DXR is a code search and navigation tool aimed at making sense of large projects. It supports full-text and regex searches as well as structural queries.

Implementation

Mercurial (fbecf6c8a86f)

VCS Links

Line Code
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863
/* -*- Mode: C++; c-basic-offset: 4; indent-tabs-mode: t; tab-width: 4 -*- */
/* ***** BEGIN LICENSE BLOCK *****
 * Version: MPL 1.1/GPL 2.0/LGPL 2.1
 *
 * The contents of this file are subject to the Mozilla Public License Version
 * 1.1 (the "License"); you may not use this file except in compliance with
 * the License. You may obtain a copy of the License at
 * http://www.mozilla.org/MPL/
 *
 * Software distributed under the License is distributed on an "AS IS" basis,
 * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
 * for the specific language governing rights and limitations under the
 * License.
 *
 * The Original Code is [Open Source Virtual Machine.].
 *
 * The Initial Developer of the Original Code is
 * Adobe System Incorporated.
 * Portions created by the Initial Developer are Copyright (C) 2004-2006
 * the Initial Developer. All Rights Reserved.
 *
 * Contributor(s):
 *   Adobe AS3 Team
 *
 * Alternatively, the contents of this file may be used under the terms of
 * either the GNU General Public License Version 2 or later (the "GPL"), or
 * the GNU Lesser General Public License Version 2.1 or later (the "LGPL"),
 * in which case the provisions of the GPL or the LGPL are applicable instead
 * of those above. If you wish to allow use of your version of this file only
 * under the terms of either the GPL or the LGPL, and not to allow others to
 * use your version of this file under the terms of the MPL, indicate your
 * decision by deleting the provisions above and replace them with the notice
 * and other provisions required by the GPL or the LGPL. If you do not delete
 * the provisions above, a recipient may use your version of this file under
 * the terms of any one of the MPL, the GPL or the LGPL.
 *
 * ***** END LICENSE BLOCK ***** */

#ifndef __GCHeap__
#define __GCHeap__

namespace MMgc
{
	class GCHeapConfig
	{
	public:
		GCHeapConfig();
		
		size_t initialSize;
		/**
		 * if the heap gets this big we stop expanding
		 */
		size_t heapLimit;

		/**
		 * if the heap gets this big we fire memory status 
		 * notifications and engage collection activities
		 */
		size_t heapSoftLimit;
		/**
		 * If the application wants the allocator to exit when memory
		 * runs out and reclamation efforts fail set this to a
		 * non-zero value.   Defaults to zero.
		 */
		uint32_t OOMExitCode;
		bool useVirtualMemory;
		bool trimVirtualMemory;
		bool mergeContiguousRegions;
		bool verbose;
		bool returnMemory;
		bool gcstats;
		bool autoGCStats;
		bool gcbehavior;		// Print gross history and policy decisions (MMGC_POLICY_PROFILING)
		bool eagerSweeping;     // Enable full-heap sweeping at the end of Sweep()
#ifdef MMGC_HEAP_GRAPH
		bool dumpFalsePositives;
#endif
		double gcLoad;			// GC load factor: policy aims for a heap size that is gcLoad*H where H is the live size following GC
		double gcLoadCeiling;	// Max multiple of gcLoad policy should use after adjusting L for various factors (0=unlimited)
		double gcEfficiency;    // Max fraction of time to spend in the collector while the incremental collector is active
		static const size_t kDefaultHeapLimit = (size_t)-1;
	};
	
	/**
	 * The GCManager centralizes management of all the memory allocators in the
	 * system, and provides iteration facilities.
	 *
	 * The GCHeap singleton holds the only instance of this manager.
	 */	 
	class GCManager
	{
	public:
		GCManager() {}
		
		/**
		 * Can't have a destructor as it'll be called too late, call destroy to
		 * free any resources.
		 */
		void destroy();

		/**
		 * Register the GC with the manager.  GC must not already be registered.
		 */
		bool tryAddGC(GC* gc) { return collectors.TryAdd(gc); }
		
