tracy/client/tracy_rpmalloc.cpp

#ifdef TRACY_ENABLE

/* rpmalloc.c  -  Memory allocator  -  Public Domain  -  2016 Mattias Jansson / Rampant Pixels
 *
 * This library provides a cross-platform lock free thread caching malloc implementation in C11.
 * The latest source code is always available at
 *
 * https://github.com/rampantpixels/rpmalloc
 *
 * This library is put in the public domain; you can redistribute it and/or modify it without any restrictions.
 *
 */

#include "tracy_rpmalloc.hpp"

/// Build time configurable limits
#ifndef HEAP_ARRAY_SIZE
//! Size of heap hashmap
#define HEAP_ARRAY_SIZE           79
#endif
#ifndef ENABLE_THREAD_CACHE
//! Enable per-thread cache
#define ENABLE_THREAD_CACHE       1
#endif
#ifndef ENABLE_GLOBAL_CACHE
//! Enable global cache shared between all threads, requires thread cache
#define ENABLE_GLOBAL_CACHE       1
#endif
#ifndef ENABLE_VALIDATE_ARGS
//! Enable validation of args to public entry points
#define ENABLE_VALIDATE_ARGS      0
#endif
#ifndef ENABLE_STATISTICS
//! Enable statistics collection
#define ENABLE_STATISTICS         0
#endif
#ifndef ENABLE_ASSERTS
//! Enable asserts
#define ENABLE_ASSERTS            0
#endif
#ifndef ENABLE_PRELOAD
//! Support preloading
#define ENABLE_PRELOAD            0
#endif
#ifndef ENABLE_GUARDS
//! Enable overwrite/underwrite guards
#define ENABLE_GUARDS             0
#endif
#ifndef ENABLE_UNLIMITED_CACHE
//! Unlimited cache disables any cache limitations
#define ENABLE_UNLIMITED_CACHE    0
#endif
#ifndef DEFAULT_SPAN_MAP_COUNT
//! Default number of spans to map in call to map more virtual memory
#define DEFAULT_SPAN_MAP_COUNT    16
#endif
//! Minimum cache size to remain after a release to global cache
#define MIN_SPAN_CACHE_SIZE 64
//! Minimum number of spans to transfer between thread and global cache
#define MIN_SPAN_CACHE_RELEASE 16
//! Maximum cache size divisor (max cache size will be max allocation count divided by this divisor)
#define MAX_SPAN_CACHE_DIVISOR 4
//! Minimum cache size to remain after a release to global cache, large spans
#define MIN_LARGE_SPAN_CACHE_SIZE 8
//! Minimum number of spans to transfer between thread and global cache, large spans
#define MIN_LARGE_SPAN_CACHE_RELEASE 4
//! Maximum cache size divisor, large spans (max cache size will be max allocation count divided by this divisor)
#define MAX_LARGE_SPAN_CACHE_DIVISOR 16
//! Multiplier for global span cache limit (max cache size will be calculated like thread cache and multiplied with this)
#define MAX_GLOBAL_CACHE_MULTIPLIER 8

#if !ENABLE_THREAD_CACHE
#  undef ENABLE_GLOBAL_CACHE
#  define ENABLE_GLOBAL_CACHE 0
#  undef MIN_SPAN_CACHE_SIZE
#  undef MIN_SPAN_CACHE_RELEASE
#  undef MAX_SPAN_CACHE_DIVISOR
#  undef MIN_LARGE_SPAN_CACHE_SIZE
#  undef MIN_LARGE_SPAN_CACHE_RELEASE
#  undef MAX_LARGE_SPAN_CACHE_DIVISOR
#endif
#if !ENABLE_GLOBAL_CACHE
#  undef MAX_GLOBAL_CACHE_MULTIPLIER
#endif

/// Platform and arch specifics
#ifdef _MSC_VER
#  pragma warning( push )
#  pragma warning( disable : 4324 )
#  define ALIGNED_STRUCT(name, alignment) __declspec(align(alignment)) struct name
#  define FORCEINLINE __forceinline
#  define atomic_thread_fence_acquire() //_ReadWriteBarrier()
#  define atomic_thread_fence_release() //_ReadWriteBarrier()
#  if ENABLE_VALIDATE_ARGS
#    include <Intsafe.h>
#  endif
#  include <intrin.h>
#else
#  include <unistd.h>
#  if defined(__APPLE__) && ENABLE_PRELOAD
#    include <pthread.h>
#  endif
#  define ALIGNED_STRUCT(name, alignment) struct __attribute__((__aligned__(alignment))) name
#  ifdef FORCEINLINE
#    undef FORCEINLINE
#  endif
#  define FORCEINLINE inline __attribute__((__always_inline__))
#  ifdef __arm__
#    define atomic_thread_fence_acquire() __asm volatile("dmb ish" ::: "memory")
#    define atomic_thread_fence_release() __asm volatile("dmb ishst" ::: "memory")
#  else
#    define atomic_thread_fence_acquire() //__asm volatile("" ::: "memory")
#    define atomic_thread_fence_release() //__asm volatile("" ::: "memory")
#  endif
#endif

#if defined( __x86_64__ ) || defined( _M_AMD64 ) || defined( _M_X64 ) || defined( _AMD64_ ) || defined( __arm64__ ) || defined( __aarch64__ )
#  define ARCH_64BIT 1
#else
#  define ARCH_64BIT 0
#endif

#if defined( _WIN32 ) || defined( __WIN32__ ) || defined( _WIN64 )
#  define PLATFORM_WINDOWS 1
#  define PLATFORM_POSIX 0
#else
#  define PLATFORM_WINDOWS 0
#  define PLATFORM_POSIX 1
#endif

#include <stdint.h>
#include <string.h>

#include <assert.h>

#if ENABLE_GUARDS
#  define MAGIC_GUARD 0xDEADBAAD
#endif

namespace tracy
{

/// Atomic access abstraction
ALIGNED_STRUCT(atomic32_t, 4) {
	volatile int32_t nonatomic;
};
typedef struct atomic32_t atomic32_t;

ALIGNED_STRUCT(atomic64_t, 8) {
	volatile int64_t nonatomic;
};
typedef struct atomic64_t atomic64_t;

ALIGNED_STRUCT(atomicptr_t, 8) {
	volatile void* nonatomic;
};
typedef struct atomicptr_t atomicptr_t;

static FORCEINLINE int32_t
atomic_load32(atomic32_t* src) {
	return src->nonatomic;
}

static FORCEINLINE void
atomic_store32(atomic32_t* dst, int32_t val) {
	dst->nonatomic = val;
}

static FORCEINLINE int32_t
atomic_incr32(atomic32_t* val) {
#ifdef _MSC_VER
	int32_t old = (int32_t)_InterlockedExchangeAdd((volatile long*)&val->nonatomic, 1);
	return (old + 1);
#else
	return __sync_add_and_fetch(&val->nonatomic, 1);
#endif
}

static FORCEINLINE int32_t
atomic_add32(atomic32_t* val, int32_t add) {
#ifdef _MSC_VER
	int32_t old = (int32_t)_InterlockedExchangeAdd((volatile long*)&val->nonatomic, add);
	return (old + add);
#else
	return __sync_add_and_fetch(&val->nonatomic, add);
#endif
}

static FORCEINLINE void*
atomic_load_ptr(atomicptr_t* src) {
	return (void*)((uintptr_t)src->nonatomic);
}

static FORCEINLINE void
atomic_store_ptr(atomicptr_t* dst, void* val) {
	dst->nonatomic = val;
}

static FORCEINLINE int
atomic_cas_ptr(atomicptr_t* dst, void* val, void* ref) {
#ifdef _MSC_VER
#  if ARCH_64BIT
	return (_InterlockedCompareExchange64((volatile long long*)&dst->nonatomic,
	                                      (long long)val, (long long)ref) == (long long)ref) ? 1 : 0;
#  else
	return (_InterlockedCompareExchange((volatile long*)&dst->nonatomic,
	                                    (long)val, (long)ref) == (long)ref) ? 1 : 0;
#  endif
#else
	return __sync_bool_compare_and_swap(&dst->nonatomic, ref, val);
#endif
}

/// Preconfigured limits and sizes
//! Granularity of a small allocation block
#define SMALL_GRANULARITY         32
//! Small granularity shift count
#define SMALL_GRANULARITY_SHIFT   5
//! Number of small block size classes
#define SMALL_CLASS_COUNT         63
//! Maximum size of a small block
#define SMALL_SIZE_LIMIT          2016
//! Granularity of a medium allocation block
#define MEDIUM_GRANULARITY        512
//! Medium granularity shift count
#define MEDIUM_GRANULARITY_SHIFT  9
//! Number of medium block size classes
#define MEDIUM_CLASS_COUNT        60
//! Total number of small + medium size classes
#define SIZE_CLASS_COUNT          (SMALL_CLASS_COUNT + MEDIUM_CLASS_COUNT)
//! Number of large block size classes
#define LARGE_CLASS_COUNT         32
//! Maximum size of a medium block
#define MEDIUM_SIZE_LIMIT         (SMALL_SIZE_LIMIT + (MEDIUM_GRANULARITY * MEDIUM_CLASS_COUNT) - SPAN_HEADER_SIZE)
//! Maximum size of a large block
#define LARGE_SIZE_LIMIT          ((LARGE_CLASS_COUNT * _memory_span_size) - SPAN_HEADER_SIZE)
//! Size of a span header
#define SPAN_HEADER_SIZE          32

#define pointer_offset(ptr, ofs) (void*)((char*)(ptr) + (ptrdiff_t)(ofs))
#define pointer_diff(first, second) (ptrdiff_t)((const char*)(first) - (const char*)(second))

#if ARCH_64BIT
typedef int64_t offset_t;
#else
typedef int32_t offset_t;
#endif
typedef uint32_t count_t;

#if ENABLE_VALIDATE_ARGS
//! Maximum allocation size to avoid integer overflow
#undef  MAX_ALLOC_SIZE
#define MAX_ALLOC_SIZE            (((size_t)-1) - _memory_span_size)
#endif

/// Data types
//! A memory heap, per thread
typedef struct heap_t heap_t;
//! Span of memory pages
typedef struct span_t span_t;
//! Size class definition
typedef struct size_class_t size_class_t;
//! Span block bookkeeping
typedef struct span_block_t span_block_t;
//! Span list bookkeeping
typedef struct span_list_t span_list_t;
//! Span data union, usage depending on span state
typedef union span_data_t span_data_t;
//! Cache data
typedef struct span_counter_t span_counter_t;
//! Global cache
typedef struct global_cache_t global_cache_t;

