blob: 74af2c9ae861df6de33bb1ba4aa5022e287050e7 [file] [log] [blame]
 // // Copyright (c) 2017-2018 Advanced Micro Devices, Inc. All rights reserved. // // Permission is hereby granted, free of charge, to any person obtaining a copy // of this software and associated documentation files (the "Software"), to deal // in the Software without restriction, including without limitation the rights // to use, copy, modify, merge, publish, distribute, sublicense, and/or sell // copies of the Software, and to permit persons to whom the Software is // furnished to do so, subject to the following conditions: // // The above copyright notice and this permission notice shall be included in // all copies or substantial portions of the Software. // // THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR // IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, // FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE // AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER // LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, // OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN // THE SOFTWARE. // #ifndef AMD_VULKAN_MEMORY_ALLOCATOR_H #define AMD_VULKAN_MEMORY_ALLOCATOR_H #ifdef __cplusplus extern "C" { #endif /** \mainpage Vulkan Memory Allocator Version 2.0.0 (2018-03-19) Copyright (c) 2017-2018 Advanced Micro Devices, Inc. All rights reserved. \n License: MIT Documentation of all members: vk_mem_alloc.h \section main_table_of_contents Table of contents - User guide - \subpage quick_start - [Project setup](@ref quick_start_project_setup) - [Initialization](@ref quick_start_initialization) - [Resource allocation](@ref quick_start_resource_allocation) - \subpage choosing_memory_type - [Usage](@ref choosing_memory_type_usage) - [Required and preferred flags](@ref choosing_memory_type_required_preferred_flags) - [Explicit memory types](@ref choosing_memory_type_explicit_memory_types) - [Custom memory pools](@ref choosing_memory_type_custom_memory_pools) - \subpage memory_mapping - [Mapping functions](@ref memory_mapping_mapping_functions) - [Persistently mapped memory](@ref memory_mapping_persistently_mapped_memory) - [Cache control](@ref memory_mapping_cache_control) - [Finding out if memory is mappable](@ref memory_mapping_finding_if_memory_mappable) - \subpage custom_memory_pools - [Choosing memory type index](@ref custom_memory_pools_MemTypeIndex) - \subpage defragmentation - \subpage lost_allocations - \subpage statistics - [Numeric statistics](@ref statistics_numeric_statistics) - [JSON dump](@ref statistics_json_dump) - \subpage allocation_annotation - [Allocation user data](@ref allocation_user_data) - [Allocation names](@ref allocation_names) - \subpage usage_patterns - [Simple patterns](@ref usage_patterns_simple) - [Advanced patterns](@ref usage_patterns_advanced) - \subpage configuration - [Pointers to Vulkan functions](@ref config_Vulkan_functions) - [Custom host memory allocator](@ref custom_memory_allocator) - [Device memory allocation callbacks](@ref allocation_callbacks) - [Device heap memory limit](@ref heap_memory_limit) - \subpage vk_khr_dedicated_allocation - \subpage general_considerations - [Thread safety](@ref general_considerations_thread_safety) - [Allocation algorithm](@ref general_considerations_allocation_algorithm) - [Features not supported](@ref general_considerations_features_not_supported) \section main_see_also See also - [Product page on GPUOpen](https://gpuopen.com/gaming-product/vulkan-memory-allocator/) - [Source repository on GitHub](https://github.com/GPUOpen-LibrariesAndSDKs/VulkanMemoryAllocator) \page quick_start Quick start \section quick_start_project_setup Project setup Vulkan Memory Allocator comes in form of a single header file. You don't need to build it as a separate library project. You can add this file directly to your project and submit it to code repository next to your other source files. "Single header" doesn't mean that everything is contained in C/C++ declarations, like it tends to be in case of inline functions or C++ templates. It means that implementation is bundled with interface in a single file and needs to be extracted using preprocessor macro. If you don't do it properly, you will get linker errors. To do it properly: -# Include "vk_mem_alloc.h" file in each CPP file where you want to use the library. This includes declarations of all members of the library. -# In exacly one CPP file define following macro before this include. It enables also internal definitions. \code #define VMA_IMPLEMENTATION #include "vk_mem_alloc.h" \endcode It may be a good idea to create dedicated CPP file just for this purpose. \section quick_start_initialization Initialization At program startup: -# Initialize Vulkan to have VkPhysicalDevice and VkDevice object. -# Fill VmaAllocatorCreateInfo structure and create #VmaAllocator object by calling vmaCreateAllocator(). \code VmaAllocatorCreateInfo allocatorInfo = {}; allocatorInfo.physicalDevice = physicalDevice; allocatorInfo.device = device; VmaAllocator allocator; vmaCreateAllocator(&allocatorInfo, &allocator); \endcode \section quick_start_resource_allocation Resource allocation When you want to create a buffer or image: -# Fill VkBufferCreateInfo / VkImageCreateInfo structure. -# Fill VmaAllocationCreateInfo structure. -# Call vmaCreateBuffer() / vmaCreateImage() to get VkBuffer/VkImage with memory already allocated and bound to it. \code VkBufferCreateInfo bufferInfo = { VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO }; bufferInfo.size = 65536; bufferInfo.usage = VK_BUFFER_USAGE_VERTEX_BUFFER_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT; VmaAllocationCreateInfo allocInfo = {}; allocInfo.usage = VMA_MEMORY_USAGE_GPU_ONLY; VkBuffer buffer; VmaAllocation allocation; vmaCreateBuffer(allocator, &bufferInfo, &allocInfo, &buffer, &allocation, nullptr); \endcode Don't forget to destroy your objects when no longer needed: \code vmaDestroyBuffer(allocator, buffer, allocation); vmaDestroyAllocator(allocator); \endcode \page choosing_memory_type Choosing memory type Physical devices in Vulkan support various combinations of memory heaps and types. Help with choosing correct and optimal memory type for your specific resource is one of the key features of this library. You can use it by filling appropriate members of VmaAllocationCreateInfo structure, as described below. You can also combine multiple methods. -# If you just want to find memory type index that meets your requirements, you can use function vmaFindMemoryTypeIndex(). -# If you want to allocate a region of device memory without association with any specific image or buffer, you can use function vmaAllocateMemory(). Usage of this function is not recommended and usually not needed. -# If you already have a buffer or an image created, you want to allocate memory for it and then you will bind it yourself, you can use function vmaAllocateMemoryForBuffer(), vmaAllocateMemoryForImage(). For binding you should use functions: vmaBindBufferMemory(), vmaBindImageMemory(). -# If you want to create a buffer or an image, allocate memory for it and bind them together, all in one call, you can use function vmaCreateBuffer(), vmaCreateImage(). This is the recommended way to use this library. When using 3. or 4., the library internally queries Vulkan for memory types supported for that buffer or image (function vkGetBufferMemoryRequirements()) and uses only one of these types. If no memory type can be found that meets all the requirements, these functions return VK_ERROR_FEATURE_NOT_PRESENT. You can leave VmaAllocationCreateInfo structure completely filled with zeros. It means no requirements are specified for memory type. It is valid, although not very useful. \section choosing_memory_type_usage Usage The easiest way to specify memory requirements is to fill member VmaAllocationCreateInfo::usage using one of the values of enum #VmaMemoryUsage. It defines high level, common usage types. For more details, see description of this enum. For example, if you want to create a uniform buffer that will be filled using transfer only once or infrequently and used for rendering every frame, you can do it using following code: \code VkBufferCreateInfo bufferInfo = { VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO }; bufferInfo.size = 65536; bufferInfo.usage = VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT; VmaAllocationCreateInfo allocInfo = {}; allocInfo.usage = VMA_MEMORY_USAGE_GPU_ONLY; VkBuffer buffer; VmaAllocation allocation; vmaCreateBuffer(allocator, &bufferInfo, &allocInfo, &buffer, &allocation, nullptr); \endcode \section choosing_memory_type_required_preferred_flags Required and preferred flags You can specify more detailed requirements by filling members VmaAllocationCreateInfo::requiredFlags and VmaAllocationCreateInfo::preferredFlags with a combination of bits from enum VkMemoryPropertyFlags. For example, if you want to create a buffer that will be persistently mapped on host (so it must be HOST_VISIBLE) and preferably will also be HOST_COHERENT and HOST_CACHED, use following code: \code VmaAllocationCreateInfo allocInfo = {}; allocInfo.requiredFlags = VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT; allocInfo.preferredFlags = VK_MEMORY_PROPERTY_HOST_COHERENT_BIT | VK_MEMORY_PROPERTY_HOST_CACHED_BIT; allocInfo.flags = VMA_ALLOCATION_CREATE_MAPPED_BIT; VkBuffer buffer; VmaAllocation allocation; vmaCreateBuffer(allocator, &bufferInfo, &allocInfo, &buffer, &allocation, nullptr); \endcode A memory type is chosen that has all the required flags and as many preferred flags set as possible. If you use VmaAllocationCreateInfo::usage, it is just internally converted to a set of required and preferred flags. \section choosing_memory_type_explicit_memory_types Explicit memory types If you inspected memory types available on the physical device and you have a preference for memory types that you want to use, you can fill member VmaAllocationCreateInfo::memoryTypeBits. It is a bit mask, where each bit set means that a memory type with that index is allowed to be used for the allocation. Special value 0, just like UINT32_MAX, means there are no restrictions to memory type index. Please note that this member is NOT just a memory type index. Still you can use it to choose just one, specific memory type. For example, if you already determined that your buffer should be created in memory type 2, use following code: \code uint32_t memoryTypeIndex = 2; VmaAllocationCreateInfo allocInfo = {}; allocInfo.memoryTypeBits = 1u << memoryTypeIndex; VkBuffer buffer; VmaAllocation allocation; vmaCreateBuffer(allocator, &bufferInfo, &allocInfo, &buffer, &allocation, nullptr); \endcode \section choosing_memory_type_custom_memory_pools Custom memory pools If you allocate from custom memory pool, all the ways of specifying memory requirements described above are not applicable and the aforementioned members of VmaAllocationCreateInfo structure are ignored. Memory type is selected explicitly when creating the pool and then used to make all the allocations from that pool. For further details, see \ref custom_memory_pools. \page memory_mapping Memory mapping To "map memory" in Vulkan means to obtain a CPU pointer to VkDeviceMemory, to be able to read from it or write to it in CPU code. Mapping is possible only of memory allocated from a memory type that has VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT flag. Functions vkMapMemory(), vkUnmapMemory() are designed for this purpose. You can use them directly with memory allocated by this library, but it is not recommended because of following issue: Mapping the same VkDeviceMemory block multiple times is illegal - only one mapping at a time is allowed. This includes mapping disjoint regions. Mapping is not reference-counted internally by Vulkan. Because of this, Vulkan Memory Allocator provides following facilities: \section memory_mapping_mapping_functions Mapping functions The library provides following functions for mapping of a specific #VmaAllocation: vmaMapMemory(), vmaUnmapMemory(). They are safer and more convenient to use than standard Vulkan functions. You can map an allocation multiple times simultaneously - mapping is reference-counted internally. You can also map different allocations simultaneously regardless of whether they use the same VkDeviceMemory block. They way it's implemented is that the library always maps entire memory block, not just region of the allocation. For further details, see description of vmaMapMemory() function. Example: \code // Having these objects initialized: struct ConstantBuffer { ... }; ConstantBuffer constantBufferData; VmaAllocator allocator; VmaBuffer constantBuffer; VmaAllocation constantBufferAllocation; // You can map and fill your buffer using following code: void* mappedData; vmaMapMemory(allocator, constantBufferAllocation, &mappedData); memcpy(mappedData, &constantBufferData, sizeof(constantBufferData)); vmaUnmapMemory(allocator, constantBufferAllocation); \endcode \section memory_mapping_persistently_mapped_memory Persistently mapped memory Kepping your memory persistently mapped is generally OK in Vulkan. You don't need to unmap it before using its data on the GPU. The library provides a special feature designed for that: Allocations made with #VMA_ALLOCATION_CREATE_MAPPED_BIT flag set in VmaAllocationCreateInfo::flags