diff options
| author | Rémi Verschelde <remi@verschelde.fr> | 2022-02-14 09:07:11 +0100 |
|---|---|---|
| committer | GitHub <noreply@github.com> | 2022-02-14 09:07:11 +0100 |
| commit | 09a61cdf53a0b34dd6bfa19da7f097c869a8e145 (patch) | |
| tree | a9159424740989ba146b89bde09843561cfb0013 /thirdparty | |
| parent | 618324cd84d7d03aff4512298d9f012f3f956a8d (diff) | |
| parent | 4e6c9d3ae979f2eb0151cf581fe61d2f3194ea72 (diff) | |
Merge pull request #57989 from RandomShaper/update_vma
Update & patch VMA, and re-implement the small buffers optimization
Diffstat (limited to 'thirdparty')
| -rw-r--r-- | thirdparty/README.md | 4 | ||||
| -rw-r--r-- | thirdparty/vulkan/patches/01-VMA-fix-nullability.patch | 80 | ||||
| -rw-r--r-- | thirdparty/vulkan/patches/02-VMA-use-volk.patch (renamed from thirdparty/vulkan/patches/VMA-use-volk.patch) | 6 | ||||
| -rw-r--r-- | thirdparty/vulkan/patches/03-VMA-universal-pools.patch | 567 | ||||
| -rw-r--r-- | thirdparty/vulkan/vk_mem_alloc.h | 20400 |
5 files changed, 10936 insertions, 10121 deletions
diff --git a/thirdparty/README.md b/thirdparty/README.md index 012087526c..7d2586b4dc 100644 --- a/thirdparty/README.md +++ b/thirdparty/README.md @@ -665,10 +665,10 @@ Files extracted from upstream source: SDK release: https://github.com/KhronosGroup/Vulkan-ValidationLayers/blob/master/layers/generated/vk_enum_string_helper.h `vk_mem_alloc.h` is taken from https://github.com/GPUOpen-LibrariesAndSDKs/VulkanMemoryAllocator -Version: 3.0.0-development (2021-07-07), branch `feature-small-buffers`, commit `cfea2f72851f9ee4a399769f18865047b83711f1` +Version: 3.0.0-development (2022-02-08), commit `a1895bc76547370564d604faa27e0b73de747df1` `vk_mem_alloc.cpp` is a Godot file and should be preserved on updates. -Patches in the `patches` directory should be re-applied after updates. +Patches in the `patches` directory should be re-applied after updates (order must be followed among the number-prefixed ones). ## wslay diff --git a/thirdparty/vulkan/patches/01-VMA-fix-nullability.patch b/thirdparty/vulkan/patches/01-VMA-fix-nullability.patch new file mode 100644 index 0000000000..7deada97b0 --- /dev/null +++ b/thirdparty/vulkan/patches/01-VMA-fix-nullability.patch @@ -0,0 +1,80 @@ +diff --git a/thirdparty/vulkan/vk_mem_alloc.h b/thirdparty/vulkan/vk_mem_alloc.h +index 52b403bede..d88c305a7c 100644 +--- a/thirdparty/vulkan/vk_mem_alloc.h ++++ b/thirdparty/vulkan/vk_mem_alloc.h +@@ -2366,7 +2366,7 @@ VMA_CALL_PRE VkBool32 VMA_CALL_POST vmaIsVirtualBlockEmpty( + */ + VMA_CALL_PRE void VMA_CALL_POST vmaGetVirtualAllocationInfo( + VmaVirtualBlock VMA_NOT_NULL virtualBlock, +- VmaVirtualAllocation allocation, VmaVirtualAllocationInfo* VMA_NOT_NULL pVirtualAllocInfo); ++ VmaVirtualAllocation VMA_NOT_NULL_NON_DISPATCHABLE allocation, VmaVirtualAllocationInfo* VMA_NOT_NULL pVirtualAllocInfo); + + /** \brief Allocates new virtual allocation inside given #VmaVirtualBlock. + +diff --git a/thirdparty/vulkan/vk_mem_alloc.h b/thirdparty/vulkan/vk_mem_alloc.h +index d1138a7bc8..74c66b9789 100644 +--- a/thirdparty/vulkan/vk_mem_alloc.h ++++ b/thirdparty/vulkan/vk_mem_alloc.h +@@ -2386,7 +2386,7 @@ If the allocation fails due to not enough free space available, `VK_ERROR_OUT_OF + VMA_CALL_PRE VkResult VMA_CALL_POST vmaVirtualAllocate( + VmaVirtualBlock VMA_NOT_NULL virtualBlock, + const VmaVirtualAllocationCreateInfo* VMA_NOT_NULL pCreateInfo, +- VmaVirtualAllocation* VMA_NOT_NULL pAllocation, ++ VmaVirtualAllocation VMA_NULLABLE_NON_DISPATCHABLE* VMA_NOT_NULL pAllocation, + VkDeviceSize* VMA_NULLABLE pOffset); + + /** \brief Frees virtual allocation inside given #VmaVirtualBlock. +@@ -2391,7 +2391,7 @@ It is correct to call this function with `allocation == VK_NULL_HANDLE` - it doe + */ + VMA_CALL_PRE void VMA_CALL_POST vmaVirtualFree( + VmaVirtualBlock VMA_NOT_NULL virtualBlock, +- VmaVirtualAllocation allocation); ++ VmaVirtualAllocation VMA_NULLABLE_NON_DISPATCHABLE allocation); + + /** \brief Frees all virtual allocations inside given #VmaVirtualBlock. + +@@ -2408,7 +2408,7 @@ VMA_CALL_PRE void VMA_CALL_POST vmaClearVirtualBlock( + */ + VMA_CALL_PRE void VMA_CALL_POST vmaSetVirtualAllocationUserData( + VmaVirtualBlock VMA_NOT_NULL virtualBlock, +- VmaVirtualAllocation allocation, ++ VmaVirtualAllocation VMA_NOT_NULL_NON_DISPATCHABLE allocation, + void* VMA_NULLABLE pUserData); + + /** \brief Calculates and returns statistics about virtual allocations and memory usage in given #VmaVirtualBlock. +@@ -17835,7 +17835,7 @@ VMA_CALL_PRE VkBool32 VMA_CALL_POST vmaIsVirtualBlockEmpty(VmaVirtualBlock VMA_N + } + + VMA_CALL_PRE void VMA_CALL_POST vmaGetVirtualAllocationInfo(VmaVirtualBlock VMA_NOT_NULL virtualBlock, +- VmaVirtualAllocation allocation, VmaVirtualAllocationInfo* VMA_NOT_NULL pVirtualAllocInfo) ++ VmaVirtualAllocation VMA_NOT_NULL_NON_DISPATCHABLE allocation, VmaVirtualAllocationInfo* VMA_NOT_NULL pVirtualAllocInfo) + { + VMA_ASSERT(virtualBlock != VK_NULL_HANDLE && pVirtualAllocInfo != VMA_NULL); + VMA_DEBUG_LOG("vmaGetVirtualAllocationInfo"); +@@ -17853,7 +17853,7 @@ VMA_CALL_PRE VkResult VMA_CALL_POST vmaVirtualAllocate(VmaVirtualBlock VMA_NOT_N + return virtualBlock->Allocate(*pCreateInfo, *pAllocation, pOffset); + } + +-VMA_CALL_PRE void VMA_CALL_POST vmaVirtualFree(VmaVirtualBlock VMA_NOT_NULL virtualBlock, VmaVirtualAllocation allocation) ++VMA_CALL_PRE void VMA_CALL_POST vmaVirtualFree(VmaVirtualBlock VMA_NOT_NULL virtualBlock, VmaVirtualAllocation VMA_NULLABLE_NON_DISPATCHABLE allocation) + { + if(allocation != VK_NULL_HANDLE) + { +@@ -17873,7 +17873,7 @@ VMA_CALL_PRE void VMA_CALL_POST vmaClearVirtualBlock(VmaVirtualBlock VMA_NOT_NUL + } + + VMA_CALL_PRE void VMA_CALL_POST vmaSetVirtualAllocationUserData(VmaVirtualBlock VMA_NOT_NULL virtualBlock, +- VmaVirtualAllocation allocation, void* VMA_NULLABLE pUserData) ++ VmaVirtualAllocation VMA_NOT_NULL_NON_DISPATCHABLE allocation, void* VMA_NULLABLE pUserData) + { + VMA_ASSERT(virtualBlock != VK_NULL_HANDLE); + VMA_DEBUG_LOG("vmaSetVirtualAllocationUserData"); +@@ -17848,7 +17848,7 @@ VMA_CALL_PRE void VMA_CALL_POST vmaGetVirtualAllocationInfo(VmaVirtualBlock VMA_ + } + + VMA_CALL_PRE VkResult VMA_CALL_POST vmaVirtualAllocate(VmaVirtualBlock VMA_NOT_NULL virtualBlock, +- const VmaVirtualAllocationCreateInfo* VMA_NOT_NULL pCreateInfo, VmaVirtualAllocation* VMA_NOT_NULL pAllocation, ++ const VmaVirtualAllocationCreateInfo* VMA_NOT_NULL pCreateInfo, VmaVirtualAllocation VMA_NULLABLE_NON_DISPATCHABLE* VMA_NOT_NULL pAllocation, + VkDeviceSize* VMA_NULLABLE pOffset) + { + VMA_ASSERT(virtualBlock != VK_NULL_HANDLE && pCreateInfo != VMA_NULL && pAllocation != VMA_NULL); diff --git a/thirdparty/vulkan/patches/VMA-use-volk.patch b/thirdparty/vulkan/patches/02-VMA-use-volk.patch index 81bfcccd89..1b6e0f04b8 100644 --- a/thirdparty/vulkan/patches/VMA-use-volk.patch +++ b/thirdparty/vulkan/patches/02-VMA-use-volk.patch @@ -1,9 +1,9 @@ diff --git a/thirdparty/vulkan/vk_mem_alloc.h b/thirdparty/vulkan/vk_mem_alloc.h -index 65d6243419..9890f20f7c 100644 +index 52b403bede..7c450be211 100644 --- a/thirdparty/vulkan/vk_mem_alloc.h +++ b/thirdparty/vulkan/vk_mem_alloc.h -@@ -2063,7 +2063,11 @@ available through VmaAllocatorCreateInfo::pRecordSettings. - #endif // #if defined(__ANDROID__) && VMA_STATIC_VULKAN_FUNCTIONS && VK_NO_PROTOTYPES +@@ -127,7 +127,11 @@ extern "C" { + #endif #ifndef VULKAN_H_ - #include <vulkan/vulkan.h> diff --git a/thirdparty/vulkan/patches/03-VMA-universal-pools.patch b/thirdparty/vulkan/patches/03-VMA-universal-pools.patch new file mode 100644 index 0000000000..a5de3aaace --- /dev/null +++ b/thirdparty/vulkan/patches/03-VMA-universal-pools.patch @@ -0,0 +1,567 @@ +diff --git a/thirdparty/vulkan/vk_mem_alloc.h b/thirdparty/vulkan/vk_mem_alloc.h +index 74c66b9789..89e00e6326 100644 +--- a/thirdparty/vulkan/vk_mem_alloc.h ++++ b/thirdparty/vulkan/vk_mem_alloc.h +@@ -1127,31 +1127,26 @@ typedef struct VmaAllocationCreateInfo + /** \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.*/ ++ Leave 0 if you specify memory requirements in other way.*/ + VkMemoryPropertyFlags requiredFlags; + /** \brief Flags that preferably should be set in a memory type chosen for an allocation. + +- Set to 0 if no additional flags are preferred. \n +- If `pool` is not null, this member is ignored. */ ++ Set to 0 if no additional flags are preferred.*/ + 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. ++ Leave `VK_NULL_HANDLE` to allocate from default pool. + */ + VmaPool VMA_NULLABLE pool; + /** \brief Custom general-purpose pointer that will be stored in #VmaAllocation, can be read as VmaAllocationInfo::pUserData and changed using vmaSetAllocationUserData(). +@@ -1173,9 +1168,6 @@ typedef struct VmaAllocationCreateInfo + /// 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; +@@ -10904,13 +10896,12 @@ struct VmaPool_T + friend struct VmaPoolListItemTraits; + VMA_CLASS_NO_COPY(VmaPool_T) + public: +- VmaBlockVector m_BlockVector; +- VmaDedicatedAllocationList m_DedicatedAllocations; ++ VmaBlockVector* m_pBlockVectors[VK_MAX_MEMORY_TYPES]; ++ VmaDedicatedAllocationList m_DedicatedAllocations[VK_MAX_MEMORY_TYPES]; + + VmaPool_T( + VmaAllocator hAllocator, +- const VmaPoolCreateInfo& createInfo, +- VkDeviceSize preferredBlockSize); ++ const VmaPoolCreateInfo& createInfo); + ~VmaPool_T(); + + uint32_t GetId() const { return m_Id; } +@@ -10924,6 +10915,7 @@ public: + #endif + + private: ++ const VmaAllocator m_hAllocator; + uint32_t m_Id; + char* m_Name; + VmaPool_T* m_PrevPool = VMA_NULL; +@@ -11405,8 +11397,10 @@ private: + + void ValidateVulkanFunctions(); + ++public: // I'm sorry + VkDeviceSize CalcPreferredBlockSize(uint32_t memTypeIndex); + ++private: + VkResult AllocateMemoryOfType( + VmaPool pool, + VkDeviceSize size, +@@ -14176,30 +14170,36 @@ void VmaDefragmentationContext_T::AddPools(uint32_t poolCount, const VmaPool* pP + { + VmaPool pool = pPools[poolIndex]; + VMA_ASSERT(pool); +- // Pools with algorithm other than default are not defragmented. +- if (pool->m_BlockVector.GetAlgorithm() == 0) ++ for(uint32_t memTypeIndex = 0; memTypeIndex < m_hAllocator->GetMemoryTypeCount(); ++memTypeIndex) + { +- VmaBlockVectorDefragmentationContext* pBlockVectorDefragCtx = VMA_NULL; +- +- for (size_t i = m_CustomPoolContexts.size(); i--; ) ++ if(pool->m_pBlockVectors[memTypeIndex]) + { +- if (m_CustomPoolContexts[i]->GetCustomPool() == pool) ++ // Pools with algorithm other than default are not defragmented. ++ if (pool->m_pBlockVectors[memTypeIndex]->GetAlgorithm() == 0) + { +- pBlockVectorDefragCtx = m_CustomPoolContexts[i]; +- break; +- } +- } ++ VmaBlockVectorDefragmentationContext* pBlockVectorDefragCtx = VMA_NULL; + +- if (!pBlockVectorDefragCtx) +- { +- pBlockVectorDefragCtx = vma_new(m_hAllocator, VmaBlockVectorDefragmentationContext)( +- m_hAllocator, +- pool, +- &pool->m_BlockVector); +- m_CustomPoolContexts.push_back(pBlockVectorDefragCtx); +- } ++ for (size_t i = m_CustomPoolContexts.size(); i--; ) ++ { ++ if (m_CustomPoolContexts[i]->GetCustomPool() == pool) ++ { ++ pBlockVectorDefragCtx = m_CustomPoolContexts[i]; ++ break; ++ } ++ } ++ ++ if (!pBlockVectorDefragCtx) ++ { ++ pBlockVectorDefragCtx = vma_new(m_hAllocator, VmaBlockVectorDefragmentationContext)( ++ m_hAllocator, ++ pool, ++ pool->m_pBlockVectors[memTypeIndex]); ++ m_CustomPoolContexts.push_back(pBlockVectorDefragCtx); ++ } + +- pBlockVectorDefragCtx->AddAll(); ++ pBlockVectorDefragCtx->AddAll(); ++ } ++ } + } + } + } +@@ -14214,6 +14214,7 @@ void VmaDefragmentationContext_T::AddAllocations( + { + const VmaAllocation hAlloc = pAllocations[allocIndex]; + VMA_ASSERT(hAlloc); ++ const uint32_t memTypeIndex = hAlloc->GetMemoryTypeIndex(); + // DedicatedAlloc cannot be defragmented. + if (hAlloc->GetType() == VmaAllocation_T::ALLOCATION_TYPE_BLOCK) + { +@@ -14224,7 +14225,7 @@ void VmaDefragmentationContext_T::AddAllocations( + if (hAllocPool != VK_NULL_HANDLE) + { + // Pools with algorithm other than default are not defragmented. +- if (hAllocPool->m_BlockVector.GetAlgorithm() == 0) ++ if (hAllocPool->m_pBlockVectors[memTypeIndex]->GetAlgorithm() == 0) + { + for (size_t i = m_CustomPoolContexts.size(); i--; ) + { +@@ -14239,7 +14240,7 @@ void VmaDefragmentationContext_T::AddAllocations( + pBlockVectorDefragCtx = vma_new(m_hAllocator, VmaBlockVectorDefragmentationContext)( + m_hAllocator, + hAllocPool, +- &hAllocPool->m_BlockVector); ++ hAllocPool->m_pBlockVectors[memTypeIndex]); + m_CustomPoolContexts.push_back(pBlockVectorDefragCtx); + } + } +@@ -14247,7 +14248,6 @@ void VmaDefragmentationContext_T::AddAllocations( + // This allocation belongs to default pool. + else + { +- const uint32_t memTypeIndex = hAlloc->GetMemoryTypeIndex(); + pBlockVectorDefragCtx = m_DefaultPoolContexts[memTypeIndex]; + if (!pBlockVectorDefragCtx) + { +@@ -14481,41 +14481,61 @@ VkResult VmaDefragmentationContext_T::DefragmentPassEnd() + #ifndef _VMA_POOL_T_FUNCTIONS + VmaPool_T::VmaPool_T( + VmaAllocator hAllocator, +- const VmaPoolCreateInfo& createInfo, +- VkDeviceSize preferredBlockSize) +- : m_BlockVector( +- hAllocator, +- this, // hParentPool +- createInfo.memoryTypeIndex, +- createInfo.blockSize != 0 ? createInfo.blockSize : preferredBlockSize, +- createInfo.minBlockCount, +- createInfo.maxBlockCount, +- (createInfo.flags& VMA_POOL_CREATE_IGNORE_BUFFER_IMAGE_GRANULARITY_BIT) != 0 ? 1 : hAllocator->GetBufferImageGranularity(), +- createInfo.blockSize != 0, // explicitBlockSize +- createInfo.flags & VMA_POOL_CREATE_ALGORITHM_MASK, // algorithm +- createInfo.priority, +- VMA_MAX(hAllocator->GetMemoryTypeMinAlignment(createInfo.memoryTypeIndex), createInfo.minAllocationAlignment), +- createInfo.pMemoryAllocateNext), ++ const VmaPoolCreateInfo& createInfo) : ++ m_hAllocator(hAllocator), ++ m_pBlockVectors{}, + m_Id(0), +- m_Name(VMA_NULL) {} ++ m_Name(VMA_NULL) ++{ ++ for(uint32_t memTypeIndex = 0; memTypeIndex < hAllocator->GetMemoryTypeCount(); ++memTypeIndex) ++ { ++ // Create only supported types ++ if((hAllocator->GetGlobalMemoryTypeBits() & (1u << memTypeIndex)) != 0) ++ { ++ m_pBlockVectors[memTypeIndex] = vma_new(hAllocator, VmaBlockVector)( ++ hAllocator, ++ this, // hParentPool ++ memTypeIndex, ++ createInfo.blockSize != 0 ? createInfo.blockSize : hAllocator->CalcPreferredBlockSize(memTypeIndex), ++ createInfo.minBlockCount, ++ createInfo.maxBlockCount, ++ (createInfo.flags& VMA_POOL_CREATE_IGNORE_BUFFER_IMAGE_GRANULARITY_BIT) != 0 ? 1 : hAllocator->GetBufferImageGranularity(), ++ false, // explicitBlockSize ++ createInfo.flags & VMA_POOL_CREATE_ALGORITHM_MASK, // algorithm ++ createInfo.priority, ++ VMA_MAX(hAllocator->GetMemoryTypeMinAlignment(memTypeIndex), createInfo.minAllocationAlignment), ++ createInfo.pMemoryAllocateNext); ++ } ++ } ++} + + VmaPool_T::~VmaPool_T() + { + VMA_ASSERT(m_PrevPool == VMA_NULL && m_NextPool == VMA_NULL); ++ for(uint32_t memTypeIndex = 0; memTypeIndex < m_hAllocator->GetMemoryTypeCount(); ++memTypeIndex) ++ { ++ vma_delete(m_hAllocator, m_pBlockVectors[memTypeIndex]); ++ } + } + + void VmaPool_T::SetName(const char* pName) + { +- const VkAllocationCallbacks* allocs = m_BlockVector.GetAllocator()->GetAllocationCallbacks(); +- VmaFreeString(allocs, m_Name); +- +- if (pName != VMA_NULL) +- { +- m_Name = VmaCreateStringCopy(allocs, pName); +- } +- else ++ for(uint32_t memTypeIndex = 0; memTypeIndex < m_hAllocator->GetMemoryTypeCount(); ++memTypeIndex) + { +- m_Name = VMA_NULL; ++ if(m_pBlockVectors[memTypeIndex]) ++ { ++ const VkAllocationCallbacks* allocs = m_pBlockVectors[memTypeIndex]->GetAllocator()->GetAllocationCallbacks(); ++ VmaFreeString(allocs, m_Name); ++ ++ if (pName != VMA_NULL) ++ { ++ m_Name = VmaCreateStringCopy(allocs, pName); ++ } ++ else ++ { ++ m_Name = VMA_NULL; ++ } ++ } + } + } + #endif // _VMA_POOL_T_FUNCTIONS +@@ -15377,15 +15397,22 @@ VkResult VmaAllocator_T::CalcAllocationParams( + inoutCreateInfo.flags |= VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT; + } + +- if(inoutCreateInfo.pool != VK_NULL_HANDLE) ++ if(inoutCreateInfo.pool != VK_NULL_HANDLE && (inoutCreateInfo.flags & VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT) != 0) + { +- if(inoutCreateInfo.pool->m_BlockVector.HasExplicitBlockSize() && +- (inoutCreateInfo.flags & VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT) != 0) ++ // Assuming here every block has the same block size and priority. ++ for(uint32_t memTypeIndex = 0; memTypeIndex < GetMemoryTypeCount(); ++memTypeIndex) + { +- VMA_ASSERT(0 && "Specifying VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT while current custom pool doesn't support dedicated allocations."); +- return VK_ERROR_FEATURE_NOT_PRESENT; ++ if(inoutCreateInfo.pool->m_pBlockVectors[memTypeIndex]) ++ { ++ if(inoutCreateInfo.pool->m_pBlockVectors[memTypeIndex]->HasExplicitBlockSize()) ++ { ++ VMA_ASSERT(0 && "Specifying VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT while current custom pool doesn't support dedicated allocations."); ++ return VK_ERROR_FEATURE_NOT_PRESENT; ++ } ++ inoutCreateInfo.priority = inoutCreateInfo.pool->m_pBlockVectors[memTypeIndex]->GetPriority(); ++ break; ++ } + } +- inoutCreateInfo.priority = inoutCreateInfo.pool->m_BlockVector.GetPriority(); + } + + if((inoutCreateInfo.flags & VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT) != 0 && +@@ -15429,67 +15456,46 @@ VkResult VmaAllocator_T::AllocateMemory( + if(res != VK_SUCCESS) + return res; + +- if(createInfoFinal.pool != VK_NULL_HANDLE) ++ // Bit mask of memory Vulkan types acceptable for this allocation. ++ uint32_t memoryTypeBits = vkMemReq.memoryTypeBits; ++ uint32_t memTypeIndex = UINT32_MAX; ++ res = vmaFindMemoryTypeIndex(this, memoryTypeBits, &createInfoFinal, &memTypeIndex); ++ // Can't find any single memory type matching requirements. res is VK_ERROR_FEATURE_NOT_PRESENT. ++ if(res != VK_SUCCESS) ++ return res; ++ do + { +- VmaBlockVector& blockVector = createInfoFinal.pool->m_BlockVector; +- return AllocateMemoryOfType( ++ VmaBlockVector* blockVector = createInfoFinal.pool == VK_NULL_HANDLE ? m_pBlockVectors[memTypeIndex] : createInfoFinal.pool->m_pBlockVectors[memTypeIndex]; ++ VMA_ASSERT(blockVector && "Trying to use unsupported memory type!"); ++ VmaDedicatedAllocationList& dedicatedAllocations = createInfoFinal.pool == VK_NULL_HANDLE ? m_DedicatedAllocations[memTypeIndex] : createInfoFinal.pool->m_DedicatedAllocations[memTypeIndex]; ++ res = AllocateMemoryOfType( + createInfoFinal.pool, + vkMemReq.size, + vkMemReq.alignment, +- prefersDedicatedAllocation, ++ requiresDedicatedAllocation || prefersDedicatedAllocation, + dedicatedBuffer, + dedicatedBufferUsage, + dedicatedImage, + createInfoFinal, +- blockVector.GetMemoryTypeIndex(), ++ memTypeIndex, + suballocType, +- createInfoFinal.pool->m_DedicatedAllocations, +- blockVector, ++ dedicatedAllocations, ++ *blockVector, + allocationCount, + pAllocations); +- } +- else +- { +- // Bit mask of memory Vulkan types acceptable for this allocation. +- uint32_t memoryTypeBits = vkMemReq.memoryTypeBits; +- uint32_t memTypeIndex = UINT32_MAX; +- res = vmaFindMemoryTypeIndex(this, memoryTypeBits, &createInfoFinal, &memTypeIndex); +- // Can't find any single memory type matching requirements. res is VK_ERROR_FEATURE_NOT_PRESENT. +- if(res != VK_SUCCESS) +- return res; +- do +- { +- VmaBlockVector* blockVector = m_pBlockVectors[memTypeIndex]; +- VMA_ASSERT(blockVector && "Trying to use unsupported memory type!"); +- res = AllocateMemoryOfType( +- VK_NULL_HANDLE, +- vkMemReq.size, +- vkMemReq.alignment, +- requiresDedicatedAllocation || prefersDedicatedAllocation, +- dedicatedBuffer, +- dedicatedBufferUsage, +- dedicatedImage, +- createInfoFinal, +- memTypeIndex, +- suballocType, +- m_DedicatedAllocations[memTypeIndex], +- *blockVector, +- allocationCount, +- pAllocations); +- // Allocation succeeded +- if(res == VK_SUCCESS) +- return VK_SUCCESS; ++ // Allocation succeeded ++ if(res == VK_SUCCESS) ++ return VK_SUCCESS; + +- // Remove old memTypeIndex from list of possibilities. +- memoryTypeBits &= ~(1u << memTypeIndex); +- // Find alternative memTypeIndex. +- res = vmaFindMemoryTypeIndex(this, memoryTypeBits, &createInfoFinal, &memTypeIndex); +- } while(res == VK_SUCCESS); ++ // Remove old memTypeIndex from list of possibilities. ++ memoryTypeBits &= ~(1u << memTypeIndex); ++ // Find alternative memTypeIndex. ++ res = vmaFindMemoryTypeIndex(this, memoryTypeBits, &createInfoFinal, &memTypeIndex); ++ } while(res == VK_SUCCESS); + +- // No other matching memory type index could be found. +- // Not returning res, which is VK_ERROR_FEATURE_NOT_PRESENT, because we already failed to allocate once. +- return VK_ERROR_OUT_OF_DEVICE_MEMORY; +- } ++ // No other matching memory type index could be found. ++ // Not returning res, which is VK_ERROR_FEATURE_NOT_PRESENT, because we already failed to allocate once. ++ return VK_ERROR_OUT_OF_DEVICE_MEMORY; + } + + void VmaAllocator_T::FreeMemory( +@@ -15515,16 +15521,16 @@ void VmaAllocator_T::FreeMemory( + { + VmaBlockVector* pBlockVector = VMA_NULL; + VmaPool hPool = allocation->GetParentPool(); ++ const uint32_t memTypeIndex = allocation->GetMemoryTypeIndex(); + if(hPool != VK_NULL_HANDLE) + { +- pBlockVector = &hPool->m_BlockVector; ++ pBlockVector = hPool->m_pBlockVectors[memTypeIndex]; + } + else + { +- const uint32_t memTypeIndex = allocation->GetMemoryTypeIndex(); + pBlockVector = m_pBlockVectors[memTypeIndex]; +- VMA_ASSERT(pBlockVector && "Trying to free memory of unsupported type!"); + } ++ VMA_ASSERT(pBlockVector && "Trying to free memory of unsupported type!"); + pBlockVector->Free(allocation); + } + break; +@@ -15564,11 +15570,17 @@ void VmaAllocator_T::CalculateStats(VmaStats* pStats) + VmaMutexLockRead lock(m_PoolsMutex, m_UseMutex); + for(VmaPool pool = m_Pools.Front(); pool != VMA_NULL; pool = m_Pools.GetNext(pool)) + { +- VmaBlockVector& blockVector = pool->m_BlockVector; +- blockVector.AddStats(pStats); +- const uint32_t memTypeIndex = blockVector.GetMemoryTypeIndex(); +- const uint32_t memHeapIndex = MemoryTypeIndexToHeapIndex(memTypeIndex); +- pool->m_DedicatedAllocations.AddStats(pStats, memTypeIndex, memHeapIndex); ++ for(uint32_t memTypeIndex = 0; memTypeIndex < GetMemoryTypeCount(); ++memTypeIndex) ++ { ++ if (pool->m_pBlockVectors[memTypeIndex]) ++ { ++ VmaBlockVector& blockVector = *pool->m_pBlockVectors[memTypeIndex]; ++ blockVector.AddStats(pStats); ++ const uint32_t memTypeIndex = blockVector.GetMemoryTypeIndex(); ++ const uint32_t memHeapIndex = MemoryTypeIndexToHeapIndex(memTypeIndex); ++ pool->m_DedicatedAllocations[memTypeIndex].AddStats(pStats, memTypeIndex, memHeapIndex); ++ } ++ } + } + } + +@@ -15720,27 +15732,26 @@ VkResult VmaAllocator_T::CreatePool(const VmaPoolCreateInfo* pCreateInfo, VmaPoo + { + return VK_ERROR_INITIALIZATION_FAILED; + } +- // Memory type index out of range or forbidden. +- if(pCreateInfo->memoryTypeIndex >= GetMemoryTypeCount() || +- ((1u << pCreateInfo->memoryTypeIndex) & m_GlobalMemoryTypeBits) == 0) +- { +- return VK_ERROR_FEATURE_NOT_PRESENT; +- } + if(newCreateInfo.minAllocationAlignment > 0) + { + VMA_ASSERT(VmaIsPow2(newCreateInfo.minAllocationAlignment)); + } + +- const VkDeviceSize preferredBlockSize = CalcPreferredBlockSize(newCreateInfo.memoryTypeIndex); +- +- *pPool = vma_new(this, VmaPool_T)(this, newCreateInfo, preferredBlockSize); ++ *pPool = vma_new(this, VmaPool_T)(this, newCreateInfo); + +- VkResult res = (*pPool)->m_BlockVector.CreateMinBlocks(); +- if(res != VK_SUCCESS) ++ for(uint32_t memTypeIndex = 0; memTypeIndex < GetMemoryTypeCount(); ++memTypeIndex) + { +- vma_delete(this, *pPool); +- *pPool = VMA_NULL; +- return res; ++ // Create only supported types ++ if((m_GlobalMemoryTypeBits & (1u << memTypeIndex)) != 0) ++ { ++ VkResult res = (*pPool)->m_pBlockVectors[memTypeIndex]->CreateMinBlocks(); ++ if(res != VK_SUCCESS) ++ { ++ vma_delete(this, *pPool); ++ *pPool = VMA_NULL; ++ return res; ++ } ++ } + } + + // Add to m_Pools. +@@ -15772,8 +15783,14 @@ void VmaAllocator_T::GetPoolStats(VmaPool pool, VmaPoolStats* pPoolStats) + pPoolStats->unusedRangeCount = 0; + pPoolStats->blockCount = 0; + +- pool->m_BlockVector.AddPoolStats(pPoolStats); +- pool->m_DedicatedAllocations.AddPoolStats(pPoolStats); ++ for(uint32_t memTypeIndex = 0; memTypeIndex < GetMemoryTypeCount(); ++memTypeIndex) ++ { ++ if((m_GlobalMemoryTypeBits & (1u << memTypeIndex)) != 0) ++ { ++ pool->m_pBlockVectors[memTypeIndex]->AddPoolStats(pPoolStats); ++ pool->m_DedicatedAllocations[memTypeIndex].AddPoolStats(pPoolStats); ++ } ++ } + } + + void VmaAllocator_T::SetCurrentFrameIndex(uint32_t frameIndex) +@@ -15790,7 +15807,13 @@ void VmaAllocator_T::SetCurrentFrameIndex(uint32_t frameIndex) + + VkResult VmaAllocator_T::CheckPoolCorruption(VmaPool hPool) + { +- return hPool->m_BlockVector.CheckCorruption(); ++ for(uint32_t memTypeIndex = 0; memTypeIndex < GetMemoryTypeCount(); ++memTypeIndex) ++ { ++ if((m_GlobalMemoryTypeBits & (1u << memTypeIndex)) != 0) ++ { ++ return hPool->m_pBlockVectors[memTypeIndex]->CheckCorruption(); ++ } ++ } + } + + VkResult VmaAllocator_T::CheckCorruption(uint32_t memoryTypeBits) +@@ -15822,18 +15845,21 @@ VkResult VmaAllocator_T::CheckCorruption(uint32_t memoryTypeBits) + VmaMutexLockRead lock(m_PoolsMutex, m_UseMutex); + for(VmaPool pool = m_Pools.Front(); pool != VMA_NULL; pool = m_Pools.GetNext(pool)) + { +- if(((1u << pool->m_BlockVector.GetMemoryTypeIndex()) & memoryTypeBits) != 0) ++ for(uint32_t memTypeIndex = 0; memTypeIndex < GetMemoryTypeCount(); ++memTypeIndex) + { +- VkResult localRes = pool->m_BlockVector.CheckCorruption(); +- switch(localRes) ++ if(pool->m_pBlockVectors[memTypeIndex] && ((1u << memTypeIndex) & memoryTypeBits) != 0) + { +- case VK_ERROR_FEATURE_NOT_PRESENT: +- break; +- case VK_SUCCESS: +- finalRes = VK_SUCCESS; +- break; +- default: +- return localRes; ++ VkResult localRes = pool->m_pBlockVectors[memTypeIndex]->CheckCorruption(); ++ switch(localRes) ++ { ++ case VK_ERROR_FEATURE_NOT_PRESENT: ++ break; ++ case VK_SUCCESS: ++ finalRes = VK_SUCCESS; ++ break; ++ default: ++ return localRes; ++ } + } + } + } +@@ -16155,7 +16181,7 @@ void VmaAllocator_T::FreeDedicatedMemory(const VmaAllocation allocation) + else + { + // Custom pool +- parentPool->m_DedicatedAllocations.Unregister(allocation); ++ parentPool->m_DedicatedAllocations[memTypeIndex].Unregister(allocation); + } + + VkDeviceMemory hMemory = allocation->GetMemory(); +@@ -16430,12 +16456,18 @@ void VmaAllocator_T::PrintDetailedMap(VmaJsonWriter& json) + json.EndString(); + + json.BeginObject(); +- pool->m_BlockVector.PrintDetailedMap(json); +- +- if (!pool->m_DedicatedAllocations.IsEmpty()) ++ for(uint32_t memTypeIndex = 0; memTypeIndex < GetMemoryTypeCount(); ++memTypeIndex) + { +- json.WriteString("DedicatedAllocations"); +- pool->m_DedicatedAllocations.BuildStatsString(json); ++ if (pool->m_pBlockVectors[memTypeIndex]) ++ { ++ pool->m_pBlockVectors[memTypeIndex]->PrintDetailedMap(json); ++ } ++ ++ if (!pool->m_DedicatedAllocations[memTypeIndex].IsEmpty()) ++ { ++ json.WriteString("DedicatedAllocations"); ++ pool->m_DedicatedAllocations->BuildStatsString(json); ++ } + } + json.EndObject(); + } diff --git a/thirdparty/vulkan/vk_mem_alloc.h b/thirdparty/vulkan/vk_mem_alloc.h index 9890f20f7c..89e00e6326 100644 --- a/thirdparty/vulkan/vk_mem_alloc.h +++ b/thirdparty/vulkan/vk_mem_alloc.h @@ -1,5 +1,5 @@ // -// Copyright (c) 2017-2021 Advanced Micro Devices, Inc. All rights reserved. +/// Copyright (c) 2017-2022 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 @@ -25,12 +25,12 @@ /** \mainpage Vulkan Memory Allocator -<b>Version 3.0.0-development</b> (2021-06-21) +<b>Version 3.0.0-development</b> -Copyright (c) 2017-2021 Advanced Micro Devices, Inc. All rights reserved. \n +Copyright (c) 2017-2022 Advanced Micro Devices, Inc. All rights reserved. \n License: MIT -Documentation of all members: vk_mem_alloc.h +<b>API documentation divided into groups:</b> [Modules](modules.html) \section main_table_of_contents Table of contents @@ -67,18 +67,18 @@ Documentation of all members: vk_mem_alloc.h - [Defragmenting GPU memory](@ref defragmentation_gpu) - [Additional notes](@ref defragmentation_additional_notes) - [Writing custom allocation algorithm](@ref defragmentation_custom_algorithm) - - \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 virtual_allocator - \subpage debugging_memory_usage - [Memory initialization](@ref debugging_memory_usage_initialization) - [Margins](@ref debugging_memory_usage_margins) - [Corruption detection](@ref debugging_memory_usage_corruption_detection) - - \subpage record_and_replay + - \subpage opengl_interop - \subpage usage_patterns - [Common mistakes](@ref usage_patterns_common_mistakes) - [Simple patterns](@ref usage_patterns_simple) @@ -102,1935 +102,51 @@ Documentation of all members: vk_mem_alloc.h - [Product page on GPUOpen](https://gpuopen.com/gaming-product/vulkan-memory-allocator/) - [Source repository on GitHub](https://github.com/GPUOpen-LibrariesAndSDKs/VulkanMemoryAllocator) +\defgroup group_init Library initialization +\brief API elements related to the initialization and management of the entire library, especially #VmaAllocator object. +\defgroup group_alloc Memory allocation -\page quick_start Quick start - -\section quick_start_project_setup Project setup - -Vulkan Memory Allocator comes in form of a "stb-style" 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 exactly 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. - -Note on language: This library is written in C++, but has C-compatible interface. -Thus you can include and use vk_mem_alloc.h in C or C++ code, but full -implementation with `VMA_IMPLEMENTATION` macro must be compiled as C++, NOT as C. - -Please note that this library includes header `<vulkan/vulkan.h>`, which in turn -includes `<windows.h>` on Windows. If you need some specific macros defined -before including these headers (like `WIN32_LEAN_AND_MEAN` or -`WINVER` for Windows, `VK_USE_PLATFORM_WIN32_KHR` for Vulkan), you must define -them before every `#include` of this library. - -You may need to configure the way you import Vulkan functions. - -- By default, VMA assumes you you link statically with Vulkan API. If this is not the case, - `#define VMA_STATIC_VULKAN_FUNCTIONS 0` before `#include` of the VMA implementation and use another way. -- You can `#define VMA_DYNAMIC_VULKAN_FUNCTIONS 1` and make sure `vkGetInstanceProcAddr` and `vkGetDeviceProcAddr` globals are defined. - All the remaining Vulkan functions will be fetched automatically. -- Finally, you can provide your own pointers to all Vulkan functions needed by VMA using structure member - VmaAllocatorCreateInfo::pVulkanFunctions, if you fetched them in some custom way e.g. using some loader like [Volk](https://github.com/zeux/volk). - - -\section quick_start_initialization Initialization - -At program startup: - --# Initialize Vulkan to have `VkPhysicalDevice`, `VkDevice` and `VkInstance` object. --# Fill VmaAllocatorCreateInfo structure and create #VmaAllocator object by - calling vmaCreateAllocator(). - -\code -VmaAllocatorCreateInfo allocatorInfo = {}; -allocatorInfo.vulkanApiVersion = VK_API_VERSION_1_2; -allocatorInfo.physicalDevice = physicalDevice; -allocatorInfo.device = device; -allocatorInfo.instance = instance; - -VmaAllocator allocator; -vmaCreateAllocator(&allocatorInfo, &allocator); -\endcode - -Only members `physicalDevice`, `device`, `instance` are required. -However, you should inform the library which Vulkan version do you use by setting -VmaAllocatorCreateInfo::vulkanApiVersion and which extensions did you enable -by setting VmaAllocatorCreateInfo::flags (like #VMA_ALLOCATOR_CREATE_BUFFER_DEVICE_ADDRESS_BIT for VK_KHR_buffer_device_address). -Otherwise, VMA would use only features of Vulkan 1.0 core with no extensions. - - -\section quick_start_resource_allocation Resource allocation - -When you want to create a buffer or image: - --# 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(), vmaFindMemoryTypeIndexForBufferInfo(), - vmaFindMemoryTypeIndexForImageInfo(). --# 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. - vmaAllocateMemoryPages() function is also provided for creating multiple allocations at once, - which may be useful for sparse binding. --# 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() - or their extended versions: vmaBindBufferMemory2(), vmaBindImageMemory2(). --# 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 easiest and 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. - -\section choosing_memory_type_dedicated_allocations Dedicated allocations - -Memory for allocations is reserved out of larger block of `VkDeviceMemory` -allocated from Vulkan internally. That's the main feature of this whole library. -You can still request a separate memory block to be created for an allocation, -just like you would do in a trivial solution without using any allocator. -In that case, a buffer or image is always bound to that memory at offset 0. -This is called a "dedicated allocation". -You can explicitly request it by using flag #VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT. -The library can also internally decide to use dedicated allocation in some cases, e.g.: - -- When the size of the allocation is large. -- When [VK_KHR_dedicated_allocation](@ref vk_khr_dedicated_allocation) extension is enabled - and it reports that dedicated allocation is required or recommended for the resource. -- When allocation of next big memory block fails due to not enough device memory, - but allocation with the exact requested size succeeds. - - -\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. -The 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; -VkBuffer 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 - -When mapping, you may see a warning from Vulkan validation layer similar to this one: - -<i>Mapping an image with layout VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL can result in undefined behavior if this memory is used by the device. Only GENERAL or PREINITIALIZED should be used.</i> - -It happens because the library maps entire `VkDeviceMemory` block, where different -types of images and buffers may end up together, especially on GPUs with unified memory like Intel. -You can safely ignore it if you are sure you access only memory of the intended -object that you wanted to map. - - -\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 flush and invalidate - -Memory in Vulkan doesn't need to be unmapped before using it on GPU, -but unless a memory types has `VK_MEMORY_PROPERTY_HOST_COHERENT_BIT` flag set, -you need to manually **invalidate** cache before reading of mapped pointer -and **flush** cache after writing to mapped pointer. -Map/unmap operations don't do that automatically. -Vulkan provides following functions for this purpose `vkFlushMappedMemoryRanges()`, -`vkInvalidateMappedMemoryRanges()`, but this library provides more convenient -functions that refer to given allocation object: vmaFlushAllocation(), -vmaInvalidateAllocation(), -or multiple objects at once: vmaFlushAllocations(), vmaInvalidateAllocations(). - -Regions of memory specified for flush/invalidate must be aligned to -`VkPhysicalDeviceLimits::nonCoherentAtomSize`. This is automatically ensured by the library. -In any memory type that is `HOST_VISIBLE` but not `HOST_COHERENT`, all allocations -within blocks are aligned to this value, so their offsets are always multiply of -`nonCoherentAtomSize` and two different allocations never share same "line" of this size. - -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 `HOST_COHERENT` flag on all memory types that are -`HOST_VISIBLE`, 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; -allocCreateInfo.preferredFlags = VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT; - -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.pMappedData != 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 staying_within_budget Staying within budget - -When developing a graphics-intensive game or program, it is important to avoid allocating -more GPU memory than it's physically available. When the memory is over-committed, -various bad things can happen, depending on the specific GPU, graphics driver, and -operating system: - -- It may just work without any problems. -- The application may slow down because some memory blocks are moved to system RAM - and the GPU has to access them through PCI Express bus. -- A new allocation may take very long time to complete, even few seconds, and possibly - freeze entire system. -- The new allocation may fail with `VK_ERROR_OUT_OF_DEVICE_MEMORY`. -- It may even result in GPU crash (TDR), observed as `VK_ERROR_DEVICE_LOST` - returned somewhere later. - -\section staying_within_budget_querying_for_budget Querying for budget - -To query for current memory usage and available budget, use function vmaGetBudget(). -Returned structure #VmaBudget contains quantities expressed in bytes, per Vulkan memory heap. - -Please note that this function returns different information and works faster than -vmaCalculateStats(). vmaGetBudget() can be called every frame or even before every -allocation, while vmaCalculateStats() is intended to be used rarely, -only to obtain statistical information, e.g. for debugging purposes. +\brief API elements related to the allocation, deallocation, and management of Vulkan memory, buffers, images. +Most basic ones being: vmaCreateBuffer(), vmaCreateImage(). -It is recommended to use <b>VK_EXT_memory_budget</b> device extension to obtain information -about the budget from Vulkan device. VMA is able to use this extension automatically. -When not enabled, the allocator behaves same way, but then it estimates current usage -and available budget based on its internal information and Vulkan memory heap sizes, -which may be less precise. In order to use this extension: +\defgroup group_virtual Virtual allocator -1. Make sure extensions VK_EXT_memory_budget and VK_KHR_get_physical_device_properties2 - required by it are available and enable them. Please note that the first is a device - extension and the second is instance extension! -2. Use flag #VMA_ALLOCATOR_CREATE_EXT_MEMORY_BUDGET_BIT when creating #VmaAllocator object. -3. Make sure to call vmaSetCurrentFrameIndex() every frame. Budget is queried from - Vulkan inside of it to avoid overhead of querying it with every allocation. - -\section staying_within_budget_controlling_memory_usage Controlling memory usage - -There are many ways in which you can try to stay within the budget. - -First, when making new allocation requires allocating a new memory block, the library -tries not to exceed the budget automatically. If a block with default recommended size -(e.g. 256 MB) would go over budget, a smaller block is allocated, possibly even -dedicated memory for just this resource. - -If the size of the requested resource plus current memory usage is more than the -budget, by default the library still tries to create it, leaving it to the Vulkan -implementation whether the allocation succeeds or fails. You can change this behavior -by using #VMA_ALLOCATION_CREATE_WITHIN_BUDGET_BIT flag. With it, the allocation is -not made if it would exceed the budget or if the budget is already exceeded. -Some other allocations become lost instead to make room for it, if the mechanism of -[lost allocations](@ref lost_allocations) is used. -If that is not possible, the allocation fails with `VK_ERROR_OUT_OF_DEVICE_MEMORY`. -Example usage pattern may be to pass the #VMA_ALLOCATION_CREATE_WITHIN_BUDGET_BIT flag -when creating resources that are not essential for the application (e.g. the texture -of a specific object) and not to pass it when creating critically important resources -(e.g. render targets). - -Finally, you can also use #VMA_ALLOCATION_CREATE_NEVER_ALLOCATE_BIT flag to make sure -a new allocation is created only when it fits inside one of the existing memory blocks. -If it would require to allocate a new block, if fails instead with `VK_ERROR_OUT_OF_DEVICE_MEMORY`. -This also ensures that the function call is very fast because it never goes to Vulkan -to obtain a new block. - -Please note that creating \ref custom_memory_pools with VmaPoolCreateInfo::minBlockCount -set to more than 0 will try to allocate memory blocks without checking whether they -fit within budget. - - -\page resource_aliasing Resource aliasing (overlap) - -New explicit graphics APIs (Vulkan and Direct3D 12), thanks to manual memory -management, give an opportunity to alias (overlap) multiple resources in the -same region of memory - a feature not available in the old APIs (Direct3D 11, OpenGL). -It can be useful to save video memory, but it must be used with caution. - -For example, if you know the flow of your whole render frame in advance, you -are going to use some intermediate textures or buffers only during a small range of render passes, -and you know these ranges don't overlap in time, you can bind these resources to -the same place in memory, even if they have completely different parameters (width, height, format etc.). - - - -Such scenario is possible using VMA, but you need to create your images manually. -Then you need to calculate parameters of an allocation to be made using formula: - -- allocation size = max(size of each image) -- allocation alignment = max(alignment of each image) -- allocation memoryTypeBits = bitwise AND(memoryTypeBits of each image) - -Following example shows two different images bound to the same place in memory, -allocated to fit largest of them. - -\code -// A 512x512 texture to be sampled. -VkImageCreateInfo img1CreateInfo = { VK_STRUCTURE_TYPE_IMAGE_CREATE_INFO }; -img1CreateInfo.imageType = VK_IMAGE_TYPE_2D; -img1CreateInfo.extent.width = 512; -img1CreateInfo.extent.height = 512; -img1CreateInfo.extent.depth = 1; -img1CreateInfo.mipLevels = 10; -img1CreateInfo.arrayLayers = 1; -img1CreateInfo.format = VK_FORMAT_R8G8B8A8_SRGB; -img1CreateInfo.tiling = VK_IMAGE_TILING_OPTIMAL; -img1CreateInfo.initialLayout = VK_IMAGE_LAYOUT_UNDEFINED; -img1CreateInfo.usage = VK_IMAGE_USAGE_TRANSFER_DST_BIT | VK_IMAGE_USAGE_SAMPLED_BIT; -img1CreateInfo.samples = VK_SAMPLE_COUNT_1_BIT; - -// A full screen texture to be used as color attachment. -VkImageCreateInfo img2CreateInfo = { VK_STRUCTURE_TYPE_IMAGE_CREATE_INFO }; -img2CreateInfo.imageType = VK_IMAGE_TYPE_2D; -img2CreateInfo.extent.width = 1920; -img2CreateInfo.extent.height = 1080; -img2CreateInfo.extent.depth = 1; -img2CreateInfo.mipLevels = 1; -img2CreateInfo.arrayLayers = 1; -img2CreateInfo.format = VK_FORMAT_R8G8B8A8_UNORM; -img2CreateInfo.tiling = VK_IMAGE_TILING_OPTIMAL; -img2CreateInfo.initialLayout = VK_IMAGE_LAYOUT_UNDEFINED; -img2CreateInfo.usage = VK_IMAGE_USAGE_SAMPLED_BIT | VK_IMAGE_USAGE_COLOR_ATTACHMENT_BIT; -img2CreateInfo.samples = VK_SAMPLE_COUNT_1_BIT; - -VkImage img1; -res = vkCreateImage(device, &img1CreateInfo, nullptr, &img1); -VkImage img2; -res = vkCreateImage(device, &img2CreateInfo, nullptr, &img2); - -VkMemoryRequirements img1MemReq; -vkGetImageMemoryRequirements(device, img1, &img1MemReq); -VkMemoryRequirements img2MemReq; -vkGetImageMemoryRequirements(device, img2, &img2MemReq); - -VkMemoryRequirements finalMemReq = {}; -finalMemReq.size = std::max(img1MemReq.size, img2MemReq.size); -finalMemReq.alignment = std::max(img1MemReq.alignment, img2MemReq.alignment); -finalMemReq.memoryTypeBits = img1MemReq.memoryTypeBits & img2MemReq.memoryTypeBits; -// Validate if(finalMemReq.memoryTypeBits != 0) - -VmaAllocationCreateInfo allocCreateInfo = {}; -allocCreateInfo.usage = VMA_MEMORY_USAGE_GPU_ONLY; - -VmaAllocation alloc; -res = vmaAllocateMemory(allocator, &finalMemReq, &allocCreateInfo, &alloc, nullptr); - -res = vmaBindImageMemory(allocator, alloc, img1); -res = vmaBindImageMemory(allocator, alloc, img2); - -// You can use img1, img2 here, but not at the same time! - -vmaFreeMemory(allocator, alloc); -vkDestroyImage(allocator, img2, nullptr); -vkDestroyImage(allocator, img1, nullptr); -\endcode - -Remember that using resources that alias in memory requires proper synchronization. -You need to issue a memory barrier to make sure commands that use `img1` and `img2` -don't overlap on GPU timeline. -You also need to treat a resource after aliasing as uninitialized - containing garbage data. -For example, if you use `img1` and then want to use `img2`, you need to issue -an image memory barrier for `img2` with `oldLayout` = `VK_IMAGE_LAYOUT_UNDEFINED`. - -Additional considerations: - -- Vulkan also allows to interpret contents of memory between aliasing resources consistently in some cases. -See chapter 11.8. "Memory Aliasing" of Vulkan specification or `VK_IMAGE_CREATE_ALIAS_BIT` flag. -- You can create more complex layout where different images and buffers are bound -at different offsets inside one large allocation. For example, one can imagine -a big texture used in some render passes, aliasing with a set of many small buffers -used between in some further passes. To bind a resource at non-zero offset of an allocation, -use vmaBindBufferMemory2() / vmaBindImageMemory2(). -- Before allocating memory for the resources you want to alias, check `memoryTypeBits` -returned in memory requirements of each resource to make sure the bits overlap. -Some GPUs may expose multiple memory types suitable e.g. only for buffers or -images with `COLOR_ATTACHMENT` usage, so the sets of memory types supported by your -resources may be disjoint. Aliasing them is not possible in that case. - - -\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. - -\section linear_algorithm Linear allocation algorithm - -Each Vulkan memory block managed by this library has accompanying metadata that -keeps track of used and unused regions. By default, the metadata structure and -algorithm tries to find best place for new allocations among free regions to -optimize memory usage. This way you can allocate and free objects in any order. - - - -Sometimes there is a need to use simpler, linear allocation algorithm. You can -create custom pool that uses such algorithm by adding flag -#VMA_POOL_CREATE_LINEAR_ALGORITHM_BIT to VmaPoolCreateInfo::flags while creating -#VmaPool object. Then an alternative metadata management is used. It always -creates new allocations after last one and doesn't reuse free regions after -allocations freed in the middle. It results in better allocation performance and -less memory consumed by metadata. - - - -With this one flag, you can create a custom pool that can be used in many ways: -free-at-once, stack, double stack, and ring buffer. See below for details. - -\subsection linear_algorithm_free_at_once Free-at-once - -In a pool that uses linear algorithm, you still need to free all the allocations -individually, e.g. by using vmaFreeMemory() or vmaDestroyBuffer(). You can free -them in any order. New allocations are always made after last one - free space -in the middle is not reused. However, when you release all the allocation and -the pool becomes empty, allocation starts from the beginning again. This way you -can use linear algorithm to speed up creation of allocations that you are going -to release all at once. - - - -This mode is also available for pools created with VmaPoolCreateInfo::maxBlockCount -value that allows multiple memory blocks. - -\subsection linear_algorithm_stack Stack - -When you free an allocation that was created last, its space can be reused. -Thanks to this, if you always release allocations in the order opposite to their -creation (LIFO - Last In First Out), you can achieve behavior of a stack. - - - -This mode is also available for pools created with VmaPoolCreateInfo::maxBlockCount -value that allows multiple memory blocks. - -\subsection linear_algorithm_double_stack Double stack - -The space reserved by a custom pool with linear algorithm may be used by two -stacks: - -- First, default one, growing up from offset 0. -- Second, "upper" one, growing down from the end towards lower offsets. - -To make allocation from upper stack, add flag #VMA_ALLOCATION_CREATE_UPPER_ADDRESS_BIT -to VmaAllocationCreateInfo::flags. - - - -Double stack is available only in pools with one memory block - -VmaPoolCreateInfo::maxBlockCount must be 1. Otherwise behavior is undefined. - -When the two stacks' ends meet so there is not enough space between them for a -new allocation, such allocation fails with usual -`VK_ERROR_OUT_OF_DEVICE_MEMORY` error. - -\subsection linear_algorithm_ring_buffer Ring buffer - -When you free some allocations from the beginning and there is not enough free space -for a new one at the end of a pool, allocator's "cursor" wraps around to the -beginning and starts allocation there. Thanks to this, if you always release -allocations in the same order as you created them (FIFO - First In First Out), -you can achieve behavior of a ring buffer / queue. - - - -Pools with linear algorithm support [lost allocations](@ref lost_allocations) when used as ring buffer. -If there is not enough free space for a new allocation, but existing allocations -from the front of the queue can become lost, they become lost and the allocation -succeeds. - - - -Ring buffer is available only in pools with one memory block - -VmaPoolCreateInfo::maxBlockCount must be 1. Otherwise behavior is undefined. - -\section buddy_algorithm Buddy allocation algorithm - -There is another allocation algorithm that can be used with custom pools, called -"buddy". Its internal data structure is based on a tree of blocks, each having -size that is a power of two and a half of its parent's size. When you want to -allocate memory of certain size, a free node in the tree is located. If it's too -large, it is recursively split into two halves (called "buddies"). However, if -requested allocation size is not a power of two, the size of a tree node is -aligned up to the nearest power of two and the remaining space is wasted. When -two buddy nodes become free, they are merged back into one larger node. - - - -The advantage of buddy allocation algorithm over default algorithm is faster -allocation and deallocation, as well as smaller external fragmentation. The -disadvantage is more wasted space (internal fragmentation). - -For more information, please read ["Buddy memory allocation" on Wikipedia](https://en.wikipedia.org/wiki/Buddy_memory_allocation) -or other sources that describe this concept in general. - -To use buddy allocation algorithm with a custom pool, add flag -#VMA_POOL_CREATE_BUDDY_ALGORITHM_BIT to VmaPoolCreateInfo::flags while creating -#VmaPool object. +\brief API elements related to the mechanism of \ref virtual_allocator - using the core allocation algorithm +for user-defined purpose without allocating any real GPU memory. -Several limitations apply to pools that use buddy algorithm: - -- It is recommended to use VmaPoolCreateInfo::blockSize that is a power of two. - Otherwise, only largest power of two smaller than the size is used for - allocations. The remaining space always stays unused. -- [Margins](@ref debugging_memory_usage_margins) and - [corruption detection](@ref debugging_memory_usage_corruption_detection) - don't work in such pools. -- [Lost allocations](@ref lost_allocations) don't work in such pools. You can - use them, but they never become lost. Support may be added in the future. -- [Defragmentation](@ref defragmentation) doesn't work with allocations made from - such pool. - -\page defragmentation Defragmentation - -Interleaved allocations and deallocations of many objects of varying size can -cause fragmentation over time, 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 defragmentation feature: -structure #VmaDefragmentationInfo2, function vmaDefragmentationBegin(), vmaDefragmentationEnd(). -Given set of allocations, -this function can move them to compact used memory, ensure more continuous free -space and possibly also free some `VkDeviceMemory` blocks. - -What the defragmentation does is: - -- Updates #VmaAllocation objects to point to new `VkDeviceMemory` and offset. - After allocation has been moved, its VmaAllocationInfo::deviceMemory and/or - VmaAllocationInfo::offset changes. You must query them again using - vmaGetAllocationInfo() if you need them. -- Moves actual data in memory. - -What it doesn't do, so you need to do it yourself: - -- Recreate buffers and images that were bound to allocations that were defragmented and - bind them with their new places in memory. - You must use `vkDestroyBuffer()`, `vkDestroyImage()`, - `vkCreateBuffer()`, `vkCreateImage()`, vmaBindBufferMemory(), vmaBindImageMemory() - for that purpose and NOT vmaDestroyBuffer(), - vmaDestroyImage(), vmaCreateBuffer(), vmaCreateImage(), because you don't need to - destroy or create allocation objects! -- Recreate views and update descriptors that point to these buffers and images. - -\section defragmentation_cpu Defragmenting CPU memory - -Following example demonstrates how you can run defragmentation on CPU. -Only allocations created in memory types that are `HOST_VISIBLE` can be defragmented. -Others are ignored. - -The way it works is: - -- It temporarily maps entire memory blocks when necessary. -- It moves data using `memmove()` function. - -\code -// Given following variables already initialized: -VkDevice device; -VmaAllocator allocator; -std::vector<VkBuffer> buffers; -std::vector<VmaAllocation> allocations; - - -const uint32_t allocCount = (uint32_t)allocations.size(); -std::vector<VkBool32> allocationsChanged(allocCount); - -VmaDefragmentationInfo2 defragInfo = {}; -defragInfo.allocationCount = allocCount; -defragInfo.pAllocations = allocations.data(); -defragInfo.pAllocationsChanged = allocationsChanged.data(); -defragInfo.maxCpuBytesToMove = VK_WHOLE_SIZE; // No limit. -defragInfo.maxCpuAllocationsToMove = UINT32_MAX; // No limit. - -VmaDefragmentationContext defragCtx; -vmaDefragmentationBegin(allocator, &defragInfo, nullptr, &defragCtx); -vmaDefragmentationEnd(allocator, defragCtx); - -for(uint32_t i = 0; i < allocCount; ++i) -{ - if(allocationsChanged[i]) - { - // Destroy buffer that is immutably bound to memory region which is no longer valid. - vkDestroyBuffer(device, buffers[i], nullptr); - - // Create new buffer with same parameters. - VkBufferCreateInfo bufferInfo = ...; - vkCreateBuffer(device, &bufferInfo, nullptr, &buffers[i]); - - // You can make dummy call to vkGetBufferMemoryRequirements here to silence validation layer warning. - - // Bind new buffer to new memory region. Data contained in it is already moved. - VmaAllocationInfo allocInfo; - vmaGetAllocationInfo(allocator, allocations[i], &allocInfo); - vmaBindBufferMemory(allocator, allocations[i], buffers[i]); - } -} -\endcode - -Setting VmaDefragmentationInfo2::pAllocationsChanged is optional. -This output array tells whether particular allocation in VmaDefragmentationInfo2::pAllocations at the same index -has been modified during defragmentation. -You can pass null, but you then need to query every allocation passed to defragmentation -for new parameters using vmaGetAllocationInfo() if you might need to recreate and rebind a buffer or image associated with it. - -If you use [Custom memory pools](@ref choosing_memory_type_custom_memory_pools), -you can fill VmaDefragmentationInfo2::poolCount and VmaDefragmentationInfo2::pPools -instead of VmaDefragmentationInfo2::allocationCount and VmaDefragmentationInfo2::pAllocations -to defragment all allocations in given pools. -You cannot use VmaDefragmentationInfo2::pAllocationsChanged in that case. -You can also combine both methods. - -\section defragmentation_gpu Defragmenting GPU memory - -It is also possible to defragment allocations created in memory types that are not `HOST_VISIBLE`. -To do that, you need to pass a command buffer that meets requirements as described in -VmaDefragmentationInfo2::commandBuffer. The way it works is: - -- It creates temporary buffers and binds them to entire memory blocks when necessary. -- It issues `vkCmdCopyBuffer()` to passed command buffer. - -Example: - -\code -// Given following variables already initialized: -VkDevice device; -VmaAllocator allocator; -VkCommandBuffer commandBuffer; -std::vector<VkBuffer> buffers; -std::vector<VmaAllocation> allocations; - - -const uint32_t allocCount = (uint32_t)allocations.size(); -std::vector<VkBool32> allocationsChanged(allocCount); - -VkCommandBufferBeginInfo cmdBufBeginInfo = ...; -vkBeginCommandBuffer(commandBuffer, &cmdBufBeginInfo); - -VmaDefragmentationInfo2 defragInfo = {}; -defragInfo.allocationCount = allocCount; -defragInfo.pAllocations = allocations.data(); -defragInfo.pAllocationsChanged = allocationsChanged.data(); -defragInfo.maxGpuBytesToMove = VK_WHOLE_SIZE; // Notice it's "GPU" this time. -defragInfo.maxGpuAllocationsToMove = UINT32_MAX; // Notice it's "GPU" this time. -defragInfo.commandBuffer = commandBuffer; - -VmaDefragmentationContext defragCtx; -vmaDefragmentationBegin(allocator, &defragInfo, nullptr, &defragCtx); - -vkEndCommandBuffer(commandBuffer); - -// Submit commandBuffer. -// Wait for a fence that ensures commandBuffer execution finished. - -vmaDefragmentationEnd(allocator, defragCtx); - -for(uint32_t i = 0; i < allocCount; ++i) -{ - if(allocationsChanged[i]) - { - // Destroy buffer that is immutably bound to memory region which is no longer valid. - vkDestroyBuffer(device, buffers[i], nullptr); - - // Create new buffer with same parameters. - VkBufferCreateInfo bufferInfo = ...; - vkCreateBuffer(device, &bufferInfo, nullptr, &buffers[i]); - - // You can make dummy call to vkGetBufferMemoryRequirements here to silence validation layer warning. - - // Bind new buffer to new memory region. Data contained in it is already moved. - VmaAllocationInfo allocInfo; - vmaGetAllocationInfo(allocator, allocations[i], &allocInfo); - vmaBindBufferMemory(allocator, allocations[i], buffers[i]); - } -} -\endcode - -You can combine these two methods by specifying non-zero `maxGpu*` as well as `maxCpu*` parameters. -The library automatically chooses best method to defragment each memory pool. - -You may try not to block your entire program to wait until defragmentation finishes, -but do it in the background, as long as you carefully fullfill requirements described -in function vmaDefragmentationBegin(). - -\section defragmentation_additional_notes Additional notes - -It is only legal to defragment allocations bound to: - -- buffers -- images created with `VK_IMAGE_CREATE_ALIAS_BIT`, `VK_IMAGE_TILING_LINEAR`, and - being currently in `VK_IMAGE_LAYOUT_GENERAL` or `VK_IMAGE_LAYOUT_PREINITIALIZED`. - -Defragmentation of images created with `VK_IMAGE_TILING_OPTIMAL` or in any other -layout may give undefined results. - -If you defragment allocations bound to images, new images to be bound to new -memory region after defragmentation should be created with `VK_IMAGE_LAYOUT_PREINITIALIZED` -and then transitioned to their original layout from before defragmentation if -needed using an image memory barrier. - -While using defragmentation, you may experience validation layer warnings, which you just need to ignore. -See [Validation layer warnings](@ref general_considerations_validation_layer_warnings). - -Please don't expect memory to be fully compacted after defragmentation. -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 maximize continuous -empty space inside remaining blocks, while minimizing the number and size of allocations that -need to be moved. Some fragmentation may still remain - this is normal. - -\section defragmentation_custom_algorithm Writing custom defragmentation algorithm - -If you want to implement your own, custom defragmentation algorithm, -there is infrastructure prepared for that, -but it is not exposed through the library API - you need to hack its source code. -Here are steps needed to do this: - --# Main thing you need to do is to define your own class derived from base abstract - class `VmaDefragmentationAlgorithm` and implement your version of its pure virtual methods. - See definition and comments of this class for details. --# Your code needs to interact with device memory block metadata. - If you need more access to its data than it's provided by its public interface, - declare your new class as a friend class e.g. in class `VmaBlockMetadata_Generic`. --# If you want to create a flag that would enable your algorithm or pass some additional - flags to configure it, add them to `VmaDefragmentationFlagBits` and use them in - VmaDefragmentationInfo2::flags. --# Modify function `VmaBlockVectorDefragmentationContext::Begin` to create object - of your new class whenever needed. - - -\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. - -<b>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?