//! Flag indicating span is the first (master) span of a split superspan
#define SPAN_FLAG_MASTER 1
//! Flag indicating span is a secondary (sub) span of a split superspan
#define SPAN_FLAG_SUBSPAN 2

//Alignment offset must match in both structures to keep the data when
//transitioning between being used for blocks and being part of a list
struct span_block_t {
	//! Free list
	uint16_t    free_list;
	//! First autolinked block
	uint16_t    first_autolink;
	//! Free count
	uint16_t    free_count;
	//! Alignment offset
	uint16_t    align_offset;
};

struct span_list_t {
	//! List size
	uint32_t    size;
	//! Unused in lists
	uint16_t    unused;
	//! Alignment offset
	uint16_t    align_offset;
};

union span_data_t {
	//! Span data when used as blocks
	span_block_t block;
	//! Span data when used in lists
	span_list_t list;
};

//A span can either represent a single span of memory pages with size declared by span_map_count configuration variable,
//or a set of spans in a continuous region, a super span. Any reference to the term "span" usually refers to both a single
//span or a super span. A super span can further be diviced into multiple spans (or this, super spans), where the first
//(super)span is the master and subsequent (super)spans are subspans. The master span keeps track of how many subspans
//that are still alive and mapped in virtual memory, and once all subspans and master have been unmapped the entire
//superspan region is released and unmapped (on Windows for example, the entire superspan range has to be released
//in the same call to release the virtual memory range, but individual subranges can be decommitted individually
//to reduce physical memory use).
struct span_t {
	//!	Heap ID
	atomic32_t  heap_id;
	//! Size class
	uint16_t    size_class;
	// TODO: If we could store remainder part of flags as an atomic counter, the entire check
	//       if master is owned by calling heap could be simplified to an atomic dec from any thread
	//       since remainder of a split super span only ever decreases, never increases
	//! Flags and counters
	uint16_t    flags;
	//! Span data
	span_data_t data;
	//! Next span
	span_t*     next_span;
	//! Previous span
	span_t*     prev_span;
};
static_assert(sizeof(span_t) <= SPAN_HEADER_SIZE, "span size mismatch");

//Adaptive cache counter of a single superspan span count
struct span_counter_t {
	//! Allocation high water mark
	uint32_t  max_allocations;
	//! Current number of allocations
	uint32_t  current_allocations;
	//! Cache limit
	uint32_t  cache_limit;
};

struct heap_t {
	//! Heap ID
	int32_t      id;
	//! Free count for each size class active span
	span_block_t active_block[SIZE_CLASS_COUNT];
	//! Active span for each size class
	span_t*      active_span[SIZE_CLASS_COUNT];
	//! List of semi-used spans with free blocks for each size class (double linked list)
	span_t*      size_cache[SIZE_CLASS_COUNT];
#if ENABLE_THREAD_CACHE
	//! List of free spans (single linked list)
	span_t*      span_cache[LARGE_CLASS_COUNT];
	//! Allocation counters
	span_counter_t span_counter[LARGE_CLASS_COUNT];
#endif
	//! Mapped but unused spans
	span_t*      span_reserve;
	//! Master span for mapped but unused spans
	span_t*      span_reserve_master;
	//! Number of mapped but unused spans
	size_t       spans_reserved;
	//! Deferred deallocation
	atomicptr_t  defer_deallocate;
	//! Deferred unmaps
	atomicptr_t  defer_unmap;
	//! Next heap in id list
	heap_t*      next_heap;
	//! Next heap in orphan list
	heap_t*      next_orphan;
	//! Memory pages alignment offset
	size_t       align_offset;
#if ENABLE_STATISTICS
	//! Number of bytes transitioned thread -> global
	size_t       thread_to_global;
	//! Number of bytes transitioned global -> thread
	size_t       global_to_thread;
#endif
};

struct size_class_t {
	//! Size of blocks in this class
	uint32_t size;
	//! Number of blocks in each chunk
	uint16_t block_count;
	//! Class index this class is merged with
	uint16_t class_idx;
};
static_assert(sizeof(size_class_t) == 8, "Size class size mismatch");

struct global_cache_t {
	//! Cache list pointer
	atomicptr_t cache;
	//! Cache size
	atomic32_t size;
	//! ABA counter
	atomic32_t counter;
};

/// Global data
//! Configuration
static rpmalloc_config_t _memory_config;
//! Memory page size
static size_t _memory_page_size;
//! Shift to divide by page size
static size_t _memory_page_size_shift;
//! Mask to get to start of a memory page
static size_t _memory_page_mask;
//! Granularity at which memory pages are mapped by OS
static size_t _memory_map_granularity;
//! Size of a span of memory pages
static size_t _memory_span_size;
//! Shift to divide by span size
static size_t _memory_span_size_shift;
//! Mask to get to start of a memory span
static uintptr_t _memory_span_mask;
//! Global size classes
static size_class_t _memory_size_class[SIZE_CLASS_COUNT];
//! Run-time size limit of medium blocks
static size_t _memory_medium_size_limit;
//! Heap ID counter
static atomic32_t _memory_heap_id;
#if ENABLE_THREAD_CACHE
//! Adaptive cache max allocation count
static uint32_t _memory_max_allocation[LARGE_CLASS_COUNT];
#endif
#if ENABLE_GLOBAL_CACHE
//! Global span cache
static global_cache_t _memory_span_cache[LARGE_CLASS_COUNT];
#endif
//! All heaps
static atomicptr_t _memory_heaps[HEAP_ARRAY_SIZE];
//! Orphaned heaps
static atomicptr_t _memory_orphan_heaps;
//! Running orphan counter to avoid ABA issues in linked list
static atomic32_t _memory_orphan_counter;
//! Active heap count
static atomic32_t _memory_active_heaps;
#if ENABLE_STATISTICS
//! Total number of currently mapped memory pages
static atomic32_t _mapped_pages;
//! Total number of currently lost spans
static atomic32_t _reserved_spans;
//! Running counter of total number of mapped memory pages since start
static atomic32_t _mapped_total;
//! Running counter of total number of unmapped memory pages since start
static atomic32_t _unmapped_total;
#endif

#define MEMORY_UNUSED(x) (void)sizeof((x))

//! Current thread heap
#if defined(__APPLE__) && ENABLE_PRELOAD
static pthread_key_t _memory_thread_heap;
#else
#  ifdef _MSC_VER
#    define _Thread_local __declspec(thread)
#    define TLS_MODEL
#  else
#    define TLS_MODEL __attribute__((tls_model("initial-exec")))
#    if !defined(__clang__) && defined(__GNUC__)
#      define _Thread_local __thread
#    endif
#  endif
static _Thread_local heap_t* _memory_thread_heap TLS_MODEL;
#endif

//! Get the current thread heap
static FORCEINLINE heap_t*
get_thread_heap(void) {
#if defined(__APPLE__) && ENABLE_PRELOAD
	return pthread_getspecific(_memory_thread_heap);
#else
	return _memory_thread_heap;
#endif
}

//! Set the current thread heap
static void
set_thread_heap(heap_t* heap) {
#if defined(__APPLE__) && ENABLE_PRELOAD
	pthread_setspecific(_memory_thread_heap, heap);
#else
	_memory_thread_heap = heap;
#endif
}

//! Default implementation to map more virtual memory
static void*
_memory_map_os(size_t size, size_t* offset);

//! Default implementation to unmap virtual memory
static void
_memory_unmap_os(void* address, size_t size, size_t offset, int release);

//! Deallocate any deferred blocks and check for the given size class
static int
_memory_deallocate_deferred(heap_t* heap, size_t size_class);

//! Lookup a memory heap from heap ID
static heap_t*
_memory_heap_lookup(int32_t id) {
	uint32_t list_idx = id % HEAP_ARRAY_SIZE;
	heap_t* heap = (heap_t*)atomic_load_ptr(&_memory_heaps[list_idx]);
	while (heap && (heap->id != id))
		heap = heap->next_heap;
	return heap;
}

#if ENABLE_THREAD_CACHE

//! Increase an allocation counter
static void
_memory_counter_increase(span_counter_t* counter, uint32_t* global_counter, size_t span_count) {
	if (++counter->current_allocations > counter->max_allocations) {
		counter->max_allocations = counter->current_allocations;
		const uint32_t cache_limit_max = (uint32_t)_memory_span_size - 2;
#if !ENABLE_UNLIMITED_CACHE
		counter->cache_limit = counter->max_allocations / ((span_count == 1) ? MAX_SPAN_CACHE_DIVISOR : MAX_LARGE_SPAN_CACHE_DIVISOR);
		const uint32_t cache_limit_min = (span_count == 1) ? (MIN_SPAN_CACHE_RELEASE + MIN_SPAN_CACHE_SIZE) : (MIN_LARGE_SPAN_CACHE_RELEASE + MIN_LARGE_SPAN_CACHE_SIZE);
		if (counter->cache_limit < cache_limit_min)
			counter->cache_limit = cache_limit_min;
		if (counter->cache_limit > cache_limit_max)
			counter->cache_limit = cache_limit_max;
#else
		counter->cache_limit = cache_limit_max;
#endif
		if (counter->max_allocations > *global_counter)
			*global_counter = counter->max_allocations;
	}
}

#else
#  define _memory_counter_increase(counter, global_counter, span_count) do {} while (0)
#endif

#if ENABLE_STATISTICS
#  define _memory_statistics_add(atomic_counter, value) atomic_add32(atomic_counter, (int32_t)(value))
#  define _memory_statistics_sub(atomic_counter, value) atomic_add32(atomic_counter, -(int32_t)(value))
#else
#  define _memory_statistics_add(atomic_counter, value) do {} while(0)
#  define _memory_statistics_sub(atomic_counter, value) do {} while(0)
#endif

//! Map more virtual memory
static void*
_memory_map(size_t size, size_t* offset) {
	assert(!(size % _memory_page_size));
	_memory_statistics_add(&_mapped_pages, (size >> _memory_page_size_shift));
	_memory_statistics_add(&_mapped_total, (size >> _memory_page_size_shift));
	return _memory_config.memory_map(size, offset);
}