stay mapped all the time, so you can just access CPU pointer to it any time without a need to call any "map" or "unmap" function. Example: \code VkBufferCreateInfo bufCreateInfo = { VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO }; bufCreateInfo.size = sizeof(ConstantBuffer); bufCreateInfo.usage = VK_BUFFER_USAGE_TRANSFER_SRC_BIT; VmaAllocationCreateInfo allocCreateInfo = {}; allocCreateInfo.usage = VMA_MEMORY_USAGE_CPU_ONLY; allocCreateInfo.flags = VMA_ALLOCATION_CREATE_MAPPED_BIT; VkBuffer buf; VmaAllocation alloc; VmaAllocationInfo allocInfo; vmaCreateBuffer(allocator, &bufCreateInfo, &allocCreateInfo, &buf, &alloc, &allocInfo); // Buffer is already mapped. You can access its memory. memcpy(allocInfo.pMappedData, &constantBufferData, sizeof(constantBufferData)); \endcode There are some exceptions though, when you should consider mapping memory only for a short period of time: - When operating system is Windows 7 or 8.x (Windows 10 is not affected because it uses WDDM2), device is discrete AMD GPU, and memory type is the special 256 MiB pool of DEVICE_LOCAL + HOST_VISIBLE memory (selected when you use #VMA_MEMORY_USAGE_CPU_TO_GPU), then whenever a memory block allocated from this memory type stays mapped for the time of any call to vkQueueSubmit() or vkQueuePresentKHR(), this block is migrated by WDDM to system RAM, which degrades performance. It doesn't matter if that particular memory block is actually used by the command buffer being submitted. - Keeping many large memory blocks mapped may impact performance or stability of some debugging tools. \section memory_mapping_cache_control Cache control 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 using function vkvkInvalidateMappedMemoryRanges() and flush cache after writing to mapped pointer using function vkFlushMappedMemoryRanges(). Example: \code memcpy(allocInfo.pMappedData, &constantBufferData, sizeof(constantBufferData)); VkMemoryPropertyFlags memFlags; vmaGetMemoryTypeProperties(allocator, allocInfo.memoryType, &memFlags); if((memFlags & VK_MEMORY_PROPERTY_HOST_COHERENT_BIT) == 0) { VkMappedMemoryRange memRange = { VK_STRUCTURE_TYPE_MAPPED_MEMORY_RANGE }; memRange.memory = allocInfo.deviceMemory; memRange.offset = allocInfo.offset; memRange.size = allocInfo.size; vkFlushMappedMemoryRanges(device, 1, &memRange); } \endcode Please note that memory allocated with #VMA_MEMORY_USAGE_CPU_ONLY is guaranteed to be host coherent. Also, Windows drivers from all 3 PC GPU vendors (AMD, Intel, NVIDIA) currently provide VK_MEMORY_PROPERTY_HOST_COHERENT_BIT flag on all memory types that are VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT, so on this platform you may not need to bother. \section memory_mapping_finding_if_memory_mappable Finding out if memory is mappable It may happen that your allocation ends up in memory that is HOST_VISIBLE (available for mapping) despite it wasn't explicitly requested. For example, application may work on integrated graphics with unified memory (like Intel) or allocation from video memory might have failed, so the library chose system memory as fallback. You can detect this case and map such allocation to access its memory on CPU directly, instead of launching a transfer operation. In order to do that: inspect allocInfo.memoryType, call vmaGetMemoryTypeProperties(), and look for VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT flag in properties of that memory type. \code VkBufferCreateInfo bufCreateInfo = { VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO }; bufCreateInfo.size = sizeof(ConstantBuffer); bufCreateInfo.usage = VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT; VmaAllocationCreateInfo allocCreateInfo = {}; allocCreateInfo.usage = VMA_MEMORY_USAGE_GPU_ONLY; VkBuffer buf; VmaAllocation alloc; VmaAllocationInfo allocInfo; vmaCreateBuffer(allocator, &bufCreateInfo, &allocCreateInfo, &buf, &alloc, &allocInfo); VkMemoryPropertyFlags memFlags; vmaGetMemoryTypeProperties(allocator, allocInfo.memoryType, &memFlags); if((memFlags & VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT) == 0) { // Allocation ended up in mappable memory. You can map it and access it directly. void* mappedData; vmaMapMemory(allocator, alloc, &mappedData); memcpy(mappedData, &constantBufferData, sizeof(constantBufferData)); vmaUnmapMemory(allocator, alloc); } else { // Allocation ended up in non-mappable memory. // You need to create CPU-side buffer in VMA_MEMORY_USAGE_CPU_ONLY and make a transfer. } \endcode You can even use #VMA_ALLOCATION_CREATE_MAPPED_BIT flag while creating allocations that are not necessarily HOST_VISIBLE (e.g. using #VMA_MEMORY_USAGE_GPU_ONLY). If the allocation ends up in memory type that is HOST_VISIBLE, it will be persistently mapped and you can use it directly. If not, the flag is just ignored. Example: \code VkBufferCreateInfo bufCreateInfo = { VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO }; bufCreateInfo.size = sizeof(ConstantBuffer); bufCreateInfo.usage = VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT; VmaAllocationCreateInfo allocCreateInfo = {}; allocCreateInfo.usage = VMA_MEMORY_USAGE_GPU_ONLY; allocCreateInfo.flags = VMA_ALLOCATION_CREATE_MAPPED_BIT; VkBuffer buf; VmaAllocation alloc; VmaAllocationInfo allocInfo; vmaCreateBuffer(allocator, &bufCreateInfo, &allocCreateInfo, &buf, &alloc, &allocInfo); if(allocInfo.pUserData != nullptr) { // Allocation ended up in mappable memory. // It's persistently mapped. You can access it directly. memcpy(allocInfo.pMappedData, &constantBufferData, sizeof(constantBufferData)); } else { // Allocation ended up in non-mappable memory. // You need to create CPU-side buffer in VMA_MEMORY_USAGE_CPU_ONLY and make a transfer. } \endcode \page custom_memory_pools Custom memory pools A memory pool contains a number of VkDeviceMemory blocks. The library automatically creates and manages default pool for each memory type available on the device. Default memory pool automatically grows in size. Size of allocated blocks is also variable and managed automatically. You can create custom pool and allocate memory out of it. It can be useful if you want to: - Keep certain kind of allocations separate from others. - Enforce particular, fixed size of Vulkan memory blocks. - Limit maximum amount of Vulkan memory allocated for that pool. - Reserve minimum or fixed amount of Vulkan memory always preallocated for that pool. To use custom memory pools: -# Fill VmaPoolCreateInfo structure. -# Call vmaCreatePool() to obtain #VmaPool handle. -# When making an allocation, set VmaAllocationCreateInfo::pool to this handle. You don't need to specify any other parameters of this structure, like usage. Example: \code // Create a pool that can have at most 2 blocks, 128 MiB each. VmaPoolCreateInfo poolCreateInfo = {}; poolCreateInfo.memoryTypeIndex = ... poolCreateInfo.blockSize = 128ull * 1024 * 1024; poolCreateInfo.maxBlockCount = 2; VmaPool pool; vmaCreatePool(allocator, &poolCreateInfo, &pool); // Allocate a buffer out of it. VkBufferCreateInfo bufCreateInfo = { VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO }; bufCreateInfo.size = 1024; bufCreateInfo.usage = VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT; VmaAllocationCreateInfo allocCreateInfo = {}; allocCreateInfo.pool = pool; VkBuffer buf; VmaAllocation alloc; VmaAllocationInfo allocInfo; vmaCreateBuffer(allocator, &bufCreateInfo, &allocCreateInfo, &buf, &alloc, &allocInfo); \endcode You have to free all allocations made from this pool before destroying it. \code vmaDestroyBuffer(allocator, buf, alloc); vmaDestroyPool(allocator, pool); \endcode \section custom_memory_pools_MemTypeIndex Choosing memory type index When creating a pool, you must explicitly specify memory type index. To find the one suitable for your buffers or images, you can use helper functions vmaFindMemoryTypeIndexForBufferInfo(), vmaFindMemoryTypeIndexForImageInfo(). You need to provide structures with example parameters of buffers or images that you are going to create in that pool. \code VkBufferCreateInfo exampleBufCreateInfo = { VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO }; exampleBufCreateInfo.size = 1024; // Whatever. exampleBufCreateInfo.usage = VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT; // Change if needed. VmaAllocationCreateInfo allocCreateInfo = {}; allocCreateInfo.usage = VMA_MEMORY_USAGE_GPU_ONLY; // Change if needed. uint32_t memTypeIndex; vmaFindMemoryTypeIndexForBufferInfo(allocator, &exampleBufCreateInfo, &allocCreateInfo, &memTypeIndex); VmaPoolCreateInfo poolCreateInfo = {}; poolCreateInfo.memoryTypeIndex = memTypeIndex; // ... \endcode When creating buffers/images allocated in that pool, provide following parameters: - VkBufferCreateInfo: Prefer to pass same parameters as above. Otherwise you risk creating resources in a memory type that is not suitable for them, which may result in undefined behavior. Using different VK_BUFFER_USAGE_ flags may work, but you shouldn't create images in a pool intended for buffers or the other way around. - VmaAllocationCreateInfo: You don't need to pass same parameters. Fill only pool member. Other members are ignored anyway. \page defragmentation Defragmentation Interleaved allocations and deallocations of many objects of varying size can cause fragmentation, which can lead to a situation where the library is unable to find a continuous range of free memory for a new allocation despite there is enough free space, just scattered across many small free ranges between existing allocations. To mitigate this problem, you can use vmaDefragment(). Given set of allocations, this function can move them to compact used memory, ensure more continuous free space and possibly also free some VkDeviceMemory. It can work only on allocations made from memory type that is HOST_VISIBLE. Allocations are modified to point to the new VkDeviceMemory and offset. Data in this memory is also memmove-ed to the new place. However, if you have images or buffers bound to these allocations (and you certainly do), you need to destroy, recreate, and bind them to the new place in memory. For further details and example code, see documentation of function vmaDefragment(). \page lost_allocations Lost allocations If your game oversubscribes video memory, if may work OK in previous-generation graphics APIs (DirectX 9, 10, 11, OpenGL) because resources are automatically paged to system RAM. In Vulkan you can't do it because when you run out of memory, an allocation just fails. If you have more data (e.g. textures) that can fit into VRAM and you don't need it all at once, you may want to upload them to GPU on demand and "push out" ones that are not used for a long time to make room for the new ones, effectively using VRAM (or a cartain memory pool) as a form of cache. Vulkan Memory Allocator can help you with that by supporting a concept of "lost allocations". To create an allocation that can become lost, include #VMA_ALLOCATION_CREATE_CAN_BECOME_LOST_BIT flag in VmaAllocationCreateInfo::flags. Before using a buffer or image bound to such allocation in every new frame, you need to query it if it's not lost. To check it, call vmaTouchAllocation(). If the allocation is lost, you should not use it or buffer/image bound to it. You mustn't forget to destroy this allocation and this buffer/image. vmaGetAllocationInfo() can also be used for checking status of the allocation. Allocation is lost when returned VmaAllocationInfo::deviceMemory == VK_NULL_HANDLE. To create an allocation that can make some other allocations lost to make room for it, use #VMA_ALLOCATION_CREATE_CAN_MAKE_OTHER_LOST_BIT flag. You will usually use both flags #VMA_ALLOCATION_CREATE_CAN_MAKE_OTHER_LOST_BIT and #VMA_ALLOCATION_CREATE_CAN_BECOME_LOST_BIT at the same time. Warning! Current implementation uses quite naive, brute force algorithm, which can make allocation calls that use #VMA_ALLOCATION_CREATE_CAN_MAKE_OTHER_LOST_BIT flag quite slow. A new, more optimal algorithm and data structure to speed this up is planned for the future. Q: When interleaving creation of new allocations with usage of existing ones, how do you make sure that an allocation won't become lost while it's used in the current frame? It is ensured because vmaTouchAllocation() / vmaGetAllocationInfo() not only returns allocation status/parameters and checks whether it's not lost, but when it's not, it also atomically marks it as used in the current frame, which makes it impossible to become lost in that frame. It uses lockless algorithm, so it works fast and doesn't involve locking any internal mutex. Q: What if my allocation may still be in use by the GPU when it's rendering a previous frame while I already submit new frame on the CPU? You can make sure that allocations "touched" by vmaTouchAllocation() / vmaGetAllocationInfo() will not become lost for a number of additional frames back from the current one by specifying this number as VmaAllocatorCreateInfo::frameInUseCount (for default memory pool) and VmaPoolCreateInfo::frameInUseCount (for custom pool). Q: How do you inform the library when new frame starts? You need to call function vmaSetCurrentFrameIndex(). Example code: \code struct MyBuffer { VkBuffer m_Buf = nullptr; VmaAllocation m_Alloc = nullptr; // Called when the buffer is really needed in the current frame. void EnsureBuffer(); }; void MyBuffer::EnsureBuffer() { // Buffer