</b> - -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. - -<b>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?</b> - -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). - -<b>Q: How do you inform the library when new frame starts?</b> - -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. - -<b>Q: Can I touch allocation that cannot become lost?</b> - -Yes, although it has no visible effect. -Calls to vmaGetAllocationInfo() and vmaTouchAllocation() update last use frame index -also for allocations that cannot become lost, but the only way to observe it is to dump -internal allocator state using vmaBuildStatsString(). -You can use this feature for debugging purposes to explicitly mark allocations that you use -in current frame and then analyze JSON dump to see for how long each allocation stays unused. - - -\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(). - -\note Passing string name to VMA allocation doesn't automatically set it to the Vulkan buffer or image created with it. -You must do it manually using an extension like VK_EXT_debug_utils, which is independent of this library. - - -\page debugging_memory_usage Debugging incorrect memory usage - -If you suspect a bug with memory usage, like usage of uninitialized memory or -memory being overwritten out of bounds of an allocation, -you can use debug features of this library to verify this. - -\section debugging_memory_usage_initialization Memory initialization - -If you experience a bug with incorrect and nondeterministic data in your program and you suspect uninitialized memory to be used, -you can enable automatic memory initialization to verify this. -To do it, define macro `VMA_DEBUG_INITIALIZE_ALLOCATIONS` to 1. - -\code -#define VMA_DEBUG_INITIALIZE_ALLOCATIONS 1 -#include "vk_mem_alloc.h" -\endcode - -It makes memory of all new allocations initialized to bit pattern `0xDCDCDCDC`. -Before an allocation is destroyed, its memory is filled with bit pattern `0xEFEFEFEF`. -Memory is automatically mapped and unmapped if necessary. - -If you find these values while debugging your program, good chances are that you incorrectly -read Vulkan memory that is allocated but not initialized, or already freed, respectively. - -Memory initialization works only with memory types that are `HOST_VISIBLE`. -It works also with dedicated allocations. -It doesn't work with allocations created with #VMA_ALLOCATION_CREATE_CAN_BECOME_LOST_BIT flag, -as they cannot be mapped. - -\section debugging_memory_usage_margins Margins - -By default, allocations are laid out in memory blocks next to each other if possible -(considering required alignment, `bufferImageGranularity`, and `nonCoherentAtomSize`). - - - -Define macro `VMA_DEBUG_MARGIN` to some non-zero value (e.g. 16) to enforce specified -number of bytes as a margin before and after every allocation. - -\code -#define VMA_DEBUG_MARGIN 16 -#include "vk_mem_alloc.h" -\endcode - - - -If your bug goes away after enabling margins, it means it may be caused by memory -being overwritten outside of allocation boundaries. It is not 100% certain though. -Change in application behavior may also be caused by different order and distribution -of allocations across memory blocks after margins are applied. - -The margin is applied also before first and after last allocation in a block. -It may occur only once between two adjacent allocations. - -Margins work with all types of memory. - -Margin is applied only to allocations made out of memory blocks and not to dedicated -allocations, which have their own memory block of specific size. -It is thus not applied to allocations made using #VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT flag -or those automatically decided to put into dedicated allocations, e.g. due to its -large size or recommended by VK_KHR_dedicated_allocation extension. -Margins are also not active in custom pools created with #VMA_POOL_CREATE_BUDDY_ALGORITHM_BIT flag. - -Margins appear in [JSON dump](@ref statistics_json_dump) as part of free space. - -Note that enabling margins increases memory usage and fragmentation. - -\section debugging_memory_usage_corruption_detection Corruption detection - -You can additionally define macro `VMA_DEBUG_DETECT_CORRUPTION` to 1 to enable validation -of contents of the margins. - -\code -#define VMA_DEBUG_MARGIN 16 -#define VMA_DEBUG_DETECT_CORRUPTION 1 -#include "vk_mem_alloc.h" -\endcode - -When this feature is enabled, number of bytes specified as `VMA_DEBUG_MARGIN` -(it must be multiply of 4) before and after every allocation is filled with a magic number. -This idea is also know as "canary". -Memory is automatically mapped and unmapped if necessary. - -This number is validated automatically when the allocation is destroyed. -If it's not equal to the expected value, `VMA_ASSERT()` is executed. -It clearly means that either CPU or GPU overwritten the memory outside of boundaries of the allocation, -which indicates a serious bug. - -You can also explicitly request checking margins of all allocations in all memory blocks -that belong to specified memory types by using function vmaCheckCorruption(), -or in memory blocks that belong to specified custom pool, by using function -vmaCheckPoolCorruption(). - -Margin validation (corruption detection) works only for memory types that are -`HOST_VISIBLE` and `HOST_COHERENT`. - - -\page record_and_replay Record and replay - -\section record_and_replay_introduction Introduction - -While using the library, sequence of calls to its functions together with their -parameters can be recorded to a file and later replayed using standalone player -application. It can be useful to: - -- Test correctness - check if same sequence of calls will not cause crash or - failures on a target platform. -- Gather statistics - see number of allocations, peak memory usage, number of - calls etc. -- Benchmark performance - see how much time it takes to replay the whole - sequence. - -\section record_and_replay_usage Usage - -Recording functionality is disabled by default. -To enable it, define following macro before every include of this library: - -\code -#define VMA_RECORDING_ENABLED 1 -\endcode - -<b>To record sequence of calls to a file:</b> Fill in -VmaAllocatorCreateInfo::pRecordSettings member while creating #VmaAllocator -object. File is opened and written during whole lifetime of the allocator. - -<b>To replay file:</b> Use VmaReplay - standalone command-line program. -Precompiled binary can be found in "bin" directory. -Its source can be found in "src/VmaReplay" directory. -Its project is generated by Premake. -Command line syntax is printed when the program is launched without parameters. -Basic usage: - - VmaReplay.exe MyRecording.csv - -<b>Documentation of file format</b> can be found in file: "docs/Recording file format.md". -It's a human-readable, text file in CSV format (Comma Separated Values). - -\section record_and_replay_additional_considerations Additional considerations - -- Replaying file that was recorded on a different GPU (with different parameters - like `bufferImageGranularity`, `nonCoherentAtomSize`, and especially different - set of memory heaps and types) may give different performance and memory usage - results, as well as issue some warnings and errors. -- Current implementation of recording in VMA, as well as VmaReplay application, is - coded and tested only on Windows. Inclusion of recording code is driven by - `VMA_RECORDING_ENABLED` macro. Support for other platforms should be easy to - add. Contributions are welcomed. - - -\page usage_patterns Recommended usage patterns - -See also slides from talk: -[Sawicki, Adam. Advanced Graphics Techniques Tutorial: Memory management in Vulkan and DX12. Game Developers Conference, 2018](https://www.gdcvault.com/play/1025458/Advanced-Graphics-Techniques-Tutorial-New) - - -\section usage_patterns_common_mistakes Common mistakes - -<b>Use of CPU_TO_GPU instead of CPU_ONLY memory</b> - -#VMA_MEMORY_USAGE_CPU_TO_GPU is recommended only for resources that will be -mapped and written by the CPU, as well as read directly by the GPU - like some -buffers or textures updated every frame (dynamic). If you create a staging copy -of a resource to be written by CPU and then used as a source of transfer to -another resource placed in the GPU memory, that staging resource should be -created with #VMA_MEMORY_USAGE_CPU_ONLY. Please read the descriptions of these -enums carefully for details. - -<b>Unnecessary use of custom pools</b> - -\ref custom_memory_pools may be useful for special purposes - when you want to -keep certain type of resources separate e.g. to reserve minimum amount of memory -for them, limit maximum amount of memory they can occupy, or make some of them -push out the other through the mechanism of \ref lost_allocations. For most -resources this is not needed and so it is not recommended to create #VmaPool -objects and allocations out of them. Allocating from the default pool is sufficient. - -\section usage_patterns_simple Simple patterns - -\subsection usage_patterns_simple_render_targets Render targets - -<b>When:</b> -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)"). - -<b>What to do:</b> -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 - -<b>When:</b> -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. - -<b>What to do:</b> -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 - -<b>When:</b> -Any resources that change frequently (aka "dynamic"), e.g. every frame or every draw call, -written on CPU, read on GPU. - -<b>What to do:</b> -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 on CPU, as it may be very slow. -Also note that textures written directly from the host through a mapped pointer need to be in LINEAR not OPTIMAL layout. - -\subsection usage_patterns_readback Readback - -<b>When:</b> -Resources that contain data written by GPU that you want to read back on CPU, -e.g. results of some computations. - -<b>What to do:</b> -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 comparably 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 transfer each time. --# Create just a single copy using #VMA_MEMORY_USAGE_CPU_TO_GPU, map it and fill it on CPU, - read it directly on GPU. --# Create just a 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 unnecessary time and memory overhead - related to using a second copy and making transfer. -- 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. - -Note that textures accessed directly from the host through a mapped pointer need to be in LINEAR layout, -which may slow down their usage on the device. -Textures accessed only by the device and transfer operations can use OPTIMAL layout. - -If you don't want to specialize your code for specific types of GPUs, you 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. - -For example, define `VMA_ASSERT(expr)` before including the library to provide -custom implementation of the assertion, compatible with your project. -By default it is defined to standard C `assert(expr)` in `_DEBUG` configuration -and empty otherwise. - -\section config_Vulkan_functions Pointers to Vulkan functions - -There are multiple ways to import pointers to Vulkan functions in the library. -In the simplest case you don't need to do anything. -If the compilation or linking of your program or the initialization of the #VmaAllocator -doesn't work for you, you can try to reconfigure it. - -First, the allocator tries to fetch pointers to Vulkan functions linked statically, -like this: - -\code -m_VulkanFunctions.vkAllocateMemory = (PFN_vkAllocateMemory)vkAllocateMemory; -\endcode - -If you want to disable this feature, set configuration macro: `#define VMA_STATIC_VULKAN_FUNCTIONS 0`. - -Second, you can provide the pointers yourself by setting member VmaAllocatorCreateInfo::pVulkanFunctions. -You can fetch them e.g. using functions `vkGetInstanceProcAddr` and `vkGetDeviceProcAddr` or -by using a helper library like [volk](https://github.com/zeux/volk). - -Third, VMA tries to fetch remaining pointers that are still null by calling -`vkGetInstanceProcAddr` and `vkGetDeviceProcAddr` on its own. -If you want to disable this feature, set configuration macro: `#define VMA_DYNAMIC_VULKAN_FUNCTIONS 0`. - -Finally, all the function pointers required by the library (considering selected -Vulkan version and enabled extensions) are checked with `VMA_ASSERT` if they are not null. - - -\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 - -When device memory of certain heap runs out of free space, new allocations may -fail (returning error code) or they may succeed, silently pushing some existing -memory blocks from GPU VRAM to system RAM (which degrades performance). This -behavior is implementation-dependent - it depends on GPU vendor and graphics -driver. - -On AMD cards it can be controlled while creating Vulkan device object by using -VK_AMD_memory_overallocation_behavior extension, if available. - -Alternatively, 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 #VmaAllocator`to 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/1.2-extensions/html/chap50.html#VK_KHR_dedicated_allocation) -- [VK_KHR_dedicated_allocation unofficial manual](http://asawicki.info/articles/VK_KHR_dedicated_allocation.php5) - - - -\page vk_amd_device_coherent_memory VK_AMD_device_coherent_memory - -VK_AMD_device_coherent_memory is a device extension that enables access to -additional memory types with `VK_MEMORY_PROPERTY_DEVICE_COHERENT_BIT_AMD` and -`VK_MEMORY_PROPERTY_DEVICE_UNCACHED_BIT_AMD` flag. It is useful mostly for -allocation of buffers intended for writing "breadcrumb markers" in between passes -or draw calls, which in turn are useful for debugging GPU crash/hang/TDR cases. - -When the extension is available but has not been enabled, Vulkan physical device -still exposes those memory types, but their usage is forbidden. VMA automatically -takes care of that - it returns `VK_ERROR_FEATURE_NOT_PRESENT` when an attempt -to allocate memory of such type is made. - -If you want to use this extension in connection with VMA, follow these steps: - -\section vk_amd_device_coherent_memory_initialization Initialization - -1) Call `vkEnumerateDeviceExtensionProperties` for the physical device. -Check if the extension is supported - if returned array of `VkExtensionProperties` contains "VK_AMD_device_coherent_memory". - -2) Call `vkGetPhysicalDeviceFeatures2` for the physical device instead of old `vkGetPhysicalDeviceFeatures`. -Attach additional structure `VkPhysicalDeviceCoherentMemoryFeaturesAMD` to `VkPhysicalDeviceFeatures2::pNext` to be returned. -Check if the device feature is really supported - check if `VkPhysicalDeviceCoherentMemoryFeaturesAMD::deviceCoherentMemory` is true. - -3) While creating device with `vkCreateDevice`, enable this extension - add "VK_AMD_device_coherent_memory" -to the list passed as `VkDeviceCreateInfo::ppEnabledExtensionNames`. - -4) While creating the device, also don't set `VkDeviceCreateInfo::pEnabledFeatures`. -Fill in `VkPhysicalDeviceFeatures2` structure instead and pass it as `VkDeviceCreateInfo::pNext`. -Enable this device feature - attach additional structure `VkPhysicalDeviceCoherentMemoryFeaturesAMD` to -`VkPhysicalDeviceFeatures2::pNext` and set its member `deviceCoherentMemory` to `VK_TRUE`. - -5) While creating #VmaAllocator with vmaCreateAllocator() inform VMA that you -have enabled this extension and feature - add #VMA_ALLOCATOR_CREATE_AMD_DEVICE_COHERENT_MEMORY_BIT -to VmaAllocatorCreateInfo::flags. - -\section vk_amd_device_coherent_memory_usage Usage - -After following steps described above, you can create VMA allocations and custom pools -out of the special `DEVICE_COHERENT` and `DEVICE_UNCACHED` memory types on eligible -devices. There are multiple ways to do it, for example: - -- You can request or prefer to allocate out of such memory types by adding - `VK_MEMORY_PROPERTY_DEVICE_COHERENT_BIT_AMD` to VmaAllocationCreateInfo::requiredFlags - or VmaAllocationCreateInfo::preferredFlags. Those flags can be freely mixed with - other ways of \ref choosing_memory_type, like setting VmaAllocationCreateInfo::usage. -- If you manually found memory type index to use for this purpose, force allocation - from this specific index by setting VmaAllocationCreateInfo::memoryTypeBits `= 1u << index`. - -\section vk_amd_device_coherent_memory_more_information More information - -To learn more about this extension, see [VK_AMD_device_coherent_memory in Vulkan specification](https://www.khronos.org/registry/vulkan/specs/1.2-extensions/html/chap44.html#VK_AMD_device_coherent_memory) - -Example use of this extension can be found in the code of the sample and test suite -accompanying this library. - - -\page enabling_buffer_device_address Enabling buffer device address - -Device extension VK_KHR_buffer_device_address -allow to fetch raw GPU pointer to a buffer and pass it for usage in a shader code. -It is promoted to core Vulkan 1.2. - -If you want to use this feature in connection with VMA, follow these steps: - -\section enabling_buffer_device_address_initialization Initialization - -1) (For Vulkan version < 1.2) Call `vkEnumerateDeviceExtensionProperties` for the physical device. -Check if the extension is supported - if returned array of `VkExtensionProperties` contains -"VK_KHR_buffer_device_address". - -2) Call `vkGetPhysicalDeviceFeatures2` for the physical device instead of old `vkGetPhysicalDeviceFeatures`. -Attach additional structure `VkPhysicalDeviceBufferDeviceAddressFeatures*` to `VkPhysicalDeviceFeatures2::pNext` to be returned. -Check if the device feature is really supported - check if `VkPhysicalDeviceBufferDeviceAddressFeatures*::bufferDeviceAddress` is true. - -3) (For Vulkan version < 1.2) While creating device with `vkCreateDevice`, enable this extension - add -"VK_KHR_buffer_device_address" to the list passed as `VkDeviceCreateInfo::ppEnabledExtensionNames`. - -4) While creating the device, also don't set `VkDeviceCreateInfo::pEnabledFeatures`. -Fill in `VkPhysicalDeviceFeatures2` structure instead and pass it as `VkDeviceCreateInfo::pNext`. -Enable this device feature - attach additional structure `VkPhysicalDeviceBufferDeviceAddressFeatures*` to -`VkPhysicalDeviceFeatures2::pNext` and set its member `bufferDeviceAddress` to `VK_TRUE`. - -5) While creating #VmaAllocator with vmaCreateAllocator() inform VMA that you -have enabled this feature - add #VMA_ALLOCATOR_CREATE_BUFFER_DEVICE_ADDRESS_BIT -to VmaAllocatorCreateInfo::flags. - -\section enabling_buffer_device_address_usage Usage - -After following steps described above, you can create buffers with `VK_BUFFER_USAGE_SHADER_DEVICE_ADDRESS_BIT*` using VMA. -The library automatically adds `VK_MEMORY_ALLOCATE_DEVICE_ADDRESS_BIT*` to -allocated memory blocks wherever it might be needed. - -Please note that the library supports only `VK_BUFFER_USAGE_SHADER_DEVICE_ADDRESS_BIT*`. -The second part of this functionality related to "capture and replay" is not supported, -as it is intended for usage in debugging tools like RenderDoc, not in everyday Vulkan usage. - -\section enabling_buffer_device_address_more_information More information - -To learn more about this extension, see [VK_KHR_buffer_device_address in Vulkan specification](https://www.khronos.org/registry/vulkan/specs/1.2-extensions/html/chap46.html#VK_KHR_buffer_device_address) - -Example use of this extension can be found in the code of the sample and test suite -accompanying this library. - -\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_validation_layer_warnings Validation layer warnings - -When using this library, you can meet following types of warnings issued by -Vulkan validation layer. They don't necessarily indicate a bug, so you may need -to just ignore them. - -- *vkBindBufferMemory(): Binding memory to buffer 0xeb8e4 but vkGetBufferMemoryRequirements() has not been called on that buffer.* - - It happens when VK_KHR_dedicated_allocation extension is enabled. - `vkGetBufferMemoryRequirements2KHR` function is used instead, while validation layer seems to be unaware of it. -- *Mapping an image with layout VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL can result in undefined behavior if this memory is used by the device. Only GENERAL or PREINITIALIZED should be used.* - - It happens when you map a buffer or image, because the library maps entire - `VkDeviceMemory` block, where different types of images and buffers may end - up together, especially on GPUs with unified memory like Intel. -- *Non-linear image 0xebc91 is aliased with linear buffer 0xeb8e4 which may indicate a bug.* - - It happens when you use lost allocations, and a new image or buffer is - created in place of an existing object that became lost. - - It may happen also when you use [defragmentation](@ref defragmentation). - -\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. Uploading (streaming) and downloading data of buffers and images - between CPU and GPU memory and related synchronization is responsibility of the user. - Defining some "texture" object that would automatically stream its data from a - staging copy in CPU memory to GPU memory would rather be a feature of another, - higher-level library implemented on top of VMA. -- Allocations for imported/exported external memory. They tend to require - explicit memory type index and dedicated allocation anyway, so they don't - interact with main features of this library. Such special purpose allocations - should be made manually, using `vkCreateBuffer()` and `vkAllocateMemory()`. -- Sub-allocation of parts of one large buffer. Although recommended as a good practice, - it is the user's responsibility to implement such logic on top of VMA. -- Recreation of buffers and images. Although the library has functions for - buffer and image creation (vmaCreateBuffer(), vmaCreateImage()), you need to - recreate these objects yourself after defragmentation. That's because the big - structures `VkBufferCreateInfo`, `VkImageCreateInfo` are not stored in - #VmaAllocation object. -- Handling CPU memory allocation failures. When dynamically creating small C++ - objects in CPU memory (not Vulkan memory), allocation failures are not checked - and handled gracefully, because that would complicate code significantly and - is usually not needed in desktop PC applications anyway. - Success of an allocation is just checked with an assert. -- Code free of any compiler warnings. Maintaining the library to compile and - work correctly on so many different platforms is hard enough. Being free of - any warnings, on any version of any compiler, is simply not feasible. -- This is a C++ library with C interface. - Bindings or ports to any other programming languages are welcomed as external projects and - are not going to be included into this repository. +\defgroup group_stats Statistics +\brief API elements that query current status of the allocator, from memory usage, budget, to full dump of the internal state in JSON format. */ + #ifdef __cplusplus extern "C" { #endif -/* -Define this macro to 0/1 to disable/enable support for recording functionality, -available through VmaAllocatorCreateInfo::pRecordSettings. -*/ -#ifndef VMA_RECORDING_ENABLED - #define VMA_RECORDING_ENABLED 0 +#ifndef VULKAN_H_ + #ifdef USE_VOLK + #include <volk.h> + #else + #include <vulkan/vulkan.h> + #endif #endif -#if !defined(NOMINMAX) && defined(VMA_IMPLEMENTATION) - #define NOMINMAX // For windows.h +// Define this macro to declare maximum supported Vulkan version in format AAABBBCCC, +// where AAA = major, BBB = minor, CCC = patch. +// If you want to use version > 1.0, it still needs to be enabled via VmaAllocatorCreateInfo::vulkanApiVersion. +#if !defined(VMA_VULKAN_VERSION) + #if defined(VK_VERSION_1_3) + #define VMA_VULKAN_VERSION 1003000 + #elif defined(VK_VERSION_1_2) + #define VMA_VULKAN_VERSION 1002000 + #elif defined(VK_VERSION_1_1) + #define VMA_VULKAN_VERSION 1001000 + #else + #define VMA_VULKAN_VERSION 1000000 + #endif #endif #if defined(__ANDROID__) && defined(VK_NO_PROTOTYPES) && VMA_STATIC_VULKAN_FUNCTIONS @@ -2062,25 +178,11 @@ available through VmaAllocatorCreateInfo::pRecordSettings. #endif // #if VMA_VULKAN_VERSION >= 1001000 #endif // #if defined(__ANDROID__) && VMA_STATIC_VULKAN_FUNCTIONS && VK_NO_PROTOTYPES -#ifndef VULKAN_H_ - #ifdef USE_VOLK - #include <volk.h> - #else - #include <vulkan/vulkan.h> - #endif -#endif - -// Define this macro to declare maximum supported Vulkan version in format AAABBBCCC, -// where AAA = major, BBB = minor, CCC = patch. -// If you want to use version > 1.0, it still needs to be enabled via VmaAllocatorCreateInfo::vulkanApiVersion. -#if !defined(VMA_VULKAN_VERSION) - #if defined(VK_VERSION_1_2) - #define VMA_VULKAN_VERSION 1002000 - #elif defined(VK_VERSION_1_1) - #define VMA_VULKAN_VERSION 1001000 - #else - #define VMA_VULKAN_VERSION 1000000 - #endif +#if !defined(VK_VERSION_1_2) + // This one is tricky. Vulkan specification defines this code as available since + // Vulkan 1.0, but doesn't actually define it in Vulkan SDK earlier than 1.2.131. + // See pull request #207. + #define VK_ERROR_UNKNOWN ((VkResult)-13) #endif #if !defined(VMA_DEDICATED_ALLOCATION) @@ -2199,50 +301,29 @@ available through VmaAllocatorCreateInfo::pRecordSettings. #endif #endif -/** \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 VMA_NOT_NULL allocator, - uint32_t memoryType, - VkDeviceMemory VMA_NOT_NULL_NON_DISPATCHABLE memory, - VkDeviceSize size, - void* VMA_NULLABLE pUserData); -/// Callback function called before vkFreeMemory. -typedef void (VKAPI_PTR *PFN_vmaFreeDeviceMemoryFunction)( - VmaAllocator VMA_NOT_NULL allocator, - uint32_t memoryType, - VkDeviceMemory VMA_NOT_NULL_NON_DISPATCHABLE memory, - VkDeviceSize size, - void* VMA_NULLABLE pUserData); +#ifndef VMA_STATS_STRING_ENABLED + #define VMA_STATS_STRING_ENABLED 1 +#endif -/** \brief Set of callbacks that the library will call for `vkAllocateMemory` and `vkFreeMemory`. +//////////////////////////////////////////////////////////////////////////////// +//////////////////////////////////////////////////////////////////////////////// +// +// INTERFACE +// +//////////////////////////////////////////////////////////////////////////////// +//////////////////////////////////////////////////////////////////////////////// -Provided for informative purpose, e.g. to gather statistics about number of -allocations or total amount of memory allocated in Vulkan. +// Sections for managing code placement in file, only for development purposes e.g. for convenient folding inside an IDE. +#ifndef _VMA_ENUM_DECLARATIONS -Used in VmaAllocatorCreateInfo::pDeviceMemoryCallbacks. +/** +\addtogroup group_init +@{ */ -typedef struct VmaDeviceMemoryCallbacks { - /// Optional, can be null. - PFN_vmaAllocateDeviceMemoryFunction VMA_NULLABLE pfnAllocate; - /// Optional, can be null. - PFN_vmaFreeDeviceMemoryFunction VMA_NULLABLE pfnFree; - /// Optional, can be null. - void* VMA_NULLABLE pUserData; -} VmaDeviceMemoryCallbacks; /// Flags for created #VmaAllocator. -typedef enum VmaAllocatorCreateFlagBits { +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. @@ -2251,7 +332,7 @@ typedef enum VmaAllocatorCreateFlagBits { /** \brief Enables usage of VK_KHR_dedicated_allocation extension. The flag works only if VmaAllocatorCreateInfo::vulkanApiVersion `== VK_API_VERSION_1_0`. - When it's `VK_API_VERSION_1_1`, the flag is ignored because the extension has been promoted to Vulkan 1.1. + When it is `VK_API_VERSION_1_1`, the flag is ignored because the extension has been promoted to Vulkan 1.1. Using this extension will automatically allocate dedicated blocks of memory for some buffers and images instead of suballocating place for them out of bigger @@ -2277,7 +358,7 @@ typedef enum VmaAllocatorCreateFlagBits { Enables usage of VK_KHR_bind_memory2 extension. The flag works only if VmaAllocatorCreateInfo::vulkanApiVersion `== VK_API_VERSION_1_0`. - When it's `VK_API_VERSION_1_1`, the flag is ignored because the extension has been promoted to Vulkan 1.1. + When it is `VK_API_VERSION_1_1`, the flag is ignored because the extension has been promoted to Vulkan 1.1. You may set this flag only if you found out that this device extension is supported, you enabled it while creating Vulkan device passed as VmaAllocatorCreateInfo::device, @@ -2358,11 +439,478 @@ typedef enum VmaAllocatorCreateFlagBits { } VmaAllocatorCreateFlagBits; typedef VkFlags VmaAllocatorCreateFlags; +/** @} */ + +/** +\addtogroup group_alloc +@{ +*/ + +/// \brief Intended usage of the allocated memory. +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 on GPU. + + 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. + Guarantees to be `HOST_VISIBLE` and `HOST_COHERENT`. + CPU access is typically uncached. Writes may be write-combined. + Resources created in this pool may still be accessible to the device, but access to them can be slow. + 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 access is typically uncached. Writes may be write-combined. + + Usage: Resources written frequently by host (dynamic), read by device. E.g. textures (with LINEAR layout), 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, + /** CPU memory - memory that is preferably not `DEVICE_LOCAL`, but also not guaranteed to be `HOST_VISIBLE`. + + Usage: Staging copy of resources moved from GPU memory to CPU memory as part + of custom paging/residency mechanism, to be moved back to GPU memory when needed. + */ + VMA_MEMORY_USAGE_CPU_COPY = 5, + /** Lazily allocated GPU memory having `VK_MEMORY_PROPERTY_LAZILY_ALLOCATED_BIT`. + Exists mostly on mobile platforms. Using it on desktop PC or other GPUs with no such memory type present will fail the allocation. + + Usage: Memory for transient attachment images (color attachments, depth attachments etc.), created with `VK_IMAGE_USAGE_TRANSIENT_ATTACHMENT_BIT`. + + Allocations with this usage are always created as dedicated - it implies #VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT. + */ + VMA_MEMORY_USAGE_GPU_LAZILY_ALLOCATED = 6, + + 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. + */ + VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT = 0x00000001, + + /** \brief Set this flag to only try to allocate from existing `VkDeviceMemory` blocks and never create new such block. + + If new allocation cannot be placed in any of the existing blocks, allocation + fails with `VK_ERROR_OUT_OF_DEVICE_MEMORY` error. + + You should not use #VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT and + #VMA_ALLOCATION_CREATE_NEVER_ALLOCATE_BIT at the same time. It makes no sense. + + If VmaAllocationCreateInfo::pool is not null, this flag is implied and ignored. */ + VMA_ALLOCATION_CREATE_NEVER_ALLOCATE_BIT = 0x00000002, + /** \brief Set this flag to use a memory that will be persistently mapped and retrieve pointer to it. + + Pointer to mapped memory will be returned through VmaAllocationInfo::pMappedData. + + It is valid to use this flag for allocation made from memory type that is not + `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). + */ + VMA_ALLOCATION_CREATE_MAPPED_BIT = 0x00000004, + /// \deprecated Removed. Do not use. + VMA_ALLOCATION_CREATE_RESERVED_1_BIT = 0x00000008, + /// \deprecated Removed. Do not use. + VMA_ALLOCATION_CREATE_RESERVED_2_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, + /** Allocation will be created from upper stack in a double stack pool. + + This flag is only allowed for custom pools created with #VMA_POOL_CREATE_LINEAR_ALGORITHM_BIT flag. + */ + VMA_ALLOCATION_CREATE_UPPER_ADDRESS_BIT = 0x00000040, + /** Create both buffer/image and allocation, but don't bind them together. + It is useful when you want to bind yourself to do some more advanced binding, e.g. using some extensions. + The flag is meaningful only with functions that bind by default: vmaCreateBuffer(), vmaCreateImage(). + Otherwise it is ignored. + */ + VMA_ALLOCATION_CREATE_DONT_BIND_BIT = 0x00000080, + /** Create allocation only if additional device memory required for it, if any, won't exceed + memory budget. Otherwise return `VK_ERROR_OUT_OF_DEVICE_MEMORY`. + */ + VMA_ALLOCATION_CREATE_WITHIN_BUDGET_BIT = 0x00000100, + /** \brief Set this flag if the allocated memory will have aliasing resources. + * + Usage of this flag prevents supplying `VkMemoryDedicatedAllocateInfoKHR` when #VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT is specified. + Otherwise created dedicated memory will not be suitable for aliasing resources, resulting in Vulkan Validation Layer errors. + */ + VMA_ALLOCATION_CREATE_CAN_ALIAS_BIT = 0x00000200, + /** Allocation strategy that chooses smallest possible free range for the allocation + to minimize memory usage and fragmentation, possibly at the expense of allocation time. + */ + VMA_ALLOCATION_CREATE_STRATEGY_MIN_MEMORY_BIT = 0x00010000, + /** Allocation strategy that chooses first suitable free range for the allocation - + not necessarily in terms of the smallest offset but the one that is easiest and fastest to find + to minimize allocation time, possibly at the expense of allocation quality. + */ + VMA_ALLOCATION_CREATE_STRATEGY_MIN_TIME_BIT = 0x00020000, + /** Alias to #VMA_ALLOCATION_CREATE_STRATEGY_MIN_MEMORY_BIT. + */ + VMA_ALLOCATION_CREATE_STRATEGY_BEST_FIT_BIT = VMA_ALLOCATION_CREATE_STRATEGY_MIN_MEMORY_BIT, + /** Alias to #VMA_ALLOCATION_CREATE_STRATEGY_MIN_TIME_BIT. + */ + VMA_ALLOCATION_CREATE_STRATEGY_FIRST_FIT_BIT = VMA_ALLOCATION_CREATE_STRATEGY_MIN_TIME_BIT, + /** A bit mask to extract only `STRATEGY` bits from entire set of flags. + */ + VMA_ALLOCATION_CREATE_STRATEGY_MASK = + VMA_ALLOCATION_CREATE_STRATEGY_MIN_MEMORY_BIT | + VMA_ALLOCATION_CREATE_STRATEGY_MIN_TIME_BIT, + + VMA_ALLOCATION_CREATE_FLAG_BITS_MAX_ENUM = 0x7FFFFFFF +} VmaAllocationCreateFlagBits; +typedef VkFlags VmaAllocationCreateFlags; + +/// 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 an 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 + faster and more optimal. + */ + VMA_POOL_CREATE_IGNORE_BUFFER_IMAGE_GRANULARITY_BIT = 0x00000002, + + /** \brief Enables alternative, linear allocation algorithm in this pool. + + Specify this flag to enable linear allocation algorithm, which always creates + new allocations after last one and doesn't reuse space from allocations freed in + between. It trades memory consumption for simplified algorithm and data + structure, which has better performance and uses less memory for metadata. + + By using this flag, you can achieve behavior of free-at-once, stack, + ring buffer, and double stack. + For details, see documentation chapter \ref linear_algorithm. + */ + VMA_POOL_CREATE_LINEAR_ALGORITHM_BIT = 0x00000004, + + /** \brief Enables alternative, buddy allocation algorithm in this pool. + + It operates on a tree of blocks, each having size that is a power of two and + a half of its parent's size. Comparing to default algorithm, this one provides + faster allocation and deallocation and decreased external fragmentation, + at the expense of more memory wasted (internal fragmentation). + For details, see documentation chapter \ref buddy_algorithm. + */ + VMA_POOL_CREATE_BUDDY_ALGORITHM_BIT = 0x00000008, + + /** \brief Enables alternative, Two-Level Segregated Fit (TLSF) allocation algorithm in this pool. + + This algorithm is based on 2-level lists dividing address space into smaller + chunks. The first level is aligned to power of two which serves as buckets for requested + memory to fall into, and the second level is lineary subdivided into lists of free memory. + This algorithm aims to achieve bounded response time even in the worst case scenario. + Allocation time can be sometimes slightly longer than compared to other algorithms + but in return the application can avoid stalls in case of fragmentation, giving + predictable results, suitable for real-time use cases. + */ + VMA_POOL_CREATE_TLSF_ALGORITHM_BIT = 0x00000010, + + /** Bit mask to extract only `ALGORITHM` bits from entire set of flags. + */ + VMA_POOL_CREATE_ALGORITHM_MASK = + VMA_POOL_CREATE_LINEAR_ALGORITHM_BIT | + VMA_POOL_CREATE_BUDDY_ALGORITHM_BIT | + VMA_POOL_CREATE_TLSF_ALGORITHM_BIT, + + VMA_POOL_CREATE_FLAG_BITS_MAX_ENUM = 0x7FFFFFFF +} VmaPoolCreateFlagBits; +/// Flags to be passed as VmaPoolCreateInfo::flags. See #VmaPoolCreateFlagBits. +typedef VkFlags VmaPoolCreateFlags; + +/// Flags to be used in vmaDefragmentationBegin(). None at the moment. Reserved for future use. +typedef enum VmaDefragmentationFlagBits +{ + VMA_DEFRAGMENTATION_FLAG_INCREMENTAL = 0x1, + VMA_DEFRAGMENTATION_FLAG_BITS_MAX_ENUM = 0x7FFFFFFF +} VmaDefragmentationFlagBits; +typedef VkFlags VmaDefragmentationFlags; + +/** @} */ + +/** +\addtogroup group_virtual +@{ +*/ + +/// Flags to be passed as VmaVirtualBlockCreateInfo::flags. +typedef enum VmaVirtualBlockCreateFlagBits +{ + /** \brief Enables alternative, linear allocation algorithm in this virtual block. + + Specify this flag to enable linear allocation algorithm, which always creates + new allocations after last one and doesn't reuse space from allocations freed in + between. It trades memory consumption for simplified algorithm and data + structure, which has better performance and uses less memory for metadata. + + By using this flag, you can achieve behavior of free-at-once, stack, + ring buffer, and double stack. + For details, see documentation chapter \ref linear_algorithm. + */ + VMA_VIRTUAL_BLOCK_CREATE_LINEAR_ALGORITHM_BIT = 0x00000001, + + /** \brief Enables alternative, buddy allocation algorithm in this virtual block. + + It operates on a tree of blocks, each having size that is a power of two and + a half of its parent's size. Comparing to default algorithm, this one provides + faster allocation and deallocation and decreased external fragmentation, + at the expense of more memory wasted (internal fragmentation). + For details, see documentation chapter \ref buddy_algorithm. + */ + VMA_VIRTUAL_BLOCK_CREATE_BUDDY_ALGORITHM_BIT = 0x00000002, + + /** \brief Enables alternative, TLSF allocation algorithm in virtual block. + + This algorithm is based on 2-level lists dividing address space into smaller + chunks. The first level is aligned to power of two which serves as buckets for requested + memory to fall into, and the second level is lineary subdivided into lists of free memory. + This algorithm aims to achieve bounded response time even in the worst case scenario. + Allocation time can be sometimes slightly longer than compared to other algorithms + but in return the application can avoid stalls in case of fragmentation, giving + predictable results, suitable for real-time use cases. + */ + VMA_VIRTUAL_BLOCK_CREATE_TLSF_ALGORITHM_BIT = 0x00000004, + + /** \brief Bit mask to extract only `ALGORITHM` bits from entire set of flags. + */ + VMA_VIRTUAL_BLOCK_CREATE_ALGORITHM_MASK = + VMA_VIRTUAL_BLOCK_CREATE_LINEAR_ALGORITHM_BIT | + VMA_VIRTUAL_BLOCK_CREATE_BUDDY_ALGORITHM_BIT | + VMA_VIRTUAL_BLOCK_CREATE_TLSF_ALGORITHM_BIT, + + VMA_VIRTUAL_BLOCK_CREATE_FLAG_BITS_MAX_ENUM = 0x7FFFFFFF +} VmaVirtualBlockCreateFlagBits; +/// Flags to be passed as VmaVirtualBlockCreateInfo::flags. See #VmaVirtualBlockCreateFlagBits. +typedef VkFlags VmaVirtualBlockCreateFlags; + +/// Flags to be passed as VmaVirtualAllocationCreateInfo::flags. +typedef enum VmaVirtualAllocationCreateFlagBits +{ + /** \brief Allocation will be created from upper stack in a double stack pool. + + This flag is only allowed for virtual blocks created with #VMA_VIRTUAL_BLOCK_CREATE_LINEAR_ALGORITHM_BIT flag. + */ + VMA_VIRTUAL_ALLOCATION_CREATE_UPPER_ADDRESS_BIT = VMA_ALLOCATION_CREATE_UPPER_ADDRESS_BIT, + /** \brief Allocation strategy that tries to minimize memory usage. + */ + VMA_VIRTUAL_ALLOCATION_CREATE_STRATEGY_MIN_MEMORY_BIT = VMA_ALLOCATION_CREATE_STRATEGY_MIN_MEMORY_BIT, + /** \brief Allocation strategy that tries to minimize allocation time. + */ + VMA_VIRTUAL_ALLOCATION_CREATE_STRATEGY_MIN_TIME_BIT = VMA_ALLOCATION_CREATE_STRATEGY_MIN_TIME_BIT, + /** \brief A bit mask to extract only `STRATEGY` bits from entire set of flags. + + These strategy flags are binary compatible with equivalent flags in #VmaAllocationCreateFlagBits. + */ + VMA_VIRTUAL_ALLOCATION_CREATE_STRATEGY_MASK = VMA_ALLOCATION_CREATE_STRATEGY_MASK, + + VMA_VIRTUAL_ALLOCATION_CREATE_FLAG_BITS_MAX_ENUM = 0x7FFFFFFF +} VmaVirtualAllocationCreateFlagBits; +/// Flags to be passed as VmaVirtualAllocationCreateInfo::flags. See #VmaVirtualAllocationCreateFlagBits. +typedef VkFlags VmaVirtualAllocationCreateFlags; + +/** @} */ + +#endif // _VMA_ENUM_DECLARATIONS + +#ifndef _VMA_DATA_TYPES_DECLARATIONS + +/** +\addtogroup group_init +@{ */ + +/** \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) + +/** @} */ + +/** +\addtogroup group_alloc +@{ +*/ + +/** \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) + +/** \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. +*/ +VK_DEFINE_HANDLE(VmaAllocation) + +/** \struct VmaDefragmentationContext +\brief Represents Opaque object that represents started defragmentation process. + +Fill structure #VmaDefragmentationInfo2 and call function vmaDefragmentationBegin() to create it. +Call function vmaDefragmentationEnd() to destroy it. +*/ +VK_DEFINE_HANDLE(VmaDefragmentationContext) + +/** @} */ + +/** +\addtogroup group_virtual +@{ +*/ + +/** \struct VmaVirtualAllocation +\brief Represents single memory allocation done inside VmaVirtualBlock. + +Use it as a unique identifier to virtual allocation within the single block. + +Use value `VK_NULL_HANDLE` to represent a null/invalid allocation. +*/ +VK_DEFINE_NON_DISPATCHABLE_HANDLE(VmaVirtualAllocation); + +/** @} */ + +/** +\addtogroup group_virtual +@{ +*/ + +/** \struct VmaVirtualBlock +\brief Handle to a virtual block object that allows to use core allocation algorithm without allocating any real GPU memory. + +Fill in #VmaVirtualBlockCreateInfo structure and use vmaCreateVirtualBlock() to create it. Use vmaDestroyVirtualBlock() to destroy it. +For more information, see documentation chapter \ref virtual_allocator. + +This object is not thread-safe - should not be used from multiple threads simultaneously, must be synchronized externally. +*/ +VK_DEFINE_HANDLE(VmaVirtualBlock) + +/** @} */ + +/** +\addtogroup group_init +@{ +*/ + +/// Callback function called after successful vkAllocateMemory. +typedef void (VKAPI_PTR* PFN_vmaAllocateDeviceMemoryFunction)( + VmaAllocator VMA_NOT_NULL allocator, + uint32_t memoryType, + VkDeviceMemory VMA_NOT_NULL_NON_DISPATCHABLE memory, + VkDeviceSize size, + void* VMA_NULLABLE pUserData); + +/// Callback function called before vkFreeMemory. +typedef void (VKAPI_PTR* PFN_vmaFreeDeviceMemoryFunction)( + VmaAllocator VMA_NOT_NULL allocator, + uint32_t memoryType, + VkDeviceMemory VMA_NOT_NULL_NON_DISPATCHABLE memory, + VkDeviceSize size, + void* VMA_NULLABLE pUserData); + +/** \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 VMA_NULLABLE pfnAllocate; + /// Optional, can be null. + PFN_vmaFreeDeviceMemoryFunction VMA_NULLABLE pfnFree; + /// Optional, can be null. + void* VMA_NULLABLE pUserData; +} VmaDeviceMemoryCallbacks; + /** \brief Pointers to some Vulkan functions - a subset used by the library. Used in VmaAllocatorCreateInfo::pVulkanFunctions. */ -typedef struct VmaVulkanFunctions { +typedef struct VmaVulkanFunctions +{ + /// Required when using VMA_DYNAMIC_VULKAN_FUNCTIONS. + PFN_vkGetInstanceProcAddr VMA_NULLABLE vkGetInstanceProcAddr; + /// Required when using VMA_DYNAMIC_VULKAN_FUNCTIONS. + PFN_vkGetDeviceProcAddr VMA_NULLABLE vkGetDeviceProcAddr; PFN_vkGetPhysicalDeviceProperties VMA_NULLABLE vkGetPhysicalDeviceProperties; PFN_vkGetPhysicalDeviceMemoryProperties VMA_NULLABLE vkGetPhysicalDeviceMemoryProperties; PFN_vkAllocateMemory VMA_NULLABLE vkAllocateMemory; @@ -2381,11 +929,15 @@ typedef struct VmaVulkanFunctions { PFN_vkDestroyImage VMA_NULLABLE vkDestroyImage; PFN_vkCmdCopyBuffer VMA_NULLABLE vkCmdCopyBuffer; #if VMA_DEDICATED_ALLOCATION || VMA_VULKAN_VERSION >= 1001000 + /// Fetch "vkGetBufferMemoryRequirements2" on Vulkan >= 1.1, fetch "vkGetBufferMemoryRequirements2KHR" when using VK_KHR_dedicated_allocation extension. PFN_vkGetBufferMemoryRequirements2KHR VMA_NULLABLE vkGetBufferMemoryRequirements2KHR; + /// Fetch "vkGetImageMemoryRequirements 2" on Vulkan >= 1.1, fetch "vkGetImageMemoryRequirements2KHR" when using VK_KHR_dedicated_allocation extension. PFN_vkGetImageMemoryRequirements2KHR VMA_NULLABLE vkGetImageMemoryRequirements2KHR; #endif #if VMA_BIND_MEMORY2 || VMA_VULKAN_VERSION >= 1001000 + /// Fetch "vkBindBufferMemory2" on Vulkan >= 1.1, fetch "vkBindBufferMemory2KHR" when using VK_KHR_bind_memory2 extension. PFN_vkBindBufferMemory2KHR VMA_NULLABLE vkBindBufferMemory2KHR; + /// Fetch "vkBindImageMemory2" on Vulkan >= 1.1, fetch "vkBindImageMemory2KHR" when using VK_KHR_bind_memory2 extension. PFN_vkBindImageMemory2KHR VMA_NULLABLE vkBindImageMemory2KHR; #endif #if VMA_MEMORY_BUDGET || VMA_VULKAN_VERSION >= 1001000 @@ -2393,34 +945,6 @@ typedef struct VmaVulkanFunctions { #endif } VmaVulkanFunctions; -/// Flags to be used in VmaRecordSettings::flags. -typedef enum VmaRecordFlagBits { - /** \brief Enables flush after recording every function call. - - Enable it if you expect your application to crash, which may leave recording file truncated. - It may degrade performance though. - */ - VMA_RECORD_FLUSH_AFTER_CALL_BIT = 0x00000001, - - VMA_RECORD_FLAG_BITS_MAX_ENUM = 0x7FFFFFFF -} VmaRecordFlagBits; -typedef VkFlags VmaRecordFlags; - -/// Parameters for recording calls to VMA functions. To be used in VmaAllocatorCreateInfo::pRecordSettings. -typedef struct VmaRecordSettings -{ - /// Flags for recording. Use #VmaRecordFlagBits enum. - VmaRecordFlags flags; - /** \brief Path to the file that should be written by the recording. - - Suggested extension: "csv". - If the file already exists, it will be overwritten. - It will be opened for the whole time #VmaAllocator object is alive. - If opening this file fails, creation of the whole allocator object fails. - */ - const char* VMA_NOT_NULL pFilePath; -} VmaRecordSettings; - /// Description of a Allocator to be created. typedef struct VmaAllocatorCreateInfo { @@ -2441,20 +965,6 @@ typedef struct VmaAllocatorCreateInfo /// Informative callbacks for `vkAllocateMemory`, `vkFreeMemory`. Optional. /** Optional, can be null. */ const VmaDeviceMemoryCallbacks* VMA_NULLABLE 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 @@ -2486,13 +996,6 @@ typedef struct VmaAllocatorCreateInfo For details see [Pointers to Vulkan functions](@ref config_Vulkan_functions). */ const VmaVulkanFunctions* VMA_NULLABLE pVulkanFunctions; - /** \brief Parameters for recording of VMA calls. Can be null. - - If not null, it enables recording of calls to VMA functions to a file. - If support for recording is not enabled using `VMA_RECORDING_ENABLED` macro, - creation of the allocator object fails with `VK_ERROR_FEATURE_NOT_PRESENT`. - */ - const VmaRecordSettings* VMA_NULLABLE pRecordSettings; /** \brief Handle to Vulkan instance object. Starting from version 3.0.0 this member is no longer optional, it must be set! @@ -2503,7 +1006,7 @@ typedef struct VmaAllocatorCreateInfo It must be a value in the format as created by macro `VK_MAKE_VERSION` or a constant like: `VK_API_VERSION_1_1`, `VK_API_VERSION_1_0`. The patch version number specified is ignored. Only the major and minor versions are considered. It must be less or equal (preferably equal) to value as passed to `vkCreateInstance` as `VkApplicationInfo::apiVersion`. - Only versions 1.0, 1.1, 1.2 are supported by the current implementation. + Only versions 1.0, 1.1, 1.2, 1.3 are supported by the current implementation. Leaving it initialized to zero is equivalent to `VK_API_VERSION_1_0`. */ uint32_t vulkanApiVersion; @@ -2521,17 +1024,7 @@ typedef struct VmaAllocatorCreateInfo #endif // #if VMA_EXTERNAL_MEMORY } VmaAllocatorCreateInfo; -/// Creates Allocator object. -VMA_CALL_PRE VkResult VMA_CALL_POST vmaCreateAllocator( - const VmaAllocatorCreateInfo* VMA_NOT_NULL pCreateInfo, - VmaAllocator VMA_NULLABLE * VMA_NOT_NULL pAllocator); - -/// Destroys allocator object. -VMA_CALL_PRE void VMA_CALL_POST vmaDestroyAllocator( - VmaAllocator VMA_NULLABLE allocator); - -/** \brief Information about existing #VmaAllocator object. -*/ +/// Information about existing #VmaAllocator object. typedef struct VmaAllocatorInfo { /** \brief Handle to Vulkan instance object. @@ -2551,54 +1044,14 @@ typedef struct VmaAllocatorInfo VkDevice VMA_NOT_NULL device; } VmaAllocatorInfo; -/** \brief Returns information about existing #VmaAllocator object - handle to Vulkan device etc. - -It might be useful if you want to keep just the #VmaAllocator handle and fetch other required handles to -`VkPhysicalDevice`, `VkDevice` etc. every time using this function. -*/ -VMA_CALL_PRE void VMA_CALL_POST vmaGetAllocatorInfo(VmaAllocator VMA_NOT_NULL allocator, VmaAllocatorInfo* VMA_NOT_NULL pAllocatorInfo); - -/** -PhysicalDeviceProperties are fetched from physicalDevice by the allocator. -You can access it here, without fetching it again on your own. -*/ -VMA_CALL_PRE void VMA_CALL_POST vmaGetPhysicalDeviceProperties( - VmaAllocator VMA_NOT_NULL allocator, - const VkPhysicalDeviceProperties* VMA_NULLABLE * VMA_NOT_NULL ppPhysicalDeviceProperties); - -/** -PhysicalDeviceMemoryProperties are fetched from physicalDevice by the allocator. -You can access it here, without fetching it again on your own. -*/ -VMA_CALL_PRE void VMA_CALL_POST vmaGetMemoryProperties( - VmaAllocator VMA_NOT_NULL allocator, - const VkPhysicalDeviceMemoryProperties* VMA_NULLABLE * VMA_NOT_NULL 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(). +\addtogroup group_stats +@{ */ -VMA_CALL_PRE void VMA_CALL_POST vmaGetMemoryTypeProperties( - VmaAllocator VMA_NOT_NULL allocator, - uint32_t memoryTypeIndex, - VkMemoryPropertyFlags* VMA_NOT_NULL 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. -*/ -VMA_CALL_PRE void VMA_CALL_POST vmaSetCurrentFrameIndex( - VmaAllocator VMA_NOT_NULL allocator, - uint32_t frameIndex); - -/** \brief Calculated statistics of memory usage in entire allocator. -*/ +/// Calculated statistics of memory usage in entire allocator. typedef struct VmaStatInfo { /// Number of `VkDeviceMemory` Vulkan memory blocks allocated. @@ -2623,21 +1076,7 @@ typedef struct VmaStats VmaStatInfo total; } VmaStats; -/** \brief Retrieves statistics from current state of the Allocator. - -This function is called "calculate" not "get" because it has to traverse all -internal data structures, so it may be quite slow. For faster but more brief statistics -suitable to be called every frame or every allocation, use vmaGetBudget(). - -Note that when using allocator from multiple threads, returned information may immediately -become outdated. -*/ -VMA_CALL_PRE void VMA_CALL_POST vmaCalculateStats( - VmaAllocator VMA_NOT_NULL allocator, - VmaStats* VMA_NOT_NULL pStats); - -/** \brief Statistics of current memory usage and available budget, in bytes, for specific memory heap. -*/ +/// Statistics of current memory usage and available budget, in bytes, for specific memory heap. typedef struct VmaBudget { /** \brief Sum size of all `VkDeviceMemory` blocks allocated from particular heap, in bytes. @@ -2649,9 +1088,6 @@ typedef struct VmaBudget Usually less or equal than `blockBytes`. Difference `blockBytes - allocationBytes` is the amount of memory allocated but unused - available for new allocations or wasted due to fragmentation. - - It might be greater than `blockBytes` if there are some allocations in lost state, as they account - to this value as well. */ VkDeviceSize allocationBytes; @@ -2677,278 +1113,496 @@ typedef struct VmaBudget VkDeviceSize budget; } VmaBudget; -/** \brief Retrieves information about current memory budget for all memory heaps. - -\param[out] pBudget Must point to array with number of elements at least equal to number of memory heaps in physical device used. - -This function is called "get" not "calculate" because it is very fast, suitable to be called -every frame or every allocation. For more detailed statistics use vmaCalculateStats(). +/** @} */ -Note that when using allocator from multiple threads, returned information may immediately -become outdated. +/** +\addtogroup group_alloc +@{ */ -VMA_CALL_PRE void VMA_CALL_POST vmaGetBudget( - VmaAllocator VMA_NOT_NULL allocator, - VmaBudget* VMA_NOT_NULL pBudget); -#ifndef VMA_STATS_STRING_ENABLED -#define VMA_STATS_STRING_ENABLED 1 -#endif +typedef struct VmaAllocationCreateInfo +{ + /// Use #VmaAllocationCreateFlagBits enum. + VmaAllocationCreateFlags flags; + /** \brief Intended usage of memory. -#if VMA_STATS_STRING_ENABLED + You can leave #VMA_MEMORY_USAGE_UNKNOWN if you specify memory requirements in other way. \n + */ + VmaMemoryUsage usage; + /** \brief Flags that must be set in a Memory Type chosen for an allocation. -/// Builds and returns statistics as string in JSON format. -/** @param[out] ppStatsString Must be freed using vmaFreeStatsString() function. -*/ -VMA_CALL_PRE void VMA_CALL_POST vmaBuildStatsString( - VmaAllocator VMA_NOT_NULL allocator, - char* VMA_NULLABLE * VMA_NOT_NULL ppStatsString, - VkBool32 detailedMap); + Leave 0 if you specify memory requirements in other way.