src/share/vm/gc_implementation/g1/g1BlockOffsetTable.hpp

changeset 777
37f87013dfd8
child 1907
c18cbe5936b8
     1.1 --- /dev/null	Thu Jan 01 00:00:00 1970 +0000
     1.2 +++ b/src/share/vm/gc_implementation/g1/g1BlockOffsetTable.hpp	Thu Jun 05 15:57:56 2008 -0700
     1.3 @@ -0,0 +1,487 @@
     1.4 +/*
     1.5 + * Copyright 2001-2007 Sun Microsystems, Inc.  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 Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
    1.23 + * CA 95054 USA or visit www.sun.com if you need additional information or
    1.24 + * have any questions.
    1.25 + *
    1.26 + */
    1.27 +
    1.28 +// The CollectedHeap type requires subtypes to implement a method
    1.29 +// "block_start".  For some subtypes, notably generational
    1.30 +// systems using card-table-based write barriers, the efficiency of this
    1.31 +// operation may be important.  Implementations of the "BlockOffsetArray"
    1.32 +// class may be useful in providing such efficient implementations.
    1.33 +//
    1.34 +// While generally mirroring the structure of the BOT for GenCollectedHeap,
    1.35 +// the following types are tailored more towards G1's uses; these should,
    1.36 +// however, be merged back into a common BOT to avoid code duplication
    1.37 +// and reduce maintenance overhead.
    1.38 +//
    1.39 +//    G1BlockOffsetTable (abstract)
    1.40 +//    -- G1BlockOffsetArray                (uses G1BlockOffsetSharedArray)
    1.41 +//       -- G1BlockOffsetArrayContigSpace
    1.42 +//
    1.43 +// A main impediment to the consolidation of this code might be the
    1.44 +// effect of making some of the block_start*() calls non-const as
    1.45 +// below. Whether that might adversely affect performance optimizations
    1.46 +// that compilers might normally perform in the case of non-G1
    1.47 +// collectors needs to be carefully investigated prior to any such
    1.48 +// consolidation.
    1.49 +
    1.50 +// Forward declarations
    1.51 +class ContiguousSpace;
    1.52 +class G1BlockOffsetSharedArray;
    1.53 +
    1.54 +class G1BlockOffsetTable VALUE_OBJ_CLASS_SPEC {
    1.55 +  friend class VMStructs;
    1.56 +protected:
    1.57 +  // These members describe the region covered by the table.
    1.58 +
    1.59 +  // The space this table is covering.
    1.60 +  HeapWord* _bottom;    // == reserved.start
    1.61 +  HeapWord* _end;       // End of currently allocated region.
    1.62 +
    1.63 +public:
    1.64 +  // Initialize the table to cover the given space.
    1.65 +  // The contents of the initial table are undefined.
    1.66 +  G1BlockOffsetTable(HeapWord* bottom, HeapWord* end) :
    1.67 +    _bottom(bottom), _end(end)
    1.68 +    {
    1.69 +      assert(_bottom <= _end, "arguments out of order");
    1.70 +    }
    1.71 +
    1.72 +  // Note that the committed size of the covered space may have changed,
    1.73 +  // so the table size might also wish to change.
    1.74 +  virtual void resize(size_t new_word_size) = 0;
    1.75 +
    1.76 +  virtual void set_bottom(HeapWord* new_bottom) {
    1.77 +    assert(new_bottom <= _end, "new_bottom > _end");
    1.78 +    _bottom = new_bottom;
    1.79 +    resize(pointer_delta(_end, _bottom));
    1.80 +  }
    1.81 +
    1.82 +  // Requires "addr" to be contained by a block, and returns the address of
    1.83 +  // the start of that block.  (May have side effects, namely updating of
    1.84 +  // shared array entries that "point" too far backwards.  This can occur,
    1.85 +  // for example, when LAB allocation is used in a space covered by the
    1.86 +  // table.)
    1.87 +  virtual HeapWord* block_start_unsafe(const void* addr) = 0;
    1.88 +  // Same as above, but does not have any of the possible side effects
    1.89 +  // discussed above.
    1.90 +  virtual HeapWord* block_start_unsafe_const(const void* addr) const = 0;
    1.91 +
    1.92 +  // Returns the address of the start of the block containing "addr", or
    1.93 +  // else "null" if it is covered by no block.  (May have side effects,
    1.94 +  // namely updating of shared array entries that "point" too far
    1.95 +  // backwards.  This can occur, for example, when lab allocation is used
    1.96 +  // in a space covered by the table.)