has been created. if(m_Buf != VK_NULL_HANDLE) { // Check if its allocation is not lost + mark it as used in current frame. if(vmaTouchAllocation(allocator, m_Alloc)) { // It's all OK - safe to use m_Buf. return; } } // Buffer not yet exists or lost - destroy and recreate it. vmaDestroyBuffer(allocator, m_Buf, m_Alloc); VkBufferCreateInfo bufCreateInfo = { VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO }; bufCreateInfo.size = 1024; bufCreateInfo.usage = VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT; VmaAllocationCreateInfo allocCreateInfo = {}; allocCreateInfo.usage = VMA_MEMORY_USAGE_GPU_ONLY; allocCreateInfo.flags = VMA_ALLOCATION_CREATE_CAN_BECOME_LOST_BIT | VMA_ALLOCATION_CREATE_CAN_MAKE_OTHER_LOST_BIT; vmaCreateBuffer(allocator, &bufCreateInfo, &allocCreateInfo, &m_Buf, &m_Alloc, nullptr); } \endcode When using lost allocations, you may see some Vulkan validation layer warnings about overlapping regions of memory bound to different kinds of buffers and images. This is still valid as long as you implement proper handling of lost allocations (like in the example above) and don't use them. You can create an allocation that is already in lost state from the beginning using function vmaCreateLostAllocation(). It may be useful if you need a "dummy" allocation that is not null. You can call function vmaMakePoolAllocationsLost() to set all eligible allocations in a specified custom pool to lost state. Allocations that have been "touched" in current frame or VmaPoolCreateInfo::frameInUseCount frames back cannot become lost. \page statistics Statistics This library contains functions that return information about its internal state, especially the amount of memory allocated from Vulkan. Please keep in mind that these functions need to traverse all internal data structures to gather these information, so they may be quite time-consuming. Don't call them too often. \section statistics_numeric_statistics Numeric statistics You can query for overall statistics of the allocator using function vmaCalculateStats(). Information are returned using structure #VmaStats. It contains #VmaStatInfo - number of allocated blocks, number of allocations (occupied ranges in these blocks), number of unused (free) ranges in these blocks, number of bytes used and unused (but still allocated from Vulkan) and other information. They are summed across memory heaps, memory types and total for whole allocator. You can query for statistics of a custom pool using function vmaGetPoolStats(). Information are returned using structure #VmaPoolStats. You can query for information about specific allocation using function vmaGetAllocationInfo(). It fill structure #VmaAllocationInfo. \section statistics_json_dump JSON dump You can dump internal state of the allocator to a string in JSON format using function vmaBuildStatsString(). The result is guaranteed to be correct JSON. It uses ANSI encoding. Any strings provided by user (see [Allocation names](@ref allocation_names)) are copied as-is and properly escaped for JSON, so if they use UTF-8, ISO-8859-2 or any other encoding, this JSON string can be treated as using this encoding. It must be freed using function vmaFreeStatsString(). The format of this JSON string is not part of official documentation of the library, but it will not change in backward-incompatible way without increasing library major version number and appropriate mention in changelog. The JSON string contains all the data that can be obtained using vmaCalculateStats(). It can also contain detailed map of allocated memory blocks and their regions - free and occupied by allocations. This allows e.g. to visualize the memory or assess fragmentation. \page allocation_annotation Allocation names and user data \section allocation_user_data Allocation user data You can annotate allocations with your own information, e.g. for debugging purposes. To do that, fill VmaAllocationCreateInfo::pUserData field when creating an allocation. It's an opaque void* pointer. You can use it e.g. as a pointer, some handle, index, key, ordinal number or any other value that would associate the allocation with your custom metadata. \code VkBufferCreateInfo bufferInfo = { VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO }; // Fill bufferInfo... MyBufferMetadata* pMetadata = CreateBufferMetadata(); VmaAllocationCreateInfo allocCreateInfo = {}; allocCreateInfo.usage = VMA_MEMORY_USAGE_GPU_ONLY; allocCreateInfo.pUserData = pMetadata; VkBuffer buffer; VmaAllocation allocation; vmaCreateBuffer(allocator, &bufferInfo, &allocCreateInfo, &buffer, &allocation, nullptr); \endcode The pointer may be later retrieved as VmaAllocationInfo::pUserData: \code VmaAllocationInfo allocInfo; vmaGetAllocationInfo(allocator, allocation, &allocInfo); MyBufferMetadata* pMetadata = (MyBufferMetadata*)allocInfo.pUserData; \endcode It can also be changed using function vmaSetAllocationUserData(). Values of (non-zero) allocations' pUserData are printed in JSON report created by vmaBuildStatsString(), in hexadecimal form. \section allocation_names Allocation names There is alternative mode available where pUserData pointer is used to point to a null-terminated string, giving a name to the allocation. To use this mode, set #VMA_ALLOCATION_CREATE_USER_DATA_COPY_STRING_BIT flag in VmaAllocationCreateInfo::flags. Then pUserData passed as VmaAllocationCreateInfo::pUserData or argument to vmaSetAllocationUserData() must be either null or pointer to a null-terminated string. The library creates internal copy of the string, so the pointer you pass doesn't need to be valid for whole lifetime of the allocation. You can free it after the call. \code VkImageCreateInfo imageInfo = { VK_STRUCTURE_TYPE_IMAGE_CREATE_INFO }; // Fill imageInfo... std::string imageName = "Texture: "; imageName += fileName; VmaAllocationCreateInfo allocCreateInfo = {}; allocCreateInfo.usage = VMA_MEMORY_USAGE_GPU_ONLY; allocCreateInfo.flags = VMA_ALLOCATION_CREATE_USER_DATA_COPY_STRING_BIT; allocCreateInfo.pUserData = imageName.c_str(); VkImage image; VmaAllocation allocation; vmaCreateImage(allocator, &imageInfo, &allocCreateInfo, &image, &allocation, nullptr); \endcode The value of pUserData pointer of the allocation will be different than the one you passed when setting allocation's name - pointing to a buffer managed internally that holds copy of the string. \code VmaAllocationInfo allocInfo; vmaGetAllocationInfo(allocator, allocation, &allocInfo); const char* imageName = (const char*)allocInfo.pUserData; printf("Image name: %s\n", imageName); \endcode That string is also printed in JSON report created by vmaBuildStatsString(). \page usage_patterns Recommended usage patterns \section usage_patterns_simple Simple patterns \subsection usage_patterns_simple_render_targets Render targets When: Any resources that you frequently write and read on GPU, e.g. images used as color attachments (aka "render targets"), depth-stencil attachments, images/buffers used as storage image/buffer (aka "Unordered Access View (UAV)"). What to do: Create them in video memory that is fastest to access from GPU using #VMA_MEMORY_USAGE_GPU_ONLY. Consider using [VK_KHR_dedicated_allocation](@ref vk_khr_dedicated_allocation) extension and/or manually creating them as dedicated allocations using #VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT, especially if they are large or if you plan to destroy and recreate them e.g. when display resolution changes. Prefer to create such resources first and all other GPU resources (like textures and vertex buffers) later. \subsection usage_patterns_simple_immutable_resources Immutable resources When: Any resources that you fill on CPU only once (aka "immutable") or infrequently and then read frequently on GPU, e.g. textures, vertex and index buffers, constant buffers that don't change often. What to do: Create them in video memory that is fastest to access from GPU using #VMA_MEMORY_USAGE_GPU_ONLY. To initialize content of such resource, create a CPU-side (aka "staging") copy of it in system memory - #VMA_MEMORY_USAGE_CPU_ONLY, map it, fill it, and submit a transfer from it to the GPU resource. You can keep the staging copy if you need it for another upload transfer in the future. If you don't, you can destroy it or reuse this buffer for uploading different resource after the transfer finishes. Prefer to create just buffers in system memory rather than images, even for uploading textures. Use vkCmdCopyBufferToImage(). Dont use images with VK_IMAGE_TILING_LINEAR. \subsection usage_patterns_dynamic_resources Dynamic resources When: Any resources that change frequently (aka "dynamic"), e.g. every frame or every draw call, written on CPU, read on GPU. What to do: Create them using #VMA_MEMORY_USAGE_CPU_TO_GPU. You can map it and write to it directly on CPU, as well as read from it on GPU. This is a more complex situation. Different solutions are possible, and the best one depends on specific GPU type, but you can use this simple approach for the start. Prefer to write to such resource sequentially (e.g. using memcpy). Don't perform random access or any reads from it, as it may be very slow. \subsection usage_patterns_readback Readback When: Resources that contain data written by GPU that you want to read back on CPU, e.g. results of some computations. What to do: Create them using #VMA_MEMORY_USAGE_GPU_TO_CPU. You can write to them directly on GPU, as well as map and read them on CPU. \section usage_patterns_advanced Advanced patterns \subsection usage_patterns_integrated_graphics Detecting integrated graphics You can support integrated graphics (like Intel HD Graphics, AMD APU) better by detecting it in Vulkan. To do it, call vkGetPhysicalDeviceProperties(), inspect VkPhysicalDeviceProperties::deviceType and look for VK_PHYSICAL_DEVICE_TYPE_INTEGRATED_GPU. When you find it, you can assume that memory is unified and all memory types are equally fast to access from GPU, regardless of VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT. You can then sum up sizes of all available memory heaps and treat them as useful for your GPU resources, instead of only DEVICE_LOCAL ones. You can also prefer to create your resources in memory types that are HOST_VISIBLE to map them directly instead of submitting explicit transfer (see below). \subsection usage_patterns_direct_vs_transfer Direct access versus transfer For resources that you frequently write on CPU and read on GPU, many solutions are possible: -# Create one copy in video memory using #VMA_MEMORY_USAGE_GPU_ONLY, second copy in system memory using #VMA_MEMORY_USAGE_CPU_ONLY and submit explicit tranfer each time. -# Create just single copy using #VMA_MEMORY_USAGE_CPU_TO_GPU, map it and fill it on CPU, read it directly on GPU. -# Create just single copy using #VMA_MEMORY_USAGE_CPU_ONLY, map it and fill it on CPU, read it directly on GPU. Which solution is the most efficient depends on your resource and especially on the GPU. It is best to measure it and then make the decision. Some general recommendations: - On integrated graphics use (2) or (3) to avoid unnecesary time and memory overhead related to using a second copy. - For small resources (e.g. constant buffers) use (2). Discrete AMD cards have special 256 MiB pool of video memory that is directly mappable. Even if the resource ends up in system memory, its data may be cached on GPU after first fetch over PCIe bus. - For larger resources (e.g. textures), decide between (1) and (2). You may want to differentiate NVIDIA and AMD, e.g. by looking for memory type that is both DEVICE_LOCAL and HOST_VISIBLE. When you find it, use (2), otherwise use (1). Similarly, for resources that you frequently write on GPU and read on CPU, multiple solutions are possible: -# Create one copy in video memory using #VMA_MEMORY_USAGE_GPU_ONLY, second copy in system memory using #VMA_MEMORY_USAGE_GPU_TO_CPU and submit explicit tranfer each time. -# Create just single copy using #VMA_MEMORY_USAGE_GPU_TO_CPU, write to it directly on GPU, map it and read it on CPU. You should take some measurements to decide which option is faster in case of your specific resource. If you don't want to specialize your code for specific types of GPUs, yon can still make an simple optimization for cases when your resource ends up in mappable memory to use it directly in this case instead of creating CPU-side staging copy. For details see [Finding out if memory is mappable](@ref memory_mapping_finding_if_memory_mappable). \page configuration Configuration Please check "CONFIGURATION SECTION" in the code to find macros that you can define before each include of this file or change directly in this file to provide your own implementation of basic facilities like assert, min() and max() functions, mutex, atomic etc. The library uses its own implementation of containers by default, but you can switch to using STL containers instead. \section config_Vulkan_functions Pointers to Vulkan functions The library uses Vulkan functions straight from the vulkan.h header by default. If you want to provide your own pointers to these functions, e.g. fetched using vkGetInstanceProcAddr() and vkGetDeviceProcAddr(): -# Define VMA_STATIC_VULKAN_FUNCTIONS 0. -# Provide valid pointers through VmaAllocatorCreateInfo::pVulkanFunctions. \section custom_memory_allocator Custom host memory allocator If you use