//! Unmap virtual memory
static void
_memory_unmap(void* address, size_t size, size_t offset, int release) {
	assert((size < _memory_span_size) || !((uintptr_t)address & ~_memory_span_mask));
	assert(!(size % _memory_page_size));
	_memory_statistics_sub(&_mapped_pages, (size >> _memory_page_size_shift));
	_memory_statistics_add(&_unmapped_total, (size >> _memory_page_size_shift));
	_memory_config.memory_unmap(address, size, offset, release);
}

//! Make flags field in a span from flags, remainder/distance and count
#define SPAN_MAKE_FLAGS(flags, remdist, count) ((uint16_t)((flags) | ((uint16_t)((remdist) - 1) << 2) | ((uint16_t)((count) - 1) << 9))); assert((flags) < 4); assert((remdist) && (remdist) < 128); assert((count) && (count) < 128)
//! Check if span has any of the given flags
#define SPAN_HAS_FLAG(flags, flag) ((flags) & (flag))
//! Get the distance from flags field
#define SPAN_DISTANCE(flags) (1 + (((flags) >> 2) & 0x7f))
//! Get the remainder from flags field
#define SPAN_REMAINS(flags) (1 + (((flags) >> 2) & 0x7f))
//! Get the count from flags field
#define SPAN_COUNT(flags) (1 + (((flags) >> 9) & 0x7f))
//! Set the remainder in the flags field (MUST be done from the owner heap thread)
#define SPAN_SET_REMAINS(flags, remains) flags = ((uint16_t)(((flags) & 0xfe03) | ((uint16_t)((remains) - 1) << 2))); assert((remains) < 128)

//! Resize the given super span to the given count of spans, store the remainder in the heap reserved spans fields
static void
_memory_set_span_remainder_as_reserved(heap_t* heap, span_t* span, size_t use_count) {
	size_t current_count = SPAN_COUNT(span->flags);

	assert(!SPAN_HAS_FLAG(span->flags, SPAN_FLAG_MASTER) || !SPAN_HAS_FLAG(span->flags, SPAN_FLAG_SUBSPAN));
	assert((current_count > 1) && (current_count < 127));
	assert(!heap->spans_reserved);
	assert((size_t)SPAN_COUNT(span->flags) == current_count);
	assert(current_count > use_count);

	heap->span_reserve = (span_t*)pointer_offset(span, use_count * _memory_span_size);
	heap->spans_reserved = current_count - use_count;
	if (!SPAN_HAS_FLAG(span->flags, SPAN_FLAG_MASTER | SPAN_FLAG_SUBSPAN)) {
		//We must store the heap id before setting as master, to force unmaps to defer to this heap thread
		atomic_store32(&span->heap_id, heap->id);
		atomic_thread_fence_release();
		heap->span_reserve_master = span;
		span->flags = SPAN_MAKE_FLAGS(SPAN_FLAG_MASTER, current_count, use_count);
		_memory_statistics_add(&_reserved_spans, current_count);
	}
	else if (SPAN_HAS_FLAG(span->flags, SPAN_FLAG_MASTER)) {
		//Only owner heap thread can modify a master span
		assert(atomic_load32(&span->heap_id) == heap->id);
		uint16_t remains = SPAN_REMAINS(span->flags);
		assert(remains >= current_count);
		heap->span_reserve_master = span;
		span->flags = SPAN_MAKE_FLAGS(SPAN_FLAG_MASTER, remains, use_count);
	}
	else { //SPAN_FLAG_SUBSPAN
		//Resizing a subspan is a safe operation in any thread
		uint16_t distance = SPAN_DISTANCE(span->flags);
		span_t* master = (span_t*)pointer_offset(span, -(int)distance * (int)_memory_span_size);
		heap->span_reserve_master = master;
		assert(SPAN_HAS_FLAG(master->flags, SPAN_FLAG_MASTER));
		assert((size_t)SPAN_REMAINS(master->flags) >= current_count);
		span->flags = SPAN_MAKE_FLAGS(SPAN_FLAG_SUBSPAN, distance, use_count);
	}
	assert((SPAN_COUNT(span->flags) + heap->spans_reserved) == current_count);
}

//! Map in memory pages for the given number of spans (or use previously reserved pages)
static span_t*
_memory_map_spans(heap_t* heap, size_t span_count) {
	if (span_count <= heap->spans_reserved) {
		span_t* span = heap->span_reserve;
		heap->span_reserve = (span_t*)pointer_offset(span, span_count * _memory_span_size);
		heap->spans_reserved -= span_count;
		//Declare the span to be a subspan with given distance from master span
		uint16_t distance = (uint16_t)((uintptr_t)pointer_diff(span, heap->span_reserve_master) >> _memory_span_size_shift);
		span->flags = SPAN_MAKE_FLAGS(SPAN_FLAG_SUBSPAN, distance, span_count);
		span->data.block.align_offset = 0;
		return span;
	}

	//We cannot request extra spans if we already have some (but not enough) pending reserved spans
	size_t request_spans = (heap->spans_reserved || (span_count > _memory_config.span_map_count)) ? span_count : _memory_config.span_map_count;
	size_t align_offset = 0;
	span_t* span = (span_t*)_memory_map(request_spans * _memory_span_size, &align_offset);
	span->flags = SPAN_MAKE_FLAGS(0, request_spans, request_spans);
	span->data.block.align_offset = (uint16_t)align_offset;
	if (request_spans > span_count) {
		//We have extra spans, store them as reserved spans in heap
		_memory_set_span_remainder_as_reserved(heap, span, span_count);
	}
	return span;
}

//! Defer unmapping of the given span to the owner heap
static int
_memory_unmap_defer(int32_t heap_id, span_t* span) {
	//Get the heap and link in pointer in list of deferred operations
	heap_t* heap = _memory_heap_lookup(heap_id);
	if (!heap)
		return 0;
	atomic_store32(&span->heap_id, heap_id);
	void* last_ptr;
	do {
		last_ptr = atomic_load_ptr(&heap->defer_unmap);
		span->next_span = (span_t*)last_ptr;
	} while (!atomic_cas_ptr(&heap->defer_unmap, span, last_ptr));
	return 1;
}

//! Unmap memory pages for the given number of spans (or mark as unused if no partial unmappings)
static void
_memory_unmap_span(heap_t* heap, span_t* span) {
	size_t span_count = SPAN_COUNT(span->flags);
	assert(!SPAN_HAS_FLAG(span->flags, SPAN_FLAG_MASTER) || !SPAN_HAS_FLAG(span->flags, SPAN_FLAG_SUBSPAN));
	//A plain run of spans can be unmapped directly
	if (!SPAN_HAS_FLAG(span->flags, SPAN_FLAG_MASTER | SPAN_FLAG_SUBSPAN)) {
		_memory_unmap(span, span_count * _memory_span_size, span->data.list.align_offset, 1);
		return;
	}

	uint32_t is_master = SPAN_HAS_FLAG(span->flags, SPAN_FLAG_MASTER);
	span_t* master = is_master ? span : (span_t*)(pointer_offset(span, -(int)SPAN_DISTANCE(span->flags) * (int)_memory_span_size));

	assert(is_master || SPAN_HAS_FLAG(span->flags, SPAN_FLAG_SUBSPAN));
	assert(SPAN_HAS_FLAG(master->flags, SPAN_FLAG_MASTER));

	//Check if we own the master span, otherwise defer (only owner of master span can modify remainder field)
	int32_t master_heap_id = atomic_load32(&master->heap_id);
	if (heap && (master_heap_id != heap->id)) {
		if (_memory_unmap_defer(master_heap_id, span))
			return;
	}
	if (!is_master) {
		//Directly unmap subspans
		assert(span->data.list.align_offset == 0);
		_memory_unmap(span, span_count * _memory_span_size, 0, 0);
		_memory_statistics_sub(&_reserved_spans, span_count);
	}
	else {
		//Special double flag to denote an unmapped master
		//It must be kept in memory since span header must be used
		span->flags |= SPAN_FLAG_MASTER | SPAN_FLAG_SUBSPAN;
	}
	//We are in owner thread of the master span
	uint32_t remains = SPAN_REMAINS(master->flags);
	assert(remains >= span_count);
	remains = ((uint32_t)span_count >= remains) ? 0 : (remains - (uint32_t)span_count);
	if (!remains) {
		//Everything unmapped, unmap the master span with release flag to unmap the entire range of the super span
		assert(SPAN_HAS_FLAG(master->flags, SPAN_FLAG_MASTER) && SPAN_HAS_FLAG(master->flags, SPAN_FLAG_SUBSPAN));
		span_count = SPAN_COUNT(master->flags);
		_memory_unmap(master, span_count * _memory_span_size, master->data.list.align_offset, 1);
		_memory_statistics_sub(&_reserved_spans, span_count);
	}
	else {
		//Set remaining spans
		SPAN_SET_REMAINS(master->flags, remains);
	}
}

//! Process pending deferred cross-thread unmaps
static span_t*
_memory_unmap_deferred(heap_t* heap, size_t wanted_count) {
	//Grab the current list of deferred unmaps
	atomic_thread_fence_acquire();
	span_t* span = (span_t*)atomic_load_ptr(&heap->defer_unmap);
	if (!span || !atomic_cas_ptr(&heap->defer_unmap, 0, span))
		return 0;
	span_t* found_span = 0;
	do {
		//Verify that we own the master span, otherwise re-defer to owner
		void* next = span->next_span;
		size_t span_count = SPAN_COUNT(span->flags);
		if (!found_span && span_count == wanted_count) {
			assert(!SPAN_HAS_FLAG(span->flags, SPAN_FLAG_MASTER) || !SPAN_HAS_FLAG(span->flags, SPAN_FLAG_SUBSPAN));
			found_span = span;
		}
		else {
			uint32_t is_master = SPAN_HAS_FLAG(span->flags, SPAN_FLAG_MASTER);
			span_t* master = is_master ? span : (span_t*)(pointer_offset(span, -(int)SPAN_DISTANCE(span->flags) * (int)_memory_span_size));
			int32_t master_heap_id = atomic_load32(&master->heap_id);
			if ((atomic_load32(&span->heap_id) == master_heap_id) ||
			        !_memory_unmap_defer(master_heap_id, span)) {
				//We own the master span (or heap merged and abandoned)
				_memory_unmap_span(heap, span);
			}
		}
		span = (span_t*)next;
	} while (span);
	return found_span;
}