		/**
		 * Unregister the GC with the manager.  The GC must be registered.
		 */
		void removeGC(GC* gc) { collectors.Remove(gc); }
		
		/**
		 * Tell every other GC that 'gc' is starting a collection (ie there may be memory pressure there).
		 */
		void signalStartCollection(GC* gc);
		
		/**
		 * Tell every other GC that 'gc' is finished with its collection.
		 */
		void signalEndCollection(GC* gc);
		
		BasicList<GC*>& gcs() { return collectors; }

	private:
		BasicList<GC*> collectors;			// array of collectors
	};

	/**
	 * GCHeap is a heap manager for the Flash Player's garbage collector.
	 *
	 * Memory is allocated from the operating system in large chunks
	 * aligned on page boundaries (4096-byte aligned).  All allocation
	 * requests are specified in numbers of 4096-byte pages.  The first
	 * 4 bytes of each page is reserved for GCHeap's use.
	 *
	 * Allocations may fail if there is no free block large enough
	 * to satisfy the request.  When this occurs, the GC class will
	 * choose to either garbage collect or expand the heap, and then
	 * call GCHeap again to retry the allocation.
	 *
	 * When blocks are freed, they are coalesced with their neighbors
	 * when possible to reduce fragmentation.  A "boundary tag" scheme
	 * similar to Doug Lea's malloc is used to make coalescing fast.
	 * To keep things simple, coalescing is always performed immediately
	 * when a block is freed, and not deferred.
	 *
	 * Allocations are performed using multiple free lists.  Blocks that
	 * are <= kUniqueThreshold pages in size get their own free list.
	 * Blocks above the threshold get mapped to more granular free lists,
	 * and all blocks above kHugeThreshold get mapped to the last free list.
	 * This approach is borrowed from Hans Boehm's C/C++ Garbage Collector.
	 * 
	 * Free lists are doubly-linked lists sorted in size order.  The
	 * sorting is done to ensure best-fit allocation.  Each free list
	 * is itself a list node, and the lists are circular, in a technique
	 * also borrowed from Doug Lea's malloc.  This makes for very simple
	 * logic to insert/remove nodes.
	 *
	 * Only Windows is currently supported, although this should be easy
	 * to get working on other platforms.  Memory is allocated in large
	 * (64K minimum) chunks using the VirtualAlloc API.  Every time the
	 * heap is expanded, a new heap region is created corresponding to
	 * the address space returned by VirtualAlloc.  Heap regions are
	 * allocated contiguously if possible to reduce fragmentation.
	 */
	class GCHeap 
	{
		friend class GC;
		friend class FixedAlloc;		
		friend class FixedMalloc;		
		friend class GCPolicyManager;
	public:
		// -- Constants
		
		/** Size of a block */
		const static uint32_t kBlockSize = 4096;

		/** 
		 * Max allowable size for any allocation = 2^32 - 1  bytes
		 * This value is based on the max value on 32-bit systems
		 * and acts as a cap on the size of an allocation request
		 * Overflow detection routines CheckForAllocSizeOverflow()
		 * use this value to check for overflows
		 */
		const static size_t kMaxObjectSize = 0xFFFFFFFF; 
		
		/** Default size of address space reserved per region in blocks */
#ifdef MMGC_64BIT
		const static uint32_t kDefaultReserve = 4096;
#else
		const static uint32_t kDefaultReserve = 512;
#endif
		
		/** Sizes up to this many blocks each have their own free list. */
		const static uint32_t kUniqueThreshold = 16;

		/**
		 * Sizes of at least this many heap blocks are mapped to a
		 * single free list.
		 */
		const static uint32_t kHugeThreshold = 128;

		/** In between sizes map this many distinct sizes to a single bin. */
		const static uint32_t kFreeListCompression = 8;

		/** Calculated number of free lists */
		const static uint32_t kNumFreeLists = (kHugeThreshold-kUniqueThreshold)/kFreeListCompression+kUniqueThreshold;