custom allocator for CPU memory rather than default operator new and delete from C++, you can make this library using your allocator as well by filling optional member VmaAllocatorCreateInfo::pAllocationCallbacks. These functions will be passed to Vulkan, as well as used by the library itself to make any CPU-side allocations. \section allocation_callbacks Device memory allocation callbacks The library makes calls to vkAllocateMemory() and vkFreeMemory() internally. You can setup callbacks to be informed about these calls, e.g. for the purpose of gathering some statistics. To do it, fill optional member VmaAllocatorCreateInfo::pDeviceMemoryCallbacks. \section heap_memory_limit Device heap memory limit If you want to test how your program behaves with limited amount of Vulkan device memory available without switching your graphics card to one that really has smaller VRAM, you can use a feature of this library intended for this purpose. To do it, fill optional member VmaAllocatorCreateInfo::pHeapSizeLimit. \page vk_khr_dedicated_allocation VK_KHR_dedicated_allocation VK_KHR_dedicated_allocation is a Vulkan extension which can be used to improve performance on some GPUs. It augments Vulkan API with possibility to query driver whether it prefers particular buffer or image to have its own, dedicated allocation (separate VkDeviceMemory block) for better efficiency - to be able to do some internal optimizations. The extension is supported by this library. It will be used automatically when enabled. To enable it: 1 . When creating Vulkan device, check if following 2 device extensions are supported (call vkEnumerateDeviceExtensionProperties()). If yes, enable them (fill VkDeviceCreateInfo::ppEnabledExtensionNames). - VK_KHR_get_memory_requirements2 - VK_KHR_dedicated_allocation If you enabled these extensions: 2 . Use #VMA_ALLOCATOR_CREATE_KHR_DEDICATED_ALLOCATION_BIT flag when creating your #VmaAllocatorto inform the library that you enabled required extensions and you want the library to use them. \code allocatorInfo.flags |= VMA_ALLOCATOR_CREATE_KHR_DEDICATED_ALLOCATION_BIT; vmaCreateAllocator(&allocatorInfo, &allocator); \endcode That's all. The extension will be automatically used whenever you create a buffer using vmaCreateBuffer() or image using vmaCreateImage(). When using the extension together with Vulkan Validation Layer, you will receive warnings like this: vkBindBufferMemory(): Binding memory to buffer 0x33 but vkGetBufferMemoryRequirements() has not been called on that buffer. It is OK, you should just ignore it. It happens because you use function vkGetBufferMemoryRequirements2KHR() instead of standard vkGetBufferMemoryRequirements(), while the validation layer seems to be unaware of it. To learn more about this extension, see: - [VK_KHR_dedicated_allocation in Vulkan specification](https://www.khronos.org/registry/vulkan/specs/1.0-extensions/html/vkspec.html#VK_KHR_dedicated_allocation) - [VK_KHR_dedicated_allocation unofficial manual](http://asawicki.info/articles/VK_KHR_dedicated_allocation.php5) \page general_considerations General considerations \section general_considerations_thread_safety Thread safety - The library has no global state, so separate #VmaAllocator objects can be used independently. There should be no need to create multiple such objects though - one per VkDevice is enough. - By default, all calls to functions that take #VmaAllocator as first parameter are safe to call from multiple threads simultaneously because they are synchronized internally when needed. - When the allocator is created with #VMA_ALLOCATOR_CREATE_EXTERNALLY_SYNCHRONIZED_BIT flag, calls to functions that take such #VmaAllocator object must be synchronized externally. - Access to a #VmaAllocation object must be externally synchronized. For example, you must not call vmaGetAllocationInfo() and vmaMapMemory() from different threads at the same time if you pass the same #VmaAllocation object to these functions. \section general_considerations_allocation_algorithm Allocation algorithm The library uses following algorithm for allocation, in order: -# Try to find free range of memory in existing blocks. -# If failed, try to create a new block of VkDeviceMemory, with preferred block size. -# If failed, try to create such block with size/2, size/4, size/8. -# If failed and #VMA_ALLOCATION_CREATE_CAN_MAKE_OTHER_LOST_BIT flag was specified, try to find space in existing blocks, possilby making some other allocations lost. -# If failed, try to allocate separate VkDeviceMemory for this allocation, just like when you use #VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT. -# If failed, choose other memory type that meets the requirements specified in VmaAllocationCreateInfo and go to point 1. -# If failed, return VK_ERROR_OUT_OF_DEVICE_MEMORY. \section general_considerations_features_not_supported Features not supported Features deliberately excluded from the scope of this library: - Data transfer - issuing commands that transfer data between buffers or images, any usage of VkCommandList or VkCommandQueue and related synchronization is responsibility of the user. - Support for any programming languages other than C/C++. Bindings to other languages are welcomed as external projects. */ // For skia we don't include vulkan.h here. Before including this header we always include // GrVkDefines which has a user defined header. Additionally we don't require vulkan/vulkan.h to be // on the include path. // #include /** \struct VmaAllocator \brief Represents main object of this library initialized. Fill structure VmaAllocatorCreateInfo and call function vmaCreateAllocator() to create it. Call function vmaDestroyAllocator() to destroy it. It is recommended to create just one object of this type per VkDevice object, right after Vulkan is initialized and keep it alive until before Vulkan device is destroyed. */ VK_DEFINE_HANDLE(VmaAllocator) /// Callback function called after successful vkAllocateMemory. typedef void (VKAPI_PTR *PFN_vmaAllocateDeviceMemoryFunction)( VmaAllocator allocator, uint32_t memoryType, VkDeviceMemory memory, VkDeviceSize size); /// Callback function called before vkFreeMemory. typedef void (VKAPI_PTR *PFN_vmaFreeDeviceMemoryFunction)( VmaAllocator allocator, uint32_t memoryType, VkDeviceMemory memory, VkDeviceSize size); /** \brief Set of callbacks that the library will call for vkAllocateMemory and vkFreeMemory. Provided for informative purpose, e.g. to gather statistics about number of allocations or total amount of memory allocated in Vulkan. Used in VmaAllocatorCreateInfo::pDeviceMemoryCallbacks. */ typedef struct VmaDeviceMemoryCallbacks { /// Optional, can be null. PFN_vmaAllocateDeviceMemoryFunction pfnAllocate; /// Optional, can be null. PFN_vmaFreeDeviceMemoryFunction pfnFree; } VmaDeviceMemoryCallbacks; /// Flags for created #VmaAllocator. typedef enum VmaAllocatorCreateFlagBits { /** \brief Allocator and all objects created from it will not be synchronized internally, so you must guarantee they are used from only one thread at a time or synchronized externally by you. Using this flag may increase performance because internal mutexes are not used. */ VMA_ALLOCATOR_CREATE_EXTERNALLY_SYNCHRONIZED_BIT = 0x00000001, /** \brief Enables usage of VK_KHR_dedicated_allocation extension. Using this extenion will automatically allocate dedicated blocks of memory for some buffers and images instead of suballocating place for them out of bigger memory blocks (as if you explicitly used #VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT flag) when it is recommended by the driver. It may improve performance on some GPUs. You may set this flag only if you found out that following device extensions are supported, you enabled them while creating Vulkan device passed as VmaAllocatorCreateInfo::device, and you want them to be used internally by this library: - VK_KHR_get_memory_requirements2 - VK_KHR_dedicated_allocation When this flag is set, you can experience following warnings reported by Vulkan validation layer. You can ignore them. > vkBindBufferMemory(): Binding memory to buffer 0x2d but vkGetBufferMemoryRequirements() has not been called on that buffer. */ VMA_ALLOCATOR_CREATE_KHR_DEDICATED_ALLOCATION_BIT = 0x00000002, VMA_ALLOCATOR_CREATE_FLAG_BITS_MAX_ENUM = 0x7FFFFFFF } VmaAllocatorCreateFlagBits; typedef VkFlags VmaAllocatorCreateFlags; /** \brief Pointers to some Vulkan functions - a subset used by the library. Used in VmaAllocatorCreateInfo::pVulkanFunctions. */ typedef struct VmaVulkanFunctions { PFN_vkGetPhysicalDeviceProperties vkGetPhysicalDeviceProperties; PFN_vkGetPhysicalDeviceMemoryProperties vkGetPhysicalDeviceMemoryProperties; PFN_vkAllocateMemory vkAllocateMemory; PFN_vkFreeMemory vkFreeMemory; PFN_vkMapMemory vkMapMemory; PFN_vkUnmapMemory vkUnmapMemory; PFN_vkBindBufferMemory vkBindBufferMemory; PFN_vkBindImageMemory vkBindImageMemory; PFN_vkGetBufferMemoryRequirements vkGetBufferMemoryRequirements; PFN_vkGetImageMemoryRequirements vkGetImageMemoryRequirements; PFN_vkCreateBuffer vkCreateBuffer; PFN_vkDestroyBuffer vkDestroyBuffer; PFN_vkCreateImage vkCreateImage; PFN_vkDestroyImage vkDestroyImage; PFN_vkGetBufferMemoryRequirements2KHR vkGetBufferMemoryRequirements2KHR; PFN_vkGetImageMemoryRequirements2KHR vkGetImageMemoryRequirements2KHR; } VmaVulkanFunctions; /// Description of a Allocator to be created. typedef struct VmaAllocatorCreateInfo { /// Flags for created allocator. Use #VmaAllocatorCreateFlagBits enum. VmaAllocatorCreateFlags flags; /// Vulkan physical device. /** It must be valid throughout whole lifetime of created allocator. */ VkPhysicalDevice physicalDevice; /// Vulkan device. /** It must be valid throughout whole lifetime of created allocator. */ VkDevice device; /// Preferred size of a single VkDeviceMemory block to be allocated from large heaps > 1 GiB. Optional. /** Set to 0 to use default, which is currently 256 MiB. */ VkDeviceSize preferredLargeHeapBlockSize; /// Custom CPU memory allocation callbacks. Optional. /** Optional, can be null. When specified, will also be used for all CPU-side memory allocations. */ const VkAllocationCallbacks* pAllocationCallbacks; /// Informative callbacks for vkAllocateMemory, vkFreeMemory. Optional. /** Optional, can be null. */ const VmaDeviceMemoryCallbacks* pDeviceMemoryCallbacks; /** \brief Maximum number of additional frames that are in use at the same time as current frame. This value is used only when you make allocations with VMA_ALLOCATION_CREATE_CAN_BECOME_LOST_BIT flag. Such allocation cannot become lost if allocation.lastUseFrameIndex >= allocator.currentFrameIndex - frameInUseCount. For example, if you double-buffer your command buffers, so resources used for rendering in previous frame may still be in use by the GPU at the moment you allocate resources needed for the current frame, set this value to 1. If you want to allow any allocations other than used in the current frame to become lost, set this value to 0. */ uint32_t frameInUseCount; /** \brief Either null or a pointer to an array of limits on maximum number of bytes that can be allocated out of particular Vulkan memory heap. If not NULL, it must be a pointer to an array of VkPhysicalDeviceMemoryProperties::memoryHeapCount elements, defining limit on maximum number of bytes that can be allocated out of particular Vulkan memory heap. Any of the elements may be equal to VK_WHOLE_SIZE, which means no limit on that heap. This is also the default in case of pHeapSizeLimit = NULL. If there is a limit defined for a heap: - If user tries to allocate more memory from that heap using this allocator, the allocation fails with VK_ERROR_OUT_OF_DEVICE_MEMORY. - If the limit is smaller than heap size reported in VkMemoryHeap::size, the value of this limit will be reported instead when using vmaGetMemoryProperties(). Warning! Using this feature may not be equivalent to installing a GPU with smaller amount of memory, because graphics driver doesn't necessary fail new allocations with VK_ERROR_OUT_OF_DEVICE_MEMORY result when memory capacity is exceeded. It may return success and just silently migrate some device memory blocks to system RAM. */ const VkDeviceSize* pHeapSizeLimit; /** \brief Pointers to Vulkan functions. Can be null if you leave define VMA_STATIC_VULKAN_FUNCTIONS 1. If you leave define VMA_STATIC_VULKAN_FUNCTIONS 1 in configuration section, you can pass null as this member, because the library will fetch pointers to Vulkan functions internally in a static way, like: vulkanFunctions.vkAllocateMemory = &vkAllocateMemory; Fill this member if you want to provide your own pointers to Vulkan functions, e.g. fetched using vkGetInstanceProcAddr() and vkGetDeviceProcAddr(). */ const VmaVulkanFunctions* pVulkanFunctions; } VmaAllocatorCreateInfo; /// Creates Allocator object. VkResult vmaCreateAllocator( const VmaAllocatorCreateInfo* pCreateInfo, VmaAllocator* pAllocator); /// Destroys allocator object. void vmaDestroyAllocator( VmaAllocator allocator); /** PhysicalDeviceProperties are fetched from physicalDevice by the allocator. You can access it here, without fetching it again on your own. */ void vmaGetPhysicalDeviceProperties( VmaAllocator allocator, const VkPhysicalDeviceProperties** ppPhysicalDeviceProperties); /** PhysicalDeviceMemoryProperties are fetched from physicalDevice by the allocator. You can access it here, without