VulkanMemoryAllocator 0.3.7.1 → 0.3.8
raw patch · 5 files changed
+286/−227 lines, 5 filesPVP ok
version bump matches the API change (PVP)
API changes (from Hackage documentation)
Files
- VulkanMemoryAllocator.cabal +2/−2
- VulkanMemoryAllocator/src/vk_mem_alloc.h +231/−203
- changelog.md +6/−0
- package.yaml +1/−1
- src/VulkanMemoryAllocator.hs +46/−21
VulkanMemoryAllocator.cabal view
@@ -4,10 +4,10 @@ -- -- see: https://github.com/sol/hpack ----- hash: bc2600808d2a3ef448255465e2fc5e1165aa7f40e9cb43ec1ab15ec0f8cce272+-- hash: fc6e88640fbfd21d257ac23c8012464900c1691882a5213592c472a5c787ac03 name: VulkanMemoryAllocator-version: 0.3.7.1+version: 0.3.8 synopsis: Bindings to the VulkanMemoryAllocator library category: Graphics homepage: https://github.com/expipiplus1/vulkan#readme
VulkanMemoryAllocator/src/vk_mem_alloc.h view
@@ -25,7 +25,7 @@ /** \mainpage Vulkan Memory Allocator -<b>Version 3.0.0-development</b> (2020-06-24) +<b>Version 3.0.0-development</b> (2020-11-03) Copyright (c) 2017-2020 Advanced Micro Devices, Inc. All rights reserved. \n License: MIT @@ -142,7 +142,16 @@ `WINVER` for Windows, `VK_USE_PLATFORM_WIN32_KHR` for Vulkan), you must define them before every `#include` of this library. +You may need to configure the way you import Vulkan functions. +- By default, VMA assumes you you link statically with Vulkan API. If this is not the case, + `#define VMA_STATIC_VULKAN_FUNCTIONS 0` before `#include` of the VMA implementation and use another way. +- You can `#define VMA_DYNAMIC_VULKAN_FUNCTIONS 1` and make sure `vkGetInstanceProcAddr` and `vkGetDeviceProcAddr` globals are defined. + All the remaining Vulkan functions will be fetched automatically. +- Finally, you can provide your own pointers to all Vulkan functions needed by VMA using structure member + VmaAllocatorCreateInfo::pVulkanFunctions, if you fetched them in some custom way e.g. using some loader like [Volk](https://github.com/zeux/volk). + + \section quick_start_initialization Initialization At program startup: @@ -153,6 +162,7 @@ \code VmaAllocatorCreateInfo allocatorInfo = {}; +allocatorInfo.vulkanApiVersion = VK_API_VERSION_1_2; allocatorInfo.physicalDevice = physicalDevice; allocatorInfo.device = device; allocatorInfo.instance = instance; @@ -161,6 +171,13 @@ vmaCreateAllocator(&allocatorInfo, &allocator); \endcode +Only members `physicalDevice`, `device`, `instance` are required. +However, you should inform the library which Vulkan version do you use by setting +VmaAllocatorCreateInfo::vulkanApiVersion and which extensions did you enable +by setting VmaAllocatorCreateInfo::flags (like #VMA_ALLOCATOR_CREATE_BUFFER_DEVICE_ADDRESS_BIT for VK_KHR_buffer_device_address). +Otherwise, VMA would use only features of Vulkan 1.0 core with no extensions. + + \section quick_start_resource_allocation Resource allocation When you want to create a buffer or image: @@ -428,7 +445,7 @@ - Keeping many large memory blocks mapped may impact performance or stability of some debugging tools. \section memory_mapping_cache_control Cache flush and invalidate - + Memory in Vulkan doesn't need to be unmapped before using it on GPU, but unless a memory types has `VK_MEMORY_PROPERTY_HOST_COHERENT_BIT` flag set, you need to manually **invalidate** cache before reading of mapped pointer @@ -932,7 +949,7 @@ To mitigate this problem, you can use defragmentation feature: structure #VmaDefragmentationInfo2, function vmaDefragmentationBegin(), vmaDefragmentationEnd(). -Given set of allocations, +Given set of allocations, this function can move them to compact used memory, ensure more continuous free space and possibly also free some `VkDeviceMemory` blocks. @@ -998,9 +1015,9 @@ // Create new buffer with same parameters. VkBufferCreateInfo bufferInfo = ...; vkCreateBuffer(device, &bufferInfo, nullptr, &buffers[i]); - + // You can make dummy call to vkGetBufferMemoryRequirements here to silence validation layer warning. - + // Bind new buffer to new memory region. Data contained in it is already moved. VmaAllocationInfo allocInfo; vmaGetAllocationInfo(allocator, allocations[i], &allocInfo); @@ -1076,9 +1093,9 @@ // Create new buffer with same parameters. VkBufferCreateInfo bufferInfo = ...; vkCreateBuffer(device, &bufferInfo, nullptr, &buffers[i]); - + // You can make dummy call to vkGetBufferMemoryRequirements here to silence validation layer warning. - + // Bind new buffer to new memory region. Data contained in it is already moved. VmaAllocationInfo allocInfo; vmaGetAllocationInfo(allocator, allocations[i], &allocInfo); @@ -1472,7 +1489,7 @@ You can also explicitly request checking margins of all allocations in all memory blocks that belong to specified memory types by using function vmaCheckCorruption(), -or in memory blocks that belong to specified custom pool, by using function +or in memory blocks that belong to specified custom pool, by using function vmaCheckPoolCorruption(). Margin validation (corruption detection) works only for memory types that are @@ -1981,6 +1998,8 @@ explicit memory type index and dedicated allocation anyway, so they don't interact with main features of this library. Such special purpose allocations should be made manually, using `vkCreateBuffer()` and `vkAllocateMemory()`. +- Sub-allocation of parts of one large buffer. Although recommended as a good practice, + it is the user's responsibility to implement such logic on top of VMA. - Recreation of buffers and images. Although the library has functions for buffer and image creation (vmaCreateBuffer(), vmaCreateImage()), you need to recreate these objects yourself after defragmentation. That's because the big @@ -1992,7 +2011,7 @@ is usually not needed in desktop PC applications anyway. Success of an allocation is just checked with an assert. - Code free of any compiler warnings. Maintaining the library to compile and - work correctly on so many different platforms is hard enough. Being free of + work correctly on so many different platforms is hard enough. Being free of any warnings, on any version of any compiler, is simply not feasible. - This is a C++ library with C interface. Bindings or ports to any other programming languages are welcomed as external projects and @@ -2263,7 +2282,7 @@ VMA_ALLOCATOR_CREATE_EXT_MEMORY_BUDGET_BIT = 0x00000008, /** Enables usage of VK_AMD_device_coherent_memory extension. - + You may set this flag only if you: - found out that this device extension is supported and enabled it while creating Vulkan device passed as VmaAllocatorCreateInfo::device, @@ -2273,7 +2292,7 @@ The extension and accompanying device feature provide access to memory types with `VK_MEMORY_PROPERTY_DEVICE_COHERENT_BIT_AMD` and `VK_MEMORY_PROPERTY_DEVICE_UNCACHED_BIT_AMD` flags. They are useful mostly for writing breadcrumb markers - a common method for debugging GPU crash/hang/TDR. - + When the extension is not enabled, such