    1.97 +  inline HeapWord* block_start(const void* addr);
    1.98 +  // Same as above, but does not have any of the possible side effects
    1.99 +  // discussed above.
   1.100 +  inline HeapWord* block_start_const(const void* addr) const;
   1.101 +};
   1.102 +
   1.103 +// This implementation of "G1BlockOffsetTable" divides the covered region
   1.104 +// into "N"-word subregions (where "N" = 2^"LogN".  An array with an entry
   1.105 +// for each such subregion indicates how far back one must go to find the
   1.106 +// start of the chunk that includes the first word of the subregion.
   1.107 +//
   1.108 +// Each BlockOffsetArray is owned by a Space.  However, the actual array
   1.109 +// may be shared by several BlockOffsetArrays; this is useful
   1.110 +// when a single resizable area (such as a generation) is divided up into
   1.111 +// several spaces in which contiguous allocation takes place,
   1.112 +// such as, for example, in G1 or in the train generation.)
   1.113 +
   1.114 +// Here is the shared array type.
   1.115 +
   1.116 +class G1BlockOffsetSharedArray: public CHeapObj {
   1.117 +  friend class G1BlockOffsetArray;
   1.118 +  friend class G1BlockOffsetArrayContigSpace;
   1.119 +  friend class VMStructs;
   1.120 +
   1.121 +private:
   1.122 +  // The reserved region covered by the shared array.
   1.123 +  MemRegion _reserved;
   1.124 +
   1.125 +  // End of the current committed region.
   1.126 +  HeapWord* _end;
   1.127 +
   1.128 +  // Array for keeping offsets for retrieving object start fast given an
   1.129 +  // address.
   1.130 +  VirtualSpace _vs;
   1.131 +  u_char* _offset_array;          // byte array keeping backwards offsets
   1.132 +
   1.133 +  // Bounds checking accessors:
   1.134 +  // For performance these have to devolve to array accesses in product builds.
   1.135 +  u_char offset_array(size_t index) const {
   1.136 +    assert(index < _vs.committed_size(), "index out of range");
   1.137 +    return _offset_array[index];
   1.138 +  }
   1.139 +
   1.140 +  void set_offset_array(size_t index, u_char offset) {
   1.141 +    assert(index < _vs.committed_size(), "index out of range");
   1.142 +    assert(offset <= N_words, "offset too large");
   1.143 +    _offset_array[index] = offset;
   1.144 +  }
   1.145 +
   1.146 +  void set_offset_array(size_t index, HeapWord* high, HeapWord* low) {
   1.147 +    assert(index < _vs.committed_size(), "index out of range");
   1.148 +    assert(high >= low, "addresses out of order");
   1.149 +    assert(pointer_delta(high, low) <= N_words, "offset too large");
   1.150 +    _offset_array[index] = (u_char) pointer_delta(high, low);
   1.151 +  }
   1.152 +
   1.153 +  void set_offset_array(HeapWord* left, HeapWord* right, u_char offset) {
   1.154 +    assert(index_for(right - 1) < _vs.committed_size(),
   1.155 +           "right address out of range");
   1.156 +    assert(left  < right, "Heap addresses out of order");
   1.157 +    size_t num_cards = pointer_delta(right, left) >> LogN_words;
   1.158 +    memset(&_offset_array[index_for(left)], offset, num_cards);
   1.159 +  }
   1.160 +
   1.161 +  void set_offset_array(size_t left, size_t right, u_char offset) {
   1.162 +    assert(right < _vs.committed_size(), "right address out of range");
   1.163 +    assert(left  <= right, "indexes out of order");
   1.164 +    size_t num_cards = right - left + 1;
   1.165 +    memset(&_offset_array[left], offset, num_cards);
   1.166 +  }
   1.167 +
   1.168 +  void check_offset_array(size_t index, HeapWord* high, HeapWord* low) const {
   1.169 +    assert(index < _vs.committed_size(), "index out of range");
   1.170 +    assert(high >= low, "addresses out of order");
   1.171 +    assert(pointer_delta(high, low) <= N_words, "offset too large");
   1.172 +    assert(_offset_array[index] == pointer_delta(high, low),
   1.173 +           "Wrong offset");
   1.174 +  }
   1.175 +
   1.176 +  bool is_card_boundary(HeapWord* p) const;
   1.177 +
   1.178 +  // Return the number of slots needed for an offset array
   1.179 +  // that covers mem_region_words words.