*/ + VkMemoryPropertyFlags requiredFlags; + /** \brief Flags that preferably should be set in a memory type chosen for an allocation. -VMA_CALL_PRE void VMA_CALL_POST vmaFreeStatsString( - VmaAllocator VMA_NOT_NULL allocator, - char* VMA_NULLABLE pStatsString); + Set to 0 if no additional flags are preferred.*/ + VkMemoryPropertyFlags preferredFlags; + /** \brief Bitmask containing one bit set for every memory type acceptable for this allocation. -#endif // #if VMA_STATS_STRING_ENABLED + 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 + */ + uint32_t memoryTypeBits; + /** \brief Pool that this allocation should be created in. -/** \struct VmaPool -\brief Represents custom memory pool + Leave `VK_NULL_HANDLE` to allocate from default pool. + */ + VmaPool VMA_NULLABLE pool; + /** \brief Custom general-purpose pointer that will be stored in #VmaAllocation, can be read as VmaAllocationInfo::pUserData and changed using vmaSetAllocationUserData(). -Fill structure VmaPoolCreateInfo and call function vmaCreatePool() to create it. -Call function vmaDestroyPool() to destroy it. + 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* VMA_NULLABLE pUserData; + /** \brief A floating-point value between 0 and 1, indicating the priority of the allocation relative to other memory allocations. -For more information see [Custom memory pools](@ref choosing_memory_type_custom_memory_pools). -*/ -VK_DEFINE_HANDLE(VmaPool) + It is used only when #VMA_ALLOCATOR_CREATE_EXT_MEMORY_PRIORITY_BIT flag was used during creation of the #VmaAllocator object + and this allocation ends up as dedicated or is explicitly forced as dedicated using #VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT. + Otherwise, it has the priority of a memory block where it is placed and this variable is ignored. + */ + float priority; +} VmaAllocationCreateInfo; -typedef enum VmaMemoryUsage +/// Describes parameter of created #VmaPool. +typedef struct VmaPoolCreateInfo { - /** No intended memory usage specified. - Use other members of VmaAllocationCreateInfo to specify your requirements. + /** \brief Use combination of #VmaPoolCreateFlagBits. */ - 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: + VmaPoolCreateFlags flags; + /** \brief Size of a single `VkDeviceMemory` block to be allocated as part of this pool, in bytes. Optional. - - 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 on GPU. + Specify nonzero to set explicit, constant size of memory blocks used by this + pool. - 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. + Leave 0 to use default and let the library manage block sizes automatically. + Sizes of particular blocks may vary. + In this case, the pool will also support dedicated allocations. */ - VMA_MEMORY_USAGE_GPU_ONLY = 1, - /** Memory will be mappable on host. - It usually means CPU (system) memory. - Guarantees to be `HOST_VISIBLE` and `HOST_COHERENT`. - CPU access is typically uncached. Writes may be write-combined. - Resources created in this pool may still be accessible to the device, but access to them can be slow. - It is roughly equivalent of `D3D12_HEAP_TYPE_UPLOAD`. + VkDeviceSize blockSize; + /** \brief Minimum number of blocks to be always allocated in this pool, even if they stay empty. - Usage: Staging copy of resources used as transfer source. + Set to 0 to have no preallocated blocks and allow the pool be completely empty. */ - 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 access is typically uncached. Writes may be write-combined. + size_t minBlockCount; + /** \brief Maximum number of blocks that can be allocated in this pool. Optional. - Usage: Resources written frequently by host (dynamic), read by device. E.g. textures (with LINEAR layout), vertex buffers, uniform buffers updated every frame or every draw call. + Set to 0 to use default, which is `SIZE_MAX`, which means no limit. + + Set to same value as VmaPoolCreateInfo::minBlockCount to have fixed amount of memory allocated + throughout whole lifetime of this pool. */ - 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`. + size_t maxBlockCount; + /** \brief A floating-point value between 0 and 1, indicating the priority of the allocations in this pool relative to other memory allocations. - Usage: + It is used only when #VMA_ALLOCATOR_CREATE_EXT_MEMORY_PRIORITY_BIT flag was used during creation of the #VmaAllocator object. + Otherwise, this variable is ignored. + */ + float priority; + /** \brief Additional minimum alignment to be used for all allocations created from this pool. Can be 0. - - 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. + Leave 0 (default) not to impose any additional alignment. If not 0, it must be a power of two. + It can be useful in cases where alignment returned by Vulkan by functions like `vkGetBufferMemoryRequirements` is not enough, + e.g. when doing interop with OpenGL. */ - VMA_MEMORY_USAGE_GPU_TO_CPU = 4, - /** CPU memory - memory that is preferably not `DEVICE_LOCAL`, but also not guaranteed to be `HOST_VISIBLE`. + VkDeviceSize minAllocationAlignment; + /** \brief Additional `pNext` chain to be attached to `VkMemoryAllocateInfo` used for every allocation made by this pool. Optional. - Usage: Staging copy of resources moved from GPU memory to CPU memory as part - of custom paging/residency mechanism, to be moved back to GPU memory when needed. + Optional, can be null. If not null, it must point to a `pNext` chain of structures that can be attached to `VkMemoryAllocateInfo`. + It can be useful for special needs such as adding `VkExportMemoryAllocateInfoKHR`. + Structures pointed by this member must remain alive and unchanged for the whole lifetime of the custom pool. + + Please note that some structures, e.g. `VkMemoryPriorityAllocateInfoEXT`, `VkMemoryDedicatedAllocateInfoKHR`, + can be attached automatically by this library when using other, more convenient of its features. */ - VMA_MEMORY_USAGE_CPU_COPY = 5, - /** Lazily allocated GPU memory having `VK_MEMORY_PROPERTY_LAZILY_ALLOCATED_BIT`. - Exists mostly on mobile platforms. Using it on desktop PC or other GPUs with no such memory type present will fail the allocation. + void* VMA_NULLABLE pMemoryAllocateNext; +} VmaPoolCreateInfo; - Usage: Memory for transient attachment images (color attachments, depth attachments etc.), created with `VK_IMAGE_USAGE_TRANSIENT_ATTACHMENT_BIT`. +/** @} */ - Allocations with this usage are always created as dedicated - it implies #VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT. +/** +\addtogroup group_stats +@{ +*/ + +/// Describes parameter of existing #VmaPool. +typedef struct VmaPoolStats +{ + /** \brief Total amount of `VkDeviceMemory` allocated from Vulkan for this pool, in bytes. */ - VMA_MEMORY_USAGE_GPU_LAZILY_ALLOCATED = 6, + 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. + */ + size_t allocationCount; + /** \brief Number of continuous memory ranges in the pool not used by any #VmaAllocation. + */ + size_t unusedRangeCount; + /** \brief Number of `VkDeviceMemory` blocks allocated for this pool. + */ + size_t blockCount; +} VmaPoolStats; - 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. +/** +\addtogroup group_alloc +@{ +*/ - Use it for special, big resources, like fullscreen images used as attachments. +/// Parameters of #VmaAllocation objects, that can be retrieved using function vmaGetAllocationInfo(). +typedef struct VmaAllocationInfo +{ + /** \brief Memory type index that this allocation was allocated from. - You should not use this flag if VmaAllocationCreateInfo::pool is not null. + It never changes. */ - 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. + uint32_t memoryType; + /** \brief Handle to Vulkan memory object. - If new allocation cannot be placed in any of the existing blocks, allocation - fails with `VK_ERROR_OUT_OF_DEVICE_MEMORY` error. + Same memory object can be shared by multiple allocations. - 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. + It can change after call to vmaDefragment() if this allocation is passed to the function. + */ + VkDeviceMemory VMA_NULLABLE_NON_DISPATCHABLE deviceMemory; + /** \brief Offset in `VkDeviceMemory` object to the beginning of this allocation, in bytes. `(deviceMemory, offset)` pair is unique to this allocation. - 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. + You usually don't need to use this offset. If you create a buffer or an image together with the allocation using e.g. function + vmaCreateBuffer(), vmaCreateImage(), functions that operate on these resources refer to the beginning of the buffer or image, + not entire device memory block. Functions like vmaMapMemory(), vmaBindBufferMemory() also refer to the beginning of the allocation + and apply this offset automatically. - Pointer to mapped memory will be returned through VmaAllocationInfo::pMappedData. + It can change after call to vmaDefragment() if this allocation is passed to the function. + */ + VkDeviceSize offset; + /** \brief Size of this allocation, in bytes. - It is 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). + It never changes. - You should not use this flag together with #VMA_ALLOCATION_CREATE_CAN_BECOME_LOST_BIT. + \note Allocation size returned in this variable may be greater than the size + requested for the resource e.g. as `VkBufferCreateInfo::size`. Whole size of the + allocation is accessible for operations on memory e.g. using a pointer after + mapping with vmaMapMemory(), but operations on the resource e.g. using + `vkCmdCopyBuffer` must be limited to the size of the resource. */ - 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. + VkDeviceSize size; + /** \brief Pointer to the beginning of this allocation as mapped data. - To check if allocation is not lost, call vmaGetAllocationInfo() and check if - VmaAllocationInfo::deviceMemory is not `VK_NULL_HANDLE`. + If the allocation hasn't been mapped using vmaMapMemory() and hasn't been + created with #VMA_ALLOCATION_CREATE_MAPPED_BIT flag, this value is null. - For details about supporting lost allocations, see Lost Allocations - chapter of User Guide on Main Page. + 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* VMA_NULLABLE pMappedData; + /** \brief Custom general-purpose pointer that was passed as VmaAllocationCreateInfo::pUserData or set using vmaSetAllocationUserData(). - You should not use this flag together with #VMA_ALLOCATION_CREATE_MAPPED_BIT. + It can change after call to vmaSetAllocationUserData() for this allocation. */ - 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. + void* VMA_NULLABLE pUserData; +} VmaAllocationInfo; + +/** \brief Parameters for defragmentation. - For details about supporting lost allocations, see Lost Allocations - chapter of User Guide on Main Page. +To be used with function vmaDefragmentationBegin(). +*/ +typedef struct VmaDefragmentationInfo2 +{ + /** \brief Reserved for future use. Should be 0. */ - 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(). + VmaDefragmentationFlags flags; + /** \brief Number of allocations in `pAllocations` array. */ - VMA_ALLOCATION_CREATE_USER_DATA_COPY_STRING_BIT = 0x00000020, - /** Allocation will be created from upper stack in a double stack pool. + uint32_t allocationCount; + /** \brief Pointer to array of allocations that can be defragmented. - This flag is only allowed for custom pools created with #VMA_POOL_CREATE_LINEAR_ALGORITHM_BIT flag. + The array should have `allocationCount` elements. + The array should not contain nulls. + Elements in the array should be unique - same allocation cannot occur twice. + All allocations not present in this array are considered non-moveable during this defragmentation. */ - VMA_ALLOCATION_CREATE_UPPER_ADDRESS_BIT = 0x00000040, - /** Create both buffer/image and allocation, but don't bind them together. - It is useful when you want to bind yourself to do some more advanced binding, e.g. using some extensions. - The flag is meaningful only with functions that bind by default: vmaCreateBuffer(), vmaCreateImage(). - Otherwise it is ignored. + const VmaAllocation VMA_NOT_NULL* VMA_NULLABLE VMA_LEN_IF_NOT_NULL(allocationCount) pAllocations; + /** \brief Optional, output. Pointer to array that will be filled with information whether the allocation at certain index has been changed during defragmentation. + + The array should have `allocationCount` elements. + You can pass null if you are not interested in this information. */ - VMA_ALLOCATION_CREATE_DONT_BIND_BIT = 0x00000080, - /** Create allocation only if additional device memory required for it, if any, won't exceed - memory budget. Otherwise return `VK_ERROR_OUT_OF_DEVICE_MEMORY`. + VkBool32* VMA_NULLABLE VMA_LEN_IF_NOT_NULL(allocationCount) pAllocationsChanged; + /** \brief Numer of pools in `pPools` array. */ - VMA_ALLOCATION_CREATE_WITHIN_BUDGET_BIT = 0x00000100, + uint32_t poolCount; + /** \brief Either null or pointer to array of pools to be defragmented. + + All the allocations in the specified pools can be moved during defragmentation + and there is no way to check if they were really moved as in `pAllocationsChanged`, + so you must query all the allocations in all these pools for new `VkDeviceMemory` + and offset using vmaGetAllocationInfo() if you might need to recreate buffers + and images bound to them. + + The array should have `poolCount` elements. + The array should not contain nulls. + Elements in the array should be unique - same pool cannot occur twice. - /** Allocation strategy that chooses smallest possible free range for the - allocation. + Using this array is equivalent to specifying all allocations from the pools in `pAllocations`. + It might be more efficient. */ - VMA_ALLOCATION_CREATE_STRATEGY_BEST_FIT_BIT = 0x00010000, - /** Allocation strategy that chooses biggest possible free range for the - allocation. + const VmaPool VMA_NOT_NULL* VMA_NULLABLE VMA_LEN_IF_NOT_NULL(poolCount) pPools; + /** \brief Maximum total numbers of bytes that can be copied while moving allocations to different places using transfers on CPU side, like `memcpy()`, `memmove()`. + + `VK_WHOLE_SIZE` means no limit. */ - VMA_ALLOCATION_CREATE_STRATEGY_WORST_FIT_BIT = 0x00020000, - /** Allocation strategy that chooses first suitable free range for the - allocation. + VkDeviceSize maxCpuBytesToMove; + /** \brief Maximum number of allocations that can be moved to a different place using transfers on CPU side, like `memcpy()`, `memmove()`. - "First" doesn't necessarily means the one with smallest offset in memory, - but rather the one that is easiest and fastest to find. + `UINT32_MAX` means no limit. */ - VMA_ALLOCATION_CREATE_STRATEGY_FIRST_FIT_BIT = 0x00040000, + uint32_t maxCpuAllocationsToMove; + /** \brief Maximum total numbers of bytes that can be copied while moving allocations to different places using transfers on GPU side, posted to `commandBuffer`. - /** Allocation strategy that tries to minimize memory usage. + `VK_WHOLE_SIZE` means no limit. */ - VMA_ALLOCATION_CREATE_STRATEGY_MIN_MEMORY_BIT = VMA_ALLOCATION_CREATE_STRATEGY_BEST_FIT_BIT, - /** Allocation strategy that tries to minimize allocation time. + VkDeviceSize maxGpuBytesToMove; + /** \brief Maximum number of allocations that can be moved to a different place using transfers on GPU side, posted to `commandBuffer`. + + `UINT32_MAX` means no limit. */ - VMA_ALLOCATION_CREATE_STRATEGY_MIN_TIME_BIT = VMA_ALLOCATION_CREATE_STRATEGY_FIRST_FIT_BIT, - /** Allocation strategy that tries to minimize memory fragmentation. + uint32_t maxGpuAllocationsToMove; + /** \brief Optional. Command buffer where GPU copy commands will be posted. + + If not null, it must be a valid command buffer handle that supports Transfer queue type. + It must be in the recording state and outside of a render pass instance. + You need to submit it and make sure it finished execution before calling vmaDefragmentationEnd(). + + Passing null means that only CPU defragmentation will be performed. */ - VMA_ALLOCATION_CREATE_STRATEGY_MIN_FRAGMENTATION_BIT = VMA_ALLOCATION_CREATE_STRATEGY_WORST_FIT_BIT, + VkCommandBuffer VMA_NULLABLE commandBuffer; +} VmaDefragmentationInfo2; - /** A bit mask to extract only `STRATEGY` bits from entire set of flags. +typedef struct VmaDefragmentationPassMoveInfo +{ + VmaAllocation VMA_NOT_NULL allocation; + VkDeviceMemory VMA_NOT_NULL_NON_DISPATCHABLE memory; + VkDeviceSize offset; +} VmaDefragmentationPassMoveInfo; + +/** \brief Parameters for incremental defragmentation steps. + +To be used with function vmaBeginDefragmentationPass(). +*/ +typedef struct VmaDefragmentationPassInfo +{ + uint32_t moveCount; + VmaDefragmentationPassMoveInfo* VMA_NOT_NULL VMA_LEN_IF_NOT_NULL(moveCount) pMoves; +} VmaDefragmentationPassInfo; + +/** \brief Deprecated. Optional configuration parameters to be passed to function vmaDefragment(). + +\deprecated This is a part of the old interface. It is recommended to use structure #VmaDefragmentationInfo2 and function vmaDefragmentationBegin() instead. +*/ +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. */ - VMA_ALLOCATION_CREATE_STRATEGY_MASK = - VMA_ALLOCATION_CREATE_STRATEGY_BEST_FIT_BIT | - VMA_ALLOCATION_CREATE_STRATEGY_WORST_FIT_BIT | - VMA_ALLOCATION_CREATE_STRATEGY_FIRST_FIT_BIT, + VkDeviceSize maxBytesToMove; + /** \brief Maximum number of allocations that can be moved to different place. - VMA_ALLOCATION_CREATE_FLAG_BITS_MAX_ENUM = 0x7FFFFFFF -} VmaAllocationCreateFlagBits; -typedef VkFlags VmaAllocationCreateFlags; + Default is `UINT32_MAX`, which means no limit. + */ + uint32_t maxAllocationsToMove; +} VmaDefragmentationInfo; -typedef struct VmaAllocationCreateInfo +/// Statistics returned by function vmaDefragment(). +typedef struct VmaDefragmentationStats { - /// Use #VmaAllocationCreateFlagBits enum. - VmaAllocationCreateFlags flags; - /** \brief Intended usage of memory. + /// 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; - 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. +/** @} */ + +/** +\addtogroup group_virtual +@{ +*/ + +/// Parameters of created #VmaVirtualBlock object to be passed to vmaCreateVirtualBlock(). +typedef struct VmaVirtualBlockCreateInfo +{ + /** \brief Total size of the virtual block. + + Sizes can be expressed in bytes or any units you want as long as you are consistent in using them. + For example, if you allocate from some array of structures, 1 can mean single instance of entire structure. */ - VmaMemoryUsage usage; - /** \brief Flags that must be set in a Memory Type chosen for an allocation. + VkDeviceSize size; - 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. + /** \brief Use combination of #VmaVirtualBlockCreateFlagBits. + */ + VmaVirtualBlockCreateFlags flags; - Set to 0 if no additional flags are preferred. \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. + /** \brief Custom CPU memory allocation callbacks. Optional. - 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. + Optional, can be null. When specified, they will be used for all CPU-side memory allocations. */ - uint32_t memoryTypeBits; - /** \brief Pool that this allocation should be created in. + const VkAllocationCallbacks* VMA_NULLABLE pAllocationCallbacks; +} VmaVirtualBlockCreateInfo; - Leave `VK_NULL_HANDLE` to allocate from default pool. If not null, members: - `usage`, `requiredFlags`, `preferredFlags`, `memoryTypeBits` are ignored. +/// Parameters of created virtual allocation to be passed to vmaVirtualAllocate(). +typedef struct VmaVirtualAllocationCreateInfo +{ + /** \brief Size of the allocation. + + Cannot be zero. */ - VmaPool VMA_NULLABLE pool; - /** \brief Custom general-purpose pointer that will be stored in #VmaAllocation, can be read as VmaAllocationInfo::pUserData and changed using vmaSetAllocationUserData(). + VkDeviceSize size; + /** \brief Required alignment of the allocation. Optional. - 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. + Must be power of two. Special value 0 has the same meaning as 1 - means no special alignment is required, so allocation can start at any offset. + */ + VkDeviceSize alignment; + /** \brief Use combination of #VmaVirtualAllocationCreateFlagBits. + */ + VmaVirtualAllocationCreateFlags flags; + /** \brief Custom pointer to be associated with the allocation. Optional. + + It can be any value and can be used for user-defined purposes. It can be fetched or changed later. */ void* VMA_NULLABLE pUserData; - /** \brief A floating-point value between 0 and 1, indicating the priority of the allocation relative to other memory allocations. +} VmaVirtualAllocationCreateInfo; - It is used only when #VMA_ALLOCATOR_CREATE_EXT_MEMORY_PRIORITY_BIT flag was used during creation of the #VmaAllocator object - and this allocation ends up as dedicated or is explicitly forced as dedicated using #VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT. - Otherwise, it has the priority of a memory block where it is placed and this variable is ignored. +/// Parameters of an existing virtual allocation, returned by vmaGetVirtualAllocationInfo(). +typedef struct VmaVirtualAllocationInfo +{ + /** \brief Offset of the allocation. + + Offset at which the allocation was made. */ - float priority; -} VmaAllocationCreateInfo; + VkDeviceSize offset; + /** \brief Size of the allocation. + + Same value as passed in VmaVirtualAllocationCreateInfo::size. + */ + VkDeviceSize size; + /** \brief Custom pointer associated with the allocation. + + Same value as passed in VmaVirtualAllocationCreateInfo::pUserData or to vmaSetVirtualAllocationUserData(). + */ + void* VMA_NULLABLE pUserData; +} VmaVirtualAllocationInfo; + +/** @} */ + +#endif // _VMA_DATA_TYPES_DECLARATIONS + +#ifndef _VMA_FUNCTION_HEADERS + +/** +\addtogroup group_init +@{ +*/ + +/// Creates #VmaAllocator object. +VMA_CALL_PRE VkResult VMA_CALL_POST vmaCreateAllocator( + const VmaAllocatorCreateInfo* VMA_NOT_NULL pCreateInfo, + VmaAllocator VMA_NULLABLE* VMA_NOT_NULL pAllocator); + +/// Destroys allocator object. +VMA_CALL_PRE void VMA_CALL_POST vmaDestroyAllocator( + VmaAllocator VMA_NULLABLE allocator); + +/** \brief Returns information about existing #VmaAllocator object - handle to Vulkan device etc. + +It might be useful if you want to keep just the #VmaAllocator handle and fetch other required handles to +`VkPhysicalDevice`, `VkDevice` etc. every time using this function. +*/ +VMA_CALL_PRE void VMA_CALL_POST vmaGetAllocatorInfo( + VmaAllocator VMA_NOT_NULL allocator, + VmaAllocatorInfo* VMA_NOT_NULL pAllocatorInfo); + +/** +PhysicalDeviceProperties are fetched from physicalDevice by the allocator. +You can access it here, without fetching it again on your own. +*/ +VMA_CALL_PRE void VMA_CALL_POST vmaGetPhysicalDeviceProperties( + VmaAllocator VMA_NOT_NULL allocator, + const VkPhysicalDeviceProperties* VMA_NULLABLE* VMA_NOT_NULL ppPhysicalDeviceProperties); + +/** +PhysicalDeviceMemoryProperties are fetched from physicalDevice by the allocator. +You can access it here, without fetching it again on your own. +*/ +VMA_CALL_PRE void VMA_CALL_POST vmaGetMemoryProperties( + VmaAllocator VMA_NOT_NULL allocator, + const VkPhysicalDeviceMemoryProperties* VMA_NULLABLE* VMA_NOT_NULL 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(). +*/ +VMA_CALL_PRE void VMA_CALL_POST vmaGetMemoryTypeProperties( + VmaAllocator VMA_NOT_NULL allocator, + uint32_t memoryTypeIndex, + VkMemoryPropertyFlags* VMA_NOT_NULL pFlags); + +/** \brief Sets index of the current frame. +*/ +VMA_CALL_PRE void VMA_CALL_POST vmaSetCurrentFrameIndex( + VmaAllocator VMA_NOT_NULL allocator, + uint32_t frameIndex); + +/** @} */ + +/** +\addtogroup group_stats +@{ +*/ + +/** \brief Retrieves statistics from current state of the Allocator. + +This function is called "calculate" not "get" because it has to traverse all +internal data structures, so it may be quite slow. For faster but more brief statistics +suitable to be called every frame or every allocation, use vmaGetHeapBudgets(). + +Note that when using allocator from multiple threads, returned information may immediately +become outdated. +*/ +VMA_CALL_PRE void VMA_CALL_POST vmaCalculateStats( + VmaAllocator VMA_NOT_NULL allocator, + VmaStats* VMA_NOT_NULL pStats); + +/** \brief Retrieves information about current memory budget for all memory heaps. + +\param allocator +\param[out] pBudgets Must point to array with number of elements at least equal to number of memory heaps in physical device used. + +This function is called "get" not "calculate" because it is very fast, suitable to be called +every frame or every allocation. For more detailed statistics use vmaCalculateStats(). + +Note that when using allocator from multiple threads, returned information may immediately +become outdated. +*/ +VMA_CALL_PRE void VMA_CALL_POST vmaGetHeapBudgets( + VmaAllocator VMA_NOT_NULL allocator, + VmaBudget* VMA_NOT_NULL VMA_LEN_IF_NOT_NULL("VkPhysicalDeviceMemoryProperties::memoryHeapCount") pBudgets); + +/** @} */ + +/** +\addtogroup group_alloc +@{ +*/ /** \brief Helps to find memoryTypeIndex, given memoryTypeBits and VmaAllocationCreateInfo. @@ -3008,172 +1662,16 @@ VMA_CALL_PRE VkResult VMA_CALL_POST vmaFindMemoryTypeIndexForImageInfo( const VmaAllocationCreateInfo* VMA_NOT_NULL pAllocationCreateInfo, uint32_t* VMA_NOT_NULL 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 an 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 - faster and more optimal. - */ - VMA_POOL_CREATE_IGNORE_BUFFER_IMAGE_GRANULARITY_BIT = 0x00000002, - - /** \brief Enables alternative, linear allocation algorithm in this pool. - - Specify this flag to enable linear allocation algorithm, which always creates - new allocations after last one and doesn't reuse space from allocations freed in - between. It trades memory consumption for simplified algorithm and data - structure, which has better performance and uses less memory for metadata. - - By using this flag, you can achieve behavior of free-at-once, stack, - ring buffer, and double stack. For details, see documentation chapter - \ref linear_algorithm. - - When using this flag, you must specify VmaPoolCreateInfo::maxBlockCount == 1 (or 0 for default). - - For more details, see [Linear allocation algorithm](@ref linear_algorithm). - */ - VMA_POOL_CREATE_LINEAR_ALGORITHM_BIT = 0x00000004, - - /** \brief Enables alternative, buddy allocation algorithm in this pool. - - It operates on a tree of blocks, each having size that is a power of two and - a half of its parent's size. Comparing to default algorithm, this one provides - faster allocation and deallocation and decreased external fragmentation, - at the expense of more memory wasted (internal fragmentation). - - For more details, see [Buddy allocation algorithm](@ref buddy_algorithm). - */ - VMA_POOL_CREATE_BUDDY_ALGORITHM_BIT = 0x00000008, - - /** Bit mask to extract only `ALGORITHM` bits from entire set of flags. - */ - VMA_POOL_CREATE_ALGORITHM_MASK = - VMA_POOL_CREATE_LINEAR_ALGORITHM_BIT | - VMA_POOL_CREATE_BUDDY_ALGORITHM_BIT, - - 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. - - Specify nonzero to set explicit, constant size of memory blocks used by this - pool. - - Leave 0 to use default and let the library manage block sizes automatically. - Sizes of particular blocks may vary. - */ - 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 allow the pool be completely empty. - */ - size_t minBlockCount; - /** \brief Maximum number of blocks that can be allocated in this pool. Optional. - - Set to 0 to use default, which is `SIZE_MAX`, which means no limit. - - Set to same value as VmaPoolCreateInfo::minBlockCount to have fixed amount of memory allocated - throughout whole lifetime of this pool. - */ - 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; - /** \brief A floating-point value between 0 and 1, indicating the priority of the allocations in this pool relative to other memory allocations. - - It is used only when #VMA_ALLOCATOR_CREATE_EXT_MEMORY_PRIORITY_BIT flag was used during creation of the #VmaAllocator object. - Otherwise, this variable is ignored. - */ - float priority; - /** \brief Additional minimum alignment to be used for all allocations created from this pool. Can be 0. - - Leave 0 (default) not to impose any additional alignment. If not 0, it must be a power of two. - It can be useful in cases where alignment returned by Vulkan by functions like `vkGetBufferMemoryRequirements` is not enough, - e.g. when doing interop with OpenGL. - */ - VkDeviceSize minAllocationAlignment; - /** \brief Additional `pNext` chain to be attached to `VkMemoryAllocateInfo` used for every allocation made by this pool. Optional. - - Optional, can be null. If not null, it must point to a `pNext` chain of structures that can be attached to `VkMemoryAllocateInfo`. - It can be useful for special needs such as adding `VkExportMemoryAllocateInfoKHR`. - Structures pointed by this member must remain alive and unchanged for the whole lifetime of the custom pool. - - Please note that some structures, e.g. `VkMemoryPriorityAllocateInfoEXT`, `VkMemoryDedicatedAllocateInfoKHR`, - can be attached automatically by this library when using other, more convenient of its features. - */ - void* VMA_NULLABLE pMemoryAllocateNext; -} 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 available for new allocation. - - 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; - /** \brief Number of `VkDeviceMemory` blocks allocated for this pool. - */ - size_t blockCount; -} 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. +\param allocator Allocator object. +\param pCreateInfo Parameters of pool to create. +\param[out] pPool Handle to created pool. */ VMA_CALL_PRE VkResult VMA_CALL_POST vmaCreatePool( VmaAllocator VMA_NOT_NULL allocator, const VmaPoolCreateInfo* VMA_NOT_NULL pCreateInfo, - VmaPool VMA_NULLABLE * VMA_NOT_NULL pPool); + VmaPool VMA_NULLABLE* VMA_NOT_NULL pPool); /** \brief Destroys #VmaPool object and frees Vulkan device memory. */ @@ -3181,27 +1679,30 @@ VMA_CALL_PRE void VMA_CALL_POST vmaDestroyPool( VmaAllocator VMA_NOT_NULL allocator, VmaPool VMA_NULLABLE pool); +/** @} */ + +/** +\addtogroup group_stats +@{ +*/ + /** \brief Retrieves statistics of existing #VmaPool object. -@param allocator Allocator object. -@param pool Pool object. -@param[out] pPoolStats Statistics of specified pool. +\param allocator Allocator object. +\param pool Pool object. +\param[out] pPoolStats Statistics of specified pool. */ VMA_CALL_PRE void VMA_CALL_POST vmaGetPoolStats( VmaAllocator VMA_NOT_NULL allocator, VmaPool VMA_NOT_NULL pool, VmaPoolStats* VMA_NOT_NULL 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. +/** +\addtogroup group_alloc +@{ */ -VMA_CALL_PRE void VMA_CALL_POST vmaMakePoolAllocationsLost( - VmaAllocator VMA_NOT_NULL allocator, - VmaPool VMA_NOT_NULL pool, - size_t* VMA_NULLABLE pLostAllocationCount); /** \brief Checks magic number in margins around all allocations in given memory pool in search for corruptions. @@ -3213,11 +1714,13 @@ Possible return values: - `VK_ERROR_FEATURE_NOT_PRESENT` - corruption detection is not enabled for specified pool. - `VK_SUCCESS` - corruption detection has been performed and succeeded. -- `VK_ERROR_VALIDATION_FAILED_EXT` - corruption detection has been performed and found memory corruptions around one of the allocations. +- `VK_ERROR_UNKNOWN` - corruption detection has been performed and found memory corruptions around one of the allocations. `VMA_ASSERT` is also fired in that case. - Other value: Error returned by Vulkan, e.g. memory mapping failure. */ -VMA_CALL_PRE VkResult VMA_CALL_POST vmaCheckPoolCorruption(VmaAllocator VMA_NOT_NULL allocator, VmaPool VMA_NOT_NULL pool); +VMA_CALL_PRE VkResult VMA_CALL_POST vmaCheckPoolCorruption( + VmaAllocator VMA_NOT_NULL allocator, + VmaPool VMA_NOT_NULL pool); /** \brief Retrieves name of a custom pool. @@ -3228,7 +1731,7 @@ destroyed or its name is changed using vmaSetPoolName(). VMA_CALL_PRE void VMA_CALL_POST vmaGetPoolName( VmaAllocator VMA_NOT_NULL allocator, VmaPool VMA_NOT_NULL pool, - const char* VMA_NULLABLE * VMA_NOT_NULL ppName); + const char* VMA_NULLABLE* VMA_NOT_NULL ppName); /** \brief Sets name of a custom pool. @@ -3240,90 +1743,13 @@ VMA_CALL_PRE void VMA_CALL_POST vmaSetPoolName( VmaPool VMA_NOT_NULL pool, const char* VMA_NULLABLE pName); -/** \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 VMA_NULLABLE_NON_DISPATCHABLE deviceMemory; - /** \brief Offset in `VkDeviceMemory` object to the beginning of this allocation, in bytes. `(deviceMemory, offset)` pair is unique to this allocation. - - You usually don't need to use this offset. If you create a buffer or an image together with the allocation using e.g. function - vmaCreateBuffer(), vmaCreateImage(), functions that operate on these resources refer to the beginning of the buffer or image, - not entire device memory block. Functions like vmaMapMemory(), vmaBindBufferMemory() also refer to the beginning of the allocation - and apply this offset automatically. - - It can change after call to vmaDefragment() if this allocation is passed to the function, or if allocation is lost. - */ - VkDeviceSize offset; - /** \brief Size of this allocation, in bytes. - - It never changes, unless allocation is lost. - - \note Allocation size returned in this variable may be greater than the size - requested for the resource e.g. as `VkBufferCreateInfo::size`. Whole size of the - allocation is accessible for operations on memory e.g. using a pointer after - mapping with vmaMapMemory(), but operations on the resource e.g. using - `vkCmdCopyBuffer` must be limited to the size of the resource. - */ - 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 is 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* VMA_NULLABLE 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* VMA_NULLABLE 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(). +\param allocator +\param pVkMemoryRequirements +\param pCreateInfo +\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() or vmaFreeMemoryPages(). @@ -3334,17 +1760,17 @@ VMA_CALL_PRE VkResult VMA_CALL_POST vmaAllocateMemory( VmaAllocator VMA_NOT_NULL allocator, const VkMemoryRequirements* VMA_NOT_NULL pVkMemoryRequirements, const VmaAllocationCreateInfo* VMA_NOT_NULL pCreateInfo, - VmaAllocation VMA_NULLABLE * VMA_NOT_NULL pAllocation, + VmaAllocation VMA_NULLABLE* VMA_NOT_NULL pAllocation, VmaAllocationInfo* VMA_NULLABLE pAllocationInfo); /** \brief General purpose memory allocation for multiple allocation objects at once. -@param allocator Allocator object. -@param pVkMemoryRequirements Memory requirements for each allocation. -@param pCreateInfo Creation parameters for each alloction. -@param allocationCount Number of allocations to make. -@param[out] pAllocations Pointer to array that will be filled with handles to created allocations. -@param[out] pAllocationInfo Optional. Pointer to array that will be filled with parameters of created allocations. +\param allocator Allocator object. +\param pVkMemoryRequirements Memory requirements for each allocation. +\param pCreateInfo Creation parameters for each allocation. +\param allocationCount Number of allocations to make. +\param[out] pAllocations Pointer to array that will be filled with handles to created allocations. +\param[out] pAllocationInfo Optional. Pointer to array that will be filled with parameters of created allocations. You should free the memory using vmaFreeMemory() or vmaFreeMemoryPages(). @@ -3361,12 +1787,15 @@ VMA_CALL_PRE VkResult VMA_CALL_POST vmaAllocateMemoryPages( const VkMemoryRequirements* VMA_NOT_NULL VMA_LEN_IF_NOT_NULL(allocationCount) pVkMemoryRequirements, const VmaAllocationCreateInfo* VMA_NOT_NULL VMA_LEN_IF_NOT_NULL(allocationCount) pCreateInfo, size_t allocationCount, - VmaAllocation VMA_NULLABLE * VMA_NOT_NULL VMA_LEN_IF_NOT_NULL(allocationCount) pAllocations, + VmaAllocation VMA_NULLABLE* VMA_NOT_NULL VMA_LEN_IF_NOT_NULL(allocationCount) pAllocations, VmaAllocationInfo* VMA_NULLABLE VMA_LEN_IF_NOT_NULL(allocationCount) pAllocationInfo); /** -@param[out] pAllocation Handle to allocated memory. -@param[out] pAllocationInfo Optional. Information about allocated memory. It can be later fetched using function vmaGetAllocationInfo(). +\param allocator +\param buffer +\param pCreateInfo +\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(). */ @@ -3374,7 +1803,7 @@ VMA_CALL_PRE VkResult VMA_CALL_POST vmaAllocateMemoryForBuffer( VmaAllocator VMA_NOT_NULL allocator, VkBuffer VMA_NOT_NULL_NON_DISPATCHABLE buffer, const VmaAllocationCreateInfo* VMA_NOT_NULL pCreateInfo, - VmaAllocation VMA_NULLABLE * VMA_NOT_NULL pAllocation, + VmaAllocation VMA_NULLABLE* VMA_NOT_NULL pAllocation, VmaAllocationInfo* VMA_NULLABLE pAllocationInfo); /// Function similar to vmaAllocateMemoryForBuffer(). @@ -3382,7 +1811,7 @@ VMA_CALL_PRE VkResult VMA_CALL_POST vmaAllocateMemoryForImage( VmaAllocator VMA_NOT_NULL allocator, VkImage VMA_NOT_NULL_NON_DISPATCHABLE image, const VmaAllocationCreateInfo* VMA_NOT_NULL pCreateInfo, - VmaAllocation VMA_NULLABLE * VMA_NOT_NULL pAllocation, + VmaAllocation VMA_NULLABLE* VMA_NOT_NULL pAllocation, VmaAllocationInfo* VMA_NULLABLE pAllocationInfo); /** \brief Frees memory previously allocated using vmaAllocateMemory(), vmaAllocateMemoryForBuffer(), or vmaAllocateMemoryForImage(). @@ -3406,47 +1835,23 @@ Passing `VK_NULL_HANDLE` as elements of `pAllocations` array is valid. Such entr VMA_CALL_PRE void VMA_CALL_POST vmaFreeMemoryPages( VmaAllocator VMA_NOT_NULL allocator, size_t allocationCount, - const VmaAllocation VMA_NULLABLE * VMA_NOT_NULL VMA_LEN_IF_NOT_NULL(allocationCount) pAllocations); + const VmaAllocation VMA_NULLABLE* VMA_NOT_NULL VMA_LEN_IF_NOT_NULL(allocationCount) pAllocations); -/** \brief Returns current information about specified allocation and atomically marks it as used in current frame. +/** \brief Returns current information about specified allocation. Current paramteres 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. +Although this function doesn't lock any mutex, so it should be quite efficient, +you should 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). */ VMA_CALL_PRE void VMA_CALL_POST vmaGetAllocationInfo( VmaAllocator VMA_NOT_NULL allocator, VmaAllocation VMA_NOT_NULL allocation, VmaAllocationInfo* VMA_NOT_NULL 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`. -*/ -VMA_CALL_PRE VkBool32 VMA_CALL_POST vmaTouchAllocation( - VmaAllocator VMA_NOT_NULL allocator, - VmaAllocation VMA_NOT_NULL allocation); - /** \brief Sets pUserData in given allocation to new value. If the allocation was created with VMA_ALLOCATION_CREATE_USER_DATA_COPY_STRING_BIT, @@ -3465,27 +1870,27 @@ VMA_CALL_PRE void VMA_CALL_POST vmaSetAllocationUserData( VmaAllocation VMA_NOT_NULL allocation, void* VMA_NULLABLE 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(). +/** +\brief Given an allocation, returns Property Flags of its memory type. -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. +This is just a convenience function. Same information can be obtained using +vmaGetAllocationInfo() + vmaGetMemoryProperties(). */ -VMA_CALL_PRE void VMA_CALL_POST vmaCreateLostAllocation( +VMA_CALL_PRE void VMA_CALL_POST vmaGetAllocationMemoryProperties( VmaAllocator VMA_NOT_NULL allocator, - VmaAllocation VMA_NULLABLE * VMA_NOT_NULL pAllocation); + VmaAllocation VMA_NOT_NULL allocation, + VkMemoryPropertyFlags* VMA_NOT_NULL pFlags); /** \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 offsetted to the beginning of region assigned to this particular -allocation. + +\warning +If the allocation is part of a bigger `VkDeviceMemory` block, returned pointer is +correctly offsetted to the beginning of region assigned to this particular allocation. +Unlike the result of `vkMapMemory`, it points to the allocation, not to the beginning of the whole block. +You should not add VmaAllocationInfo::offset to it! Mapping is internally reference-counted and synchronized, so despite raw Vulkan function `vkMapMemory()` cannot be used to map same block of `VkDeviceMemory` @@ -3509,10 +1914,6 @@ vmaMapMemory(). You must not call vmaUnmapMemory() additional time to free the 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. - This function doesn't automatically flush or invalidate caches. If the allocation is made from a memory types that is not `HOST_COHERENT`, you also need to use vmaInvalidateAllocation() / vmaFlushAllocation(), as required by Vulkan specification. @@ -3520,7 +1921,7 @@ you also need to use vmaInvalidateAllocation() / vmaFlushAllocation(), as requir VMA_CALL_PRE VkResult VMA_CALL_POST vmaMapMemory( VmaAllocator VMA_NOT_NULL allocator, VmaAllocation VMA_NOT_NULL allocation, - void* VMA_NULLABLE * VMA_NOT_NULL ppData); + void* VMA_NULLABLE* VMA_NOT_NULL ppData); /** \brief Unmaps memory represented by given allocation, mapped previously using vmaMapMemory(). @@ -3605,7 +2006,7 @@ called, otherwise `VK_SUCCESS`. VMA_CALL_PRE VkResult VMA_CALL_POST vmaFlushAllocations( VmaAllocator VMA_NOT_NULL allocator, uint32_t allocationCount, - const VmaAllocation VMA_NOT_NULL * VMA_NULLABLE VMA_LEN_IF_NOT_NULL(allocationCount) allocations, + const VmaAllocation VMA_NOT_NULL* VMA_NULLABLE VMA_LEN_IF_NOT_NULL(allocationCount) allocations, const VkDeviceSize* VMA_NULLABLE VMA_LEN_IF_NOT_NULL(allocationCount) offsets, const VkDeviceSize* VMA_NULLABLE VMA_LEN_IF_NOT_NULL(allocationCount) sizes); @@ -3626,13 +2027,14 @@ called, otherwise `VK_SUCCESS`. VMA_CALL_PRE VkResult VMA_CALL_POST vmaInvalidateAllocations( VmaAllocator VMA_NOT_NULL allocator, uint32_t allocationCount, - const VmaAllocation VMA_NOT_NULL * VMA_NULLABLE VMA_LEN_IF_NOT_NULL(allocationCount) allocations, + const VmaAllocation VMA_NOT_NULL* VMA_NULLABLE VMA_LEN_IF_NOT_NULL(allocationCount) allocations, const VkDeviceSize* VMA_NULLABLE VMA_LEN_IF_NOT_NULL(allocationCount) offsets, const VkDeviceSize* VMA_NULLABLE VMA_LEN_IF_NOT_NULL(allocationCount) sizes); /** \brief Checks magic number in margins around all allocations in given memory types (in both default and custom pools) in search for corruptions. -@param memoryTypeBits Bit mask, where each bit set means that a memory type with that index should be checked. +\param allocator +\param memoryTypeBits Bit mask, where each bit set means that a memory type with that index should be checked. Corruption detection is enabled only when `VMA_DEBUG_DETECT_CORRUPTION` macro is defined to nonzero, `VMA_DEBUG_MARGIN` is defined to nonzero and only for memory types that are @@ -3642,154 +2044,21 @@ Possible return values: - `VK_ERROR_FEATURE_NOT_PRESENT` - corruption detection is not enabled for any of specified memory types. - `VK_SUCCESS` - corruption detection has been performed and succeeded. -- `VK_ERROR_VALIDATION_FAILED_EXT` - corruption detection has been performed and found memory corruptions around one of the allocations. +- `VK_ERROR_UNKNOWN` - corruption detection has been performed and found memory corruptions around one of the allocations. `VMA_ASSERT` is also fired in that case. - Other value: Error returned by Vulkan, e.g. memory mapping failure. */ -VMA_CALL_PRE VkResult VMA_CALL_POST vmaCheckCorruption(VmaAllocator VMA_NOT_NULL allocator, uint32_t memoryTypeBits); - -/** \struct VmaDefragmentationContext -\brief Represents Opaque object that represents started defragmentation process. - -Fill structure #VmaDefragmentationInfo2 and call function vmaDefragmentationBegin() to create it. -Call function vmaDefragmentationEnd() to destroy it. -*/ -VK_DEFINE_HANDLE(VmaDefragmentationContext) - -/// Flags to be used in vmaDefragmentationBegin(). None at the moment. Reserved for future use. -typedef enum VmaDefragmentationFlagBits { - VMA_DEFRAGMENTATION_FLAG_INCREMENTAL = 0x1, - VMA_DEFRAGMENTATION_FLAG_BITS_MAX_ENUM = 0x7FFFFFFF -} VmaDefragmentationFlagBits; -typedef VkFlags VmaDefragmentationFlags; - -/** \brief Parameters for defragmentation. - -To be used with function vmaDefragmentationBegin(). -*/ -typedef struct VmaDefragmentationInfo2 { - /** \brief Reserved for future use. Should be 0. - */ - VmaDefragmentationFlags flags; - /** \brief Number of allocations in `pAllocations` array. - */ - uint32_t allocationCount; - /** \brief Pointer to array of allocations that can be defragmented. - - The array should have `allocationCount` elements. - The array should not contain nulls. - Elements in the array should be unique - same allocation cannot occur twice. - It is safe to pass allocations that are in the lost state - they are ignored. - All allocations not present in this array are considered non-moveable during this defragmentation. - */ - const VmaAllocation VMA_NOT_NULL * VMA_NULLABLE VMA_LEN_IF_NOT_NULL(allocationCount) pAllocations; - /** \brief Optional, output. Pointer to array that will be filled with information whether the allocation at certain index has been changed during defragmentation. - - The array should have `allocationCount` elements. - You can pass null if you are not interested in this information. - */ - VkBool32* VMA_NULLABLE VMA_LEN_IF_NOT_NULL(allocationCount) pAllocationsChanged; - /** \brief Numer of pools in `pPools` array. - */ - uint32_t poolCount; - /** \brief Either null or pointer to array of pools to be defragmented. - - All the allocations in the specified pools can be moved during defragmentation - and there is no way to check if they were really moved as in `pAllocationsChanged`, - so you must query all the allocations in all these pools for new `VkDeviceMemory` - and offset using vmaGetAllocationInfo() if you might need to recreate buffers - and images bound to them. - - The array should have `poolCount` elements. - The array should not contain nulls. - Elements in the array should be unique - same pool cannot occur twice. - - Using this array is equivalent to specifying all allocations from the pools in `pAllocations`. - It might be more efficient. - */ - const VmaPool VMA_NOT_NULL * VMA_NULLABLE VMA_LEN_IF_NOT_NULL(poolCount) pPools; - /** \brief Maximum total numbers of bytes that can be copied while moving allocations to different places using transfers on CPU side, like `memcpy()`, `memmove()`. - - `VK_WHOLE_SIZE` means no limit. - */ - VkDeviceSize maxCpuBytesToMove; - /** \brief Maximum number of allocations that can be moved to a different place using transfers on CPU side, like `memcpy()`, `memmove()`. - - `UINT32_MAX` means no limit. - */ - uint32_t maxCpuAllocationsToMove; - /** \brief Maximum total numbers of bytes that can be copied while moving allocations to different places using transfers on GPU side, posted to `commandBuffer`. - - `VK_WHOLE_SIZE` means no limit. - */ - VkDeviceSize maxGpuBytesToMove; - /** \brief Maximum number of allocations that can be moved to a different place using transfers on GPU side, posted to `commandBuffer`. - - `UINT32_MAX` means no limit. - */ - uint32_t maxGpuAllocationsToMove; - /** \brief Optional. Command buffer where GPU copy commands will be posted. - - If not null, it must be a valid command buffer handle that supports Transfer queue type. - It must be in the recording state and outside of a render pass instance. - You need to submit it and make sure it finished execution before calling vmaDefragmentationEnd(). - - Passing null means that only CPU defragmentation will be performed. - */ - VkCommandBuffer VMA_NULLABLE commandBuffer; -} VmaDefragmentationInfo2; - -typedef struct VmaDefragmentationPassMoveInfo { - VmaAllocation VMA_NOT_NULL allocation; - VkDeviceMemory VMA_NOT_NULL_NON_DISPATCHABLE memory; - VkDeviceSize offset; -} VmaDefragmentationPassMoveInfo; - -/** \brief Parameters for incremental defragmentation steps. - -To be used with function vmaBeginDefragmentationPass(). -*/ -typedef struct VmaDefragmentationPassInfo { - uint32_t moveCount; - VmaDefragmentationPassMoveInfo* VMA_NOT_NULL VMA_LEN_IF_NOT_NULL(moveCount) pMoves; -} VmaDefragmentationPassInfo; - -/** \brief Deprecated. Optional configuration parameters to be passed to function vmaDefragment(). - -\deprecated This is a part of the old interface. It is recommended to use structure #VmaDefragmentationInfo2 and function vmaDefragmentationBegin() instead. -*/ -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; +VMA_CALL_PRE VkResult VMA_CALL_POST vmaCheckCorruption( + VmaAllocator VMA_NOT_NULL allocator, + uint32_t memoryTypeBits); /** \brief Begins defragmentation process. -@param allocator Allocator object. -@param pInfo Structure filled with parameters of defragmentation. -@param[out] pStats Optional. Statistics of defragmentation. You can pass null if you are not interested in this information. -@param[out] pContext Context object that must be passed to vmaDefragmentationEnd() to finish defragmentation. -@return `VK_SUCCESS` and `*pContext == null` if defragmentation finished within this function call. `VK_NOT_READY` and `*pContext != null` if defragmentation has been started and you need to call vmaDefragmentationEnd() to finish it. Negative value in case of error. +\param allocator Allocator object. +\param pInfo Structure filled with parameters of defragmentation. +\param[out] pStats Optional. Statistics of defragmentation. You can pass null if you are not interested in this information. +\param[out] pContext Context object that must be passed to vmaDefragmentationEnd() to finish defragmentation. +\return `VK_SUCCESS` and `*pContext == null` if defragmentation finished within this function call. `VK_NOT_READY` and `*pContext != null` if defragmentation has been started and you need to call vmaDefragmentationEnd() to finish it. Negative value in case of error. Use this function instead of old, deprecated vmaDefragment(). @@ -3797,7 +2066,7 @@ Warning! Between the call to vmaDefragmentationBegin() and vmaDefragmentationEnd - You should not use any of allocations passed as `pInfo->pAllocations` or any allocations that belong to pools passed as `pInfo->pPools`, - including calling vmaGetAllocationInfo(), vmaTouchAllocation(), or access + including calling vmaGetAllocationInfo(), or access their data. - Some mutexes protecting internal data structures may be locked, so trying to make or free any allocations, bind buffers or images, map memory, or launch @@ -3816,7 +2085,7 @@ VMA_CALL_PRE VkResult VMA_CALL_POST vmaDefragmentationBegin( VmaAllocator VMA_NOT_NULL allocator, const VmaDefragmentationInfo2* VMA_NOT_NULL pInfo, VmaDefragmentationStats* VMA_NULLABLE pStats, - VmaDefragmentationContext VMA_NULLABLE * VMA_NOT_NULL pContext); + VmaDefragmentationContext VMA_NULLABLE* VMA_NOT_NULL pContext); /** \brief Ends defragmentation process. @@ -3830,21 +2099,21 @@ VMA_CALL_PRE VkResult VMA_CALL_POST vmaDefragmentationEnd( VMA_CALL_PRE VkResult VMA_CALL_POST vmaBeginDefragmentationPass( VmaAllocator VMA_NOT_NULL allocator, VmaDefragmentationContext VMA_NULLABLE context, - VmaDefragmentationPassInfo* VMA_NOT_NULL pInfo -); + VmaDefragmentationPassInfo* VMA_NOT_NULL pInfo); + VMA_CALL_PRE VkResult VMA_CALL_POST vmaEndDefragmentationPass( VmaAllocator VMA_NOT_NULL allocator, - VmaDefragmentationContext VMA_NULLABLE context -); + VmaDefragmentationContext VMA_NULLABLE context); /** \brief Deprecated. 