		/** Minimum heap increment, in blocks */
		const static uint32_t kMinHeapIncrement = 32;

		/** if this much of the heap is free decommit some memory */
		const static uint32_t kDecommitThresholdPercentage = 25;
		/** if this much of the heap is free un-reserve it */
		const static uint32_t kReleaseThresholdPercentage = 50;

		/**
		 * Init must be called to set up the GCHeap singleton
		 */
		static void Init(const GCHeapConfig& props);

		/* legacy API */
		static void Init(GCMallocFuncPtr malloc = NULL, GCFreeFuncPtr free = NULL, int initialSize=128)
		{
			(void)malloc;
			(void)free;
			GCHeapConfig props;
			props.initialSize = initialSize;
			Init(props);
		}

		/**
		 * Destroy the GCHeap singleton, returns the number of bytes still allocated in FixedMalloc (ie leaked)
		 */
		static size_t Destroy();

		/**
		 * Get the GCHeap singleton
		 */
		inline static GCHeap *GetGCHeap() 
		{
			// when OOM occurs the last EnterFrame macro destroys the
			// heap so this can be NULL, clients should detect NULL
			// and re-create the heap if desired
			return instance; 
		}

		static void EnterLockInit();
		static void EnterLockDestroy();

		inline static bool EnterLock()
		{
			GCAssert(instanceEnterLockInitialized); 
			return VMPI_lockAcquire(&instanceEnterLock);
		}

		inline static bool EnterRelease()
		{
			GCAssert(instanceEnterLockInitialized); 
			return VMPI_lockRelease(&instanceEnterLock);
		}

		/**
		 * Signal a too-large allocation request.  This /will/ cause an immediate shutdown of
		 * the entire system.  (The alternative is to return a NULL pointer, which has the
		 * same effect but with considerably less grace.)  Clients that allocate arbitrarily
		 * large objects based on application or end-user data may wish to be concerned about
		 * checking whether the object might be too large.
		 */
#if defined (__GNUC__)
		static void SignalObjectTooLarge() __attribute__((noreturn));
#else
		static void SignalObjectTooLarge();
#endif

		/**
		 * Signal an inconsistent heap state (an unrecoverable program error).  This /will/
		 * cause an immediate shutdown of the entire system.  (The alternative is to return
		 * a NULL pointer, which has the same effect but with considerably less grace.)
		 */
#if defined (__GNUC__)
		static void SignalInconsistentHeapState(const char* reason) __attribute__((noreturn));
#else
		static void SignalInconsistentHeapState(const char* reason);
#endif

		/**
		 * Signal that the caller is about to longjmp to or past an MMGC_ENTER, and that
		 * actions should be taken to leave the heap in a consistent state.
		 *
		 * MMgc code does not use this, it's intended for use by external agents.  It can
		 * be called without ill effect even if there is no active GC or GCHeap.
		 *
		 * longjmp'ing to or past an MMGC_ENTER is not supported in all contexts: host code
		 * should never do it on call-backs on allocation, deallocation, and finalization hooks.
		 */
		static void SignalImminentAbort();
		
		/**
		 * Signal that client code has performed an allocation from memory known not to be
		 * controlled by MMgc, and wants this memory accounted for.  (A typical case is when
		 * system memory is allocated for rendering buffers.)
		 *
		 * Let 'total memory volume' be the sum of the memory volume controlled by MMgc,
		 * plus the volume added by SignalExternalAllocation, minus the volume subtracted
		 * by SignalExternalDeallocation.  MMgc performs its out-of-memory avoidance
		 * actions (actions for crossing the soft and hard limits) based on total memory
		 * volume.
		 */
		static void SignalExternalAllocation(size_t nbytes);
		
		/**
		 * Signal that client code has performed a deallocation of memory known not to be
		 * controlled by MMgc, and wants this memory accounted for.
		 *
		 * @see SignalExternalAllocation
		 */
		static void SignalExternalDeallocation(size_t nbytes);
		
		inline FixedMalloc* GetFixedMalloc() { return FixedMalloc::GetFixedMalloc(); }

		/**
		* flags to be passed as second argument to alloc
		*/
		enum AllocFlags
		{
			kExpand=1,
			kZero=2,
			kProfile=4,
			kCanFail=8
		};