memory types are still enumerated, but their usage is illegal. To protect from this error, if you don't create the allocator with this flag, it will refuse to allocate any memory or create a custom pool in such memory type, returning `VK_ERROR_FEATURE_NOT_PRESENT`. @@ -2345,7 +2364,7 @@ It may degrade performance though. */ VMA_RECORD_FLUSH_AFTER_CALL_BIT = 0x00000001, - + VMA_RECORD_FLAG_BITS_MAX_ENUM = 0x7FFFFFFF } VmaRecordFlagBits; typedef VkFlags VmaRecordFlags; @@ -2443,11 +2462,11 @@ */ VkInstance VMA_NOT_NULL instance; /** \brief Optional. The highest version of Vulkan that the application is designed to use. - + It must be a value in the format as created by macro `VK_MAKE_VERSION` or a constant like: `VK_API_VERSION_1_1`, `VK_API_VERSION_1_0`. The patch version number specified is ignored. Only the major and minor versions are considered. It must be less or equal (preferably equal) to value as passed to `vkCreateInstance` as `VkApplicationInfo::apiVersion`. - Only versions 1.0 and 1.1 are supported by the current implementation. + Only versions 1.0, 1.1, 1.2 are supported by the current implementation. Leaving it initialized to zero is equivalent to `VK_API_VERSION_1_0`. */ uint32_t vulkanApiVersion; @@ -2575,32 +2594,32 @@ /** \brief Sum size of all `VkDeviceMemory` blocks allocated from particular heap, in bytes. */ VkDeviceSize blockBytes; - + /** \brief Sum size of all allocations created in particular heap, in bytes. - + Usually less or equal than `blockBytes`. Difference `blockBytes - allocationBytes` is the amount of memory allocated but unused - available for new allocations or wasted due to fragmentation. - + It might be greater than `blockBytes` if there are some allocations in lost state, as they account to this value as well. */ VkDeviceSize allocationBytes; - + /** \brief Estimated current memory usage of the program, in bytes. - + Fetched from system using `VK_EXT_memory_budget` extension if enabled. - + It might be different than `blockBytes` (usually higher) due to additional implicit objects also occupying the memory, like swapchain, pipelines, descriptor heaps, command buffers, or `VkDeviceMemory` blocks allocated outside of this library, if any. */ VkDeviceSize usage; - + /** \brief Estimated amount of memory available to the program, in bytes. - + Fetched from system using `VK_EXT_memory_budget` extension if enabled. - + It might be different (most probably smaller) than `VkMemoryHeap::size[heapIndex]` due to factors external to the program, like other programs also consuming system resources. Difference `budget - usage` is the amount of additional memory that can probably @@ -2665,7 +2684,7 @@ It is roughly equivalent of `D3D12_HEAP_TYPE_DEFAULT`. Usage: - + - Resources written and read by device, e.g. images used as attachments. - Resources transferred from host once (immutable) or infrequently and read by device multiple times, e.g. textures to be sampled, vertex buffers, uniform @@ -2710,7 +2729,7 @@ VMA_MEMORY_USAGE_CPU_COPY = 5, /** Lazily allocated GPU memory having `VK_MEMORY_PROPERTY_LAZILY_ALLOCATED_BIT`. Exists mostly on mobile platforms. Using it on desktop PC or other GPUs with no such memory type present will fail the allocation. - + Usage: Memory for transient attachment images (color attachments, depth attachments etc.), created with `VK_IMAGE_USAGE_TRANSIENT_ATTACHMENT_BIT`. Allocations with this usage are always created as dedicated - it implies #VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT. @@ -2723,25 +2742,25 @@ /// Flags to be passed as VmaAllocationCreateInfo::flags. typedef enum VmaAllocationCreateFlagBits { /** \brief Set this flag if the allocation should have its own memory block. - + Use it for special, big resources, like fullscreen images used as attachments. - + You should not use this flag if VmaAllocationCreateInfo::pool is not null. */ VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT = 0x00000001, /** \brief Set this flag to only try to allocate from existing `VkDeviceMemory` blocks and never create new such block. - + If new allocation cannot be placed in any of the existing blocks, allocation fails with `VK_ERROR_OUT_OF_DEVICE_MEMORY` error. - + You should not use #VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT and #VMA_ALLOCATION_CREATE_NEVER_ALLOCATE_BIT at the same time. It makes no sense. - + If VmaAllocationCreateInfo::pool is not null, this flag is implied and ignored. */ VMA_ALLOCATION_CREATE_NEVER_ALLOCATE_BIT = 0x00000002, /** \brief Set this flag to use a memory that will be persistently mapped and retrieve pointer to it. - + Pointer to mapped memory will be returned through VmaAllocationInfo::pMappedData. It is valid to use this flag for allocation made from memory type that is not @@ -2837,18 +2856,18 @@ /// Use #VmaAllocationCreateFlagBits enum. VmaAllocationCreateFlags flags; /** \brief Intended usage of memory. - + You can leave #VMA_MEMORY_USAGE_UNKNOWN if you specify memory requirements in other way. \n If `pool` is not null, this member is ignored. */ VmaMemoryUsage usage; /** \brief Flags that must be set in a Memory Type chosen for an allocation. - + Leave 0 if you specify memory requirements in other way. \n If `pool` is not null, this member is ignored.*/ VkMemoryPropertyFlags requiredFlags; /** \brief Flags that preferably should be set in a memory type chosen for an allocation. - + Set to 0 if no additional flags are preferred. \n If `pool` is not null, this member is ignored. */ VkMemoryPropertyFlags preferredFlags; @@ -2867,7 +2886,7 @@ */ VmaPool VMA_NULLABLE pool; /** \brief Custom general-purpose pointer that will be stored in #VmaAllocation, can be read as VmaAllocationInfo::pUserData and changed using vmaSetAllocationUserData(). - + If #VMA_ALLOCATION_CREATE_USER_DATA_COPY_STRING_BIT is used, it must be either null or pointer to a null-terminated string. The string will be then copied to internal buffer, so it doesn't need to be valid after allocation call. @@ -3018,7 +3037,7 @@ /** \brief Maximum number of blocks that can be allocated in this pool. Optional. Set to 0 to use default, which is `SIZE_MAX`, which means no limit. - + Set to same value as VmaPoolCreateInfo::minBlockCount to have fixed amount of memory allocated throughout whole lifetime of this pool. */ @@ -3172,21 +3191,26 @@ */ typedef struct VmaAllocationInfo { /** \brief Memory type index that this allocation was allocated from. - + It never changes. */ uint32_t memoryType; /** \brief Handle to Vulkan memory object. Same memory object can be shared by multiple allocations. - + It can change after call to vmaDefragment() if this allocation is passed to the function, or if allocation is lost. If the allocation is lost, it is equal to `VK_NULL_HANDLE`. */ VkDeviceMemory VMA_NULLABLE_NON_DISPATCHABLE deviceMemory; - /** \brief Offset into deviceMemory object to the beginning