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, negative error code in case of error. +\param allocator +\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, negative error code in case of error. \deprecated This is a part of the old interface. It is recommended to use structure #VmaDefragmentationInfo2 and function vmaDefragmentationBegin() instead. @@ -3879,7 +2148,7 @@ For more information, see [Defragmentation](@ref defragmentation) chapter. */ VMA_CALL_PRE VkResult VMA_CALL_POST vmaDefragment( VmaAllocator VMA_NOT_NULL allocator, - const VmaAllocation VMA_NOT_NULL * VMA_NOT_NULL VMA_LEN_IF_NOT_NULL(allocationCount) pAllocations, + const VmaAllocation VMA_NOT_NULL* VMA_NOT_NULL VMA_LEN_IF_NOT_NULL(allocationCount) pAllocations, size_t allocationCount, VkBool32* VMA_NULLABLE VMA_LEN_IF_NOT_NULL(allocationCount) pAllocationsChanged, const VmaDefragmentationInfo* VMA_NULLABLE pDefragmentationInfo, @@ -3904,8 +2173,11 @@ VMA_CALL_PRE VkResult VMA_CALL_POST vmaBindBufferMemory( /** \brief Binds buffer to allocation with additional parameters. -@param allocationLocalOffset Additional offset to be added while binding, relative to the beginning of the `allocation`. Normally it should be 0. -@param pNext A chain of structures to be attached to `VkBindBufferMemoryInfoKHR` structure used internally. Normally it should be null. +\param allocator +\param allocation +\param allocationLocalOffset Additional offset to be added while binding, relative to the beginning of the `allocation`. Normally it should be 0. +\param buffer +\param pNext A chain of structures to be attached to `VkBindBufferMemoryInfoKHR` structure used internally. Normally it should be null. This function is similar to vmaBindBufferMemory(), but it provides additional parameters. @@ -3938,8 +2210,11 @@ VMA_CALL_PRE VkResult VMA_CALL_POST vmaBindImageMemory( /** \brief Binds image to allocation with additional parameters. -@param allocationLocalOffset Additional offset to be added while binding, relative to the beginning of the `allocation`. Normally it should be 0. -@param pNext A chain of structures to be attached to `VkBindImageMemoryInfoKHR` structure used internally. Normally it should be null. +\param allocator +\param allocation +\param allocationLocalOffset Additional offset to be added while binding, relative to the beginning of the `allocation`. Normally it should be 0. +\param image +\param pNext A chain of structures to be attached to `VkBindImageMemoryInfoKHR` structure used internally. Normally it should be null. This function is similar to vmaBindImageMemory(), but it provides additional parameters. @@ -3954,9 +2229,12 @@ VMA_CALL_PRE VkResult VMA_CALL_POST vmaBindImageMemory2( const void* VMA_NULLABLE pNext); /** -@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(). +\param allocator +\param pBufferCreateInfo +\param pAllocationCreateInfo +\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: @@ -3974,8 +2252,8 @@ 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 +and if dedicated allocation is possible +(#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. @@ -3987,8 +2265,23 @@ VMA_CALL_PRE VkResult VMA_CALL_POST vmaCreateBuffer( VmaAllocator VMA_NOT_NULL allocator, const VkBufferCreateInfo* VMA_NOT_NULL pBufferCreateInfo, const VmaAllocationCreateInfo* VMA_NOT_NULL pAllocationCreateInfo, - VkBuffer VMA_NULLABLE_NON_DISPATCHABLE * VMA_NOT_NULL pBuffer, - VmaAllocation VMA_NULLABLE * VMA_NOT_NULL pAllocation, + VkBuffer VMA_NULLABLE_NON_DISPATCHABLE* VMA_NOT_NULL pBuffer, + VmaAllocation VMA_NULLABLE* VMA_NOT_NULL pAllocation, + VmaAllocationInfo* VMA_NULLABLE pAllocationInfo); + +/** \brief Creates a buffer with additional minimum alignment. + +Similar to vmaCreateBuffer() but provides additional parameter `minAlignment` which allows to specify custom, +minimum alignment to be used when placing the buffer inside a larger memory block, which may be needed e.g. +for interop with OpenGL. +*/ +VMA_CALL_PRE VkResult VMA_CALL_POST vmaCreateBufferWithAlignment( + VmaAllocator VMA_NOT_NULL allocator, + const VkBufferCreateInfo* VMA_NOT_NULL pBufferCreateInfo, + const VmaAllocationCreateInfo* VMA_NOT_NULL pAllocationCreateInfo, + VkDeviceSize minAlignment, + VkBuffer VMA_NULLABLE_NON_DISPATCHABLE* VMA_NOT_NULL pBuffer, + VmaAllocation VMA_NULLABLE* VMA_NOT_NULL pAllocation, VmaAllocationInfo* VMA_NULLABLE pAllocationInfo); /** \brief Destroys Vulkan buffer and frees allocated memory. @@ -4012,8 +2305,8 @@ VMA_CALL_PRE VkResult VMA_CALL_POST vmaCreateImage( VmaAllocator VMA_NOT_NULL allocator, const VkImageCreateInfo* VMA_NOT_NULL pImageCreateInfo, const VmaAllocationCreateInfo* VMA_NOT_NULL pAllocationCreateInfo, - VkImage VMA_NULLABLE_NON_DISPATCHABLE * VMA_NOT_NULL pImage, - VmaAllocation VMA_NULLABLE * VMA_NOT_NULL pAllocation, + VkImage VMA_NULLABLE_NON_DISPATCHABLE* VMA_NOT_NULL pImage, + VmaAllocation VMA_NULLABLE* VMA_NOT_NULL pAllocation, VmaAllocationInfo* VMA_NULLABLE pAllocationInfo); /** \brief Destroys Vulkan image and frees allocated memory. @@ -4032,12 +2325,153 @@ VMA_CALL_PRE void VMA_CALL_POST vmaDestroyImage( VkImage VMA_NULLABLE_NON_DISPATCHABLE image, VmaAllocation VMA_NULLABLE allocation); +/** @} */ + +/** +\addtogroup group_virtual +@{ +*/ + +/** \brief Creates new #VmaVirtualBlock object. + +\param pCreateInfo Parameters for creation. +\param[out] pVirtualBlock Returned virtual block object or `VMA_NULL` if creation failed. +*/ +VMA_CALL_PRE VkResult VMA_CALL_POST vmaCreateVirtualBlock( + const VmaVirtualBlockCreateInfo* VMA_NOT_NULL pCreateInfo, + VmaVirtualBlock VMA_NULLABLE* VMA_NOT_NULL pVirtualBlock); + +/** \brief Destroys #VmaVirtualBlock object. + +Please note that you should consciously handle virtual allocations that could remain unfreed in the block. +You should either free them individually using vmaVirtualFree() or call vmaClearVirtualBlock() +if you are sure this is what you want. If you do neither, an assert is called. + +If you keep pointers to some additional metadata associated with your virtual allocations in their `pUserData`, +don't forget to free them. +*/ +VMA_CALL_PRE void VMA_CALL_POST vmaDestroyVirtualBlock( + VmaVirtualBlock VMA_NULLABLE virtualBlock); + +/** \brief Returns true of the #VmaVirtualBlock is empty - contains 0 virtual allocations and has all its space available for new allocations. +*/ +VMA_CALL_PRE VkBool32 VMA_CALL_POST vmaIsVirtualBlockEmpty( + VmaVirtualBlock VMA_NOT_NULL virtualBlock); + +/** \brief Returns information about a specific virtual allocation within a virtual block, like its size and `pUserData` pointer. +*/ +VMA_CALL_PRE void VMA_CALL_POST vmaGetVirtualAllocationInfo( + VmaVirtualBlock VMA_NOT_NULL virtualBlock, + VmaVirtualAllocation VMA_NOT_NULL_NON_DISPATCHABLE allocation, VmaVirtualAllocationInfo* VMA_NOT_NULL pVirtualAllocInfo); + +/** \brief Allocates new virtual allocation inside given #VmaVirtualBlock. + +If the allocation fails due to not enough free space available, `VK_ERROR_OUT_OF_DEVICE_MEMORY` is returned +(despite the function doesn't ever allocate actual GPU memory). +`pAllocation` is then set to `VK_NULL_HANDLE` and `pOffset`, if not null, it set to `UINT64_MAX`. + +\param virtualBlock Virtual block +\param pCreateInfo Parameters for the allocation +\param[out] pAllocation Returned handle of the new allocation +\param[out] pOffset Returned offset of the new allocation. Optional, can be null. +*/ +VMA_CALL_PRE VkResult VMA_CALL_POST vmaVirtualAllocate( + VmaVirtualBlock VMA_NOT_NULL virtualBlock, + const VmaVirtualAllocationCreateInfo* VMA_NOT_NULL pCreateInfo, + VmaVirtualAllocation VMA_NULLABLE_NON_DISPATCHABLE* VMA_NOT_NULL pAllocation, + VkDeviceSize* VMA_NULLABLE pOffset); + +/** \brief Frees virtual allocation inside given #VmaVirtualBlock. + +It is correct to call this function with `allocation == VK_NULL_HANDLE` - it does nothing. +*/ +VMA_CALL_PRE void VMA_CALL_POST vmaVirtualFree( + VmaVirtualBlock VMA_NOT_NULL virtualBlock, + VmaVirtualAllocation VMA_NULLABLE_NON_DISPATCHABLE allocation); + +/** \brief Frees all virtual allocations inside given #VmaVirtualBlock. + +You must either call this function or free each virtual allocation individually with vmaVirtualFree() +before destroying a virtual block. Otherwise, an assert is called. + +If you keep pointer to some additional metadata associated with your virtual allocation in its `pUserData`, +don't forget to free it as well. +*/ +VMA_CALL_PRE void VMA_CALL_POST vmaClearVirtualBlock( + VmaVirtualBlock VMA_NOT_NULL virtualBlock); + +/** \brief Changes custom pointer associated with given virtual allocation. +*/ +VMA_CALL_PRE void VMA_CALL_POST vmaSetVirtualAllocationUserData( + VmaVirtualBlock VMA_NOT_NULL virtualBlock, + VmaVirtualAllocation VMA_NOT_NULL_NON_DISPATCHABLE allocation, + void* VMA_NULLABLE pUserData); + +/** \brief Calculates and returns statistics about virtual allocations and memory usage in given #VmaVirtualBlock. +*/ +VMA_CALL_PRE void VMA_CALL_POST vmaCalculateVirtualBlockStats( + VmaVirtualBlock VMA_NOT_NULL virtualBlock, + VmaStatInfo* VMA_NOT_NULL pStatInfo); + +/** @} */ + +#if VMA_STATS_STRING_ENABLED +/** +\addtogroup group_stats +@{ +*/ + +/** \brief Builds and returns a null-terminated string in JSON format with information about given #VmaVirtualBlock. +\param virtualBlock Virtual block. +\param[out] ppStatsString Returned string. +\param detailedMap Pass `VK_FALSE` to only obtain statistics as returned by vmaCalculateVirtualBlockStats(). Pass `VK_TRUE` to also obtain full list of allocations and free spaces. + +Returned string must be freed using vmaFreeVirtualBlockStatsString(). +*/ +VMA_CALL_PRE void VMA_CALL_POST vmaBuildVirtualBlockStatsString( + VmaVirtualBlock VMA_NOT_NULL virtualBlock, + char* VMA_NULLABLE* VMA_NOT_NULL ppStatsString, + VkBool32 detailedMap); + +/// Frees a string returned by vmaBuildVirtualBlockStatsString(). +VMA_CALL_PRE void VMA_CALL_POST vmaFreeVirtualBlockStatsString( + VmaVirtualBlock VMA_NOT_NULL virtualBlock, + char* VMA_NULLABLE pStatsString); + +/** \brief Builds and returns statistics as a null-terminated string in JSON format. +\param allocator +\param[out] ppStatsString Must be freed using vmaFreeStatsString() function. +\param detailedMap +*/ +VMA_CALL_PRE void VMA_CALL_POST vmaBuildStatsString( + VmaAllocator VMA_NOT_NULL allocator, + char* VMA_NULLABLE* VMA_NOT_NULL ppStatsString, + VkBool32 detailedMap); + +VMA_CALL_PRE void VMA_CALL_POST vmaFreeStatsString( + VmaAllocator VMA_NOT_NULL allocator, + char* VMA_NULLABLE pStatsString); + +/** @} */ + +#endif // VMA_STATS_STRING_ENABLED + +#endif // _VMA_FUNCTION_HEADERS + #ifdef __cplusplus } #endif #endif // AMD_VULKAN_MEMORY_ALLOCATOR_H +//////////////////////////////////////////////////////////////////////////////// +//////////////////////////////////////////////////////////////////////////////// +// +// IMPLEMENTATION +// +//////////////////////////////////////////////////////////////////////////////// +//////////////////////////////////////////////////////////////////////////////// + // For Visual Studio IntelliSense. #if defined(__cplusplus) && defined(__INTELLISENSE__) #define VMA_IMPLEMENTATION @@ -4051,22 +2485,13 @@ VMA_CALL_PRE void VMA_CALL_POST vmaDestroyImage( #include <cstring> #include <utility> -#if VMA_RECORDING_ENABLED - #include <chrono> - #if defined(_WIN32) - #include <windows.h> - #else - #include <sstream> - #include <thread> - #endif -#endif - /******************************************************************************* 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. */ +#ifndef _VMA_CONFIGURATION /* Define this macro to 1 to make the library fetch pointers to Vulkan functions @@ -4082,29 +2507,14 @@ internally, like: Define this macro to 1 to make the library fetch pointers to Vulkan functions internally, like: - vulkanFunctions.vkAllocateMemory = (PFN_vkAllocateMemory)vkGetDeviceProcAddr(m_hDevice, vkAllocateMemory); + vulkanFunctions.vkAllocateMemory = (PFN_vkAllocateMemory)vkGetDeviceProcAddr(device, "vkAllocateMemory"); + +To use this feature in new versions of VMA you now have to pass +VmaVulkanFunctions::vkGetInstanceProcAddr and vkGetDeviceProcAddr as +VmaAllocatorCreateInfo::pVulkanFunctions. Other members can be null. */ #if !defined(VMA_DYNAMIC_VULKAN_FUNCTIONS) #define VMA_DYNAMIC_VULKAN_FUNCTIONS 1 - #if defined(VK_NO_PROTOTYPES) - extern PFN_vkGetInstanceProcAddr vkGetInstanceProcAddr; - extern PFN_vkGetDeviceProcAddr vkGetDeviceProcAddr; - #endif -#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 #ifndef VMA_USE_STL_SHARED_MUTEX @@ -4112,8 +2522,7 @@ the containers. #if __cplusplus >= 201703L #define VMA_USE_STL_SHARED_MUTEX 1 // Visual studio defines __cplusplus properly only when passed additional parameter: /Zc:__cplusplus - // Otherwise it's always 199711L, despite shared_mutex works since Visual Studio 2015 Update 2. - // See: https://blogs.msdn.microsoft.com/vcblog/2018/04/09/msvc-now-correctly-reports-__cplusplus/ + // Otherwise it is always 199711L, despite shared_mutex works since Visual Studio 2015 Update 2. #elif defined(_MSC_FULL_VER) && _MSC_FULL_VER >= 190023918 && __cplusplus == 199711L && _MSVC_LANG >= 201703L #define VMA_USE_STL_SHARED_MUTEX 1 #else @@ -4122,28 +2531,33 @@ the containers. #endif /* -THESE INCLUDES ARE NOT ENABLED BY DEFAULT. -Library has its own container implementation. -*/ -#if VMA_USE_STL_VECTOR - #include <vector> -#endif +Define this macro to include custom header files without having to edit this file directly, e.g.: -#if VMA_USE_STL_UNORDERED_MAP - #include <unordered_map> -#endif + // Inside of "my_vma_configuration_user_includes.h": -#if VMA_USE_STL_LIST - #include <list> -#endif + #include "my_custom_assert.h" // for MY_CUSTOM_ASSERT + #include "my_custom_min.h" // for my_custom_min + #include <algorithm> + #include <mutex> -/* -Following headers are used in this CONFIGURATION section only, so feel free to + // Inside a different file, which includes "vk_mem_alloc.h": + + #define VMA_CONFIGURATION_USER_INCLUDES_H "my_vma_configuration_user_includes.h" + #define VMA_ASSERT(expr) MY_CUSTOM_ASSERT(expr) + #define VMA_MIN(v1, v2) (my_custom_min(v1, v2)) + #include "vk_mem_alloc.h" + ... + +The following headers are used in this CONFIGURATION section only, so feel free to remove them if not needed. */ -#include <cassert> // for assert -#include <algorithm> // for min, max -#include <mutex> +#if !defined(VMA_CONFIGURATION_USER_INCLUDES_H) + #include <cassert> // for assert + #include <algorithm> // for min, max + #include <mutex> +#else + #include VMA_CONFIGURATION_USER_INCLUDES_H +#endif #ifndef VMA_NULL // Value used as null pointer. Define it to e.g.: nullptr, NULL, 0, (void*)0. @@ -4171,18 +2585,21 @@ static void* vma_aligned_alloc(size_t alignment, size_t size) static void* vma_aligned_alloc(size_t alignment, size_t size) { -#if defined(__APPLE__) && (defined(MAC_OS_X_VERSION_10_16) || defined(__IPHONE_14_0)) -#if MAC_OS_X_VERSION_MAX_ALLOWED >= MAC_OS_X_VERSION_10_16 || __IPHONE_OS_VERSION_MAX_ALLOWED >= __IPHONE_14_0 - // For C++14, usr/include/malloc/_malloc.h declares aligned_alloc()) only - // with the MacOSX11.0 SDK in Xcode 12 (which is what adds - // MAC_OS_X_VERSION_10_16), even though the function is marked - // availabe for 10.15. That's why the preprocessor checks for 10.16 but - // the __builtin_available checks for 10.15. - // People who use C++17 could call aligned_alloc with the 10.15 SDK already. - if (__builtin_available(macOS 10.15, iOS 13, *)) - return aligned_alloc(alignment, size); -#endif -#endif + // Unfortunately, aligned_alloc causes VMA to crash due to it returning null pointers. (At least under 11.4) + // Therefore, for now disable this specific exception until a proper solution is found. + //#if defined(__APPLE__) && (defined(MAC_OS_X_VERSION_10_16) || defined(__IPHONE_14_0)) + //#if MAC_OS_X_VERSION_MAX_ALLOWED >= MAC_OS_X_VERSION_10_16 || __IPHONE_OS_VERSION_MAX_ALLOWED >= __IPHONE_14_0 + // // For C++14, usr/include/malloc/_malloc.h declares aligned_alloc()) only + // // with the MacOSX11.0 SDK in Xcode 12 (which is what adds + // // MAC_OS_X_VERSION_10_16), even though the function is marked + // // availabe for 10.15. That is why the preprocessor checks for 10.16 but + // // the __builtin_available checks for 10.15. + // // People who use C++17 could call aligned_alloc with the 10.15 SDK already. + // if (__builtin_available(macOS 10.15, iOS 13, *)) + // return aligned_alloc(alignment, size); + //#endif + //#endif + // alignment must be >= sizeof(void*) if(alignment < sizeof(void*)) { @@ -4259,12 +2676,22 @@ static void vma_aligned_free(void* VMA_NULLABLE ptr) #endif #endif +#ifndef VMA_BITSCAN_LSB + // Scans integer for index of first nonzero value from the Least Significant Bit (LSB). If mask is 0 then returns UINT8_MAX + #define VMA_BITSCAN_LSB(mask) VmaBitScanLSB(mask) +#endif + +#ifndef VMA_BITSCAN_MSB + // Scans integer for index of first nonzero value from the Most Significant Bit (MSB). If mask is 0 then returns UINT8_MAX + #define VMA_BITSCAN_MSB(mask) VmaBitScanMSB(mask) +#endif + #ifndef VMA_MIN - #define VMA_MIN(v1, v2) (std::min((v1), (v2))) + #define VMA_MIN(v1, v2) ((std::min)((v1), (v2))) #endif #ifndef VMA_MAX - #define VMA_MAX(v1, v2) (std::max((v1), (v2))) + #define VMA_MAX(v1, v2) ((std::max)((v1), (v2))) #endif #ifndef VMA_SWAP @@ -4402,7 +2829,7 @@ If providing your own implementation, you need to implement a subset of std::ato #ifndef VMA_DEBUG_MARGIN /** - Minimum margin before and after every allocation, in bytes. + Minimum margin after every allocation, in bytes. Set nonzero for debugging purposes only. */ #define VMA_DEBUG_MARGIN (0) @@ -4419,7 +2846,7 @@ If providing your own implementation, you need to implement a subset of std::ato #ifndef VMA_DEBUG_DETECT_CORRUPTION /** Define this macro to 1 together with non-zero value of VMA_DEBUG_MARGIN to - enable writing magic value to the margin before and after every allocation and + enable writing magic value to the margin after every allocation and validating it, so that memory corruptions (out-of-bounds writes) are detected. */ #define VMA_DEBUG_DETECT_CORRUPTION (0) @@ -4466,38 +2893,273 @@ If providing your own implementation, you need to implement a subset of std::ato className& operator=(const className&) = delete; #endif -static const uint32_t VMA_FRAME_INDEX_LOST = UINT32_MAX; - -// Decimal 2139416166, float NaN, little-endian binary 66 E6 84 7F. -static const uint32_t VMA_CORRUPTION_DETECTION_MAGIC_VALUE = 0x7F84E666; - -static const uint8_t VMA_ALLOCATION_FILL_PATTERN_CREATED = 0xDC; -static const uint8_t VMA_ALLOCATION_FILL_PATTERN_DESTROYED = 0xEF; +#define VMA_VALIDATE(cond) do { if(!(cond)) { \ + VMA_ASSERT(0 && "Validation failed: " #cond); \ + return false; \ + } } while(false) /******************************************************************************* END OF CONFIGURATION */ +#endif // _VMA_CONFIGURATION + -// # Copy of some Vulkan definitions so we don't need to check their existence just to handle few constants. +static const uint8_t VMA_ALLOCATION_FILL_PATTERN_CREATED = 0xDC; +static const uint8_t VMA_ALLOCATION_FILL_PATTERN_DESTROYED = 0xEF; +// Decimal 2139416166, float NaN, little-endian binary 66 E6 84 7F. +static const uint32_t VMA_CORRUPTION_DETECTION_MAGIC_VALUE = 0x7F84E666; +// Copy of some Vulkan definitions so we don't need to check their existence just to handle few constants. static const uint32_t VK_MEMORY_PROPERTY_DEVICE_COHERENT_BIT_AMD_COPY = 0x00000040; static const uint32_t VK_MEMORY_PROPERTY_DEVICE_UNCACHED_BIT_AMD_COPY = 0x00000080; static const uint32_t VK_BUFFER_USAGE_SHADER_DEVICE_ADDRESS_BIT_COPY = 0x00020000; - static const uint32_t VMA_ALLOCATION_INTERNAL_STRATEGY_MIN_OFFSET = 0x10000000u; +static const uint32_t VMA_ALLOCATION_TRY_COUNT = 32; +static const uint32_t VMA_VENDOR_ID_AMD = 4098; + + +#if VMA_STATS_STRING_ENABLED +// Correspond to values of enum VmaSuballocationType. +static const char* VMA_SUBALLOCATION_TYPE_NAMES[] = +{ + "FREE", + "UNKNOWN", + "BUFFER", + "IMAGE_UNKNOWN", + "IMAGE_LINEAR", + "IMAGE_OPTIMAL", +}; +#endif + +static VkAllocationCallbacks VmaEmptyAllocationCallbacks = + { VMA_NULL, VMA_NULL, VMA_NULL, VMA_NULL, VMA_NULL, VMA_NULL }; -static VkAllocationCallbacks VmaEmptyAllocationCallbacks = { - VMA_NULL, VMA_NULL, VMA_NULL, VMA_NULL, VMA_NULL, VMA_NULL }; +#ifndef _VMA_ENUM_DECLARATIONS + +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 +}; + +enum VMA_CACHE_OPERATION +{ + VMA_CACHE_FLUSH, + VMA_CACHE_INVALIDATE +}; + +enum class VmaAllocationRequestType +{ + Normal, + TLSF, + // Used by "Linear" algorithm. + UpperAddress, + EndOf1st, + EndOf2nd, +}; + +#endif // _VMA_ENUM_DECLARATIONS + +#ifndef _VMA_FORWARD_DECLARATIONS +// Opaque handle used by allocation algorithms to identify single allocation in any conforming way. +VK_DEFINE_NON_DISPATCHABLE_HANDLE(VmaAllocHandle); + +struct VmaMutexLock; +struct VmaMutexLockRead; +struct VmaMutexLockWrite; + +template<typename T> +struct AtomicTransactionalIncrement; + +template<typename T> +struct VmaStlAllocator; + +template<typename T, typename AllocatorT> +class VmaVector; + +template<typename T, typename AllocatorT, size_t N> +class VmaSmallVector; + +template<typename T> +class VmaPoolAllocator; + +template<typename T> +struct VmaListItem; + +template<typename T> +class VmaRawList; + +template<typename T, typename AllocatorT> +class VmaList; + +template<typename ItemTypeTraits> +class VmaIntrusiveLinkedList; + +// Unused in this version +#if 0 +template<typename T1, typename T2> +struct VmaPair; +template<typename FirstT, typename SecondT> +struct VmaPairFirstLess; + +template<typename KeyT, typename ValueT> +class VmaMap; +#endif + +#if VMA_STATS_STRING_ENABLED +class VmaStringBuilder; +class VmaJsonWriter; +#endif + +class VmaDeviceMemoryBlock; + +struct VmaDedicatedAllocationListItemTraits; +class VmaDedicatedAllocationList; + +struct VmaSuballocation; +struct VmaSuballocationOffsetLess; +struct VmaSuballocationOffsetGreater; +struct VmaSuballocationItemSizeLess; + +typedef VmaList<VmaSuballocation, VmaStlAllocator<VmaSuballocation>> VmaSuballocationList; + +struct VmaAllocationRequest; + +class VmaBlockMetadata; +class VmaBlockMetadata_Generic; +class VmaBlockMetadata_Linear; +class VmaBlockMetadata_Buddy; +class VmaBlockMetadata_TLSF; + +class VmaBlockVector; + +struct VmaDefragmentationMove; +class VmaDefragmentationAlgorithm; +class VmaDefragmentationAlgorithm_Generic; +class VmaDefragmentationAlgorithm_Fast; + +struct VmaPoolListItemTraits; + +struct VmaBlockDefragmentationContext; +class VmaBlockVectorDefragmentationContext; + +struct VmaCurrentBudgetData; + +class VmaAllocationObjectAllocator; + +#endif // _VMA_FORWARD_DECLARATIONS + + +#ifndef _VMA_FUNCTIONS // Returns number of bits set to 1 in (v). static inline uint32_t VmaCountBitsSet(uint32_t v) { +#ifdef _MSC_VER + return __popcnt(v); +#elif defined __GNUC__ || defined __clang__ + return static_cast<uint32_t>(__builtin_popcount(v)); +#else 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 >> 2) & 0x33333333) + (c & 0x33333333); + c = ((c >> 4) + c) & 0x0F0F0F0F; + c = ((c >> 8) + c) & 0x00FF00FF; c = ((c >> 16) + c) & 0x0000FFFF; return c; +#endif +} + +static inline uint8_t VmaBitScanLSB(uint64_t mask) +{ +#if defined(_MSC_VER) && defined(_WIN64) + unsigned long pos; + if (_BitScanForward64(&pos, mask)) + return static_cast<uint8_t>(pos); + return UINT8_MAX; +#elif defined __GNUC__ || defined __clang__ + return static_cast<uint8_t>(__builtin_ffsll(mask)) - 1U; +#else + uint8_t pos = 0; + uint64_t bit = 1; + do + { + if (mask & bit) + return pos; + bit <<= 1; + } while (pos++ < 63); + return UINT8_MAX; +#endif +} + +static inline uint8_t VmaBitScanLSB(uint32_t mask) +{ +#ifdef _MSC_VER + unsigned long pos; + if (_BitScanForward(&pos, mask)) + return static_cast<uint8_t>(pos); + return UINT8_MAX; +#elif defined __GNUC__ || defined __clang__ + return static_cast<uint8_t>(__builtin_ffs(mask)) - 1U; +#else + uint8_t pos = 0; + uint32_t bit = 1; + do + { + if (mask & bit) + return pos; + bit <<= 1; + } while (pos++ < 31); + return UINT8_MAX; +#endif +} + +static inline uint8_t VmaBitScanMSB(uint64_t mask) +{ +#if defined(_MSC_VER) && defined(_WIN64) + unsigned long pos; + if (_BitScanReverse64(&pos, mask)) + return static_cast<uint8_t>(pos); +#elif defined __GNUC__ || defined __clang__ + if (mask) + return 63 - static_cast<uint8_t>(__builtin_clzll(mask)); +#else + uint8_t pos = 63; + uint64_t bit = 1ULL << 63; + do + { + if (mask & bit) + return pos; + bit >>= 1; + } while (pos-- > 0); +#endif + return UINT8_MAX; +} + +static inline uint8_t VmaBitScanMSB(uint32_t mask) +{ +#ifdef _MSC_VER + unsigned long pos; + if (_BitScanReverse(&pos, mask)) + return static_cast<uint8_t>(pos); +#elif defined __GNUC__ || defined __clang__ + if (mask) + return 31 - static_cast<uint8_t>(__builtin_clz(mask)); +#else + uint8_t pos = 31; + uint32_t bit = 1UL << 31; + do + { + if (mask & bit) + return pos; + bit >>= 1; + } while (pos-- > 0); +#endif + return UINT8_MAX; } /* @@ -4508,7 +3170,7 @@ For 0 returns true. template <typename T> inline bool VmaIsPow2(T x) { - return (x & (x-1)) == 0; + return (x & (x - 1)) == 0; } // Aligns given value up to nearest multiply of align value. For example: VmaAlignUp(11, 8) = 16. @@ -4519,6 +3181,7 @@ static inline T VmaAlignUp(T val, T alignment) VMA_HEAVY_ASSERT(VmaIsPow2(alignment)); return (val + alignment - 1) & ~(alignment - 1); } + // Aligns given value down to nearest multiply of align value. For example: VmaAlignUp(11, 8) = 8. // Use types like uint32_t, uint64_t as T. template <typename T> @@ -4535,6 +3198,13 @@ static inline T VmaRoundDiv(T x, T y) return (x + (y / (T)2)) / y; } +// Divide by 'y' and round up to nearest integer. +template <typename T> +static inline T VmaDivideRoundingUp(T x, T y) +{ + return (x + y - (T)1) / y; +} + // Returns smallest power of 2 greater or equal to v. static inline uint32_t VmaNextPow2(uint32_t v) { @@ -4547,6 +3217,7 @@ static inline uint32_t VmaNextPow2(uint32_t v) v++; return v; } + static inline uint64_t VmaNextPow2(uint64_t v) { v--; @@ -4571,6 +3242,7 @@ static inline uint32_t VmaPrevPow2(uint32_t v) v = v ^ (v >> 1); return v; } + static inline uint64_t VmaPrevPow2(uint64_t v) { v |= v >> 1; @@ -4589,15 +3261,16 @@ static inline bool VmaStrIsEmpty(const char* pStr) } #if VMA_STATS_STRING_ENABLED - static const char* VmaAlgorithmToStr(uint32_t algorithm) { - switch(algorithm) + switch (algorithm) { case VMA_POOL_CREATE_LINEAR_ALGORITHM_BIT: return "Linear"; case VMA_POOL_CREATE_BUDDY_ALGORITHM_BIT: return "Buddy"; + case VMA_POOL_CREATE_TLSF_ALGORITHM_BIT: + return "TLSF"; case 0: return "Default"; default: @@ -4605,28 +3278,26 @@ static const char* VmaAlgorithmToStr(uint32_t algorithm) return ""; } } - -#endif // #if VMA_STATS_STRING_ENABLED +#endif // VMA_STATS_STRING_ENABLED #ifndef VMA_SORT - template<typename Iterator, typename Compare> Iterator VmaQuickSortPartition(Iterator beg, Iterator end, Compare cmp) { Iterator centerValue = end; --centerValue; Iterator insertIndex = beg; - for(Iterator memTypeIndex = beg; memTypeIndex < centerValue; ++memTypeIndex) + for (Iterator memTypeIndex = beg; memTypeIndex < centerValue; ++memTypeIndex) { - if(cmp(*memTypeIndex, *centerValue)) + if (cmp(*memTypeIndex, *centerValue)) { - if(insertIndex != memTypeIndex) + if (insertIndex != memTypeIndex) { VMA_SWAP(*memTypeIndex, *insertIndex); } ++insertIndex; } } - if(insertIndex != centerValue) + if (insertIndex != centerValue) { VMA_SWAP(*insertIndex, *centerValue); } @@ -4636,7 +3307,7 @@ Iterator VmaQuickSortPartition(Iterator beg, Iterator end, Compare cmp) template<typename Iterator, typename Compare> void VmaQuickSort(Iterator beg, Iterator end, Compare cmp) { - if(beg < end) + if (beg < end) { Iterator it = VmaQuickSortPartition<Iterator, Compare>(beg, end, cmp); VmaQuickSort<Iterator, Compare>(beg, it, cmp); @@ -4645,8 +3316,7 @@ void VmaQuickSort(Iterator beg, Iterator end, Compare cmp) } #define VMA_SORT(beg, end, cmp) VmaQuickSort(beg, end, cmp) - -#endif // #ifndef VMA_SORT +#endif // VMA_SORT /* Returns true if two memory blocks occupy overlapping pages. @@ -4669,17 +3339,6 @@ static inline bool VmaBlocksOnSamePage( 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 @@ -4690,12 +3349,12 @@ static inline bool VmaIsBufferImageGranularityConflict( VmaSuballocationType suballocType1, VmaSuballocationType suballocType2) { - if(suballocType1 > suballocType2) + if (suballocType1 > suballocType2) { VMA_SWAP(suballocType1, suballocType2); } - switch(suballocType1) + switch (suballocType1) { case VMA_SUBALLOCATION_TYPE_FREE: return false; @@ -4726,7 +3385,7 @@ static void VmaWriteMagicValue(void* pData, VkDeviceSize offset) #if VMA_DEBUG_MARGIN > 0 && VMA_DEBUG_DETECT_CORRUPTION uint32_t* pDst = (uint32_t*)((char*)pData + offset); const size_t numberCount = VMA_DEBUG_MARGIN / sizeof(uint32_t); - for(size_t i = 0; i < numberCount; ++i, ++pDst) + for (size_t i = 0; i < numberCount; ++i, ++pDst) { *pDst = VMA_CORRUPTION_DETECTION_MAGIC_VALUE; } @@ -4740,9 +3399,9 @@ static bool VmaValidateMagicValue(const void* pData, VkDeviceSize offset) #if VMA_DEBUG_MARGIN > 0 && VMA_DEBUG_DETECT_CORRUPTION const uint32_t* pSrc = (const uint32_t*)((const char*)pData + offset); const size_t numberCount = VMA_DEBUG_MARGIN / sizeof(uint32_t); - for(size_t i = 0; i < numberCount; ++i, ++pSrc) + for (size_t i = 0; i < numberCount; ++i, ++pSrc) { - if(*pSrc != VMA_CORRUPTION_DETECTION_MAGIC_VALUE) + if (*pSrc != VMA_CORRUPTION_DETECTION_MAGIC_VALUE) { return false; } @@ -4763,55 +3422,6 @@ static void VmaFillGpuDefragmentationBufferCreateInfo(VkBufferCreateInfo& outBuf outBufCreateInfo.size = (VkDeviceSize)VMA_DEFAULT_LARGE_HEAP_BLOCK_SIZE; // Example size. } -// Helper RAII class to lock a mutex in constructor and unlock it in destructor (at the end of scope). -struct VmaMutexLock -{ - VMA_CLASS_NO_COPY(VmaMutexLock) -public: - VmaMutexLock(VMA_MUTEX& mutex, bool useMutex = true) : - m_pMutex(useMutex ? &mutex : VMA_NULL) - { if(m_pMutex) { m_pMutex->Lock(); } } - ~VmaMutexLock() - { if(m_pMutex) { m_pMutex->Unlock(); } } -private: - VMA_MUTEX* m_pMutex; -}; - -// Helper RAII class to lock a RW mutex in constructor and unlock it in destructor (at the end of scope), for reading. -struct VmaMutexLockRead -{ - VMA_CLASS_NO_COPY(VmaMutexLockRead) -public: - VmaMutexLockRead(VMA_RW_MUTEX& mutex, bool useMutex) : - m_pMutex(useMutex ? &mutex : VMA_NULL) - { if(m_pMutex) { m_pMutex->LockRead(); } } - ~VmaMutexLockRead() { if(m_pMutex) { m_pMutex->UnlockRead(); } } -private: - VMA_RW_MUTEX* m_pMutex; -}; - -// Helper RAII class to lock a RW mutex in constructor and unlock it in destructor (at the end of scope), for writing. -struct VmaMutexLockWrite -{ - VMA_CLASS_NO_COPY(VmaMutexLockWrite) -public: - VmaMutexLockWrite(VMA_RW_MUTEX& mutex, bool useMutex) : - m_pMutex(useMutex ? &mutex : VMA_NULL) - { if(m_pMutex) { m_pMutex->LockWrite(); } } - ~VmaMutexLockWrite() { if(m_pMutex) { m_pMutex->UnlockWrite(); } } -private: - VMA_RW_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 @@ -4823,13 +3433,13 @@ Returned value is the found element, if present in the collection or place where new element with value (key) should be inserted. */ template <typename CmpLess, typename IterT, typename KeyT> -static IterT VmaBinaryFindFirstNotLess(IterT beg, IterT end, const KeyT &key, const CmpLess& cmp) +static IterT VmaBinaryFindFirstNotLess(IterT beg, IterT end, const KeyT& key, const CmpLess& cmp) { size_t down = 0, up = (end - beg); - while(down < up) + while (down < up) { const size_t mid = down + (up - down) / 2; // Overflow-safe midpoint calculation - if(cmp(*(beg+mid), key)) + if (cmp(*(beg + mid), key)) { down = mid + 1; } @@ -4846,7 +3456,7 @@ IterT VmaBinaryFindSorted(const IterT& beg, const IterT& end, const KeyT& value, { IterT it = VmaBinaryFindFirstNotLess<CmpLess, IterT, KeyT>( beg, end, value, cmp); - if(it == end || + if (it == end || (!cmp(*it, value) && !cmp(value, *it))) { return it; @@ -4862,16 +3472,16 @@ T must be pointer type, e.g. VmaAllocation, VmaPool. template<typename T> static bool VmaValidatePointerArray(uint32_t count, const T* arr) { - for(uint32_t i = 0; i < count; ++i) + for (uint32_t i = 0; i < count; ++i) { const T iPtr = arr[i]; - if(iPtr == VMA_NULL) + if (iPtr == VMA_NULL) { return false; } - for(uint32_t j = i + 1; j < count; ++j) + for (uint32_t j = i + 1; j < count; ++j) { - if(iPtr == arr[j]) + if (iPtr == arr[j]) { return false; } @@ -4893,7 +3503,7 @@ static inline void VmaPnextChainPushFront(MainT* mainStruct, NewT* newStruct) static void* VmaMalloc(const VkAllocationCallbacks* pAllocationCallbacks, size_t size, size_t alignment) { void* result = VMA_NULL; - if((pAllocationCallbacks != VMA_NULL) && + if ((pAllocationCallbacks != VMA_NULL) && (pAllocationCallbacks->pfnAllocation != VMA_NULL)) { result = (*pAllocationCallbacks->pfnAllocation)( @@ -4912,7 +3522,7 @@ static void* VmaMalloc(const VkAllocationCallbacks* pAllocationCallbacks, size_t static void VmaFree(const VkAllocationCallbacks* pAllocationCallbacks, void* ptr) { - if((pAllocationCallbacks != VMA_NULL) && + if ((pAllocationCallbacks != VMA_NULL) && (pAllocationCallbacks->pfnFree != VMA_NULL)) { (*pAllocationCallbacks->pfnFree)(pAllocationCallbacks->pUserData, ptr); @@ -4949,9 +3559,9 @@ static void vma_delete(const VkAllocationCallbacks* pAllocationCallbacks, T* ptr template<typename T> static void vma_delete_array(const VkAllocationCallbacks* pAllocationCallbacks, T* ptr, size_t count) { - if(ptr != VMA_NULL) + if (ptr != VMA_NULL) { - for(size_t i = count; i--; ) + for (size_t i = count; i--; ) { ptr[i].~T(); } @@ -4961,302 +3571,442 @@ static void vma_delete_array(const VkAllocationCallbacks* pAllocationCallbacks, static char* VmaCreateStringCopy(const VkAllocationCallbacks* allocs, const char* srcStr) { - if(srcStr != VMA_NULL) + if (srcStr != VMA_NULL) { const size_t len = strlen(srcStr); char* const result = vma_new_array(allocs, char, len + 1); memcpy(result, srcStr, len + 1); return result; } - else + return VMA_NULL; +} + +#if VMA_STATS_STRING_ENABLED +static char* VmaCreateStringCopy(const VkAllocationCallbacks* allocs, const char* srcStr, size_t strLen) +{ + if (srcStr != VMA_NULL) { - return VMA_NULL; + char* const result = vma_new_array(allocs, char, strLen + 1); + memcpy(result, srcStr, strLen); + result[strLen] = '\0'; + return result; } + return VMA_NULL; } +#endif // VMA_STATS_STRING_ENABLED static void VmaFreeString(const VkAllocationCallbacks* allocs, char* str) { - if(str != VMA_NULL) + if (str != VMA_NULL) { const size_t len = strlen(str); vma_delete_array(allocs, str, len + 1); } } -// STL-compatible allocator. -template<typename T> -class VmaStlAllocator +template<typename CmpLess, typename VectorT> +size_t VmaVectorInsertSorted(VectorT& vector, const typename VectorT::value_type& value) { -public: - const VkAllocationCallbacks* const m_pCallbacks; - typedef T value_type; + const size_t indexToInsert = VmaBinaryFindFirstNotLess( + vector.data(), + vector.data() + vector.size(), + value, + CmpLess()) - vector.data(); + VmaVectorInsert(vector, indexToInsert, value); + return indexToInsert; +} - VmaStlAllocator(const VkAllocationCallbacks* pCallbacks) : m_pCallbacks(pCallbacks) { } - template<typename U> VmaStlAllocator(const VmaStlAllocator<U>& src) : m_pCallbacks(src.m_pCallbacks) { } +template<typename CmpLess, typename VectorT> +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; +} +#endif // _VMA_FUNCTIONS - T* allocate(size_t n) { return VmaAllocateArray<T>(m_pCallbacks, n); } - void deallocate(T* p, size_t n) { VmaFree(m_pCallbacks, p); } +#ifndef _VMA_STAT_INFO_FUNCTIONS +static void VmaInitStatInfo(VmaStatInfo& outInfo) +{ + memset(&outInfo, 0, sizeof(outInfo)); + outInfo.allocationSizeMin = UINT64_MAX; + outInfo.unusedRangeSizeMin = UINT64_MAX; +} - template<typename U> - bool operator==(const VmaStlAllocator<U>& rhs) const +// Adds statistics srcInfo into inoutInfo, like: inoutInfo += srcInfo. +static void VmaAddStatInfo(VmaStatInfo& inoutInfo, const VmaStatInfo& srcInfo) +{ + inoutInfo.blockCount += srcInfo.blockCount; + inoutInfo.allocationCount += srcInfo.allocationCount; + inoutInfo.unusedRangeCount += srcInfo.unusedRangeCount; + inoutInfo.usedBytes += srcInfo.usedBytes; + inoutInfo.unusedBytes += srcInfo.unusedBytes; + inoutInfo.allocationSizeMin = VMA_MIN(inoutInfo.allocationSizeMin, srcInfo.allocationSizeMin); + inoutInfo.allocationSizeMax = VMA_MAX(inoutInfo.allocationSizeMax, srcInfo.allocationSizeMax); + inoutInfo.unusedRangeSizeMin = VMA_MIN(inoutInfo.unusedRangeSizeMin, srcInfo.unusedRangeSizeMin); + inoutInfo.unusedRangeSizeMax = VMA_MAX(inoutInfo.unusedRangeSizeMax, srcInfo.unusedRangeSizeMax); +} + +static void VmaAddStatInfoAllocation(VmaStatInfo& inoutInfo, VkDeviceSize size) +{ + ++inoutInfo.allocationCount; + inoutInfo.usedBytes += size; + if (size < inoutInfo.allocationSizeMin) { - return m_pCallbacks == rhs.m_pCallbacks; + inoutInfo.allocationSizeMin = size; } - template<typename U> - bool operator!=(const VmaStlAllocator<U>& rhs) const + if (size > inoutInfo.allocationSizeMax) { - return m_pCallbacks != rhs.m_pCallbacks; + inoutInfo.allocationSizeMax = size; } +} - VmaStlAllocator& operator=(const VmaStlAllocator& x) = delete; - VmaStlAllocator(const VmaStlAllocator&) = default; -}; - -#if VMA_USE_STL_VECTOR - -#define VmaVector std::vector - -template<typename T, typename allocatorT> -static void VmaVectorInsert(std::vector<T, allocatorT>& vec, size_t index, const T& item) +static void VmaAddStatInfoUnusedRange(VmaStatInfo& inoutInfo, VkDeviceSize size) { - vec.insert(vec.begin() + index, item); + ++inoutInfo.unusedRangeCount; + inoutInfo.unusedBytes += size; + if (size < inoutInfo.unusedRangeSizeMin) + { + inoutInfo.unusedRangeSizeMin = size; + } + if (size > inoutInfo.unusedRangeSizeMax) + { + inoutInfo.unusedRangeSizeMax = size; + } } -template<typename T, typename allocatorT> -static void VmaVectorRemove(std::vector<T, allocatorT>& vec, size_t index) +static void VmaPostprocessCalcStatInfo(VmaStatInfo& inoutInfo) { - vec.erase(vec.begin() + index); + inoutInfo.allocationSizeAvg = (inoutInfo.allocationCount > 0) ? + VmaRoundDiv<VkDeviceSize>(inoutInfo.usedBytes, inoutInfo.allocationCount) : 0; + inoutInfo.unusedRangeSizeAvg = (inoutInfo.unusedRangeCount > 0) ? + VmaRoundDiv<VkDeviceSize>(inoutInfo.unusedBytes, inoutInfo.unusedRangeCount) : 0; } +#endif // _VMA_STAT_INFO_FUNCTIONS -#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<typename T, typename AllocatorT> -class VmaVector +#ifndef _VMA_MUTEX_LOCK +// Helper RAII class to lock a mutex in constructor and unlock it in destructor (at the end of scope). +struct VmaMutexLock { + VMA_CLASS_NO_COPY(VmaMutexLock) public: - typedef T value_type; + VmaMutexLock(VMA_MUTEX& mutex, bool useMutex = true) : + m_pMutex(useMutex ? &mutex : VMA_NULL) + { + if (m_pMutex) { m_pMutex->Lock(); } + } + ~VmaMutexLock() { if (m_pMutex) { m_pMutex->Unlock(); } } - VmaVector(const AllocatorT& allocator) : - m_Allocator(allocator), - m_pArray(VMA_NULL), - m_Count(0), - m_Capacity(0) +private: + VMA_MUTEX* m_pMutex; +}; + +// Helper RAII class to lock a RW mutex in constructor and unlock it in destructor (at the end of scope), for reading. +struct VmaMutexLockRead +{ + VMA_CLASS_NO_COPY(VmaMutexLockRead) +public: + VmaMutexLockRead(VMA_RW_MUTEX& mutex, bool useMutex) : + m_pMutex(useMutex ? &mutex : VMA_NULL) { + if (m_pMutex) { m_pMutex->LockRead(); } } + ~VmaMutexLockRead() { if (m_pMutex) { m_pMutex->UnlockRead(); } } + +private: + VMA_RW_MUTEX* m_pMutex; +}; - VmaVector(size_t count, const AllocatorT& allocator) : - m_Allocator(allocator), - m_pArray(count ? (T*)VmaAllocateArray<T>(allocator.m_pCallbacks, count) : VMA_NULL), - m_Count(count), - m_Capacity(count) +// Helper RAII class to lock a RW mutex in constructor and unlock it in destructor (at the end of scope), for writing. +struct VmaMutexLockWrite +{ + VMA_CLASS_NO_COPY(VmaMutexLockWrite) +public: + VmaMutexLockWrite(VMA_RW_MUTEX& mutex, bool useMutex) + : m_pMutex(useMutex ? &mutex : VMA_NULL) { + if (m_pMutex) { m_pMutex->LockWrite(); } } + ~VmaMutexLockWrite() { if (m_pMutex) { m_pMutex->UnlockWrite(); } } - // This version of the constructor is here for compatibility with pre-C++14 std::vector. - // value is unused. - VmaVector(size_t count, const T& value, const AllocatorT& allocator) - : VmaVector(count, allocator) {} +private: + VMA_RW_MUTEX* m_pMutex; +}; - VmaVector(const VmaVector<T, AllocatorT>& src) : - m_Allocator(src.m_Allocator), - m_pArray(src.m_Count ? (T*)VmaAllocateArray<T>(src.m_Allocator.m_pCallbacks, src.m_Count) : VMA_NULL), - m_Count(src.m_Count), - m_Capacity(src.m_Count) +#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 +#endif // _VMA_MUTEX_LOCK + +#ifndef _VMA_ATOMIC_TRANSACTIONAL_INCREMENT +// An object that increments given atomic but decrements it back in the destructor unless Commit() is called. +template<typename T> +struct AtomicTransactionalIncrement +{ +public: + typedef std::atomic<T> AtomicT; + + ~AtomicTransactionalIncrement() { - if(m_Count != 0) - { - memcpy(m_pArray, src.m_pArray, m_Count * sizeof(T)); - } + if(m_Atomic) + --(*m_Atomic); } - ~VmaVector() + void Commit() { m_Atomic = nullptr; } + T Increment(AtomicT* atomic) { - VmaFree(m_Allocator.m_pCallbacks, m_pArray); + m_Atomic = atomic; + return m_Atomic->fetch_add(1); } - VmaVector& operator=(const VmaVector<T, AllocatorT>& rhs) +private: + AtomicT* m_Atomic = nullptr; +}; +#endif // _VMA_ATOMIC_TRANSACTIONAL_INCREMENT + +#ifndef _VMA_STL_ALLOCATOR +// STL-compatible allocator. +template<typename T> +struct VmaStlAllocator +{ + const VkAllocationCallbacks* const m_pCallbacks; + typedef T value_type; + + VmaStlAllocator(const VkAllocationCallbacks* pCallbacks) : m_pCallbacks(pCallbacks) {} + template<typename U> + VmaStlAllocator(const VmaStlAllocator<U>& src) : m_pCallbacks(src.m_pCallbacks) {} + VmaStlAllocator(const VmaStlAllocator&) = default; + VmaStlAllocator& operator=(const VmaStlAllocator&) = delete; + + T* allocate(size_t n) { return VmaAllocateArray<T>(m_pCallbacks, n); } + void deallocate(T* p, size_t n) { VmaFree(m_pCallbacks, p); } + + template<typename U> + bool operator==(const VmaStlAllocator<U>& rhs) const { - if(&rhs != this) - { - resize(rhs.m_Count); - if(m_Count != 0) - { - memcpy(m_pArray, rhs.m_pArray, m_Count * sizeof(T)); - } - } - return *this; + return m_pCallbacks == rhs.m_pCallbacks; + } + template<typename U> + bool operator!=(const VmaStlAllocator<U>& rhs) const + { + return m_pCallbacks != rhs.m_pCallbacks; } +}; +#endif // _VMA_STL_ALLOCATOR + +#ifndef _VMA_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<typename T, typename AllocatorT> +class VmaVector +{ +public: + typedef T value_type; + typedef T* iterator; + typedef const T* const_iterator; + + VmaVector(const AllocatorT& allocator); + VmaVector(size_t count, const AllocatorT& allocator); + // This version of the constructor is here for compatibility with pre-C++14 std::vector. + // value is unused. + VmaVector(size_t count, const T& value, const AllocatorT& allocator) : VmaVector(count, allocator) {} + VmaVector(const VmaVector<T, AllocatorT>& src); + VmaVector& operator=(const VmaVector& rhs); + ~VmaVector() { VmaFree(m_Allocator.m_pCallbacks, m_pArray); } bool empty() const { return m_Count == 0; } size_t size() const { return m_Count; } T* data() { return m_pArray; } + T& front() { 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* data() const { return m_pArray; } + const T& front() const { VMA_HEAVY_ASSERT(m_Count > 0); return m_pArray[0]; } + const T& back() const { VMA_HEAVY_ASSERT(m_Count > 0); return m_pArray[m_Count - 1]; } - 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]; - } + iterator begin() { return m_pArray; } + iterator end() { return m_pArray + m_Count; } + const_iterator cbegin() const { return m_pArray; } + const_iterator cend() const { return m_pArray + m_Count; } + const_iterator begin() const { return cbegin(); } + const_iterator end() const { return cend(); } - 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 pop_front() { VMA_HEAVY_ASSERT(m_Count > 0); remove(0); } + void pop_back() { VMA_HEAVY_ASSERT(m_Count > 0); resize(size() - 1); } + void push_front(const T& src) { insert(0, src); } - void reserve(size_t newCapacity, bool freeMemory = false) - { - newCapacity = VMA_MAX(newCapacity, m_Count); + void push_back(const T& src); + void reserve(size_t newCapacity, bool freeMemory = false); + void resize(size_t newCount); + void clear() { resize(0); } + void shrink_to_fit(); + void insert(size_t index, const T& src); + void remove(size_t index); - if((newCapacity < m_Capacity) && !freeMemory) - { - newCapacity = m_Capacity; - } + 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]; } - if(newCapacity != m_Capacity) - { - T* const newArray = newCapacity ? VmaAllocateArray<T>(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; - } +private: + AllocatorT m_Allocator; + T* m_pArray; + size_t m_Count; + size_t m_Capacity; +}; + +#ifndef _VMA_VECTOR_FUNCTIONS +template<typename T, typename AllocatorT> +VmaVector<T, AllocatorT>::VmaVector(const AllocatorT& allocator) + : m_Allocator(allocator), + m_pArray(VMA_NULL), + m_Count(0), + m_Capacity(0) {} + +template<typename T, typename AllocatorT> +VmaVector<T, AllocatorT>::VmaVector(size_t count, const AllocatorT& allocator) + : m_Allocator(allocator), + m_pArray(count ? (T*)VmaAllocateArray<T>(allocator.m_pCallbacks, count) : VMA_NULL), + m_Count(count), + m_Capacity(count) {} + +template<typename T, typename AllocatorT> +VmaVector<T, AllocatorT>::VmaVector(const VmaVector& src) + : m_Allocator(src.m_Allocator), + m_pArray(src.m_Count ? (T*)VmaAllocateArray<T>(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)); } +} - void resize(size_t newCount) +template<typename T, typename AllocatorT> +VmaVector<T, AllocatorT>& VmaVector<T, AllocatorT>::operator=(const VmaVector& rhs) +{ + if (&rhs != this) { - size_t newCapacity = m_Capacity; - if(newCount > m_Capacity) + resize(rhs.m_Count); + if (m_Count != 0) { - newCapacity = VMA_MAX(newCount, VMA_MAX(m_Capacity * 3 / 2, (size_t)8)); + memcpy(m_pArray, rhs.m_pArray, m_Count * sizeof(T)); } + } + return *this; +} - if(newCapacity != m_Capacity) - { - T* const newArray = newCapacity ? VmaAllocateArray<T>(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; - } +template<typename T, typename AllocatorT> +void VmaVector<T, AllocatorT>::push_back(const T& src) +{ + const size_t newIndex = size(); + resize(newIndex + 1); + m_pArray[newIndex] = src; +} - m_Count = newCount; - } +template<typename T, typename AllocatorT> +void VmaVector<T, AllocatorT>::reserve(size_t newCapacity, bool freeMemory) +{ + newCapacity = VMA_MAX(newCapacity, m_Count); - void clear() + if ((newCapacity < m_Capacity) && !freeMemory) { - resize(0); + newCapacity = m_Capacity; } - void shrink_to_fit() + if (newCapacity != m_Capacity) { - if(m_Capacity > m_Count) + T* const newArray = newCapacity ? VmaAllocateArray<T>(m_Allocator, newCapacity) : VMA_NULL; + if (m_Count != 0) { - T* newArray = VMA_NULL; - if(m_Count > 0) - { - newArray = VmaAllocateArray<T>(m_Allocator.m_pCallbacks, m_Count); - memcpy(newArray, m_pArray, m_Count * sizeof(T)); - } - VmaFree(m_Allocator.m_pCallbacks, m_pArray); - m_Capacity = m_Count; - m_pArray = newArray; + memcpy(newArray, m_pArray, m_Count * sizeof(T)); } + VmaFree(m_Allocator.m_pCallbacks, m_pArray); + m_Capacity = newCapacity; + m_pArray = newArray; } +} - void insert(size_t index, const T& src) +template<typename T, typename AllocatorT> +void VmaVector<T, AllocatorT>::resize(size_t newCount) +{ + size_t newCapacity = m_Capacity; + if (newCount > m_Capacity) { - 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; + newCapacity = VMA_MAX(newCount, VMA_MAX(m_Capacity * 3 / 2, (size_t)8)); } - void remove(size_t index) + if (newCapacity != m_Capacity) { - VMA_HEAVY_ASSERT(index < m_Count); - const size_t oldCount = size(); - if(index < oldCount - 1) + T* const newArray = newCapacity ? VmaAllocateArray<T>(m_Allocator.m_pCallbacks, newCapacity) : VMA_NULL; + const size_t elementsToCopy = VMA_MIN(m_Count, newCount); + if (elementsToCopy != 0) { - memmove(m_pArray + index, m_pArray + (index + 1), (oldCount - index - 1) * sizeof(T)); + memcpy(newArray, m_pArray, elementsToCopy * sizeof(T)); } - resize(oldCount - 1); + VmaFree(m_Allocator.m_pCallbacks, m_pArray); + m_Capacity = newCapacity; + m_pArray = newArray; } - void push_back(const T& src) - { - const size_t newIndex = size(); - resize(newIndex + 1); - m_pArray[newIndex] = src; - } + m_Count = newCount; +} - void pop_back() +template<typename T, typename AllocatorT> +void VmaVector<T, AllocatorT>::shrink_to_fit() +{ + if (m_Capacity > m_Count) { - VMA_HEAVY_ASSERT(m_Count > 0); - resize(size() - 1); + T* newArray = VMA_NULL; + if (m_Count > 0) + { + newArray = VmaAllocateArray<T>(m_Allocator.m_pCallbacks, m_Count); + memcpy(newArray, m_pArray, m_Count * sizeof(T)); + } + VmaFree(m_Allocator.m_pCallbacks, m_pArray); + m_Capacity = m_Count; + m_pArray = newArray; } +} - void push_front(const T& src) +template<typename T, typename AllocatorT> +void VmaVector<T, AllocatorT>::insert(size_t index, const T& src) +{ + VMA_HEAVY_ASSERT(index <= m_Count); + const size_t oldCount = size(); + resize(oldCount + 1); + if (index < oldCount) { - insert(0, src); + memmove(m_pArray + (index + 1), m_pArray + index, (oldCount - index) * sizeof(T)); } + m_pArray[index] = src; +} - void pop_front() +template<typename T, typename AllocatorT> +void VmaVector<T, AllocatorT>::remove(size_t index) +{ + VMA_HEAVY_ASSERT(index < m_Count); + const size_t oldCount = size(); + if (index < oldCount - 1) { - VMA_HEAVY_ASSERT(m_Count > 0); - remove(0); + memmove(m_pArray + index, m_pArray + (index + 1), (oldCount - index - 1) * sizeof(T)); } - - typedef T* iterator; - typedef const T* const_iterator; - - iterator begin() { return m_pArray; } - iterator end() { return m_pArray + m_Count; } - const_iterator cbegin() const { return m_pArray; } - const_iterator cend() const { return m_pArray + m_Count; } - const_iterator begin() const { return cbegin(); } - const_iterator end() const { return cend(); } - -private: - AllocatorT m_Allocator; - T* m_pArray; - size_t m_Count; - size_t m_Capacity; -}; + resize(oldCount - 1); +} +#endif // _VMA_VECTOR_FUNCTIONS template<typename T, typename allocatorT> static void VmaVectorInsert(VmaVector<T, allocatorT>& vec, size_t index, const T& item) @@ -5269,42 +4019,9 @@ static void VmaVectorRemove(VmaVector<T, allocatorT>& vec, size_t index) { vec.remove(index); } +#endif // _VMA_VECTOR -#endif // #if VMA_USE_STL_VECTOR - -template<typename CmpLess, typename VectorT> -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<typename CmpLess, typename VectorT> -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; -} - -//////////////////////////////////////////////////////////////////////////////// -// class VmaSmallVector - +#ifndef _VMA_SMALL_VECTOR /* This is a vector (a variable-sized array), optimized for the case when the array is small. @@ -5312,180 +4029,154 @@ It contains some number of elements in-place, which allows it to avoid heap allo when the actual number of elements is below that threshold. This allows normal "small" cases to be fast without losing generality for large inputs. */ - template<typename T, typename AllocatorT, size_t N> class VmaSmallVector { public: typedef T value_type; + typedef T* iterator; - VmaSmallVector(const AllocatorT& allocator) : - m_Count(0), - m_DynamicArray(allocator) - { - } - VmaSmallVector(size_t count, const AllocatorT& allocator) : - m_Count(count), - m_DynamicArray(count > N ? count : 0, allocator) - { - } + VmaSmallVector(const AllocatorT& allocator); + VmaSmallVector(size_t count, const AllocatorT& allocator); template<typename SrcT, typename SrcAllocatorT, size_t SrcN> - VmaSmallVector(const VmaSmallVector<SrcT, SrcAllocatorT, SrcN>& src) = delete; + VmaSmallVector(const VmaSmallVector<SrcT, SrcAllocatorT, SrcN>&) = delete; template<typename SrcT, typename SrcAllocatorT, size_t SrcN> - VmaSmallVector<T, AllocatorT, N>& operator=(const VmaSmallVector<SrcT, SrcAllocatorT, SrcN>& rhs) = delete; + VmaSmallVector<T, AllocatorT, N>& operator=(const VmaSmallVector<SrcT, SrcAllocatorT, SrcN>&) = delete; + ~VmaSmallVector() = default; bool empty() const { return m_Count == 0; } size_t size() const { return m_Count; } T* data() { return m_Count > N ? m_DynamicArray.data() : m_StaticArray; } + T& front() { VMA_HEAVY_ASSERT(m_Count > 0); return data()[0]; } + T& back() { VMA_HEAVY_ASSERT(m_Count > 0); return data()[m_Count - 1]; } const T* data() const { return m_Count > N ? m_DynamicArray.data() : m_StaticArray; } + const T& front() const { VMA_HEAVY_ASSERT(m_Count > 0); return data()[0]; } + const T& back() const { VMA_HEAVY_ASSERT(m_Count > 0); return data()[m_Count - 1]; } - T& operator[](size_t index) - { - VMA_HEAVY_ASSERT(index < m_Count); - return data()[index]; - } - const T& operator[](size_t index) const - { - VMA_HEAVY_ASSERT(index < m_Count); - return data()[index]; - } + iterator begin() { return data(); } + iterator end() { return data() + m_Count; } - T& front() - { - VMA_HEAVY_ASSERT(m_Count > 0); - return data()[0]; - } - const T& front() const - { - VMA_HEAVY_ASSERT(m_Count > 0); - return data()[0]; - } - T& back() - { - VMA_HEAVY_ASSERT(m_Count > 0); - return data()[m_Count - 1]; - } - const T& back() const - { - VMA_HEAVY_ASSERT(m_Count > 0); - return data()[m_Count - 1]; - } + void pop_front() { VMA_HEAVY_ASSERT(m_Count > 0); remove(0); } + void pop_back() { VMA_HEAVY_ASSERT(m_Count > 0); resize(size() - 1); } + void push_front(const T& src) { insert(0, src); } - void resize(size_t newCount, bool freeMemory = false) - { - if(newCount > N && m_Count > N) - { - // Any direction, staying in m_DynamicArray - m_DynamicArray.resize(newCount); - if(freeMemory) - { - m_DynamicArray.shrink_to_fit(); - } - } - else if(newCount > N && m_Count <= N) - { - // Growing, moving from m_StaticArray to m_DynamicArray - m_DynamicArray.resize(newCount); - if(m_Count > 0) - { - memcpy(m_DynamicArray.data(), m_StaticArray, m_Count * sizeof(T)); - } - } - else if(newCount <= N && m_Count > N) - { - // Shrinking, moving from m_DynamicArray to m_StaticArray - if(newCount > 0) - { - memcpy(m_StaticArray, m_DynamicArray.data(), newCount * sizeof(T)); - } - m_DynamicArray.resize(0); - if(freeMemory) - { - m_DynamicArray.shrink_to_fit(); - } - } - else - { - // Any direction, staying in m_StaticArray - nothing to do here - } - m_Count = newCount; - } + void push_back(const T& src); + void resize(size_t newCount, bool freeMemory = false); + void clear(bool freeMemory = false); + void insert(size_t index, const T& src); + void remove(size_t index); - void clear(bool freeMemory = false) + T& operator[](size_t index) { VMA_HEAVY_ASSERT(index < m_Count); return data()[index]; } + const T& operator[](size_t index) const { VMA_HEAVY_ASSERT(index < m_Count); return data()[index]; } + +private: + size_t m_Count; + T m_StaticArray[N]; // Used when m_Size <= N + VmaVector<T, AllocatorT> m_DynamicArray; // Used when m_Size > N +}; + +#ifndef _VMA_SMALL_VECTOR_FUNCTIONS +template<typename T, typename AllocatorT, size_t N> +VmaSmallVector<T, AllocatorT, N>::VmaSmallVector(const AllocatorT& allocator) + : m_Count(0), + m_DynamicArray(allocator) {} + +template<typename T, typename AllocatorT, size_t N> +VmaSmallVector<T, AllocatorT, N>::VmaSmallVector(size_t count, const AllocatorT& allocator) + : m_Count(count), + m_DynamicArray(count > N ? count : 0, allocator) {} + +template<typename T, typename AllocatorT, size_t N> +void VmaSmallVector<T, AllocatorT, N>::push_back(const T& src) +{ + resize(m_Count + 1); + data()[m_Count] = src; +} + +template<typename T, typename AllocatorT, size_t N> +void VmaSmallVector<T, AllocatorT, N>::resize(size_t newCount, bool freeMemory) +{ + if (newCount > N && m_Count > N) { - m_DynamicArray.clear(); - if(freeMemory) + // Any direction, staying in m_DynamicArray + m_DynamicArray.resize(newCount); + if (freeMemory) { m_DynamicArray.shrink_to_fit(); } - m_Count = 0; } - - void insert(size_t index, const T& src) + else if (newCount > N && m_Count <= N) { - VMA_HEAVY_ASSERT(index <= m_Count); - const size_t oldCount = size(); - resize(oldCount + 1); - T* const dataPtr = data(); - if(index < oldCount) + // Growing, moving from m_StaticArray to m_DynamicArray + m_DynamicArray.resize(newCount); + if (m_Count > 0) { - // I know, this could be more optimal for case where memmove can be memcpy directly from m_StaticArray to m_DynamicArray. - memmove(dataPtr + (index + 1), dataPtr + index, (oldCount - index) * sizeof(T)); + memcpy(m_DynamicArray.data(), m_StaticArray, m_Count * sizeof(T)); } - dataPtr[index] = src; } - - void remove(size_t index) + else if (newCount <= N && m_Count > N) { - VMA_HEAVY_ASSERT(index < m_Count); - const size_t oldCount = size(); - if(index < oldCount - 1) + // Shrinking, moving from m_DynamicArray to m_StaticArray + if (newCount > 0) + { + memcpy(m_StaticArray, m_DynamicArray.data(), newCount * sizeof(T)); + } + m_DynamicArray.resize(0); + if (freeMemory) { - // I know, this could be more optimal for case where memmove can be memcpy directly from m_DynamicArray to m_StaticArray. - T* const dataPtr = data(); - memmove(dataPtr + index, dataPtr + (index + 1), (oldCount - index - 1) * sizeof(T)); + m_DynamicArray.shrink_to_fit(); } - resize(oldCount - 1); } - - void push_back(const T& src) + else { - const size_t newIndex = size(); - resize(newIndex + 1); - data()[newIndex] = src; + // Any direction, staying in m_StaticArray - nothing to do here } + m_Count = newCount; +} - void pop_back() +template<typename T, typename AllocatorT, size_t N> +void VmaSmallVector<T, AllocatorT, N>::clear(bool freeMemory) +{ + m_DynamicArray.clear(); + if (freeMemory) { - VMA_HEAVY_ASSERT(m_Count > 0); - resize(size() - 1); + m_DynamicArray.shrink_to_fit(); } + m_Count = 0; +} - void push_front(const T& src) +template<typename T, typename AllocatorT, size_t N> +void VmaSmallVector<T, AllocatorT, N>::insert(size_t index, const T& src) +{ + VMA_HEAVY_ASSERT(index <= m_Count); + const size_t oldCount = size(); + resize(oldCount + 1); + T* const dataPtr = data(); + if (index < oldCount) { - insert(0, src); + // I know, this could be more optimal for case where memmove can be memcpy directly from m_StaticArray to m_DynamicArray. + memmove(dataPtr + (index + 1), dataPtr + index, (oldCount - index) * sizeof(T)); } + dataPtr[index] = src; +} - void pop_front() +template<typename T, typename AllocatorT, size_t N> +void VmaSmallVector<T, AllocatorT, N>::remove(size_t index) +{ + VMA_HEAVY_ASSERT(index < m_Count); + const size_t oldCount = size(); + if (index < oldCount - 1) { - VMA_HEAVY_ASSERT(m_Count > 0); - remove(0); + // I know, this could be more optimal for case where memmove can be memcpy directly from m_DynamicArray to m_StaticArray. + T* const dataPtr = data(); + memmove(dataPtr + index, dataPtr + (index + 1), (oldCount - index - 1) * sizeof(T)); } + resize(oldCount - 1); +} +#endif // _VMA_SMALL_VECTOR_FUNCTIONS +#endif // _VMA_SMALL_VECTOR - typedef T* iterator; - - iterator begin() { return data(); } - iterator end() { return data() + m_Count; } - -private: - size_t m_Count; - T m_StaticArray[N]; // Used when m_Size <= N - VmaVector<T, AllocatorT> m_DynamicArray; // Used when m_Size > N -}; - -//////////////////////////////////////////////////////////////////////////////// -// class VmaPoolAllocator - +#ifndef _VMA_POOL_ALLOCATOR /* 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 @@ -5498,7 +4189,7 @@ class VmaPoolAllocator