		/**
		 * Allocates a block from the heap.
		 * @param size the number of pages (kBlockSize bytes apiece)
		 *             to allocate.
		 * @return pointer to beginning of block, or NULL if failed.
		 */
		void *Alloc(size_t size, uint32_t flags=kExpand | kZero | kProfile);

		/**
		 * Allocate memory that the JIT will use for object code.
		 * @param size the number of blocks
		 */
		void *AllocCodeMemory(size_t size)
		{
			codeMemory += size;
			return Alloc(size);
		}
		
		/**
		 * Frees a block.
		 * @param item the block to free.  This must be the same
		 *             pointer that was previously returned by
		 *             a call to Alloc.
		 */
		void Free(void *item) { FreeInternal(item, true); }
		void FreeNoProfile(void *item) { FreeInternal(item, false); }

		/**
		 * Free memory allocated from AllocCodeMemory.
		 */
		void FreeCodeMemory(void* item)
		{
			HeapBlock *block = AddrToBlock(item);
			codeMemory -= block->size;
			Free(item);
		}
		
		/**
		 * Added for NJ's portability needs cause it doesn't always MMgc 
		 */
		void Free(void *item, size_t /*ignore*/) { FreeInternal(item, true); }


		size_t Size(const void *item);

		/**
		 * Returns the used heap size, that is, the total
		 * space actually used by allocated objects.
		 * @return the minimum heap size in pages (kBlockSize bytes apiece)
		 */
		size_t GetUsedHeapSize() const { return numAlloc; }

		/**
		 * Returns the "free heap size", that is, the difference in the
		 * total heap size and the used heap size
		 * @return the minimum heap size in pages (kBlockSize bytes apiece)
		 */
		size_t GetFreeHeapSize() const { return GetTotalHeapSize()-numAlloc; }

		/**
		 * @return the amount of code memory currently allocated.  GCHeap does not
		 * have a notion of "free" and "used" code memory.
		 */
		size_t GetTotalCodeSize() const { return codeMemory; }

#ifdef MMGC_POLICY_PROFILING
		/**
		 * Returns the peak value for the total amount of space managed by the heap
		 * and the amount of private memory at the point where the maximum heap
		 * number was recorded.
		 */
		void GetMaxTotalHeapSize(size_t& heapBlocks, size_t& privateBlocks)
		{
			heapBlocks = maxTotalHeapSize / kBlockSize;
			privateBlocks = maxPrivateMemory / kBlockSize;
		}
#endif

		/**
		 * Returns the total heap size, that is, the total amount
		 * of space managed by the heap, including all used and
		 * free space.
		 * @return the total heap size in pages (kBlockSize bytes apiece)
		 */
		size_t GetTotalHeapSize() const { return blocksLen - numDecommitted; }
		
		/**
		 * gives memory back to the OS when there hasn't been any memory activity in a while
		 * and we have lots of free memory
		 */
		void Decommit();

		static size_t SizeToBlocks(size_t bytes) { return ((bytes + kBlockSize - 1) & ~(kBlockSize-1)) / kBlockSize; }

#ifdef MMGC_HOOKS
		/* Hooks are normally disabled in RELEASE builds, as there is a slight cost added
		   to some hot paths.  */
		/* controls whether AllocHook and FreeHook are called */
		void EnableHooks() { hooksEnabled = true; }
		REALLY_INLINE bool HooksEnabled() const { return hooksEnabled; }
		void AllocHook(const void *item, size_t askSize, size_t gotSize);
		// called when object is determined to be garbage but we can't write to it yet
		void FinalizeHook(const void *item, size_t size);
		// called when object is really dead and can be poisoned
		void FreeHook(const void *item, size_t size, int poison);
#endif

