert(!_g1h->is_obj_ill(obj), "invariant");
1.280 + assert(_nextMarkBitMap->isMarked(objAddr), "invariant");
1.281 +
1.282 + if (_cm->verbose_high()) {
1.283 + gclog_or_tty->print_cr("[%u] pushing " PTR_FORMAT, _worker_id, p2i((void*) obj));
1.284 + }
1.285 +
1.286 + if (!_task_queue->push(obj)) {
1.287 + // The local task queue looks full. We need to push some entries
1.288 + // to the global stack.
1.289 +
1.290 + if (_cm->verbose_medium()) {
1.291 + gclog_or_tty->print_cr("[%u] task queue overflow, "
1.292 + "moving entries to the global stack",
1.293 + _worker_id);
1.294 + }
1.295 + move_entries_to_global_stack();
1.296 +
1.297 + // this should succeed since, even if we overflow the global
1.298 + // stack, we should have definitely removed some entries from the
1.299 + // local queue. So, there must be space on it.
1.300 + bool success = _task_queue->push(obj);
1.301 + assert(success, "invariant");
1.302 + }
1.303 +
1.304 + statsOnly( int tmp_size = _task_queue->size();
1.305 + if (tmp_size > _local_max_size) {
1.306 + _local_max_size = tmp_size;
1.307 + }
1.308 + ++_local_pushes );
1.309 +}
1.310 +
1.311 +// This determines whether the method below will check both the local
1.312 +// and global fingers when determining whether to push on the stack a
1.313 +// gray object (value 1) or whether it will only check the global one
1.314 +// (value 0). The tradeoffs are that the former will be a bit more
1.315 +// accurate and possibly push less on the stack, but it might also be
1.316 +// a little bit slower.
1.317 +
1.318 +#define _CHECK_BOTH_FINGERS_ 1
1.319 +
1.320 +inline void CMTask::deal_with_reference(oop obj) {
1.321 + if (_cm->verbose_high()) {
1.322 + gclog_or_tty->print_cr("[%u] we're dealing with reference = "PTR_FORMAT,
1.323 + _worker_id, p2i((void*) obj));
1.324 + }
1.325 +
1.326 + ++_refs_reached;
1.327 +
1.328 + HeapWord* objAddr = (HeapWord*) obj;
1.329 + assert(obj->is_oop_or_null(true /* ignore mark word */), "Error");
1.330 + if (_g1h->is_in_g1_reserved(objAddr)) {
1.331 + assert(obj != NULL, "null check is implicit");
1.332 + if (!_nextMarkBitMap->isMarked(objAddr)) {
1.333 + // Only get the containing region if the object is not marked on the
1.334 + // bitmap (otherwise, it's a waste of time since we won't do
1.335 + // anything with it).
1.336 + HeapRegion* hr = _g1h->heap_region_containing_raw(obj);
1.337 + if (!hr->obj_allocated_since_next_marking(obj)) {
1.338 + if (_cm->verbose_high()) {
1.339 + gclog_or_tty->print_cr("[%u] "PTR_FORMAT" is not considered marked",
1.340 + _worker_id, p2i((void*) obj));
1.341 + }
1.342 +
1.343 + // we need to mark it first
1.344 + if (_cm->par_mark_and_count(obj, hr, _marked_bytes_array, _card_bm)) {
1.345 + // No OrderAccess:store_load() is needed. It is implicit in the
1.346 + // CAS done in CMBitMap::parMark() call in the routine above.