of this allocation, in bytes. (deviceMemory, offset) pair is unique to this allocation. + /** \brief Offset in `VkDeviceMemory` object to the beginning of this allocation, in bytes. `(deviceMemory, offset)` pair is unique to this allocation. + You usually don't need to use this offset. If you create a buffer or an image together with the allocation using e.g. function + vmaCreateBuffer(), vmaCreateImage(), functions that operate on these resources refer to the beginning of the buffer or image, + not entire device memory block. Functions like vmaMapMemory(), vmaBindBufferMemory() also refer to the beginning of the allocation + and apply this offset automatically. + It can change after call to vmaDefragment() if this allocation is passed to the function, or if allocation is lost. */ VkDeviceSize offset; @@ -3618,7 +3642,7 @@ */ const VmaPool VMA_NOT_NULL * VMA_NULLABLE VMA_LEN_IF_NOT_NULL(poolCount) pPools; /** \brief Maximum total numbers of bytes that can be copied while moving allocations to different places using transfers on CPU side, like `memcpy()`, `memmove()`. - + `VK_WHOLE_SIZE` means no limit. */ VkDeviceSize maxCpuBytesToMove; @@ -3628,7 +3652,7 @@ */ uint32_t maxCpuAllocationsToMove; /** \brief Maximum total numbers of bytes that can be copied while moving allocations to different places using transfers on GPU side, posted to `commandBuffer`. - + `VK_WHOLE_SIZE` means no limit. */ VkDeviceSize maxGpuBytesToMove; @@ -3669,7 +3693,7 @@ */ typedef struct VmaDefragmentationInfo { /** \brief Maximum total numbers of bytes that can be copied while moving allocations to different places. - + Default is `VK_WHOLE_SIZE`, which means no limit. */ VkDeviceSize maxBytesToMove; @@ -3880,13 +3904,17 @@ no longer need them using either convenience function vmaDestroyBuffer() or separately, using `vkDestroyBuffer()` and vmaFreeMemory(). -If VMA_ALLOCATOR_CREATE_KHR_DEDICATED_ALLOCATION_BIT flag was used, +If #VMA_ALLOCATOR_CREATE_KHR_DEDICATED_ALLOCATION_BIT flag was used, VK_KHR_dedicated_allocation extension is used internally to query driver whether it requires or prefers the new buffer to have dedicated allocation. If yes, and if dedicated allocation is possible (VmaAllocationCreateInfo::pool is null -and VMA_ALLOCATION_CREATE_NEVER_ALLOCATE_BIT is not used), it creates dedicated +and #VMA_ALLOCATION_CREATE_NEVER_ALLOCATE_BIT is not used), it creates dedicated allocation for this buffer, just like when using -VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT. +#VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT. + +\note This function creates a new `VkBuffer`. Sub-allocation of parts of one large buffer, +although recommended as a good practice, is out of scope of this library and could be implemented +by the user as a higher-level logic on top of VMA. */ VMA_CALL_PRE VkResult VMA_CALL_POST vmaCreateBuffer( VmaAllocator VMA_NOT_NULL allocator, @@ -4079,7 +4107,7 @@ #if defined(__APPLE__) && (defined(MAC_OS_X_VERSION_10_16) || defined(__IPHONE_14_0)) #if MAC_OS_X_VERSION_MAX_ALLOWED >= MAC_OS_X_VERSION_10_16 || __IPHONE_OS_VERSION_MAX_ALLOWED >= __IPHONE_14_0 // For C++14, usr/include/malloc/_malloc.h declares aligned_alloc()) only - // with the MacOSX11.0 SDK in Xcode 12 (which is what adds + // with the MacOSX11.0 SDK in Xcode 12 (which is what adds // MAC_OS_X_VERSION_10_16), even though the function is marked // availabe for 10.15. That's why the preprocessor checks for 10.16 but // the __builtin_available checks for 10.15. @@ -4587,7 +4615,7 @@ { VMA_SWAP(suballocType1, suballocType2); } - + switch(suballocType1) { case VMA_SUBALLOCATION_TYPE_FREE: @@ -4721,7 +4749,7 @@ size_t down = 0, up = (end - beg); while(down < up) { - const size_t mid = (down + up) / 2; + const size_t mid = down + (up - down) / 2; // Overflow-safe midpoint calculation if(cmp(*(beg+mid), key)) { down = mid + 1; @@ -4883,7 +4911,7 @@ public: const VkAllocationCallbacks* const m_pCallbacks; typedef T value_type; - + VmaStlAllocator(const VkAllocationCallbacks* pCallbacks) : m_pCallbacks(pCallbacks) { } template<typename U> VmaStlAllocator(const VmaStlAllocator<U>& src) : m_pCallbacks(src.m_pCallbacks) { } @@ -4946,12 +4974,12 @@ m_Capacity(count) { } - + // This version of the constructor is here for compatibility with pre-C++14 std::vector. // value is unused. VmaVector(size_t count, const T& value, const AllocatorT& allocator) : VmaVector(count, allocator) {} - + VmaVector(const VmaVector<T, AllocatorT>& src) : m_Allocator(src.m_Allocator), m_pArray(src.m_Count ? (T*)VmaAllocateArray<T>(src.m_Allocator.m_pCallbacks, src.m_Count) : VMA_NULL), @@ -4963,7 +4991,7 @@ memcpy(m_pArray, src.m_pArray, m_Count * sizeof(T)); } } - + ~VmaVector() { VmaFree(m_Allocator.m_pCallbacks, m_pArray); @@ -4981,12 +5009,12 @@ } return *this; } - + bool empty() const { return m_Count == 0; } size_t size() const { return m_Count; } T* data() { return m_pArray; } const T* data() const { return m_pArray; } - + T& operator[](size_t index) { VMA_HEAVY_ASSERT(index < m_Count); @@ -5022,12 +5050,12 @@ void reserve(size_t newCapacity, bool freeMemory = false) { newCapacity = VMA_MAX(newCapacity, m_Count); - + if((newCapacity < m_Capacity) && !freeMemory) { newCapacity = m_Capacity; } - + if(newCapacity != m_Capacity) { T* const newArray = newCapacity ? VmaAllocateArray<T>(m_Allocator, newCapacity) : VMA_NULL; @@ -5377,7 +5405,7 @@ uint32_t Capacity; uint32_t FirstFreeIndex; }; - + const VkAllocationCallbacks* m_pAllocationCallbacks; const uint32_t m_FirstBlockCapacity; VmaVector< ItemBlock, VmaStlAllocator<ItemBlock> > m_ItemBlocks; @@ -5435,11 +5463,11 @@ for(size_t i = m_ItemBlocks.size(); i--; ) { ItemBlock& block = m_ItemBlocks[i]; - + // Casting to union. Item* pItemPtr; memcpy(&pItemPtr, &ptr, sizeof(pItemPtr)); - + // Check if pItemPtr is in address range of this block. if((pItemPtr >= block.pItems) && (pItemPtr < block.pItems + block.Capacity)) { @@ -5516,7 +5544,7 @@ ItemType* PushFront(const T& value); void PopBack(); void PopFront(); - + // Item can be null - it means PushBack. ItemType* InsertBefore(ItemType* pItem); // Item can be null - it means PushFront. @@ -5826,7 +5854,7 @@ VMA_HEAVY_ASSERT(m_pList == rhs.m_pList); return m_pItem != rhs.m_pItem; } - + private: VmaRawList<T>* m_pList; VmaListItem<T>* m_pItem; @@ -5854,7 +5882,7 @@ m_pItem(src.m_pItem) { } - + const T& operator*() const { VMA_HEAVY_ASSERT(m_pItem != VMA_NULL); @@ -5909,7 +5937,7 @@ VMA_HEAVY_ASSERT(m_pList == rhs.m_pList); return m_pItem != rhs.m_pItem; } - + private: const_iterator(const VmaRawList<T>* pList, const VmaListItem<T>* pItem) : m_pList(pList), @@ -5988,7 +6016,7 @@ void insert(const PairType& pair); iterator find(const KeyT& key); void erase(iterator it); - + private: VmaVector< PairType, VmaStlAllocator<PairType> > m_Vector; }; @@ -6137,7 +6165,7 @@ void ChangeBlockAllocation( VmaAllocator