#ifdef MMGC_MEMORY_PROFILER
		MemoryProfiler *GetProfiler() { return profiler; }
		void DumpFatties() { profiler->DumpFatties(); }
#endif

		// Every new GC must register itself with the GCHeap.
		void AddGC(GC *gc);
		
		// When the GC is destroyed it must remove itself from the GCHeap.
		void RemoveGC(GC *gc);

		void AddOOMCallback(OOMCallback *p);

		void RemoveOOMCallback(OOMCallback *p);
		
#ifdef MMGC_USE_SYSTEM_MALLOC
		// Signal a failure to allocate 'size' bytes from the system heap (VMPI_alloc).
		// The value 'attempt' denotes the number of previous attempts made to satisfy
		// this particular memory request; the implementation is at liberty to have
		// a cutoff for the number of attempts and must signal an abort if the number
		// of attempts exceeds the cutoff.  (Normally the cutoff would be one previous
		// attempt.)
		void SystemOOMEvent(size_t size, int attempt);
#endif

#if defined (__GNUC__)
		void Abort() __attribute__((noreturn));
#else
		void Abort();
#endif

		MemoryStatus GetStatus() { return status; }

		/**
		 * CheckForAllocSizeOverflow checks whether an allocation request
		 * would request an object larger than what MMgc can accomodate.
		 *
		 * Overflow detection logic: 
		 * Currently, in MMgc all object allocations are capped to 2^32-1 (=kMaxObjectSize)
		 * which is the largest object size on 32-bit systems.  To detect overflow
		 * the standard way is to add the values and look for wraparound by checking if
		 * the result is less than either of the operands.  However, on 64-bit systems
		 * sizeof(size_t) == 8 bytes and the wraparound check would not work.  So this
		 * method also checks if the result exceeds 2^32-1 to conform our allocation size cap.
		 * Requirement:
		 * All allocation routines and methods that add extra payload (such as headers) to 
		 * allocation requests should call this method for overflow detection.
		 * @param size requested size pertaining to object being allocated
		 * @param extra amount of extra bytes accompanying an allocation
		 * @return:  The sum of size and extra.
		 * @note This method may not return.  It is designed to terminate
		 * the program if an overflow is detected.
		 */
		REALLY_INLINE static size_t CheckForAllocSizeOverflow(size_t size, size_t extra)
		{
			//calculate the total requested size
			uint64_t total = (uint64_t)size + (uint64_t)extra;
			
			//check if request size exceeds kMaxObjectSize
			// or for sizeof(size_t) = 8 bytes check for wraparound on the total value
#ifdef MMGC_64BIT
			if ((total > (uint64_t)kMaxObjectSize) || (total < size) || (total < extra))
				SignalObjectTooLarge();
#else
			// This is the 32-bit implementation, it avoids unnecessary checks for overflow.
			if (total > (uint64_t)kMaxObjectSize)
				SignalObjectTooLarge();
#endif
			return size_t(total);
		}
		
		/**
		 * CheckForCallocSizeOverflow
		 * This method is designed to check whether an allocation request
		 * of N objects for a given size might result in numeric overflow.  
		 * Overflow detection logic:
		 * In this method we detect overflow occurring from result of (N * M) where
		 * N = number of objects and M = size of one object.
		 * Currently, in MMgc all object allocations are capped to 2^32-1 (=kMaxObjectSize)
		 * which is the largest object size on 32-bit systems.  
		 * To detect overflow, we first check if either of N or M exceeds kMaxObjectSize.  
		 * This check is a guard against overflow on 64-bit systems where sizeof(size_t) is
		 * greater than 4 bytes.  If this check succeeds then we perform a 64-bit based
		 * product i.e. (N * M) and check if the result exceeds kMaxObjectSize.
		 * Requirement:
		 * All allocation routines and methods that are servicing allocation requests
		 * based on a product of object size and a number of objects should call this method prior to allocation
		 * @param count number of objects being allocated
		 * @param elsize requested size pertaining to a single object being allocated
		 * @return: None.  This method is designed to call Abort() which in turn terminates
		 * the program if an overflow is detected.
		 */		
		REALLY_INLINE static size_t CheckForCallocSizeOverflow(size_t count, size_t elsize)
		{
			//If either of the size of requested bytes
			//or the number of requested size 
			//or if their product exceeds kMaxObjectSize 
			//we treat that as overflow and abort
			uint64_t total = (uint64_t)elsize * (uint64_t)count;
#ifdef MMGC_64BIT
			if(   elsize > kMaxObjectSize 
			   || count >= kMaxObjectSize
			   || total > (uint64_t)kMaxObjectSize)
				SignalObjectTooLarge();
#else
			if(total > (uint64_t)kMaxObjectSize)
				SignalObjectTooLarge();
#endif
			return size_t(total);
		}