fetching it again on your own. */ void vmaGetMemoryProperties( VmaAllocator allocator, const VkPhysicalDeviceMemoryProperties** ppPhysicalDeviceMemoryProperties); /** \brief Given Memory Type Index, returns Property Flags of this memory type. This is just a convenience function. Same information can be obtained using vmaGetMemoryProperties(). */ void vmaGetMemoryTypeProperties( VmaAllocator allocator, uint32_t memoryTypeIndex, VkMemoryPropertyFlags* pFlags); /** \brief Sets index of the current frame. This function must be used if you make allocations with #VMA_ALLOCATION_CREATE_CAN_BECOME_LOST_BIT and #VMA_ALLOCATION_CREATE_CAN_MAKE_OTHER_LOST_BIT flags to inform the allocator when a new frame begins. Allocations queried using vmaGetAllocationInfo() cannot become lost in the current frame. */ void vmaSetCurrentFrameIndex( VmaAllocator allocator, uint32_t frameIndex); /** \brief Calculated statistics of memory usage in entire allocator. */ typedef struct VmaStatInfo { /// Number of VkDeviceMemory Vulkan memory blocks allocated. uint32_t blockCount; /// Number of #VmaAllocation allocation objects allocated. uint32_t allocationCount; /// Number of free ranges of memory between allocations. uint32_t unusedRangeCount; /// Total number of bytes occupied by all allocations. VkDeviceSize usedBytes; /// Total number of bytes occupied by unused ranges. VkDeviceSize unusedBytes; VkDeviceSize allocationSizeMin, allocationSizeAvg, allocationSizeMax; VkDeviceSize unusedRangeSizeMin, unusedRangeSizeAvg, unusedRangeSizeMax; } VmaStatInfo; /// General statistics from current state of Allocator. typedef struct VmaStats { VmaStatInfo memoryType[VK_MAX_MEMORY_TYPES]; VmaStatInfo memoryHeap[VK_MAX_MEMORY_HEAPS]; VmaStatInfo total; } VmaStats; /// Retrieves statistics from current state of the Allocator. void vmaCalculateStats( VmaAllocator allocator, VmaStats* pStats); #define VMA_STATS_STRING_ENABLED 1 #if VMA_STATS_STRING_ENABLED /// Builds and returns statistics as string in JSON format. /** @param[out] ppStatsString Must be freed using vmaFreeStatsString() function. */ void vmaBuildStatsString( VmaAllocator allocator, char** ppStatsString, VkBool32 detailedMap); void vmaFreeStatsString( VmaAllocator allocator, char* pStatsString); #endif // #if VMA_STATS_STRING_ENABLED /** \struct VmaPool \brief Represents custom memory pool Fill structure VmaPoolCreateInfo and call function vmaCreatePool() to create it. Call function vmaDestroyPool() to destroy it. For more information see [Custom memory pools](@ref choosing_memory_type_custom_memory_pools). */ VK_DEFINE_HANDLE(VmaPool) typedef enum VmaMemoryUsage { /** No intended memory usage specified. Use other members of VmaAllocationCreateInfo to specify your requirements. */ VMA_MEMORY_USAGE_UNKNOWN = 0, /** Memory will be used on device only, so fast access from the device is preferred. It usually means device-local GPU (video) memory. No need to be mappable on host. 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 (constant) buffers, and majority of other types of resources used by device. Allocation may still end up in HOST_VISIBLE memory on some implementations. In such case, you are free to map it. You can use #VMA_ALLOCATION_CREATE_MAPPED_BIT with this usage type. */ VMA_MEMORY_USAGE_GPU_ONLY = 1, /** Memory will be mappable on host. It usually means CPU (system) memory. Resources created in this pool may still be accessible to the device, but access to them can be slower. Guarantees to be HOST_VISIBLE and HOST_COHERENT. CPU read may be uncached. It is roughly equivalent of D3D12_HEAP_TYPE_UPLOAD. Usage: Staging copy of resources used as transfer source. */ VMA_MEMORY_USAGE_CPU_ONLY = 2, /** Memory that is both mappable on host (guarantees to be HOST_VISIBLE) and preferably fast to access by GPU. CPU reads may be uncached and very slow. Usage: Resources written frequently by host (dynamic), read by device. E.g. textures, vertex buffers, uniform buffers updated every frame or every draw call. */ VMA_MEMORY_USAGE_CPU_TO_GPU = 3, /** Memory mappable on host (guarantees to be HOST_VISIBLE) and cached. It is roughly equivalent of D3D12_HEAP_TYPE_READBACK. Usage: - Resources written by device, read by host - results of some computations, e.g. screen capture, average scene luminance for HDR tone mapping. - Any resources read or accessed randomly on host, e.g. CPU-side copy of vertex buffer used as source of transfer, but also used for collision detection. */ VMA_MEMORY_USAGE_GPU_TO_CPU = 4, VMA_MEMORY_USAGE_MAX_ENUM = 0x7FFFFFFF } VmaMemoryUsage; /// 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. This flag must also be used for host visible resources that you want to map simultaneously because otherwise they might end up as regions of the same VkDeviceMemory, while mapping same VkDeviceMemory multiple times simultaneously is illegal. 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. Is it valid to use this flag for allocation made from memory type that is not HOST_VISIBLE. This flag is then ignored and memory is not mapped. This is useful if you need an allocation that is efficient to use on GPU (DEVICE_LOCAL) and still want to map it directly if possible on platforms that support it (e.g. Intel GPU). You should not use this flag together with #VMA_ALLOCATION_CREATE_CAN_BECOME_LOST_BIT. */ VMA_ALLOCATION_CREATE_MAPPED_BIT = 0x00000004, /** Allocation created with this flag can become lost as a result of another allocation with #VMA_ALLOCATION_CREATE_CAN_MAKE_OTHER_LOST_BIT flag, so you must check it before use. To check if allocation is not lost, call vmaGetAllocationInfo() and check if VmaAllocationInfo::deviceMemory is not VK_NULL_HANDLE. For details about supporting lost allocations, see Lost Allocations chapter of User Guide on Main Page. You should not use this flag together with #VMA_ALLOCATION_CREATE_MAPPED_BIT. */ VMA_ALLOCATION_CREATE_CAN_BECOME_LOST_BIT = 0x00000008, /** While creating allocation using this flag, other allocations that were created with flag #VMA_ALLOCATION_CREATE_CAN_BECOME_LOST_BIT can become lost. For details about supporting lost allocations, see Lost Allocations chapter of User Guide on Main Page. */ VMA_ALLOCATION_CREATE_CAN_MAKE_OTHER_LOST_BIT = 0x00000010, /** Set this flag to treat VmaAllocationCreateInfo::pUserData as pointer to a null-terminated string. Instead of copying pointer value, a local copy of the string is made and stored in allocation's pUserData. The string is automatically freed together with the allocation. It is also used in vmaBuildStatsString(). */ VMA_ALLOCATION_CREATE_USER_DATA_COPY_STRING_BIT = 0x00000020, VMA_ALLOCATION_CREATE_FLAG_BITS_MAX_ENUM = 0x7FFFFFFF } VmaAllocationCreateFlagBits; typedef VkFlags VmaAllocationCreateFlags; typedef struct VmaAllocationCreateInfo { /// 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 prefered. \n If pool is not null, this member is ignored. */ VkMemoryPropertyFlags preferredFlags; /** \brief 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. \n If pool is not null, this member is ignored. */ uint32_t memoryTypeBits; /** \brief Pool that this allocation should be created in. Leave VK_NULL_HANDLE to allocate from default pool. If not null, members: usage, requiredFlags, preferredFlags, memoryTypeBits are ignored. */ VmaPool 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. */ void* pUserData; } VmaAllocationCreateInfo; /** \brief Helps to find memoryTypeIndex, given memoryTypeBits and VmaAllocationCreateInfo. This algorithm tries to find a memory type that: - Is allowed by memoryTypeBits. - Contains all the flags from pAllocationCreateInfo->requiredFlags. - Matches intended usage. - Has as many flags from pAllocationCreateInfo->preferredFlags as possible. \return Returns VK_ERROR_FEATURE_NOT_PRESENT if not found. Receiving such result from this function or any other allocating function probably means that your device doesn't support any memory type with requested features for the specific type of resource you want to use it for. Please check parameters of your resource, like image layout (OPTIMAL versus LINEAR) or mip level count. */ VkResult vmaFindMemoryTypeIndex( VmaAllocator allocator, uint32_t memoryTypeBits, const VmaAllocationCreateInfo* pAllocationCreateInfo, uint32_t* pMemoryTypeIndex); /** \brief Helps to find memoryTypeIndex, given VkBufferCreateInfo and VmaAllocationCreateInfo. It can be useful e.g. to determine value to be used as VmaPoolCreateInfo::memoryTypeIndex. It internally creates a temporary, dummy buffer that never has memory bound. It is just a convenience function, equivalent to calling: - vkCreateBuffer - vkGetBufferMemoryRequirements - vmaFindMemoryTypeIndex - vkDestroyBuffer */ VkResult vmaFindMemoryTypeIndexForBufferInfo( VmaAllocator allocator, const VkBufferCreateInfo* pBufferCreateInfo, const VmaAllocationCreateInfo* pAllocationCreateInfo, uint32_t* pMemoryTypeIndex); /** \brief Helps to find memoryTypeIndex, given VkImageCreateInfo and VmaAllocationCreateInfo. It can be useful e.g. to determine value to be used as VmaPoolCreateInfo::memoryTypeIndex. It internally creates a temporary, dummy image that never has memory bound. It is just a convenience function, equivalent to calling: - vkCreateImage - vkGetImageMemoryRequirements - vmaFindMemoryTypeIndex - vkDestroyImage */ VkResult vmaFindMemoryTypeIndexForImageInfo( VmaAllocator allocator, const VkImageCreateInfo* pImageCreateInfo, const VmaAllocationCreateInfo* pAllocationCreateInfo, uint32_t* pMemoryTypeIndex); /// Flags to be passed as VmaPoolCreateInfo::flags. typedef enum VmaPoolCreateFlagBits { /** \brief Use this flag if you always allocate only buffers and linear images or only optimal images out of this pool and so Buffer-Image Granularity can be ignored. This is na optional optimization flag. If you always allocate using vmaCreateBuffer(), vmaCreateImage(), vmaAllocateMemoryForBuffer(), then you don't need to use it because allocator knows exact type of your allocations so it can handle Buffer-Image Granularity in the optimal way. If you also allocate using vmaAllocateMemoryForImage() or vmaAllocateMemory(), exact type of such allocations is not known, so allocator must be conservative in handling Buffer-Image Granularity, which can lead to suboptimal allocation (wasted memory). In that case, if you can make sure you always allocate only buffers and linear images or only optimal images out of this pool, use this flag to make allocator disregard Buffer-Image Granularity and so make allocations more optimal. */ VMA_POOL_CREATE_IGNORE_BUFFER_IMAGE_GRANULARITY_BIT = 0x00000002, VMA_POOL_CREATE_FLAG_BITS_MAX_ENUM = 0x7FFFFFFF } VmaPoolCreateFlagBits; typedef VkFlags VmaPoolCreateFlags; /** \brief Describes parameter of created #VmaPool. */ typedef struct VmaPoolCreateInfo { /** \brief Vulkan memory type index to allocate this pool from. */ uint32_t memoryTypeIndex; /** \brief Use combination of #VmaPoolCreateFlagBits. */ VmaPoolCreateFlags flags; /** \brief Size of a single VkDeviceMemory block to be allocated as part of this pool, in bytes. Optional. Leave 0 to use default. */ VkDeviceSize blockSize; /** \brief Minimum number of blocks to be always allocated in this pool, even if they stay empty. Set to 0 to have no preallocated blocks and let the pool be completely empty. */ size_t minBlockCount; /** \brief Maximum number of blocks that can be allocated in this pool. Optional. Optional. Set to 0 to use SIZE_MAX, which means no limit. Set to same value as minBlockCount to have fixed amount of memory allocated throuout whole lifetime of this pool. */ size_t maxBlockCount; /** \brief Maximum number of additional frames that are in use at the same time as current frame. This value is used only when you make allocations with #VMA_ALLOCATION_CREATE_CAN_BECOME_LOST_BIT flag. Such allocation cannot become lost if allocation.lastUseFrameIndex >= allocator.currentFrameIndex - frameInUseCount. For example, if you double-buffer your command buffers, so resources used for rendering in previous frame may still be in use by the GPU at the moment you allocate resources needed for the current frame, set this value to 1. If you want to allow any allocations other than used in the current frame to become lost, set this value to 0. */ uint32_t frameInUseCount; } VmaPoolCreateInfo; /** \brief Describes parameter of existing #VmaPool. */ typedef struct VmaPoolStats { /** \brief Total amount of VkDeviceMemory allocated from Vulkan for this pool, in bytes. */ VkDeviceSize size; /** \brief Total number of bytes in the pool not used by any #VmaAllocation. */ VkDeviceSize unusedSize; /** \brief Number of #VmaAllocation objects created from this pool that were not destroyed or lost. */ size_t allocationCount; /** \brief Number of continuous memory ranges in the pool not used by any #VmaAllocation. */ size_t unusedRangeCount; /** \brief Size of the largest continuous free memory region. Making a new allocation of that size is not guaranteed to succeed because of possible additional margin required to respect alignment and buffer/image granularity. */ VkDeviceSize unusedRangeSizeMax; } VmaPoolStats; /** \brief Allocates Vulkan device memory and creates #VmaPool object. @param allocator Allocator object. @param pCreateInfo Parameters of pool to create. @param[out] pPool Handle to created pool. */ VkResult vmaCreatePool( VmaAllocator allocator, const