hAllocator, VmaDeviceMemoryBlock* block, - VkDeviceSize offset); + VkDeviceSize offset); void ChangeOffset(VkDeviceSize newOffset); @@ -6180,7 +6208,7 @@ bool IsPersistentMap() const { return (m_MapCount & MAP_COUNT_FLAG_PERSISTENT_MAP) != 0; } void* GetMappedData() const; bool CanBecomeLost() const; - + uint32_t GetLastUseFrameIndex() const { return m_LastUseFrameIndex.load(); @@ -6193,7 +6221,7 @@ - If hAllocation.LastUseFrameIndex + frameInUseCount < allocator.CurrentFrameIndex, makes it lost by setting LastUseFrameIndex = VMA_FRAME_INDEX_LOST and returns true. - Else, returns false. - + If hAllocation is already lost, assert - you should not call it then. If hAllocation was not created with CAN_BECOME_LOST_BIT, assert. */ @@ -6489,7 +6517,7 @@ //////////////////////////////////////////////////////////////////////////////// // For defragmentation - + bool IsBufferImageGranularityConflictPossible( VkDeviceSize bufferImageGranularity, VmaSuballocationType& inOutPrevSuballocType) const; @@ -6700,7 +6728,7 @@ SuballocationVectorType& AccessSuballocations2nd() { return m_1stVectorIndex ? m_Suballocations0 : m_Suballocations1; } const SuballocationVectorType& AccessSuballocations1st() const { return m_1stVectorIndex ? m_Suballocations1 : m_Suballocations0; } const SuballocationVectorType& AccessSuballocations2nd() const { return m_1stVectorIndex ? m_Suballocations0 : m_Suballocations1; } - + // Number of items in 1st vector with hAllocation = null at the beginning. size_t m_1stNullItemsBeginCount; // Number of other items in 1st vector with hAllocation = null somewhere in the middle. @@ -6911,7 +6939,7 @@ uint32_t algorithm); // Always call before destruction. void Destroy(VmaAllocator allocator); - + VmaPool GetParentPool() const { return m_hParentPool; } VkDeviceMemory GetDeviceMemory() const { return m_hMemory; } uint32_t GetMemoryTypeIndex() const { return m_MemoryTypeIndex; } @@ -7456,7 +7484,7 @@ } } } - + if(bestIndex != SIZE_MAX) { outBlockInfoIndex = m_FreeSpaces[bestIndex].blockInfoIndex; @@ -7833,7 +7861,7 @@ VkAllocationCallbacks m_AllocationCallbacks; VmaDeviceMemoryCallbacks m_DeviceMemoryCallbacks; VmaAllocationObjectAllocator m_AllocationObjectAllocator; - + // Each bit (1 << i) is set if HeapSizeLimit is enabled for that heap, so cannot allocate more than the heap size. uint32_t m_HeapSizeLimitMask; @@ -8033,7 +8061,7 @@ VkPhysicalDevice m_PhysicalDevice; VMA_ATOMIC_UINT32 m_CurrentFrameIndex; VMA_ATOMIC_UINT32 m_GpuDefragmentationMemoryTypeBits; // UINT32_MAX means uninitialized. - + VMA_RW_MUTEX m_PoolsMutex; // Protected by m_PoolsMutex. Sorted by pointer value. VmaVector<VmaPool, VmaStlAllocator<VmaPool> > m_Pools; @@ -8254,10 +8282,10 @@ void BeginObject(bool singleLine = false); void EndObject(); - + void BeginArray(bool singleLine = false); void EndArray(); - + void WriteString(const char* pStr); void BeginString(const char* pStr = VMA_NULL); void ContinueString(const char* pStr); @@ -8265,7 +8293,7 @@ void ContinueString(uint64_t n); void ContinueString_Pointer(const void* ptr); void EndString(const char* pStr = VMA_NULL); - + void WriteNumber(uint32_t n); void WriteNumber(uint64_t n); void WriteBool(bool b); @@ -8513,7 +8541,7 @@ if(!m_Stack.empty() && !m_Stack.back().singleLineMode) { m_SB.AddNewLine(); - + size_t count = m_Stack.size(); if(count > 0 && oneLess) { @@ -8927,7 +8955,7 @@ VmaAllocation hAllocation) const { json.BeginObject(true); - + json.WriteString("Offset"); json.WriteNumber(offset); @@ -8941,7 +8969,7 @@ VkDeviceSize size) const { json.BeginObject(true); - + json.WriteString("Offset"); json.WriteNumber(offset); @@ -9001,7 +9029,7 @@ bool VmaBlockMetadata_Generic::Validate() const { VMA_VALIDATE(!m_Suballocations.empty()); - + // Expected offset of new suballocation as calculated from previous ones. VkDeviceSize calculatedOffset = 0; // Expected number of free suballocations as calculated from traversing their list. @@ -9019,7 +9047,7 @@ ++suballocItem) { const VmaSuballocation& subAlloc = *suballocItem; - + // Actual offset of this suballocation doesn't match expected one. VMA_VALIDATE(subAlloc.offset == calculatedOffset); @@ -9062,7 +9090,7 @@ for(size_t i = 0; i < m_FreeSuballocationsBySize.size(); ++i) { VmaSuballocationList::iterator suballocItem = m_FreeSuballocationsBySize[i]; - + // Only free suballocations can be registered in m_FreeSuballocationsBySize. VMA_VALIDATE(suballocItem->type == VMA_SUBALLOCATION_TYPE_FREE); // They must be sorted by size ascending. @@ -9104,7 +9132,7 @@ const uint32_t rangeCount = (uint32_t)m_Suballocations.size(); outInfo.allocationCount = rangeCount - m_FreeCount; outInfo.unusedRangeCount = m_FreeCount; - + outInfo.unusedBytes = m_SumFreeSize; outInfo.usedBytes = GetSize() - outInfo.unusedBytes; @@ -9363,7 +9391,7 @@ VMA_HEAVY_ASSERT(Validate()); VMA_ASSERT(pAllocationRequest->item != m_Suballocations.end()); VMA_ASSERT(pAllocationRequest->item->type == VMA_SUBALLOCATION_TYPE_FREE); - + return true; } @@ -9540,7 +9568,7 @@ VMA_ASSERT(allocType != VMA_SUBALLOCATION_TYPE_FREE); VMA_ASSERT(suballocItem != m_Suballocations.cend()); VMA_ASSERT(pOffset != VMA_NULL); - + *itemsToMakeLostCount = 0; *pSumFreeSize = 0; *pSumItemSize = 0; @@ -9573,13 +9601,13 @@ // Start from offset equal to beginning of this suballocation. *pOffset = suballocItem->offset; - + // Apply VMA_DEBUG_MARGIN at the beginning. if(VMA_DEBUG_MARGIN > 0) { *pOffset += VMA_DEBUG_MARGIN; } - + // Apply alignment. *pOffset = VmaAlignUp(*pOffset, allocAlignment); @@ -9610,14 +9638,14 @@ *pOffset = VmaAlignUp(*pOffset, bufferImageGranularity); } } - + // Now that we have final *pOffset, check if we are past suballocItem. // If yes, return false - this function should be called for another suballocItem as starting point. if(*pOffset >= suballocItem->offset + suballocItem->size) { return false; } - + // Calculate padding at the beginning based on current offset. const VkDeviceSize paddingBegin = *pOffset - suballocItem->offset; @@ -9716,16 +9744,16 @@ // Start from offset equal to beginning of this suballocation. *pOffset = suballoc.offset; - + // Apply VMA_DEBUG_MARGIN at the beginning. if(VMA_DEBUG_MARGIN > 0) { *pOffset += VMA_DEBUG_MARGIN; } - + // Apply alignment. *pOffset = VmaAlignUp(*pOffset, allocAlignment); - + // Check previous suballocations for BufferImageGranularity conflicts. // Make bigger alignment if necessary. if(bufferImageGranularity > 1) @@ -9753,7 +9781,7 @@ *pOffset = VmaAlignUp(*pOffset, bufferImageGranularity); } } - + // Calculate padding at the beginning based on current offset. const VkDeviceSize paddingBegin = *pOffset - suballoc.offset; @@ -9800,7 +9828,7 @@ { VMA_ASSERT(item != m_Suballocations.end()); VMA_ASSERT(item->type == VMA_SUBALLOCATION_TYPE_FREE); - + VmaSuballocationList::iterator nextItem = item; ++nextItem; VMA_ASSERT(nextItem != m_Suballocations.end()); @@ -9817,7 +9845,7 @@ VmaSuballocation& suballoc = *suballocItem; suballoc.type = VMA_SUBALLOCATION_TYPE_FREE; suballoc.hAllocation = VK_NULL_HANDLE; - + // Update totals. ++m_FreeCount; m_SumFreeSize += suballoc.size; @@ -9825,7 +9853,7 @@ // Merge with previous and/or next suballocation if it's also free. bool mergeWithNext = false; bool mergeWithPrev = false; - + VmaSuballocationList::iterator nextItem = suballocItem; ++nextItem; if((nextItem != m_Suballocations.end()) && (nextItem->type == VMA_SUBALLOCATION_TYPE_FREE)) @@ -10123,7 +10151,7 @@ { return size; } - + const SuballocationVectorType& suballocations1st = AccessSuballocations1st(); switch(m_2ndVectorMode) @@ -10210,7 +10238,7 @@ if(nextAlloc2ndIndex < suballoc2ndCount) { const VmaSuballocation& suballoc = suballocations2nd[nextAlloc2ndIndex]; - + // 1. Process free space before this allocation. if(lastOffset < suballoc.offset) { @@ -10221,13 +10249,13 @@ outInfo.unusedRangeSizeMin = VMA_MIN(outInfo.unusedRangeSizeMin, unusedRangeSize); outInfo.unusedRangeSizeMax = VMA_MIN(outInfo.unusedRangeSizeMax, unusedRangeSize); } - + // 2. Process this allocation. // There is allocation with suballoc.offset, suballoc.size. outInfo.usedBytes += suballoc.size; outInfo.allocationSizeMin = VMA_MIN(outInfo.allocationSizeMin, suballoc.size); outInfo.allocationSizeMax = VMA_MIN(outInfo.allocationSizeMax, suballoc.size); - + // 3. Prepare for next iteration. lastOffset = suballoc.offset + suballoc.size; ++nextAlloc2ndIndex; @@ -10267,7 +10295,7 @@ if(nextAlloc1stIndex < suballoc1stCount) { const VmaSuballocation& suballoc = suballocations1st[nextAlloc1stIndex]; - + // 1. Process free space before this allocation. if(lastOffset < suballoc.offset) { @@ -10278,13 +10306,13 @@ outInfo.unusedRangeSizeMin = VMA_MIN(outInfo.unusedRangeSizeMin, unusedRangeSize); outInfo.unusedRangeSizeMax = VMA_MIN(outInfo.unusedRangeSizeMax, unusedRangeSize); } - + // 2. Process this allocation. // There is allocation with suballoc.offset, suballoc.size. outInfo.usedBytes += suballoc.size; outInfo.allocationSizeMin = VMA_MIN(outInfo.allocationSizeMin, suballoc.size); outInfo.allocationSizeMax = VMA_MIN(outInfo.allocationSizeMax, suballoc.size); - + // 3. Prepare for next iteration. lastOffset = suballoc.offset + suballoc.size; ++nextAlloc1stIndex; @@ -10323,7 +10351,7 @@ if(nextAlloc2ndIndex != SIZE_MAX) { const VmaSuballocation& suballoc = suballocations2nd[nextAlloc2ndIndex]; - + // 1. Process free space before this allocation. if(lastOffset < suballoc.offset) { @@ -10334,13 +10362,13 @@ outInfo.unusedRangeSizeMin = VMA_MIN(outInfo.unusedRangeSizeMin, unusedRangeSize); outInfo.unusedRangeSizeMax = VMA_MIN(outInfo.unusedRangeSizeMax, unusedRangeSize); } - + // 2. Process this allocation. // There is allocation with suballoc.offset, suballoc.size. outInfo.usedBytes += suballoc.size; outInfo.allocationSizeMin = VMA_MIN(outInfo.allocationSizeMin, suballoc.size); outInfo.allocationSizeMax = VMA_MIN(outInfo.allocationSizeMax, suballoc.size); - + // 3. Prepare for next iteration. lastOffset = suballoc.offset + suballoc.size; --nextAlloc2ndIndex; @@ -10396,7 +10424,7 @@ if(nextAlloc2ndIndex < suballoc2ndCount) { const VmaSuballocation& suballoc = suballocations2nd[nextAlloc2ndIndex]; - + // 1. Process free space before this allocation. if(lastOffset < suballoc.offset) { @@ -10406,11 +10434,11 @@ ++inoutStats.unusedRangeCount; inoutStats.unusedRangeSizeMax = VMA_MAX(inoutStats.unusedRangeSizeMax, unusedRangeSize); } - + // 2. Process this allocation. // There is allocation with suballoc.offset, suballoc.size. ++inoutStats.allocationCount; - + // 3. Prepare for next iteration. lastOffset = suballoc.offset + suballoc.size; ++nextAlloc2ndIndex; @@ -10449,7 +10477,7 @@ if(nextAlloc1stIndex < suballoc1stCount) { const VmaSuballocation& suballoc = suballocations1st[nextAlloc1stIndex]; - + // 1. Process free space before this allocation. if(lastOffset < suballoc.offset) { @@ -10459,11 +10487,11 @@ ++inoutStats.unusedRangeCount; inoutStats.unusedRangeSizeMax = VMA_MAX(inoutStats.unusedRangeSizeMax, unusedRangeSize); } - + // 2. Process this allocation. // There is allocation with suballoc.offset, suballoc.size. ++inoutStats.allocationCount; - + // 3. Prepare for next iteration. lastOffset = suballoc.offset + suballoc.size; ++nextAlloc1stIndex; @@ -10501,7 +10529,7 @@ if(nextAlloc2ndIndex != SIZE_MAX) { const VmaSuballocation& suballoc = suballocations2nd[nextAlloc2ndIndex]; - + // 1. Process free space before this allocation. if(lastOffset < suballoc.offset) { @@ -10511,11 +10539,11 @@ ++inoutStats.unusedRangeCount; inoutStats.unusedRangeSizeMax = VMA_MAX(inoutStats.unusedRangeSizeMax, unusedRangeSize); } - + // 2. Process this allocation. // There is allocation with suballoc.offset, suballoc.size. ++inoutStats.allocationCount; - + // 3. Prepare for next iteration. lastOffset = suballoc.offset + suballoc.size; --nextAlloc2ndIndex; @@ -10573,19 +10601,19 @@ if(nextAlloc2ndIndex < suballoc2ndCount) { const VmaSuballocation& suballoc = suballocations2nd[nextAlloc2ndIndex]; - + // 1. Process free space before this allocation. if(lastOffset < suballoc.offset) { // There is free space from lastOffset to suballoc.offset. ++unusedRangeCount; } - + // 2. Process this allocation. // There is allocation with suballoc.offset, suballoc.size. ++alloc2ndCount; usedBytes += suballoc.size; - + // 3. Prepare for next iteration. lastOffset = suballoc.offset + suballoc.size; ++nextAlloc2ndIndex; @@ -10622,19 +10650,19 @@ if(nextAlloc1stIndex < suballoc1stCount) { const VmaSuballocation& suballoc = suballocations1st[nextAlloc1stIndex]; - + // 1. Process free space before this allocation. if(lastOffset < suballoc.offset) { // There is free space from lastOffset to suballoc.offset. ++unusedRangeCount; } - + // 2. Process this allocation. // There is allocation with suballoc.offset, suballoc.size. ++alloc1stCount; usedBytes += suballoc.size; - + // 3. Prepare for next iteration. lastOffset = suballoc.offset + suballoc.size; ++nextAlloc1stIndex; @@ -10669,19 +10697,19 @@ if(nextAlloc2ndIndex != SIZE_MAX) { const VmaSuballocation& suballoc = suballocations2nd[nextAlloc2ndIndex]; - + // 1. Process free space before this allocation. if(lastOffset < suballoc.offset) { // There is free space from lastOffset to suballoc.offset. ++unusedRangeCount; } - + // 2. Process this allocation. // There is allocation with suballoc.offset, suballoc.size. ++alloc2ndCount; usedBytes += suballoc.size; - + // 3. Prepare for next iteration. lastOffset = suballoc.offset + suballoc.size; --nextAlloc2ndIndex; @@ -10724,7 +10752,7 @@ if(nextAlloc2ndIndex < suballoc2ndCount) { const VmaSuballocation& suballoc = suballocations2nd[nextAlloc2ndIndex]; - + // 1. Process free space before this allocation. if(lastOffset < suballoc.offset) { @@ -10732,11 +10760,11 @@ const VkDeviceSize unusedRangeSize = suballoc.offset - lastOffset; PrintDetailedMap_UnusedRange(json, lastOffset, unusedRangeSize); } - + // 2. Process this allocation. // There is allocation with suballoc.offset, suballoc.size. PrintDetailedMap_Allocation(json, suballoc.offset, suballoc.hAllocation); - + // 3. Prepare for next iteration. lastOffset = suballoc.offset + suballoc.size; ++nextAlloc2ndIndex; @@ -10771,7 +10799,7 @@ if(nextAlloc1stIndex < suballoc1stCount) { const VmaSuballocation& suballoc = suballocations1st[nextAlloc1stIndex]; - + // 1. Process free space before this allocation. if(lastOffset < suballoc.offset) { @@ -10779,11 +10807,11 @@ const VkDeviceSize unusedRangeSize = suballoc.offset - lastOffset; PrintDetailedMap_UnusedRange(json, lastOffset, unusedRangeSize); } - + // 2. Process this allocation. // There is allocation with suballoc.offset, suballoc.size. PrintDetailedMap_Allocation(json, suballoc.offset, suballoc.hAllocation); - + // 3. Prepare for next iteration. lastOffset = suballoc.offset + suballoc.size; ++nextAlloc1stIndex; @@ -10819,7 +10847,7 @@ if(nextAlloc2ndIndex != SIZE_MAX) { const VmaSuballocation& suballoc = suballocations2nd[nextAlloc2ndIndex]; - + // 1. Process free space before this allocation. if(lastOffset < suballoc.offset) { @@ -10827,11 +10855,11 @@ const VkDeviceSize unusedRangeSize = suballoc.offset - lastOffset; PrintDetailedMap_UnusedRange(json, lastOffset, unusedRangeSize); } - + // 2. Process this allocation. // There is allocation with suballoc.offset, suballoc.size. PrintDetailedMap_Allocation(json, suballoc.offset, suballoc.hAllocation); - + // 3. Prepare for next iteration. lastOffset = suballoc.offset + suballoc.size; --nextAlloc2ndIndex; @@ -11283,7 +11311,7 @@ } VMA_ASSERT(m_2ndVectorMode == SECOND_VECTOR_EMPTY || m_2ndVectorMode == SECOND_VECTOR_RING_BUFFER); - + // We always start from 1st. SuballocationVectorType* suballocations = &AccessSuballocations1st(); size_t index = m_1stNullItemsBeginCount; @@ -11332,14 +11360,14 @@ CleanupAfterFree(); //VMA_HEAVY_ASSERT(Validate()); // Already called by ClanupAfterFree(). - + return true; } uint32_t VmaBlockMetadata_Linear::MakeAllocationsLost(uint32_t currentFrameIndex, uint32_t frameInUseCount) { uint32_t lostAllocationCount = 0; - + SuballocationVectorType& suballocations1st = AccessSuballocations1st(); for(size_t i = m_1stNullItemsBeginCount, count = suballocations1st.size(); i < count; ++i) { @@ -11763,7 +11791,7 @@ node = node->free.next) { VMA_VALIDATE(node->type == Node::TYPE_FREE); - + if(node->free.next == VMA_NULL) { VMA_VALIDATE(m_FreeList[level].back == node); @@ -11949,7 +11977,7 @@ const uint32_t targetLevel = AllocSizeToLevel(allocSize); uint32_t currLevel = (uint32_t)(uintptr_t)request.customData; - + Node* currNode = m_FreeList[currLevel].front; VMA_ASSERT(currNode != VMA_NULL && currNode->type == Node::TYPE_FREE); while(currNode->offset != request.offset) @@ -11957,14 +11985,14 @@ currNode = currNode->free.next; VMA_ASSERT(currNode != VMA_NULL && currNode->type == Node::TYPE_FREE); } - + // Go down, splitting free nodes. while(currLevel < targetLevel) { // currNode is already first free node at currLevel. // Remove it from list of free nodes at this currLevel. RemoveFromFreeList(currLevel, currNode); - + const uint32_t childrenLevel = currLevel + 1; // Create two free sub-nodes. @@ -12126,7 +12154,7 @@ vma_delete(GetAllocationCallbacks(), node->buddy); vma_delete(GetAllocationCallbacks(), node); parent->type = Node::TYPE_FREE; - + node = parent; --level; //m_SumFreeSize += LevelToNodeSize(level) % 2; // Useful only when level node sizes can be non power of 2. @@ -12240,7 +12268,7 @@ PrintDetailedMap_UnusedRange(json, node->offset, levelNodeSize); break; case Node::TYPE_ALLOCATION: - { + { PrintDetailedMap_Allocation(json, node->offset, node->allocation.alloc); const VkDeviceSize allocSize = node->allocation.alloc->GetSize(); if(allocSize < levelNodeSize) @@ -12329,7 +12357,7 @@ { VMA_VALIDATE((m_hMemory != VK_NULL_HANDLE) && (m_pMetadata->GetSize() != 0)); - + return m_pMetadata->Validate(); } @@ -12548,7 +12576,7 @@ { const VkAllocationCallbacks* allocs = m_BlockVector.GetAllocator()->GetAllocationCallbacks(); VmaFreeString(allocs, m_Name); - + if(pName != VMA_NULL) { m_Name = VmaCreateStringCopy(allocs, pName); @@ -12712,7 +12740,7 @@ bool canMakeOtherLost = (createInfo.flags & VMA_ALLOCATION_CREATE_CAN_MAKE_OTHER_LOST_BIT) != 0; const bool mapped = (createInfo.flags & VMA_ALLOCATION_CREATE_MAPPED_BIT) != 0; const bool isUserDataString = (createInfo.flags & VMA_ALLOCATION_CREATE_USER_DATA_COPY_STRING_BIT) != 0; - + VkDeviceSize freeMemory; { const uint32_t heapIndex = m_hAllocator->MemoryTypeIndexToHeapIndex(m_MemoryTypeIndex); @@ -12720,7 +12748,7 @@ m_hAllocator->GetBudget(&heapBudget, heapIndex, 1); freeMemory = (heapBudget.usage < heapBudget.budget) ? (heapBudget.budget - heapBudget.usage) : 0; } - + const bool canFallbackToDedicated = !IsCustomPool(); const bool canCreateNewBlock = ((createInfo.flags & VMA_ALLOCATION_CREATE_NEVER_ALLOCATE_BIT) == 0) && @@ -13234,7 +13262,7 @@ return res; } } - + *pAllocation = m_hAllocator->m_AllocationObjectAllocator.Allocate(currentFrameIndex, isUserDataString); pBlock->m_pMetadata->Alloc(currRequest, suballocType, size, *pAllocation); UpdateHasEmptyBlock(); @@ -13629,7 +13657,7 @@ VkCommandBuffer commandBuffer) { pCtx->res = VK_SUCCESS; - + const VkMemoryPropertyFlags memPropFlags = m_hAllocator->m_MemProps.memoryTypes[m_MemoryTypeIndex].propertyFlags; const bool isHostVisible = (memPropFlags & VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT) != 0; @@ -13708,13 +13736,13 @@ { if(m_hAllocator->m_UseMutex) m_Mutex.UnlockWrite(); - + if(pCtx->res >= VK_SUCCESS && !pCtx->defragmentationMoves.empty()) pCtx->res = VK_NOT_READY; return; } - + if(pCtx->res >= VK_SUCCESS) { if(defragmentOnGpu) @@ -13775,7 +13803,7 @@ VmaDefragmentationPassMoveInfo* pMove, uint32_t maxMoves) { VmaMutexLockWrite lock(m_Mutex, m_hAllocator->m_UseMutex); - + const uint32_t moveCount = VMA_MIN(uint32_t(pCtx->defragmentationMoves.size()) - pCtx->defragmentationMovesProcessed, maxMoves); for(uint32_t i = 0; i < moveCount; ++ i) @@ -13799,7 +13827,7 @@ VmaDefragmentationStats* pStats) { VmaMutexLockWrite lock(m_Mutex, m_hAllocator->m_UseMutex); - + for(uint32_t i = pCtx->defragmentationMovesCommitted; i < pCtx->defragmentationMovesProcessed; ++ i) { const VmaDefragmentationMove &move = pCtx->defragmentationMoves[i]; @@ -14018,7 +14046,7 @@ srcAllocIndex = m_Blocks[srcBlockIndex]->m_Allocations.size() - 1; } } - + BlockInfo* pSrcBlockInfo = m_Blocks[srcBlockIndex]; AllocationInfo& allocInfo = pSrcBlockInfo->m_Allocations[srcAllocIndex]; @@ -14078,7 +14106,7 @@ pSrcBlockInfo->m_pBlock->m_pMetadata->FreeAtOffset(srcOffset); allocInfo.m_hAllocation->ChangeBlockAllocation(m_hAllocator, pDstBlockInfo->m_pBlock, dstAllocRequest.offset); } - + if(allocInfo.m_pChanged != VMA_NULL) { *allocInfo.m_pChanged = VK_TRUE; @@ -14159,7 +14187,7 @@ } pBlockInfo->CalcHasNonMovableAllocations(); - + // This is a choice based on research. // Option 1: pBlockInfo->SortAllocationsByOffsetDescending(); @@ -14309,7 +14337,7 @@ suballoc.hAllocation->ChangeOffset(dstAllocOffset); m_BytesMoved += srcAllocSize; ++m_AllocationsMoved; - + VmaSuballocationList::iterator nextSuballocIt = srcSuballocIt; ++nextSuballocIt; pSrcMetadata->m_Suballocations.erase(srcSuballocIt); @@ -14322,7 +14350,7 @@ move.srcOffset = srcAllocOffset; move.dstOffset = dstAllocOffset; move.size = srcAllocSize; - + moves.push_back(move); } // Different block @@ -14350,7 +14378,7 @@ move.srcOffset = srcAllocOffset; move.dstOffset = dstAllocOffset; move.size = srcAllocSize; - + moves.push_back(move); } } @@ -14405,13 +14433,13 @@ m_BytesMoved += srcAllocSize; ++m_AllocationsMoved; ++srcSuballocIt; - + move.srcBlockIndex = srcOrigBlockIndex; move.dstBlockIndex = dstOrigBlockIndex; move.srcOffset = srcAllocOffset; move.dstOffset = dstAllocOffset; move.size = srcAllocSize; - + moves.push_back(move); } } @@ -14442,7 +14470,7 @@ move.srcOffset = srcAllocOffset; move.dstOffset = dstAllocOffset; move.size = srcAllocSize; - + moves.push_back(move); } } @@ -14450,7 +14478,7 @@ } m_BlockInfos.clear(); - + PostprocessMetadata(); return VK_SUCCESS; @@ -14492,7 +14520,7 @@ VmaBlockMetadata_Generic* const pMetadata = (VmaBlockMetadata_Generic*)m_pBlockVector->GetBlock(blockIndex)->m_pMetadata; const VkDeviceSize blockSize = pMetadata->GetSize(); - + // No allocations in this block - entire area is free. if(pMetadata->m_Suballocations.empty()) { @@ -14706,7 +14734,7 @@ if(pool->m_BlockVector.GetAlgorithm() == 0) { VmaBlockVectorDefragmentationContext* pBlockVectorDefragCtx = VMA_NULL; - + for(size_t i = m_CustomPoolContexts.size(); i--; ) { if(m_CustomPoolContexts[i]->GetCustomPool() == pool) @@ -14715,7 +14743,7 @@ break; } } - + if(!pBlockVectorDefragCtx) { pBlockVectorDefragCtx = vma_new(m_hAllocator, VmaBlockVectorDefragmentationContext)( @@ -15579,7 +15607,7 @@ string thread_id_as_string = thread_id_to_string_converter.str(); outParams.threadId = static_cast<uint32_t>(std::stoi(thread_id_as_string.c_str())); #endif - + auto current_time = std::chrono::high_resolution_clock::now(); outParams.time = std::chrono::duration<double, std::chrono::seconds::period>(current_time - m_RecordingStartTime).count(); @@ -15713,7 +15741,7 @@ memset(&m_DeviceMemoryCallbacks, 0 ,sizeof(m_DeviceMemoryCallbacks)); memset(&m_PhysicalDeviceProperties, 0, sizeof(m_PhysicalDeviceProperties)); memset(&m_MemProps, 0, sizeof(m_MemProps)); - + memset(&m_pBlockVectors, 0, sizeof(m_pBlockVectors)); memset(&m_pDedicatedAllocations, 0, sizeof(m_pDedicatedAllocations)); memset(&m_VulkanFunctions, 0, sizeof(m_VulkanFunctions)); @@ -15826,7 +15854,7 @@ vma_delete(this, m_pRecorder); } #endif - + VMA_ASSERT(m_Pools.empty()); for(size_t i = GetMemoryTypeCount(); i--; ) @@ -16295,7 +16323,7 @@ { VmaAllocation currAlloc = pAllocations[allocIndex]; VkDeviceMemory hMemory = currAlloc->GetMemory(); - + /* There is no need to call this, because Vulkan spec allows to skip vkUnmapMemory before vkFreeMemory. @@ -16305,7 +16333,7 @@ (*m_VulkanFunctions.vkUnmapMemory)(m_hDevice, hMemory); } */ - + FreeVulkanMemory(memTypeIndex, currAlloc->GetSize(), hMemory); m_Budget.RemoveAllocation(MemoryTypeIndexToHeapIndex(memTypeIndex), currAlloc->GetSize()); currAlloc->SetUserData(this, VMA_NULL); @@ -16548,7 +16576,7 @@ alignmentForMemType = VMA_MAX( vkMemReq.alignment, GetMemoryTypeMinAlignment(memTypeIndex)); - + res = AllocateMemoryOfType( vkMemReq.size, alignmentForMemType, @@ -16660,7 +16688,7 @@ InitStatInfo(pStats->memoryType[i]); for(size_t i = 0; i < VK_MAX_MEMORY_HEAPS; ++i) InitStatInfo(pStats->memoryHeap[i]); - + // Process default pools. for(uint32_t memTypeIndex = 0; memTypeIndex < GetMemoryTypeCount(); ++memTypeIndex) { @@ -16805,7 +16833,7 @@ VmaDefragmentationContext context) { return context->DefragmentPassEnd(); - + } void VmaAllocator_T::GetAllocationInfo(VmaAllocation hAllocation, VmaAllocationInfo* pAllocationInfo) @@ -17333,7 +17361,7 @@ typedef VmaStlAllocator<VkMappedMemoryRange> RangeAllocator; typedef VmaSmallVector<VkMappedMemoryRange, RangeAllocator, 16> RangeVector; RangeVector ranges = RangeVector(RangeAllocator(GetAllocationCallbacks())); - + for(uint32_t allocIndex = 0; allocIndex < allocationCount; ++allocIndex) { const VmaAllocation alloc = allocations[allocIndex]; @@ -17379,7 +17407,7 @@ } VkDeviceMemory hMemory = allocation->GetMemory(); - + /* There is no need to call this, because Vulkan spec allows to skip vkUnmapMemory before vkFreeMemory. @@ -17389,7 +17417,7 @@ (*m_VulkanFunctions.vkUnmapMemory)(m_hDevice, hMemory); } */ - + FreeVulkanMemory(memTypeIndex, allocation->GetSize(), hMemory); VMA_DEBUG_LOG(" Freed DedicatedMemory MemoryTypeIndex=%u", memTypeIndex); @@ -17601,7 +17629,7 @@ json.BeginString("Type "); json.ContinueString(memTypeIndex); json.EndString(); - + json.BeginArray(); for(size_t i = 0; i < pDedicatedAllocVector->size(); ++i) @@ -17782,7 +17810,7 @@ json.WriteString("Total"); VmaPrintStatInfo(json, stats.total); - + for(uint32_t heapIndex = 0; heapIndex < allocator->GetMemoryHeapCount(); ++heapIndex) { json.BeginString("Heap "); @@ -17931,7 +17959,7 @@ { memoryTypeBits &= pAllocationCreateInfo->memoryTypeBits; } - + uint32_t requiredFlags = pAllocationCreateInfo->requiredFlags; uint32_t preferredFlags = pAllocationCreateInfo->preferredFlags; uint32_t notPreferredFlags = 0; @@ -18084,20 +18112,20 @@ VmaPool* pPool) { VMA_ASSERT(allocator && pCreateInfo && pPool); - + VMA_DEBUG_LOG("vmaCreatePool"); - + VMA_DEBUG_GLOBAL_MUTEX_LOCK - + VkResult res = allocator->CreatePool(pCreateInfo, pPool); - + #if VMA_RECORDING_ENABLED if(allocator->GetRecorder() != VMA_NULL) { allocator->GetRecorder()->RecordCreatePool(allocator->GetCurrentFrameIndex(), *pCreateInfo, *pPool); } #endif - + return res; } @@ -18106,16 +18134,16 @@ VmaPool pool) { VMA_ASSERT(allocator); - + if(pool == VK_NULL_HANDLE) { return; } - + VMA_DEBUG_LOG("vmaDestroyPool"); - + VMA_DEBUG_GLOBAL_MUTEX_LOCK - + #if VMA_RECORDING_ENABLED if(allocator->GetRecorder() != VMA_NULL) { @@ -18174,7 +18202,7 @@ const char** ppName) { VMA_ASSERT(allocator && pool && ppName); - + VMA_DEBUG_LOG("vmaGetPoolName"); VMA_DEBUG_GLOBAL_MUTEX_LOCK @@ -18238,7 +18266,7 @@ *pAllocation); } #endif - + if(pAllocationInfo != VMA_NULL && result == VK_SUCCESS) { allocator->GetAllocationInfo(*pAllocation, pAllocationInfo); @@ -18289,7 +18317,7 @@ pAllocations); } #endif - + if(pAllocationInfo != VMA_NULL && result == VK_SUCCESS) { for(size_t i = 0; i < allocationCount; ++i) @@ -18411,14 +18439,14 @@ VmaAllocation allocation) { VMA_ASSERT(allocator); - + if(allocation == VK_NULL_HANDLE) { return; } - + VMA_DEBUG_LOG("vmaFreeMemory"); - + VMA_DEBUG_GLOBAL_MUTEX_LOCK #if VMA_RECORDING_ENABLED @@ -18429,7 +18457,7 @@ allocation); } #endif - + allocator->FreeMemory( 1, // allocationCount &allocation); @@ -18446,9 +18474,9 @@ } VMA_ASSERT(allocator); - + VMA_DEBUG_LOG("vmaFreeMemoryPages"); - + VMA_DEBUG_GLOBAL_MUTEX_LOCK #if VMA_RECORDING_ENABLED @@ -18460,7 +18488,7 @@ pAllocations); } #endif - + allocator->FreeMemory(allocationCount, pAllocations); } @@ -18470,9 +18498,9 @@ VkDeviceSize newSize) { VMA_ASSERT(allocator && allocation); - + VMA_DEBUG_LOG("vmaResizeAllocation"); - + VMA_DEBUG_GLOBAL_MUTEX_LOCK return allocator->ResizeAllocation(allocation, newSize); @@ -18938,9 +18966,9 @@ VMA_ASSERT(0 && "Creating a buffer with VK_BUFFER_USAGE_SHADER_DEVICE_ADDRESS_BIT is not valid if VMA_ALLOCATOR_CREATE_BUFFER_DEVICE_ADDRESS_BIT was not used."); return VK_ERROR_VALIDATION_FAILED_EXT; } - + VMA_DEBUG_LOG("vmaCreateBuffer"); - + VMA_DEBUG_GLOBAL_MUTEX_LOCK *pBuffer = VK_NULL_HANDLE; @@ -19095,7 +19123,7 @@ VmaSuballocationType suballocType = pImageCreateInfo->tiling == VK_IMAGE_TILING_OPTIMAL ? VMA_SUBALLOCATION_TYPE_IMAGE_OPTIMAL : VMA_SUBALLOCATION_TYPE_IMAGE_LINEAR; - + // 2. Allocate memory using allocator. VkMemoryRequirements vkMemReq = {}; bool requiresDedicatedAllocation = false;
changelog.md view
@@ -2,6 +2,12 @@ ## WIP +## [0.3.8] - 2020-11-12++- Bump VMA+ - Documentation chages+ - Potential overflow fix https://github.com/GPUOpen-LibrariesAndSDKs/VulkanMemoryAllocator/issues/153+ ## [0.3.7.1] - 2020-11-01 - Bump VMA, just documentation changes
package.yaml view
@@ -1,5 +1,5 @@ name: VulkanMemoryAllocator-version: "0.3.7.1"+version: "0.3.8" synopsis: Bindings to the VulkanMemoryAllocator library category: Graphics maintainer: Joe Hermaszewski <live.long.and.prosper@monoid.al>
src/VulkanMemoryAllocator.hs view
@@ -2236,14 +2236,21 @@ -- 'destroyBuffer' or separately, using @vkDestroyBuffer()@ and -- 'freeMemory'. ----- If VMA_ALLOCATOR_CREATE_KHR_DEDICATED_ALLOCATION_BIT flag was used,+-- If 'ALLOCATOR_CREATE_KHR_DEDICATED_ALLOCATION_BIT' flag was used, -- VK_KHR_dedicated_allocation extension is used internally to query driver -- whether it requires or prefers the new buffer to have dedicated -- allocation. If yes, and if dedicated allocation is possible -- (/VmaAllocationCreateInfo::pool/ is null and--- VMA_ALLOCATION_CREATE_NEVER_ALLOCATE_BIT is not used), it creates+-- 'ALLOCATION_CREATE_NEVER_ALLOCATE_BIT' is not used), it creates -- dedicated allocation for this buffer, just like when using--- VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT.+-- 'ALLOCATION_CREATE_DEDICATED_MEMORY_BIT'.+--+-- Note+--+-- This function creates a new @VkBuffer@. Sub-allocation of parts of one+-- large buffer, although recommended as a good practice, is out of scope+-- of this library and could be implemented by the user as a higher-level+-- logic on top of VMA. createBuffer :: forall a io . (Extendss BufferCreateInfo a, PokeChain a, MonadIO io) => -- No documentation found for Nested "vmaCreateBuffer" "allocator"@@ -2291,9 +2298,8 @@ -- -- This is just a convenience function equivalent to: ----- @vkDestroyBuffer(device, buffer, allocationCallbacks);--- vmaFreeMemory(allocator, allocation);--- @+-- > vkDestroyBuffer(device, buffer, allocationCallbacks);+-- > vmaFreeMemory(allocator, allocation); -- -- It it safe to pass null as buffer and\/or allocation. destroyBuffer :: forall io@@ -2365,9 +2371,8 @@ -- -- This is just a convenience function equivalent to: ----- @vkDestroyImage(device, image, allocationCallbacks);--- vmaFreeMemory(allocator, allocation);--- @+-- > vkDestroyImage(device, image, allocationCallbacks);+-- > vmaFreeMemory(allocator, allocation); -- -- It it safe to pass null as image and\/or allocation. destroyImage :: forall io@@ -3036,9 +3041,9 @@ -- patch version number specified is ignored. Only the major and minor -- versions are considered. It must be less or equal (preferably equal) to -- value as passed to @vkCreateInstance@ as- -- @VkApplicationInfo::apiVersion@. Only versions 1.0 and 1.1 are supported- -- by the current implementation. Leaving it initialized to zero is- -- equivalent to @VK_API_VERSION_1_0@.+ -- @VkApplicationInfo::apiVersion@. Only versions 1.0, 1.1, 1.2 are+ -- supported by the current implementation. Leaving it initialized to zero+ -- is equivalent to @VK_API_VERSION_1_0@. vulkanApiVersion :: Word32 } deriving (Typeable)@@ -3732,26 +3737,39 @@ , -- | Intended usage of memory. -- -- You can leave 'MEMORY_USAGE_UNKNOWN' if you specify memory requirements- -- in other way. If @pool@ is not null, this member is ignored.+ -- in other way.+ --+ -- > + --+ -- If @pool@ is not null, this member is ignored. usage :: MemoryUsage , -- | Flags that must be set in a Memory Type chosen for an allocation. --- -- Leave 0 if you specify memory requirements in other way. If @pool@ is- -- not null, this member is ignored.+ -- Leave 0 if you specify memory requirements in other way.+ --+ -- > + --+ -- If @pool@ is not null, this member is ignored. requiredFlags :: MemoryPropertyFlags , -- | Flags that preferably should be set in a memory type chosen for an -- allocation. --- -- Set to 0 if no additional flags are preferred. If @pool@ is not null,- -- this member is ignored.+ -- Set to 0 if no additional flags are preferred.+ --+ -- > + --+ -- If @pool@ is not null, this member is ignored. preferredFlags :: MemoryPropertyFlags , -- | Bitmask containing one bit set for every memory type acceptable for this -- allocation. -- -- Value 0 is equivalent to @UINT32_MAX@ - it means any memory type is -- accepted if it meets other requirements specified by this structure,- -- with no further restrictions on memory type index. If @pool@ is not- -- null, this member is ignored.+ -- with no further restrictions on memory type index.+ --+ -- > + --+ -- If @pool@ is not null, this member is ignored. memoryTypeBits :: Word32 , -- | Pool that this allocation should be created in. --@@ -4130,8 +4148,15 @@ -- -- If the allocation is lost, it is equal to @VK_NULL_HANDLE@. deviceMemory :: DeviceMemory- , -- | Offset into deviceMemory object to the beginning of this allocation, in- -- bytes. (deviceMemory, offset) pair is unique to this allocation.+ , -- | Offset in @VkDeviceMemory@ object to the beginning of this allocation,+ -- in bytes. @(deviceMemory, offset)@ pair is unique to this allocation.+ --+ -- You usually don\'t need to use this offset. If you create a buffer or an+ -- image together with the allocation using e.g. function 'createBuffer',+ -- 'createImage', functions that operate on these resources refer to the+ -- beginning of the buffer or image, not entire device memory block.+ -- Functions like 'mapMemory', 'bindBufferMemory' also refer to the+ -- beginning of the allocation and apply this offset automatically. -- -- It can change after call to 'defragment' if this allocation is passed to -- the function, or if allocation is lost.