		/** The native VM page size (in bytes) for the current architecture */
		const size_t kNativePageSize;

		GCHeapConfig &Config() { return config; }

		void log_percentage(const char *, size_t bytes, size_t relativeTo);

		void DumpMemoryInfo();
		void DumpMemoryInfoLocked();

#ifdef MMGC_MEMORY_PROFILER
#ifdef MMGC_USE_SYSTEM_MALLOC
		void TrackSystemAlloc(void *addr, size_t askSize);
		void TrackSystemFree(void *addr);
#endif //MMGC_USE_SYSTEM_MALLOC
#endif

		void *GetStackEntryAddress() { return (void*)GetEnterFrame(); }
		EnterFrame *GetEnterFrame() { return enterFrame; }

		void SetActiveGC(GC *gc) { if(enterFrame) enterFrame->SetActiveGC(gc); }
		inline bool StackEnteredCheck() { return !entryChecksEnabled || GetEnterFrame() != NULL; }
		
		// remove this and make them always enabled once its possible
		inline void SetEntryChecks(bool to) { entryChecksEnabled = to; }

		// Heap regions
		// (ought to be private but some VMPI implementations 
		// currently need to peek at it)
		class Region
		{
		public:
			Region(GCHeap *heap, char *baseAddr, char *rTop, char *cTop, size_t blockId);
			Region *prev;
			char *baseAddr;
			char *reserveTop;
			char *commitTop;
			size_t blockId;
		};
		Region *lastRegion;
		
		static bool ShouldNotEnter();

		bool IsAddressInHeap(void *);

	private:

		GCHeap(const GCHeapConfig &config);
		void DestroyInstance();

#ifdef MMGC_MEMORY_PROFILER
		static void InitProfiler();
		inline static bool IsProfilerInitialized()
		{
			return profiler != (MemoryProfiler*)-1;
		}
#endif