1.347 + HeapWord* global_finger = _cm->finger();
1.348 +
1.349 +#if _CHECK_BOTH_FINGERS_
1.350 + // we will check both the local and global fingers
1.351 +
1.352 + if (_finger != NULL && objAddr < _finger) {
1.353 + if (_cm->verbose_high()) {
1.354 + gclog_or_tty->print_cr("[%u] below the local finger ("PTR_FORMAT"), "
1.355 + "pushing it", _worker_id, p2i(_finger));
1.356 + }
1.357 + push(obj);
1.358 + } else if (_curr_region != NULL && objAddr < _region_limit) {
1.359 + // do nothing
1.360 + } else if (objAddr < global_finger) {
1.361 + // Notice that the global finger might be moving forward
1.362 + // concurrently. This is not a problem. In the worst case, we
1.363 + // mark the object while it is above the global finger and, by
1.364 + // the time we read the global finger, it has moved forward
1.365 + // passed this object. In this case, the object will probably
1.366 + // be visited when a task is scanning the region and will also
1.367 + // be pushed on the stack. So, some duplicate work, but no
1.368 + // correctness problems.
1.369 +
1.370 + if (_cm->verbose_high()) {
1.371 + gclog_or_tty->print_cr("[%u] below the global finger "
1.372 + "("PTR_FORMAT"), pushing it",
1.373 + _worker_id, p2i(global_finger));
1.374 + }
1.375 + push(obj);
1.376 + } else {
1.377 + // do nothing
1.378 + }
1.379 +#else // _CHECK_BOTH_FINGERS_
1.380 + // we will only check the global finger
1.381 +
1.382 + if (objAddr < global_finger) {
1.383 + // see long comment above
1.384 +
1.385 + if (_cm->verbose_high()) {
1.386 + gclog_or_tty->print_cr("[%u] below the global finger "
1.387 + "("PTR_FORMAT"), pushing it",
1.388 + _worker_id, p2i(global_finger));
1.389 + }
1.390 + push(obj);
1.391 + }
1.392 +#endif // _CHECK_BOTH_FINGERS_
1.393 + }
1.394 + }
1.395 + }
1.396 + }
1.397 +}
1.398 +
1.399 +inline void ConcurrentMark::markPrev(oop p) {
1.400 + assert(!_prevMarkBitMap->isMarked((HeapWord*) p), "sanity");
1.401 + // Note we are overriding the read-only view of the prev map here, via
1.402 + // the cast.
1.403 + ((CMBitMap*)_prevMarkBitMap)->mark((HeapWord*) p);
1.404 +}
1.405 +
1.406 +inline void ConcurrentMark::grayRoot(oop obj, size_t word_size,
1.407 + uint worker_id, HeapRegion* hr) {
1.408 + assert(obj != NULL, "pre-condition");
1.409 + HeapWord* addr = (HeapWord*) obj;
1.410 + if (hr == NULL) {
1.411 + hr = _g1h->heap_region_containing_raw(addr);
1.412 + } else {
1.413 + assert(hr->is_in(addr), "pre-condition");
1.414 + }
1.415 + assert(hr != NULL, "sanity");
1.416 + // Given that we're looking for a region that contains an object
1.417 + // header it's impossible to get back a HC region.
1.418 + assert(!hr->continuesHumongous(), "sanity");
1.419 +
1.420 + // We cannot assert that word_size == obj->size() given that obj
1.421 + // might not be in a consistent state (another thread might be in
1.422 + // the process of copying it). So the best thing we can do is to
1.423 + // assert that word_size is under an upper bound which is its
1.424 + // containing region's capacity.
1.425 + assert(word_size * HeapWordSize <= hr->capacity(),
1.426 + err_msg("size: "SIZE_FORMAT" capacity: "SIZE_FORMAT" "HR_FORMAT,
1.427 + word_size * HeapWordSize, hr->capacity(),
1.428 + HR_FORMAT_PARAMS(hr)));
1.429 +
1.430 + if (addr < hr->next_top_at_mark_start()) {
1.431 + if (!_nextMarkBitMap->isMarked(addr)) {
1.432 + par_mark_and_count(obj, word_size, hr, worker_id);
1.433 + }
1.434 + }
1.435 +}
1.436 +
1.437 +#endif // SHARE_VM_GC_IMPLEMENTATION_G1_CONCURRENTMARK_INLINE_HPP