public: VmaPoolAllocator(const VkAllocationCallbacks* pAllocationCallbacks, uint32_t firstBlockCapacity); ~VmaPoolAllocator(); - template<typename... Types> T* Alloc(Types... args); + template<typename... Types> T* Alloc(Types&&... args); void Free(T* ptr); private: @@ -5507,7 +4198,6 @@ private: uint32_t NextFreeIndex; alignas(T) char Value[sizeof(T)]; }; - struct ItemBlock { Item* pItems; @@ -5517,14 +4207,15 @@ private: const VkAllocationCallbacks* m_pAllocationCallbacks; const uint32_t m_FirstBlockCapacity; - VmaVector< ItemBlock, VmaStlAllocator<ItemBlock> > m_ItemBlocks; + VmaVector<ItemBlock, VmaStlAllocator<ItemBlock>> m_ItemBlocks; ItemBlock& CreateNewBlock(); }; +#ifndef _VMA_POOL_ALLOCATOR_FUNCTIONS template<typename T> -VmaPoolAllocator<T>::VmaPoolAllocator(const VkAllocationCallbacks* pAllocationCallbacks, uint32_t firstBlockCapacity) : - m_pAllocationCallbacks(pAllocationCallbacks), +VmaPoolAllocator<T>::VmaPoolAllocator(const VkAllocationCallbacks* pAllocationCallbacks, uint32_t firstBlockCapacity) + : m_pAllocationCallbacks(pAllocationCallbacks), m_FirstBlockCapacity(firstBlockCapacity), m_ItemBlocks(VmaStlAllocator<ItemBlock>(pAllocationCallbacks)) { @@ -5534,19 +4225,19 @@ VmaPoolAllocator<T>::VmaPoolAllocator(const VkAllocationCallbacks* pAllocationCa template<typename T> VmaPoolAllocator<T>::~VmaPoolAllocator() { - for(size_t i = m_ItemBlocks.size(); i--; ) + for (size_t i = m_ItemBlocks.size(); i--;) vma_delete_array(m_pAllocationCallbacks, m_ItemBlocks[i].pItems, m_ItemBlocks[i].Capacity); m_ItemBlocks.clear(); } template<typename T> -template<typename... Types> T* VmaPoolAllocator<T>::Alloc(Types... args) +template<typename... Types> T* VmaPoolAllocator<T>::Alloc(Types&&... args) { - for(size_t i = m_ItemBlocks.size(); i--; ) + 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) + if (block.FirstFreeIndex != UINT32_MAX) { Item* const pItem = &block.pItems[block.FirstFreeIndex]; block.FirstFreeIndex = pItem->NextFreeIndex; @@ -5561,7 +4252,7 @@ template<typename... Types> T* VmaPoolAllocator<T>::Alloc(Types... args) Item* const pItem = &newBlock.pItems[0]; newBlock.FirstFreeIndex = pItem->NextFreeIndex; T* result = (T*)&pItem->Value; - new(result)T(std::forward<Types>(args)...); // Explicit constructor call. + new(result) T(std::forward<Types>(args)...); // Explicit constructor call. return result; } @@ -5569,7 +4260,7 @@ template<typename T> void VmaPoolAllocator<T>::Free(T* ptr) { // Search all memory blocks to find ptr. - for(size_t i = m_ItemBlocks.size(); i--; ) + for (size_t i = m_ItemBlocks.size(); i--; ) { ItemBlock& block = m_ItemBlocks[i]; @@ -5578,7 +4269,7 @@ void VmaPoolAllocator<T>::Free(T* ptr) memcpy(&pItemPtr, &ptr, sizeof(pItemPtr)); // Check if pItemPtr is in address range of this block. - if((pItemPtr >= block.pItems) && (pItemPtr < block.pItems + block.Capacity)) + if ((pItemPtr >= block.pItems) && (pItemPtr < block.pItems + block.Capacity)) { ptr->~T(); // Explicit destructor call. const uint32_t index = static_cast<uint32_t>(pItemPtr - block.pItems); @@ -5596,29 +4287,25 @@ typename VmaPoolAllocator<T>::ItemBlock& VmaPoolAllocator<T>::CreateNewBlock() const uint32_t newBlockCapacity = m_ItemBlocks.empty() ? m_FirstBlockCapacity : m_ItemBlocks.back().Capacity * 3 / 2; - const ItemBlock newBlock = { + const ItemBlock newBlock = + { vma_new_array(m_pAllocationCallbacks, Item, newBlockCapacity), newBlockCapacity, - 0 }; + 0 + }; m_ItemBlocks.push_back(newBlock); // Setup singly-linked list of all free items in this block. - for(uint32_t i = 0; i < newBlockCapacity - 1; ++i) + for (uint32_t i = 0; i < newBlockCapacity - 1; ++i) newBlock.pItems[i].NextFreeIndex = i + 1; newBlock.pItems[newBlockCapacity - 1].NextFreeIndex = UINT32_MAX; return m_ItemBlocks.back(); } +#endif // _VMA_POOL_ALLOCATOR_FUNCTIONS +#endif // _VMA_POOL_ALLOCATOR -//////////////////////////////////////////////////////////////////////////////// -// class VmaRawList, VmaList - -#if VMA_USE_STL_LIST - -#define VmaList std::list - -#else // #if VMA_USE_STL_LIST - +#ifndef _VMA_RAW_LIST template<typename T> struct VmaListItem { @@ -5636,32 +4323,33 @@ public: typedef VmaListItem<T> ItemType; VmaRawList(const VkAllocationCallbacks* pAllocationCallbacks); - ~VmaRawList(); - void Clear(); + // Intentionally not calling Clear, because that would be unnecessary + // computations to return all items to m_ItemAllocator as free. + ~VmaRawList() = default; size_t GetCount() const { return m_Count; } bool IsEmpty() const { return m_Count == 0; } ItemType* Front() { return m_pFront; } - const ItemType* Front() const { return m_pFront; } ItemType* Back() { return m_pBack; } + const ItemType* Front() const { return m_pFront; } const ItemType* Back() const { return m_pBack; } - ItemType* PushBack(); ItemType* PushFront(); - ItemType* PushBack(const T& value); + ItemType* PushBack(); ItemType* PushFront(const T& value); - void PopBack(); + ItemType* PushBack(const T& value); void PopFront(); + void PopBack(); // Item can be null - it means PushBack. ItemType* InsertBefore(ItemType* pItem); // Item can be null - it means PushFront. ItemType* InsertAfter(ItemType* pItem); - ItemType* InsertBefore(ItemType* pItem, const T& value); ItemType* InsertAfter(ItemType* pItem, const T& value); + void Clear(); void Remove(ItemType* pItem); private: @@ -5672,39 +4360,35 @@ private: size_t m_Count; }; +#ifndef _VMA_RAW_LIST_FUNCTIONS template<typename T> -VmaRawList<T>::VmaRawList(const VkAllocationCallbacks* pAllocationCallbacks) : - m_pAllocationCallbacks(pAllocationCallbacks), +VmaRawList<T>::VmaRawList(const VkAllocationCallbacks* pAllocationCallbacks) + : m_pAllocationCallbacks(pAllocationCallbacks), m_ItemAllocator(pAllocationCallbacks, 128), m_pFront(VMA_NULL), m_pBack(VMA_NULL), - m_Count(0) -{ -} - -template<typename T> -VmaRawList<T>::~VmaRawList() -{ - // Intentionally not calling Clear, because that would be unnecessary - // computations to return all items to m_ItemAllocator as free. -} + m_Count(0) {} template<typename T> -void VmaRawList<T>::Clear() +VmaListItem<T>* VmaRawList<T>::PushFront() { - if(IsEmpty() == false) + ItemType* const pNewItem = m_ItemAllocator.Alloc(); + pNewItem->pPrev = VMA_NULL; + if (IsEmpty()) { - ItemType* pItem = m_pBack; - while(pItem != VMA_NULL) - { - ItemType* const pPrevItem = pItem->pPrev; - m_ItemAllocator.Free(pItem); - pItem = pPrevItem; - } - m_pFront = VMA_NULL; - m_pBack = VMA_NULL; - m_Count = 0; + pNewItem->pNext = VMA_NULL; + m_pFront = pNewItem; + m_pBack = pNewItem; + m_Count = 1; + } + else + { + pNewItem->pNext = m_pFront; + m_pFront->pPrev = pNewItem; + m_pFront = pNewItem; + ++m_Count; } + return pNewItem; } template<typename T> @@ -5730,24 +4414,10 @@ VmaListItem<T>* VmaRawList<T>::PushBack() } template<typename T> -VmaListItem<T>* VmaRawList<T>::PushFront() +VmaListItem<T>* VmaRawList<T>::PushFront(const T& value) { - ItemType* const pNewItem = m_ItemAllocator.Alloc(); - pNewItem->pPrev = VMA_NULL; - if(IsEmpty()) - { - pNewItem->pNext = VMA_NULL; - m_pFront = pNewItem; - m_pBack = pNewItem; - m_Count = 1; - } - else - { - pNewItem->pNext = m_pFront; - m_pFront->pPrev = pNewItem; - m_pFront = pNewItem; - ++m_Count; - } + ItemType* const pNewItem = PushFront(); + pNewItem->Value = value; return pNewItem; } @@ -5760,11 +4430,18 @@ VmaListItem<T>* VmaRawList<T>::PushBack(const T& value) } template<typename T> -VmaListItem<T>* VmaRawList<T>::PushFront(const T& value) +void VmaRawList<T>::PopFront() { - ItemType* const pNewItem = PushFront(); - pNewItem->Value = value; - return pNewItem; + VMA_HEAVY_ASSERT(m_Count > 0); + ItemType* const pFrontItem = m_pFront; + ItemType* const pNextItem = pFrontItem->pNext; + if (pNextItem != VMA_NULL) + { + pNextItem->pPrev = VMA_NULL; + } + m_pFront = pNextItem; + m_ItemAllocator.Free(pFrontItem); + --m_Count; } template<typename T> @@ -5783,18 +4460,21 @@ void VmaRawList<T>::PopBack() } template<typename T> -void VmaRawList<T>::PopFront() +void VmaRawList<T>::Clear() { - VMA_HEAVY_ASSERT(m_Count > 0); - ItemType* const pFrontItem = m_pFront; - ItemType* const pNextItem = pFrontItem->pNext; - if(pNextItem != VMA_NULL) + if (IsEmpty() == false) { - pNextItem->pPrev = VMA_NULL; + ItemType* pItem = m_pBack; + while (pItem != VMA_NULL) + { + ItemType* const pPrevItem = pItem->pPrev; + m_ItemAllocator.Free(pItem); + pItem = pPrevItem; + } + m_pFront = VMA_NULL; + m_pBack = VMA_NULL; + m_Count = 0; } - m_pFront = pNextItem; - m_ItemAllocator.Free(pFrontItem); - --m_Count; } template<typename T> @@ -5894,173 +4574,123 @@ VmaListItem<T>* VmaRawList<T>::InsertAfter(ItemType* pItem, const T& value) newItem->Value = value; return newItem; } +#endif // _VMA_RAW_LIST_FUNCTIONS +#endif // _VMA_RAW_LIST +#ifndef _VMA_LIST template<typename T, typename AllocatorT> class VmaList { VMA_CLASS_NO_COPY(VmaList) public: + class reverse_iterator; + class const_iterator; + class const_reverse_iterator; + class iterator { + friend class VmaList<T, AllocatorT>; public: - iterator() : - m_pList(VMA_NULL), - m_pItem(VMA_NULL) - { - } + iterator() : m_pList(VMA_NULL), m_pItem(VMA_NULL) {} + iterator(const reverse_iterator& src) : m_pList(src.m_pList), m_pItem(src.m_pItem) {} - T& operator*() const - { - VMA_HEAVY_ASSERT(m_pItem != VMA_NULL); - return m_pItem->Value; - } - T* operator->() const - { - VMA_HEAVY_ASSERT(m_pItem != VMA_NULL); - return &m_pItem->Value; - } + T& operator*() const { VMA_HEAVY_ASSERT(m_pItem != VMA_NULL); return m_pItem->Value; } + T* operator->() const { VMA_HEAVY_ASSERT(m_pItem != VMA_NULL); return &m_pItem->Value; } - iterator& operator++() - { - VMA_HEAVY_ASSERT(m_pItem != VMA_NULL); - m_pItem = m_pItem->pNext; - return *this; - } - iterator& operator--() - { - if(m_pItem != VMA_NULL) - { - m_pItem = m_pItem->pPrev; - } - else - { - VMA_HEAVY_ASSERT(!m_pList->IsEmpty()); - m_pItem = m_pList->Back(); - } - return *this; - } + bool operator==(const iterator& rhs) const { VMA_HEAVY_ASSERT(m_pList == rhs.m_pList); return m_pItem == rhs.m_pItem; } + bool operator!=(const iterator& rhs) const { VMA_HEAVY_ASSERT(m_pList == rhs.m_pList); return m_pItem != rhs.m_pItem; } - iterator operator++(int) - { - iterator result = *this; - ++*this; - return result; - } - iterator operator--(int) - { - iterator result = *this; - --*this; - return result; - } + iterator operator++(int) { iterator result = *this; ++*this; return result; } + iterator operator--(int) { iterator result = *this; --*this; return result; } - bool operator==(const iterator& rhs) const - { - VMA_HEAVY_ASSERT(m_pList == rhs.m_pList); - return m_pItem == rhs.m_pItem; - } - bool operator!=(const iterator& rhs) const - { - VMA_HEAVY_ASSERT(m_pList == rhs.m_pList); - return m_pItem != rhs.m_pItem; - } + iterator& operator++() { VMA_HEAVY_ASSERT(m_pItem != VMA_NULL); m_pItem = m_pItem->pNext; return *this; } + iterator& operator--(); private: VmaRawList<T>* m_pList; VmaListItem<T>* m_pItem; - iterator(VmaRawList<T>* pList, VmaListItem<T>* pItem) : - m_pList(pList), - m_pItem(pItem) - { - } - - friend class VmaList<T, AllocatorT>; + iterator(VmaRawList<T>* pList, VmaListItem<T>* pItem) : m_pList(pList), m_pItem(pItem) {} }; + class reverse_iterator + { + friend class VmaList<T, AllocatorT>; + public: + reverse_iterator() : m_pList(VMA_NULL), m_pItem(VMA_NULL) {} + reverse_iterator(const iterator& src) : m_pList(src.m_pList), m_pItem(src.m_pItem) {} + + T& operator*() const { VMA_HEAVY_ASSERT(m_pItem != VMA_NULL); return m_pItem->Value; } + T* operator->() const { VMA_HEAVY_ASSERT(m_pItem != VMA_NULL); return &m_pItem->Value; } + + bool operator==(const reverse_iterator& rhs) const { VMA_HEAVY_ASSERT(m_pList == rhs.m_pList); return m_pItem == rhs.m_pItem; } + bool operator!=(const reverse_iterator& rhs) const { VMA_HEAVY_ASSERT(m_pList == rhs.m_pList); return m_pItem != rhs.m_pItem; } + reverse_iterator operator++(int) { reverse_iterator result = *this; ++* this; return result; } + reverse_iterator operator--(int) { reverse_iterator result = *this; --* this; return result; } + + reverse_iterator& operator++() { VMA_HEAVY_ASSERT(m_pItem != VMA_NULL); m_pItem = m_pItem->pPrev; return *this; } + reverse_iterator& operator--(); + + private: + VmaRawList<T>* m_pList; + VmaListItem<T>* m_pItem; + + reverse_iterator(VmaRawList<T>* pList, VmaListItem<T>* pItem) : m_pList(pList), m_pItem(pItem) {} + }; class const_iterator { + friend class VmaList<T, AllocatorT>; public: - const_iterator() : - m_pList(VMA_NULL), - m_pItem(VMA_NULL) - { - } + const_iterator() : m_pList(VMA_NULL), m_pItem(VMA_NULL) {} + const_iterator(const iterator& src) : m_pList(src.m_pList), m_pItem(src.m_pItem) {} + const_iterator(const reverse_iterator& src) : m_pList(src.m_pList), m_pItem(src.m_pItem) {} - const_iterator(const iterator& src) : - m_pList(src.m_pList), - m_pItem(src.m_pItem) - { - } + const T& operator*() const { VMA_HEAVY_ASSERT(m_pItem != VMA_NULL); return m_pItem->Value; } + const T* operator->() const { VMA_HEAVY_ASSERT(m_pItem != VMA_NULL); return &m_pItem->Value; } - const T& operator*() const - { - VMA_HEAVY_ASSERT(m_pItem != VMA_NULL); - return m_pItem->Value; - } - const T* operator->() const - { - VMA_HEAVY_ASSERT(m_pItem != VMA_NULL); - return &m_pItem->Value; - } + bool operator==(const const_iterator& rhs) const { VMA_HEAVY_ASSERT(m_pList == rhs.m_pList); return m_pItem == rhs.m_pItem; } + bool operator!=(const const_iterator& rhs) const { VMA_HEAVY_ASSERT(m_pList == rhs.m_pList); return m_pItem != rhs.m_pItem; } - const_iterator& operator++() - { - VMA_HEAVY_ASSERT(m_pItem != VMA_NULL); - m_pItem = m_pItem->pNext; - return *this; - } - const_iterator& operator--() - { - if(m_pItem != VMA_NULL) - { - m_pItem = m_pItem->pPrev; - } - else - { - VMA_HEAVY_ASSERT(!m_pList->IsEmpty()); - m_pItem = m_pList->Back(); - } - return *this; - } + const_iterator operator++(int) { const_iterator result = *this; ++* this; return result; } + const_iterator operator--(int) { const_iterator result = *this; --* this; return result; } - const_iterator operator++(int) - { - const_iterator result = *this; - ++*this; - return result; - } - const_iterator operator--(int) - { - const_iterator result = *this; - --*this; - return result; - } - - bool operator==(const const_iterator& rhs) const - { - VMA_HEAVY_ASSERT(m_pList == rhs.m_pList); - return m_pItem == rhs.m_pItem; - } - bool operator!=(const const_iterator& rhs) const - { - VMA_HEAVY_ASSERT(m_pList == rhs.m_pList); - return m_pItem != rhs.m_pItem; - } + const_iterator& operator++() { VMA_HEAVY_ASSERT(m_pItem != VMA_NULL); m_pItem = m_pItem->pNext; return *this; } + const_iterator& operator--(); private: - const_iterator(const VmaRawList<T>* pList, const VmaListItem<T>* pItem) : - m_pList(pList), - m_pItem(pItem) - { - } - const VmaRawList<T>* m_pList; const VmaListItem<T>* m_pItem; + const_iterator(const VmaRawList<T>* pList, const VmaListItem<T>* pItem) : m_pList(pList), m_pItem(pItem) {} + }; + class const_reverse_iterator + { friend class VmaList<T, AllocatorT>; + public: + const_reverse_iterator() : m_pList(VMA_NULL), m_pItem(VMA_NULL) {} + const_reverse_iterator(const reverse_iterator& src) : m_pList(src.m_pList), m_pItem(src.m_pItem) {} + const_reverse_iterator(const iterator& src) : m_pList(src.m_pList), m_pItem(src.m_pItem) {} + + const T& operator*() const { VMA_HEAVY_ASSERT(m_pItem != VMA_NULL); return m_pItem->Value; } + const T* operator->() const { VMA_HEAVY_ASSERT(m_pItem != VMA_NULL); return &m_pItem->Value; } + + bool operator==(const const_reverse_iterator& rhs) const { VMA_HEAVY_ASSERT(m_pList == rhs.m_pList); return m_pItem == rhs.m_pItem; } + bool operator!=(const const_reverse_iterator& rhs) const { VMA_HEAVY_ASSERT(m_pList == rhs.m_pList); return m_pItem != rhs.m_pItem; } + + const_reverse_iterator operator++(int) { const_reverse_iterator result = *this; ++* this; return result; } + const_reverse_iterator operator--(int) { const_reverse_iterator result = *this; --* this; return result; } + + const_reverse_iterator& operator++() { VMA_HEAVY_ASSERT(m_pItem != VMA_NULL); m_pItem = m_pItem->pPrev; return *this; } + const_reverse_iterator& operator--(); + + private: + const VmaRawList<T>* m_pList; + const VmaListItem<T>* m_pItem; + + const_reverse_iterator(const VmaRawList<T>* pList, const VmaListItem<T>* pItem) : m_pList(pList), m_pItem(pItem) {} }; - VmaList(const AllocatorT& allocator) : m_RawList(allocator.m_pCallbacks) { } + VmaList(const AllocatorT& allocator) : m_RawList(allocator.m_pCallbacks) {} bool empty() const { return m_RawList.IsEmpty(); } size_t size() const { return m_RawList.GetCount(); } @@ -6074,20 +4704,89 @@ public: const_iterator begin() const { return cbegin(); } const_iterator end() const { return cend(); } - void clear() { m_RawList.Clear(); } + reverse_iterator rbegin() { return reverse_iterator(&m_RawList, m_RawList.Back()); } + reverse_iterator rend() { return reverse_iterator(&m_RawList, VMA_NULL); } + + const_reverse_iterator crbegin() { return const_reverse_iterator(&m_RawList, m_RawList.Back()); } + const_reverse_iterator crend() { return const_reverse_iterator(&m_RawList, VMA_NULL); } + + const_reverse_iterator rbegin() const { return crbegin(); } + const_reverse_iterator rend() const { return crend(); } + void push_back(const T& value) { m_RawList.PushBack(value); } - void erase(iterator it) { m_RawList.Remove(it.m_pItem); } iterator insert(iterator it, const T& value) { return iterator(&m_RawList, m_RawList.InsertBefore(it.m_pItem, value)); } + void clear() { m_RawList.Clear(); } + void erase(iterator it) { m_RawList.Remove(it.m_pItem); } + private: VmaRawList<T> m_RawList; }; -#endif // #if VMA_USE_STL_LIST +#ifndef _VMA_LIST_FUNCTIONS +template<typename T, typename AllocatorT> +typename VmaList<T, AllocatorT>::iterator& VmaList<T, AllocatorT>::iterator::operator--() +{ + if (m_pItem != VMA_NULL) + { + m_pItem = m_pItem->pPrev; + } + else + { + VMA_HEAVY_ASSERT(!m_pList->IsEmpty()); + m_pItem = m_pList->Back(); + } + return *this; +} + +template<typename T, typename AllocatorT> +typename VmaList<T, AllocatorT>::reverse_iterator& VmaList<T, AllocatorT>::reverse_iterator::operator--() +{ + if (m_pItem != VMA_NULL) + { + m_pItem = m_pItem->pNext; + } + else + { + VMA_HEAVY_ASSERT(!m_pList->IsEmpty()); + m_pItem = m_pList->Front(); + } + return *this; +} -//////////////////////////////////////////////////////////////////////////////// -// class VmaIntrusiveLinkedList +template<typename T, typename AllocatorT> +typename VmaList<T, AllocatorT>::const_iterator& VmaList<T, AllocatorT>::const_iterator::operator--() +{ + if (m_pItem != VMA_NULL) + { + m_pItem = m_pItem->pPrev; + } + else + { + VMA_HEAVY_ASSERT(!m_pList->IsEmpty()); + m_pItem = m_pList->Back(); + } + return *this; +} +template<typename T, typename AllocatorT> +typename VmaList<T, AllocatorT>::const_reverse_iterator& VmaList<T, AllocatorT>::const_reverse_iterator::operator--() +{ + if (m_pItem != VMA_NULL) + { + m_pItem = m_pItem->pNext; + } + else + { + VMA_HEAVY_ASSERT(!m_pList->IsEmpty()); + m_pItem = m_pList->Back(); + } + return *this; +} +#endif // _VMA_LIST_FUNCTIONS +#endif // _VMA_LIST + +#ifndef _VMA_INTRUSIVE_LINKED_LIST /* Expected interface of ItemTypeTraits: struct MyItemTypeTraits @@ -6106,226 +4805,264 @@ public: typedef typename ItemTypeTraits::ItemType ItemType; static ItemType* GetPrev(const ItemType* item) { return ItemTypeTraits::GetPrev(item); } static ItemType* GetNext(const ItemType* item) { return ItemTypeTraits::GetNext(item); } + // Movable, not copyable. - VmaIntrusiveLinkedList() { } - VmaIntrusiveLinkedList(const VmaIntrusiveLinkedList<ItemTypeTraits>& src) = delete; - VmaIntrusiveLinkedList(VmaIntrusiveLinkedList<ItemTypeTraits>&& src) : - m_Front(src.m_Front), m_Back(src.m_Back), m_Count(src.m_Count) + VmaIntrusiveLinkedList() = default; + VmaIntrusiveLinkedList(VmaIntrusiveLinkedList && src); + VmaIntrusiveLinkedList(const VmaIntrusiveLinkedList&) = delete; + VmaIntrusiveLinkedList& operator=(VmaIntrusiveLinkedList&& src); + VmaIntrusiveLinkedList& operator=(const VmaIntrusiveLinkedList&) = delete; + ~VmaIntrusiveLinkedList() { VMA_HEAVY_ASSERT(IsEmpty()); } + + size_t GetCount() const { return m_Count; } + bool IsEmpty() const { return m_Count == 0; } + ItemType* Front() { return m_Front; } + ItemType* Back() { return m_Back; } + const ItemType* Front() const { return m_Front; } + const ItemType* Back() const { return m_Back; } + + void PushBack(ItemType* item); + void PushFront(ItemType* item); + ItemType* PopBack(); + ItemType* PopFront(); + + // MyItem can be null - it means PushBack. + void InsertBefore(ItemType* existingItem, ItemType* newItem); + // MyItem can be null - it means PushFront. + void InsertAfter(ItemType* existingItem, ItemType* newItem); + void Remove(ItemType* item); + void RemoveAll(); + +private: + ItemType* m_Front = VMA_NULL; + ItemType* m_Back = VMA_NULL; + size_t m_Count = 0; +}; + +#ifndef _VMA_INTRUSIVE_LINKED_LIST_FUNCTIONS +template<typename ItemTypeTraits> +VmaIntrusiveLinkedList<ItemTypeTraits>::VmaIntrusiveLinkedList(VmaIntrusiveLinkedList&& src) + : m_Front(src.m_Front), m_Back(src.m_Back), m_Count(src.m_Count) +{ + src.m_Front = src.m_Back = VMA_NULL; + src.m_Count = 0; +} + +template<typename ItemTypeTraits> +VmaIntrusiveLinkedList<ItemTypeTraits>& VmaIntrusiveLinkedList<ItemTypeTraits>::operator=(VmaIntrusiveLinkedList&& src) +{ + if (&src != this) { + VMA_HEAVY_ASSERT(IsEmpty()); + m_Front = src.m_Front; + m_Back = src.m_Back; + m_Count = src.m_Count; src.m_Front = src.m_Back = VMA_NULL; src.m_Count = 0; } - ~VmaIntrusiveLinkedList() + return *this; +} + +template<typename ItemTypeTraits> +void VmaIntrusiveLinkedList<ItemTypeTraits>::PushBack(ItemType* item) +{ + VMA_HEAVY_ASSERT(ItemTypeTraits::GetPrev(item) == VMA_NULL && ItemTypeTraits::GetNext(item) == VMA_NULL); + if (IsEmpty()) { - VMA_HEAVY_ASSERT(IsEmpty()); + m_Front = item; + m_Back = item; + m_Count = 1; } - VmaIntrusiveLinkedList<ItemTypeTraits>& operator=(const VmaIntrusiveLinkedList<ItemTypeTraits>& src) = delete; - VmaIntrusiveLinkedList<ItemTypeTraits>& operator=(VmaIntrusiveLinkedList<ItemTypeTraits>&& src) + else { - if(&src != this) - { - VMA_HEAVY_ASSERT(IsEmpty()); - m_Front = src.m_Front; - m_Back = src.m_Back; - m_Count = src.m_Count; - src.m_Front = src.m_Back = VMA_NULL; - src.m_Count = 0; - } - return *this; + ItemTypeTraits::AccessPrev(item) = m_Back; + ItemTypeTraits::AccessNext(m_Back) = item; + m_Back = item; + ++m_Count; } - void RemoveAll() +} + +template<typename ItemTypeTraits> +void VmaIntrusiveLinkedList<ItemTypeTraits>::PushFront(ItemType* item) +{ + VMA_HEAVY_ASSERT(ItemTypeTraits::GetPrev(item) == VMA_NULL && ItemTypeTraits::GetNext(item) == VMA_NULL); + if (IsEmpty()) { - if(!IsEmpty()) - { - ItemType* item = m_Back; - while(item != VMA_NULL) - { - ItemType* const prevItem = ItemTypeTraits::AccessPrev(item); - ItemTypeTraits::AccessPrev(item) = VMA_NULL; - ItemTypeTraits::AccessNext(item) = VMA_NULL; - item = prevItem; - } - m_Front = VMA_NULL; - m_Back = VMA_NULL; - m_Count = 0; - } + m_Front = item; + m_Back = item; + m_Count = 1; } - size_t GetCount() const { return m_Count; } - bool IsEmpty() const { return m_Count == 0; } - ItemType* Front() { return m_Front; } - const ItemType* Front() const { return m_Front; } - ItemType* Back() { return m_Back; } - const ItemType* Back() const { return m_Back; } - void PushBack(ItemType* item) + else + { + ItemTypeTraits::AccessNext(item) = m_Front; + ItemTypeTraits::AccessPrev(m_Front) = item; + m_Front = item; + ++m_Count; + } +} + +template<typename ItemTypeTraits> +typename VmaIntrusiveLinkedList<ItemTypeTraits>::ItemType* VmaIntrusiveLinkedList<ItemTypeTraits>::PopBack() +{ + VMA_HEAVY_ASSERT(m_Count > 0); + ItemType* const backItem = m_Back; + ItemType* const prevItem = ItemTypeTraits::GetPrev(backItem); + if (prevItem != VMA_NULL) { - VMA_HEAVY_ASSERT(ItemTypeTraits::GetPrev(item) == VMA_NULL && ItemTypeTraits::GetNext(item) == VMA_NULL); - if(IsEmpty()) + ItemTypeTraits::AccessNext(prevItem) = VMA_NULL; + } + m_Back = prevItem; + --m_Count; + ItemTypeTraits::AccessPrev(backItem) = VMA_NULL; + ItemTypeTraits::AccessNext(backItem) = VMA_NULL; + return backItem; +} + +template<typename ItemTypeTraits> +typename VmaIntrusiveLinkedList<ItemTypeTraits>::ItemType* VmaIntrusiveLinkedList<ItemTypeTraits>::PopFront() +{ + VMA_HEAVY_ASSERT(m_Count > 0); + ItemType* const frontItem = m_Front; + ItemType* const nextItem = ItemTypeTraits::GetNext(frontItem); + if (nextItem != VMA_NULL) + { + ItemTypeTraits::AccessPrev(nextItem) = VMA_NULL; + } + m_Front = nextItem; + --m_Count; + ItemTypeTraits::AccessPrev(frontItem) = VMA_NULL; + ItemTypeTraits::AccessNext(frontItem) = VMA_NULL; + return frontItem; +} + +template<typename ItemTypeTraits> +void VmaIntrusiveLinkedList<ItemTypeTraits>::InsertBefore(ItemType* existingItem, ItemType* newItem) +{ + VMA_HEAVY_ASSERT(newItem != VMA_NULL && ItemTypeTraits::GetPrev(newItem) == VMA_NULL && ItemTypeTraits::GetNext(newItem) == VMA_NULL); + if (existingItem != VMA_NULL) + { + ItemType* const prevItem = ItemTypeTraits::GetPrev(existingItem); + ItemTypeTraits::AccessPrev(newItem) = prevItem; + ItemTypeTraits::AccessNext(newItem) = existingItem; + ItemTypeTraits::AccessPrev(existingItem) = newItem; + if (prevItem != VMA_NULL) { - m_Front = item; - m_Back = item; - m_Count = 1; + ItemTypeTraits::AccessNext(prevItem) = newItem; } else { - ItemTypeTraits::AccessPrev(item) = m_Back; - ItemTypeTraits::AccessNext(m_Back) = item; - m_Back = item; - ++m_Count; + VMA_HEAVY_ASSERT(m_Front == existingItem); + m_Front = newItem; } + ++m_Count; } - void PushFront(ItemType* item) + else + PushBack(newItem); +} + +template<typename ItemTypeTraits> +void VmaIntrusiveLinkedList<ItemTypeTraits>::InsertAfter(ItemType* existingItem, ItemType* newItem) +{ + VMA_HEAVY_ASSERT(newItem != VMA_NULL && ItemTypeTraits::GetPrev(newItem) == VMA_NULL && ItemTypeTraits::GetNext(newItem) == VMA_NULL); + if (existingItem != VMA_NULL) { - VMA_HEAVY_ASSERT(ItemTypeTraits::GetPrev(item) == VMA_NULL && ItemTypeTraits::GetNext(item) == VMA_NULL); - if(IsEmpty()) + ItemType* const nextItem = ItemTypeTraits::GetNext(existingItem); + ItemTypeTraits::AccessNext(newItem) = nextItem; + ItemTypeTraits::AccessPrev(newItem) = existingItem; + ItemTypeTraits::AccessNext(existingItem) = newItem; + if (nextItem != VMA_NULL) { - m_Front = item; - m_Back = item; - m_Count = 1; + ItemTypeTraits::AccessPrev(nextItem) = newItem; } else { - ItemTypeTraits::AccessNext(item) = m_Front; - ItemTypeTraits::AccessPrev(m_Front) = item; - m_Front = item; - ++m_Count; + VMA_HEAVY_ASSERT(m_Back == existingItem); + m_Back = newItem; } + ++m_Count; } - ItemType* PopBack() + else + return PushFront(newItem); +} + +template<typename ItemTypeTraits> +void VmaIntrusiveLinkedList<ItemTypeTraits>::Remove(ItemType* item) +{ + VMA_HEAVY_ASSERT(item != VMA_NULL && m_Count > 0); + if (ItemTypeTraits::GetPrev(item) != VMA_NULL) { - VMA_HEAVY_ASSERT(m_Count > 0); - ItemType* const backItem = m_Back; - ItemType* const prevItem = ItemTypeTraits::GetPrev(backItem); - if(prevItem != VMA_NULL) - { - ItemTypeTraits::AccessNext(prevItem) = VMA_NULL; - } - m_Back = prevItem; - --m_Count; - ItemTypeTraits::AccessPrev(backItem) = VMA_NULL; - ItemTypeTraits::AccessNext(backItem) = VMA_NULL; - return backItem; + ItemTypeTraits::AccessNext(ItemTypeTraits::AccessPrev(item)) = ItemTypeTraits::GetNext(item); } - ItemType* PopFront() + else { - VMA_HEAVY_ASSERT(m_Count > 0); - ItemType* const frontItem = m_Front; - ItemType* const nextItem = ItemTypeTraits::GetNext(frontItem); - if(nextItem != VMA_NULL) - { - ItemTypeTraits::AccessPrev(nextItem) = VMA_NULL; - } - m_Front = nextItem; - --m_Count; - ItemTypeTraits::AccessPrev(frontItem) = VMA_NULL; - ItemTypeTraits::AccessNext(frontItem) = VMA_NULL; - return frontItem; + VMA_HEAVY_ASSERT(m_Front == item); + m_Front = ItemTypeTraits::GetNext(item); } - // MyItem can be null - it means PushBack. - void InsertBefore(ItemType* existingItem, ItemType* newItem) + if (ItemTypeTraits::GetNext(item) != VMA_NULL) { - VMA_HEAVY_ASSERT(newItem != VMA_NULL && ItemTypeTraits::GetPrev(newItem) == VMA_NULL && ItemTypeTraits::GetNext(newItem) == VMA_NULL); - if(existingItem != VMA_NULL) - { - ItemType* const prevItem = ItemTypeTraits::GetPrev(existingItem); - ItemTypeTraits::AccessPrev(newItem) = prevItem; - ItemTypeTraits::AccessNext(newItem) = existingItem; - ItemTypeTraits::AccessPrev(existingItem) = newItem; - if(prevItem != VMA_NULL) - { - ItemTypeTraits::AccessNext(prevItem) = newItem; - } - else - { - VMA_HEAVY_ASSERT(m_Front == existingItem); - m_Front = newItem; - } - ++m_Count; - } - else - PushBack(newItem); + ItemTypeTraits::AccessPrev(ItemTypeTraits::AccessNext(item)) = ItemTypeTraits::GetPrev(item); } - // MyItem can be null - it means PushFront. - void InsertAfter(ItemType* existingItem, ItemType* newItem) + else { - VMA_HEAVY_ASSERT(newItem != VMA_NULL && ItemTypeTraits::GetPrev(newItem) == VMA_NULL && ItemTypeTraits::GetNext(newItem) == VMA_NULL); - if(existingItem != VMA_NULL) - { - ItemType* const nextItem = ItemTypeTraits::GetNext(existingItem); - ItemTypeTraits::AccessNext(newItem) = nextItem; - ItemTypeTraits::AccessPrev(newItem) = existingItem; - ItemTypeTraits::AccessNext(existingItem) = newItem; - if(nextItem != VMA_NULL) - { - ItemTypeTraits::AccessPrev(nextItem) = newItem; - } - else - { - VMA_HEAVY_ASSERT(m_Back == existingItem); - m_Back = newItem; - } - ++m_Count; - } - else - return PushFront(newItem); + VMA_HEAVY_ASSERT(m_Back == item); + m_Back = ItemTypeTraits::GetPrev(item); } - void Remove(ItemType* item) - { - VMA_HEAVY_ASSERT(item != VMA_NULL && m_Count > 0); - if(ItemTypeTraits::GetPrev(item) != VMA_NULL) - { - ItemTypeTraits::AccessNext(ItemTypeTraits::AccessPrev(item)) = ItemTypeTraits::GetNext(item); - } - else - { - VMA_HEAVY_ASSERT(m_Front == item); - m_Front = ItemTypeTraits::GetNext(item); - } + ItemTypeTraits::AccessPrev(item) = VMA_NULL; + ItemTypeTraits::AccessNext(item) = VMA_NULL; + --m_Count; +} - if(ItemTypeTraits::GetNext(item) != VMA_NULL) - { - ItemTypeTraits::AccessPrev(ItemTypeTraits::AccessNext(item)) = ItemTypeTraits::GetPrev(item); - } - else +template<typename ItemTypeTraits> +void VmaIntrusiveLinkedList<ItemTypeTraits>::RemoveAll() +{ + if (!IsEmpty()) + { + ItemType* item = m_Back; + while (item != VMA_NULL) { - VMA_HEAVY_ASSERT(m_Back == item); - m_Back = ItemTypeTraits::GetPrev(item); + ItemType* const prevItem = ItemTypeTraits::AccessPrev(item); + ItemTypeTraits::AccessPrev(item) = VMA_NULL; + ItemTypeTraits::AccessNext(item) = VMA_NULL; + item = prevItem; } - ItemTypeTraits::AccessPrev(item) = VMA_NULL; - ItemTypeTraits::AccessNext(item) = VMA_NULL; - --m_Count; + m_Front = VMA_NULL; + m_Back = VMA_NULL; + m_Count = 0; } -private: - ItemType* m_Front = VMA_NULL; - ItemType* m_Back = VMA_NULL; - size_t m_Count = 0; -}; - -//////////////////////////////////////////////////////////////////////////////// -// class VmaMap +} +#endif // _VMA_INTRUSIVE_LINKED_LIST_FUNCTIONS +#endif // _VMA_INTRUSIVE_LINKED_LIST // Unused in this version. #if 0 -#if VMA_USE_STL_UNORDERED_MAP - -#define VmaPair std::pair - -#define VMA_MAP_TYPE(KeyT, ValueT) \ - std::unordered_map< KeyT, ValueT, std::hash<KeyT>, std::equal_to<KeyT>, VmaStlAllocator< std::pair<KeyT, ValueT> > > - -#else // #if VMA_USE_STL_UNORDERED_MAP - +#ifndef _VMA_PAIR template<typename T1, typename T2> struct VmaPair { T1 first; T2 second; - VmaPair() : first(), second() { } - VmaPair(const T1& firstSrc, const T2& secondSrc) : first(firstSrc), second(secondSrc) { } + VmaPair() : first(), second() {} + VmaPair(const T1& firstSrc, const T2& secondSrc) : first(firstSrc), second(secondSrc) {} +}; + +template<typename FirstT, typename SecondT> +struct VmaPairFirstLess +{ + bool operator()(const VmaPair<FirstT, SecondT>& lhs, const VmaPair<FirstT, SecondT>& rhs) const + { + return lhs.first < rhs.first; + } + bool operator()(const VmaPair<FirstT, SecondT>& lhs, const FirstT& rhsFirst) const + { + return lhs.first < rhsFirst; + } }; +#endif // _VMA_PAIR +#ifndef _VMA_MAP /* Class compatible with subset of interface of std::unordered_map. KeyT, ValueT must be POD because they will be stored in VmaVector. */ @@ -6336,7 +5073,7 @@ public: typedef VmaPair<KeyT, ValueT> PairType; typedef PairType* iterator; - VmaMap(const VmaStlAllocator<PairType>& allocator) : m_Vector(allocator) { } + VmaMap(const VmaStlAllocator<PairType>& allocator) : m_Vector(allocator) {} iterator begin() { return m_Vector.begin(); } iterator end() { return m_Vector.end(); } @@ -6349,21 +5086,7 @@ private: VmaVector< PairType, VmaStlAllocator<PairType> > m_Vector; }; -#define VMA_MAP_TYPE(KeyT, ValueT) VmaMap<KeyT, ValueT> - -template<typename FirstT, typename SecondT> -struct VmaPairFirstLess -{ - bool operator()(const VmaPair<FirstT, SecondT>& lhs, const VmaPair<FirstT, SecondT>& rhs) const - { - return lhs.first < rhs.first; - } - bool operator()(const VmaPair<FirstT, SecondT>& lhs, const FirstT& rhsFirst) const - { - return lhs.first < rhsFirst; - } -}; - +#ifndef _VMA_MAP_FUNCTIONS template<typename KeyT, typename ValueT> void VmaMap<KeyT, ValueT>::insert(const PairType& pair) { @@ -6383,7 +5106,7 @@ VmaPair<KeyT, ValueT>* VmaMap<KeyT, ValueT>::find(const KeyT& key) m_Vector.data() + m_Vector.size(), key, VmaPairFirstLess<KeyT, ValueT>()); - if((it != m_Vector.end()) && (it->first == key)) + if ((it != m_Vector.end()) && (it->first == key)) { return it; } @@ -6398,2215 +5121,37 @@ void VmaMap<KeyT, ValueT>::erase(iterator it) { VmaVectorRemove(m_Vector, it - m_Vector.begin()); } - -#endif // #if VMA_USE_STL_UNORDERED_MAP +#endif // _VMA_MAP_FUNCTIONS +#endif // _VMA_MAP #endif // #if 0 -//////////////////////////////////////////////////////////////////////////////// - -class VmaDeviceMemoryBlock; - -enum VMA_CACHE_OPERATION { VMA_CACHE_FLUSH, VMA_CACHE_INVALIDATE }; - -struct VmaAllocation_T -{ -private: - static const uint8_t MAP_COUNT_FLAG_PERSISTENT_MAP = 0x80; - - enum FLAGS - { - FLAG_USER_DATA_STRING = 0x01, - }; - -public: - enum ALLOCATION_TYPE - { - ALLOCATION_TYPE_NONE, - ALLOCATION_TYPE_BLOCK, - ALLOCATION_TYPE_DEDICATED, - }; - - /* - This struct is allocated using VmaPoolAllocator. - */ - - VmaAllocation_T(uint32_t currentFrameIndex, bool userDataString) : - m_Alignment{1}, - m_Size{0}, - m_pUserData{VMA_NULL}, - m_LastUseFrameIndex{currentFrameIndex}, - m_MemoryTypeIndex{0}, - m_Type{(uint8_t)ALLOCATION_TYPE_NONE}, - m_SuballocationType{(uint8_t)VMA_SUBALLOCATION_TYPE_UNKNOWN}, - m_MapCount{0}, - m_Flags{userDataString ? (uint8_t)FLAG_USER_DATA_STRING : (uint8_t)0} - { -#if VMA_STATS_STRING_ENABLED - m_CreationFrameIndex = currentFrameIndex; - m_BufferImageUsage = 0; -#endif - } - - ~VmaAllocation_T() - { - VMA_ASSERT((m_MapCount & ~MAP_COUNT_FLAG_PERSISTENT_MAP) == 0 && "Allocation was not unmapped before destruction."); - - // Check if owned string was freed. - VMA_ASSERT(m_pUserData == VMA_NULL); - } - - void InitBlockAllocation( - VmaDeviceMemoryBlock* block, - VkDeviceSize offset, - VkDeviceSize alignment, - VkDeviceSize size, - uint32_t memoryTypeIndex, - VmaSuballocationType suballocationType, - bool mapped, - bool canBecomeLost) - { - VMA_ASSERT(m_Type == ALLOCATION_TYPE_NONE); - VMA_ASSERT(block != VMA_NULL); - m_Type = (uint8_t)ALLOCATION_TYPE_BLOCK; - m_Alignment = alignment; - m_Size = size; - m_MemoryTypeIndex = memoryTypeIndex; - m_MapCount = mapped ? MAP_COUNT_FLAG_PERSISTENT_MAP : 0; - m_SuballocationType = (uint8_t)suballocationType; - m_BlockAllocation.m_Block = block; - m_BlockAllocation.m_Offset = offset; - m_BlockAllocation.m_CanBecomeLost = canBecomeLost; - } - - void InitLost() - { - VMA_ASSERT(m_Type == ALLOCATION_TYPE_NONE); - VMA_ASSERT(m_LastUseFrameIndex.load() == VMA_FRAME_INDEX_LOST); - m_Type = (uint8_t)ALLOCATION_TYPE_BLOCK; - m_MemoryTypeIndex = 0; - m_BlockAllocation.m_Block = VMA_NULL; - m_BlockAllocation.m_Offset = 0; - m_BlockAllocation.m_CanBecomeLost = true; - } - - void ChangeBlockAllocation( - VmaAllocator hAllocator, - VmaDeviceMemoryBlock* block, - VkDeviceSize offset); - - void ChangeOffset(VkDeviceSize newOffset); - - // pMappedData not null means allocation is created with MAPPED flag. - void InitDedicatedAllocation( - uint32_t memoryTypeIndex, - VkDeviceMemory hMemory, - VmaSuballocationType suballocationType, - void* pMappedData, - VkDeviceSize size) - { - VMA_ASSERT(m_Type == ALLOCATION_TYPE_NONE); - VMA_ASSERT(hMemory != VK_NULL_HANDLE); - m_Type = (uint8_t)ALLOCATION_TYPE_DEDICATED; - m_Alignment = 0; - m_Size = size; - m_MemoryTypeIndex = memoryTypeIndex; - m_SuballocationType = (uint8_t)suballocationType; - m_MapCount = (pMappedData != VMA_NULL) ? MAP_COUNT_FLAG_PERSISTENT_MAP : 0; - m_DedicatedAllocation.m_hMemory = hMemory; - m_DedicatedAllocation.m_pMappedData = pMappedData; - m_DedicatedAllocation.m_Prev = VMA_NULL; - m_DedicatedAllocation.m_Next = VMA_NULL; - } - - ALLOCATION_TYPE GetType() const { return (ALLOCATION_TYPE)m_Type; } - VkDeviceSize GetAlignment() const { return m_Alignment; } - VkDeviceSize GetSize() const { return m_Size; } - bool IsUserDataString() const { return (m_Flags & FLAG_USER_DATA_STRING) != 0; } - void* GetUserData() const { return m_pUserData; } - void SetUserData(VmaAllocator hAllocator, void* pUserData); - VmaSuballocationType GetSuballocationType() const { return (VmaSuballocationType)m_SuballocationType; } - - VmaDeviceMemoryBlock* GetBlock() const - { - VMA_ASSERT(m_Type == ALLOCATION_TYPE_BLOCK); - return m_BlockAllocation.m_Block; - } - VkDeviceSize GetOffset() const; - VkDeviceMemory GetMemory() const; - uint32_t GetMemoryTypeIndex() const { return m_MemoryTypeIndex; } - bool IsPersistentMap() const { return (m_MapCount & MAP_COUNT_FLAG_PERSISTENT_MAP) != 0; } - void* GetMappedData() const; - bool CanBecomeLost() const; - - uint32_t GetLastUseFrameIndex() const - { - return m_LastUseFrameIndex.load(); - } - bool CompareExchangeLastUseFrameIndex(uint32_t& expected, uint32_t desired) - { - return m_LastUseFrameIndex.compare_exchange_weak(expected, desired); - } - /* - - If hAllocation.LastUseFrameIndex + frameInUseCount < allocator.CurrentFrameIndex, - makes it lost by setting LastUseFrameIndex = VMA_FRAME_INDEX_LOST and returns true. - - Else, returns false. - - If hAllocation is already lost, assert - you should not call it then. - If hAllocation was not created with CAN_BECOME_LOST_BIT, assert. - */ - bool MakeLost(uint32_t currentFrameIndex, uint32_t frameInUseCount); - - void DedicatedAllocCalcStatsInfo(VmaStatInfo& outInfo) - { - VMA_ASSERT(m_Type == ALLOCATION_TYPE_DEDICATED); - outInfo.blockCount = 1; - outInfo.allocationCount = 1; - outInfo.unusedRangeCount = 0; - outInfo.usedBytes = m_Size; - outInfo.unusedBytes = 0; - outInfo.allocationSizeMin = outInfo.allocationSizeMax = m_Size; - outInfo.unusedRangeSizeMin = UINT64_MAX; - outInfo.unusedRangeSizeMax = 0; - } - - void BlockAllocMap(); - void BlockAllocUnmap(); - VkResult DedicatedAllocMap(VmaAllocator hAllocator, void** ppData); - void DedicatedAllocUnmap(VmaAllocator hAllocator); - -#if VMA_STATS_STRING_ENABLED - uint32_t GetCreationFrameIndex() const { return m_CreationFrameIndex; } - uint32_t GetBufferImageUsage() const { return m_BufferImageUsage; } - - void InitBufferImageUsage(uint32_t bufferImageUsage) - { - VMA_ASSERT(m_BufferImageUsage == 0); - m_BufferImageUsage = bufferImageUsage; - } - - void PrintParameters(class VmaJsonWriter& json) const; -#endif - -private: - VkDeviceSize m_Alignment; - VkDeviceSize m_Size; - void* m_pUserData; - VMA_ATOMIC_UINT32 m_LastUseFrameIndex; - uint32_t m_MemoryTypeIndex; - uint8_t m_Type; // ALLOCATION_TYPE - uint8_t m_SuballocationType; // VmaSuballocationType - // Bit 0x80 is set when allocation was created with VMA_ALLOCATION_CREATE_MAPPED_BIT. - // Bits with mask 0x7F are reference counter for vmaMapMemory()/vmaUnmapMemory(). - uint8_t m_MapCount; - uint8_t m_Flags; // enum FLAGS - - // Allocation out of VmaDeviceMemoryBlock. - struct BlockAllocation - { - VmaDeviceMemoryBlock* m_Block; - VkDeviceSize m_Offset; - bool m_CanBecomeLost; - }; - - // Allocation for an object that has its own private VkDeviceMemory. - struct DedicatedAllocation - { - VkDeviceMemory m_hMemory; - void* m_pMappedData; // Not null means memory is mapped. - VmaAllocation_T* m_Prev; - VmaAllocation_T* m_Next; - }; - - union - { - // Allocation out of VmaDeviceMemoryBlock. - BlockAllocation m_BlockAllocation; - // Allocation for an object that has its own private VkDeviceMemory. - DedicatedAllocation m_DedicatedAllocation; - }; - -#if VMA_STATS_STRING_ENABLED - uint32_t m_CreationFrameIndex; - uint32_t m_BufferImageUsage; // 0 if unknown. -#endif - - void FreeUserDataString(VmaAllocator hAllocator); - - friend struct VmaDedicatedAllocationListItemTraits; -}; - -struct VmaDedicatedAllocationListItemTraits -{ - typedef VmaAllocation_T ItemType; - static ItemType* GetPrev(const ItemType* item) - { - VMA_HEAVY_ASSERT(item->GetType() == VmaAllocation_T::ALLOCATION_TYPE_DEDICATED); - return item->m_DedicatedAllocation.m_Prev; - } - static ItemType* GetNext(const ItemType* item) - { - VMA_HEAVY_ASSERT(item->GetType() == VmaAllocation_T::ALLOCATION_TYPE_DEDICATED); - return item->m_DedicatedAllocation.m_Next; - } - static ItemType*& AccessPrev(ItemType* item) - { - VMA_HEAVY_ASSERT(item->GetType() == VmaAllocation_T::ALLOCATION_TYPE_DEDICATED); - return item->m_DedicatedAllocation.m_Prev; - } - static ItemType*& AccessNext(ItemType* item){ - VMA_HEAVY_ASSERT(item->GetType() == VmaAllocation_T::ALLOCATION_TYPE_DEDICATED); - return item->m_DedicatedAllocation.m_Next; - } -}; - -/* -Represents a region of VmaDeviceMemoryBlock that is either assigned and returned as -allocated memory block or free. -*/ -struct VmaSuballocation -{ - VkDeviceSize offset; - VkDeviceSize size; - VmaAllocation hAllocation; - VmaSuballocationType type; -}; - -// Comparator for offsets. -struct VmaSuballocationOffsetLess -{ - bool operator()(const VmaSuballocation& lhs, const VmaSuballocation& rhs) const - { - return lhs.offset < rhs.offset; - } -}; -struct VmaSuballocationOffsetGreater -{ - bool operator()(const VmaSuballocation& lhs, const VmaSuballocation& rhs) const - { - return lhs.offset > rhs.offset; - } -}; - -typedef VmaList< VmaSuballocation, VmaStlAllocator<VmaSuballocation> > VmaSuballocationList; - -// Cost of one additional allocation lost, as equivalent in bytes. -static const VkDeviceSize VMA_LOST_ALLOCATION_COST = 1048576; - -enum class VmaAllocationRequestType -{ - Normal, - // Used by "Linear" algorithm. - UpperAddress, - EndOf1st, - EndOf2nd, -}; - -/* -Parameters of planned allocation inside a VmaDeviceMemoryBlock. - -If canMakeOtherLost was false: -- item points to a FREE suballocation. -- itemsToMakeLostCount is 0. - -If canMakeOtherLost was true: -- item points to first of sequence of suballocations, which are either FREE, - or point to VmaAllocations that can become lost. -- itemsToMakeLostCount is the number of VmaAllocations that need to be made lost for - the requested allocation to succeed. -*/ -struct VmaAllocationRequest -{ - VkDeviceSize offset; - VkDeviceSize sumFreeSize; // Sum size of free items that overlap with proposed allocation. - VkDeviceSize sumItemSize; // Sum size of items to make lost that overlap with proposed allocation. - VmaSuballocationList::iterator item; - size_t itemsToMakeLostCount; - void* customData; - VmaAllocationRequestType type; - - VkDeviceSize CalcCost() const - { - return sumItemSize + itemsToMakeLostCount * VMA_LOST_ALLOCATION_COST; - } -}; - -/* -Data structure used for bookkeeping of allocations and unused ranges of memory -in a single VkDeviceMemory block. -*/ -class VmaBlockMetadata -{ -public: - VmaBlockMetadata(VmaAllocator hAllocator); - virtual ~VmaBlockMetadata() { } - virtual void Init(VkDeviceSize size) { m_Size = size; } - - // Validates all data structures inside this object. If not valid, returns false. - virtual bool Validate() const = 0; - VkDeviceSize GetSize() const { return m_Size; } - virtual size_t GetAllocationCount() const = 0; - virtual VkDeviceSize GetSumFreeSize() const = 0; - virtual VkDeviceSize GetUnusedRangeSizeMax() const = 0; - // Returns true if this block is empty - contains only single free suballocation. - virtual bool IsEmpty() const = 0; - - virtual void CalcAllocationStatInfo(VmaStatInfo& outInfo) const = 0; - // Shouldn't modify blockCount. - virtual void AddPoolStats(VmaPoolStats& inoutStats) const = 0; - -#if VMA_STATS_STRING_ENABLED - virtual void PrintDetailedMap(class VmaJsonWriter& json) const = 0; -#endif - - // Tries to find a place for suballocation with given parameters inside this block. - // If succeeded, fills pAllocationRequest and returns true. - // If failed, returns false. - virtual bool CreateAllocationRequest( - uint32_t currentFrameIndex, - uint32_t frameInUseCount, - VkDeviceSize bufferImageGranularity, - VkDeviceSize allocSize, - VkDeviceSize allocAlignment, - bool upperAddress, - VmaSuballocationType allocType, - bool canMakeOtherLost, - // Always one of VMA_ALLOCATION_CREATE_STRATEGY_* or VMA_ALLOCATION_INTERNAL_STRATEGY_* flags. - uint32_t strategy, - VmaAllocationRequest* pAllocationRequest) = 0; - - virtual bool MakeRequestedAllocationsLost( - uint32_t currentFrameIndex, - uint32_t frameInUseCount, - VmaAllocationRequest* pAllocationRequest) = 0; - - virtual uint32_t MakeAllocationsLost(uint32_t currentFrameIndex, uint32_t frameInUseCount) = 0; - - virtual VkResult CheckCorruption(const void* pBlockData) = 0; - - // Makes actual allocation based on request. Request must already be checked and valid. - virtual void Alloc( - const VmaAllocationRequest& request, - VmaSuballocationType type, - VkDeviceSize allocSize, - VmaAllocation hAllocation) = 0; - - // Frees suballocation assigned to given memory region. - virtual void Free(const VmaAllocation allocation) = 0; - virtual void FreeAtOffset(VkDeviceSize offset) = 0; - -protected: - const VkAllocationCallbacks* GetAllocationCallbacks() const { return m_pAllocationCallbacks; } - -#if VMA_STATS_STRING_ENABLED - void PrintDetailedMap_Begin(class VmaJsonWriter& json, - VkDeviceSize unusedBytes, - size_t allocationCount, - size_t unusedRangeCount) const; - void PrintDetailedMap_Allocation(class VmaJsonWriter& json, - VkDeviceSize offset, - VmaAllocation hAllocation) const; - void PrintDetailedMap_UnusedRange(class VmaJsonWriter& json, - VkDeviceSize offset, - VkDeviceSize size) const; - void PrintDetailedMap_End(class VmaJsonWriter& json) const; -#endif - -private: - VkDeviceSize m_Size; - const VkAllocationCallbacks* m_pAllocationCallbacks; -}; - -#define VMA_VALIDATE(cond) do { if(!(cond)) { \ - VMA_ASSERT(0 && "Validation failed: " #cond); \ - return false; \ - } } while(false) - -class VmaBlockMetadata_Generic : public VmaBlockMetadata -{ - VMA_CLASS_NO_COPY(VmaBlockMetadata_Generic) -public: - VmaBlockMetadata_Generic(VmaAllocator hAllocator); - virtual ~VmaBlockMetadata_Generic(); - virtual void Init(VkDeviceSize size); - - virtual bool Validate() const; - virtual size_t GetAllocationCount() const { return m_Suballocations.size() - m_FreeCount; } - virtual VkDeviceSize GetSumFreeSize() const { return m_SumFreeSize; } - virtual VkDeviceSize GetUnusedRangeSizeMax() const; - virtual bool IsEmpty() const; - - virtual void CalcAllocationStatInfo(VmaStatInfo& outInfo) const; - virtual void AddPoolStats(VmaPoolStats& inoutStats) const; - -#if VMA_STATS_STRING_ENABLED - virtual void PrintDetailedMap(class VmaJsonWriter& json) const; -#endif - - virtual bool CreateAllocationRequest( - uint32_t currentFrameIndex, - uint32_t frameInUseCount, - VkDeviceSize bufferImageGranularity, - VkDeviceSize allocSize, - VkDeviceSize allocAlignment, - bool upperAddress, - VmaSuballocationType allocType, - bool canMakeOtherLost, - uint32_t strategy, - VmaAllocationRequest* pAllocationRequest); - - virtual bool MakeRequestedAllocationsLost( - uint32_t currentFrameIndex, - uint32_t frameInUseCount, - VmaAllocationRequest* pAllocationRequest); - - virtual uint32_t MakeAllocationsLost(uint32_t currentFrameIndex, uint32_t frameInUseCount); - - virtual VkResult CheckCorruption(const void* pBlockData); - - virtual void Alloc( - const VmaAllocationRequest& request, - VmaSuballocationType type, - VkDeviceSize allocSize, - VmaAllocation hAllocation); - - virtual void Free(const VmaAllocation allocation); - virtual void FreeAtOffset(VkDeviceSize offset); - - //////////////////////////////////////////////////////////////////////////////// - // For defragmentation - - bool IsBufferImageGranularityConflictPossible( - VkDeviceSize bufferImageGranularity, - VmaSuballocationType& inOutPrevSuballocType) const; - -private: - friend class VmaDefragmentationAlgorithm_Generic; - friend class VmaDefragmentationAlgorithm_Fast; - - uint32_t m_FreeCount; - VkDeviceSize m_SumFreeSize; - VmaSuballocationList m_Suballocations; - // Suballocations that are free and have size greater than certain threshold. - // Sorted by size, ascending. - VmaVector< VmaSuballocationList::iterator, VmaStlAllocator< VmaSuballocationList::iterator > > m_FreeSuballocationsBySize; - - bool ValidateFreeSuballocationList() const; - - // Checks if requested suballocation with given parameters can be placed in given pFreeSuballocItem. - // If yes, fills pOffset and returns true. If no, returns false. - bool CheckAllocation( - uint32_t currentFrameIndex, - uint32_t frameInUseCount, - VkDeviceSize bufferImageGranularity, - VkDeviceSize allocSize, - VkDeviceSize allocAlignment, - VmaSuballocationType allocType, - VmaSuballocationList::const_iterator suballocItem, - bool canMakeOtherLost, - VkDeviceSize* pOffset, - size_t* itemsToMakeLostCount, - VkDeviceSize* pSumFreeSize, - VkDeviceSize* pSumItemSize) const; - // Given free suballocation, it merges it with following one, which must also be free. - void MergeFreeWithNext(VmaSuballocationList::iterator item); - // Releases given suballocation, making it free. - // Merges it with adjacent free suballocations if applicable. - // Returns iterator to new free suballocation at this place. - VmaSuballocationList::iterator FreeSuballocation(VmaSuballocationList::iterator suballocItem); - // Given free suballocation, it inserts it into sorted list of - // m_FreeSuballocationsBySize if it's suitable. - void RegisterFreeSuballocation(VmaSuballocationList::iterator item); - // Given free suballocation, it removes it from sorted list of - // m_FreeSuballocationsBySize if it's suitable. - void UnregisterFreeSuballocation(VmaSuballocationList::iterator item); -}; - -/* -Allocations and their references in internal data structure look like this: - -if(m_2ndVectorMode == SECOND_VECTOR_EMPTY): - - 0 +-------+ - | | - | | - | | - +-------+ - | Alloc | 1st[m_1stNullItemsBeginCount] - +-------+ - | Alloc | 1st[m_1stNullItemsBeginCount + 1] - +-------+ - | ... | - +-------+ - | Alloc | 1st[1st.size() - 1] - +-------+ - | | - | | - | | -GetSize() +-------+ - -if(m_2ndVectorMode == SECOND_VECTOR_RING_BUFFER): - - 0 +-------+ - | Alloc | 2nd[0] - +-------+ - | Alloc | 2nd[1] - +-------+ - | ... | - +-------+ - | Alloc | 2nd[2nd.size() - 1] - +-------+ - | | - | | - | | - +-------+ - | Alloc | 1st[m_1stNullItemsBeginCount] - +-------+ - | Alloc | 1st[m_1stNullItemsBeginCount + 1] - +-------+ - | ... | - +-------+ - | Alloc | 1st[1st.size() - 1] - +-------+ - | | -GetSize() +-------+ - -if(m_2ndVectorMode == SECOND_VECTOR_DOUBLE_STACK): - - 0 +-------+ - | | - | | - | | - +-------+ - | Alloc | 1st[m_1stNullItemsBeginCount] - +-------+ - | Alloc | 1st[m_1stNullItemsBeginCount + 1] - +-------+ - | ... | - +-------+ - | Alloc | 1st[1st.size() - 1] - +-------+ - | | - | | - | | - +-------+ - | Alloc | 2nd[2nd.size() - 1] - +-------+ - | ... | - +-------+ - | Alloc | 2nd[1] - +-------+ - | Alloc | 2nd[0] -GetSize() +-------+ - -*/ -class VmaBlockMetadata_Linear : public VmaBlockMetadata -{ - VMA_CLASS_NO_COPY(VmaBlockMetadata_Linear) -public: - VmaBlockMetadata_Linear(VmaAllocator hAllocator); - virtual ~VmaBlockMetadata_Linear(); - virtual void Init(VkDeviceSize size); - - virtual bool Validate() const; - virtual size_t GetAllocationCount() const; - virtual VkDeviceSize GetSumFreeSize() const { return m_SumFreeSize; } - virtual VkDeviceSize GetUnusedRangeSizeMax() const; - virtual bool IsEmpty() const { return GetAllocationCount() == 0; } - - virtual void CalcAllocationStatInfo(VmaStatInfo& outInfo) const; - virtual void AddPoolStats(VmaPoolStats& inoutStats) const; - -#if VMA_STATS_STRING_ENABLED - virtual void PrintDetailedMap(class VmaJsonWriter& json) const; -#endif - - virtual bool CreateAllocationRequest( - uint32_t currentFrameIndex, - uint32_t frameInUseCount, - VkDeviceSize bufferImageGranularity, - VkDeviceSize allocSize, - VkDeviceSize allocAlignment, - bool upperAddress, - VmaSuballocationType allocType, - bool canMakeOtherLost, - uint32_t strategy, - VmaAllocationRequest* pAllocationRequest); - - virtual bool MakeRequestedAllocationsLost( - uint32_t currentFrameIndex, - uint32_t frameInUseCount, - VmaAllocationRequest* pAllocationRequest); - - virtual uint32_t MakeAllocationsLost(uint32_t currentFrameIndex, uint32_t frameInUseCount); - - virtual VkResult CheckCorruption(const void* pBlockData); - - virtual void Alloc( - const VmaAllocationRequest& request, - VmaSuballocationType type, - VkDeviceSize allocSize, - VmaAllocation hAllocation); - - virtual void Free(const VmaAllocation allocation); - virtual void FreeAtOffset(VkDeviceSize offset); - -private: - /* - There are two suballocation vectors, used in ping-pong way. - The one with index m_1stVectorIndex is called 1st. - The one with index (m_1stVectorIndex ^ 1) is called 2nd. - 2nd can be non-empty only when 1st is not empty. - When 2nd is not empty, m_2ndVectorMode indicates its mode of operation. - */ - typedef VmaVector< VmaSuballocation, VmaStlAllocator<VmaSuballocation> > SuballocationVectorType; - - enum SECOND_VECTOR_MODE - { - SECOND_VECTOR_EMPTY, - /* - Suballocations in 2nd vector are created later than the ones in 1st, but they - all have smaller offset. - */ - SECOND_VECTOR_RING_BUFFER, - /* - Suballocations in 2nd vector are upper side of double stack. - They all have offsets higher than those in 1st vector. - Top of this stack means smaller offsets, but higher indices in this vector. - */ - SECOND_VECTOR_DOUBLE_STACK, - }; - - VkDeviceSize m_SumFreeSize; - SuballocationVectorType m_Suballocations0, m_Suballocations1; - uint32_t m_1stVectorIndex; - SECOND_VECTOR_MODE m_2ndVectorMode; - - SuballocationVectorType& AccessSuballocations1st() { return m_1stVectorIndex ? m_Suballocations1 : m_Suballocations0; } - SuballocationVectorType& AccessSuballocations2nd() { return m_1stVectorIndex ? m_Suballocations0 : m_Suballocations1; } - const SuballocationVectorType& AccessSuballocations1st() const { return m_1stVectorIndex ? m_Suballocations1 : m_Suballocations0; } - const SuballocationVectorType& AccessSuballocations2nd() const { return m_1stVectorIndex ? m_Suballocations0 : m_Suballocations1; } - - // Number of items in 1st vector with hAllocation = null at the beginning. - size_t m_1stNullItemsBeginCount; - // Number of other items in 1st vector with hAllocation = null somewhere in the middle. - size_t m_1stNullItemsMiddleCount; - // Number of items in 2nd vector with hAllocation = null. - size_t m_2ndNullItemsCount; - - bool ShouldCompact1st() const; - void CleanupAfterFree(); - - bool CreateAllocationRequest_LowerAddress( - uint32_t currentFrameIndex, - uint32_t frameInUseCount, - VkDeviceSize bufferImageGranularity, - VkDeviceSize allocSize, - VkDeviceSize allocAlignment, - VmaSuballocationType allocType, - bool canMakeOtherLost, - uint32_t strategy, - VmaAllocationRequest* pAllocationRequest); - bool CreateAllocationRequest_UpperAddress( - uint32_t currentFrameIndex, - uint32_t frameInUseCount, - VkDeviceSize bufferImageGranularity, - VkDeviceSize allocSize, - VkDeviceSize allocAlignment, - VmaSuballocationType allocType, - bool canMakeOtherLost, - uint32_t strategy, - VmaAllocationRequest* pAllocationRequest); -}; - -/* -- GetSize() is the original size of allocated memory block. -- m_UsableSize is this size aligned down to a power of two. - All allocations and calculations happen relative to