//! Unmap a single linked list of spans
static void
_memory_unmap_span_list(heap_t* heap, span_t* span) {
	size_t list_size = span->data.list.size;
	for (size_t ispan = 0; ispan < list_size; ++ispan) {
		span_t* next_span = span->next_span;
		_memory_unmap_span(heap, span);
		span = next_span;
	}
	assert(!span);
}

#if ENABLE_THREAD_CACHE

//! Split a super span in two
static span_t*
_memory_span_split(heap_t* heap, span_t* span, size_t use_count) {
	uint16_t distance = 0;
	size_t current_count = SPAN_COUNT(span->flags);
	assert(current_count > use_count);
	assert(!SPAN_HAS_FLAG(span->flags, SPAN_FLAG_MASTER) || !SPAN_HAS_FLAG(span->flags, SPAN_FLAG_SUBSPAN));
	if (!SPAN_HAS_FLAG(span->flags, SPAN_FLAG_MASTER | SPAN_FLAG_SUBSPAN)) {
		//Must store heap in master span before use, to avoid issues when unmapping subspans
		atomic_store32(&span->heap_id, heap->id);
		atomic_thread_fence_release();
		span->flags = SPAN_MAKE_FLAGS(SPAN_FLAG_MASTER, current_count, use_count);
		_memory_statistics_add(&_reserved_spans, current_count);
	}
	else if (SPAN_HAS_FLAG(span->flags, SPAN_FLAG_MASTER)) {
		//Only valid to call on master span if we own it
		assert(atomic_load32(&span->heap_id) == heap->id);
		uint16_t remains = SPAN_REMAINS(span->flags);
		assert(remains >= current_count);
		span->flags = SPAN_MAKE_FLAGS(SPAN_FLAG_MASTER, remains, use_count);
	}
	else { //SPAN_FLAG_SUBSPAN
		distance = SPAN_DISTANCE(span->flags);
		span->flags = SPAN_MAKE_FLAGS(SPAN_FLAG_SUBSPAN, distance, use_count);
	}
	//Setup remainder as a subspan
	span_t* subspan = (span_t*)pointer_offset(span, use_count * _memory_span_size);
	subspan->flags = SPAN_MAKE_FLAGS(SPAN_FLAG_SUBSPAN, distance + use_count, current_count - use_count);
	subspan->data.list.align_offset = 0;
	return subspan;
}

//! Add span to head of single linked span list
static size_t
_memory_span_list_push(span_t** head, span_t* span) {
	span->next_span = *head;
	if (*head)
		span->data.list.size = (*head)->data.list.size + 1;
	else
		span->data.list.size = 1;
	*head = span;
	return span->data.list.size;
}

//! Remove span from head of single linked span list, returns the new list head
static span_t*
_memory_span_list_pop(span_t** head) {
	span_t* span = *head;
	span_t* next_span = 0;
	if (span->data.list.size > 1) {
		next_span = span->next_span;
		assert(next_span);
		next_span->data.list.size = span->data.list.size - 1;
	}
	*head = next_span;
	return span;
}

//! Split a single linked span list
static span_t*
_memory_span_list_split(span_t* span, size_t limit) {
	span_t* next = 0;
	if (limit < 2)
		limit = 2;
	if (span->data.list.size > limit) {
		count_t list_size = 1;
		span_t* last = span;
		next = span->next_span;
		while (list_size < limit) {
			last = next;
			next = next->next_span;
			++list_size;
		}
		last->next_span = 0;
		assert(next);
		next->data.list.size = span->data.list.size - list_size;
		span->data.list.size = list_size;
		span->prev_span = 0;
	}
	return next;
}

#endif

//! Add a span to a double linked list
static void
_memory_span_list_doublelink_add(span_t** head, span_t* span) {
	if (*head) {
		(*head)->prev_span = span;
		span->next_span = *head;
	}
	else {
		span->next_span = 0;
	}
	*head = span;
}

//! Remove a span from a double linked list
static void
_memory_span_list_doublelink_remove(span_t** head, span_t* span) {
	if (*head == span) {
		*head = span->next_span;
	}
	else {
		span_t* next_span = span->next_span;
		span_t* prev_span = span->prev_span;
		if (next_span)
			next_span->prev_span = prev_span;
		prev_span->next_span = next_span;
	}
}

#if ENABLE_GLOBAL_CACHE

//! Insert the given list of memory page spans in the global cache
static void
_memory_cache_insert(heap_t* heap, global_cache_t* cache, span_t* span, size_t cache_limit) {
	assert((span->data.list.size == 1) || (span->next_span != 0));
	int32_t list_size = (int32_t)span->data.list.size;
	//Unmap if cache has reached the limit
	if (atomic_add32(&cache->size, list_size) > (int32_t)cache_limit) {
		_memory_unmap_span_list(heap, span);
		atomic_add32(&cache->size, -list_size);
		return;
	}
	void* current_cache, *new_cache;
	do {
		current_cache = atomic_load_ptr(&cache->cache);
		span->prev_span = (span_t*)(void*)((uintptr_t)current_cache & _memory_span_mask);
		new_cache = (void*)((uintptr_t)span | ((uintptr_t)atomic_incr32(&cache->counter) & ~_memory_span_mask));
	} while (!atomic_cas_ptr(&cache->cache, new_cache, current_cache));
}

//! Extract a number of memory page spans from the global cache
static span_t*
_memory_cache_extract(global_cache_t* cache) {
	uintptr_t span_ptr;
	do {
		void* global_span = atomic_load_ptr(&cache->cache);
		span_ptr = (uintptr_t)global_span & _memory_span_mask;
		if (span_ptr) {
			span_t* span = (span_t*)(void*)span_ptr;
			//By accessing the span ptr before it is swapped out of list we assume that a contending thread
			//does not manage to traverse the span to being unmapped before we access it
			void* new_cache = (void*)((uintptr_t)span->prev_span | ((uintptr_t)atomic_incr32(&cache->counter) & ~_memory_span_mask));
			if (atomic_cas_ptr(&cache->cache, new_cache, global_span)) {
				atomic_add32(&cache->size, -(int32_t)span->data.list.size);
				return span;
			}
		}
	} while (span_ptr);
	return 0;
}

//! Finalize a global cache, only valid from allocator finalization (not thread safe)
static void
_memory_cache_finalize(global_cache_t* cache) {
	void* current_cache = atomic_load_ptr(&cache->cache);
	span_t* span = (span_t*)(void*)((uintptr_t)current_cache & _memory_span_mask);
	while (span) {
		span_t* skip_span = (span_t*)(void*)((uintptr_t)span->prev_span & _memory_span_mask);
		atomic_add32(&cache->size, -(int32_t)span->data.list.size);
		_memory_unmap_span_list(0, span);
		span = skip_span;
	}
	assert(!atomic_load32(&cache->size));
	atomic_store_ptr(&cache->cache, 0);
	atomic_store32(&cache->size, 0);
}

//! Insert the given list of memory page spans in the global cache
static void
_memory_global_cache_insert(heap_t* heap, span_t* span) {
	//Calculate adaptive limits
	size_t span_count = SPAN_COUNT(span->flags);
	const size_t cache_divisor = (span_count == 1) ? MAX_SPAN_CACHE_DIVISOR : (MAX_LARGE_SPAN_CACHE_DIVISOR * span_count * 2);
	const size_t cache_limit = (MAX_GLOBAL_CACHE_MULTIPLIER * _memory_max_allocation[span_count - 1]) / cache_divisor;
	const size_t cache_limit_min = MAX_GLOBAL_CACHE_MULTIPLIER * (span_count == 1 ? MIN_SPAN_CACHE_SIZE : MIN_LARGE_SPAN_CACHE_SIZE);
	_memory_cache_insert(heap, &_memory_span_cache[span_count - 1], span, cache_limit > cache_limit_min ? cache_limit : cache_limit_min);
}

//! Extract a number of memory page spans from the global cache for large blocks
static span_t*
_memory_global_cache_extract(size_t span_count) {
	span_t* span = _memory_cache_extract(&_memory_span_cache[span_count - 1]);
	assert(!span || ((size_t)SPAN_COUNT(span->flags) == span_count));
	return span;
}

#endif

//! Insert a single span into thread heap cache, releasing to global cache if overflow
static void
_memory_heap_cache_insert(heap_t* heap, span_t* span) {
#if ENABLE_THREAD_CACHE
	size_t span_count = SPAN_COUNT(span->flags);
	size_t idx = span_count - 1;
	if (_memory_span_list_push(&heap->span_cache[idx], span) <= heap->span_counter[idx].cache_limit)
		return;
	heap->span_cache[idx] = _memory_span_list_split(span, heap->span_counter[idx].cache_limit);
	assert(span->data.list.size == heap->span_counter[idx].cache_limit);
#if ENABLE_STATISTICS
	heap->thread_to_global += (size_t)span->data.list.size * span_count * _memory_span_size;
#endif
#if ENABLE_GLOBAL_CACHE
	_memory_global_cache_insert(heap, span);
#else
	_memory_unmap_span_list(heap, span);
#endif
#else
	_memory_unmap_span(heap, span);
#endif
}

//! Extract the given number of spans from the different cache levels
static span_t*
_memory_heap_cache_extract(heap_t* heap, size_t span_count) {
#if ENABLE_THREAD_CACHE
	size_t idx = span_count - 1;
	//Step 1: check thread cache
	if (heap->span_cache[idx])
		return _memory_span_list_pop(&heap->span_cache[idx]);
#endif
	//Step 2: Check reserved spans
	if (heap->spans_reserved >= span_count)
		return _memory_map_spans(heap, span_count);
	//Step 3: Try processing deferred unmappings
	span_t* span = _memory_unmap_deferred(heap, span_count);
	if (span)
		return span;
#if ENABLE_THREAD_CACHE
	//Step 4: Check larger super spans and split if we find one
	for (++idx; idx < LARGE_CLASS_COUNT; ++idx) {
		if (heap->span_cache[idx]) {
			span = _memory_span_list_pop(&heap->span_cache[idx]);
			break;
		}
	}
	if (span) {
		//Mark the span as owned by this heap before splitting
		size_t got_count = SPAN_COUNT(span->flags);
		assert(got_count > span_count);
		atomic_store32(&span->heap_id, heap->id);
		atomic_thread_fence_release();