   1.180 +  // We always add an extra slot because if an object
   1.181 +  // ends on a card boundary we put a 0 in the next
   1.182 +  // offset array slot, so we want that slot always
   1.183 +  // to be reserved.
   1.184 +
   1.185 +  size_t compute_size(size_t mem_region_words) {
   1.186 +    size_t number_of_slots = (mem_region_words / N_words) + 1;
   1.187 +    return ReservedSpace::page_align_size_up(number_of_slots);
   1.188 +  }
   1.189 +
   1.190 +public:
   1.191 +  enum SomePublicConstants {
   1.192 +    LogN = 9,
   1.193 +    LogN_words = LogN - LogHeapWordSize,
   1.194 +    N_bytes = 1 << LogN,
   1.195 +    N_words = 1 << LogN_words
   1.196 +  };
   1.197 +
   1.198 +  // Initialize the table to cover from "base" to (at least)
   1.199 +  // "base + init_word_size".  In the future, the table may be expanded
   1.200 +  // (see "resize" below) up to the size of "_reserved" (which must be at
   1.201 +  // least "init_word_size".) The contents of the initial table are
   1.202 +  // undefined; it is the responsibility of the constituent
   1.203 +  // G1BlockOffsetTable(s) to initialize cards.
   1.204 +  G1BlockOffsetSharedArray(MemRegion reserved, size_t init_word_size);
   1.205 +
   1.206 +  // Notes a change in the committed size of the region covered by the
   1.207 +  // table.  The "new_word_size" may not be larger than the size of the
   1.208 +  // reserved region this table covers.
   1.209 +  void resize(size_t new_word_size);
   1.210 +
   1.211 +  void set_bottom(HeapWord* new_bottom);
   1.212 +
   1.213 +  // Updates all the BlockOffsetArray's sharing this shared array to
   1.214 +  // reflect the current "top"'s of their spaces.
   1.215 +  void update_offset_arrays();
   1.216 +
   1.217 +  // Return the appropriate index into "_offset_array" for "p".
   1.218 +  inline size_t index_for(const void* p) const;
   1.219 +
   1.220 +  // Return the address indicating the start of the region corresponding to
   1.221 +  // "index" in "_offset_array".
   1.222 +  inline HeapWord* address_for_index(size_t index) const;
   1.223 +};
   1.224 +
   1.225 +// And here is the G1BlockOffsetTable subtype that uses the array.
   1.226 +
   1.227 +class G1BlockOffsetArray: public G1BlockOffsetTable {
   1.228 +  friend class G1BlockOffsetSharedArray;
   1.229 +  friend class G1BlockOffsetArrayContigSpace;
   1.230 +  friend class VMStructs;
   1.231 +private:
   1.232 +  enum SomePrivateConstants {
   1.233 +    N_words = G1BlockOffsetSharedArray::N_words,
   1.234 +    LogN    = G1BlockOffsetSharedArray::LogN
   1.235 +  };
   1.236 +
   1.237 +  // The following enums are used by do_block_helper
   1.238 +  enum Action {
   1.239 +    Action_single,      // BOT records a single block (see single_block())
   1.240 +    Action_mark,        // BOT marks the start of a block (see mark_block())
   1.241 +    Action_check        // Check that BOT records block correctly
   1.242 +                        // (see verify_single_block()).
   1.243 +  };
   1.244 +
   1.245 +  // This is the array, which can be shared by several BlockOffsetArray's
   1.246 +  // servicing different
   1.247 +  G1BlockOffsetSharedArray* _array;
   1.248 +
   1.249 +  // The space that owns this subregion.
   1.250 +  Space* _sp;
   1.251 +
   1.252 +  // If "_sp" is a contiguous space, the field below is the view of "_sp"
   1.253 +  // as a contiguous space, else NULL.
   1.254 +  ContiguousSpace* _csp;
   1.255 +
   1.256 +  // If true, array entries are initialized to 0; otherwise, they are
   1.257 +  // initialized to point backwards to the beginning of the covered region.
   1.258 +  bool _init_to_zero;
   1.259 +
   1.260 +  // The portion [_unallocated_block, _sp.end()) of the space that
   1.261 +  // is a single block known not to contain any objects.