m_UsableSize. -- GetUnusableSize() is the difference between them. - It is reported as separate, unused range, not available for allocations. - -Node at level 0 has size = m_UsableSize. -Each next level contains nodes with size 2 times smaller than current level. -m_LevelCount is the maximum number of levels to use in the current object. -*/ -class VmaBlockMetadata_Buddy : public VmaBlockMetadata -{ - VMA_CLASS_NO_COPY(VmaBlockMetadata_Buddy) -public: - VmaBlockMetadata_Buddy(VmaAllocator hAllocator); - virtual ~VmaBlockMetadata_Buddy(); - virtual void Init(VkDeviceSize size); - - virtual bool Validate() const; - virtual size_t GetAllocationCount() const { return m_AllocationCount; } - virtual VkDeviceSize GetSumFreeSize() const { return m_SumFreeSize + GetUnusableSize(); } - virtual VkDeviceSize GetUnusedRangeSizeMax() const; - virtual bool IsEmpty() const { return m_Root->type == Node::TYPE_FREE; } - - virtual void CalcAllocationStatInfo(VmaStatInfo& outInfo) const; - virtual void AddPoolStats(VmaPoolStats& inoutStats) const; - -#if VMA_STATS_STRING_ENABLED - virtual void PrintDetailedMap(class VmaJsonWriter& json) const; -#endif - - virtual bool CreateAllocationRequest( - uint32_t currentFrameIndex, - uint32_t frameInUseCount, - VkDeviceSize bufferImageGranularity, - VkDeviceSize allocSize, - VkDeviceSize allocAlignment, - bool upperAddress, - VmaSuballocationType allocType, - bool canMakeOtherLost, - uint32_t strategy, - VmaAllocationRequest* pAllocationRequest); - - virtual bool MakeRequestedAllocationsLost( - uint32_t currentFrameIndex, - uint32_t frameInUseCount, - VmaAllocationRequest* pAllocationRequest); - - virtual uint32_t MakeAllocationsLost(uint32_t currentFrameIndex, uint32_t frameInUseCount); - - virtual VkResult CheckCorruption(const void* pBlockData) { return VK_ERROR_FEATURE_NOT_PRESENT; } - - virtual void Alloc( - const VmaAllocationRequest& request, - VmaSuballocationType type, - VkDeviceSize allocSize, - VmaAllocation hAllocation); - - virtual void Free(const VmaAllocation allocation) { FreeAtOffset(allocation, allocation->GetOffset()); } - virtual void FreeAtOffset(VkDeviceSize offset) { FreeAtOffset(VMA_NULL, offset); } - -private: - static const VkDeviceSize MIN_NODE_SIZE = 32; - static const size_t MAX_LEVELS = 30; - - struct ValidationContext - { - size_t calculatedAllocationCount; - size_t calculatedFreeCount; - VkDeviceSize calculatedSumFreeSize; - - ValidationContext() : - calculatedAllocationCount(0), - calculatedFreeCount(0), - calculatedSumFreeSize(0) { } - }; - - struct Node - { - VkDeviceSize offset; - enum TYPE - { - TYPE_FREE, - TYPE_ALLOCATION, - TYPE_SPLIT, - TYPE_COUNT - } type; - Node* parent; - Node* buddy; - - union - { - struct - { - Node* prev; - Node* next; - } free; - struct - { - VmaAllocation alloc; - } allocation; - struct - { - Node* leftChild; - } split; - }; - }; - - // Size of the memory block aligned down to a power of two. - VkDeviceSize m_UsableSize; - uint32_t m_LevelCount; - - Node* m_Root; - struct { - Node* front; - Node* back; - } m_FreeList[MAX_LEVELS]; - // Number of nodes in the tree with type == TYPE_ALLOCATION. - size_t m_AllocationCount; - // Number of nodes in the tree with type == TYPE_FREE. - size_t m_FreeCount; - // This includes space wasted due to internal fragmentation. Doesn't include unusable size. - VkDeviceSize m_SumFreeSize; - - VkDeviceSize GetUnusableSize() const { return GetSize() - m_UsableSize; } - void DeleteNode(Node* node); - bool ValidateNode(ValidationContext& ctx, const Node* parent, const Node* curr, uint32_t level, VkDeviceSize levelNodeSize) const; - uint32_t AllocSizeToLevel(VkDeviceSize allocSize) const; - inline VkDeviceSize LevelToNodeSize(uint32_t level) const { return m_UsableSize >> level; } - // Alloc passed just for validation. Can be null. - void FreeAtOffset(VmaAllocation alloc, VkDeviceSize offset); - void CalcAllocationStatInfoNode(VmaStatInfo& outInfo, const Node* node, VkDeviceSize levelNodeSize) const; - // Adds node to the front of FreeList at given level. - // node->type must be FREE. - // node->free.prev, next can be undefined. - void AddToFreeListFront(uint32_t level, Node* node); - // Removes node from FreeList at given level. - // node->type must be FREE. - // node->free.prev, next stay untouched. - void RemoveFromFreeList(uint32_t level, Node* node); - -#if VMA_STATS_STRING_ENABLED - void PrintDetailedMapNode(class VmaJsonWriter& json, const Node* node, VkDeviceSize levelNodeSize) const; -#endif -}; - -struct VmaBlockVector; - -/* -Represents a single block of device memory (`VkDeviceMemory`) with all the -data about its regions (aka suballocations, #VmaAllocation), assigned and free. - -Thread-safety: This class must be externally synchronized. -*/ -class VmaDeviceMemoryBlock -{ - VMA_CLASS_NO_COPY(VmaDeviceMemoryBlock) -public: - VmaBlockMetadata* m_pMetadata; - - VmaDeviceMemoryBlock(VmaAllocator hAllocator); - - ~VmaDeviceMemoryBlock() - { - VMA_ASSERT(m_MapCount == 0 && "VkDeviceMemory block is being destroyed while it is still mapped."); - VMA_ASSERT(m_hMemory == VK_NULL_HANDLE); - } - - // Always call after construction. - void Init( - VmaAllocator hAllocator, - VmaBlockVector* parentBlockVector, - VmaPool hParentPool, - uint32_t newMemoryTypeIndex, - VkDeviceMemory newMemory, - VkDeviceSize newSize, - uint32_t id, - uint32_t algorithm); - // Always call before destruction. - void Destroy(VmaAllocator allocator); - - VmaBlockVector* GetParentBlockVector() const { return m_ParentBlockVector; } - VmaPool GetParentPool() const { return m_hParentPool; } - VkDeviceMemory GetDeviceMemory() const { return m_hMemory; } - uint32_t GetMemoryTypeIndex() const { return m_MemoryTypeIndex; } - uint32_t GetId() const { return m_Id; } - void* GetMappedData() const { return m_pMappedData; } - - // Validates all data structures inside this object. If not valid, returns false. - bool Validate() const; - - VkResult CheckCorruption(VmaAllocator hAllocator); - - // ppData can be null. - VkResult Map(VmaAllocator hAllocator, uint32_t count, void** ppData); - void Unmap(VmaAllocator hAllocator, uint32_t count); - - VkResult WriteMagicValueAroundAllocation(VmaAllocator hAllocator, VkDeviceSize allocOffset, VkDeviceSize allocSize); - VkResult ValidateMagicValueAroundAllocation(VmaAllocator hAllocator, VkDeviceSize allocOffset, VkDeviceSize allocSize); - - VkResult BindBufferMemory( - const VmaAllocator hAllocator, - const VmaAllocation hAllocation, - VkDeviceSize allocationLocalOffset, - VkBuffer hBuffer, - const void* pNext); - VkResult BindImageMemory( - const VmaAllocator hAllocator, - const VmaAllocation hAllocation, - VkDeviceSize allocationLocalOffset, - VkImage hImage, - const void* pNext); - -private: - VmaBlockVector* m_ParentBlockVector = VMA_NULL; - VmaPool m_hParentPool = VK_NULL_HANDLE; // VK_NULL_HANDLE if not belongs to custom pool. - uint32_t m_MemoryTypeIndex = UINT32_MAX; - uint32_t m_Id = 0; - VkDeviceMemory m_hMemory = VK_NULL_HANDLE; - - /* - Protects access to m_hMemory so it's not used by multiple threads simultaneously, e.g. vkMapMemory, vkBindBufferMemory. - Also protects m_MapCount, m_pMappedData. - Allocations, deallocations, any change in m_pMetadata is protected by parent's VmaBlockVector::m_Mutex. - */ - VMA_MUTEX m_Mutex; - uint32_t m_MapCount = 0; - void* m_pMappedData = VMA_NULL; -}; - -struct VmaDefragmentationMove -{ - size_t srcBlockIndex; - size_t dstBlockIndex; - VkDeviceSize srcOffset; - VkDeviceSize dstOffset; - VkDeviceSize size; - VmaAllocation hAllocation; - VmaDeviceMemoryBlock* pSrcBlock; - VmaDeviceMemoryBlock* pDstBlock; -}; - -class VmaDefragmentationAlgorithm; - -/* -Sequence of VmaDeviceMemoryBlock. Represents memory blocks allocated for a specific -Vulkan memory type. - -Synchronized internally with a mutex. -*/ -struct VmaBlockVector -{ - VMA_CLASS_NO_COPY(VmaBlockVector) -public: - VmaBlockVector( - VmaAllocator hAllocator, - VmaPool hParentPool, - uint32_t memoryTypeIndex, - VkDeviceSize preferredBlockSize, - size_t minBlockCount, - size_t maxBlockCount, - VkDeviceSize bufferImageGranularity, - uint32_t frameInUseCount, - bool explicitBlockSize, - uint32_t algorithm, - float priority, - VkDeviceSize minAllocationAlignment, - void* pMemoryAllocateNext); - ~VmaBlockVector(); - - VkResult CreateMinBlocks(); - - VmaAllocator GetAllocator() const { return m_hAllocator; } - VmaPool GetParentPool() const { return m_hParentPool; } - bool IsCustomPool() const { return m_hParentPool != VMA_NULL; } - uint32_t GetMemoryTypeIndex() const { return m_MemoryTypeIndex; } - VkDeviceSize GetPreferredBlockSize() const { return m_PreferredBlockSize; } - VkDeviceSize GetBufferImageGranularity() const { return m_BufferImageGranularity; } - uint32_t GetFrameInUseCount() const { return m_FrameInUseCount; } - uint32_t GetAlgorithm() const { return m_Algorithm; } - - void GetPoolStats(VmaPoolStats* pStats); - - bool IsEmpty(); - bool IsCorruptionDetectionEnabled() const; - - VkResult Allocate( - uint32_t currentFrameIndex, - VkDeviceSize size, - VkDeviceSize alignment, - const VmaAllocationCreateInfo& createInfo, - VmaSuballocationType suballocType, - size_t allocationCount, - VmaAllocation* pAllocations); - - void Free(const VmaAllocation hAllocation); - - // Adds statistics of this BlockVector to pStats. - void AddStats(VmaStats* pStats); - -#if VMA_STATS_STRING_ENABLED - void PrintDetailedMap(class VmaJsonWriter& json); -#endif - - void MakePoolAllocationsLost( - uint32_t currentFrameIndex, - size_t* pLostAllocationCount); - VkResult CheckCorruption(); - - // Saves results in pCtx->res. - void Defragment( - class VmaBlockVectorDefragmentationContext* pCtx, - VmaDefragmentationStats* pStats, VmaDefragmentationFlags flags, - VkDeviceSize& maxCpuBytesToMove, uint32_t& maxCpuAllocationsToMove, - VkDeviceSize& maxGpuBytesToMove, uint32_t& maxGpuAllocationsToMove, - VkCommandBuffer commandBuffer); - void DefragmentationEnd( - class VmaBlockVectorDefragmentationContext* pCtx, - uint32_t flags, - VmaDefragmentationStats* pStats); - - uint32_t ProcessDefragmentations( - class VmaBlockVectorDefragmentationContext *pCtx, - VmaDefragmentationPassMoveInfo* pMove, uint32_t maxMoves); - - void CommitDefragmentations( - class VmaBlockVectorDefragmentationContext *pCtx, - VmaDefragmentationStats* pStats); - - //////////////////////////////////////////////////////////////////////////////// - // To be used only while the m_Mutex is locked. Used during defragmentation. - - size_t GetBlockCount() const { return m_Blocks.size(); } - VmaDeviceMemoryBlock* GetBlock(size_t index) const { return m_Blocks[index]; } - size_t CalcAllocationCount() const; - bool IsBufferImageGranularityConflictPossible() const; - -private: - friend class VmaDefragmentationAlgorithm_Generic; - - const VmaAllocator m_hAllocator; - const VmaPool m_hParentPool; - const uint32_t m_MemoryTypeIndex; - const VkDeviceSize m_PreferredBlockSize; - const size_t m_MinBlockCount; - const size_t m_MaxBlockCount; - const VkDeviceSize m_BufferImageGranularity; - const uint32_t m_FrameInUseCount; - const bool m_ExplicitBlockSize; - const uint32_t m_Algorithm; - const float m_Priority; - const VkDeviceSize m_MinAllocationAlignment; - void* const m_pMemoryAllocateNext; - VMA_RW_MUTEX m_Mutex; - - /* There can be at most one allocation that is completely empty (except when minBlockCount > 0) - - a hysteresis to avoid pessimistic case of alternating creation and destruction of a VkDeviceMemory. */ - bool m_HasEmptyBlock; - // Incrementally sorted by sumFreeSize, ascending. - VmaVector< VmaDeviceMemoryBlock*, VmaStlAllocator<VmaDeviceMemoryBlock*> > m_Blocks; - uint32_t m_NextBlockId; - - VkDeviceSize CalcMaxBlockSize() const; - - // Finds and removes given block from vector. - void Remove(VmaDeviceMemoryBlock* pBlock); - - // Performs single step in sorting m_Blocks. They may not be fully sorted - // after this call. - void IncrementallySortBlocks(); - - VkResult AllocatePage( - uint32_t currentFrameIndex, - VkDeviceSize size, - VkDeviceSize alignment, - const VmaAllocationCreateInfo& createInfo, - VmaSuballocationType suballocType, - VmaAllocation* pAllocation); - - // To be used only without CAN_MAKE_OTHER_LOST flag. - VkResult AllocateFromBlock( - VmaDeviceMemoryBlock* pBlock, - uint32_t currentFrameIndex, - VkDeviceSize size, - VkDeviceSize alignment, - VmaAllocationCreateFlags allocFlags, - void* pUserData, - VmaSuballocationType suballocType, - uint32_t strategy, - VmaAllocation* pAllocation); - - VkResult CreateBlock(VkDeviceSize blockSize, size_t* pNewBlockIndex); - - // Saves result to pCtx->res. - void ApplyDefragmentationMovesCpu( - class VmaBlockVectorDefragmentationContext* pDefragCtx, - const VmaVector< VmaDefragmentationMove, VmaStlAllocator<VmaDefragmentationMove> >& moves); - // Saves result to pCtx->res. - void ApplyDefragmentationMovesGpu( - class VmaBlockVectorDefragmentationContext* pDefragCtx, - VmaVector< VmaDefragmentationMove, VmaStlAllocator<VmaDefragmentationMove> >& moves, - VkCommandBuffer commandBuffer); - - /* - Used during defragmentation. pDefragmentationStats is optional. It's in/out - - updated with new data. - */ - void FreeEmptyBlocks(VmaDefragmentationStats* pDefragmentationStats); - - void UpdateHasEmptyBlock(); -}; - -struct VmaPool_T -{ - VMA_CLASS_NO_COPY(VmaPool_T) -public: - VmaBlockVector m_BlockVector; - - VmaPool_T( - VmaAllocator hAllocator, - const VmaPoolCreateInfo& createInfo, - VkDeviceSize preferredBlockSize); - ~VmaPool_T(); - - uint32_t GetId() const { return m_Id; } - void SetId(uint32_t id) { VMA_ASSERT(m_Id == 0); m_Id = id; } - - const char* GetName() const { return m_Name; } - void SetName(const char* pName); - -#if VMA_STATS_STRING_ENABLED - //void PrintDetailedMap(class VmaStringBuilder& sb); -#endif - -private: - uint32_t m_Id; - char* m_Name; - VmaPool_T* m_PrevPool = VMA_NULL; - VmaPool_T* m_NextPool = VMA_NULL; - friend struct VmaPoolListItemTraits; -}; - -struct VmaPoolListItemTraits -{ - typedef VmaPool_T ItemType; - static ItemType* GetPrev(const ItemType* item) { return item->m_PrevPool; } - static ItemType* GetNext(const ItemType* item) { return item->m_NextPool; } - static ItemType*& AccessPrev(ItemType* item) { return item->m_PrevPool; } - static ItemType*& AccessNext(ItemType* item) { return item->m_NextPool; } -}; - -/* -Performs defragmentation: - -- Updates `pBlockVector->m_pMetadata`. -- Updates allocations by calling ChangeBlockAllocation() or ChangeOffset(). -- Does not move actual data, only returns requested moves as `moves`. -*/ -class VmaDefragmentationAlgorithm -{ - VMA_CLASS_NO_COPY(VmaDefragmentationAlgorithm) -public: - VmaDefragmentationAlgorithm( - VmaAllocator hAllocator, - VmaBlockVector* pBlockVector, - uint32_t currentFrameIndex) : - m_hAllocator(hAllocator), - m_pBlockVector(pBlockVector), - m_CurrentFrameIndex(currentFrameIndex) - { - } - virtual ~VmaDefragmentationAlgorithm() - { - } - - virtual void AddAllocation(VmaAllocation hAlloc, VkBool32* pChanged) = 0; - virtual void AddAll() = 0; - - virtual VkResult Defragment( - VmaVector< VmaDefragmentationMove, VmaStlAllocator<VmaDefragmentationMove> >& moves, - VkDeviceSize maxBytesToMove, - uint32_t maxAllocationsToMove, - VmaDefragmentationFlags flags) = 0; - - virtual VkDeviceSize GetBytesMoved() const = 0; - virtual uint32_t GetAllocationsMoved() const = 0; - -protected: - VmaAllocator const m_hAllocator; - VmaBlockVector* const m_pBlockVector; - const uint32_t m_CurrentFrameIndex; - - struct AllocationInfo - { - VmaAllocation m_hAllocation; - VkBool32* m_pChanged; - - AllocationInfo() : - m_hAllocation(VK_NULL_HANDLE), - m_pChanged(VMA_NULL) - { - } - AllocationInfo(VmaAllocation hAlloc, VkBool32* pChanged) : - m_hAllocation(hAlloc), - m_pChanged(pChanged) - { - } - }; -}; - -class VmaDefragmentationAlgorithm_Generic : public VmaDefragmentationAlgorithm -{ - VMA_CLASS_NO_COPY(VmaDefragmentationAlgorithm_Generic) -public: - VmaDefragmentationAlgorithm_Generic( - VmaAllocator hAllocator, - VmaBlockVector* pBlockVector, - uint32_t currentFrameIndex, - bool overlappingMoveSupported); - virtual ~VmaDefragmentationAlgorithm_Generic(); - - virtual void AddAllocation(VmaAllocation hAlloc, VkBool32* pChanged); - virtual void AddAll() { m_AllAllocations = true; } - - virtual VkResult Defragment( - VmaVector< VmaDefragmentationMove, VmaStlAllocator<VmaDefragmentationMove> >& moves, - VkDeviceSize maxBytesToMove, - uint32_t maxAllocationsToMove, - VmaDefragmentationFlags flags); - - virtual VkDeviceSize GetBytesMoved() const { return m_BytesMoved; } - virtual uint32_t GetAllocationsMoved() const { return m_AllocationsMoved; } - -private: - uint32_t m_AllocationCount; - bool m_AllAllocations; - - VkDeviceSize m_BytesMoved; - uint32_t m_AllocationsMoved; - - struct AllocationInfoSizeGreater - { - bool operator()(const AllocationInfo& lhs, const AllocationInfo& rhs) const - { - return lhs.m_hAllocation->GetSize() > rhs.m_hAllocation->GetSize(); - } - }; - - struct AllocationInfoOffsetGreater - { - bool operator()(const AllocationInfo& lhs, const AllocationInfo& rhs) const - { - return lhs.m_hAllocation->GetOffset() > rhs.m_hAllocation->GetOffset(); - } - }; - - struct BlockInfo - { - size_t m_OriginalBlockIndex; - VmaDeviceMemoryBlock* m_pBlock; - bool m_HasNonMovableAllocations; - VmaVector< AllocationInfo, VmaStlAllocator<AllocationInfo> > m_Allocations; - - BlockInfo(const VkAllocationCallbacks* pAllocationCallbacks) : - m_OriginalBlockIndex(SIZE_MAX), - m_pBlock(VMA_NULL), - m_HasNonMovableAllocations(true), - m_Allocations(pAllocationCallbacks) - { - } - - void CalcHasNonMovableAllocations() - { - const size_t blockAllocCount = m_pBlock->m_pMetadata->GetAllocationCount(); - const size_t defragmentAllocCount = m_Allocations.size(); - m_HasNonMovableAllocations = blockAllocCount != defragmentAllocCount; - } - - void SortAllocationsBySizeDescending() - { - VMA_SORT(m_Allocations.begin(), m_Allocations.end(), AllocationInfoSizeGreater()); - } - - void SortAllocationsByOffsetDescending() - { - VMA_SORT(m_Allocations.begin(), m_Allocations.end(), AllocationInfoOffsetGreater()); - } - }; - - struct BlockPointerLess - { - bool operator()(const BlockInfo* pLhsBlockInfo, const VmaDeviceMemoryBlock* pRhsBlock) const - { - return pLhsBlockInfo->m_pBlock < pRhsBlock; - } - bool operator()(const BlockInfo* pLhsBlockInfo, const BlockInfo* pRhsBlockInfo) const - { - return pLhsBlockInfo->m_pBlock < pRhsBlockInfo->m_pBlock; - } - }; - - // 1. Blocks with some non-movable allocations go first. - // 2. Blocks with smaller sumFreeSize go first. - struct BlockInfoCompareMoveDestination - { - bool operator()(const BlockInfo* pLhsBlockInfo, const BlockInfo* pRhsBlockInfo) const - { - if(pLhsBlockInfo->m_HasNonMovableAllocations && !pRhsBlockInfo->m_HasNonMovableAllocations) - { - return true; - } - if(!pLhsBlockInfo->m_HasNonMovableAllocations && pRhsBlockInfo->m_HasNonMovableAllocations) - { - return false; - } - if(pLhsBlockInfo->m_pBlock->m_pMetadata->GetSumFreeSize() < pRhsBlockInfo->m_pBlock->m_pMetadata->GetSumFreeSize()) - { - return true; - } - return false; - } - }; - - typedef VmaVector< BlockInfo*, VmaStlAllocator<BlockInfo*> > BlockInfoVector; - BlockInfoVector m_Blocks; - - VkResult DefragmentRound( - VmaVector< VmaDefragmentationMove, VmaStlAllocator<VmaDefragmentationMove> >& moves, - VkDeviceSize maxBytesToMove, - uint32_t maxAllocationsToMove, - bool freeOldAllocations); - - size_t CalcBlocksWithNonMovableCount() const; - - static bool MoveMakesSense( - size_t dstBlockIndex, VkDeviceSize dstOffset, - size_t srcBlockIndex, VkDeviceSize srcOffset); -}; - -class VmaDefragmentationAlgorithm_Fast : public VmaDefragmentationAlgorithm -{ - VMA_CLASS_NO_COPY(VmaDefragmentationAlgorithm_Fast) -public: - VmaDefragmentationAlgorithm_Fast( - VmaAllocator hAllocator, - VmaBlockVector* pBlockVector, - uint32_t currentFrameIndex, - bool overlappingMoveSupported); - virtual ~VmaDefragmentationAlgorithm_Fast(); - - virtual void AddAllocation(VmaAllocation hAlloc, VkBool32* pChanged) { ++m_AllocationCount; } - virtual void AddAll() { m_AllAllocations = true; } - - virtual VkResult Defragment( - VmaVector< VmaDefragmentationMove, VmaStlAllocator<VmaDefragmentationMove> >& moves, - VkDeviceSize maxBytesToMove, - uint32_t maxAllocationsToMove, - VmaDefragmentationFlags flags); - - virtual VkDeviceSize GetBytesMoved() const { return m_BytesMoved; } - virtual uint32_t GetAllocationsMoved() const { return m_AllocationsMoved; } - -private: - struct BlockInfo - { - size_t origBlockIndex; - }; - - class FreeSpaceDatabase - { - public: - FreeSpaceDatabase() - { - FreeSpace s = {}; - s.blockInfoIndex = SIZE_MAX; - for(size_t i = 0; i < MAX_COUNT; ++i) - { - m_FreeSpaces[i] = s; - } - } - - void Register(size_t blockInfoIndex, VkDeviceSize offset, VkDeviceSize size) - { - if(size < VMA_MIN_FREE_SUBALLOCATION_SIZE_TO_REGISTER) - { - return; - } - - // Find first invalid or the smallest structure. - size_t bestIndex = SIZE_MAX; - for(size_t i = 0; i < MAX_COUNT; ++i) - { - // Empty structure. - if(m_FreeSpaces[i].blockInfoIndex == SIZE_MAX) - { - bestIndex = i; - break; - } - if(m_FreeSpaces[i].size < size && - (bestIndex == SIZE_MAX || m_FreeSpaces[bestIndex].size > m_FreeSpaces[i].size)) - { - bestIndex = i; - } - } - - if(bestIndex != SIZE_MAX) - { - m_FreeSpaces[bestIndex].blockInfoIndex = blockInfoIndex; - m_FreeSpaces[bestIndex].offset = offset; - m_FreeSpaces[bestIndex].size = size; - } - } - - bool Fetch(VkDeviceSize alignment, VkDeviceSize size, - size_t& outBlockInfoIndex, VkDeviceSize& outDstOffset) - { - size_t bestIndex = SIZE_MAX; - VkDeviceSize bestFreeSpaceAfter = 0; - for(size_t i = 0; i < MAX_COUNT; ++i) - { - // Structure is valid. - if(m_FreeSpaces[i].blockInfoIndex != SIZE_MAX) - { - const VkDeviceSize dstOffset = VmaAlignUp(m_FreeSpaces[i].offset, alignment); - // Allocation fits into this structure. - if(dstOffset + size <= m_FreeSpaces[i].offset + m_FreeSpaces[i].size) - { - const VkDeviceSize freeSpaceAfter = (m_FreeSpaces[i].offset + m_FreeSpaces[i].size) - - (dstOffset + size); - if(bestIndex == SIZE_MAX || freeSpaceAfter > bestFreeSpaceAfter) - { - bestIndex = i; - bestFreeSpaceAfter = freeSpaceAfter; - } - } - } - } - - if(bestIndex != SIZE_MAX) - { - outBlockInfoIndex = m_FreeSpaces[bestIndex].blockInfoIndex; - outDstOffset = VmaAlignUp(m_FreeSpaces[bestIndex].offset, alignment); - - if(bestFreeSpaceAfter >= VMA_MIN_FREE_SUBALLOCATION_SIZE_TO_REGISTER) - { - // Leave this structure for remaining empty space. - const VkDeviceSize alignmentPlusSize = (outDstOffset - m_FreeSpaces[bestIndex].offset) + size; - m_FreeSpaces[bestIndex].offset += alignmentPlusSize; - m_FreeSpaces[bestIndex].size -= alignmentPlusSize; - } - else - { - // This structure becomes invalid. - m_FreeSpaces[bestIndex].blockInfoIndex = SIZE_MAX; - } - - return true; - } - - return false; - } - - private: - static const size_t MAX_COUNT = 4; - - struct FreeSpace - { - size_t blockInfoIndex; // SIZE_MAX means this structure is invalid. - VkDeviceSize offset; - VkDeviceSize size; - } m_FreeSpaces[MAX_COUNT]; - }; - - const bool m_OverlappingMoveSupported; - - uint32_t m_AllocationCount; - bool m_AllAllocations; - - VkDeviceSize m_BytesMoved; - uint32_t m_AllocationsMoved; - - VmaVector< BlockInfo, VmaStlAllocator<BlockInfo> > m_BlockInfos; - - void PreprocessMetadata(); - void PostprocessMetadata(); - void InsertSuballoc(VmaBlockMetadata_Generic* pMetadata, const VmaSuballocation& suballoc); -}; - -struct VmaBlockDefragmentationContext -{ - enum BLOCK_FLAG - { - BLOCK_FLAG_USED = 0x00000001, - }; - uint32_t flags; - VkBuffer hBuffer; -}; - -class VmaBlockVectorDefragmentationContext -{ - VMA_CLASS_NO_COPY(VmaBlockVectorDefragmentationContext) -public: - VkResult res; - bool mutexLocked; - VmaVector< VmaBlockDefragmentationContext, VmaStlAllocator<VmaBlockDefragmentationContext> > blockContexts; - VmaVector< VmaDefragmentationMove, VmaStlAllocator<VmaDefragmentationMove> > defragmentationMoves; - uint32_t defragmentationMovesProcessed; - uint32_t defragmentationMovesCommitted; - bool hasDefragmentationPlan; - - VmaBlockVectorDefragmentationContext( - VmaAllocator hAllocator, - VmaPool hCustomPool, // Optional. - VmaBlockVector* pBlockVector, - uint32_t currFrameIndex); - ~VmaBlockVectorDefragmentationContext(); - - VmaPool GetCustomPool() const { return m_hCustomPool; } - VmaBlockVector* GetBlockVector() const { return m_pBlockVector; } - VmaDefragmentationAlgorithm* GetAlgorithm() const { return m_pAlgorithm; } - - void AddAllocation(VmaAllocation hAlloc, VkBool32* pChanged); - void AddAll() { m_AllAllocations = true; } - - void Begin(bool overlappingMoveSupported, VmaDefragmentationFlags flags); - -private: - const VmaAllocator m_hAllocator; - // Null if not from custom pool. - const VmaPool m_hCustomPool; - // Redundant, for convenience not to fetch from m_hCustomPool->m_BlockVector or m_hAllocator->m_pBlockVectors. - VmaBlockVector* const m_pBlockVector; - const uint32_t m_CurrFrameIndex; - // Owner of this object. - VmaDefragmentationAlgorithm* m_pAlgorithm; - - struct AllocInfo - { - VmaAllocation hAlloc; - VkBool32* pChanged; - }; - // Used between constructor and Begin. - VmaVector< AllocInfo, VmaStlAllocator<AllocInfo> > m_Allocations; - bool m_AllAllocations; -}; - -struct VmaDefragmentationContext_T -{ -private: - VMA_CLASS_NO_COPY(VmaDefragmentationContext_T) -public: - VmaDefragmentationContext_T( - VmaAllocator hAllocator, - uint32_t currFrameIndex, - uint32_t flags, - VmaDefragmentationStats* pStats); - ~VmaDefragmentationContext_T(); - - void AddPools(uint32_t poolCount, const VmaPool* pPools); - void AddAllocations( - uint32_t allocationCount, - const VmaAllocation* pAllocations, - VkBool32* pAllocationsChanged); - - /* - Returns: - - `VK_SUCCESS` if succeeded and object can be destroyed immediately. - - `VK_NOT_READY` if succeeded but the object must remain alive until vmaDefragmentationEnd(). - - Negative value if error occurred and object can be destroyed immediately. - */ - VkResult Defragment( - VkDeviceSize maxCpuBytesToMove, uint32_t maxCpuAllocationsToMove, - VkDeviceSize maxGpuBytesToMove, uint32_t maxGpuAllocationsToMove, - VkCommandBuffer commandBuffer, VmaDefragmentationStats* pStats, VmaDefragmentationFlags flags); - - VkResult DefragmentPassBegin(VmaDefragmentationPassInfo* pInfo); - VkResult DefragmentPassEnd(); - -private: - const VmaAllocator m_hAllocator; - const uint32_t m_CurrFrameIndex; - const uint32_t m_Flags; - VmaDefragmentationStats* const m_pStats; - - VkDeviceSize m_MaxCpuBytesToMove; - uint32_t m_MaxCpuAllocationsToMove; - VkDeviceSize m_MaxGpuBytesToMove; - uint32_t m_MaxGpuAllocationsToMove; - - // Owner of these objects. - VmaBlockVectorDefragmentationContext* m_DefaultPoolContexts[VK_MAX_MEMORY_TYPES]; - // Owner of these objects. - VmaVector< VmaBlockVectorDefragmentationContext*, VmaStlAllocator<VmaBlockVectorDefragmentationContext*> > m_CustomPoolContexts; -}; - -#if VMA_RECORDING_ENABLED - -class VmaRecorder -{ -public: - VmaRecorder(); - VkResult Init(const VmaRecordSettings& settings, bool useMutex); - void WriteConfiguration( - const VkPhysicalDeviceProperties& devProps, - const VkPhysicalDeviceMemoryProperties& memProps, - uint32_t vulkanApiVersion, - bool dedicatedAllocationExtensionEnabled, - bool bindMemory2ExtensionEnabled, - bool memoryBudgetExtensionEnabled, - bool deviceCoherentMemoryExtensionEnabled); - ~VmaRecorder(); - - void RecordCreateAllocator(uint32_t frameIndex); - void RecordDestroyAllocator(uint32_t frameIndex); - void RecordCreatePool(uint32_t frameIndex, - const VmaPoolCreateInfo& createInfo, - VmaPool pool); - void RecordDestroyPool(uint32_t frameIndex, VmaPool pool); - void RecordAllocateMemory(uint32_t frameIndex, - const VkMemoryRequirements& vkMemReq, - const VmaAllocationCreateInfo& createInfo, - VmaAllocation allocation); - void RecordAllocateMemoryPages(uint32_t frameIndex, - const VkMemoryRequirements& vkMemReq, - const VmaAllocationCreateInfo& createInfo, - uint64_t allocationCount, - const VmaAllocation* pAllocations); - void RecordAllocateMemoryForBuffer(uint32_t frameIndex, - const VkMemoryRequirements& vkMemReq, - bool requiresDedicatedAllocation, - bool prefersDedicatedAllocation, - const VmaAllocationCreateInfo& createInfo, - VmaAllocation allocation); - void RecordAllocateMemoryForImage(uint32_t frameIndex, - const VkMemoryRequirements& vkMemReq, - bool requiresDedicatedAllocation, - bool prefersDedicatedAllocation, - const VmaAllocationCreateInfo& createInfo, - VmaAllocation allocation); - void RecordFreeMemory(uint32_t frameIndex, - VmaAllocation allocation); - void RecordFreeMemoryPages(uint32_t frameIndex, - uint64_t allocationCount, - const VmaAllocation* pAllocations); - void RecordSetAllocationUserData(uint32_t frameIndex, - VmaAllocation allocation, - const void* pUserData); - void RecordCreateLostAllocation(uint32_t frameIndex, - VmaAllocation allocation); - void RecordMapMemory(uint32_t frameIndex, - VmaAllocation allocation); - void RecordUnmapMemory(uint32_t frameIndex, - VmaAllocation allocation); - void RecordFlushAllocation(uint32_t frameIndex, - VmaAllocation allocation, VkDeviceSize offset, VkDeviceSize size); - void RecordInvalidateAllocation(uint32_t frameIndex, - VmaAllocation allocation, VkDeviceSize offset, VkDeviceSize size); - void RecordCreateBuffer(uint32_t frameIndex, - const VkBufferCreateInfo& bufCreateInfo, - const VmaAllocationCreateInfo& allocCreateInfo, - VmaAllocation allocation); - void RecordCreateImage(uint32_t frameIndex, - const VkImageCreateInfo& imageCreateInfo, - const VmaAllocationCreateInfo& allocCreateInfo, - VmaAllocation allocation); - void RecordDestroyBuffer(uint32_t frameIndex, - VmaAllocation allocation); - void RecordDestroyImage(uint32_t frameIndex, - VmaAllocation allocation); - void RecordTouchAllocation(uint32_t frameIndex, - VmaAllocation allocation); - void RecordGetAllocationInfo(uint32_t frameIndex, - VmaAllocation allocation); - void RecordMakePoolAllocationsLost(uint32_t frameIndex, - VmaPool pool); - void RecordDefragmentationBegin(uint32_t frameIndex, - const VmaDefragmentationInfo2& info, - VmaDefragmentationContext ctx); - void RecordDefragmentationEnd(uint32_t frameIndex, - VmaDefragmentationContext ctx); - void RecordSetPoolName(uint32_t frameIndex, - VmaPool pool, - const char* name); - -private: - struct CallParams - { - uint32_t threadId; - double time; - }; - - class UserDataString - { - public: - UserDataString(VmaAllocationCreateFlags allocFlags, const void* pUserData); - const char* GetString() const { return m_Str; } - - private: - char m_PtrStr[17]; - const char* m_Str; - }; - - bool m_UseMutex; - VmaRecordFlags m_Flags; - FILE* m_File; - VMA_MUTEX m_FileMutex; - std::chrono::time_point<std::chrono::high_resolution_clock> m_RecordingStartTime; - - void GetBasicParams(CallParams& outParams); - - // T must be a pointer type, e.g. VmaAllocation, VmaPool. - template<typename T> - void PrintPointerList(uint64_t count, const T* pItems) - { - if(count) - { - fprintf(m_File, "%p", pItems[0]); - for(uint64_t i = 1; i < count; ++i) - { - fprintf(m_File, " %p", pItems[i]); - } - } - } - - void PrintPointerList(uint64_t count, const VmaAllocation* pItems); - void Flush(); -}; - -#endif // #if VMA_RECORDING_ENABLED - -/* -Thread-safe wrapper over VmaPoolAllocator free list, for allocation of VmaAllocation_T objects. -*/ -class VmaAllocationObjectAllocator -{ - VMA_CLASS_NO_COPY(VmaAllocationObjectAllocator) -public: - VmaAllocationObjectAllocator(const VkAllocationCallbacks* pAllocationCallbacks); - - template<typename... Types> VmaAllocation Allocate(Types... args); - void Free(VmaAllocation hAlloc); - -private: - VMA_MUTEX m_Mutex; - VmaPoolAllocator<VmaAllocation_T> m_Allocator; -}; - -struct VmaCurrentBudgetData -{ - VMA_ATOMIC_UINT64 m_BlockBytes[VK_MAX_MEMORY_HEAPS]; - VMA_ATOMIC_UINT64 m_AllocationBytes[VK_MAX_MEMORY_HEAPS]; - -#if VMA_MEMORY_BUDGET - VMA_ATOMIC_UINT32 m_OperationsSinceBudgetFetch; - VMA_RW_MUTEX m_BudgetMutex; - uint64_t m_VulkanUsage[VK_MAX_MEMORY_HEAPS]; - uint64_t m_VulkanBudget[VK_MAX_MEMORY_HEAPS]; - uint64_t m_BlockBytesAtBudgetFetch[VK_MAX_MEMORY_HEAPS]; -#endif // #if VMA_MEMORY_BUDGET - - VmaCurrentBudgetData() - { - for(uint32_t heapIndex = 0; heapIndex < VK_MAX_MEMORY_HEAPS; ++heapIndex) - { - m_BlockBytes[heapIndex] = 0; - m_AllocationBytes[heapIndex] = 0; -#if VMA_MEMORY_BUDGET - m_VulkanUsage[heapIndex] = 0; - m_VulkanBudget[heapIndex] = 0; - m_BlockBytesAtBudgetFetch[heapIndex] = 0; -#endif - } - -#if VMA_MEMORY_BUDGET - m_OperationsSinceBudgetFetch = 0; -#endif - } - - void AddAllocation(uint32_t heapIndex, VkDeviceSize allocationSize) - { - m_AllocationBytes[heapIndex] += allocationSize; -#if VMA_MEMORY_BUDGET - ++m_OperationsSinceBudgetFetch; -#endif - } - - void RemoveAllocation(uint32_t heapIndex, VkDeviceSize allocationSize) - { - VMA_ASSERT(m_AllocationBytes[heapIndex] >= allocationSize); // DELME - m_AllocationBytes[heapIndex] -= allocationSize; -#if VMA_MEMORY_BUDGET - ++m_OperationsSinceBudgetFetch; -#endif - } -}; - -// Main allocator object. -struct VmaAllocator_T -{ - VMA_CLASS_NO_COPY(VmaAllocator_T) -public: - bool m_UseMutex; - uint32_t m_VulkanApiVersion; - bool m_UseKhrDedicatedAllocation; // Can be set only if m_VulkanApiVersion < VK_MAKE_VERSION(1, 1, 0). - bool m_UseKhrBindMemory2; // Can be set only if m_VulkanApiVersion < VK_MAKE_VERSION(1, 1, 0). - bool m_UseExtMemoryBudget; - bool m_UseAmdDeviceCoherentMemory; - bool m_UseKhrBufferDeviceAddress; - bool m_UseExtMemoryPriority; - VkDevice m_hDevice; - VkInstance m_hInstance; - bool m_AllocationCallbacksSpecified; - VkAllocationCallbacks m_AllocationCallbacks; - VmaDeviceMemoryCallbacks m_DeviceMemoryCallbacks; - VmaAllocationObjectAllocator m_AllocationObjectAllocator; - - // Each bit (1 << i) is set if HeapSizeLimit is enabled for that heap, so cannot allocate more than the heap size. - uint32_t m_HeapSizeLimitMask; - - VkPhysicalDeviceProperties m_PhysicalDeviceProperties; - VkPhysicalDeviceMemoryProperties m_MemProps; - - // Default pools. - VmaBlockVector* m_pBlockVectors[VK_MAX_MEMORY_TYPES]; - VmaBlockVector* m_pSmallBufferBlockVectors[VK_MAX_MEMORY_TYPES]; - - typedef VmaIntrusiveLinkedList<VmaDedicatedAllocationListItemTraits> DedicatedAllocationLinkedList; - DedicatedAllocationLinkedList m_DedicatedAllocations[VK_MAX_MEMORY_TYPES]; - VMA_RW_MUTEX m_DedicatedAllocationsMutex[VK_MAX_MEMORY_TYPES]; - - VmaCurrentBudgetData m_Budget; - VMA_ATOMIC_UINT32 m_DeviceMemoryCount; // Total number of VkDeviceMemory objects. - - VmaAllocator_T(const VmaAllocatorCreateInfo* pCreateInfo); - VkResult Init(const VmaAllocatorCreateInfo* pCreateInfo); - ~VmaAllocator_T(); - - const VkAllocationCallbacks* GetAllocationCallbacks() const - { - return m_AllocationCallbacksSpecified ? &m_AllocationCallbacks : 0; - } - const VmaVulkanFunctions& GetVulkanFunctions() const - { - return m_VulkanFunctions; - } - - VkPhysicalDevice GetPhysicalDevice() const { return m_PhysicalDevice; } - - VkDeviceSize GetBufferImageGranularity() const - { - return VMA_MAX( - static_cast<VkDeviceSize>(VMA_DEBUG_MIN_BUFFER_IMAGE_GRANULARITY), - m_PhysicalDeviceProperties.limits.bufferImageGranularity); - } - - uint32_t GetMemoryHeapCount() const { return m_MemProps.memoryHeapCount; } - uint32_t GetMemoryTypeCount() const { return m_MemProps.memoryTypeCount; } - - uint32_t MemoryTypeIndexToHeapIndex(uint32_t memTypeIndex) const - { - VMA_ASSERT(memTypeIndex < m_MemProps.memoryTypeCount); - return m_MemProps.memoryTypes[memTypeIndex].heapIndex; - } - // True when specific memory type is HOST_VISIBLE but not HOST_COHERENT. - bool IsMemoryTypeNonCoherent(uint32_t memTypeIndex) const - { - return (m_MemProps.memoryTypes[memTypeIndex].propertyFlags & (VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT | VK_MEMORY_PROPERTY_HOST_COHERENT_BIT)) == - VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT; - } - // Minimum alignment for all allocations in specific memory type. - VkDeviceSize GetMemoryTypeMinAlignment(uint32_t memTypeIndex) const - { - return IsMemoryTypeNonCoherent(memTypeIndex) ? - VMA_MAX((VkDeviceSize)VMA_MIN_ALIGNMENT, m_PhysicalDeviceProperties.limits.nonCoherentAtomSize) : - (VkDeviceSize)VMA_MIN_ALIGNMENT; - } - - bool IsIntegratedGpu() const - { - return m_PhysicalDeviceProperties.deviceType == VK_PHYSICAL_DEVICE_TYPE_INTEGRATED_GPU; - } - - uint32_t GetGlobalMemoryTypeBits() const { return m_GlobalMemoryTypeBits; } - -#if VMA_RECORDING_ENABLED - VmaRecorder* GetRecorder() const { return m_pRecorder; } -#endif - - void GetBufferMemoryRequirements( - VkBuffer hBuffer, - VkMemoryRequirements& memReq, - bool& requiresDedicatedAllocation, - bool& prefersDedicatedAllocation) const; - void GetImageMemoryRequirements( - VkImage hImage, - VkMemoryRequirements& memReq, - bool& requiresDedicatedAllocation, - bool& prefersDedicatedAllocation) const; - - // Main allocation function. - VkResult AllocateMemory( - const VkMemoryRequirements& vkMemReq, - bool requiresDedicatedAllocation, - bool prefersDedicatedAllocation, - VkBuffer dedicatedBuffer, - VkBufferUsageFlags dedicatedBufferUsage, // UINT32_MAX when unknown. - VkImage dedicatedImage, - const VmaAllocationCreateInfo& createInfo, - VmaSuballocationType suballocType, - size_t allocationCount, - VmaAllocation* pAllocations); - - // Main deallocation function. - void FreeMemory( - size_t allocationCount, - const VmaAllocation* pAllocations); - - void CalculateStats(VmaStats* pStats); - - void GetBudget( - VmaBudget* outBudget, uint32_t firstHeap, uint32_t heapCount); - -#if VMA_STATS_STRING_ENABLED - void PrintDetailedMap(class VmaJsonWriter& json); -#endif - - VkResult DefragmentationBegin( - const VmaDefragmentationInfo2& info, - VmaDefragmentationStats* pStats, - VmaDefragmentationContext* pContext); - VkResult DefragmentationEnd( - VmaDefragmentationContext context); - - VkResult DefragmentationPassBegin( - VmaDefragmentationPassInfo* pInfo, - VmaDefragmentationContext context); - VkResult DefragmentationPassEnd( - VmaDefragmentationContext context); - - void GetAllocationInfo(VmaAllocation hAllocation, VmaAllocationInfo* pAllocationInfo); - bool TouchAllocation(VmaAllocation hAllocation); - - VkResult CreatePool(const VmaPoolCreateInfo* pCreateInfo, VmaPool* pPool); - void DestroyPool(VmaPool pool); - void GetPoolStats(VmaPool pool, VmaPoolStats* pPoolStats); - - void SetCurrentFrameIndex(uint32_t frameIndex); - uint32_t GetCurrentFrameIndex() const { return m_CurrentFrameIndex.load(); } - - void MakePoolAllocationsLost( - VmaPool hPool, - size_t* pLostAllocationCount); - VkResult CheckPoolCorruption(VmaPool hPool); - VkResult CheckCorruption(uint32_t memoryTypeBits); - - void CreateLostAllocation(VmaAllocation* pAllocation); - - // Call to Vulkan function vkAllocateMemory with accompanying bookkeeping. - VkResult AllocateVulkanMemory(const VkMemoryAllocateInfo* pAllocateInfo, VkDeviceMemory* pMemory); - // Call to Vulkan function vkFreeMemory with accompanying bookkeeping. - void FreeVulkanMemory(uint32_t memoryType, VkDeviceSize size, VkDeviceMemory hMemory); - // Call to Vulkan function vkBindBufferMemory or vkBindBufferMemory2KHR. - VkResult BindVulkanBuffer( - VkDeviceMemory memory, - VkDeviceSize memoryOffset, - VkBuffer buffer, - const void* pNext); - // Call to Vulkan function vkBindImageMemory or vkBindImageMemory2KHR. - VkResult BindVulkanImage( - VkDeviceMemory memory, - VkDeviceSize memoryOffset, - VkImage image, - const void* pNext); - - VkResult Map(VmaAllocation hAllocation, void** ppData); - void Unmap(VmaAllocation hAllocation); - - VkResult BindBufferMemory( - VmaAllocation hAllocation, - VkDeviceSize allocationLocalOffset, - VkBuffer hBuffer, - const void* pNext); - VkResult BindImageMemory( - VmaAllocation hAllocation, - VkDeviceSize allocationLocalOffset, - VkImage hImage, - const void* pNext); - - VkResult FlushOrInvalidateAllocation( - VmaAllocation hAllocation, - VkDeviceSize offset, VkDeviceSize size, - VMA_CACHE_OPERATION op); - VkResult FlushOrInvalidateAllocations( - uint32_t allocationCount, - const VmaAllocation* allocations, - const VkDeviceSize* offsets, const VkDeviceSize* sizes, - VMA_CACHE_OPERATION op); - - void FillAllocation(const VmaAllocation hAllocation, uint8_t pattern); - - /* - Returns bit mask of memory types that can support defragmentation on GPU as - they support creation of required buffer for copy operations. - */ - uint32_t GetGpuDefragmentationMemoryTypeBits(); - -#if VMA_EXTERNAL_MEMORY - VkExternalMemoryHandleTypeFlagsKHR GetExternalMemoryHandleTypeFlags(uint32_t memTypeIndex) const - { - return m_TypeExternalMemoryHandleTypes[memTypeIndex]; - } -#endif // #if VMA_EXTERNAL_MEMORY - -private: - VkDeviceSize m_PreferredLargeHeapBlockSize; - - VkPhysicalDevice m_PhysicalDevice; - VMA_ATOMIC_UINT32 m_CurrentFrameIndex; - VMA_ATOMIC_UINT32 m_GpuDefragmentationMemoryTypeBits; // UINT32_MAX means uninitialized. -#if VMA_EXTERNAL_MEMORY - VkExternalMemoryHandleTypeFlagsKHR m_TypeExternalMemoryHandleTypes[VK_MAX_MEMORY_TYPES]; -#endif // #if VMA_EXTERNAL_MEMORY - - VMA_RW_MUTEX m_PoolsMutex; - typedef VmaIntrusiveLinkedList<VmaPoolListItemTraits> PoolList; - // Protected by m_PoolsMutex. - PoolList m_Pools; - uint32_t m_NextPoolId; - - VmaVulkanFunctions m_VulkanFunctions; - - // Global bit mask AND-ed with any memoryTypeBits to disallow certain memory types. - uint32_t m_GlobalMemoryTypeBits; - -#if VMA_RECORDING_ENABLED - VmaRecorder* m_pRecorder; -#endif - - void ImportVulkanFunctions(const VmaVulkanFunctions* pVulkanFunctions); - -#if VMA_STATIC_VULKAN_FUNCTIONS == 1 - void ImportVulkanFunctions_Static(); -#endif - - void ImportVulkanFunctions_Custom(const VmaVulkanFunctions* pVulkanFunctions); - -#if VMA_DYNAMIC_VULKAN_FUNCTIONS == 1 - void ImportVulkanFunctions_Dynamic(); -#endif - - void ValidateVulkanFunctions(); - - VkDeviceSize CalcPreferredBlockSize(uint32_t memTypeIndex); - - VkResult AllocateMemoryOfType( - VkDeviceSize size, - VkDeviceSize alignment, - bool dedicatedAllocation, - VkBuffer dedicatedBuffer, - VkBufferUsageFlags dedicatedBufferUsage, - VkImage dedicatedImage, - const VmaAllocationCreateInfo& createInfo, - uint32_t memTypeIndex, - VmaSuballocationType suballocType, - size_t allocationCount, - VmaAllocation* pAllocations); - - // Helper function only to be used inside AllocateDedicatedMemory. - VkResult AllocateDedicatedMemoryPage( - VkDeviceSize size, - VmaSuballocationType suballocType, - uint32_t memTypeIndex, - const VkMemoryAllocateInfo& allocInfo, - bool map, - bool isUserDataString, - void* pUserData, - VmaAllocation* pAllocation); - - // Allocates and registers new VkDeviceMemory specifically for dedicated allocations. - VkResult AllocateDedicatedMemory( - VkDeviceSize size, - VmaSuballocationType suballocType, - uint32_t memTypeIndex, - bool withinBudget, - bool map, - bool isUserDataString, - void* pUserData, - float priority, - VkBuffer dedicatedBuffer, - VkBufferUsageFlags dedicatedBufferUsage, - VkImage dedicatedImage, - size_t allocationCount, - VmaAllocation* pAllocations); - - void FreeDedicatedMemory(const VmaAllocation allocation); - - /* - Calculates and returns bit mask of memory types that can support defragmentation - on GPU as they support creation of required buffer for copy operations. - */ - uint32_t CalculateGpuDefragmentationMemoryTypeBits() const; - - uint32_t CalculateGlobalMemoryTypeBits() const; - - bool GetFlushOrInvalidateRange( - VmaAllocation allocation, - VkDeviceSize offset, VkDeviceSize size, - VkMappedMemoryRange& outRange) const; - -#if VMA_MEMORY_BUDGET - void UpdateVulkanBudget(); -#endif // #if VMA_MEMORY_BUDGET -}; - -//////////////////////////////////////////////////////////////////////////////// -// Memory allocation #2 after VmaAllocator_T definition - -static void* VmaMalloc(VmaAllocator hAllocator, size_t size, size_t alignment) -{ - return VmaMalloc(&hAllocator->m_AllocationCallbacks, size, alignment); -} - -static void VmaFree(VmaAllocator hAllocator, void* ptr) -{ - VmaFree(&hAllocator->m_AllocationCallbacks, ptr); -} - -template<typename T> -static T* VmaAllocate(VmaAllocator hAllocator) -{ - return (T*)VmaMalloc(hAllocator, sizeof(T), VMA_ALIGN_OF(T)); -} - -template<typename T> -static T* VmaAllocateArray(VmaAllocator hAllocator, size_t count) -{ - return (T*)VmaMalloc(hAllocator, sizeof(T) * count, VMA_ALIGN_OF(T)); -} - -template<typename T> -static void vma_delete(VmaAllocator hAllocator, T* ptr) -{ - if(ptr != VMA_NULL) - { - ptr->~T(); - VmaFree(hAllocator, ptr); - } -} - -template<typename T> -static void vma_delete_array(VmaAllocator hAllocator, T* ptr, size_t count) -{ - if(ptr != VMA_NULL) - { - for(size_t i = count; i--; ) - ptr[i].~T(); - VmaFree(hAllocator, ptr); - } -} - -//////////////////////////////////////////////////////////////////////////////// -// VmaStringBuilder - -#if VMA_STATS_STRING_ENABLED - +#if !defined(_VMA_STRING_BUILDER) && VMA_STATS_STRING_ENABLED class VmaStringBuilder { public: - VmaStringBuilder(VmaAllocator alloc) : m_Data(VmaStlAllocator<char>(alloc->GetAllocationCallbacks())) { } + VmaStringBuilder(const VkAllocationCallbacks* allocationCallbacks) : m_Data(VmaStlAllocator<char>(allocationCallbacks)) {} + ~VmaStringBuilder() = default; + size_t GetLength() const { return m_Data.size(); } const char* GetData() const { return m_Data.data(); } - + void AddNewLine() { Add('\n'); } void Add(char ch) { m_Data.push_back(ch); } + void Add(const char* pStr); - void AddNewLine() { Add('\n'); } void AddNumber(uint32_t num); void AddNumber(uint64_t num); void AddPointer(const void* ptr); private: - VmaVector< char, VmaStlAllocator<char> > m_Data; + VmaVector<char, VmaStlAllocator<char>> m_Data; }; +#ifndef _VMA_STRING_BUILDER_FUNCTIONS void VmaStringBuilder::Add(const char* pStr) { const size_t strLen = strlen(pStr); - if(strLen > 0) + if (strLen > 0) { const size_t oldCount = m_Data.size(); m_Data.resize(oldCount + strLen); @@ -8618,13 +5163,12 @@ void VmaStringBuilder::AddNumber(uint32_t num) { char buf[11]; buf[10] = '\0'; - char *p = &buf[10]; + char* p = &buf[10]; do { *--p = '0' + (num % 10); num /= 10; - } - while(num); + } while (num); Add(p); } @@ -8632,13 +5176,12 @@ void VmaStringBuilder::AddNumber(uint64_t num) { char buf[21]; buf[20] = '\0'; - char *p = &buf[20]; + char* p = &buf[20]; do { *--p = '0' + (num % 10); num /= 10; - } - while(num); + } while (num); Add(p); } @@ -8648,43 +5191,65 @@ void VmaStringBuilder::AddPointer(const void* ptr) VmaPtrToStr(buf, sizeof(buf), ptr); Add(buf); } +#endif //_VMA_STRING_BUILDER_FUNCTIONS +#endif // _VMA_STRING_BUILDER -#endif // #if VMA_STATS_STRING_ENABLED - -//////////////////////////////////////////////////////////////////////////////// -// VmaJsonWriter - -#if VMA_STATS_STRING_ENABLED - +#if !defined(_VMA_JSON_WRITER) && VMA_STATS_STRING_ENABLED +/* +Allows to conveniently build a correct JSON document to be written to the +VmaStringBuilder passed to the constructor. +*/ class VmaJsonWriter { VMA_CLASS_NO_COPY(VmaJsonWriter) public: + // sb - string builder to write the document to. Must remain alive for the whole lifetime of this object. VmaJsonWriter(const VkAllocationCallbacks* pAllocationCallbacks, VmaStringBuilder& sb); ~VmaJsonWriter(); + // Begins object by writing "{". + // Inside an object, you must call pairs of WriteString and a value, e.g.: + // j.BeginObject(true); j.WriteString("A"); j.WriteNumber(1); j.WriteString("B"); j.WriteNumber(2); j.EndObject(); + // Will write: { "A": 1, "B": 2 } void BeginObject(bool singleLine = false); + // Ends object by writing "}". void EndObject(); + // Begins array by writing "[". + // Inside an array, you can write a sequence of any values. void BeginArray(bool singleLine = false); + // Ends array by writing "[". void EndArray(); + // Writes a string value inside "". + // pStr can contain any ANSI characters, including '"', new line etc. - they will be properly escaped. void WriteString(const char* pStr); + + // Begins writing a string value. + // Call BeginString, ContinueString, ContinueString, ..., EndString instead of + // WriteString to conveniently build the string content incrementally, made of + // parts including numbers. void BeginString(const char* pStr = VMA_NULL); + // Posts next part of an open string. void ContinueString(const char* pStr); + // Posts next part of an open string. The number is converted to decimal characters. void ContinueString(uint32_t n); void ContinueString(uint64_t n); + // Posts next part of an open string. Pointer value is converted to characters + // using "%p" formatting - shown as hexadecimal number, e.g.: 000000081276Ad00 void ContinueString_Pointer(const void* ptr); + // Ends writing a string value by writing '"'. void EndString(const char* pStr = VMA_NULL); + // Writes a number value. void WriteNumber(uint32_t n); void WriteNumber(uint64_t n); + // Writes a boolean value - false or true. void WriteBool(bool b); + // Writes a null value. void WriteNull(); private: - static const char* const INDENT; - enum COLLECTION_TYPE { COLLECTION_TYPE_OBJECT, @@ -8697,6 +5262,8 @@ private: bool singleLineMode; }; + static const char* const INDENT; + VmaStringBuilder& m_SB; VmaVector< StackItem, VmaStlAllocator<StackItem> > m_Stack; bool m_InsideString; @@ -8704,15 +5271,13 @@ private: void BeginValue(bool isString); void WriteIndent(bool oneLess = false); }; - const char* const VmaJsonWriter::INDENT = " "; -VmaJsonWriter::VmaJsonWriter(const VkAllocationCallbacks* pAllocationCallbacks, VmaStringBuilder& sb) : - m_SB(sb), +#ifndef _VMA_JSON_WRITER_FUNCTIONS +VmaJsonWriter::VmaJsonWriter(const VkAllocationCallbacks* pAllocationCallbacks, VmaStringBuilder& sb) + : m_SB(sb), m_Stack(VmaStlAllocator<StackItem>(pAllocationCallbacks)), - m_InsideString(false) -{ -} + m_InsideString(false) {} VmaJsonWriter::~VmaJsonWriter() { @@ -8783,7 +5348,7 @@ void VmaJsonWriter::BeginString(const char* pStr) BeginValue(true); m_SB.Add('"'); m_InsideString = true; - if(pStr != VMA_NULL && pStr[0] != '\0') + if (pStr != VMA_NULL && pStr[0] != '\0') { ContinueString(pStr); } @@ -8794,22 +5359,22 @@ void VmaJsonWriter::ContinueString(const char* pStr) VMA_ASSERT(m_InsideString); const size_t strLen = strlen(pStr); - for(size_t i = 0; i < strLen; ++i) + for (size_t i = 0; i < strLen; ++i) { char ch = pStr[i]; - if(ch == '\\') + if (ch == '\\') { m_SB.Add("\\\\"); } - else if(ch == '"') + else if (ch == '"') { m_SB.Add("\\\""); } - else if(ch >= 32) + else if (ch >= 32) { m_SB.Add(ch); } - else switch(ch) + else switch (ch) { case '\b': m_SB.Add("\\b"); @@ -8854,7 +5419,7 @@ void VmaJsonWriter::ContinueString_Pointer(const void* ptr) void VmaJsonWriter::EndString(const char* pStr) { VMA_ASSERT(m_InsideString); - if(pStr != VMA_NULL && pStr[0] != '\0') + if (pStr != VMA_NULL && pStr[0] != '\0') { ContinueString(pStr); } @@ -8892,21 +5457,21 @@ void VmaJsonWriter::WriteNull() void VmaJsonWriter::BeginValue(bool isString) { - if(!m_Stack.empty()) + if (!m_Stack.empty()) { StackItem& currItem = m_Stack.back(); - if(currItem.type == COLLECTION_TYPE_OBJECT && + if (currItem.type == COLLECTION_TYPE_OBJECT && currItem.valueCount % 2 == 0) { VMA_ASSERT(isString); } - if(currItem.type == COLLECTION_TYPE_OBJECT && + if (currItem.type == COLLECTION_TYPE_OBJECT && currItem.valueCount % 2 != 0) { m_SB.Add(": "); } - else if(currItem.valueCount > 0) + else if (currItem.valueCount > 0) { m_SB.Add(", "); WriteIndent(); @@ -8921,399 +5486,624 @@ void VmaJsonWriter::BeginValue(bool isString) void VmaJsonWriter::WriteIndent(bool oneLess) { - if(!m_Stack.empty() && !m_Stack.back().singleLineMode) + if (!m_Stack.empty() && !m_Stack.back().singleLineMode) { m_SB.AddNewLine(); size_t count = m_Stack.size(); - if(count > 0 && oneLess) + if (count > 0 && oneLess) { --count; } - for(size_t i = 0; i < count; ++i) + for (size_t i = 0; i < count; ++i) { m_SB.Add(INDENT); } } } +#endif // _VMA_JSON_WRITER_FUNCTIONS -#endif // #if VMA_STATS_STRING_ENABLED +static void VmaPrintStatInfo(VmaJsonWriter& json, const VmaStatInfo& stat) +{ + json.BeginObject(); -//////////////////////////////////////////////////////////////////////////////// + json.WriteString("Blocks"); + json.WriteNumber(stat.blockCount); -void VmaAllocation_T::SetUserData(VmaAllocator hAllocator, void* pUserData) -{ - if(IsUserDataString()) - { - VMA_ASSERT(pUserData == VMA_NULL || pUserData != m_pUserData); + json.WriteString("Allocations"); + json.WriteNumber(stat.allocationCount); - FreeUserDataString(hAllocator); + json.WriteString("UnusedRanges"); + json.WriteNumber(stat.unusedRangeCount); - if(pUserData != VMA_NULL) - { - m_pUserData = VmaCreateStringCopy(hAllocator->GetAllocationCallbacks(), (const char*)pUserData); - } - } - else + json.WriteString("UsedBytes"); + json.WriteNumber(stat.usedBytes); + + json.WriteString("UnusedBytes"); + json.WriteNumber(stat.unusedBytes); + + if (stat.allocationCount > 1) { - m_pUserData = pUserData; + json.WriteString("AllocationSize"); + json.BeginObject(true); + json.WriteString("Min"); + json.WriteNumber(stat.allocationSizeMin); + json.WriteString("Avg"); + json.WriteNumber(stat.allocationSizeAvg); + json.WriteString("Max"); + json.WriteNumber(stat.allocationSizeMax); + json.EndObject(); } -} -void VmaAllocation_T::ChangeBlockAllocation( - VmaAllocator hAllocator, - VmaDeviceMemoryBlock* block, - VkDeviceSize offset) -{ - VMA_ASSERT(block != VMA_NULL); - VMA_ASSERT(m_Type == ALLOCATION_TYPE_BLOCK); - - // Move mapping reference counter from old block to new block. - if(block != m_BlockAllocation.m_Block) + if (stat.unusedRangeCount > 1) { - uint32_t mapRefCount = m_MapCount & ~MAP_COUNT_FLAG_PERSISTENT_MAP; - if(IsPersistentMap()) - ++mapRefCount; - m_BlockAllocation.m_Block->Unmap(hAllocator, mapRefCount); - block->Map(hAllocator, mapRefCount, VMA_NULL); + json.WriteString("UnusedRangeSize"); + json.BeginObject(true); + json.WriteString("Min"); + json.WriteNumber(stat.unusedRangeSizeMin); + json.WriteString("Avg"); + json.WriteNumber(stat.unusedRangeSizeAvg); + json.WriteString("Max"); + json.WriteNumber(stat.unusedRangeSizeMax); + json.EndObject(); } - m_BlockAllocation.m_Block = block; - m_BlockAllocation.m_Offset = offset; + json.EndObject(); } +#endif // _VMA_JSON_WRITER -void VmaAllocation_T::ChangeOffset(VkDeviceSize newOffset) -{ - VMA_ASSERT(m_Type == ALLOCATION_TYPE_BLOCK); - m_BlockAllocation.m_Offset = newOffset; -} +#ifndef _VMA_DEVICE_MEMORY_BLOCK +/* +Represents a single block of device memory (`VkDeviceMemory`) with all the +data about its regions (aka suballocations, #VmaAllocation), assigned and free. -VkDeviceSize VmaAllocation_T::GetOffset() const +Thread-safety: This class must be externally synchronized. +*/ +class VmaDeviceMemoryBlock { - switch(m_Type) - { - case ALLOCATION_TYPE_BLOCK: - return m_BlockAllocation.m_Offset; - case ALLOCATION_TYPE_DEDICATED: - return 0; - default: - VMA_ASSERT(0); - return 0; - } -} + VMA_CLASS_NO_COPY(VmaDeviceMemoryBlock) +public: + VmaBlockMetadata* m_pMetadata; -VkDeviceMemory VmaAllocation_T::GetMemory() const -{ - switch(m_Type) - { - case ALLOCATION_TYPE_BLOCK: - return m_BlockAllocation.m_Block->GetDeviceMemory(); - case ALLOCATION_TYPE_DEDICATED: - return m_DedicatedAllocation.m_hMemory; - default: - VMA_ASSERT(0); - return VK_NULL_HANDLE; - } -} + VmaDeviceMemoryBlock(VmaAllocator hAllocator); + ~VmaDeviceMemoryBlock(); -void* VmaAllocation_T::GetMappedData() const -{ - switch(m_Type) - { - case ALLOCATION_TYPE_BLOCK: - if(m_MapCount != 0) - { - void* pBlockData = m_BlockAllocation.m_Block->GetMappedData(); - VMA_ASSERT(pBlockData != VMA_NULL); - return (char*)pBlockData + m_BlockAllocation.m_Offset; - } - else - { - return VMA_NULL; - } - break; - case ALLOCATION_TYPE_DEDICATED: - VMA_ASSERT((m_DedicatedAllocation.m_pMappedData != VMA_NULL) == (m_MapCount != 0)); - return m_DedicatedAllocation.m_pMappedData; - default: - VMA_ASSERT(0); - return VMA_NULL; - } -} + // Always call after construction. + void Init( + VmaAllocator hAllocator, + VmaPool hParentPool, + uint32_t newMemoryTypeIndex, + VkDeviceMemory newMemory, + VkDeviceSize newSize, + uint32_t id, + uint32_t algorithm, + VkDeviceSize bufferImageGranularity); + // Always call before destruction. + void Destroy(VmaAllocator