		//Split the span and store as reserved if no previously reserved spans, or in thread cache otherwise
		span_t* subspan = _memory_span_split(heap, span, span_count);
		assert((size_t)(SPAN_COUNT(span->flags) + SPAN_COUNT(subspan->flags)) == got_count);
		assert((size_t)SPAN_COUNT(span->flags) == span_count);
		if (!heap->spans_reserved) {
			heap->spans_reserved = got_count - span_count;
			heap->span_reserve = subspan;
			heap->span_reserve_master = (span_t*)pointer_offset(subspan, -(int32_t)SPAN_DISTANCE(subspan->flags) * (int32_t)_memory_span_size);
		}
		else {
			_memory_heap_cache_insert(heap, subspan);
		}
		return span;
	}
#if ENABLE_GLOBAL_CACHE
	//Step 5: Extract from global cache
	idx = span_count - 1;
	heap->span_cache[idx] = _memory_global_cache_extract(span_count);
	if (heap->span_cache[idx]) {
#if ENABLE_STATISTICS
		heap->global_to_thread += (size_t)heap->span_cache[idx]->data.list.size * span_count * _memory_span_size;
#endif
		return _memory_span_list_pop(&heap->span_cache[idx]);
	}
#endif
#endif
	return 0;
}

//! Allocate a small/medium sized memory block from the given heap
static void*
_memory_allocate_from_heap(heap_t* heap, size_t size) {
	//Calculate the size class index and do a dependent lookup of the final class index (in case of merged classes)
	const size_t base_idx = (size <= SMALL_SIZE_LIMIT) ?
	                        ((size + (SMALL_GRANULARITY - 1)) >> SMALL_GRANULARITY_SHIFT) :
	                        SMALL_CLASS_COUNT + ((size - SMALL_SIZE_LIMIT + (MEDIUM_GRANULARITY - 1)) >> MEDIUM_GRANULARITY_SHIFT);
	assert(!base_idx || ((base_idx - 1) < SIZE_CLASS_COUNT));
	const size_t class_idx = _memory_size_class[base_idx ? (base_idx - 1) : 0].class_idx;

	span_block_t* active_block = heap->active_block + class_idx;
	size_class_t* size_class = _memory_size_class + class_idx;
	const count_t class_size = size_class->size;

	//Step 1: Try to get a block from the currently active span. The span block bookkeeping
	//        data for the active span is stored in the heap for faster access
use_active:
	if (active_block->free_count) {
		//Happy path, we have a span with at least one free block
		span_t* span = heap->active_span[class_idx];
		count_t offset = class_size * active_block->free_list;
		uint32_t* block = (uint32_t*)pointer_offset(span, SPAN_HEADER_SIZE + offset);
		assert(span);

		--active_block->free_count;
		if (!active_block->free_count) {
			//Span is now completely allocated, set the bookkeeping data in the
			//span itself and reset the active span pointer in the heap
			span->data.block.free_count = 0;
			span->data.block.first_autolink = (uint16_t)size_class->block_count;
			heap->active_span[class_idx] = 0;
		}
		else {
			//Get the next free block, either from linked list or from auto link
			if (active_block->free_list < active_block->first_autolink) {
				active_block->free_list = (uint16_t)(*block);
			}
			else {
				++active_block->free_list;
				++active_block->first_autolink;
			}
			assert(active_block->free_list < size_class->block_count);
		}

		return block;
	}

	//Step 2: No active span, try executing deferred deallocations and try again if there
	//        was at least one of the requested size class
	if (_memory_deallocate_deferred(heap, class_idx)) {
		if (active_block->free_count)
			goto use_active;
	}

	//Step 3: Check if there is a semi-used span of the requested size class available
	if (heap->size_cache[class_idx]) {
		//Promote a pending semi-used span to be active, storing bookkeeping data in
		//the heap structure for faster access
		span_t* span = heap->size_cache[class_idx];
		*active_block = span->data.block;
		assert(active_block->free_count > 0);
		heap->size_cache[class_idx] = span->next_span;
		heap->active_span[class_idx] = span;

		//Mark span as owned by this heap
		atomic_store32(&span->heap_id, heap->id);
		atomic_thread_fence_release();

		goto use_active;
	}

	//Step 4: Find a span in one of the cache levels
	span_t* span = _memory_heap_cache_extract(heap, 1);
	if (!span) {
		//Step 5: Map in more virtual memory
		span = _memory_map_spans(heap, 1);
	}

	//Mark span as owned by this heap and set base data
	assert(SPAN_COUNT(span->flags) == 1);
	span->size_class = (uint16_t)class_idx;
	atomic_store32(&span->heap_id, heap->id);
	atomic_thread_fence_release();

	//If we only have one block we will grab it, otherwise
	//set span as new span to use for next allocation
	if (size_class->block_count > 1) {
		//Reset block order to sequential auto linked order
		active_block->free_count = (uint16_t)(size_class->block_count - 1);
		active_block->free_list = 1;
		active_block->first_autolink = 1;
		heap->active_span[class_idx] = span;
	}
	else {
		span->data.block.free_count = 0;
		span->data.block.first_autolink = (uint16_t)size_class->block_count;
	}

	//Track counters
	_memory_counter_increase(&heap->span_counter[0], &_memory_max_allocation[0], 1);

	//Return first block if memory page span
	return pointer_offset(span, SPAN_HEADER_SIZE);
}

//! Allocate a large sized memory block from the given heap
static void*
_memory_allocate_large_from_heap(heap_t* heap, size_t size) {
	//Calculate number of needed max sized spans (including header)
	//Since this function is never called if size > LARGE_SIZE_LIMIT
	//the span_count is guaranteed to be <= LARGE_CLASS_COUNT
	size += SPAN_HEADER_SIZE;
	size_t span_count = size >> _memory_span_size_shift;
	if (size & (_memory_span_size - 1))
		++span_count;
	size_t idx = span_count - 1;

#if ENABLE_THREAD_CACHE
	if (!heap->span_cache[idx])
		_memory_deallocate_deferred(heap, SIZE_CLASS_COUNT + idx);
#else
	_memory_deallocate_deferred(heap, SIZE_CLASS_COUNT + idx);
#endif
	//Step 1: Find span in one of the cache levels
	span_t* span = _memory_heap_cache_extract(heap, span_count);
	if (!span) {
		//Step 2: Map in more virtual memory
		span = _memory_map_spans(heap, span_count);
	}

	//Mark span as owned by this heap and set base data
	assert((size_t)SPAN_COUNT(span->flags) == span_count);
	span->size_class = (uint16_t)(SIZE_CLASS_COUNT + idx);
	atomic_store32(&span->heap_id, heap->id);
	atomic_thread_fence_release();

	//Increase counter
	_memory_counter_increase(&heap->span_counter[idx], &_memory_max_allocation[idx], span_count);

	return pointer_offset(span, SPAN_HEADER_SIZE);
}

//! Allocate a new heap
static heap_t*
_memory_allocate_heap(void) {
	void* raw_heap;
	void* next_raw_heap;
	uintptr_t orphan_counter;
	heap_t* heap;
	heap_t* next_heap;
	//Try getting an orphaned heap
	atomic_thread_fence_acquire();
	do {
		raw_heap = atomic_load_ptr(&_memory_orphan_heaps);
		heap = (heap_t*)(void*)((uintptr_t)raw_heap & _memory_page_mask);
		if (!heap)
			break;
		next_heap = heap->next_orphan;
		orphan_counter = (uintptr_t)atomic_incr32(&_memory_orphan_counter);
		next_raw_heap = (void*)((uintptr_t)next_heap | (orphan_counter & ~_memory_page_mask));
	}
	while (!atomic_cas_ptr(&_memory_orphan_heaps, next_raw_heap, raw_heap));

	if (!heap) {
		//Map in pages for a new heap
		size_t align_offset = 0;
		heap = (heap_t*)_memory_map((1 + (sizeof(heap_t) >> _memory_page_size_shift)) * _memory_page_size, &align_offset);
		memset(heap, 0, sizeof(heap_t));
		heap->align_offset = align_offset;

		//Get a new heap ID
		do {
			heap->id = atomic_incr32(&_memory_heap_id);
			if (_memory_heap_lookup(heap->id))
				heap->id = 0;
		} while (!heap->id);

		//Link in heap in heap ID map
		size_t list_idx = heap->id % HEAP_ARRAY_SIZE;
		do {
			next_heap = (heap_t*)atomic_load_ptr(&_memory_heaps[list_idx]);
			heap->next_heap = next_heap;
		} while (!atomic_cas_ptr(&_memory_heaps[list_idx], heap, next_heap));
	}

#if ENABLE_THREAD_CACHE
	heap->span_counter[0].cache_limit = MIN_SPAN_CACHE_RELEASE + MIN_SPAN_CACHE_SIZE;
	for (size_t idx = 1; idx < LARGE_CLASS_COUNT; ++idx)
		heap->span_counter[idx].cache_limit = MIN_LARGE_SPAN_CACHE_RELEASE + MIN_LARGE_SPAN_CACHE_SIZE;
#endif

	//Clean up any deferred operations
	_memory_unmap_deferred(heap, 0);
	_memory_deallocate_deferred(heap, 0);

	return heap;
}

//! Deallocate the given small/medium memory block from the given heap
static void
_memory_deallocate_to_heap(heap_t* heap, span_t* span, void* p) {
	//Check if span is the currently active span in order to operate
	//on the correct bookkeeping data
	assert(SPAN_COUNT(span->flags) == 1);
	const count_t class_idx = span->size_class;
	size_class_t* size_class = _memory_size_class + class_idx;
	int is_active = (heap->active_span[class_idx] == span);
	span_block_t* block_data = is_active ?
	                           heap->active_block + class_idx :
	                           &span->data.block;

	//Check if the span will become completely free
	if (block_data->free_count == ((count_t)size_class->block_count - 1)) {
#if ENABLE_THREAD_CACHE
		//Track counters
		assert(heap->span_counter[0].current_allocations > 0);
		if (heap->span_counter[0].current_allocations)
			--heap->span_counter[0].current_allocations;
#endif

		//If it was active, reset counter. Otherwise, if not active, remove from
		//partial free list if we had a previous free block (guard for classes with only 1 block)
		if (is_active)
			block_data->free_count = 0;
		else if (block_data->free_count > 0)
			_memory_span_list_doublelink_remove(&heap->size_cache[class_idx], span);