VmaPoolCreateInfo* pCreateInfo, VmaPool* pPool); /** \brief Destroys #VmaPool object and frees Vulkan device memory. */ void vmaDestroyPool( VmaAllocator allocator, VmaPool pool); /** \brief Retrieves statistics of existing #VmaPool object. @param allocator Allocator object. @param pool Pool object. @param[out] pPoolStats Statistics of specified pool. */ void vmaGetPoolStats( VmaAllocator allocator, VmaPool pool, VmaPoolStats* pPoolStats); /** \brief Marks all allocations in given pool as lost if they are not used in current frame or VmaPoolCreateInfo::frameInUseCount back from now. @param allocator Allocator object. @param pool Pool. @param[out] pLostAllocationCount Number of allocations marked as lost. Optional - pass null if you don't need this information. */ void vmaMakePoolAllocationsLost( VmaAllocator allocator, VmaPool pool, size_t* pLostAllocationCount); /** \struct VmaAllocation \brief Represents single memory allocation. It may be either dedicated block of VkDeviceMemory or a specific region of a bigger block of this type plus unique offset. There are multiple ways to create such object. You need to fill structure VmaAllocationCreateInfo. For more information see [Choosing memory type](@ref choosing_memory_type). Although the library provides convenience functions that create Vulkan buffer or image, allocate memory for it and bind them together, binding of the allocation to a buffer or an image is out of scope of the allocation itself. Allocation object can exist without buffer/image bound, binding can be done manually by the user, and destruction of it can be done independently of destruction of the allocation. The object also remembers its size and some other information. To retrieve this information, use function vmaGetAllocationInfo() and inspect returned structure VmaAllocationInfo. Some kinds allocations can be in lost state. For more information, see [Lost allocations](@ref lost_allocations). */ VK_DEFINE_HANDLE(VmaAllocation) /** \brief Parameters of #VmaAllocation objects, that can be retrieved using function vmaGetAllocationInfo(). */ 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 deviceMemory; /** \brief Offset into deviceMemory object to the beginning of this allocation, in bytes. (deviceMemory, offset) pair is unique to this allocation. It can change after call to vmaDefragment() if this allocation is passed to the function, or if allocation is lost. */ VkDeviceSize offset; /** \brief Size of this allocation, in bytes. It never changes, unless allocation is lost. */ VkDeviceSize size; /** \brief Pointer to the beginning of this allocation as mapped data. If the allocation hasn't been mapped using vmaMapMemory() and hasn't been created with #VMA_ALLOCATION_CREATE_MAPPED_BIT flag, this value null. It can change after call to vmaMapMemory(), vmaUnmapMemory(). It can also change after call to vmaDefragment() if this allocation is passed to the function. */ void* pMappedData; /** \brief Custom general-purpose pointer that was passed as VmaAllocationCreateInfo::pUserData or set using vmaSetAllocationUserData(). It can change after call to vmaSetAllocationUserData() for this allocation. */ void* pUserData; } VmaAllocationInfo; /** \brief General purpose memory allocation. @param[out] pAllocation Handle to allocated memory. @param[out] pAllocationInfo Optional. Information about allocated memory. It can be later fetched using function vmaGetAllocationInfo(). You should free the memory using vmaFreeMemory(). It is recommended to use vmaAllocateMemoryForBuffer(), vmaAllocateMemoryForImage(), vmaCreateBuffer(), vmaCreateImage() instead whenever possible. */ VkResult vmaAllocateMemory( VmaAllocator allocator, const VkMemoryRequirements* pVkMemoryRequirements, const VmaAllocationCreateInfo* pCreateInfo, VmaAllocation* pAllocation, VmaAllocationInfo* pAllocationInfo); /** @param[out] pAllocation Handle to allocated memory. @param[out] pAllocationInfo Optional. Information about allocated memory. It can be later fetched using function vmaGetAllocationInfo(). You should free the memory using vmaFreeMemory(). */ VkResult vmaAllocateMemoryForBuffer( VmaAllocator allocator, VkBuffer buffer, const VmaAllocationCreateInfo* pCreateInfo, VmaAllocation* pAllocation, VmaAllocationInfo* pAllocationInfo); /// Function similar to vmaAllocateMemoryForBuffer(). VkResult vmaAllocateMemoryForImage( VmaAllocator allocator, VkImage image, const VmaAllocationCreateInfo* pCreateInfo, VmaAllocation* pAllocation, VmaAllocationInfo* pAllocationInfo); /// Frees memory previously allocated using vmaAllocateMemory(), vmaAllocateMemoryForBuffer(), or vmaAllocateMemoryForImage(). void vmaFreeMemory( VmaAllocator allocator, VmaAllocation allocation); /** \brief Returns current information about specified allocation and atomically marks it as used in current frame. Current paramters of given allocation are returned in pAllocationInfo. This function also atomically "touches" allocation - marks it as used in current frame, just like vmaTouchAllocation(). If the allocation is in lost state, pAllocationInfo->deviceMemory == VK_NULL_HANDLE. Although this function uses atomics and doesn't lock any mutex, so it should be quite efficient, you can avoid calling it too often. - You can retrieve same VmaAllocationInfo structure while creating your resource, from function vmaCreateBuffer(), vmaCreateImage(). You can remember it if you are sure parameters don't change (e.g. due to defragmentation or allocation becoming lost). - If you just want to check if allocation is not lost, vmaTouchAllocation() will work faster. */ void vmaGetAllocationInfo( VmaAllocator allocator, VmaAllocation allocation, VmaAllocationInfo* pAllocationInfo); /** \brief Returns VK_TRUE if allocation is not lost and atomically marks it as used in current frame. If the allocation has been created with #VMA_ALLOCATION_CREATE_CAN_BECOME_LOST_BIT flag, this function returns VK_TRUE if it's not in lost state, so it can still be used. It then also atomically "touches" the allocation - marks it as used in current frame, so that you can be sure it won't become lost in current frame or next frameInUseCount frames. If the allocation is in lost state, the function returns VK_FALSE. Memory of such allocation, as well as buffer or image bound to it, should not be used. Lost allocation and the buffer/image still need to be destroyed. If the allocation has been created without #VMA_ALLOCATION_CREATE_CAN_BECOME_LOST_BIT flag, this function always returns VK_TRUE. */ VkBool32 vmaTouchAllocation( VmaAllocator allocator, VmaAllocation allocation); /** \brief Sets pUserData in given allocation to new value. If the allocation was created with VMA_ALLOCATION_CREATE_USER_DATA_COPY_STRING_BIT, pUserData must be either null, or pointer to a null-terminated string. The function makes local copy of the string and sets it as allocation's pUserData. String passed as pUserData doesn't need to be valid for whole lifetime of the allocation - you can free it after this call. String previously pointed by allocation's pUserData is freed from memory. If the flag was not used, the value of pointer pUserData is just copied to allocation's pUserData. It is opaque, so you can use it however you want - e.g. as a pointer, ordinal number or some handle to you own data. */ void vmaSetAllocationUserData( VmaAllocator allocator, VmaAllocation allocation, void* pUserData); /** \brief Creates new allocation that is in lost state from the beginning. It can be useful if you need a dummy, non-null allocation. You still need to destroy created object using vmaFreeMemory(). Returned allocation is not tied to any specific memory pool or memory type and not bound to any image or buffer. It has size = 0. It cannot be turned into a real, non-empty allocation. */ void vmaCreateLostAllocation( VmaAllocator allocator, VmaAllocation* pAllocation); /** \brief Maps memory represented by given allocation and returns pointer to it. Maps memory represented by given allocation to make it accessible to CPU code. When succeeded, *ppData contains pointer to first byte of this memory. If the allocation is part of bigger VkDeviceMemory block, the pointer is correctly offseted to the beginning of region assigned to this particular allocation. Mapping is internally reference-counted and synchronized, so despite raw Vulkan function vkMapMemory() cannot be used to map same block of VkDeviceMemory multiple times simultaneously, it is safe to call this function on allocations assigned to the same memory block. Actual Vulkan memory will be mapped on first mapping and unmapped on last unmapping. If the function succeeded, you must call vmaUnmapMemory() to unmap the allocation when mapping is no longer needed or before freeing the allocation, at the latest. It also safe to call this function multiple times on the same allocation. You must call vmaUnmapMemory() same number of times as you called vmaMapMemory(). It is also safe to call this function on allocation created with #VMA_ALLOCATION_CREATE_MAPPED_BIT flag. Its memory stays mapped all the time. You must still call vmaUnmapMemory() same number of times as you called vmaMapMemory(). You must not call vmaUnmapMemory() additional time to free the "0-th" mapping made automatically due to #VMA_ALLOCATION_CREATE_MAPPED_BIT flag. This function fails when used on allocation made in memory type that is not HOST_VISIBLE. This function always fails when called for allocation that was created with #VMA_ALLOCATION_CREATE_CAN_BECOME_LOST_BIT flag. Such allocations cannot be mapped. */ VkResult vmaMapMemory( VmaAllocator allocator, VmaAllocation allocation, void** ppData); /** \brief Unmaps memory represented by given allocation, mapped previously using vmaMapMemory(). For details, see description of vmaMapMemory(). */ void vmaUnmapMemory( VmaAllocator allocator, VmaAllocation allocation); /** \brief Optional configuration parameters to be passed to function vmaDefragment(). */ 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; /** \brief Maximum number of allocations that can be moved to different place. Default is UINT32_MAX, which means no limit. */ uint32_t maxAllocationsToMove; } VmaDefragmentationInfo; /** \brief Statistics returned by function vmaDefragment(). */ typedef struct VmaDefragmentationStats { /// Total number of bytes that have been copied while moving allocations to different places. VkDeviceSize bytesMoved; /// Total number of bytes that have been released to the system by freeing empty VkDeviceMemory objects. VkDeviceSize bytesFreed; /// Number of allocations that have been moved to different places. uint32_t allocationsMoved; /// Number of empty VkDeviceMemory objects that have been released to the system. uint32_t deviceMemoryBlocksFreed; } VmaDefragmentationStats; /** \brief Compacts memory by moving allocations. @param pAllocations Array of allocations that can be moved during this compation. @param allocationCount Number of elements in pAllocations and pAllocationsChanged arrays. @param[out] pAllocationsChanged Array of boolean values that will indicate whether matching allocation in pAllocations array has been moved. This parameter is optional. Pass null if you don't need this information. @param pDefragmentationInfo Configuration parameters. Optional - pass null to use default values. @param[out] pDefragmentationStats Statistics returned by the function. Optional - pass null if you don't need this information. @return VK_SUCCESS if completed, VK_INCOMPLETE if succeeded but didn't make all possible optimizations because limits specified in pDefragmentationInfo have been reached, negative error code in case of error. This function works by moving allocations to different places (different VkDeviceMemory objects and/or different offsets) in order to optimize memory usage. Only allocations that are in pAllocations array can be moved. All other allocations are considered nonmovable in this call. Basic rules: - Only allocations made in memory types that have VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT flag can be compacted. You may pass other allocations but it makes no sense - these will never be moved. - You may pass allocations made with #VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT but it makes no sense - they will never be moved. - Both allocations made with or without #VMA_ALLOCATION_CREATE_MAPPED_BIT flag can be compacted. If not persistently mapped, memory will be mapped temporarily inside this function if needed. - You must not pass same #VmaAllocation object multiple times in pAllocations array. The function also frees empty VkDeviceMemory blocks. After allocation has been moved, its VmaAllocationInfo::deviceMemory and/or VmaAllocationInfo::offset changes. You must query them again using vmaGetAllocationInfo() if you need them. If an allocation has been moved, data in memory is copied to new place automatically, but if it was bound to a buffer or an image, you must destroy that object yourself, create new one and bind it to the new memory pointed by the allocation. You must use vkDestroyBuffer(), vkDestroyImage(), vkCreateBuffer(), vkCreateImage() for that purpose and NOT vmaDestroyBuffer(), vmaDestroyImage(), vmaCreateBuffer(), vmaCreateImage()! Example: \code VkDevice device = ...; VmaAllocator allocator = ...; std::vector buffers = ...; std::vector allocations = ...; std::vector