allocator); -bool VmaAllocation_T::CanBecomeLost() const -{ - switch(m_Type) - { - case ALLOCATION_TYPE_BLOCK: - return m_BlockAllocation.m_CanBecomeLost; - case ALLOCATION_TYPE_DEDICATED: - return false; - default: - VMA_ASSERT(0); - return false; - } -} + VmaPool GetParentPool() const { return m_hParentPool; } + VkDeviceMemory GetDeviceMemory() const { return m_hMemory; } + uint32_t GetMemoryTypeIndex() const { return m_MemoryTypeIndex; } + uint32_t GetId() const { return m_Id; } + void* GetMappedData() const { return m_pMappedData; } -bool VmaAllocation_T::MakeLost(uint32_t currentFrameIndex, uint32_t frameInUseCount) -{ - VMA_ASSERT(CanBecomeLost()); + // Validates all data structures inside this object. If not valid, returns false. + bool Validate() const; + VkResult CheckCorruption(VmaAllocator hAllocator); - /* - Warning: This is a carefully designed algorithm. - Do not modify unless you really know what you're doing :) - */ - uint32_t localLastUseFrameIndex = GetLastUseFrameIndex(); - for(;;) - { - if(localLastUseFrameIndex == VMA_FRAME_INDEX_LOST) - { - VMA_ASSERT(0); - return false; - } - else if(localLastUseFrameIndex + frameInUseCount >= currentFrameIndex) - { - return false; - } - else // Last use time earlier than current time. - { - if(CompareExchangeLastUseFrameIndex(localLastUseFrameIndex, VMA_FRAME_INDEX_LOST)) - { - // Setting hAllocation.LastUseFrameIndex atomic to VMA_FRAME_INDEX_LOST is enough to mark it as LOST. - // Calling code just needs to unregister this allocation in owning VmaDeviceMemoryBlock. - return true; - } - } - } -} + // ppData can be null. + VkResult Map(VmaAllocator hAllocator, uint32_t count, void** ppData); + void Unmap(VmaAllocator hAllocator, uint32_t count); -#if VMA_STATS_STRING_ENABLED + VkResult WriteMagicValueAfterAllocation(VmaAllocator hAllocator, VkDeviceSize allocOffset, VkDeviceSize allocSize); + VkResult ValidateMagicValueAfterAllocation(VmaAllocator hAllocator, VkDeviceSize allocOffset, VkDeviceSize allocSize); -// Correspond to values of enum VmaSuballocationType. -static const char* VMA_SUBALLOCATION_TYPE_NAMES[] = { - "FREE", - "UNKNOWN", - "BUFFER", - "IMAGE_UNKNOWN", - "IMAGE_LINEAR", - "IMAGE_OPTIMAL", + VkResult BindBufferMemory( + const VmaAllocator hAllocator, + const VmaAllocation hAllocation, + VkDeviceSize allocationLocalOffset, + VkBuffer hBuffer, + const void* pNext); + VkResult BindImageMemory( + const VmaAllocator hAllocator, + const VmaAllocation hAllocation, + VkDeviceSize allocationLocalOffset, + VkImage hImage, + const void* pNext); + +private: + VmaPool m_hParentPool; // VK_NULL_HANDLE if not belongs to custom pool. + uint32_t m_MemoryTypeIndex; + uint32_t m_Id; + VkDeviceMemory m_hMemory; + + /* + Protects access to m_hMemory so it is not used by multiple threads simultaneously, e.g. vkMapMemory, vkBindBufferMemory. + Also protects m_MapCount, m_pMappedData. + Allocations, deallocations, any change in m_pMetadata is protected by parent's VmaBlockVector::m_Mutex. + */ + VMA_MUTEX m_Mutex; + uint32_t m_MapCount; + void* m_pMappedData; }; +#endif // _VMA_DEVICE_MEMORY_BLOCK -void VmaAllocation_T::PrintParameters(class VmaJsonWriter& json) const +#ifndef _VMA_ALLOCATION_T +struct VmaAllocation_T { - json.WriteString("Type"); - json.WriteString(VMA_SUBALLOCATION_TYPE_NAMES[m_SuballocationType]); + friend struct VmaDedicatedAllocationListItemTraits; - json.WriteString("Size"); - json.WriteNumber(m_Size); + static const uint8_t MAP_COUNT_FLAG_PERSISTENT_MAP = 0x80; - if(m_pUserData != VMA_NULL) + enum FLAGS { FLAG_USER_DATA_STRING = 0x01 }; + +public: + enum ALLOCATION_TYPE { - json.WriteString("UserData"); - if(IsUserDataString()) - { - json.WriteString((const char*)m_pUserData); - } - else - { - json.BeginString(); - json.ContinueString_Pointer(m_pUserData); - json.EndString(); - } - } + ALLOCATION_TYPE_NONE, + ALLOCATION_TYPE_BLOCK, + ALLOCATION_TYPE_DEDICATED, + }; + + // This struct is allocated using VmaPoolAllocator. + VmaAllocation_T(bool userDataString); + ~VmaAllocation_T(); + + void InitBlockAllocation( + VmaDeviceMemoryBlock* block, + VmaAllocHandle allocHandle, + VkDeviceSize alignment, + VkDeviceSize size, + uint32_t memoryTypeIndex, + VmaSuballocationType suballocationType, + bool mapped); + // pMappedData not null means allocation is created with MAPPED flag. + void InitDedicatedAllocation( + VmaPool hParentPool, + uint32_t memoryTypeIndex, + VkDeviceMemory hMemory, + VmaSuballocationType suballocationType, + void* pMappedData, + VkDeviceSize size); + + ALLOCATION_TYPE GetType() const { return (ALLOCATION_TYPE)m_Type; } + VkDeviceSize GetAlignment() const { return m_Alignment; } + VkDeviceSize GetSize() const { return m_Size; } + bool IsUserDataString() const { return (m_Flags & FLAG_USER_DATA_STRING) != 0; } + void* GetUserData() const { return m_pUserData; } + VmaSuballocationType GetSuballocationType() const { return (VmaSuballocationType)m_SuballocationType; } + + VmaDeviceMemoryBlock* GetBlock() const { VMA_ASSERT(m_Type == ALLOCATION_TYPE_BLOCK); return m_BlockAllocation.m_Block; } + uint32_t GetMemoryTypeIndex() const { return m_MemoryTypeIndex; } + bool IsPersistentMap() const { return (m_MapCount & MAP_COUNT_FLAG_PERSISTENT_MAP) != 0; } + + void SetUserData(VmaAllocator hAllocator, void* pUserData); + void ChangeBlockAllocation(VmaAllocator hAllocator, VmaDeviceMemoryBlock* block, VmaAllocHandle allocHandle); + void ChangeAllocHandle(VmaAllocHandle newAllocHandle); + VmaAllocHandle GetAllocHandle() const; + VkDeviceSize GetOffset() const; + VmaPool GetParentPool() const; + VkDeviceMemory GetMemory() const; + void* GetMappedData() const; + + void DedicatedAllocCalcStatsInfo(VmaStatInfo& outInfo); + + void BlockAllocMap(); + void BlockAllocUnmap(); + VkResult DedicatedAllocMap(VmaAllocator hAllocator, void** ppData); + void DedicatedAllocUnmap(VmaAllocator hAllocator); - json.WriteString("CreationFrameIndex"); - json.WriteNumber(m_CreationFrameIndex); +#if VMA_STATS_STRING_ENABLED + uint32_t GetBufferImageUsage() const { return m_BufferImageUsage; } - json.WriteString("LastUseFrameIndex"); - json.WriteNumber(GetLastUseFrameIndex()); + void InitBufferImageUsage(uint32_t bufferImageUsage); + void PrintParameters(class VmaJsonWriter& json) const; +#endif - if(m_BufferImageUsage != 0) +private: + // Allocation out of VmaDeviceMemoryBlock. + struct BlockAllocation { - json.WriteString("Usage"); - json.WriteNumber(m_BufferImageUsage); - } -} + VmaDeviceMemoryBlock* m_Block; + VmaAllocHandle m_AllocHandle; + }; + // Allocation for an object that has its own private VkDeviceMemory. + struct DedicatedAllocation + { + VmaPool m_hParentPool; // VK_NULL_HANDLE if not belongs to custom pool. + VkDeviceMemory m_hMemory; + void* m_pMappedData; // Not null means memory is mapped. + VmaAllocation_T* m_Prev; + VmaAllocation_T* m_Next; + }; + union + { + // Allocation out of VmaDeviceMemoryBlock. + BlockAllocation m_BlockAllocation; + // Allocation for an object that has its own private VkDeviceMemory. + DedicatedAllocation m_DedicatedAllocation; + }; + VkDeviceSize m_Alignment; + VkDeviceSize m_Size; + void* m_pUserData; + uint32_t m_MemoryTypeIndex; + uint8_t m_Type; // ALLOCATION_TYPE + uint8_t m_SuballocationType; // VmaSuballocationType + // Bit 0x80 is set when allocation was created with VMA_ALLOCATION_CREATE_MAPPED_BIT. + // Bits with mask 0x7F are reference counter for vmaMapMemory()/vmaUnmapMemory(). + uint8_t m_MapCount; + uint8_t m_Flags; // enum FLAGS +#if VMA_STATS_STRING_ENABLED + uint32_t m_BufferImageUsage; // 0 if unknown. #endif -void VmaAllocation_T::FreeUserDataString(VmaAllocator hAllocator) -{ - VMA_ASSERT(IsUserDataString()); - VmaFreeString(hAllocator->GetAllocationCallbacks(), (char*)m_pUserData); - m_pUserData = VMA_NULL; -} + void FreeUserDataString(VmaAllocator hAllocator); +}; +#endif // _VMA_ALLOCATION_T -void VmaAllocation_T::BlockAllocMap() +#ifndef _VMA_DEDICATED_ALLOCATION_LIST_ITEM_TRAITS +struct VmaDedicatedAllocationListItemTraits { - VMA_ASSERT(GetType() == ALLOCATION_TYPE_BLOCK); + typedef VmaAllocation_T ItemType; - if((m_MapCount & ~MAP_COUNT_FLAG_PERSISTENT_MAP) < 0x7F) + static ItemType* GetPrev(const ItemType* item) { - ++m_MapCount; + VMA_HEAVY_ASSERT(item->GetType() == VmaAllocation_T::ALLOCATION_TYPE_DEDICATED); + return item->m_DedicatedAllocation.m_Prev; } - else + static ItemType* GetNext(const ItemType* item) { - VMA_ASSERT(0 && "Allocation mapped too many times simultaneously."); + VMA_HEAVY_ASSERT(item->GetType() == VmaAllocation_T::ALLOCATION_TYPE_DEDICATED); + return item->m_DedicatedAllocation.m_Next; } -} - -void VmaAllocation_T::BlockAllocUnmap() -{ - VMA_ASSERT(GetType() == ALLOCATION_TYPE_BLOCK); - - if((m_MapCount & ~MAP_COUNT_FLAG_PERSISTENT_MAP) != 0) + static ItemType*& AccessPrev(ItemType* item) { - --m_MapCount; + VMA_HEAVY_ASSERT(item->GetType() == VmaAllocation_T::ALLOCATION_TYPE_DEDICATED); + return item->m_DedicatedAllocation.m_Prev; } - else + static ItemType*& AccessNext(ItemType* item) { - VMA_ASSERT(0 && "Unmapping allocation not previously mapped."); + VMA_HEAVY_ASSERT(item->GetType() == VmaAllocation_T::ALLOCATION_TYPE_DEDICATED); + return item->m_DedicatedAllocation.m_Next; } -} +}; +#endif // _VMA_DEDICATED_ALLOCATION_LIST_ITEM_TRAITS -VkResult VmaAllocation_T::DedicatedAllocMap(VmaAllocator hAllocator, void** ppData) +#ifndef _VMA_DEDICATED_ALLOCATION_LIST +/* +Stores linked list of VmaAllocation_T objects. +Thread-safe, synchronized internally. +*/ +class VmaDedicatedAllocationList { - VMA_ASSERT(GetType() == ALLOCATION_TYPE_DEDICATED); +public: + VmaDedicatedAllocationList() {} + ~VmaDedicatedAllocationList(); - if(m_MapCount != 0) + void Init(bool useMutex) { m_UseMutex = useMutex; } + bool Validate(); + + void AddStats(VmaStats* stats, uint32_t memTypeIndex, uint32_t memHeapIndex); + void AddPoolStats(VmaPoolStats* stats); +#if VMA_STATS_STRING_ENABLED + // Writes JSON array with the list of allocations. + void BuildStatsString(VmaJsonWriter& json); +#endif + + bool IsEmpty(); + void Register(VmaAllocation alloc); + void Unregister(VmaAllocation alloc); + +private: + typedef VmaIntrusiveLinkedList<VmaDedicatedAllocationListItemTraits> DedicatedAllocationLinkedList; + + bool m_UseMutex = true; + VMA_RW_MUTEX m_Mutex; + DedicatedAllocationLinkedList m_AllocationList; +}; + +#ifndef _VMA_DEDICATED_ALLOCATION_LIST_FUNCTIONS + +VmaDedicatedAllocationList::~VmaDedicatedAllocationList() +{ + VMA_HEAVY_ASSERT(Validate()); + + if (!m_AllocationList.IsEmpty()) { - if((m_MapCount & ~MAP_COUNT_FLAG_PERSISTENT_MAP) < 0x7F) - { - VMA_ASSERT(m_DedicatedAllocation.m_pMappedData != VMA_NULL); - *ppData = m_DedicatedAllocation.m_pMappedData; - ++m_MapCount; - return VK_SUCCESS; - } - else - { - VMA_ASSERT(0 && "Dedicated allocation mapped too many times simultaneously."); - return VK_ERROR_MEMORY_MAP_FAILED; - } + VMA_ASSERT(false && "Unfreed dedicated allocations found!"); } - else +} + +bool VmaDedicatedAllocationList::Validate() +{ + const size_t declaredCount = m_AllocationList.GetCount(); + size_t actualCount = 0; + VmaMutexLockRead lock(m_Mutex, m_UseMutex); + for (VmaAllocation alloc = m_AllocationList.Front(); + alloc != VMA_NULL; alloc = m_AllocationList.GetNext(alloc)) { - VkResult result = (*hAllocator->GetVulkanFunctions().vkMapMemory)( - hAllocator->m_hDevice, - m_DedicatedAllocation.m_hMemory, - 0, // offset - VK_WHOLE_SIZE, - 0, // flags - ppData); - if(result == VK_SUCCESS) - { - m_DedicatedAllocation.m_pMappedData = *ppData; - m_MapCount = 1; - } - return result; + ++actualCount; } + VMA_VALIDATE(actualCount == declaredCount); + + return true; } -void VmaAllocation_T::DedicatedAllocUnmap(VmaAllocator hAllocator) +void VmaDedicatedAllocationList::AddStats(VmaStats* stats, uint32_t memTypeIndex, uint32_t memHeapIndex) { - VMA_ASSERT(GetType() == ALLOCATION_TYPE_DEDICATED); - - if((m_MapCount & ~MAP_COUNT_FLAG_PERSISTENT_MAP) != 0) + VmaMutexLockRead lock(m_Mutex, m_UseMutex); + for (VmaAllocation alloc = m_AllocationList.Front(); + alloc != VMA_NULL; alloc = m_AllocationList.GetNext(alloc)) { - --m_MapCount; - if(m_MapCount == 0) - { - m_DedicatedAllocation.m_pMappedData = VMA_NULL; - (*hAllocator->GetVulkanFunctions().vkUnmapMemory)( - hAllocator->m_hDevice, - m_DedicatedAllocation.m_hMemory); - } + VmaStatInfo allocationStatInfo; + alloc->DedicatedAllocCalcStatsInfo(allocationStatInfo); + VmaAddStatInfo(stats->total, allocationStatInfo); + VmaAddStatInfo(stats->memoryType[memTypeIndex], allocationStatInfo); + VmaAddStatInfo(stats->memoryHeap[memHeapIndex], allocationStatInfo); } - else +} + +void VmaDedicatedAllocationList::AddPoolStats(VmaPoolStats* stats) +{ + VmaMutexLockRead lock(m_Mutex, m_UseMutex); + + const size_t allocCount = m_AllocationList.GetCount(); + stats->allocationCount += allocCount; + stats->blockCount += allocCount; + + for(auto* item = m_AllocationList.Front(); item != nullptr; item = DedicatedAllocationLinkedList::GetNext(item)) { - VMA_ASSERT(0 && "Unmapping dedicated allocation not previously mapped."); + stats->size += item->GetSize(); } } #if VMA_STATS_STRING_ENABLED - -static void VmaPrintStatInfo(VmaJsonWriter& json, const VmaStatInfo& stat) +void VmaDedicatedAllocationList::BuildStatsString(VmaJsonWriter& json) { - json.BeginObject(); - - json.WriteString("Blocks"); - json.WriteNumber(stat.blockCount); + VmaMutexLockRead lock(m_Mutex, m_UseMutex); + json.BeginArray(); + for (VmaAllocation alloc = m_AllocationList.Front(); + alloc != VMA_NULL; alloc = m_AllocationList.GetNext(alloc)) + { + json.BeginObject(true); + alloc->PrintParameters(json); + json.EndObject(); + } + json.EndArray(); +} +#endif // VMA_STATS_STRING_ENABLED - json.WriteString("Allocations"); - json.WriteNumber(stat.allocationCount); +bool VmaDedicatedAllocationList::IsEmpty() +{ + VmaMutexLockRead lock(m_Mutex, m_UseMutex); + return m_AllocationList.IsEmpty(); +} - json.WriteString("UnusedRanges"); - json.WriteNumber(stat.unusedRangeCount); +void VmaDedicatedAllocationList::Register(VmaAllocation alloc) +{ + VmaMutexLockWrite lock(m_Mutex, m_UseMutex); + m_AllocationList.PushBack(alloc); +} - json.WriteString("UsedBytes"); - json.WriteNumber(stat.usedBytes); +void VmaDedicatedAllocationList::Unregister(VmaAllocation alloc) +{ + VmaMutexLockWrite lock(m_Mutex, m_UseMutex); + m_AllocationList.Remove(alloc); +} +#endif // _VMA_DEDICATED_ALLOCATION_LIST_FUNCTIONS +#endif // _VMA_DEDICATED_ALLOCATION_LIST - json.WriteString("UnusedBytes"); - json.WriteNumber(stat.unusedBytes); +#ifndef _VMA_SUBALLOCATION +/* +Represents a region of VmaDeviceMemoryBlock that is either assigned and returned as +allocated memory block or free. +*/ +struct VmaSuballocation +{ + VkDeviceSize offset; + VkDeviceSize size; + void* userData; + VmaSuballocationType type; +}; - if(stat.allocationCount > 1) +// Comparator for offsets. +struct VmaSuballocationOffsetLess +{ + bool operator()(const VmaSuballocation& lhs, const VmaSuballocation& rhs) const { - json.WriteString("AllocationSize"); - json.BeginObject(true); - json.WriteString("Min"); - json.WriteNumber(stat.allocationSizeMin); - json.WriteString("Avg"); - json.WriteNumber(stat.allocationSizeAvg); - json.WriteString("Max"); - json.WriteNumber(stat.allocationSizeMax); - json.EndObject(); + return lhs.offset < rhs.offset; } +}; - if(stat.unusedRangeCount > 1) +struct VmaSuballocationOffsetGreater +{ + bool operator()(const VmaSuballocation& lhs, const VmaSuballocation& rhs) const { - json.WriteString("UnusedRangeSize"); - json.BeginObject(true); - json.WriteString("Min"); - json.WriteNumber(stat.unusedRangeSizeMin); - json.WriteString("Avg"); - json.WriteNumber(stat.unusedRangeSizeAvg); - json.WriteString("Max"); - json.WriteNumber(stat.unusedRangeSizeMax); - json.EndObject(); + return lhs.offset > rhs.offset; } - - json.EndObject(); -} - -#endif // #if VMA_STATS_STRING_ENABLED +}; struct VmaSuballocationItemSizeLess { - bool operator()( - const VmaSuballocationList::iterator lhs, + bool operator()(const VmaSuballocationList::iterator lhs, const VmaSuballocationList::iterator rhs) const { return lhs->size < rhs->size; } - bool operator()( - const VmaSuballocationList::iterator lhs, + + bool operator()(const VmaSuballocationList::iterator lhs, VkDeviceSize rhsSize) const { return lhs->size < rhsSize; } }; +#endif // _VMA_SUBALLOCATION +#ifndef _VMA_ALLOCATION_REQUEST +/* +Parameters of planned allocation inside a VmaDeviceMemoryBlock. +item points to a FREE suballocation. +*/ +struct VmaAllocationRequest +{ + VmaAllocHandle allocHandle; + VkDeviceSize size; + VmaSuballocationList::iterator item; + void* customData; + uint64_t algorithmData; + VmaAllocationRequestType type; +}; +#endif // _VMA_ALLOCATION_REQUEST -//////////////////////////////////////////////////////////////////////////////// -// class VmaBlockMetadata +#ifndef _VMA_BLOCK_METADATA +/* +Data structure used for bookkeeping of allocations and unused ranges of memory +in a single VkDeviceMemory block. +*/ +class VmaBlockMetadata +{ +public: + // pAllocationCallbacks, if not null, must be owned externally - alive and unchanged for the whole lifetime of this object. + VmaBlockMetadata(const VkAllocationCallbacks* pAllocationCallbacks, + VkDeviceSize bufferImageGranularity, bool isVirtual); + virtual ~VmaBlockMetadata() = default; + + virtual void Init(VkDeviceSize size) { m_Size = size; } + bool IsVirtual() const { return m_IsVirtual; } + VkDeviceSize GetSize() const { return m_Size; } + + // Validates all data structures inside this object. If not valid, returns false. + virtual bool Validate() const = 0; + virtual size_t GetAllocationCount() const = 0; + virtual VkDeviceSize GetSumFreeSize() const = 0; + // Returns true if this block is empty - contains only single free suballocation. + virtual bool IsEmpty() const = 0; + virtual void GetAllocationInfo(VmaAllocHandle allocHandle, VmaVirtualAllocationInfo& outInfo) = 0; + virtual VkDeviceSize GetAllocationOffset(VmaAllocHandle allocHandle) const = 0; + + // Must set blockCount to 1. + virtual void CalcAllocationStatInfo(VmaStatInfo& outInfo) const = 0; + // Shouldn't modify blockCount. + virtual void AddPoolStats(VmaPoolStats& inoutStats) const = 0; + +#if VMA_STATS_STRING_ENABLED + virtual void PrintDetailedMap(class VmaJsonWriter& json) const = 0; +#endif -VmaBlockMetadata::VmaBlockMetadata(VmaAllocator hAllocator) : - m_Size(0), - m_pAllocationCallbacks(hAllocator->GetAllocationCallbacks()) + // Tries to find a place for suballocation with given parameters inside this block. + // If succeeded, fills pAllocationRequest and returns true. + // If failed, returns false. + virtual bool CreateAllocationRequest( + VkDeviceSize allocSize, + VkDeviceSize allocAlignment, + bool upperAddress, + VmaSuballocationType allocType, + // Always one of VMA_ALLOCATION_CREATE_STRATEGY_* or VMA_ALLOCATION_INTERNAL_STRATEGY_* flags. + uint32_t strategy, + VmaAllocationRequest* pAllocationRequest) = 0; + + virtual VkResult CheckCorruption(const void* pBlockData) = 0; + + // Makes actual allocation based on request. Request must already be checked and valid. + virtual void Alloc( + const VmaAllocationRequest& request, + VmaSuballocationType type, + void* userData) = 0; + + // Frees suballocation assigned to given memory region. + virtual void Free(VmaAllocHandle allocHandle) = 0; + + // Frees all allocations. + // Careful! Don't call it if there are VmaAllocation objects owned by userData of cleared allocations! + virtual void Clear() = 0; + + virtual void SetAllocationUserData(VmaAllocHandle allocHandle, void* userData) = 0; + virtual void DebugLogAllAllocations() const = 0; + +protected: + const VkAllocationCallbacks* GetAllocationCallbacks() const { return m_pAllocationCallbacks; } + VkDeviceSize GetBufferImageGranularity() const { return m_BufferImageGranularity; } + VkDeviceSize GetDebugMargin() const { return IsVirtual() ? 0 : VMA_DEBUG_MARGIN; } + + void DebugLogAllocation(VkDeviceSize offset, VkDeviceSize size, void* userData) const; +#if VMA_STATS_STRING_ENABLED + void PrintDetailedMap_Begin(class VmaJsonWriter& json, + VkDeviceSize unusedBytes, + size_t allocationCount, + size_t unusedRangeCount) const; + void PrintDetailedMap_Allocation(class VmaJsonWriter& json, + VkDeviceSize offset, VkDeviceSize size, void* userData) const; + void PrintDetailedMap_UnusedRange(class VmaJsonWriter& json, + VkDeviceSize offset, + VkDeviceSize size) const; + void PrintDetailedMap_End(class VmaJsonWriter& json) const; +#endif + +private: + VkDeviceSize m_Size; + const VkAllocationCallbacks* m_pAllocationCallbacks; + const VkDeviceSize m_BufferImageGranularity; + const bool m_IsVirtual; +}; + +#ifndef _VMA_BLOCK_METADATA_FUNCTIONS +VmaBlockMetadata::VmaBlockMetadata(const VkAllocationCallbacks* pAllocationCallbacks, + VkDeviceSize bufferImageGranularity, bool isVirtual) + : m_Size(0), + m_pAllocationCallbacks(pAllocationCallbacks), + m_BufferImageGranularity(bufferImageGranularity), + m_IsVirtual(isVirtual) {} + +void VmaBlockMetadata::DebugLogAllocation(VkDeviceSize offset, VkDeviceSize size, void* userData) const { -} + if (IsVirtual()) + { + VMA_DEBUG_LOG("UNFREED VIRTUAL ALLOCATION; Offset: %llu; Size: %llu; UserData: %p", offset, size, userData); + } + else + { + VMA_ASSERT(userData != VMA_NULL); + VmaAllocation allocation = reinterpret_cast<VmaAllocation>(userData); + + userData = allocation->GetUserData(); #if VMA_STATS_STRING_ENABLED + if (userData != VMA_NULL && allocation->IsUserDataString()) + { + VMA_DEBUG_LOG("UNFREED ALLOCATION; Offset: %llu; Size: %llu; UserData: %s; Type: %s; Usage: %u", + offset, size, reinterpret_cast<const char*>(userData), + VMA_SUBALLOCATION_TYPE_NAMES[allocation->GetSuballocationType()], + allocation->GetBufferImageUsage()); + } + else + { + VMA_DEBUG_LOG("UNFREED ALLOCATION; Offset: %llu; Size: %llu; UserData: %p; Type: %s; Usage: %u", + offset, size, userData, + VMA_SUBALLOCATION_TYPE_NAMES[allocation->GetSuballocationType()], + allocation->GetBufferImageUsage()); + } +#else + if (userData != VMA_NULL && allocation->IsUserDataString()) + { + VMA_DEBUG_LOG("UNFREED ALLOCATION; Offset: %llu; Size: %llu; UserData: %s; Type: %u", + offset, size, reinterpret_cast<const char*>(userData), + (uint32_t)allocation->GetSuballocationType()); + } + else + { + VMA_DEBUG_LOG("UNFREED ALLOCATION; Offset: %llu; Size: %llu; UserData: %p; Type: %u", + offset, size, userData, + (uint32_t)allocation->GetSuballocationType()); + } +#endif // VMA_STATS_STRING_ENABLED + } + +} +#if VMA_STATS_STRING_ENABLED void VmaBlockMetadata::PrintDetailedMap_Begin(class VmaJsonWriter& json, - VkDeviceSize unusedBytes, - size_t allocationCount, - size_t unusedRangeCount) const + VkDeviceSize unusedBytes, size_t allocationCount, size_t unusedRangeCount) const { json.BeginObject(); @@ -9334,22 +6124,39 @@ void VmaBlockMetadata::PrintDetailedMap_Begin(class VmaJsonWriter& json, } void VmaBlockMetadata::PrintDetailedMap_Allocation(class VmaJsonWriter& json, - VkDeviceSize offset, - VmaAllocation hAllocation) const + VkDeviceSize offset, VkDeviceSize size, void* userData) const { json.BeginObject(true); json.WriteString("Offset"); json.WriteNumber(offset); - hAllocation->PrintParameters(json); + if (IsVirtual()) + { + json.WriteString("Type"); + json.WriteString("VirtualAllocation"); + + json.WriteString("Size"); + json.WriteNumber(size); + + if (userData != VMA_NULL) + { + json.WriteString("UserData"); + json.BeginString(); + json.ContinueString_Pointer(userData); + json.EndString(); + } + } + else + { + ((VmaAllocation)userData)->PrintParameters(json); + } json.EndObject(); } void VmaBlockMetadata::PrintDetailedMap_UnusedRange(class VmaJsonWriter& json, - VkDeviceSize offset, - VkDeviceSize size) const + VkDeviceSize offset, VkDeviceSize size) const { json.BeginObject(true); @@ -9370,25 +6177,341 @@ void VmaBlockMetadata::PrintDetailedMap_End(class VmaJsonWriter& json) const json.EndArray(); json.EndObject(); } +#endif // VMA_STATS_STRING_ENABLED +#endif // _VMA_BLOCK_METADATA_FUNCTIONS +#endif // _VMA_BLOCK_METADATA + +#ifndef _VMA_BLOCK_BUFFER_IMAGE_GRANULARITY +// Before deleting object of this class remember to call 'Destroy()' +class VmaBlockBufferImageGranularity final +{ +public: + struct ValidationContext + { + const VkAllocationCallbacks* allocCallbacks; + uint16_t* pageAllocs; + }; -#endif // #if VMA_STATS_STRING_ENABLED + VmaBlockBufferImageGranularity(VkDeviceSize bufferImageGranularity); + ~VmaBlockBufferImageGranularity(); -//////////////////////////////////////////////////////////////////////////////// -// class VmaBlockMetadata_Generic + bool IsEnabled() const { return m_BufferImageGranularity > MAX_LOW_BUFFER_IMAGE_GRANULARITY; } -VmaBlockMetadata_Generic::VmaBlockMetadata_Generic(VmaAllocator hAllocator) : - VmaBlockMetadata(hAllocator), - m_FreeCount(0), - m_SumFreeSize(0), - m_Suballocations(VmaStlAllocator<VmaSuballocation>(hAllocator->GetAllocationCallbacks())), - m_FreeSuballocationsBySize(VmaStlAllocator<VmaSuballocationList::iterator>(hAllocator->GetAllocationCallbacks())) + void Init(const VkAllocationCallbacks* pAllocationCallbacks, VkDeviceSize size); + // Before destroying object you must call free it's memory + void Destroy(const VkAllocationCallbacks* pAllocationCallbacks); + + void RoundupAllocRequest(VmaSuballocationType allocType, + VkDeviceSize& inOutAllocSize, + VkDeviceSize& inOutAllocAlignment) const; + + bool CheckConflictAndAlignUp(VkDeviceSize& inOutAllocOffset, + VkDeviceSize allocSize, + VkDeviceSize blockOffset, + VkDeviceSize blockSize, + VmaSuballocationType allocType) const; + + void AllocPages(uint8_t allocType, VkDeviceSize offset, VkDeviceSize size); + void FreePages(VkDeviceSize offset, VkDeviceSize size); + void Clear(); + + ValidationContext StartValidation(const VkAllocationCallbacks* pAllocationCallbacks, + bool isVirutal) const; + bool Validate(ValidationContext& ctx, VkDeviceSize offset, VkDeviceSize size) const; + bool FinishValidation(ValidationContext& ctx) const; + +private: + static const uint16_t MAX_LOW_BUFFER_IMAGE_GRANULARITY = 256; + + struct RegionInfo + { + uint8_t allocType; + uint16_t allocCount; + }; + + VkDeviceSize m_BufferImageGranularity; + uint32_t m_RegionCount; + RegionInfo* m_RegionInfo; + + uint32_t GetStartPage(VkDeviceSize offset) const { return OffsetToPageIndex(offset & ~(m_BufferImageGranularity - 1)); } + uint32_t GetEndPage(VkDeviceSize offset, VkDeviceSize size) const { return OffsetToPageIndex((offset + size - 1) & ~(m_BufferImageGranularity - 1)); } + + uint32_t OffsetToPageIndex(VkDeviceSize offset) const; + void AllocPage(RegionInfo& page, uint8_t allocType); +}; + +#ifndef _VMA_BLOCK_BUFFER_IMAGE_GRANULARITY_FUNCTIONS +VmaBlockBufferImageGranularity::VmaBlockBufferImageGranularity(VkDeviceSize bufferImageGranularity) + : m_BufferImageGranularity(bufferImageGranularity), + m_RegionCount(0), + m_RegionInfo(VMA_NULL) {} + +VmaBlockBufferImageGranularity::~VmaBlockBufferImageGranularity() +{ + VMA_ASSERT(m_RegionInfo == VMA_NULL && "Free not called before destroying object!"); +} + +void VmaBlockBufferImageGranularity::Init(const VkAllocationCallbacks* pAllocationCallbacks, VkDeviceSize size) +{ + if (IsEnabled()) + { + m_RegionCount = static_cast<uint32_t>(VmaDivideRoundingUp(size, m_BufferImageGranularity)); + m_RegionInfo = vma_new_array(pAllocationCallbacks, RegionInfo, m_RegionCount); + memset(m_RegionInfo, 0, m_RegionCount * sizeof(RegionInfo)); + } +} + +void VmaBlockBufferImageGranularity::Destroy(const VkAllocationCallbacks* pAllocationCallbacks) +{ + if (m_RegionInfo) + { + vma_delete_array(pAllocationCallbacks, m_RegionInfo, m_RegionCount); + m_RegionInfo = VMA_NULL; + } +} + +void VmaBlockBufferImageGranularity::RoundupAllocRequest(VmaSuballocationType allocType, + VkDeviceSize& inOutAllocSize, + VkDeviceSize& inOutAllocAlignment) const +{ + if (m_BufferImageGranularity > 1 && + m_BufferImageGranularity <= MAX_LOW_BUFFER_IMAGE_GRANULARITY) + { + if (allocType == VMA_SUBALLOCATION_TYPE_UNKNOWN || + allocType == VMA_SUBALLOCATION_TYPE_IMAGE_UNKNOWN || + allocType == VMA_SUBALLOCATION_TYPE_IMAGE_OPTIMAL) + { + inOutAllocAlignment = VMA_MAX(inOutAllocAlignment, m_BufferImageGranularity); + inOutAllocSize = VmaAlignUp(inOutAllocSize, m_BufferImageGranularity); + } + } +} + +bool VmaBlockBufferImageGranularity::CheckConflictAndAlignUp(VkDeviceSize& inOutAllocOffset, + VkDeviceSize allocSize, + VkDeviceSize blockOffset, + VkDeviceSize blockSize, + VmaSuballocationType allocType) const +{ + if (IsEnabled()) + { + uint32_t startPage = GetStartPage(inOutAllocOffset); + if (m_RegionInfo[startPage].allocCount > 0 && + VmaIsBufferImageGranularityConflict(static_cast<VmaSuballocationType>(m_RegionInfo[startPage].allocType), allocType)) + { + inOutAllocOffset = VmaAlignUp(inOutAllocOffset, m_BufferImageGranularity); + if (blockSize < allocSize + inOutAllocOffset - blockOffset) + return true; + ++startPage; + } + uint32_t endPage = GetEndPage(inOutAllocOffset, allocSize); + if (endPage != startPage && + m_RegionInfo[endPage].allocCount > 0 && + VmaIsBufferImageGranularityConflict(static_cast<VmaSuballocationType>(m_RegionInfo[endPage].allocType), allocType)) + { + return true; + } + } + return false; +} + +void VmaBlockBufferImageGranularity::AllocPages(uint8_t allocType, VkDeviceSize offset, VkDeviceSize size) +{ + if (IsEnabled()) + { + uint32_t startPage = GetStartPage(offset); + AllocPage(m_RegionInfo[startPage], allocType); + + uint32_t endPage = GetEndPage(offset, size); + if (startPage != endPage) + AllocPage(m_RegionInfo[endPage], allocType); + } +} + +void VmaBlockBufferImageGranularity::FreePages(VkDeviceSize offset, VkDeviceSize size) { + if (IsEnabled()) + { + uint32_t startPage = GetStartPage(offset); + --m_RegionInfo[startPage].allocCount; + if (m_RegionInfo[startPage].allocCount == 0) + m_RegionInfo[startPage].allocType = VMA_SUBALLOCATION_TYPE_FREE; + uint32_t endPage = GetEndPage(offset, size); + if (startPage != endPage) + { + --m_RegionInfo[endPage].allocCount; + if (m_RegionInfo[endPage].allocCount == 0) + m_RegionInfo[endPage].allocType = VMA_SUBALLOCATION_TYPE_FREE; + } + } } -VmaBlockMetadata_Generic::~VmaBlockMetadata_Generic() +void VmaBlockBufferImageGranularity::Clear() { + if (m_RegionInfo) + memset(m_RegionInfo, 0, m_RegionCount * sizeof(RegionInfo)); } +VmaBlockBufferImageGranularity::ValidationContext VmaBlockBufferImageGranularity::StartValidation( + const VkAllocationCallbacks* pAllocationCallbacks, bool isVirutal) const +{ + ValidationContext ctx{ pAllocationCallbacks, VMA_NULL }; + if (!isVirutal && IsEnabled()) + { + ctx.pageAllocs = vma_new_array(pAllocationCallbacks, uint16_t, m_RegionCount); + memset(ctx.pageAllocs, 0, m_RegionCount * sizeof(uint16_t)); + } + return ctx; +} + +bool VmaBlockBufferImageGranularity::Validate(ValidationContext& ctx, + VkDeviceSize offset, VkDeviceSize size) const +{ + if (IsEnabled()) + { + uint32_t start = GetStartPage(offset); + ++ctx.pageAllocs[start]; + VMA_VALIDATE(m_RegionInfo[start].allocCount > 0); + + uint32_t end = GetEndPage(offset, size); + if (start != end) + { + ++ctx.pageAllocs[end]; + VMA_VALIDATE(m_RegionInfo[end].allocCount > 0); + } + } + return true; +} + +bool VmaBlockBufferImageGranularity::FinishValidation(ValidationContext& ctx) const +{ + // Check proper page structure + if (IsEnabled()) + { + VMA_ASSERT(ctx.pageAllocs != VMA_NULL && "Validation context not initialized!"); + + for (uint32_t page = 0; page < m_RegionCount; ++page) + { + VMA_VALIDATE(ctx.pageAllocs[page] == m_RegionInfo[page].allocCount); + } + vma_delete_array(ctx.allocCallbacks, ctx.pageAllocs, m_RegionCount); + ctx.pageAllocs = VMA_NULL; + } + return true; +} + +uint32_t VmaBlockBufferImageGranularity::OffsetToPageIndex(VkDeviceSize offset) const +{ + return static_cast<uint32_t>(offset >> VMA_BITSCAN_MSB(m_BufferImageGranularity)); +} + +void VmaBlockBufferImageGranularity::AllocPage(RegionInfo& page, uint8_t allocType) +{ + // When current alloc type is free then it can be overriden by new type + if (page.allocCount == 0 || page.allocCount > 0 && page.allocType == VMA_SUBALLOCATION_TYPE_FREE) + page.allocType = allocType; + + ++page.allocCount; +} +#endif // _VMA_BLOCK_BUFFER_IMAGE_GRANULARITY_FUNCTIONS +#endif // _VMA_BLOCK_BUFFER_IMAGE_GRANULARITY + +#ifndef _VMA_BLOCK_METADATA_GENERIC +class VmaBlockMetadata_Generic : public VmaBlockMetadata +{ + friend class VmaDefragmentationAlgorithm_Generic; + friend class VmaDefragmentationAlgorithm_Fast; + VMA_CLASS_NO_COPY(VmaBlockMetadata_Generic) +public: + VmaBlockMetadata_Generic(const VkAllocationCallbacks* pAllocationCallbacks, + VkDeviceSize bufferImageGranularity, bool isVirtual); + virtual ~VmaBlockMetadata_Generic() = default; + + size_t GetAllocationCount() const override { return m_Suballocations.size() - m_FreeCount; } + VkDeviceSize GetSumFreeSize() const override { return m_SumFreeSize; } + bool IsEmpty() const override { return (m_Suballocations.size() == 1) && (m_FreeCount == 1); } + void Free(VmaAllocHandle allocHandle) override { FreeSuballocation(FindAtOffset((VkDeviceSize)allocHandle - 1)); } + VkDeviceSize GetAllocationOffset(VmaAllocHandle allocHandle) const override { return (VkDeviceSize)allocHandle - 1; }; + + void Init(VkDeviceSize size) override; + bool Validate() const override; + + void CalcAllocationStatInfo(VmaStatInfo& outInfo) const override; + void AddPoolStats(VmaPoolStats& inoutStats) const override; + +#if VMA_STATS_STRING_ENABLED + void PrintDetailedMap(class VmaJsonWriter& json) const override; +#endif + + bool CreateAllocationRequest( + VkDeviceSize allocSize, + VkDeviceSize allocAlignment, + bool upperAddress, + VmaSuballocationType allocType, + uint32_t strategy, + VmaAllocationRequest* pAllocationRequest) override; + + VkResult CheckCorruption(const void* pBlockData) override; + + void Alloc( + const VmaAllocationRequest& request, + VmaSuballocationType type, + void* userData) override; + + void GetAllocationInfo(VmaAllocHandle allocHandle, VmaVirtualAllocationInfo& outInfo) override; + void Clear() override; + void SetAllocationUserData(VmaAllocHandle allocHandle, void* userData) override; + void DebugLogAllAllocations() const override; + + // For defragmentation + bool IsBufferImageGranularityConflictPossible( + VkDeviceSize bufferImageGranularity, + VmaSuballocationType& inOutPrevSuballocType) const; + +private: + uint32_t m_FreeCount; + VkDeviceSize m_SumFreeSize; + VmaSuballocationList m_Suballocations; + // Suballocations that are free. Sorted by size, ascending. + VmaVector<VmaSuballocationList::iterator, VmaStlAllocator<VmaSuballocationList::iterator>> m_FreeSuballocationsBySize; + + VkDeviceSize AlignAllocationSize(VkDeviceSize size) const { return IsVirtual() ? size : VmaAlignUp(size, (VkDeviceSize)16); } + + VmaSuballocationList::iterator FindAtOffset(VkDeviceSize offset); + bool ValidateFreeSuballocationList() const; + + // Checks if requested suballocation with given parameters can be placed in given pFreeSuballocItem. + // If yes, fills pOffset and returns true. If no, returns false. + bool CheckAllocation( + VkDeviceSize allocSize, + VkDeviceSize allocAlignment, + VmaSuballocationType allocType, + VmaSuballocationList::const_iterator suballocItem, + VmaAllocHandle* pAllocHandle) const; + + // Given free suballocation, it merges it with following one, which must also be free. + void MergeFreeWithNext(VmaSuballocationList::iterator item); + // Releases given suballocation, making it free. + // Merges it with adjacent free suballocations if applicable. + // Returns iterator to new free suballocation at this place. + VmaSuballocationList::iterator FreeSuballocation(VmaSuballocationList::iterator suballocItem); + // Given free suballocation, it inserts it into sorted list of + // m_FreeSuballocationsBySize if it is suitable. + void RegisterFreeSuballocation(VmaSuballocationList::iterator item); + // Given free suballocation, it removes it from sorted list of + // m_FreeSuballocationsBySize if it is suitable. + void UnregisterFreeSuballocation(VmaSuballocationList::iterator item); +}; + +#ifndef _VMA_BLOCK_METADATA_GENERIC_FUNCTIONS +VmaBlockMetadata_Generic::VmaBlockMetadata_Generic(const VkAllocationCallbacks* pAllocationCallbacks, + VkDeviceSize bufferImageGranularity, bool isVirtual) + : VmaBlockMetadata(pAllocationCallbacks, bufferImageGranularity, isVirtual), + m_FreeCount(0), + m_SumFreeSize(0), + m_Suballocations(VmaStlAllocator<VmaSuballocation>(pAllocationCallbacks)), + m_FreeSuballocationsBySize(VmaStlAllocator<VmaSuballocationList::iterator>(pAllocationCallbacks)) {} + void VmaBlockMetadata_Generic::Init(VkDeviceSize size) { VmaBlockMetadata::Init(size); @@ -9400,13 +6523,9 @@ void VmaBlockMetadata_Generic::Init(VkDeviceSize size) suballoc.offset = 0; suballoc.size = size; suballoc.type = VMA_SUBALLOCATION_TYPE_FREE; - suballoc.hAllocation = VK_NULL_HANDLE; - VMA_ASSERT(size > VMA_MIN_FREE_SUBALLOCATION_SIZE_TO_REGISTER); m_Suballocations.push_back(suballoc); - VmaSuballocationList::iterator suballocItem = m_Suballocations.end(); - --suballocItem; - m_FreeSuballocationsBySize.push_back(suballocItem); + m_FreeSuballocationsBySize.push_back(m_Suballocations.begin()); } bool VmaBlockMetadata_Generic::Validate() const @@ -9425,7 +6544,9 @@ bool VmaBlockMetadata_Generic::Validate() const // True if previous visited suballocation was free. bool prevFree = false; - for(const auto& subAlloc : m_Suballocations) + const VkDeviceSize debugMargin = GetDebugMargin(); + + for (const auto& subAlloc : m_Suballocations) { // Actual offset of this suballocation doesn't match expected one. VMA_VALIDATE(subAlloc.offset == calculatedOffset); @@ -9434,27 +6555,31 @@ bool VmaBlockMetadata_Generic::Validate() const // Two adjacent free suballocations are invalid. They should be merged. VMA_VALIDATE(!prevFree || !currFree); - VMA_VALIDATE(currFree == (subAlloc.hAllocation == VK_NULL_HANDLE)); + VmaAllocation alloc = (VmaAllocation)subAlloc.userData; + if (!IsVirtual()) + { + VMA_VALIDATE(currFree == (alloc == VK_NULL_HANDLE)); + } - if(currFree) + if (currFree) { calculatedSumFreeSize += subAlloc.size; ++calculatedFreeCount; - if(subAlloc.size >= VMA_MIN_FREE_SUBALLOCATION_SIZE_TO_REGISTER) - { - ++freeSuballocationsToRegister; - } + ++freeSuballocationsToRegister; // Margin required between allocations - every free space must be at least that large. - VMA_VALIDATE(subAlloc.size >= VMA_DEBUG_MARGIN); + VMA_VALIDATE(subAlloc.size >= debugMargin); } else { - VMA_VALIDATE(subAlloc.hAllocation->GetOffset() == subAlloc.offset); - VMA_VALIDATE(subAlloc.hAllocation->GetSize() == subAlloc.size); + if (!IsVirtual()) + { + VMA_VALIDATE((VkDeviceSize)alloc->GetAllocHandle() == subAlloc.offset + 1); + VMA_VALIDATE(alloc->GetSize() == subAlloc.size); + } // Margin required between allocations - previous allocation must be free. - VMA_VALIDATE(VMA_DEBUG_MARGIN == 0 || prevFree); + VMA_VALIDATE(debugMargin == 0 || prevFree); } calculatedOffset += subAlloc.size; @@ -9466,7 +6591,7 @@ bool VmaBlockMetadata_Generic::Validate() const VMA_VALIDATE(m_FreeSuballocationsBySize.size() == freeSuballocationsToRegister); VkDeviceSize lastSize = 0; - for(size_t i = 0; i < m_FreeSuballocationsBySize.size(); ++i) + for (size_t i = 0; i < m_FreeSuballocationsBySize.size(); ++i) { VmaSuballocationList::iterator suballocItem = m_FreeSuballocationsBySize[i]; @@ -9487,50 +6612,21 @@ bool VmaBlockMetadata_Generic::Validate() const return true; } -VkDeviceSize VmaBlockMetadata_Generic::GetUnusedRangeSizeMax() const -{ - if(!m_FreeSuballocationsBySize.empty()) - { - return m_FreeSuballocationsBySize.back()->size; - } - else - { - return 0; - } -} - -bool VmaBlockMetadata_Generic::IsEmpty() const -{ - return (m_Suballocations.size() == 1) && (m_FreeCount == 1); -} - void VmaBlockMetadata_Generic::CalcAllocationStatInfo(VmaStatInfo& outInfo) const { - outInfo.blockCount = 1; - const uint32_t rangeCount = (uint32_t)m_Suballocations.size(); - outInfo.allocationCount = rangeCount - m_FreeCount; - outInfo.unusedRangeCount = m_FreeCount; - - outInfo.unusedBytes = m_SumFreeSize; - outInfo.usedBytes = GetSize() - outInfo.unusedBytes; - - outInfo.allocationSizeMin = UINT64_MAX; - outInfo.allocationSizeMax = 0; - outInfo.unusedRangeSizeMin = UINT64_MAX; - outInfo.unusedRangeSizeMax = 0; + VmaInitStatInfo(outInfo); + outInfo.blockCount = 1; - for(const auto& suballoc : m_Suballocations) + for (const auto& suballoc : m_Suballocations) { - if(suballoc.type != VMA_SUBALLOCATION_TYPE_FREE) + if (suballoc.type != VMA_SUBALLOCATION_TYPE_FREE) { - outInfo.allocationSizeMin = VMA_MIN(outInfo.allocationSizeMin, suballoc.size); - outInfo.allocationSizeMax = VMA_MAX(outInfo.allocationSizeMax, suballoc.size); + VmaAddStatInfoAllocation(outInfo, suballoc.size); } else { - outInfo.unusedRangeSizeMin = VMA_MIN(outInfo.unusedRangeSizeMin, suballoc.size); - outInfo.unusedRangeSizeMax = VMA_MAX(outInfo.unusedRangeSizeMax, suballoc.size); + VmaAddStatInfoUnusedRange(outInfo, suballoc.size); } } } @@ -9543,11 +6639,9 @@ void VmaBlockMetadata_Generic::AddPoolStats(VmaPoolStats& inoutStats) const inoutStats.unusedSize += m_SumFreeSize; inoutStats.allocationCount += rangeCount - m_FreeCount; inoutStats.unusedRangeCount += m_FreeCount; - inoutStats.unusedRangeSizeMax = VMA_MAX(inoutStats.unusedRangeSizeMax, GetUnusedRangeSizeMax()); } #if VMA_STATS_STRING_ENABLED - void VmaBlockMetadata_Generic::PrintDetailedMap(class VmaJsonWriter& json) const { PrintDetailedMap_Begin(json, @@ -9555,33 +6649,27 @@ void VmaBlockMetadata_Generic::PrintDetailedMap(class VmaJsonWriter& json) const m_Suballocations.size() - (size_t)m_FreeCount, // allocationCount m_FreeCount); // unusedRangeCount - size_t i = 0; - for(const auto& suballoc : m_Suballocations) + for (const auto& suballoc : m_Suballocations) { - if(suballoc.type == VMA_SUBALLOCATION_TYPE_FREE) + if (suballoc.type == VMA_SUBALLOCATION_TYPE_FREE) { PrintDetailedMap_UnusedRange(json, suballoc.offset, suballoc.size); } else { - PrintDetailedMap_Allocation(json, suballoc.offset, suballoc.hAllocation); + PrintDetailedMap_Allocation(json, suballoc.offset, suballoc.size, suballoc.userData); } } PrintDetailedMap_End(json); } - -#endif // #if VMA_STATS_STRING_ENABLED +#endif // VMA_STATS_STRING_ENABLED bool VmaBlockMetadata_Generic::CreateAllocationRequest( - uint32_t currentFrameIndex, - uint32_t frameInUseCount, - VkDeviceSize bufferImageGranularity, VkDeviceSize allocSize, VkDeviceSize allocAlignment, bool upperAddress, VmaSuballocationType allocType, - bool canMakeOtherLost, uint32_t strategy, VmaAllocationRequest* pAllocationRequest) { @@ -9591,92 +6679,77 @@ bool VmaBlockMetadata_Generic::CreateAllocationRequest( VMA_ASSERT(pAllocationRequest != VMA_NULL); VMA_HEAVY_ASSERT(Validate()); + allocSize = AlignAllocationSize(allocSize); + pAllocationRequest->type = VmaAllocationRequestType::Normal; + pAllocationRequest->size = allocSize; + + const VkDeviceSize debugMargin = GetDebugMargin(); - // There is not enough total free space in this block to fullfill the request: Early return. - if(canMakeOtherLost == false && - m_SumFreeSize < allocSize + 2 * VMA_DEBUG_MARGIN) + // There is not enough total free space in this block to fulfill the request: Early return. + if (m_SumFreeSize < allocSize + debugMargin) { return false; } // New algorithm, efficiently searching freeSuballocationsBySize. const size_t freeSuballocCount = m_FreeSuballocationsBySize.size(); - if(freeSuballocCount > 0) + if (freeSuballocCount > 0) { - if(strategy == VMA_ALLOCATION_CREATE_STRATEGY_BEST_FIT_BIT) + if (strategy == 0 || + strategy == VMA_ALLOCATION_CREATE_STRATEGY_MIN_MEMORY_BIT) { - // Find first free suballocation with size not less than allocSize + 2 * VMA_DEBUG_MARGIN. + // Find first free suballocation with size not less than allocSize + debugMargin. VmaSuballocationList::iterator* const it = VmaBinaryFindFirstNotLess( m_FreeSuballocationsBySize.data(), m_FreeSuballocationsBySize.data() + freeSuballocCount, - allocSize + 2 * VMA_DEBUG_MARGIN, + allocSize + debugMargin, VmaSuballocationItemSizeLess()); size_t index = it - m_FreeSuballocationsBySize.data(); - for(; index < freeSuballocCount; ++index) + for (; index < freeSuballocCount; ++index) { - if(CheckAllocation( - currentFrameIndex, - frameInUseCount, - bufferImageGranularity, + if (CheckAllocation( allocSize, allocAlignment, allocType, m_FreeSuballocationsBySize[index], - false, // canMakeOtherLost - &pAllocationRequest->offset, - &pAllocationRequest->itemsToMakeLostCount, - &pAllocationRequest->sumFreeSize, - &pAllocationRequest->sumItemSize)) + &pAllocationRequest->allocHandle)) { pAllocationRequest->item = m_FreeSuballocationsBySize[index]; return true; } } } - else if(strategy == VMA_ALLOCATION_INTERNAL_STRATEGY_MIN_OFFSET) + else if (strategy == VMA_ALLOCATION_INTERNAL_STRATEGY_MIN_OFFSET) { - for(VmaSuballocationList::iterator it = m_Suballocations.begin(); + for (VmaSuballocationList::iterator it = m_Suballocations.begin(); it != m_Suballocations.end(); ++it) { - if(it->type == VMA_SUBALLOCATION_TYPE_FREE && CheckAllocation( - currentFrameIndex, - frameInUseCount, - bufferImageGranularity, + if (it->type == VMA_SUBALLOCATION_TYPE_FREE && CheckAllocation( allocSize, allocAlignment, allocType, it, - false, // canMakeOtherLost - &pAllocationRequest->offset, - &pAllocationRequest->itemsToMakeLostCount, - &pAllocationRequest->sumFreeSize, - &pAllocationRequest->sumItemSize)) + &pAllocationRequest->allocHandle)) { pAllocationRequest->item = it; return true; } } } - else // WORST_FIT, FIRST_FIT + else { + VMA_ASSERT(strategy == VMA_ALLOCATION_CREATE_STRATEGY_MIN_TIME_BIT); // Search staring from biggest suballocations. - for(size_t index = freeSuballocCount; index--; ) + for (size_t index = freeSuballocCount; index--; ) { - if(CheckAllocation( - currentFrameIndex, - frameInUseCount, - bufferImageGranularity, + if (CheckAllocation( allocSize, allocAlignment, allocType, m_FreeSuballocationsBySize[index], - false, // canMakeOtherLost - &pAllocationRequest->offset, - &pAllocationRequest->itemsToMakeLostCount, - &pAllocationRequest->sumFreeSize, - &pAllocationRequest->sumItemSize)) + &pAllocationRequest->allocHandle)) { pAllocationRequest->item = m_FreeSuballocationsBySize[index]; return true; @@ -9685,123 +6758,19 @@ bool VmaBlockMetadata_Generic::CreateAllocationRequest( } } - if(canMakeOtherLost) - { - // Brute-force algorithm. TODO: Come up with something better. - - bool found = false; - VmaAllocationRequest tmpAllocRequest = {}; - tmpAllocRequest.type = VmaAllocationRequestType::Normal; - for(VmaSuballocationList::iterator suballocIt = m_Suballocations.begin(); - suballocIt != m_Suballocations.end(); - ++suballocIt) - { - if(suballocIt->type == VMA_SUBALLOCATION_TYPE_FREE || - suballocIt->hAllocation->CanBecomeLost()) - { - if(CheckAllocation( - currentFrameIndex, - frameInUseCount, - bufferImageGranularity, - allocSize, - allocAlignment, - allocType, - suballocIt, - canMakeOtherLost, - &tmpAllocRequest.offset, - &tmpAllocRequest.itemsToMakeLostCount, - &tmpAllocRequest.sumFreeSize, - &tmpAllocRequest.sumItemSize)) - { - if(strategy == VMA_ALLOCATION_CREATE_STRATEGY_FIRST_FIT_BIT) - { - *pAllocationRequest = tmpAllocRequest; - pAllocationRequest->item = suballocIt; - break; - } - if(!found || tmpAllocRequest.CalcCost() < pAllocationRequest->CalcCost()) - { - *pAllocationRequest = tmpAllocRequest; - pAllocationRequest->item = suballocIt; - found = true; - } - } - } - } - - return found; - } - return false; } -bool VmaBlockMetadata_Generic::MakeRequestedAllocationsLost( - uint32_t currentFrameIndex, - uint32_t frameInUseCount, - VmaAllocationRequest* pAllocationRequest) -{ - VMA_ASSERT(pAllocationRequest && pAllocationRequest->type == VmaAllocationRequestType::Normal); - - while(pAllocationRequest->itemsToMakeLostCount > 0) - { - if(pAllocationRequest->item->type == VMA_SUBALLOCATION_TYPE_FREE) - { - ++pAllocationRequest->item; - } - VMA_ASSERT(pAllocationRequest->item != m_Suballocations.end()); - VMA_ASSERT(pAllocationRequest->item->hAllocation != VK_NULL_HANDLE); - VMA_ASSERT(pAllocationRequest->item->hAllocation->CanBecomeLost()); - if(pAllocationRequest->item->hAllocation->MakeLost(currentFrameIndex, frameInUseCount)) - { - pAllocationRequest->item = FreeSuballocation(pAllocationRequest->item); - --pAllocationRequest->itemsToMakeLostCount; - } - else - { - return false; - } - } - - VMA_HEAVY_ASSERT(Validate()); - VMA_ASSERT(pAllocationRequest->item != m_Suballocations.end()); - VMA_ASSERT(pAllocationRequest->item->type == VMA_SUBALLOCATION_TYPE_FREE); - - return true; -} - -uint32_t VmaBlockMetadata_Generic::MakeAllocationsLost(uint32_t currentFrameIndex, uint32_t frameInUseCount) -{ - uint32_t lostAllocationCount = 0; - for(VmaSuballocationList::iterator it = m_Suballocations.begin(); - it != m_Suballocations.end(); - ++it) - { - if(it->type != VMA_SUBALLOCATION_TYPE_FREE && - it->hAllocation->CanBecomeLost() && - it->hAllocation->MakeLost(currentFrameIndex, frameInUseCount)) - { - it = FreeSuballocation(it); - ++lostAllocationCount; - } - } - return lostAllocationCount; -} - VkResult VmaBlockMetadata_Generic::CheckCorruption(const void* pBlockData) { - for(auto& suballoc : m_Suballocations) + for (auto& suballoc : m_Suballocations) { - if(suballoc.type != VMA_SUBALLOCATION_TYPE_FREE) + if (suballoc.type != VMA_SUBALLOCATION_TYPE_FREE) { - if(!VmaValidateMagicValue(pBlockData, suballoc.offset - VMA_DEBUG_MARGIN)) - { - VMA_ASSERT(0 && "MEMORY CORRUPTION DETECTED BEFORE VALIDATED ALLOCATION!"); - return VK_ERROR_VALIDATION_FAILED_EXT; - } - if(!VmaValidateMagicValue(pBlockData, suballoc.offset + suballoc.size)) + if (!VmaValidateMagicValue(pBlockData, suballoc.offset + suballoc.size)) { VMA_ASSERT(0 && "MEMORY CORRUPTION DETECTED AFTER VALIDATED ALLOCATION!"); - return VK_ERROR_VALIDATION_FAILED_EXT; + return VK_ERROR_UNKNOWN; } } } @@ -9812,34 +6781,34 @@ VkResult VmaBlockMetadata_Generic::CheckCorruption(const void* pBlockData) void VmaBlockMetadata_Generic::Alloc( const VmaAllocationRequest& request, VmaSuballocationType type, - VkDeviceSize allocSize, - VmaAllocation hAllocation) + void* userData) { VMA_ASSERT(request.type == VmaAllocationRequestType::Normal); VMA_ASSERT(request.item != m_Suballocations.end()); VmaSuballocation& suballoc = *request.item; // Given suballocation is a free block. VMA_ASSERT(suballoc.type == VMA_SUBALLOCATION_TYPE_FREE); + // Given offset is inside this suballocation. - VMA_ASSERT(request.offset >= suballoc.offset); - const VkDeviceSize paddingBegin = request.offset - suballoc.offset; - VMA_ASSERT(suballoc.size >= paddingBegin + allocSize); - const VkDeviceSize paddingEnd = suballoc.size - paddingBegin - allocSize; + VMA_ASSERT((VkDeviceSize)request.allocHandle - 1 >= suballoc.offset); + const VkDeviceSize paddingBegin = (VkDeviceSize)request.allocHandle - suballoc.offset - 1; + VMA_ASSERT(suballoc.size >= paddingBegin + request.size); + const VkDeviceSize paddingEnd = suballoc.size - paddingBegin - request.size; // Unregister this free suballocation from m_FreeSuballocationsBySize and update // it to become used. UnregisterFreeSuballocation(request.item); - suballoc.offset = request.offset; - suballoc.size = allocSize; + suballoc.offset = (VkDeviceSize)request.allocHandle - 1; + suballoc.size = request.size; suballoc.type = type; - suballoc.hAllocation = hAllocation; + suballoc.userData = userData; // If there are any free bytes remaining at the end, insert new free suballocation after current one. - if(paddingEnd) + if (paddingEnd) { VmaSuballocation paddingSuballoc = {}; - paddingSuballoc.offset = request.offset + allocSize; + paddingSuballoc.offset = suballoc.offset + suballoc.size; paddingSuballoc.size = paddingEnd; paddingSuballoc.type = VMA_SUBALLOCATION_TYPE_FREE; VmaSuballocationList::iterator next = request.item; @@ -9850,10 +6819,10 @@ void VmaBlockMetadata_Generic::Alloc( } // If there are any free bytes remaining at the beginning, insert new free suballocation before current one. - if(paddingBegin) + if (paddingBegin) { VmaSuballocation paddingSuballoc = {}; - paddingSuballoc.offset = request.offset - paddingBegin; + paddingSuballoc.offset = suballoc.offset - paddingBegin; paddingSuballoc.size = paddingBegin; paddingSuballoc.type = VMA_SUBALLOCATION_TYPE_FREE; const VmaSuballocationList::iterator paddingBeginItem = @@ -9863,59 +6832,100 @@ void VmaBlockMetadata_Generic::Alloc( // Update totals. m_FreeCount = m_FreeCount - 1; - if(paddingBegin > 0) + if (paddingBegin > 0) { ++m_FreeCount; } - if(paddingEnd > 0) + if (paddingEnd > 0) { ++m_FreeCount; } - m_SumFreeSize -= allocSize; + m_SumFreeSize -= request.size; +} + +void VmaBlockMetadata_Generic::GetAllocationInfo(VmaAllocHandle allocHandle, VmaVirtualAllocationInfo& outInfo) +{ + outInfo.offset = (VkDeviceSize)allocHandle - 1; + const VmaSuballocation& suballoc = *FindAtOffset(outInfo.offset); + outInfo.size = suballoc.size; + outInfo.pUserData = suballoc.userData; +} + +void VmaBlockMetadata_Generic::Clear() +{ + const VkDeviceSize size = GetSize(); + + VMA_ASSERT(IsVirtual()); + m_FreeCount = 1; + m_SumFreeSize = size; + m_Suballocations.clear(); + m_FreeSuballocationsBySize.clear(); + + VmaSuballocation suballoc = {}; + suballoc.offset = 0; + suballoc.size = size; + suballoc.type = VMA_SUBALLOCATION_TYPE_FREE; + m_Suballocations.push_back(suballoc); + + m_FreeSuballocationsBySize.push_back(m_Suballocations.begin()); } -void VmaBlockMetadata_Generic::Free(const VmaAllocation allocation) +void VmaBlockMetadata_Generic::SetAllocationUserData(VmaAllocHandle allocHandle, void* userData) { - for(VmaSuballocationList::iterator suballocItem = m_Suballocations.begin(); - suballocItem != m_Suballocations.end(); - ++suballocItem) + VmaSuballocation& suballoc = *FindAtOffset((VkDeviceSize)allocHandle - 1); + suballoc.userData = userData; +} + +void VmaBlockMetadata_Generic::DebugLogAllAllocations() const +{ + for (const auto& suballoc : m_Suballocations) { - VmaSuballocation& suballoc = *suballocItem; - if(suballoc.hAllocation == allocation) - { - FreeSuballocation(suballocItem); - VMA_HEAVY_ASSERT(Validate()); - return; - } + if (suballoc.type != VMA_SUBALLOCATION_TYPE_FREE) + DebugLogAllocation(suballoc.offset, suballoc.size, suballoc.userData); } - VMA_ASSERT(0 && "Not found!"); } -void VmaBlockMetadata_Generic::FreeAtOffset(VkDeviceSize offset) +VmaSuballocationList::iterator VmaBlockMetadata_Generic::FindAtOffset(VkDeviceSize offset) { - for(VmaSuballocationList::iterator suballocItem = m_Suballocations.begin(); - suballocItem != m_Suballocations.end(); - ++suballocItem) + VMA_HEAVY_ASSERT(!m_Suballocations.empty()); + const VkDeviceSize last = m_Suballocations.rbegin()->offset; + if (last == offset) + return m_Suballocations.rbegin(); + const VkDeviceSize first = m_Suballocations.begin()->offset; + if (first == offset) + return m_Suballocations.begin(); + + const size_t suballocCount = m_Suballocations.size(); + const VkDeviceSize step = (last - first + m_Suballocations.begin()->size) / suballocCount; + auto findSuballocation = [&](auto begin, auto end) -> VmaSuballocationList::iterator { - VmaSuballocation& suballoc = *suballocItem; - if(suballoc.offset == offset) + for (auto suballocItem = begin; + suballocItem != end; + ++suballocItem) { - FreeSuballocation(suballocItem); - return; + VmaSuballocation& suballoc = *suballocItem; + if (suballoc.offset == offset) + return suballocItem; } + VMA_ASSERT(false && "Not found!"); + return m_Suballocations.end(); + }; + // If requested offset is closer to the end of range, search from the end + if (offset - first > suballocCount * step / 2) + { + return findSuballocation(m_Suballocations.rbegin(), m_Suballocations.rend()); } - VMA_ASSERT(0 && "Not found!"); + return findSuballocation(m_Suballocations.begin(), m_Suballocations.end()); } bool VmaBlockMetadata_Generic::ValidateFreeSuballocationList() const { VkDeviceSize lastSize = 0; - for(size_t i = 0, count = m_FreeSuballocationsBySize.size(); i < count; ++i) + for (size_t i = 0, count = m_FreeSuballocationsBySize.size(); i < count; ++i) { const VmaSuballocationList::iterator it = m_FreeSuballocationsBySize[i]; VMA_VALIDATE(it->type == VMA_SUBALLOCATION_TYPE_FREE); - VMA_VALIDATE(it->size >= VMA_MIN_FREE_SUBALLOCATION_SIZE_TO_REGISTER); VMA_VALIDATE(it->size >= lastSize); lastSize = it->size; } @@ -9923,276 +6933,105 @@ bool VmaBlockMetadata_Generic::ValidateFreeSuballocationList() const } bool VmaBlockMetadata_Generic::CheckAllocation( - uint32_t currentFrameIndex, - uint32_t frameInUseCount, - VkDeviceSize bufferImageGranularity, VkDeviceSize allocSize, VkDeviceSize allocAlignment, VmaSuballocationType allocType, VmaSuballocationList::const_iterator suballocItem, - bool canMakeOtherLost, - VkDeviceSize* pOffset, - size_t* itemsToMakeLostCount, - VkDeviceSize* pSumFreeSize, - VkDeviceSize* pSumItemSize) const + VmaAllocHandle* pAllocHandle) const { VMA_ASSERT(allocSize > 0); VMA_ASSERT(allocType != VMA_SUBALLOCATION_TYPE_FREE); VMA_ASSERT(suballocItem != m_Suballocations.cend()); - VMA_ASSERT(pOffset != VMA_NULL); + VMA_ASSERT(pAllocHandle != VMA_NULL); - *itemsToMakeLostCount = 0; - *pSumFreeSize = 0; - *pSumItemSize = 0; + const VkDeviceSize debugMargin = GetDebugMargin(); + const VkDeviceSize bufferImageGranularity = GetBufferImageGranularity(); - if(canMakeOtherLost) - { - if(suballocItem->type == VMA_SUBALLOCATION_TYPE_FREE) - { - *pSumFreeSize = suballocItem->size; - } - else - { - if(suballocItem->hAllocation->CanBecomeLost() && - suballocItem->hAllocation->GetLastUseFrameIndex() + frameInUseCount < currentFrameIndex) - { - ++*itemsToMakeLostCount; - *pSumItemSize = suballocItem->size; - } - else - { - return false; - } - } + const VmaSuballocation& suballoc = *suballocItem; + VMA_ASSERT(suballoc.type == VMA_SUBALLOCATION_TYPE_FREE); - // Remaining size is too small for this request: Early return. - if(GetSize() - suballocItem->offset < allocSize) - { - return false; - } + // Size of this suballocation is too small for this request: Early return. + if (suballoc.size < allocSize) + { + return false; + } - // Start from offset equal to beginning of this suballocation. - *pOffset = suballocItem->offset; + // Start from offset equal to beginning of this suballocation. + VkDeviceSize offset = suballoc.offset + (suballocItem == m_Suballocations.cbegin() ? 