		//Add to heap span cache
		_memory_heap_cache_insert(heap, span);
		return;
	}

	//Check if first free block for this span (previously fully allocated)
	if (block_data->free_count == 0) {
		//add to free list and disable autolink
		_memory_span_list_doublelink_add(&heap->size_cache[class_idx], span);
		block_data->first_autolink = (uint16_t)size_class->block_count;
	}
	++block_data->free_count;
	//Span is not yet completely free, so add block to the linked list of free blocks
	void* blocks_start = pointer_offset(span, SPAN_HEADER_SIZE);
	count_t block_offset = (count_t)pointer_diff(p, blocks_start);
	count_t block_idx = block_offset / (count_t)size_class->size;
	uint32_t* block = (uint32_t*)pointer_offset(blocks_start, block_idx * size_class->size);
	*block = block_data->free_list;
	block_data->free_list = (uint16_t)block_idx;
}

//! Deallocate the given large memory block from the given heap
static void
_memory_deallocate_large_to_heap(heap_t* heap, span_t* span) {
	//Decrease counter
	size_t idx = (size_t)span->size_class - SIZE_CLASS_COUNT;
	size_t span_count = idx + 1;
	assert((size_t)SPAN_COUNT(span->flags) == span_count);
	assert(span->size_class >= SIZE_CLASS_COUNT);
	assert(idx < LARGE_CLASS_COUNT);
#if ENABLE_THREAD_CACHE
	assert(heap->span_counter[idx].current_allocations > 0);
	if (heap->span_counter[idx].current_allocations)
		--heap->span_counter[idx].current_allocations;
#endif
	if (!heap->spans_reserved && (span_count > 1)) {
		//Split the span and store remainder as reserved spans
		//Must split to a dummy 1-span master since we cannot have master spans as reserved
		_memory_set_span_remainder_as_reserved(heap, span, 1);
		span_count = 1;
	}

	//Insert into cache list
	_memory_heap_cache_insert(heap, span);
}

//! Process pending deferred cross-thread deallocations
static int
_memory_deallocate_deferred(heap_t* heap, size_t size_class) {
	//Grab the current list of deferred deallocations
	atomic_thread_fence_acquire();
	void* p = atomic_load_ptr(&heap->defer_deallocate);
	if (!p || !atomic_cas_ptr(&heap->defer_deallocate, 0, p))
		return 0;
	//Keep track if we deallocate in the given size class
	int got_class = 0;
	do {
		void* next = *(void**)p;
		//Get span and check which type of block
		span_t* span = (span_t*)(void*)((uintptr_t)p & _memory_span_mask);
		if (span->size_class < SIZE_CLASS_COUNT) {
			//Small/medium block
			got_class |= (span->size_class == size_class);
			_memory_deallocate_to_heap(heap, span, p);
		}
		else {
			//Large block
			got_class |= ((span->size_class >= size_class) && (span->size_class <= (size_class + 2)));
			_memory_deallocate_large_to_heap(heap, span);
		}
		//Loop until all pending operations in list are processed
		p = next;
	} while (p);
	return got_class;
}

//! Defer deallocation of the given block to the given heap
static void
_memory_deallocate_defer(int32_t heap_id, void* p) {
	//Get the heap and link in pointer in list of deferred operations
	heap_t* heap = _memory_heap_lookup(heap_id);
	if (!heap)
		return;
	void* last_ptr;
	do {
		last_ptr = atomic_load_ptr(&heap->defer_deallocate);
		*(void**)p = last_ptr; //Safe to use block, it's being deallocated
	} while (!atomic_cas_ptr(&heap->defer_deallocate, p, last_ptr));
}

//! Allocate a block of the given size
static void*
_memory_allocate(size_t size) {
	if (size <= _memory_medium_size_limit)
		return _memory_allocate_from_heap(get_thread_heap(), size);
	else if (size <= LARGE_SIZE_LIMIT)
		return _memory_allocate_large_from_heap(get_thread_heap(), size);

	//Oversized, allocate pages directly
	size += SPAN_HEADER_SIZE;
	size_t num_pages = size >> _memory_page_size_shift;
	if (size & (_memory_page_size - 1))
		++num_pages;
	size_t align_offset = 0;
	span_t* span = (span_t*)_memory_map(num_pages * _memory_page_size, &align_offset);
	atomic_store32(&span->heap_id, 0);
	//Store page count in next_span
	span->next_span = (span_t*)((uintptr_t)num_pages);
	span->data.list.align_offset = (uint16_t)align_offset;

	return pointer_offset(span, SPAN_HEADER_SIZE);
}

//! Deallocate the given block
static void
_memory_deallocate(void* p) {
	if (!p)
		return;

	//Grab the span (always at start of span, using 64KiB alignment)
	span_t* span = (span_t*)(void*)((uintptr_t)p & _memory_span_mask);
	int32_t heap_id = atomic_load32(&span->heap_id);
	heap_t* heap = get_thread_heap();
	//Check if block belongs to this heap or if deallocation should be deferred
	if (heap_id == heap->id) {
		if (span->size_class < SIZE_CLASS_COUNT)
			_memory_deallocate_to_heap(heap, span, p);
		else
			_memory_deallocate_large_to_heap(heap, span);
	}
	else if (heap_id > 0) {
		_memory_deallocate_defer(heap_id, p);
	}
	else {
		//Oversized allocation, page count is stored in next_span
		size_t num_pages = (size_t)span->next_span;
		_memory_unmap(span, num_pages * _memory_page_size, span->data.list.align_offset, 1);
	}
}

//! Reallocate the given block to the given size
static void*
_memory_reallocate(void* p, size_t size, size_t oldsize, unsigned int flags) {
	if (p) {
		//Grab the span using guaranteed span alignment
		span_t* span = (span_t*)(void*)((uintptr_t)p & _memory_span_mask);
		int32_t heap_id = atomic_load32(&span->heap_id);
		if (heap_id) {
			if (span->size_class < SIZE_CLASS_COUNT) {
				//Small/medium sized block
				size_class_t* size_class = _memory_size_class + span->size_class;
				if ((size_t)size_class->size >= size)
					return p; //Still fits in block, never mind trying to save memory
				if (!oldsize)
					oldsize = size_class->size;
			}
			else {
				//Large block
				size_t total_size = size + SPAN_HEADER_SIZE;
				size_t num_spans = total_size >> _memory_span_size_shift;
				if (total_size & (_memory_span_mask - 1))
					++num_spans;
				size_t current_spans = (span->size_class - SIZE_CLASS_COUNT) + 1;
				if ((current_spans >= num_spans) && (num_spans >= (current_spans / 2)))
					return p; //Still fits and less than half of memory would be freed
				if (!oldsize)
					oldsize = (current_spans * _memory_span_size) - SPAN_HEADER_SIZE;
			}
		}
		else {
			//Oversized block
			size_t total_size = size + SPAN_HEADER_SIZE;
			size_t num_pages = total_size >> _memory_page_size_shift;
			if (total_size & (_memory_page_size - 1))
				++num_pages;
			//Page count is stored in next_span
			size_t current_pages = (size_t)span->next_span;
			if ((current_pages >= num_pages) && (num_pages >= (current_pages / 2)))
				return p; //Still fits and less than half of memory would be freed
			if (!oldsize)
				oldsize = (current_pages * _memory_page_size) - SPAN_HEADER_SIZE;
		}
	}

	//Size is greater than block size, need to allocate a new block and deallocate the old
	//Avoid hysteresis by overallocating if increase is small (below 37%)
	size_t lower_bound = oldsize + (oldsize >> 2) + (oldsize >> 3);
	void* block = _memory_allocate((size > lower_bound) ? size : ((size > oldsize) ? lower_bound : size));
	if (p) {
		if (!(flags & RPMALLOC_NO_PRESERVE))
			memcpy(block, p, oldsize < size ? oldsize : size);
		_memory_deallocate(p);
	}

	return block;
}

//! Get the usable size of the given block
static size_t
_memory_usable_size(void* p) {
	//Grab the span using guaranteed span alignment
	span_t* span = (span_t*)(void*)((uintptr_t)p & _memory_span_mask);
	int32_t heap_id = atomic_load32(&span->heap_id);
	if (heap_id) {
		//Small/medium block
		if (span->size_class < SIZE_CLASS_COUNT)
			return _memory_size_class[span->size_class].size;

		//Large block
		size_t current_spans = (span->size_class - SIZE_CLASS_COUNT) + 1;
		return (current_spans * _memory_span_size) - SPAN_HEADER_SIZE;
	}

	//Oversized block, page count is stored in next_span
	size_t current_pages = (size_t)span->next_span;
	return (current_pages * _memory_page_size) - SPAN_HEADER_SIZE;
}

//! Adjust and optimize the size class properties for the given class
static void
_memory_adjust_size_class(size_t iclass) {
	size_t block_size = _memory_size_class[iclass].size;
	size_t block_count = (_memory_span_size - SPAN_HEADER_SIZE) / block_size;

	_memory_size_class[iclass].block_count = (uint16_t)block_count;
	_memory_size_class[iclass].class_idx = (uint16_t)iclass;

	//Check if previous size classes can be merged
	size_t prevclass = iclass;
	while (prevclass > 0) {
		--prevclass;
		//A class can be merged if number of pages and number of blocks are equal
		if (_memory_size_class[prevclass].block_count == _memory_size_class[iclass].block_count) {
			memcpy(_memory_size_class + prevclass, _memory_size_class + iclass, sizeof(_memory_size_class[iclass]));
		}
		else {
			break;
		}
	}
}

}

#if defined( _WIN32 ) || defined( __WIN32__ ) || defined( _WIN64 )
#  include <windows.h>
#else
#  include <sys/mman.h>
#  include <sched.h>
#  ifndef MAP_UNINITIALIZED
#    define MAP_UNINITIALIZED 0
#  endif
#endif
#include <errno.h>

namespace tracy
{

//! Initialize the allocator and setup global data
int
rpmalloc_initialize(void) {
	memset(&_memory_config, 0, sizeof(rpmalloc_config_t));
	return rpmalloc_initialize_config(0);
}

int
rpmalloc_initialize_config(const rpmalloc_config_t* config) {
	if (config)
		memcpy(&_memory_config, config, sizeof(rpmalloc_config_t));

	if (!_memory_config.memory_map || !_memory_config.memory_unmap) {
		_memory_config.memory_map = _memory_map_os;
		_memory_config.memory_unmap = _memory_unmap_os;
	}