allocationsChanged(allocations.size()); vmaDefragment(allocator, allocations.data(), allocations.size(), allocationsChanged.data(), nullptr, nullptr); for(size_t i = 0; i < allocations.size(); ++i) { if(allocationsChanged[i]) { VmaAllocationInfo allocInfo; vmaGetAllocationInfo(allocator, allocations[i], &allocInfo); vkDestroyBuffer(device, buffers[i], nullptr); VkBufferCreateInfo bufferInfo = ...; vkCreateBuffer(device, &bufferInfo, nullptr, &buffers[i]); // You can make dummy call to vkGetBufferMemoryRequirements here to silence validation layer warning. vkBindBufferMemory(device, buffers[i], allocInfo.deviceMemory, allocInfo.offset); } } \endcode Note: Please don't expect memory to be fully compacted after this call. Algorithms inside are based on some heuristics that try to maximize number of Vulkan memory blocks to make totally empty to release them, as well as to maximimze continuous empty space inside remaining blocks, while minimizing the number and size of data that needs to be moved. Some fragmentation still remains after this call. This is normal. Warning: This function is not 100% correct according to Vulkan specification. Use it at your own risk. That's because Vulkan doesn't guarantee that memory requirements (size and alignment) for a new buffer or image are consistent. They may be different even for subsequent calls with the same parameters. It really does happen on some platforms, especially with images. Warning: This function may be time-consuming, so you shouldn't call it too often (like every frame or after every resource creation/destruction). You can call it on special occasions (like when reloading a game level or when you just destroyed a lot of objects). */ VkResult vmaDefragment( VmaAllocator allocator, VmaAllocation* pAllocations, size_t allocationCount, VkBool32* pAllocationsChanged, const VmaDefragmentationInfo *pDefragmentationInfo, VmaDefragmentationStats* pDefragmentationStats); /** \brief Binds buffer to allocation. Binds specified buffer to region of memory represented by specified allocation. Gets VkDeviceMemory handle and offset from the allocation. If you want to create a buffer, allocate memory for it and bind them together separately, you should use this function for binding instead of standard vkBindBufferMemory(), because it ensures proper synchronization so that when a VkDeviceMemory object is used by multiple allocations, calls to vkBind*Memory() or vkMapMemory() won't happen from multiple threads simultaneously (which is illegal in Vulkan). It is recommended to use function vmaCreateBuffer() instead of this one. */ VkResult vmaBindBufferMemory( VmaAllocator allocator, VmaAllocation allocation, VkBuffer buffer); /** \brief Binds image to allocation. Binds specified image to region of memory represented by specified allocation. Gets VkDeviceMemory handle and offset from the allocation. If you want to create an image, allocate memory for it and bind them together separately, you should use this function for binding instead of standard vkBindImageMemory(), because it ensures proper synchronization so that when a VkDeviceMemory object is used by multiple allocations, calls to vkBind*Memory() or vkMapMemory() won't happen from multiple threads simultaneously (which is illegal in Vulkan). It is recommended to use function vmaCreateImage() instead of this one. */ VkResult vmaBindImageMemory( VmaAllocator allocator, VmaAllocation allocation, VkImage image); /** @param[out] pBuffer Buffer that was created. @param[out] pAllocation Allocation that was created. @param[out] pAllocationInfo Optional. Information about allocated memory. It can be later fetched using function vmaGetAllocationInfo(). This function automatically: -# Creates buffer. -# Allocates appropriate memory for it. -# Binds the buffer with the memory. If any of these operations fail, buffer and allocation are not created, returned value is negative error code, *pBuffer and *pAllocation are null. If the function succeeded, you must destroy both buffer and allocation when you 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, 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 allocation for this buffer, just like when using VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT. */ VkResult vmaCreateBuffer( VmaAllocator allocator, const VkBufferCreateInfo* pBufferCreateInfo, const VmaAllocationCreateInfo* pAllocationCreateInfo, VkBuffer* pBuffer, VmaAllocation* pAllocation, VmaAllocationInfo* pAllocationInfo); /** \brief Destroys Vulkan buffer and frees allocated memory. This is just a convenience function equivalent to: \code vkDestroyBuffer(device, buffer, allocationCallbacks); vmaFreeMemory(allocator, allocation); \endcode It it safe to pass null as buffer and/or allocation. */ void vmaDestroyBuffer( VmaAllocator allocator, VkBuffer buffer, VmaAllocation allocation); /// Function similar to vmaCreateBuffer(). VkResult vmaCreateImage( VmaAllocator allocator, const VkImageCreateInfo* pImageCreateInfo, const VmaAllocationCreateInfo* pAllocationCreateInfo, VkImage* pImage, VmaAllocation* pAllocation, VmaAllocationInfo* pAllocationInfo); /** \brief Destroys Vulkan image and frees allocated memory. This is just a convenience function equivalent to: \code vkDestroyImage(device, image, allocationCallbacks); vmaFreeMemory(allocator, allocation); \endcode It it safe to pass null as image and/or allocation. */ void vmaDestroyImage( VmaAllocator allocator, VkImage image, VmaAllocation allocation); #ifdef __cplusplus } #endif #endif // AMD_VULKAN_MEMORY_ALLOCATOR_H // For Visual Studio IntelliSense. #ifdef __INTELLISENSE__ #define VMA_IMPLEMENTATION #endif #ifdef VMA_IMPLEMENTATION #undef VMA_IMPLEMENTATION #include #include #include /******************************************************************************* CONFIGURATION SECTION Define some of these macros before each #include of this header or change them here if you need other then default behavior depending on your environment. */ /* Define this macro to 1 to make the library fetch pointers to Vulkan functions internally, like: vulkanFunctions.vkAllocateMemory = &vkAllocateMemory; Define to 0 if you are going to provide you own pointers to Vulkan functions via VmaAllocatorCreateInfo::pVulkanFunctions. */ #if !defined(VMA_STATIC_VULKAN_FUNCTIONS) && !defined(VK_NO_PROTOTYPES) #define VMA_STATIC_VULKAN_FUNCTIONS 1 #endif // Define this macro to 1 to make the library use STL containers instead of its own implementation. //#define VMA_USE_STL_CONTAINERS 1 /* Set this macro to 1 to make the library including and using STL containers: std::pair, std::vector, std::list, std::unordered_map. Set it to 0 or undefined to make the library using its own implementation of the containers. */ #if VMA_USE_STL_CONTAINERS #define VMA_USE_STL_VECTOR 1 #define VMA_USE_STL_UNORDERED_MAP 1 #define VMA_USE_STL_LIST 1 #endif #if VMA_USE_STL_VECTOR #include #endif #if VMA_USE_STL_UNORDERED_MAP #include #endif #if VMA_USE_STL_LIST #include #endif /* Following headers are used in this CONFIGURATION section only, so feel free to remove them if not needed. */ #include // for assert #include // for min, max #include // for std::mutex #include // for std::atomic #if !defined(_WIN32) && !defined(__APPLE__) #include // for aligned_alloc() #endif #ifndef VMA_NULL // Value used as null pointer. Define it to e.g.: nullptr, NULL, 0, (void*)0. #define VMA_NULL nullptr #endif #if defined(__APPLE__) || defined(__ANDROID__) #include void *aligned_alloc(size_t alignment, size_t size) { // alignment must be >= sizeof(void*) if(alignment < sizeof(void*)) { alignment = sizeof(void*); } void *pointer; if(posix_memalign(&pointer, alignment, size) == 0) return pointer; return VMA_NULL; } #endif // Normal assert to check for programmer's errors, especially in Debug configuration. #ifndef VMA_ASSERT #ifdef _DEBUG #define VMA_ASSERT(expr) assert(expr) #else #define VMA_ASSERT(expr) #endif #endif // Assert that will be called very often, like inside data structures e.g. operator[]. // Making it non-empty can make program slow. #ifndef VMA_HEAVY_ASSERT #ifdef _DEBUG #define VMA_HEAVY_ASSERT(expr) //VMA_ASSERT(expr) #else #define VMA_HEAVY_ASSERT(expr) #endif #endif #ifndef VMA_ALIGN_OF #define VMA_ALIGN_OF(type) (__alignof(type)) #endif #ifndef VMA_SYSTEM_ALIGNED_MALLOC #if defined(_WIN32) #define VMA_SYSTEM_ALIGNED_MALLOC(size, alignment) (_aligned_malloc((size), (alignment))) #else #define VMA_SYSTEM_ALIGNED_MALLOC(size, alignment) (aligned_alloc((alignment), (size) )) #endif #endif #ifndef VMA_SYSTEM_FREE #if defined(_WIN32) #define VMA_SYSTEM_FREE(ptr) _aligned_free(ptr) #else #define VMA_SYSTEM_FREE(ptr) free(ptr) #endif #endif #ifndef VMA_MIN #define VMA_MIN(v1, v2) (std::min((v1), (v2))) #endif #ifndef VMA_MAX #define VMA_MAX(v1, v2) (std::max((v1), (v2))) #endif #ifndef VMA_SWAP #define VMA_SWAP(v1, v2) std::swap((v1), (v2)) #endif #ifndef VMA_SORT #define VMA_SORT(beg, end, cmp) std::sort(beg, end, cmp) #endif #ifndef VMA_DEBUG_LOG #define VMA_DEBUG_LOG(format, ...) /* #define VMA_DEBUG_LOG(format, ...) do { \ printf(format, __VA_ARGS__); \ printf("\n"); \ } while(false) */ #endif // Define this macro to 1 to enable functions: vmaBuildStatsString, vmaFreeStatsString. #if VMA_STATS_STRING_ENABLED static inline void VmaUint32ToStr(char* outStr, size_t strLen, uint32_t num) { snprintf(outStr, strLen, "%u", static_cast(num)); } static inline void VmaUint64ToStr(char* outStr, size_t strLen, uint64_t num) { snprintf(outStr, strLen, "%llu", static_cast(num)); } static inline void VmaPtrToStr(char* outStr, size_t strLen, const void* ptr) { snprintf(outStr, strLen, "%p", ptr); } #endif #ifndef VMA_MUTEX class VmaMutex { public: VmaMutex() { } ~VmaMutex() { } void Lock() { m_Mutex.lock(); } void Unlock() { m_Mutex.unlock(); } private: std::mutex m_Mutex; }; #define VMA_MUTEX VmaMutex #endif /* If providing your own implementation, you need to implement a subset of std::atomic: - Constructor(uint32_t desired) - uint32_t load() const - void store(uint32_t desired) - bool compare_exchange_weak(uint32_t& expected, uint32_t desired) */ #ifndef VMA_ATOMIC_UINT32 #define VMA_ATOMIC_UINT32 std::atomic #endif #ifndef VMA_BEST_FIT /** Main parameter for function assessing how good is a free suballocation for a new allocation request. - Set to 1 to use Best-Fit algorithm - prefer smaller blocks, as close to the size of requested allocations as possible. - Set to 0 to use Worst-Fit algorithm - prefer larger blocks, as large as possible. Experiments in special testing environment showed that Best-Fit algorithm is better. */ #define VMA_BEST_FIT (1) #endif #ifndef VMA_DEBUG_ALWAYS_DEDICATED_MEMORY /** Every allocation will have its own memory block. Define to 1 for debugging purposes only. */ #define VMA_DEBUG_ALWAYS_DEDICATED_MEMORY (0) #endif #ifndef VMA_DEBUG_ALIGNMENT /** Minimum alignment of all suballocations, in bytes. Set to more than 1 for debugging purposes only. Must be power of two. */ #define VMA_DEBUG_ALIGNMENT (1) #endif #ifndef VMA_DEBUG_MARGIN /** Minimum margin between suballocations, in bytes. Set nonzero for debugging purposes only. */ #define VMA_DEBUG_MARGIN (0) #endif #ifndef VMA_DEBUG_GLOBAL_MUTEX /** Set this to 1 for debugging purposes only, to enable single mutex protecting all entry calls to the library. Can be useful for debugging multithreading issues. */ #define VMA_DEBUG_GLOBAL_MUTEX (0) #endif #ifndef VMA_DEBUG_MIN_BUFFER_IMAGE_GRANULARITY /** Minimum value for VkPhysicalDeviceLimits::bufferImageGranularity. Set to more than 1 for debugging purposes only. Must be power of two. */ #define VMA_DEBUG_MIN_BUFFER_IMAGE_GRANULARITY (1) #endif #ifndef VMA_SMALL_HEAP_MAX_SIZE /// Maximum size of a memory heap in Vulkan to consider it "small". #define VMA_SMALL_HEAP_MAX_SIZE (1024ull * 1024 * 1024) #endif #ifndef VMA_DEFAULT_LARGE_HEAP_BLOCK_SIZE /// Default size of a block allocated as single VkDeviceMemory from a "large" heap. #define VMA_DEFAULT_LARGE_HEAP_BLOCK_SIZE (256ull * 1024 * 1024) #endif static const uint32_t VMA_FRAME_INDEX_LOST = UINT32_MAX; /******************************************************************************* END OF CONFIGURATION */ static VkAllocationCallbacks VmaEmptyAllocationCallbacks = { VMA_NULL, VMA_NULL, VMA_NULL, VMA_NULL, VMA_NULL, VMA_NULL }; // Returns number of bits set to 1 in (v). static inline uint32_t VmaCountBitsSet(uint32_t v) { uint32_t c = v - ((v >> 1) & 0x55555555); c = ((c >> 2) & 0x33333333) + (c & 0x33333333); c = ((c >> 4) + c) & 0x0F0F0F0F; c = ((c >> 8) + c) & 0x00FF00FF; c = ((c >> 16) + c) & 0x0000FFFF; return c; } // Aligns given value up to nearest multiply of align value. For example: VmaAlignUp(11, 8) = 16. // Use types like uint32_t, uint64_t as T. template static inline T VmaAlignUp(T val, T align) { return (val + align - 1) / align * align; } // Division with mathematical rounding to nearest number. template inline T VmaRoundDiv(T x, T y) { return (x + (y / (T)2)) / y; } #ifndef VMA_SORT template Iterator VmaQuickSortPartition(Iterator beg, Iterator end, Compare cmp) { Iterator centerValue = end; --centerValue; Iterator insertIndex = beg; for(Iterator memTypeIndex = beg; memTypeIndex < centerValue; ++memTypeIndex) { if(cmp(*memTypeIndex, *centerValue)) { if(insertIndex != memTypeIndex) { VMA_SWAP(*memTypeIndex, *insertIndex); } ++insertIndex; } } if(insertIndex != centerValue) { VMA_SWAP(*insertIndex, *centerValue); } return