   1.262 +  // NOTE: See BlockOffsetArrayUseUnallocatedBlock flag.
   1.263 +  HeapWord* _unallocated_block;
   1.264 +
   1.265 +  // Sets the entries
   1.266 +  // corresponding to the cards starting at "start" and ending at "end"
   1.267 +  // to point back to the card before "start": the interval [start, end)
   1.268 +  // is right-open.
   1.269 +  void set_remainder_to_point_to_start(HeapWord* start, HeapWord* end);
   1.270 +  // Same as above, except that the args here are a card _index_ interval
   1.271 +  // that is closed: [start_index, end_index]
   1.272 +  void set_remainder_to_point_to_start_incl(size_t start, size_t end);
   1.273 +
   1.274 +  // A helper function for BOT adjustment/verification work
   1.275 +  void do_block_internal(HeapWord* blk_start, HeapWord* blk_end, Action action);
   1.276 +
   1.277 +protected:
   1.278 +
   1.279 +  ContiguousSpace* csp() const { return _csp; }
   1.280 +
   1.281 +  // Returns the address of a block whose start is at most "addr".
   1.282 +  // If "has_max_index" is true, "assumes "max_index" is the last valid one
   1.283 +  // in the array.
   1.284 +  inline HeapWord* block_at_or_preceding(const void* addr,
   1.285 +                                         bool has_max_index,
   1.286 +                                         size_t max_index) const;
   1.287 +
   1.288 +  // "q" is a block boundary that is <= "addr"; "n" is the address of the
   1.289 +  // next block (or the end of the space.)  Return the address of the
   1.290 +  // beginning of the block that contains "addr".  Does so without side
   1.291 +  // effects (see, e.g., spec of  block_start.)
   1.292 +  inline HeapWord*
   1.293 +  forward_to_block_containing_addr_const(HeapWord* q, HeapWord* n,
   1.294 +                                         const void* addr) const;
   1.295 +
   1.296 +  // "q" is a block boundary that is <= "addr"; return the address of the
   1.297 +  // beginning of the block that contains "addr".  May have side effects
   1.298 +  // on "this", by updating imprecise entries.
   1.299 +  inline HeapWord* forward_to_block_containing_addr(HeapWord* q,
   1.300 +                                                    const void* addr);
   1.301 +
   1.302 +  // "q" is a block boundary that is <= "addr"; "n" is the address of the
   1.303 +  // next block (or the end of the space.)  Return the address of the
   1.304 +  // beginning of the block that contains "addr".  May have side effects
   1.305 +  // on "this", by updating imprecise entries.
   1.306 +  HeapWord* forward_to_block_containing_addr_slow(HeapWord* q,
   1.307 +                                                  HeapWord* n,
   1.308 +                                                  const void* addr);
   1.309 +
   1.310 +  // Requires that "*threshold_" be the first array entry boundary at or
   1.311 +  // above "blk_start", and that "*index_" be the corresponding array
   1.312 +  // index.  If the block starts at or crosses "*threshold_", records
   1.313 +  // "blk_start" as the appropriate block start for the array index
   1.314 +  // starting at "*threshold_", and for any other indices crossed by the
   1.315 +  // block.  Updates "*threshold_" and "*index_" to correspond to the first
   1.316 +  // index after the block end.
   1.317 +  void alloc_block_work2(HeapWord** threshold_, size_t* index_,
   1.318 +                         HeapWord* blk_start, HeapWord* blk_end);
   1.319 +
   1.320 +public:
   1.321 +  // The space may not have it's bottom and top set yet, which is why the
   1.322 +  // region is passed as a parameter.  If "init_to_zero" is true, the
   1.323 +  // elements of the array are initialized to zero.  Otherwise, they are
   1.324 +  // initialized to point backwards to the beginning.
   1.325 +  G1BlockOffsetArray(G1BlockOffsetSharedArray* array, MemRegion mr,
   1.326 +                     bool init_to_zero);
   1.327 +
   1.328 +  // Note: this ought to be part of the constructor, but that would require
   1.329 +  // "this" to be passed as a parameter to a member constructor for
   1.330 +  // the containing concrete subtype of Space.
   1.331 +  // This would be legal C++, but MS VC++ doesn't allow it.
   1.332 +  void set_space(Space* sp);
   1.333 +
   1.334 +  // Resets the covered region to the given "mr".