	_memory_page_size = _memory_config.page_size;
	if (!_memory_page_size) {
#if PLATFORM_WINDOWS
		SYSTEM_INFO system_info;
		memset(&system_info, 0, sizeof(system_info));
		GetSystemInfo(&system_info);
		_memory_page_size = system_info.dwPageSize;
		_memory_map_granularity = system_info.dwAllocationGranularity;
#else
		_memory_page_size = (size_t)sysconf(_SC_PAGESIZE);
		_memory_map_granularity = _memory_page_size;
#endif
	}

	if (_memory_page_size < 512)
		_memory_page_size = 512;
	if (_memory_page_size > (16 * 1024))
		_memory_page_size = (16 * 1024);

	_memory_page_size_shift = 0;
	size_t page_size_bit = _memory_page_size;
	while (page_size_bit != 1) {
		++_memory_page_size_shift;
		page_size_bit >>= 1;
	}
	_memory_page_size = ((size_t)1 << _memory_page_size_shift);
	_memory_page_mask = ~(uintptr_t)(_memory_page_size - 1);

	size_t span_size = _memory_config.span_size;
	if (!span_size)
		span_size = (64 * 1024);
	if (span_size > (256 * 1024))
		span_size = (256 * 1024);
	_memory_span_size = 4096;
	_memory_span_size_shift = 12;
	while ((_memory_span_size < span_size) || (_memory_span_size < _memory_page_size)) {
		_memory_span_size <<= 1;
		++_memory_span_size_shift;
	}
	_memory_span_mask = ~(uintptr_t)(_memory_span_size - 1);

	_memory_config.page_size = _memory_page_size;
	_memory_config.span_size = _memory_span_size;

	if (!_memory_config.span_map_count)
		_memory_config.span_map_count = DEFAULT_SPAN_MAP_COUNT;
	if (_memory_config.span_size * _memory_config.span_map_count < _memory_config.page_size)
		_memory_config.span_map_count = (_memory_config.page_size / _memory_config.span_size);
	if (_memory_config.span_map_count > 128)
		_memory_config.span_map_count = 128;

#if defined(__APPLE__) && ENABLE_PRELOAD
	if (pthread_key_create(&_memory_thread_heap, 0))
		return -1;
#endif

	atomic_store32(&_memory_heap_id, 0);
	atomic_store32(&_memory_orphan_counter, 0);
	atomic_store32(&_memory_active_heaps, 0);

	//Setup all small and medium size classes
	size_t iclass;
	for (iclass = 0; iclass < SMALL_CLASS_COUNT; ++iclass) {
		size_t size = (iclass + 1) * SMALL_GRANULARITY;
		_memory_size_class[iclass].size = (uint16_t)size;
		_memory_adjust_size_class(iclass);
	}

	_memory_medium_size_limit = _memory_span_size - SPAN_HEADER_SIZE;
	if (_memory_medium_size_limit > MEDIUM_SIZE_LIMIT)
		_memory_medium_size_limit = MEDIUM_SIZE_LIMIT;
	for (iclass = 0; iclass < MEDIUM_CLASS_COUNT; ++iclass) {
		size_t size = SMALL_SIZE_LIMIT + ((iclass + 1) * MEDIUM_GRANULARITY);
		if (size > _memory_medium_size_limit)
			size = _memory_medium_size_limit;
		_memory_size_class[SMALL_CLASS_COUNT + iclass].size = (uint16_t)size;
		_memory_adjust_size_class(SMALL_CLASS_COUNT + iclass);
	}

	//Initialize this thread
	rpmalloc_thread_initialize();
	return 0;
}

//! Finalize the allocator
void
rpmalloc_finalize(void) {
	atomic_thread_fence_acquire();

	rpmalloc_thread_finalize();
	//If you hit this assert, you still have active threads or forgot to finalize some thread(s)
	assert(atomic_load32(&_memory_active_heaps) == 0);

	//Free all thread caches
	for (size_t list_idx = 0; list_idx < HEAP_ARRAY_SIZE; ++list_idx) {
		heap_t* heap = (heap_t*)atomic_load_ptr(&_memory_heaps[list_idx]);
		while (heap) {
			_memory_deallocate_deferred(heap, 0);

			//Free span caches (other thread might have deferred after the thread using this heap finalized)
#if ENABLE_THREAD_CACHE
			for (size_t iclass = 0; iclass < LARGE_CLASS_COUNT; ++iclass) {
				if (heap->span_cache[iclass])
					_memory_unmap_span_list(0, heap->span_cache[iclass]);
			}
#endif
			heap = heap->next_heap;
		}
	}

#if ENABLE_GLOBAL_CACHE
	//Free global caches
	for (size_t iclass = 0; iclass < LARGE_CLASS_COUNT; ++iclass)
		_memory_cache_finalize(&_memory_span_cache[iclass]);
#endif

	for (size_t list_idx = 0; list_idx < HEAP_ARRAY_SIZE; ++list_idx) {
		heap_t* heap = (heap_t*)atomic_load_ptr(&_memory_heaps[list_idx]);
		atomic_store_ptr(&_memory_heaps[list_idx], 0);
		while (heap) {
			if (heap->spans_reserved) {
				span_t* span = heap->span_reserve;
				span_t* master = heap->span_reserve_master;
				uint32_t remains = SPAN_REMAINS(master->flags);

				assert(master != span);
				assert(remains >= heap->spans_reserved);
				_memory_unmap(span, heap->spans_reserved * _memory_span_size, 0, 0);
				_memory_statistics_sub(&_reserved_spans, heap->spans_reserved);
				remains = ((uint32_t)heap->spans_reserved >= remains) ? 0 : (remains - (uint32_t)heap->spans_reserved);
				if (!remains) {
					uint32_t master_span_count = SPAN_COUNT(master->flags);
					_memory_statistics_sub(&_reserved_spans, master_span_count);
					_memory_unmap(master, master_span_count * _memory_span_size, master->data.list.align_offset, 1);
				}
				else {
					SPAN_SET_REMAINS(master->flags, remains);
				}
			}

			_memory_unmap_deferred(heap, 0);

			heap_t* next_heap = heap->next_heap;
			_memory_unmap(heap, (1 + (sizeof(heap_t) >> _memory_page_size_shift)) * _memory_page_size, heap->align_offset, 1);
			heap = next_heap;
		}
	}
	atomic_store_ptr(&_memory_orphan_heaps, 0);
	atomic_thread_fence_release();

#if ENABLE_STATISTICS
	//If you hit these asserts you probably have memory leaks or double frees in your code
	assert(!atomic_load32(&_mapped_pages));
	assert(!atomic_load32(&_reserved_spans));
#endif

#if defined(__APPLE__) && ENABLE_PRELOAD
	pthread_key_delete(_memory_thread_heap);
#endif
}

//! Initialize thread, assign heap
void
rpmalloc_thread_initialize(void) {
	if (!get_thread_heap()) {
		atomic_incr32(&_memory_active_heaps);
		heap_t* heap = _memory_allocate_heap();
#if ENABLE_STATISTICS
		heap->thread_to_global = 0;
		heap->global_to_thread = 0;
#endif
		set_thread_heap(heap);
	}
}

//! Finalize thread, orphan heap
void
rpmalloc_thread_finalize(void) {
	heap_t* heap = get_thread_heap();
	if (!heap)
		return;

	_memory_deallocate_deferred(heap, 0);
	_memory_unmap_deferred(heap, 0);

	//Release thread cache spans back to global cache
#if ENABLE_THREAD_CACHE
	for (size_t iclass = 0; iclass < LARGE_CLASS_COUNT; ++iclass) {
		span_t* span = heap->span_cache[iclass];
#if ENABLE_GLOBAL_CACHE
		const size_t span_count = iclass + 1;
		while (span) {
			assert((size_t)SPAN_COUNT(span->flags) == span_count);
			span_t* next = _memory_span_list_split(span, !iclass ? MIN_SPAN_CACHE_RELEASE : (MIN_LARGE_SPAN_CACHE_RELEASE / span_count));
			_memory_global_cache_insert(0, span);
			span = next;
		}
#else
		if (span)
			_memory_unmap_span_list(heap, span);
#endif
		heap->span_cache[iclass] = 0;
	}
#endif

	//Orphan the heap
	void* raw_heap;
	uintptr_t orphan_counter;
	heap_t* last_heap;
	do {
		last_heap = (heap_t*)atomic_load_ptr(&_memory_orphan_heaps);
		heap->next_orphan = (heap_t*)(void*)((uintptr_t)last_heap & _memory_page_mask);
		orphan_counter = (uintptr_t)atomic_incr32(&_memory_orphan_counter);
		raw_heap = (void*)((uintptr_t)heap | (orphan_counter & ~_memory_page_mask));
	}
	while (!atomic_cas_ptr(&_memory_orphan_heaps, raw_heap, last_heap));

	set_thread_heap(0);
	atomic_add32(&_memory_active_heaps, -1);
}

int
rpmalloc_is_thread_initialized(void) {
	return (get_thread_heap() != 0) ? 1 : 0;
}

const rpmalloc_config_t*
rpmalloc_config(void) {
	return &_memory_config;
}

//! Map new pages to virtual memory
static void*
_memory_map_os(size_t size, size_t* offset) {
	//Either size is a heap (a single page) or a (multiple) span - we only need to align spans
	size_t padding = ((size >= _memory_span_size) && (_memory_span_size > _memory_map_granularity)) ? _memory_span_size : 0;