insertIndex; } template void VmaQuickSort(Iterator beg, Iterator end, Compare cmp) { if(beg < end) { Iterator it = VmaQuickSortPartition(beg, end, cmp); VmaQuickSort(beg, it, cmp); VmaQuickSort(it + 1, end, cmp); } } #define VMA_SORT(beg, end, cmp) VmaQuickSort(beg, end, cmp) #endif // #ifndef VMA_SORT /* Returns true if two memory blocks occupy overlapping pages. ResourceA must be in less memory offset than ResourceB. Algorithm is based on "Vulkan 1.0.39 - A Specification (with all registered Vulkan extensions)" chapter 11.6 "Resource Memory Association", paragraph "Buffer-Image Granularity". */ static inline bool VmaBlocksOnSamePage( VkDeviceSize resourceAOffset, VkDeviceSize resourceASize, VkDeviceSize resourceBOffset, VkDeviceSize pageSize) { VMA_ASSERT(resourceAOffset + resourceASize <= resourceBOffset && resourceASize > 0 && pageSize > 0); VkDeviceSize resourceAEnd = resourceAOffset + resourceASize - 1; VkDeviceSize resourceAEndPage = resourceAEnd & ~(pageSize - 1); VkDeviceSize resourceBStart = resourceBOffset; VkDeviceSize resourceBStartPage = resourceBStart & ~(pageSize - 1); return resourceAEndPage == resourceBStartPage; } enum VmaSuballocationType { VMA_SUBALLOCATION_TYPE_FREE = 0, VMA_SUBALLOCATION_TYPE_UNKNOWN = 1, VMA_SUBALLOCATION_TYPE_BUFFER = 2, VMA_SUBALLOCATION_TYPE_IMAGE_UNKNOWN = 3, VMA_SUBALLOCATION_TYPE_IMAGE_LINEAR = 4, VMA_SUBALLOCATION_TYPE_IMAGE_OPTIMAL = 5, VMA_SUBALLOCATION_TYPE_MAX_ENUM = 0x7FFFFFFF }; /* Returns true if given suballocation types could conflict and must respect VkPhysicalDeviceLimits::bufferImageGranularity. They conflict if one is buffer or linear image and another one is optimal image. If type is unknown, behave conservatively. */ static inline bool VmaIsBufferImageGranularityConflict( VmaSuballocationType suballocType1, VmaSuballocationType suballocType2) { if(suballocType1 > suballocType2) { VMA_SWAP(suballocType1, suballocType2); } switch(suballocType1) { case VMA_SUBALLOCATION_TYPE_FREE: return false; case VMA_SUBALLOCATION_TYPE_UNKNOWN: return true; case VMA_SUBALLOCATION_TYPE_BUFFER: return suballocType2 == VMA_SUBALLOCATION_TYPE_IMAGE_UNKNOWN || suballocType2 == VMA_SUBALLOCATION_TYPE_IMAGE_OPTIMAL; case VMA_SUBALLOCATION_TYPE_IMAGE_UNKNOWN: return suballocType2 == VMA_SUBALLOCATION_TYPE_IMAGE_UNKNOWN || suballocType2 == VMA_SUBALLOCATION_TYPE_IMAGE_LINEAR || suballocType2 == VMA_SUBALLOCATION_TYPE_IMAGE_OPTIMAL; case VMA_SUBALLOCATION_TYPE_IMAGE_LINEAR: return suballocType2 == VMA_SUBALLOCATION_TYPE_IMAGE_OPTIMAL; case VMA_SUBALLOCATION_TYPE_IMAGE_OPTIMAL: return false; default: VMA_ASSERT(0); return true; } } // Helper RAII class to lock a mutex in constructor and unlock it in destructor (at the end of scope). struct VmaMutexLock { public: VmaMutexLock(VMA_MUTEX& mutex, bool useMutex) : m_pMutex(useMutex ? &mutex : VMA_NULL) { if(m_pMutex) { m_pMutex->Lock(); } } ~VmaMutexLock() { if(m_pMutex) { m_pMutex->Unlock(); } } private: VMA_MUTEX* m_pMutex; }; #if VMA_DEBUG_GLOBAL_MUTEX static VMA_MUTEX gDebugGlobalMutex; #define VMA_DEBUG_GLOBAL_MUTEX_LOCK VmaMutexLock debugGlobalMutexLock(gDebugGlobalMutex, true); #else #define VMA_DEBUG_GLOBAL_MUTEX_LOCK #endif // Minimum size of a free suballocation to register it in the free suballocation collection. static const VkDeviceSize VMA_MIN_FREE_SUBALLOCATION_SIZE_TO_REGISTER = 16; /* Performs binary search and returns iterator to first element that is greater or equal to (key), according to comparison (cmp). Cmp should return true if first argument is less than second argument. Returned value is the found element, if present in the collection or place where new element with value (key) should be inserted. */ template static IterT VmaBinaryFindFirstNotLess(IterT beg, IterT end, const KeyT &key, CmpT cmp) { size_t down = 0, up = (end - beg); while(down < up) { const size_t mid = (down + up) / 2; if(cmp(*(beg+mid), key)) { down = mid + 1; } else { up = mid; } } return beg + down; } //////////////////////////////////////////////////////////////////////////////// // Memory allocation static void* VmaMalloc(const VkAllocationCallbacks* pAllocationCallbacks, size_t size, size_t alignment) { if((pAllocationCallbacks != VMA_NULL) && (pAllocationCallbacks->pfnAllocation != VMA_NULL)) { return (*pAllocationCallbacks->pfnAllocation)( pAllocationCallbacks->pUserData, size, alignment, VK_SYSTEM_ALLOCATION_SCOPE_OBJECT); } else { return VMA_SYSTEM_ALIGNED_MALLOC(size, alignment); } } static void VmaFree(const VkAllocationCallbacks* pAllocationCallbacks, void* ptr) { if((pAllocationCallbacks != VMA_NULL) && (pAllocationCallbacks->pfnFree != VMA_NULL)) { (*pAllocationCallbacks->pfnFree)(pAllocationCallbacks->pUserData, ptr); } else { VMA_SYSTEM_FREE(ptr); } } template static T* VmaAllocate(const VkAllocationCallbacks* pAllocationCallbacks) { return (T*)VmaMalloc(pAllocationCallbacks, sizeof(T), VMA_ALIGN_OF(T)); } template static T* VmaAllocateArray(const VkAllocationCallbacks* pAllocationCallbacks, size_t count) { return (T*)VmaMalloc(pAllocationCallbacks, sizeof(T) * count, VMA_ALIGN_OF(T)); } #define vma_new(allocator, type) new(VmaAllocate(allocator))(type) #define vma_new_array(allocator, type, count) new(VmaAllocateArray((allocator), (count)))(type) template static void vma_delete(const VkAllocationCallbacks* pAllocationCallbacks, T* ptr) { ptr->~T(); VmaFree(pAllocationCallbacks, ptr); } template static void vma_delete_array(const VkAllocationCallbacks* pAllocationCallbacks, T* ptr, size_t count) { if(ptr != VMA_NULL) { for(size_t i = count; i--; ) { ptr[i].~T(); } VmaFree(pAllocationCallbacks, ptr); } } // STL-compatible allocator. template class VmaStlAllocator { public: const VkAllocationCallbacks* const m_pCallbacks; typedef T value_type; VmaStlAllocator(const VkAllocationCallbacks* pCallbacks) : m_pCallbacks(pCallbacks) { } template VmaStlAllocator(const VmaStlAllocator& src) : m_pCallbacks(src.m_pCallbacks) { } T* allocate(size_t n) { return VmaAllocateArray(m_pCallbacks, n); } void deallocate(T* p, size_t n) { VmaFree(m_pCallbacks, p); } template bool operator==(const VmaStlAllocator& rhs) const { return m_pCallbacks == rhs.m_pCallbacks; } template bool operator!=(const VmaStlAllocator& rhs) const { return m_pCallbacks != rhs.m_pCallbacks; } VmaStlAllocator& operator=(const VmaStlAllocator& x) = delete; }; #if VMA_USE_STL_VECTOR #define VmaVector std::vector template static void VmaVectorInsert(std::vector& vec, size_t index, const T& item) { vec.insert(vec.begin() + index, item); } template static void VmaVectorRemove(std::vector& vec, size_t index) { vec.erase(vec.begin() + index); } #else // #if VMA_USE_STL_VECTOR /* Class with interface compatible with subset of std::vector. T must be POD because constructors and destructors are not called and memcpy is used for these objects. */ template class VmaVector { public: typedef T value_type; VmaVector(const AllocatorT& allocator) : m_Allocator(allocator), m_pArray(VMA_NULL), m_Count(0), m_Capacity(0) { } VmaVector(size_t count, const AllocatorT& allocator) : m_Allocator(allocator), m_pArray(count ? (T*)VmaAllocateArray(allocator.m_pCallbacks, count) : VMA_NULL), m_Count(count), m_Capacity(count) { } VmaVector(const VmaVector& src) : m_Allocator(src.m_Allocator), m_pArray(src.m_Count ? (T*)VmaAllocateArray(src.m_Allocator.m_pCallbacks, src.m_Count) : VMA_NULL), m_Count(src.m_Count), m_Capacity(src.m_Count) { if(m_Count != 0) { memcpy(m_pArray, src.m_pArray, m_Count * sizeof(T)); } } ~VmaVector() { VmaFree(m_Allocator.m_pCallbacks, m_pArray); } VmaVector& operator=(const VmaVector& rhs) { if(&rhs != this) { resize(rhs.m_Count); if(m_Count != 0) { memcpy(m_pArray, rhs.m_pArray, m_Count * sizeof(T)); } } 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); return m_pArray[index]; } const T& operator[](size_t index) const { VMA_HEAVY_ASSERT(index < m_Count); return m_pArray[index]; } T& front() { VMA_HEAVY_ASSERT(m_Count > 0); return m_pArray[0]; } const T& front() const { VMA_HEAVY_ASSERT(m_Count > 0); return m_pArray[0]; } T& back() { VMA_HEAVY_ASSERT(m_Count > 0); return m_pArray[m_Count - 1]; } const T& back() const { VMA_HEAVY_ASSERT(m_Count > 0); return m_pArray[m_Count - 1]; } 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(m_Allocator, newCapacity) : VMA_NULL; if(m_Count != 0) { memcpy(newArray, m_pArray, m_Count * sizeof(T)); } VmaFree(m_Allocator.m_pCallbacks, m_pArray); m_Capacity = newCapacity; m_pArray = newArray; } } void resize(size_t newCount, bool freeMemory = false) { size_t newCapacity = m_Capacity; if(newCount > m_Capacity) { newCapacity = VMA_MAX(newCount, VMA_MAX(m_Capacity * 3 / 2, (size_t)8)); } else if(freeMemory) { newCapacity = newCount; } if(newCapacity != m_Capacity) { T* const newArray = newCapacity ? VmaAllocateArray(m_Allocator.m_pCallbacks, newCapacity) : VMA_NULL; const size_t elementsToCopy = VMA_MIN(m_Count, newCount); if(elementsToCopy != 0) { memcpy(newArray, m_pArray, elementsToCopy * sizeof(T)); } VmaFree(m_Allocator.m_pCallbacks, m_pArray); m_Capacity = newCapacity; m_pArray = newArray; } m_Count = newCount; } void clear(bool freeMemory = false) { resize(0, freeMemory); } void insert(size_t index, const T& src) { VMA_HEAVY_ASSERT(index <= m_Count); const size_t oldCount = size(); resize(oldCount + 1); if(index < oldCount) { memmove(m_pArray + (index + 1), m_pArray + index, (oldCount - index) * sizeof(T)); } m_pArray[index] = src; } void remove(size_t index) { VMA_HEAVY_ASSERT(index < m_Count); const size_t oldCount = size(); if(index < oldCount - 1) { memmove(m_pArray + index, m_pArray + (index + 1), (oldCount - index - 1) * sizeof(T)); } resize(oldCount - 1); } void push_back(const T& src) { const size_t newIndex = size(); resize(newIndex + 1); m_pArray[newIndex] = src; } void pop_back() { VMA_HEAVY_ASSERT(m_Count > 0); resize(size() - 1); } void push_front(const T& src) { insert(0, src); } void pop_front() { VMA_HEAVY_ASSERT(m_Count > 0); remove(0); } typedef T* iterator; iterator begin() { return m_pArray; } iterator end() { return m_pArray + m_Count; } private: AllocatorT m_Allocator; T* m_pArray; size_t m_Count; size_t m_Capacity; }; template static void VmaVectorInsert(VmaVector& vec, size_t index, const T& item) { vec.insert(index, item); } template static void VmaVectorRemove(VmaVector& vec, size_t index) { vec.remove(index); } #endif // #if VMA_USE_STL_VECTOR template size_t VmaVectorInsertSorted(VectorT& vector, const typename VectorT::value_type& value) { const size_t indexToInsert = VmaBinaryFindFirstNotLess( vector.data(), vector.data() + vector.size(), value, CmpLess()) - vector.data(); VmaVectorInsert(vector, indexToInsert, value); return indexToInsert; } template bool VmaVectorRemoveSorted(VectorT& vector, const typename VectorT::value_type& value) { CmpLess comparator; typename VectorT::iterator it = VmaBinaryFindFirstNotLess( vector.begin(), vector.end(), value, comparator); if((it != vector.end()) && !comparator(*it, value) && !comparator(value, *it)) { size_t indexToRemove = it - vector.begin(); VmaVectorRemove(vector, indexToRemove); return true; } return false; } template size_t VmaVectorFindSorted(const VectorT& vector, const typename VectorT::value_type& value) { CmpLess comparator; typename VectorT::iterator it = VmaBinaryFindFirstNotLess( vector.data(), vector.data() + vector.size(), value, comparator); if(it != vector.size() && !comparator(*it, value) && !comparator(value, *it)) { return it - vector.begin(); } else { return vector.size(); } } //////////////////////////////////////////////////////////////////////////////// // class VmaPoolAllocator /* Allocator for objects of type T using a list of arrays (pools) to speed up allocation. Number of elements that can be allocated is not bounded because allocator can create multiple blocks. */ template class VmaPoolAllocator { public: VmaPoolAllocator(const VkAllocationCallbacks* pAllocationCallbacks, size_t itemsPerBlock); ~VmaPoolAllocator(); void Clear(); T* Alloc(); void Free(T* ptr); private: union Item { uint32_t NextFreeIndex; T Value; }; struct ItemBlock { Item* pItems; uint32_t FirstFreeIndex; }; const VkAllocationCallbacks* m_pAllocationCallbacks; size_t m_ItemsPerBlock; VmaVector< ItemBlock, VmaStlAllocator > m_ItemBlocks; ItemBlock& CreateNewBlock(); }; template VmaPoolAllocator::VmaPoolAllocator(const VkAllocationCallbacks* pAllocationCallbacks, size_t itemsPerBlock) : m_pAllocationCallbacks(pAllocationCallbacks), m_ItemsPerBlock(itemsPerBlock), m_ItemBlocks(VmaStlAllocator(pAllocationCallbacks)) { VMA_ASSERT(itemsPerBlock > 0); } template VmaPoolAllocator::~VmaPoolAllocator() { Clear(); } template void VmaPoolAllocator::Clear() { for(size_t i = m_ItemBlocks.size(); i--; ) vma_delete_array(m_pAllocationCallbacks, m_ItemBlocks[i].pItems, m_ItemsPerBlock); m_ItemBlocks.clear(); } template T* VmaPoolAllocator::Alloc() { for(size_t i = m_ItemBlocks.size(); i--; ) { ItemBlock& block = m_ItemBlocks[i]; // This block has some free items: Use first one. if(block.FirstFreeIndex != UINT32_MAX) { Item* const pItem = &block.pItems[block.FirstFreeIndex]; block.FirstFreeIndex = pItem->NextFreeIndex; return &pItem->Value; } } // No block has free item: Create new one and use it. ItemBlock& newBlock = CreateNewBlock(); Item* const pItem = &newBlock.pItems[0]; newBlock.FirstFreeIndex = pItem->NextFreeIndex; return &pItem->Value; } template void VmaPoolAllocator::Free(T* ptr) { // Search all memory blocks to f