   1.335 +  void set_region(MemRegion mr);
   1.336 +
   1.337 +  // Resets the covered region to one with the same _bottom as before but
   1.338 +  // the "new_word_size".
   1.339 +  void resize(size_t new_word_size);
   1.340 +
   1.341 +  // These must be guaranteed to work properly (i.e., do nothing)
   1.342 +  // when "blk_start" ("blk" for second version) is "NULL".
   1.343 +  virtual void alloc_block(HeapWord* blk_start, HeapWord* blk_end);
   1.344 +  virtual void alloc_block(HeapWord* blk, size_t size) {
   1.345 +    alloc_block(blk, blk + size);
   1.346 +  }
   1.347 +
   1.348 +  // The following methods are useful and optimized for a
   1.349 +  // general, non-contiguous space.
   1.350 +
   1.351 +  // The given arguments are required to be the starts of adjacent ("blk1"
   1.352 +  // before "blk2") well-formed blocks covered by "this".  After this call,
   1.353 +  // they should be considered to form one block.
   1.354 +  virtual void join_blocks(HeapWord* blk1, HeapWord* blk2);
   1.355 +
   1.356 +  // Given a block [blk_start, blk_start + full_blk_size), and
   1.357 +  // a left_blk_size < full_blk_size, adjust the BOT to show two
   1.358 +  // blocks [blk_start, blk_start + left_blk_size) and
   1.359 +  // [blk_start + left_blk_size, blk_start + full_blk_size).
   1.360 +  // It is assumed (and verified in the non-product VM) that the
   1.361 +  // BOT was correct for the original block.
   1.362 +  void split_block(HeapWord* blk_start, size_t full_blk_size,
   1.363 +                           size_t left_blk_size);
   1.364 +
   1.365 +  // Adjust the BOT to show that it has a single block in the
   1.366 +  // range [blk_start, blk_start + size). All necessary BOT
   1.367 +  // cards are adjusted, but _unallocated_block isn't.
   1.368 +  void single_block(HeapWord* blk_start, HeapWord* blk_end);
   1.369 +  void single_block(HeapWord* blk, size_t size) {
   1.370 +    single_block(blk, blk + size);
   1.371 +  }
   1.372 +
   1.373 +  // Adjust BOT to show that it has a block in the range
   1.374 +  // [blk_start, blk_start + size). Only the first card
   1.375 +  // of BOT is touched. It is assumed (and verified in the
   1.376 +  // non-product VM) that the remaining cards of the block
   1.377 +  // are correct.
   1.378 +  void mark_block(HeapWord* blk_start, HeapWord* blk_end);
   1.379 +  void mark_block(HeapWord* blk, size_t size) {
   1.380 +    mark_block(blk, blk + size);
   1.381 +  }
   1.382 +
   1.383 +  // Adjust _unallocated_block to indicate that a particular
   1.384 +  // block has been newly allocated or freed. It is assumed (and
   1.385 +  // verified in the non-product VM) that the BOT is correct for
   1.386 +  // the given block.
   1.387 +  inline void allocated(HeapWord* blk_start, HeapWord* blk_end) {
   1.388 +    // Verify that the BOT shows [blk, blk + blk_size) to be one block.
   1.389 +    verify_single_block(blk_start, blk_end);
   1.390 +    if (BlockOffsetArrayUseUnallocatedBlock) {
   1.391 +      _unallocated_block = MAX2(_unallocated_block, blk_end);
   1.392 +    }
   1.393 +  }
   1.394 +
   1.395 +  inline void allocated(HeapWord* blk, size_t size) {
   1.396 +    allocated(blk, blk + size);
   1.397 +  }
   1.398 +
   1.399 +  inline void freed(HeapWord* blk_start, HeapWord* blk_end);
   1.400 +
   1.401 +  inline void freed(HeapWord* blk, size_t size);
   1.402 +
   1.403 +  virtual HeapWord* block_start_unsafe(const void* addr);
   1.404 +  virtual HeapWord* block_start_unsafe_const(const void* addr) const;
   1.405 +
   1.406 +  // Requires "addr" to be the start of a card and returns the
   1.407 +  // start of the block that contains the given address.
   1.408 +  HeapWord* block_start_careful(const void* addr) const;
   1.409 +
   1.410 +  // If true, initialize array slots with no allocated blocks to zero.