#if PLATFORM_WINDOWS
	//Ok to MEM_COMMIT - according to MSDN, "actual physical pages are not allocated unless/until the virtual addresses are actually accessed"
	void* ptr = VirtualAlloc(0, size + padding, MEM_RESERVE | MEM_COMMIT, PAGE_READWRITE);
	if (!ptr) {
		assert("Failed to map virtual memory block" && 0);
		return 0;
	}
#else
	void* ptr = mmap(0, size + padding, PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS | MAP_UNINITIALIZED, -1, 0);
	if ((ptr == MAP_FAILED) || !ptr) {
		assert("Failed to map virtual memory block" && 0);
		return 0;
	}
#endif

	if (padding) {
		size_t final_padding = padding - ((uintptr_t)ptr & ~_memory_span_mask);
#if PLATFORM_POSIX
		//Unmap the last unused pages, for Windows this is done with the final VirtualFree with MEM_RELEASE call
		size_t remains = padding - final_padding;
		if (remains)
			munmap(pointer_offset(ptr, final_padding + size), remains);
#endif
		ptr = pointer_offset(ptr, final_padding);
		assert(final_padding <= _memory_span_size);
		assert(!(final_padding & 5));
		assert(!((uintptr_t)ptr & ~_memory_span_mask));
		*offset = final_padding >> 3;
		assert(*offset < 65536);
	}

	return ptr;
}

//! Unmap pages from virtual memory
static void
_memory_unmap_os(void* address, size_t size, size_t offset, int release) {
	assert(release || (offset == 0));
	if (release && offset) {
		offset <<= 3;
#if PLATFORM_POSIX
		size += offset;
#endif
		address = pointer_offset(address, -(int32_t)offset);
	}
#if PLATFORM_WINDOWS
	if (!VirtualFree(address, release ? 0 : size, release ? MEM_RELEASE : MEM_DECOMMIT)) {
		DWORD err = GetLastError();
		(void)err;
		assert("Failed to unmap virtual memory block" && 0);
	}
#else
	MEMORY_UNUSED(release);
	if (munmap(address, size)) {
		assert("Failed to unmap virtual memory block" && 0);
	}
#endif
}

#if ENABLE_GUARDS
static void
_memory_guard_validate(void* p) {
	if (!p)
		return;
	void* block_start;
	size_t block_size = _memory_usable_size(p);
	span_t* span = (void*)((uintptr_t)p & _memory_span_mask);
	int32_t heap_id = atomic_load32(&span->heap_id);
	if (heap_id) {
		if (span->size_class < SIZE_CLASS_COUNT) {
			void* span_blocks_start = pointer_offset(span, SPAN_HEADER_SIZE);
			size_class_t* size_class = _memory_size_class + span->size_class;
			count_t block_offset = (count_t)pointer_diff(p, span_blocks_start);
			count_t block_idx = block_offset / (count_t)size_class->size;
			block_start = pointer_offset(span_blocks_start, block_idx * size_class->size);
		}
		else {
			block_start = pointer_offset(span, SPAN_HEADER_SIZE);
		}
	}
	else {
		block_start = pointer_offset(span, SPAN_HEADER_SIZE);
	}
	uint32_t* deadzone = block_start;
	//If these asserts fire, you have written to memory before the block start
	for (int i = 0; i < 8; ++i) {
		if (deadzone[i] != MAGIC_GUARD) {
			if (_memory_config.memory_overwrite)
				_memory_config.memory_overwrite(p);
			else
				assert("Memory overwrite before block start" && 0);
			return;
		}
		deadzone[i] = 0;
	}
	deadzone = (uint32_t*)pointer_offset(block_start, block_size - 32);
	//If these asserts fire, you have written to memory after the block end
	for (int i = 0; i < 8; ++i) {
		if (deadzone[i] != MAGIC_GUARD) {
			if (_memory_config.memory_overwrite)
				_memory_config.memory_overwrite(p);
			else
				assert("Memory overwrite after block end" && 0);
			return;
		}
		deadzone[i] = 0;
	}
}
#else
#define _memory_guard_validate(block)
#endif

#if ENABLE_GUARDS
static void
_memory_guard_block(void* block) {
	if (block) {
		size_t block_size = _memory_usable_size(block);
		uint32_t* deadzone = block;
		deadzone[0] = deadzone[1] = deadzone[2] = deadzone[3] =
		deadzone[4] = deadzone[5] = deadzone[6] = deadzone[7] = MAGIC_GUARD;
		deadzone = (uint32_t*)pointer_offset(block, block_size - 32);
		deadzone[0] = deadzone[1] = deadzone[2] = deadzone[3] =
		deadzone[4] = deadzone[5] = deadzone[6] = deadzone[7] = MAGIC_GUARD;
	}
}
#define _memory_guard_pre_alloc(size) size += 64
#define _memory_guard_pre_realloc(block, size) block = pointer_offset(block, -32); size += 64
#define _memory_guard_post_alloc(block, size) _memory_guard_block(block); block = pointer_offset(block, 32); size -= 64
#else
#define _memory_guard_pre_alloc(size)
#define _memory_guard_pre_realloc(block, size)
#define _memory_guard_post_alloc(block, size)
#endif

// Extern interface

RPMALLOC_RESTRICT void*
rpmalloc(size_t size) {
#if ENABLE_VALIDATE_ARGS
	if (size >= MAX_ALLOC_SIZE) {
		errno = EINVAL;
		return 0;
	}
#endif
	_memory_guard_pre_alloc(size);
	void* block = _memory_allocate(size);
	_memory_guard_post_alloc(block, size);
	return block;
}

void
rpfree(void* ptr) {
	_memory_guard_validate(ptr);
	_memory_deallocate(ptr);
}

RPMALLOC_RESTRICT void*
rpcalloc(size_t num, size_t size) {
	size_t total;
#if ENABLE_VALIDATE_ARGS
#if PLATFORM_WINDOWS
	int err = SizeTMult(num, size, &total);
	if ((err != S_OK) || (total >= MAX_ALLOC_SIZE)) {
		errno = EINVAL;
		return 0;
	}
#else
	int err = __builtin_umull_overflow(num, size, &total);
	if (err || (total >= MAX_ALLOC_SIZE)) {
		errno = EINVAL;
		return 0;
	}
#endif
#else
	total = num * size;
#endif
	_memory_guard_pre_alloc(total);
	void* block = _memory_allocate(total);
	_memory_guard_post_alloc(block, total);
	memset(block, 0, total);
	return block;
}

void*
rprealloc(void* ptr, size_t size) {
#if ENABLE_VALIDATE_ARGS
	if (size >= MAX_ALLOC_SIZE) {
		errno = EINVAL;
		return ptr;
	}
#endif
	_memory_guard_validate(ptr);
	_memory_guard_pre_realloc(ptr, size);
	void* block = _memory_reallocate(ptr, size, 0, 0);
	_memory_guard_post_alloc(block, size);
	return block;
}

void*
rpaligned_realloc(void* ptr, size_t alignment, size_t size, size_t oldsize,
                  unsigned int flags) {
#if ENABLE_VALIDATE_ARGS
	if ((size + alignment < size) || (alignment > _memory_page_size)) {
		errno = EINVAL;
		return 0;
	}
#endif
	void* block;
	if (alignment > 32) {
		block = rpaligned_alloc(alignment, size);
		if (!(flags & RPMALLOC_NO_PRESERVE))
			memcpy(block, ptr, oldsize < size ? oldsize : size);
		rpfree(ptr);
	}
	else {
		_memory_guard_validate(ptr);
		_memory_guard_pre_realloc(ptr, size);
		block = _memory_reallocate(ptr, size, oldsize, flags);
		_memory_guard_post_alloc(block, size);
	}
	return block;
}

RPMALLOC_RESTRICT void*
rpaligned_alloc(size_t alignment, size_t size) {
	if (alignment <= 32)
		return rpmalloc(size);

#if ENABLE_VALIDATE_ARGS
	if ((size + alignment < size) || (alignment > _memory_page_size)) {
		errno = EINVAL;
		return 0;
	}
#endif

	void* ptr = rpmalloc(size + alignment);
	if ((uintptr_t)ptr & (alignment - 1))
		ptr = (void*)(((uintptr_t)ptr & ~((uintptr_t)alignment - 1)) + alignment);
	return ptr;
}

RPMALLOC_RESTRICT void*
rpmemalign(size_t alignment, size_t size) {
	return rpaligned_alloc(alignment, size);
}

int
rpposix_memalign(void **memptr, size_t alignment, size_t size) {
	if (memptr)
		*memptr = rpaligned_alloc(alignment, size);
	else
		return EINVAL;
	return *memptr ? 0 : ENOMEM;
}

size_t
rpmalloc_usable_size(void* ptr) {
	size_t size = 0;
	if (ptr) {
		size = _memory_usable_size(ptr);
#if ENABLE_GUARDS
		size -= 64;
#endif
	}
	return size;
}

void
rpmalloc_thread_collect(void) {
	heap_t* heap = get_thread_heap();
	_memory_unmap_deferred(heap, 0);
	_memory_deallocate_deferred(0, 0);
}

void
rpmalloc_thread_statistics(rpmalloc_thread_statistics_t* stats) {
	memset(stats, 0, sizeof(rpmalloc_thread_statistics_t));
	heap_t* heap = get_thread_heap();
	void* p = atomic_load_ptr(&heap->defer_deallocate);
	while (p) {
		void* next = *(void**)p;
		span_t* span = (span_t*)(void*)((uintptr_t)p & _memory_span_mask);
		stats->deferred += _memory_size_class[span->size_class].size;
		p = next;
	}

	for (size_t isize = 0; isize < SIZE_CLASS_COUNT; ++isize) {
		if (heap->active_block[isize].free_count)
			stats->active += heap->active_block[isize].free_count * _memory_size_class[heap->active_span[isize]->size_class].size;

		span_t* cache = heap->size_cache[isize];
		while (cache) {
			stats->sizecache = cache->data.block.free_count * _memory_size_class[cache->size_class].size;
			cache = cache->next_span;
		}
	}

#if ENABLE_THREAD_CACHE
	for (size_t iclass = 0; iclass < LARGE_CLASS_COUNT; ++iclass) {
		if (heap->span_cache[iclass])
			stats->spancache = (size_t)heap->span_cache[iclass]->data.list.size * (iclass + 1) * _memory_span_size;
	}
#endif
}

void
rpmalloc_global_statistics(rpmalloc_global_statistics_t* stats) {
	memset(stats, 0, sizeof(rpmalloc_global_statistics_t));
#if ENABLE_STATISTICS
	stats->mapped = (size_t)atomic_load32(&_mapped_pages) * _memory_page_size;
	stats->mapped_total = (size_t)atomic_load32(&_mapped_total) * _memory_page_size;
	stats->unmapped_total = (size_t)atomic_load32(&_unmapped_total) * _memory_page_size;
#endif
#if ENABLE_GLOBAL_CACHE
	for (size_t iclass = 0; iclass < LARGE_CLASS_COUNT; ++iclass) {
		stats->cached += (size_t)atomic_load32(&_memory_span_cache[iclass].size) * (iclass + 1) * _memory_span_size;
	}
#endif
}

}

#ifdef _MSC_VER
#  pragma warning( pop )
#endif

#endif