		/**
		 * Expands the heap by size pages.
		 *
		 * Expands the heap by "size" blocks, such that a single contiguous
		 * allocation of "size" blocks can be performed.  This method is
		 * also called to create the initial heap.
		 *
		 * On Windows, this uses the VirtualAlloc API to obtain memory.
		 * VirtualAlloc can _reserve_ memory, _commit_ memory or both at
		 * the same time.  Reserved memory is just virtual address space.
		 * It consumes the address space of the process but isn't really
		 * allocated yet; there are no pages committed to it yet.
		 * Memory allocation really occurs when reserved pages are
		 * committed.  Our strategy in GCHeap is to reserve a fairly large
		 * chunk of address space, and then commit pages from it as needed.
		 * By doing this, we're more likely to get contiguous regions in
		 * memory for our heap.
		 *
		 * By default, we reserve 16MB (4096 pages) per heap region.
		 * The amount to reserve by default is controlled by kDefaultReserve.
		 * That shouldn't be a big deal, as the process address space is 2GB.
		 * As we're usually a plug-in, however, we don't want to make it too
		 * big because it's not all our memory.
		 *
		 * The goal of reserving so much address space is so that subsequent
		 * expansions of the heap are able to obtain contiguous memory blocks.
		 * If we can keep the heap contiguous, that reduces fragmentation
		 * and the possibility of many small "Balkanized" heap regions.
		 *
		 * Algorithm: When an allocation is requested,
		 * 1. If there is enough reserved but uncommitted memory in the
		 *    last-created region to satisfy the request, commit that memory
		 *    and exit with success, also check decommitted list
		 * 2. Try to reserve a new region contiguous with the last-created
		 *    region.  Go for a 16MB reservation or the requested size,
		 *    whichever is bigger.
		 * 3. If we tried for 16MB reserved space and it didn't work, try
		 *    to reserve again, but for the requested size.
		 * 4. If we were able to retrieve a contiguous region in Step 2 or 3,
		 *    commit any leftover memory from the last-created region,
		 *    commit the remainer from the newly created region, and exit
		 *    with success.
		 * 5. OK, the contiguous region didn't work out, so allocate a
		 *    non-contiguous region.  Go for 16MB or the requested size
		 *    again, whichever is bigger.
		 * 6. If we tried for 16MB reserved space and it didn't work, try
		 *    to reserve again, but for the requested size.
		 * 7. Commit the requested size out of the newly created region
		 *    and exit with success.
		 *
		 * If we are able to reserve memory but can't commit it, then, well
		 * there isn't enough memory.  We free the reserved memory and
		 * exit with failure.
		 *
		 * @param size the number of pages to expand the heap by
		 */	 
		void ExpandHeap(size_t size, bool canFail=false);
		bool ExpandHeapInternal(size_t size);
		
		// Block struct used for free lists and memory traversal
		class HeapBlock 
		{
		public:
			char *baseAddr;   // base address of block's memory
			size_t size;         // size of this block
			size_t sizePrevious; // size of previous block
			HeapBlock *prev;      // prev entry on free list
			HeapBlock *next;      // next entry on free list
			bool committed;   // is block fully committed?
			bool dirty;		  // needs zero'ing, only valid if committed
#if defined(MMGC_MEMORY_PROFILER) && defined(MMGC_MEMORY_INFO)
			StackTrace *allocTrace;
			StackTrace *freeTrace;
#endif
			bool inUse() const { return prev == NULL; }
			char *endAddr() const { return baseAddr + size*kBlockSize; }
			void FreelistInit()
			{
				baseAddr     = NULL;
				size         = 0;
				sizePrevious = 0;
				prev         = this;
				next         = this;
				committed    = true;
				dirty 	    = true;
			}
		};

		// Core methods
		void AddToFreeList(HeapBlock *block);
		void AddToFreeList(HeapBlock *block, HeapBlock* pointToInsert);
		HeapBlock *AllocBlock(size_t size, bool& zero);
		void FreeBlock(HeapBlock *block);
		void FreeAll();

		void FreeInternal(const void *item, bool profile);
	
		HeapBlock *Split(HeapBlock *block, size_t size);

		// abandon a block of memory that may maps completely to the committed portion of region
		void RemoveBlock(HeapBlock *block);
		
#ifdef MMGC_MAC
		// Abandon a block of memory that may be in the middle of a
		// region.  On mac decommit is a two step process, release and
		// reserve, another thread could steal the memory between the
		// two operations so we have to be prepared to ditch a block
		// we try to decommit.  This is a horrible hack that can go
		// away if OS X fixes its mmap impl to be like windows, linux
		// and solaris (atomic decommit with VirtualFree/mmap)
		void RemovePartialBlock(HeapBlock *block);
#endif

		void Commit(HeapBlock *block);

		HeapBlock *AddrToBlock(const void *item) const;
		Region *AddrToRegion(const void *item) const;
		void RemoveRegion(Region *r);

		// debug only freelist consistency checks
#ifdef _DEBUG
		void CheckFreelist();
#else
		REALLY_INLINE void CheckFreelist()
		{
			// nothing
		}
#endif

		bool BlocksAreContiguous(void *item1, void *item2);
		
		// textual heap representation, very nice!
		void DumpHeapRep();
		