   1.411 +  // Otherwise, make them point back to the front.
   1.412 +  bool init_to_zero() { return _init_to_zero; }
   1.413 +
   1.414 +  // Verification & debugging - ensure that the offset table reflects the fact
   1.415 +  // that the block [blk_start, blk_end) or [blk, blk + size) is a
   1.416 +  // single block of storage. NOTE: can;t const this because of
   1.417 +  // call to non-const do_block_internal() below.
   1.418 +  inline void verify_single_block(HeapWord* blk_start, HeapWord* blk_end) {
   1.419 +    if (VerifyBlockOffsetArray) {
   1.420 +      do_block_internal(blk_start, blk_end, Action_check);
   1.421 +    }
   1.422 +  }
   1.423 +
   1.424 +  inline void verify_single_block(HeapWord* blk, size_t size) {
   1.425 +    verify_single_block(blk, blk + size);
   1.426 +  }
   1.427 +
   1.428 +  // Verify that the given block is before _unallocated_block
   1.429 +  inline void verify_not_unallocated(HeapWord* blk_start,
   1.430 +                                     HeapWord* blk_end) const {
   1.431 +    if (BlockOffsetArrayUseUnallocatedBlock) {
   1.432 +      assert(blk_start < blk_end, "Block inconsistency?");
   1.433 +      assert(blk_end <= _unallocated_block, "_unallocated_block problem");
   1.434 +    }
   1.435 +  }
   1.436 +
   1.437 +  inline void verify_not_unallocated(HeapWord* blk, size_t size) const {
   1.438 +    verify_not_unallocated(blk, blk + size);
   1.439 +  }
   1.440 +
   1.441 +  void check_all_cards(size_t left_card, size_t right_card) const;
   1.442 +};
   1.443 +
   1.444 +// A subtype of BlockOffsetArray that takes advantage of the fact
   1.445 +// that its underlying space is a ContiguousSpace, so that its "active"
   1.446 +// region can be more efficiently tracked (than for a non-contiguous space).
   1.447 +class G1BlockOffsetArrayContigSpace: public G1BlockOffsetArray {
   1.448 +  friend class VMStructs;
   1.449 +
   1.450 +  // allocation boundary at which offset array must be updated
   1.451 +  HeapWord* _next_offset_threshold;
   1.452 +  size_t    _next_offset_index;      // index corresponding to that boundary
   1.453 +
   1.454 +  // Work function to be called when allocation start crosses the next
   1.455 +  // threshold in the contig space.
   1.456 +  void alloc_block_work1(HeapWord* blk_start, HeapWord* blk_end) {
   1.457 +    alloc_block_work2(&_next_offset_threshold, &_next_offset_index,
   1.458 +                      blk_start, blk_end);
   1.459 +  }
   1.460 +
   1.461 +
   1.462 + public:
   1.463 +  G1BlockOffsetArrayContigSpace(G1BlockOffsetSharedArray* array, MemRegion mr);
   1.464 +
   1.465 +  // Initialize the threshold to reflect the first boundary after the
   1.466 +  // bottom of the covered region.
   1.467 +  HeapWord* initialize_threshold();
   1.468 +
   1.469 +  // Zero out the entry for _bottom (offset will be zero).
   1.470 +  void      zero_bottom_entry();
   1.471 +
   1.472 +  // Return the next threshold, the point at which the table should be
   1.473 +  // updated.
   1.474 +  HeapWord* threshold() const { return _next_offset_threshold; }
   1.475 +
   1.476 +  // These must be guaranteed to work properly (i.e., do nothing)
   1.477 +  // when "blk_start" ("blk" for second version) is "NULL".  In this
   1.478 +  // implementation, that's true because NULL is represented as 0, and thus
   1.479 +  // never exceeds the "_next_offset_threshold".
   1.480 +  void alloc_block(HeapWord* blk_start, HeapWord* blk_end) {
   1.481 +    if (blk_end > _next_offset_threshold)
   1.482 +      alloc_block_work1(blk_start, blk_end);
   1.483 +  }
   1.484 +  void alloc_block(HeapWord* blk, size_t size) {
   1.485 +     alloc_block(blk, blk+size);
   1.486 +  }
   1.487 +
   1.488 +  HeapWord* block_start_unsafe(const void* addr);
   1.489 +  HeapWord* block_start_unsafe_const(const void* addr) const;
   1.490 +};

mercurial