		//log a character for "count" times
		void LogChar(char c, size_t count);

		// Remove a block from a free list (inlined for speed)
		inline void RemoveFromList(HeapBlock *block)
		{
			GCAssert(!block->inUse());
			block->prev->next = block->next;
			block->next->prev = block->prev;
			block->next = block->prev = 0;
		}			

		// Map a number of blocks to the appropriate large block free list index
		// (inlined for speed)
		inline uint32_t GetFreeListIndex(size_t size)
		{
			if (size <= kUniqueThreshold) {
				return (uint32_t)size-1;
			} else if (size >= kHugeThreshold) {
				return kNumFreeLists-1;
			} else {
				return (uint32_t) ((size-kUniqueThreshold)/kFreeListCompression+kUniqueThreshold-1);
			}
		}

		bool HardLimitExceeded();
		bool SoftLimitExceeded();
		void StatusChangeNotify(MemoryStatus to);
		void CheckForStatusReturnToNormal();
		void CheckForHardLimitExceeded();
		void CheckForSoftLimitExceeded(size_t request);

		void ValidateHeapBlocks();

		void ReleaseMemory(char *address, size_t size);

 		void Enter(EnterFrame *frame);
 		void Leave();

		bool statusNotNormalOrAbort()
		{
			return status != kMemNormal && status != kMemAbort;
		}

 		size_t numHeapBlocksToNumBlocks(size_t numBlocks)
 		{
 			size_t bytes = numBlocks * sizeof(HeapBlock);
 			// round up to nearest block
 			bytes = (bytes + kBlockSize - 1) & ~(kBlockSize-1);
 			return bytes / kBlockSize;
 		}

		/**
		 * Regions are allocated from the blocks GCHeap manages
		 * similar to the HeapBlocks.  Regions can come and go so we
		 * maintain a freelist although in practice they come and go
		 * rarely we want don't want any longevity bugs
		 */
		Region *NewRegion(char *baseAddr, char *rTop, char *cTop, size_t blockId);
		void FreeRegion(Region *r);

		// data section
		static GCHeap *instance;
		static size_t leakedBytes;

		static vmpi_spin_lock_t instanceEnterLock;
		static bool instanceEnterLockInitialized;
	
		FixedMalloc fixedMalloc;
		Region *freeRegion;
		Region *nextRegion;
		HeapBlock *blocks;
		size_t blocksLen;
		size_t numDecommitted;
		size_t numRegionBlocks;
		HeapBlock freelists[kNumFreeLists];
		size_t numAlloc;
		size_t codeMemory;
		size_t externalPressure;
		vmpi_spin_lock_t m_spinlock;
		GCHeapConfig config;
		GCManager gcManager;		
 		BasicList<OOMCallback*> callbacks;
 		vmpi_spin_lock_t list_lock;

		GCThreadLocal<EnterFrame*> enterFrame;
		friend class EnterFrame;
		MemoryStatus status;
		bool statusNotificationBeingSent;
 		uint32_t enterCount;

		vmpi_spin_lock_t gclog_spinlock;	// a lock used by GC::gclog for exclusive access to GCHeap::DumpMemoryInfo

#ifdef MMGC_MEMORY_PROFILER
		static MemoryProfiler *profiler;
		bool hasSpy; //flag indicating whether profiler spy is active or not.  If active, AllocHook will call VMPI_spyCallback
#endif

		size_t maxTotalHeapSize;	// in bytes
#ifdef MMGC_POLICY_PROFILING
		size_t maxPrivateMemory;	// in bytes
#endif

#ifdef MMGC_HOOKS
		bool hooksEnabled;
#endif

		bool entryChecksEnabled;
 		bool abortStatusNotificationSent;
	};
	
	// Move the following to GCHeap-inlines.h when we have it.

	REALLY_INLINE size_t GCHeap::Size(const void *item)
	{
		MMGC_LOCK(m_spinlock);
		HeapBlock *block = AddrToBlock(item);
		return block->size;
	}
}

#endif /* __GCHeap__ */