0 : GetDebugMargin()); - // Apply VMA_DEBUG_MARGIN at the beginning. - if(VMA_DEBUG_MARGIN > 0) - { - *pOffset += VMA_DEBUG_MARGIN; - } + // Apply debugMargin from the end of previous alloc. + if (debugMargin > 0) + { + offset += debugMargin; + } - // Apply alignment. - *pOffset = VmaAlignUp(*pOffset, allocAlignment); + // Apply alignment. + offset = VmaAlignUp(offset, allocAlignment); - // Check previous suballocations for BufferImageGranularity conflicts. - // Make bigger alignment if necessary. - if(bufferImageGranularity > 1 && bufferImageGranularity != allocAlignment) + // Check previous suballocations for BufferImageGranularity conflicts. + // Make bigger alignment if necessary. + if (bufferImageGranularity > 1 && bufferImageGranularity != allocAlignment) + { + bool bufferImageGranularityConflict = false; + VmaSuballocationList::const_iterator prevSuballocItem = suballocItem; + while (prevSuballocItem != m_Suballocations.cbegin()) { - bool bufferImageGranularityConflict = false; - VmaSuballocationList::const_iterator prevSuballocItem = suballocItem; - while(prevSuballocItem != m_Suballocations.cbegin()) + --prevSuballocItem; + const VmaSuballocation& prevSuballoc = *prevSuballocItem; + if (VmaBlocksOnSamePage(prevSuballoc.offset, prevSuballoc.size, offset, bufferImageGranularity)) { - --prevSuballocItem; - const VmaSuballocation& prevSuballoc = *prevSuballocItem; - if(VmaBlocksOnSamePage(prevSuballoc.offset, prevSuballoc.size, *pOffset, bufferImageGranularity)) + if (VmaIsBufferImageGranularityConflict(prevSuballoc.type, allocType)) { - if(VmaIsBufferImageGranularityConflict(prevSuballoc.type, allocType)) - { - bufferImageGranularityConflict = true; - break; - } - } - else - // Already on previous page. + bufferImageGranularityConflict = true; break; - } - if(bufferImageGranularityConflict) - { - *pOffset = VmaAlignUp(*pOffset, bufferImageGranularity); - } - } - - // Now that we have final *pOffset, check if we are past suballocItem. - // If yes, return false - this function should be called for another suballocItem as starting point. - if(*pOffset >= suballocItem->offset + suballocItem->size) - { - return false; - } - - // Calculate padding at the beginning based on current offset. - const VkDeviceSize paddingBegin = *pOffset - suballocItem->offset; - - // Calculate required margin at the end. - const VkDeviceSize requiredEndMargin = VMA_DEBUG_MARGIN; - - const VkDeviceSize totalSize = paddingBegin + allocSize + requiredEndMargin; - // Another early return check. - if(suballocItem->offset + totalSize > GetSize()) - { - return false; - } - - // Advance lastSuballocItem until desired size is reached. - // Update itemsToMakeLostCount. - VmaSuballocationList::const_iterator lastSuballocItem = suballocItem; - if(totalSize > suballocItem->size) - { - VkDeviceSize remainingSize = totalSize - suballocItem->size; - while(remainingSize > 0) - { - ++lastSuballocItem; - if(lastSuballocItem == m_Suballocations.cend()) - { - return false; - } - if(lastSuballocItem->type == VMA_SUBALLOCATION_TYPE_FREE) - { - *pSumFreeSize += lastSuballocItem->size; - } - else - { - VMA_ASSERT(lastSuballocItem->hAllocation != VK_NULL_HANDLE); - if(lastSuballocItem->hAllocation->CanBecomeLost() && - lastSuballocItem->hAllocation->GetLastUseFrameIndex() + frameInUseCount < currentFrameIndex) - { - ++*itemsToMakeLostCount; - *pSumItemSize += lastSuballocItem->size; - } - else - { - return false; - } } - remainingSize = (lastSuballocItem->size < remainingSize) ? - remainingSize - lastSuballocItem->size : 0; } + else + // Already on previous page. + break; } - - // Check next suballocations for BufferImageGranularity conflicts. - // If conflict exists, we must mark more allocations lost or fail. - if(allocSize % bufferImageGranularity || *pOffset % bufferImageGranularity) + if (bufferImageGranularityConflict) { - VmaSuballocationList::const_iterator nextSuballocItem = lastSuballocItem; - ++nextSuballocItem; - while(nextSuballocItem != m_Suballocations.cend()) - { - const VmaSuballocation& nextSuballoc = *nextSuballocItem; - if(VmaBlocksOnSamePage(*pOffset, allocSize, nextSuballoc.offset, bufferImageGranularity)) - { - if(VmaIsBufferImageGranularityConflict(allocType, nextSuballoc.type)) - { - VMA_ASSERT(nextSuballoc.hAllocation != VK_NULL_HANDLE); - if(nextSuballoc.hAllocation->CanBecomeLost() && - nextSuballoc.hAllocation->GetLastUseFrameIndex() + frameInUseCount < currentFrameIndex) - { - ++*itemsToMakeLostCount; - } - else - { - return false; - } - } - } - else - { - // Already on next page. - break; - } - ++nextSuballocItem; - } + offset = VmaAlignUp(offset, bufferImageGranularity); } } - else - { - const VmaSuballocation& suballoc = *suballocItem; - VMA_ASSERT(suballoc.type == VMA_SUBALLOCATION_TYPE_FREE); - *pSumFreeSize = suballoc.size; + // Calculate padding at the beginning based on current offset. + const VkDeviceSize paddingBegin = offset - suballoc.offset; - // Size of this suballocation is too small for this request: Early return. - if(suballoc.size < allocSize) - { - return false; - } - - // Start from offset equal to beginning of this suballocation. - *pOffset = suballoc.offset; - - // Apply VMA_DEBUG_MARGIN at the beginning. - if(VMA_DEBUG_MARGIN > 0) - { - *pOffset += VMA_DEBUG_MARGIN; - } - - // Apply alignment. - *pOffset = VmaAlignUp(*pOffset, allocAlignment); + // Fail if requested size plus margin after is bigger than size of this suballocation. + if (paddingBegin + allocSize + debugMargin > suballoc.size) + { + return false; + } - // Check previous suballocations for BufferImageGranularity conflicts. - // Make bigger alignment if necessary. - if(bufferImageGranularity > 1 && bufferImageGranularity != allocAlignment) + // Check next suballocations for BufferImageGranularity conflicts. + // If conflict exists, allocation cannot be made here. + if (allocSize % bufferImageGranularity || offset % bufferImageGranularity) + { + VmaSuballocationList::const_iterator nextSuballocItem = suballocItem; + ++nextSuballocItem; + while (nextSuballocItem != m_Suballocations.cend()) { - bool bufferImageGranularityConflict = false; - VmaSuballocationList::const_iterator prevSuballocItem = suballocItem; - while(prevSuballocItem != m_Suballocations.cbegin()) + const VmaSuballocation& nextSuballoc = *nextSuballocItem; + if (VmaBlocksOnSamePage(offset, allocSize, nextSuballoc.offset, bufferImageGranularity)) { - --prevSuballocItem; - const VmaSuballocation& prevSuballoc = *prevSuballocItem; - if(VmaBlocksOnSamePage(prevSuballoc.offset, prevSuballoc.size, *pOffset, bufferImageGranularity)) + if (VmaIsBufferImageGranularityConflict(allocType, nextSuballoc.type)) { - if(VmaIsBufferImageGranularityConflict(prevSuballoc.type, allocType)) - { - bufferImageGranularityConflict = true; - break; - } + return false; } - else - // Already on previous page. - break; } - if(bufferImageGranularityConflict) + else { - *pOffset = VmaAlignUp(*pOffset, bufferImageGranularity); + // Already on next page. + break; } - } - - // Calculate padding at the beginning based on current offset. - const VkDeviceSize paddingBegin = *pOffset - suballoc.offset; - - // Calculate required margin at the end. - const VkDeviceSize requiredEndMargin = VMA_DEBUG_MARGIN; - - // Fail if requested size plus margin before and after is bigger than size of this suballocation. - if(paddingBegin + allocSize + requiredEndMargin > suballoc.size) - { - return false; - } - - // Check next suballocations for BufferImageGranularity conflicts. - // If conflict exists, allocation cannot be made here. - if(allocSize % bufferImageGranularity || *pOffset % bufferImageGranularity) - { - VmaSuballocationList::const_iterator nextSuballocItem = suballocItem; ++nextSuballocItem; - while(nextSuballocItem != m_Suballocations.cend()) - { - const VmaSuballocation& nextSuballoc = *nextSuballocItem; - if(VmaBlocksOnSamePage(*pOffset, allocSize, nextSuballoc.offset, bufferImageGranularity)) - { - if(VmaIsBufferImageGranularityConflict(allocType, nextSuballoc.type)) - { - return false; - } - } - else - { - // Already on next page. - break; - } - ++nextSuballocItem; - } } } - // All tests passed: Success. pOffset is already filled. + *pAllocHandle = (VmaAllocHandle)(offset + 1); + // All tests passed: Success. pAllocHandle is already filled. return true; } @@ -10216,7 +7055,7 @@ VmaSuballocationList::iterator VmaBlockMetadata_Generic::FreeSuballocation(VmaSu // Change this suballocation to be marked as free. VmaSuballocation& suballoc = *suballocItem; suballoc.type = VMA_SUBALLOCATION_TYPE_FREE; - suballoc.hAllocation = VK_NULL_HANDLE; + suballoc.userData = VMA_NULL; // Update totals. ++m_FreeCount; @@ -10228,28 +7067,28 @@ VmaSuballocationList::iterator VmaBlockMetadata_Generic::FreeSuballocation(VmaSu VmaSuballocationList::iterator nextItem = suballocItem; ++nextItem; - if((nextItem != m_Suballocations.end()) && (nextItem->type == VMA_SUBALLOCATION_TYPE_FREE)) + if ((nextItem != m_Suballocations.end()) && (nextItem->type == VMA_SUBALLOCATION_TYPE_FREE)) { mergeWithNext = true; } VmaSuballocationList::iterator prevItem = suballocItem; - if(suballocItem != m_Suballocations.begin()) + if (suballocItem != m_Suballocations.begin()) { --prevItem; - if(prevItem->type == VMA_SUBALLOCATION_TYPE_FREE) + if (prevItem->type == VMA_SUBALLOCATION_TYPE_FREE) { mergeWithPrev = true; } } - if(mergeWithNext) + if (mergeWithNext) { UnregisterFreeSuballocation(nextItem); MergeFreeWithNext(suballocItem); } - if(mergeWithPrev) + if (mergeWithPrev) { UnregisterFreeSuballocation(prevItem); MergeFreeWithNext(prevItem); @@ -10272,22 +7111,18 @@ void VmaBlockMetadata_Generic::RegisterFreeSuballocation(VmaSuballocationList::i // this function, depending on what do you want to check. VMA_HEAVY_ASSERT(ValidateFreeSuballocationList()); - if(item->size >= VMA_MIN_FREE_SUBALLOCATION_SIZE_TO_REGISTER) + if (m_FreeSuballocationsBySize.empty()) { - if(m_FreeSuballocationsBySize.empty()) - { - m_FreeSuballocationsBySize.push_back(item); - } - else - { - VmaVectorInsertSorted<VmaSuballocationItemSizeLess>(m_FreeSuballocationsBySize, item); - } + m_FreeSuballocationsBySize.push_back(item); + } + else + { + VmaVectorInsertSorted<VmaSuballocationItemSizeLess>(m_FreeSuballocationsBySize, item); } //VMA_HEAVY_ASSERT(ValidateFreeSuballocationList()); } - void VmaBlockMetadata_Generic::UnregisterFreeSuballocation(VmaSuballocationList::iterator item) { VMA_ASSERT(item->type == VMA_SUBALLOCATION_TYPE_FREE); @@ -10297,26 +7132,23 @@ void VmaBlockMetadata_Generic::UnregisterFreeSuballocation(VmaSuballocationList: // this function, depending on what do you want to check. VMA_HEAVY_ASSERT(ValidateFreeSuballocationList()); - if(item->size >= VMA_MIN_FREE_SUBALLOCATION_SIZE_TO_REGISTER) + VmaSuballocationList::iterator* const it = VmaBinaryFindFirstNotLess( + m_FreeSuballocationsBySize.data(), + m_FreeSuballocationsBySize.data() + m_FreeSuballocationsBySize.size(), + item, + VmaSuballocationItemSizeLess()); + for (size_t index = it - m_FreeSuballocationsBySize.data(); + index < m_FreeSuballocationsBySize.size(); + ++index) { - VmaSuballocationList::iterator* const it = VmaBinaryFindFirstNotLess( - m_FreeSuballocationsBySize.data(), - m_FreeSuballocationsBySize.data() + m_FreeSuballocationsBySize.size(), - item, - VmaSuballocationItemSizeLess()); - for(size_t index = it - m_FreeSuballocationsBySize.data(); - index < m_FreeSuballocationsBySize.size(); - ++index) + if (m_FreeSuballocationsBySize[index] == item) { - if(m_FreeSuballocationsBySize[index] == item) - { - VmaVectorRemove(m_FreeSuballocationsBySize, index); - return; - } - VMA_ASSERT((m_FreeSuballocationsBySize[index]->size == item->size) && "Not found."); + VmaVectorRemove(m_FreeSuballocationsBySize, index); + return; } - VMA_ASSERT(0 && "Not found."); + VMA_ASSERT((m_FreeSuballocationsBySize[index]->size == item->size) && "Not found."); } + VMA_ASSERT(0 && "Not found."); //VMA_HEAVY_ASSERT(ValidateFreeSuballocationList()); } @@ -10325,20 +7157,21 @@ bool VmaBlockMetadata_Generic::IsBufferImageGranularityConflictPossible( VkDeviceSize bufferImageGranularity, VmaSuballocationType& inOutPrevSuballocType) const { - if(bufferImageGranularity == 1 || IsEmpty()) + if (bufferImageGranularity == 1 || IsEmpty() || IsVirtual()) { return false; } VkDeviceSize minAlignment = VK_WHOLE_SIZE; bool typeConflictFound = false; - for(const auto& suballoc : m_Suballocations) + for (const auto& suballoc : m_Suballocations) { const VmaSuballocationType suballocType = suballoc.type; - if(suballocType != VMA_SUBALLOCATION_TYPE_FREE) + if (suballocType != VMA_SUBALLOCATION_TYPE_FREE) { - minAlignment = VMA_MIN(minAlignment, suballoc.hAllocation->GetAlignment()); - if(VmaIsBufferImageGranularityConflict(inOutPrevSuballocType, suballocType)) + VmaAllocation const alloc = (VmaAllocation)suballoc.userData; + minAlignment = VMA_MIN(minAlignment, alloc->GetAlignment()); + if (VmaIsBufferImageGranularityConflict(inOutPrevSuballocType, suballocType)) { typeConflictFound = true; } @@ -10348,26 +7181,204 @@ bool VmaBlockMetadata_Generic::IsBufferImageGranularityConflictPossible( return typeConflictFound || minAlignment >= bufferImageGranularity; } +#endif // _VMA_BLOCK_METADATA_GENERIC_FUNCTIONS +#endif // _VMA_BLOCK_METADATA_GENERIC -//////////////////////////////////////////////////////////////////////////////// -// class VmaBlockMetadata_Linear +#ifndef _VMA_BLOCK_METADATA_LINEAR +/* +Allocations and their references in internal data structure look like this: + +if(m_2ndVectorMode == SECOND_VECTOR_EMPTY): + + 0 +-------+ + | | + | | + | | + +-------+ + | Alloc | 1st[m_1stNullItemsBeginCount] + +-------+ + | Alloc | 1st[m_1stNullItemsBeginCount + 1] + +-------+ + | ... | + +-------+ + | Alloc | 1st[1st.size() - 1] + +-------+ + | | + | | + | | +GetSize() +-------+ + +if(m_2ndVectorMode == SECOND_VECTOR_RING_BUFFER): + + 0 +-------+ + | Alloc | 2nd[0] + +-------+ + | Alloc | 2nd[1] + +-------+ + | ... | + +-------+ + | Alloc | 2nd[2nd.size() - 1] + +-------+ + | | + | | + | | + +-------+ + | Alloc | 1st[m_1stNullItemsBeginCount] + +-------+ + | Alloc | 1st[m_1stNullItemsBeginCount + 1] + +-------+ + | ... | + +-------+ + | Alloc | 1st[1st.size() - 1] + +-------+ + | | +GetSize() +-------+ -VmaBlockMetadata_Linear::VmaBlockMetadata_Linear(VmaAllocator hAllocator) : - VmaBlockMetadata(hAllocator), +if(m_2ndVectorMode == SECOND_VECTOR_DOUBLE_STACK): + + 0 +-------+ + | | + | | + | | + +-------+ + | Alloc | 1st[m_1stNullItemsBeginCount] + +-------+ + | Alloc | 1st[m_1stNullItemsBeginCount + 1] + +-------+ + | ... | + +-------+ + | Alloc | 1st[1st.size() - 1] + +-------+ + | | + | | + | | + +-------+ + | Alloc | 2nd[2nd.size() - 1] + +-------+ + | ... | + +-------+ + | Alloc | 2nd[1] + +-------+ + | Alloc | 2nd[0] +GetSize() +-------+ + +*/ +class VmaBlockMetadata_Linear : public VmaBlockMetadata +{ + VMA_CLASS_NO_COPY(VmaBlockMetadata_Linear) +public: + VmaBlockMetadata_Linear(const VkAllocationCallbacks* pAllocationCallbacks, + VkDeviceSize bufferImageGranularity, bool isVirtual); + virtual ~VmaBlockMetadata_Linear() = default; + + VkDeviceSize GetSumFreeSize() const override { return m_SumFreeSize; } + bool IsEmpty() const override { return GetAllocationCount() == 0; } + VkDeviceSize GetAllocationOffset(VmaAllocHandle allocHandle) const override { return (VkDeviceSize)allocHandle - 1; }; + + void Init(VkDeviceSize size) override; + bool Validate() const override; + size_t GetAllocationCount() const override; + + void CalcAllocationStatInfo(VmaStatInfo& outInfo) const override; + void AddPoolStats(VmaPoolStats& inoutStats) const override; + +#if VMA_STATS_STRING_ENABLED + void PrintDetailedMap(class VmaJsonWriter& json) const override; +#endif + + bool CreateAllocationRequest( + VkDeviceSize allocSize, + VkDeviceSize allocAlignment, + bool upperAddress, + VmaSuballocationType allocType, + uint32_t strategy, + VmaAllocationRequest* pAllocationRequest) override; + + VkResult CheckCorruption(const void* pBlockData) override; + + void Alloc( + const VmaAllocationRequest& request, + VmaSuballocationType type, + void* userData) override; + + void Free(VmaAllocHandle allocHandle) override; + void GetAllocationInfo(VmaAllocHandle allocHandle, VmaVirtualAllocationInfo& outInfo) override; + void Clear() override; + void SetAllocationUserData(VmaAllocHandle allocHandle, void* userData) override; + void DebugLogAllAllocations() const override; + +private: + /* + There are two suballocation vectors, used in ping-pong way. + The one with index m_1stVectorIndex is called 1st. + The one with index (m_1stVectorIndex ^ 1) is called 2nd. + 2nd can be non-empty only when 1st is not empty. + When 2nd is not empty, m_2ndVectorMode indicates its mode of operation. + */ + typedef VmaVector<VmaSuballocation, VmaStlAllocator<VmaSuballocation>> SuballocationVectorType; + + enum SECOND_VECTOR_MODE + { + SECOND_VECTOR_EMPTY, + /* + Suballocations in 2nd vector are created later than the ones in 1st, but they + all have smaller offset. + */ + SECOND_VECTOR_RING_BUFFER, + /* + Suballocations in 2nd vector are upper side of double stack. + They all have offsets higher than those in 1st vector. + Top of this stack means smaller offsets, but higher indices in this vector. + */ + SECOND_VECTOR_DOUBLE_STACK, + }; + + VkDeviceSize m_SumFreeSize; + SuballocationVectorType m_Suballocations0, m_Suballocations1; + uint32_t m_1stVectorIndex; + SECOND_VECTOR_MODE m_2ndVectorMode; + // Number of items in 1st vector with hAllocation = null at the beginning. + size_t m_1stNullItemsBeginCount; + // Number of other items in 1st vector with hAllocation = null somewhere in the middle. + size_t m_1stNullItemsMiddleCount; + // Number of items in 2nd vector with hAllocation = null. + size_t m_2ndNullItemsCount; + + SuballocationVectorType& AccessSuballocations1st() { return m_1stVectorIndex ? m_Suballocations1 : m_Suballocations0; } + SuballocationVectorType& AccessSuballocations2nd() { return m_1stVectorIndex ? m_Suballocations0 : m_Suballocations1; } + const SuballocationVectorType& AccessSuballocations1st() const { return m_1stVectorIndex ? m_Suballocations1 : m_Suballocations0; } + const SuballocationVectorType& AccessSuballocations2nd() const { return m_1stVectorIndex ? m_Suballocations0 : m_Suballocations1; } + + VmaSuballocation& FindSuballocation(VkDeviceSize offset); + bool ShouldCompact1st() const; + void CleanupAfterFree(); + + bool CreateAllocationRequest_LowerAddress( + VkDeviceSize allocSize, + VkDeviceSize allocAlignment, + VmaSuballocationType allocType, + uint32_t strategy, + VmaAllocationRequest* pAllocationRequest); + bool CreateAllocationRequest_UpperAddress( + VkDeviceSize allocSize, + VkDeviceSize allocAlignment, + VmaSuballocationType allocType, + uint32_t strategy, + VmaAllocationRequest* pAllocationRequest); +}; + +#ifndef _VMA_BLOCK_METADATA_LINEAR_FUNCTIONS +VmaBlockMetadata_Linear::VmaBlockMetadata_Linear(const VkAllocationCallbacks* pAllocationCallbacks, + VkDeviceSize bufferImageGranularity, bool isVirtual) + : VmaBlockMetadata(pAllocationCallbacks, bufferImageGranularity, isVirtual), m_SumFreeSize(0), - m_Suballocations0(VmaStlAllocator<VmaSuballocation>(hAllocator->GetAllocationCallbacks())), - m_Suballocations1(VmaStlAllocator<VmaSuballocation>(hAllocator->GetAllocationCallbacks())), + m_Suballocations0(VmaStlAllocator<VmaSuballocation>(pAllocationCallbacks)), + m_Suballocations1(VmaStlAllocator<VmaSuballocation>(pAllocationCallbacks)), m_1stVectorIndex(0), m_2ndVectorMode(SECOND_VECTOR_EMPTY), m_1stNullItemsBeginCount(0), m_1stNullItemsMiddleCount(0), - m_2ndNullItemsCount(0) -{ -} - -VmaBlockMetadata_Linear::~VmaBlockMetadata_Linear() -{ -} + m_2ndNullItemsCount(0) {} void VmaBlockMetadata_Linear::Init(VkDeviceSize size) { @@ -10385,17 +7396,17 @@ bool VmaBlockMetadata_Linear::Validate() const suballocations2nd.empty() || m_2ndVectorMode != SECOND_VECTOR_RING_BUFFER); - if(!suballocations1st.empty()) + if (!suballocations1st.empty()) { // Null item at the beginning should be accounted into m_1stNullItemsBeginCount. - VMA_VALIDATE(suballocations1st[m_1stNullItemsBeginCount].hAllocation != VK_NULL_HANDLE); + VMA_VALIDATE(suballocations1st[m_1stNullItemsBeginCount].type != VMA_SUBALLOCATION_TYPE_FREE); // Null item at the end should be just pop_back(). - VMA_VALIDATE(suballocations1st.back().hAllocation != VK_NULL_HANDLE); + VMA_VALIDATE(suballocations1st.back().type != VMA_SUBALLOCATION_TYPE_FREE); } - if(!suballocations2nd.empty()) + if (!suballocations2nd.empty()) { // Null item at the end should be just pop_back(). - VMA_VALIDATE(suballocations2nd.back().hAllocation != VK_NULL_HANDLE); + VMA_VALIDATE(suballocations2nd.back().type != VMA_SUBALLOCATION_TYPE_FREE); } VMA_VALIDATE(m_1stNullItemsBeginCount + m_1stNullItemsMiddleCount <= suballocations1st.size()); @@ -10403,24 +7414,32 @@ bool VmaBlockMetadata_Linear::Validate() const VkDeviceSize sumUsedSize = 0; const size_t suballoc1stCount = suballocations1st.size(); - VkDeviceSize offset = VMA_DEBUG_MARGIN; + const VkDeviceSize debugMargin = GetDebugMargin(); + VkDeviceSize offset = 0; - if(m_2ndVectorMode == SECOND_VECTOR_RING_BUFFER) + if (m_2ndVectorMode == SECOND_VECTOR_RING_BUFFER) { const size_t suballoc2ndCount = suballocations2nd.size(); size_t nullItem2ndCount = 0; - for(size_t i = 0; i < suballoc2ndCount; ++i) + for (size_t i = 0; i < suballoc2ndCount; ++i) { const VmaSuballocation& suballoc = suballocations2nd[i]; const bool currFree = (suballoc.type == VMA_SUBALLOCATION_TYPE_FREE); - VMA_VALIDATE(currFree == (suballoc.hAllocation == VK_NULL_HANDLE)); + VmaAllocation const alloc = (VmaAllocation)suballoc.userData; + if (!IsVirtual()) + { + VMA_VALIDATE(currFree == (alloc == VK_NULL_HANDLE)); + } VMA_VALIDATE(suballoc.offset >= offset); - if(!currFree) + if (!currFree) { - VMA_VALIDATE(suballoc.hAllocation->GetOffset() == suballoc.offset); - VMA_VALIDATE(suballoc.hAllocation->GetSize() == suballoc.size); + if (!IsVirtual()) + { + VMA_VALIDATE((VkDeviceSize)alloc->GetAllocHandle() == suballoc.offset + 1); + VMA_VALIDATE(alloc->GetSize() == suballoc.size); + } sumUsedSize += suballoc.size; } else @@ -10428,34 +7447,41 @@ bool VmaBlockMetadata_Linear::Validate() const ++nullItem2ndCount; } - offset = suballoc.offset + suballoc.size + VMA_DEBUG_MARGIN; + offset = suballoc.offset + suballoc.size + debugMargin; } VMA_VALIDATE(nullItem2ndCount == m_2ndNullItemsCount); } - for(size_t i = 0; i < m_1stNullItemsBeginCount; ++i) + for (size_t i = 0; i < m_1stNullItemsBeginCount; ++i) { const VmaSuballocation& suballoc = suballocations1st[i]; VMA_VALIDATE(suballoc.type == VMA_SUBALLOCATION_TYPE_FREE && - suballoc.hAllocation == VK_NULL_HANDLE); + suballoc.userData == VMA_NULL); } size_t nullItem1stCount = m_1stNullItemsBeginCount; - for(size_t i = m_1stNullItemsBeginCount; i < suballoc1stCount; ++i) + for (size_t i = m_1stNullItemsBeginCount; i < suballoc1stCount; ++i) { const VmaSuballocation& suballoc = suballocations1st[i]; const bool currFree = (suballoc.type == VMA_SUBALLOCATION_TYPE_FREE); - VMA_VALIDATE(currFree == (suballoc.hAllocation == VK_NULL_HANDLE)); + VmaAllocation const alloc = (VmaAllocation)suballoc.userData; + if (!IsVirtual()) + { + VMA_VALIDATE(currFree == (alloc == VK_NULL_HANDLE)); + } VMA_VALIDATE(suballoc.offset >= offset); VMA_VALIDATE(i >= m_1stNullItemsBeginCount || currFree); - if(!currFree) + if (!currFree) { - VMA_VALIDATE(suballoc.hAllocation->GetOffset() == suballoc.offset); - VMA_VALIDATE(suballoc.hAllocation->GetSize() == suballoc.size); + if (!IsVirtual()) + { + VMA_VALIDATE((VkDeviceSize)alloc->GetAllocHandle() == suballoc.offset + 1); + VMA_VALIDATE(alloc->GetSize() == suballoc.size); + } sumUsedSize += suballoc.size; } else @@ -10463,26 +7489,33 @@ bool VmaBlockMetadata_Linear::Validate() const ++nullItem1stCount; } - offset = suballoc.offset + suballoc.size + VMA_DEBUG_MARGIN; + offset = suballoc.offset + suballoc.size + debugMargin; } VMA_VALIDATE(nullItem1stCount == m_1stNullItemsBeginCount + m_1stNullItemsMiddleCount); - if(m_2ndVectorMode == SECOND_VECTOR_DOUBLE_STACK) + if (m_2ndVectorMode == SECOND_VECTOR_DOUBLE_STACK) { const size_t suballoc2ndCount = suballocations2nd.size(); size_t nullItem2ndCount = 0; - for(size_t i = suballoc2ndCount; i--; ) + for (size_t i = suballoc2ndCount; i--; ) { const VmaSuballocation& suballoc = suballocations2nd[i]; const bool currFree = (suballoc.type == VMA_SUBALLOCATION_TYPE_FREE); - VMA_VALIDATE(currFree == (suballoc.hAllocation == VK_NULL_HANDLE)); + VmaAllocation const alloc = (VmaAllocation)suballoc.userData; + if (!IsVirtual()) + { + VMA_VALIDATE(currFree == (alloc == VK_NULL_HANDLE)); + } VMA_VALIDATE(suballoc.offset >= offset); - if(!currFree) + if (!currFree) { - VMA_VALIDATE(suballoc.hAllocation->GetOffset() == suballoc.offset); - VMA_VALIDATE(suballoc.hAllocation->GetSize() == suballoc.size); + if (!IsVirtual()) + { + VMA_VALIDATE((VkDeviceSize)alloc->GetAllocHandle() == suballoc.offset + 1); + VMA_VALIDATE(alloc->GetSize() == suballoc.size); + } sumUsedSize += suballoc.size; } else @@ -10490,7 +7523,7 @@ bool VmaBlockMetadata_Linear::Validate() const ++nullItem2ndCount; } - offset = suballoc.offset + suballoc.size + VMA_DEBUG_MARGIN; + offset = suballoc.offset + suballoc.size + debugMargin; } VMA_VALIDATE(nullItem2ndCount == m_2ndNullItemsCount); @@ -10504,74 +7537,10 @@ bool VmaBlockMetadata_Linear::Validate() const size_t VmaBlockMetadata_Linear::GetAllocationCount() const { - return AccessSuballocations1st().size() - (m_1stNullItemsBeginCount + m_1stNullItemsMiddleCount) + + return AccessSuballocations1st().size() - m_1stNullItemsBeginCount - m_1stNullItemsMiddleCount + AccessSuballocations2nd().size() - m_2ndNullItemsCount; } -VkDeviceSize VmaBlockMetadata_Linear::GetUnusedRangeSizeMax() const -{ - const VkDeviceSize size = GetSize(); - - /* - We don't consider gaps inside allocation vectors with freed allocations because - they are not suitable for reuse in linear allocator. We consider only space that - is available for new allocations. - */ - if(IsEmpty()) - { - return size; - } - - const SuballocationVectorType& suballocations1st = AccessSuballocations1st(); - - switch(m_2ndVectorMode) - { - case SECOND_VECTOR_EMPTY: - /* - Available space is after end of 1st, as well as before beginning of 1st (which - would make it a ring buffer). - */ - { - const size_t suballocations1stCount = suballocations1st.size(); - VMA_ASSERT(suballocations1stCount > m_1stNullItemsBeginCount); - const VmaSuballocation& firstSuballoc = suballocations1st[m_1stNullItemsBeginCount]; - const VmaSuballocation& lastSuballoc = suballocations1st[suballocations1stCount - 1]; - return VMA_MAX( - firstSuballoc.offset, - size - (lastSuballoc.offset + lastSuballoc.size)); - } - break; - - case SECOND_VECTOR_RING_BUFFER: - /* - Available space is only between end of 2nd and beginning of 1st. - */ - { - const SuballocationVectorType& suballocations2nd = AccessSuballocations2nd(); - const VmaSuballocation& lastSuballoc2nd = suballocations2nd.back(); - const VmaSuballocation& firstSuballoc1st = suballocations1st[m_1stNullItemsBeginCount]; - return firstSuballoc1st.offset - (lastSuballoc2nd.offset + lastSuballoc2nd.size); - } - break; - - case SECOND_VECTOR_DOUBLE_STACK: - /* - Available space is only between end of 1st and top of 2nd. - */ - { - const SuballocationVectorType& suballocations2nd = AccessSuballocations2nd(); - const VmaSuballocation& topSuballoc2nd = suballocations2nd.back(); - const VmaSuballocation& lastSuballoc1st = suballocations1st.back(); - return topSuballoc2nd.offset - (lastSuballoc1st.offset + lastSuballoc1st.size); - } - break; - - default: - VMA_ASSERT(0); - return 0; - } -} - void VmaBlockMetadata_Linear::CalcAllocationStatInfo(VmaStatInfo& outInfo) const { const VkDeviceSize size = GetSize(); @@ -10580,51 +7549,40 @@ void VmaBlockMetadata_Linear::CalcAllocationStatInfo(VmaStatInfo& outInfo) const const size_t suballoc1stCount = suballocations1st.size(); const size_t suballoc2ndCount = suballocations2nd.size(); + VmaInitStatInfo(outInfo); outInfo.blockCount = 1; - outInfo.allocationCount = (uint32_t)GetAllocationCount(); - outInfo.unusedRangeCount = 0; - outInfo.usedBytes = 0; - outInfo.allocationSizeMin = UINT64_MAX; - outInfo.allocationSizeMax = 0; - outInfo.unusedRangeSizeMin = UINT64_MAX; - outInfo.unusedRangeSizeMax = 0; VkDeviceSize lastOffset = 0; - if(m_2ndVectorMode == SECOND_VECTOR_RING_BUFFER) + if (m_2ndVectorMode == SECOND_VECTOR_RING_BUFFER) { const VkDeviceSize freeSpace2ndTo1stEnd = suballocations1st[m_1stNullItemsBeginCount].offset; size_t nextAlloc2ndIndex = 0; - while(lastOffset < freeSpace2ndTo1stEnd) + while (lastOffset < freeSpace2ndTo1stEnd) { // Find next non-null allocation or move nextAllocIndex to the end. - while(nextAlloc2ndIndex < suballoc2ndCount && - suballocations2nd[nextAlloc2ndIndex].hAllocation == VK_NULL_HANDLE) + while (nextAlloc2ndIndex < suballoc2ndCount && + suballocations2nd[nextAlloc2ndIndex].userData == VMA_NULL) { ++nextAlloc2ndIndex; } // Found non-null allocation. - if(nextAlloc2ndIndex < suballoc2ndCount) + if (nextAlloc2ndIndex < suballoc2ndCount) { const VmaSuballocation& suballoc = suballocations2nd[nextAlloc2ndIndex]; // 1. Process free space before this allocation. - if(lastOffset < suballoc.offset) + if (lastOffset < suballoc.offset) { // There is free space from lastOffset to suballoc.offset. const VkDeviceSize unusedRangeSize = suballoc.offset - lastOffset; - ++outInfo.unusedRangeCount; - outInfo.unusedBytes += unusedRangeSize; - outInfo.unusedRangeSizeMin = VMA_MIN(outInfo.unusedRangeSizeMin, unusedRangeSize); - outInfo.unusedRangeSizeMax = VMA_MIN(outInfo.unusedRangeSizeMax, unusedRangeSize); + VmaAddStatInfoUnusedRange(outInfo, unusedRangeSize); } // 2. Process this allocation. // There is allocation with suballoc.offset, suballoc.size. - outInfo.usedBytes += suballoc.size; - outInfo.allocationSizeMin = VMA_MIN(outInfo.allocationSizeMin, suballoc.size); - outInfo.allocationSizeMax = VMA_MIN(outInfo.allocationSizeMax, suballoc.size); + VmaAddStatInfoAllocation(outInfo, suballoc.size); // 3. Prepare for next iteration. lastOffset = suballoc.offset + suballoc.size; @@ -10634,14 +7592,11 @@ void VmaBlockMetadata_Linear::CalcAllocationStatInfo(VmaStatInfo& outInfo) const else { // There is free space from lastOffset to freeSpace2ndTo1stEnd. - if(lastOffset < freeSpace2ndTo1stEnd) + if (lastOffset < freeSpace2ndTo1stEnd) { const VkDeviceSize unusedRangeSize = freeSpace2ndTo1stEnd - lastOffset; - ++outInfo.unusedRangeCount; - outInfo.unusedBytes += unusedRangeSize; - outInfo.unusedRangeSizeMin = VMA_MIN(outInfo.unusedRangeSizeMin, unusedRangeSize); - outInfo.unusedRangeSizeMax = VMA_MIN(outInfo.unusedRangeSizeMax, unusedRangeSize); - } + VmaAddStatInfoUnusedRange(outInfo, unusedRangeSize); + } // End of loop. lastOffset = freeSpace2ndTo1stEnd; @@ -10652,36 +7607,31 @@ void VmaBlockMetadata_Linear::CalcAllocationStatInfo(VmaStatInfo& outInfo) const size_t nextAlloc1stIndex = m_1stNullItemsBeginCount; const VkDeviceSize freeSpace1stTo2ndEnd = m_2ndVectorMode == SECOND_VECTOR_DOUBLE_STACK ? suballocations2nd.back().offset : size; - while(lastOffset < freeSpace1stTo2ndEnd) + while (lastOffset < freeSpace1stTo2ndEnd) { // Find next non-null allocation or move nextAllocIndex to the end. - while(nextAlloc1stIndex < suballoc1stCount && - suballocations1st[nextAlloc1stIndex].hAllocation == VK_NULL_HANDLE) + while (nextAlloc1stIndex < suballoc1stCount && + suballocations1st[nextAlloc1stIndex].userData == VMA_NULL) { ++nextAlloc1stIndex; } // Found non-null allocation. - if(nextAlloc1stIndex < suballoc1stCount) + if (nextAlloc1stIndex < suballoc1stCount) { const VmaSuballocation& suballoc = suballocations1st[nextAlloc1stIndex]; // 1. Process free space before this allocation. - if(lastOffset < suballoc.offset) + if (lastOffset < suballoc.offset) { // There is free space from lastOffset to suballoc.offset. const VkDeviceSize unusedRangeSize = suballoc.offset - lastOffset; - ++outInfo.unusedRangeCount; - outInfo.unusedBytes += unusedRangeSize; - outInfo.unusedRangeSizeMin = VMA_MIN(outInfo.unusedRangeSizeMin, unusedRangeSize); - outInfo.unusedRangeSizeMax = VMA_MIN(outInfo.unusedRangeSizeMax, unusedRangeSize); + VmaAddStatInfoUnusedRange(outInfo, unusedRangeSize); } // 2. Process this allocation. // There is allocation with suballoc.offset, suballoc.size. - outInfo.usedBytes += suballoc.size; - outInfo.allocationSizeMin = VMA_MIN(outInfo.allocationSizeMin, suballoc.size); - outInfo.allocationSizeMax = VMA_MIN(outInfo.allocationSizeMax, suballoc.size); + VmaAddStatInfoAllocation(outInfo, suballoc.size); // 3. Prepare for next iteration. lastOffset = suballoc.offset + suballoc.size; @@ -10691,53 +7641,45 @@ void VmaBlockMetadata_Linear::CalcAllocationStatInfo(VmaStatInfo& outInfo) const else { // There is free space from lastOffset to freeSpace1stTo2ndEnd. - if(lastOffset < freeSpace1stTo2ndEnd) + if (lastOffset < freeSpace1stTo2ndEnd) { const VkDeviceSize unusedRangeSize = freeSpace1stTo2ndEnd - lastOffset; - ++outInfo.unusedRangeCount; - outInfo.unusedBytes += unusedRangeSize; - outInfo.unusedRangeSizeMin = VMA_MIN(outInfo.unusedRangeSizeMin, unusedRangeSize); - outInfo.unusedRangeSizeMax = VMA_MIN(outInfo.unusedRangeSizeMax, unusedRangeSize); - } + VmaAddStatInfoUnusedRange(outInfo, unusedRangeSize); + } // End of loop. lastOffset = freeSpace1stTo2ndEnd; } } - if(m_2ndVectorMode == SECOND_VECTOR_DOUBLE_STACK) + if (m_2ndVectorMode == SECOND_VECTOR_DOUBLE_STACK) { size_t nextAlloc2ndIndex = suballocations2nd.size() - 1; - while(lastOffset < size) + while (lastOffset < size) { // Find next non-null allocation or move nextAllocIndex to the end. - while(nextAlloc2ndIndex != SIZE_MAX && - suballocations2nd[nextAlloc2ndIndex].hAllocation == VK_NULL_HANDLE) + while (nextAlloc2ndIndex != SIZE_MAX && + suballocations2nd[nextAlloc2ndIndex].userData == VMA_NULL) { --nextAlloc2ndIndex; } // Found non-null allocation. - if(nextAlloc2ndIndex != SIZE_MAX) + if (nextAlloc2ndIndex != SIZE_MAX) { const VmaSuballocation& suballoc = suballocations2nd[nextAlloc2ndIndex]; // 1. Process free space before this allocation. - if(lastOffset < suballoc.offset) + if (lastOffset < suballoc.offset) { // There is free space from lastOffset to suballoc.offset. const VkDeviceSize unusedRangeSize = suballoc.offset - lastOffset; - ++outInfo.unusedRangeCount; - outInfo.unusedBytes += unusedRangeSize; - outInfo.unusedRangeSizeMin = VMA_MIN(outInfo.unusedRangeSizeMin, unusedRangeSize); - outInfo.unusedRangeSizeMax = VMA_MIN(outInfo.unusedRangeSizeMax, unusedRangeSize); + VmaAddStatInfoUnusedRange(outInfo, unusedRangeSize); } // 2. Process this allocation. // There is allocation with suballoc.offset, suballoc.size. - outInfo.usedBytes += suballoc.size; - outInfo.allocationSizeMin = VMA_MIN(outInfo.allocationSizeMin, suballoc.size); - outInfo.allocationSizeMax = VMA_MIN(outInfo.allocationSizeMax, suballoc.size); + VmaAddStatInfoAllocation(outInfo, suballoc.size); // 3. Prepare for next iteration. lastOffset = suballoc.offset + suballoc.size; @@ -10747,14 +7689,11 @@ void VmaBlockMetadata_Linear::CalcAllocationStatInfo(VmaStatInfo& outInfo) const else { // There is free space from lastOffset to size. - if(lastOffset < size) + if (lastOffset < size) { const VkDeviceSize unusedRangeSize = size - lastOffset; - ++outInfo.unusedRangeCount; - outInfo.unusedBytes += unusedRangeSize; - outInfo.unusedRangeSizeMin = VMA_MIN(outInfo.unusedRangeSizeMin, unusedRangeSize); - outInfo.unusedRangeSizeMax = VMA_MIN(outInfo.unusedRangeSizeMax, unusedRangeSize); - } + VmaAddStatInfoUnusedRange(outInfo, unusedRangeSize); + } // End of loop. lastOffset = size; @@ -10777,32 +7716,31 @@ void VmaBlockMetadata_Linear::AddPoolStats(VmaPoolStats& inoutStats) const VkDeviceSize lastOffset = 0; - if(m_2ndVectorMode == SECOND_VECTOR_RING_BUFFER) + if (m_2ndVectorMode == SECOND_VECTOR_RING_BUFFER) { const VkDeviceSize freeSpace2ndTo1stEnd = suballocations1st[m_1stNullItemsBeginCount].offset; size_t nextAlloc2ndIndex = m_1stNullItemsBeginCount; - while(lastOffset < freeSpace2ndTo1stEnd) + while (lastOffset < freeSpace2ndTo1stEnd) { // Find next non-null allocation or move nextAlloc2ndIndex to the end. - while(nextAlloc2ndIndex < suballoc2ndCount && - suballocations2nd[nextAlloc2ndIndex].hAllocation == VK_NULL_HANDLE) + while (nextAlloc2ndIndex < suballoc2ndCount && + suballocations2nd[nextAlloc2ndIndex].userData == VMA_NULL) { ++nextAlloc2ndIndex; } // Found non-null allocation. - if(nextAlloc2ndIndex < suballoc2ndCount) + if (nextAlloc2ndIndex < suballoc2ndCount) { const VmaSuballocation& suballoc = suballocations2nd[nextAlloc2ndIndex]; // 1. Process free space before this allocation. - if(lastOffset < suballoc.offset) + if (lastOffset < suballoc.offset) { // There is free space from lastOffset to suballoc.offset. const VkDeviceSize unusedRangeSize = suballoc.offset - lastOffset; inoutStats.unusedSize += unusedRangeSize; ++inoutStats.unusedRangeCount; - inoutStats.unusedRangeSizeMax = VMA_MAX(inoutStats.unusedRangeSizeMax, unusedRangeSize); } // 2. Process this allocation. @@ -10816,13 +7754,12 @@ void VmaBlockMetadata_Linear::AddPoolStats(VmaPoolStats& inoutStats) const // We are at the end. else { - if(lastOffset < freeSpace2ndTo1stEnd) + if (lastOffset < freeSpace2ndTo1stEnd) { // There is free space from lastOffset to freeSpace2ndTo1stEnd. const VkDeviceSize unusedRangeSize = freeSpace2ndTo1stEnd - lastOffset; inoutStats.unusedSize += unusedRangeSize; ++inoutStats.unusedRangeCount; - inoutStats.unusedRangeSizeMax = VMA_MAX(inoutStats.unusedRangeSizeMax, unusedRangeSize); } // End of loop. @@ -10834,28 +7771,27 @@ void VmaBlockMetadata_Linear::AddPoolStats(VmaPoolStats& inoutStats) const size_t nextAlloc1stIndex = m_1stNullItemsBeginCount; const VkDeviceSize freeSpace1stTo2ndEnd = m_2ndVectorMode == SECOND_VECTOR_DOUBLE_STACK ? suballocations2nd.back().offset : size; - while(lastOffset < freeSpace1stTo2ndEnd) + while (lastOffset < freeSpace1stTo2ndEnd) { // Find next non-null allocation or move nextAllocIndex to the end. - while(nextAlloc1stIndex < suballoc1stCount && - suballocations1st[nextAlloc1stIndex].hAllocation == VK_NULL_HANDLE) + while (nextAlloc1stIndex < suballoc1stCount && + suballocations1st[nextAlloc1stIndex].userData == VMA_NULL) { ++nextAlloc1stIndex; } // Found non-null allocation. - if(nextAlloc1stIndex < suballoc1stCount) + if (nextAlloc1stIndex < suballoc1stCount) { const VmaSuballocation& suballoc = suballocations1st[nextAlloc1stIndex]; // 1. Process free space before this allocation. - if(lastOffset < suballoc.offset) + if (lastOffset < suballoc.offset) { // There is free space from lastOffset to suballoc.offset. const VkDeviceSize unusedRangeSize = suballoc.offset - lastOffset; inoutStats.unusedSize += unusedRangeSize; ++inoutStats.unusedRangeCount; - inoutStats.unusedRangeSizeMax = VMA_MAX(inoutStats.unusedRangeSizeMax, unusedRangeSize); } // 2. Process this allocation. @@ -10869,13 +7805,12 @@ void VmaBlockMetadata_Linear::AddPoolStats(VmaPoolStats& inoutStats) const // We are at the end. else { - if(lastOffset < freeSpace1stTo2ndEnd) + if (lastOffset < freeSpace1stTo2ndEnd) { // There is free space from lastOffset to freeSpace1stTo2ndEnd. const VkDeviceSize unusedRangeSize = freeSpace1stTo2ndEnd - lastOffset; inoutStats.unusedSize += unusedRangeSize; ++inoutStats.unusedRangeCount; - inoutStats.unusedRangeSizeMax = VMA_MAX(inoutStats.unusedRangeSizeMax, unusedRangeSize); } // End of loop. @@ -10883,31 +7818,30 @@ void VmaBlockMetadata_Linear::AddPoolStats(VmaPoolStats& inoutStats) const } } - if(m_2ndVectorMode == SECOND_VECTOR_DOUBLE_STACK) + if (m_2ndVectorMode == SECOND_VECTOR_DOUBLE_STACK) { size_t nextAlloc2ndIndex = suballocations2nd.size() - 1; - while(lastOffset < size) + while (lastOffset < size) { // Find next non-null allocation or move nextAlloc2ndIndex to the end. - while(nextAlloc2ndIndex != SIZE_MAX && - suballocations2nd[nextAlloc2ndIndex].hAllocation == VK_NULL_HANDLE) + while (nextAlloc2ndIndex != SIZE_MAX && + suballocations2nd[nextAlloc2ndIndex].userData == VMA_NULL) { --nextAlloc2ndIndex; } // Found non-null allocation. - if(nextAlloc2ndIndex != SIZE_MAX) + if (nextAlloc2ndIndex != SIZE_MAX) { const VmaSuballocation& suballoc = suballocations2nd[nextAlloc2ndIndex]; // 1. Process free space before this allocation. - if(lastOffset < suballoc.offset) + if (lastOffset < suballoc.offset) { // There is free space from lastOffset to suballoc.offset. const VkDeviceSize unusedRangeSize = suballoc.offset - lastOffset; inoutStats.unusedSize += unusedRangeSize; ++inoutStats.unusedRangeCount; - inoutStats.unusedRangeSizeMax = VMA_MAX(inoutStats.unusedRangeSizeMax, unusedRangeSize); } // 2. Process this allocation. @@ -10921,13 +7855,12 @@ void VmaBlockMetadata_Linear::AddPoolStats(VmaPoolStats& inoutStats) const // We are at the end. else { - if(lastOffset < size) + if (lastOffset < size) { // There is free space from lastOffset to size. const VkDeviceSize unusedRangeSize = size - lastOffset; inoutStats.unusedSize += unusedRangeSize; ++inoutStats.unusedRangeCount; - inoutStats.unusedRangeSizeMax = VMA_MAX(inoutStats.unusedRangeSizeMax, unusedRangeSize); } // End of loop. @@ -10954,26 +7887,26 @@ void VmaBlockMetadata_Linear::PrintDetailedMap(class VmaJsonWriter& json) const VkDeviceSize lastOffset = 0; size_t alloc2ndCount = 0; - if(m_2ndVectorMode == SECOND_VECTOR_RING_BUFFER) + if (m_2ndVectorMode == SECOND_VECTOR_RING_BUFFER) { const VkDeviceSize freeSpace2ndTo1stEnd = suballocations1st[m_1stNullItemsBeginCount].offset; size_t nextAlloc2ndIndex = 0; - while(lastOffset < freeSpace2ndTo1stEnd) + while (lastOffset < freeSpace2ndTo1stEnd) { // Find next non-null allocation or move nextAlloc2ndIndex to the end. - while(nextAlloc2ndIndex < suballoc2ndCount && - suballocations2nd[nextAlloc2ndIndex].hAllocation == VK_NULL_HANDLE) + while (nextAlloc2ndIndex < suballoc2ndCount && + suballocations2nd[nextAlloc2ndIndex].userData == VMA_NULL) { ++nextAlloc2ndIndex; } // Found non-null allocation. - if(nextAlloc2ndIndex < suballoc2ndCount) + if (nextAlloc2ndIndex < suballoc2ndCount) { const VmaSuballocation& suballoc = suballocations2nd[nextAlloc2ndIndex]; // 1. Process free space before this allocation. - if(lastOffset < suballoc.offset) + if (lastOffset < suballoc.offset) { // There is free space from lastOffset to suballoc.offset. ++unusedRangeCount; @@ -10991,7 +7924,7 @@ void VmaBlockMetadata_Linear::PrintDetailedMap(class VmaJsonWriter& json) const // We are at the end. else { - if(lastOffset < freeSpace2ndTo1stEnd) + if (lastOffset < freeSpace2ndTo1stEnd) { // There is free space from lastOffset to freeSpace2ndTo1stEnd. ++unusedRangeCount; @@ -11007,22 +7940,22 @@ void VmaBlockMetadata_Linear::PrintDetailedMap(class VmaJsonWriter& json) const size_t alloc1stCount = 0; const VkDeviceSize freeSpace1stTo2ndEnd = m_2ndVectorMode == SECOND_VECTOR_DOUBLE_STACK ? suballocations2nd.back().offset : size; - while(lastOffset < freeSpace1stTo2ndEnd) + while (lastOffset < freeSpace1stTo2ndEnd) { // Find next non-null allocation or move nextAllocIndex to the end. - while(nextAlloc1stIndex < suballoc1stCount && - suballocations1st[nextAlloc1stIndex].hAllocation == VK_NULL_HANDLE) + while (nextAlloc1stIndex < suballoc1stCount && + suballocations1st[nextAlloc1stIndex].userData == VMA_NULL) { ++nextAlloc1stIndex; } // Found non-null allocation. - if(nextAlloc1stIndex < suballoc1stCount) + if (nextAlloc1stIndex < suballoc1stCount) { const VmaSuballocation& suballoc = suballocations1st[nextAlloc1stIndex]; // 1. Process free space before this allocation. - if(lastOffset < suballoc.offset) + if (lastOffset < suballoc.offset) { // There is free space from lastOffset to suballoc.offset. ++unusedRangeCount; @@ -11040,7 +7973,7 @@ void VmaBlockMetadata_Linear::PrintDetailedMap(class VmaJsonWriter& json) const // We are at the end. else { - if(lastOffset < size) + if (lastOffset < size) { // There is free space from lastOffset to freeSpace1stTo2ndEnd. ++unusedRangeCount; @@ -11051,25 +7984,25 @@ void VmaBlockMetadata_Linear::PrintDetailedMap(class VmaJsonWriter& json) const } } - if(m_2ndVectorMode == SECOND_VECTOR_DOUBLE_STACK) + if (m_2ndVectorMode == SECOND_VECTOR_DOUBLE_STACK) { size_t nextAlloc2ndIndex = suballocations2nd.size() - 1; - while(lastOffset < size) + while (lastOffset < size) { // Find next non-null allocation or move nextAlloc2ndIndex to the end. - while(nextAlloc2ndIndex != SIZE_MAX && - suballocations2nd[nextAlloc2ndIndex].hAllocation == VK_NULL_HANDLE) + while (nextAlloc2ndIndex != SIZE_MAX && + suballocations2nd[nextAlloc2ndIndex].userData == VMA_NULL) { --nextAlloc2ndIndex; } // Found non-null allocation. - if(nextAlloc2ndIndex != SIZE_MAX) + if (nextAlloc2ndIndex != SIZE_MAX) { const VmaSuballocation& suballoc = suballocations2nd[nextAlloc2ndIndex]; // 1. Process free space before this allocation. - if(lastOffset < suballoc.offset) + if (lastOffset < suballoc.offset) { // There is free space from lastOffset to suballoc.offset. ++unusedRangeCount; @@ -11087,7 +8020,7 @@ void VmaBlockMetadata_Linear::PrintDetailedMap(class VmaJsonWriter& json) const // We are at the end. else { - if(lastOffset < size) + if (lastOffset < size) { // There is free space from lastOffset to size. ++unusedRangeCount; @@ -11105,26 +8038,26 @@ void VmaBlockMetadata_Linear::PrintDetailedMap(class VmaJsonWriter& json) const // SECOND PASS lastOffset = 0; - if(m_2ndVectorMode == SECOND_VECTOR_RING_BUFFER) + if (m_2ndVectorMode == SECOND_VECTOR_RING_BUFFER) { const VkDeviceSize freeSpace2ndTo1stEnd = suballocations1st[m_1stNullItemsBeginCount].offset; size_t nextAlloc2ndIndex = 0; - while(lastOffset < freeSpace2ndTo1stEnd) + while (lastOffset < freeSpace2ndTo1stEnd) { // Find next non-null allocation or move nextAlloc2ndIndex to the end. - while(nextAlloc2ndIndex < suballoc2ndCount && - suballocations2nd[nextAlloc2ndIndex].hAllocation == VK_NULL_HANDLE) + while (nextAlloc2ndIndex < suballoc2ndCount && + suballocations2nd[nextAlloc2ndIndex].userData == VMA_NULL) { ++nextAlloc2ndIndex; } // Found non-null allocation. - if(nextAlloc2ndIndex < suballoc2ndCount) + if (nextAlloc2ndIndex < suballoc2ndCount) { const VmaSuballocation& suballoc = suballocations2nd[nextAlloc2ndIndex]; // 1. Process free space before this allocation. - if(lastOffset < suballoc.offset) + if (lastOffset < suballoc.offset) { // There is free space from lastOffset to suballoc.offset. const VkDeviceSize unusedRangeSize = suballoc.offset - lastOffset; @@ -11133,7 +8066,7 @@ void VmaBlockMetadata_Linear::PrintDetailedMap(class VmaJsonWriter& json) const // 2. Process this allocation. // There is allocation with suballoc.offset, suballoc.size. - PrintDetailedMap_Allocation(json, suballoc.offset, suballoc.hAllocation); + PrintDetailedMap_Allocation(json, suballoc.offset, suballoc.size, suballoc.userData); // 3. Prepare for next iteration. lastOffset = suballoc.offset + suballoc.size; @@ -11142,7 +8075,7 @@ void VmaBlockMetadata_Linear::PrintDetailedMap(class VmaJsonWriter& json) const // We are at the end. else { - if(lastOffset < freeSpace2ndTo1stEnd) + if (lastOffset < freeSpace2ndTo1stEnd) { // There is free space from lastOffset to freeSpace2ndTo1stEnd. const VkDeviceSize unusedRangeSize = freeSpace2ndTo1stEnd - lastOffset; @@ -11156,22 +8089,22 @@ void VmaBlockMetadata_Linear::PrintDetailedMap(class VmaJsonWriter& json) const } nextAlloc1stIndex = m_1stNullItemsBeginCount; - while(lastOffset < freeSpace1stTo2ndEnd) + while (lastOffset < freeSpace1stTo2ndEnd) { // Find next non-null allocation or move nextAllocIndex to the end. - while(nextAlloc1stIndex < suballoc1stCount && - suballocations1st[nextAlloc1stIndex].hAllocation == VK_NULL_HANDLE) + while (nextAlloc1stIndex < suballoc1stCount && + suballocations1st[nextAlloc1stIndex].userData == VMA_NULL) { ++nextAlloc1stIndex; } // Found non-null allocation. - if(nextAlloc1stIndex < suballoc1stCount) + if (nextAlloc1stIndex < suballoc1stCount) { const VmaSuballocation& suballoc = suballocations1st[nextAlloc1stIndex]; // 1. Process free space before this allocation. - if(lastOffset < suballoc.offset) + if (lastOffset < suballoc.offset) { // There is free space from lastOffset to suballoc.offset. const VkDeviceSize unusedRangeSize = suballoc.offset - lastOffset; @@ -11180,7 +8113,7 @@ void VmaBlockMetadata_Linear::PrintDetailedMap(class VmaJsonWriter& json) const // 2. Process this allocation. // There is allocation with suballoc.offset, suballoc.size. - PrintDetailedMap_Allocation(json, suballoc.offset, suballoc.hAllocation); + PrintDetailedMap_Allocation(json, suballoc.offset, suballoc.size, suballoc.userData); // 3. Prepare for next iteration. lastOffset = suballoc.offset + suballoc.size; @@ -11189,7 +8122,7 @@ void VmaBlockMetadata_Linear::PrintDetailedMap(class VmaJsonWriter& json) const // We are at the end. else { - if(lastOffset < freeSpace1stTo2ndEnd) + if (lastOffset < freeSpace1stTo2ndEnd) { // There is free space from lastOffset to freeSpace1stTo2ndEnd. const VkDeviceSize unusedRangeSize = freeSpace1stTo2ndEnd - lastOffset; @@ -11201,25 +8134,25 @@ void VmaBlockMetadata_Linear::PrintDetailedMap(class VmaJsonWriter& json) const } } - if(m_2ndVectorMode == SECOND_VECTOR_DOUBLE_STACK) + if (m_2ndVectorMode == SECOND_VECTOR_DOUBLE_STACK) { size_t nextAlloc2ndIndex = suballocations2nd.size() - 1; - while(lastOffset < size) + while (lastOffset < size) { // Find next non-null allocation or move nextAlloc2ndIndex to the end. - while(nextAlloc2ndIndex != SIZE_MAX && - suballocations2nd[nextAlloc2ndIndex].hAllocation == VK_NULL_HANDLE) + while (nextAlloc2ndIndex != SIZE_MAX && + suballocations2nd[nextAlloc2ndIndex].userData == VMA_NULL) { --nextAlloc2ndIndex; } // Found non-null allocation. - if(nextAlloc2ndIndex != SIZE_MAX) + if (nextAlloc2ndIndex != SIZE_MAX) { const VmaSuballocation& suballoc = suballocations2nd[nextAlloc2ndIndex]; // 1. Process free space before this allocation. - if(lastOffset < suballoc.offset) + if (lastOffset < suballoc.offset) { // There is free space from lastOffset to suballoc.offset. const VkDeviceSize unusedRangeSize = suballoc.offset - lastOffset; @@ -11228,7 +8161,7 @@ void VmaBlockMetadata_Linear::PrintDetailedMap(class VmaJsonWriter& json) const // 2. Process this allocation. // There is allocation with suballoc.offset, suballoc.size. - PrintDetailedMap_Allocation(json, suballoc.offset, suballoc.hAllocation); + PrintDetailedMap_Allocation(json, suballoc.offset, suballoc.size, suballoc.userData); // 3. Prepare for next iteration. lastOffset = suballoc.offset + suballoc.size; @@ -11237,7 +8170,7 @@ void VmaBlockMetadata_Linear::PrintDetailedMap(class VmaJsonWriter& json) const // We are at the end. else { - if(lastOffset < size) + if (lastOffset < size) { // There is free space from lastOffset to size. const VkDeviceSize unusedRangeSize = size - lastOffset; @@ -11252,17 +8185,13 @@ void VmaBlockMetadata_Linear::PrintDetailedMap(class VmaJsonWriter& json) const PrintDetailedMap_End(json); } -#endif // #if VMA_STATS_STRING_ENABLED +#endif // VMA_STATS_STRING_ENABLED bool VmaBlockMetadata_Linear::CreateAllocationRequest( - uint32_t currentFrameIndex, - uint32_t frameInUseCount, - VkDeviceSize bufferImageGranularity, VkDeviceSize allocSize, VkDeviceSize allocAlignment, bool upperAddress, VmaSuballocationType allocType, - bool canMakeOtherLost, uint32_t strategy, VmaAllocationRequest* pAllocationRequest) { @@ -11270,184 +8199,434 @@ bool VmaBlockMetadata_Linear::CreateAllocationRequest( VMA_ASSERT(allocType != VMA_SUBALLOCATION_TYPE_FREE); VMA_ASSERT(pAllocationRequest != VMA_NULL); VMA_HEAVY_ASSERT(Validate()); + pAllocationRequest->size = allocSize; return upperAddress ? CreateAllocationRequest_UpperAddress( - currentFrameIndex, frameInUseCount, bufferImageGranularity, - allocSize, allocAlignment, allocType, canMakeOtherLost, strategy, pAllocationRequest) : + allocSize, allocAlignment, allocType, strategy, pAllocationRequest) : CreateAllocationRequest_LowerAddress( - currentFrameIndex, frameInUseCount, bufferImageGranularity, - allocSize, allocAlignment, allocType, canMakeOtherLost, strategy, pAllocationRequest); + allocSize, allocAlignment, allocType, strategy, pAllocationRequest); } -bool VmaBlockMetadata_Linear::CreateAllocationRequest_UpperAddress( - uint32_t currentFrameIndex, - uint32_t frameInUseCount, - VkDeviceSize bufferImageGranularity, - VkDeviceSize allocSize, - VkDeviceSize allocAlignment, - VmaSuballocationType allocType, - bool canMakeOtherLost, - uint32_t strategy, - VmaAllocationRequest* pAllocationRequest) +VkResult VmaBlockMetadata_Linear::CheckCorruption(const void* pBlockData) { - const VkDeviceSize size = GetSize(); + VMA_ASSERT(!IsVirtual()); SuballocationVectorType& suballocations1st = AccessSuballocations1st(); + for (size_t i = m_1stNullItemsBeginCount, count = suballocations1st.size(); i < count; ++i) + { + const VmaSuballocation& suballoc = suballocations1st[i]; + if (suballoc.type != VMA_SUBALLOCATION_TYPE_FREE) + { + if (!VmaValidateMagicValue(pBlockData, suballoc.offset + suballoc.size)) + { + VMA_ASSERT(0 && "MEMORY CORRUPTION DETECTED AFTER VALIDATED ALLOCATION!"); + return VK_ERROR_UNKNOWN; + } + } + } + SuballocationVectorType& suballocations2nd = AccessSuballocations2nd(); + for (size_t i = 0, count = suballocations2nd.size(); i < count; ++i) + { + const VmaSuballocation& suballoc = suballocations2nd[i]; + if (suballoc.type != VMA_SUBALLOCATION_TYPE_FREE) + { + if (!VmaValidateMagicValue(pBlockData, suballoc.offset + suballoc.size)) + { + VMA_ASSERT(0 && "MEMORY CORRUPTION DETECTED AFTER VALIDATED ALLOCATION!"); + return VK_ERROR_UNKNOWN; + } + } + } + + return VK_SUCCESS; +} - if(m_2ndVectorMode == SECOND_VECTOR_RING_BUFFER) +void VmaBlockMetadata_Linear::Alloc( + const VmaAllocationRequest& request, + VmaSuballocationType type, + void* userData) +{ + const VkDeviceSize offset = (VkDeviceSize)request.allocHandle - 1; + const VmaSuballocation newSuballoc = { offset, request.size, userData, type }; + + switch (request.type) { - VMA_ASSERT(0 && "Trying to use pool with linear algorithm as double stack, while it is already being used as ring buffer."); - return false; + case VmaAllocationRequestType::UpperAddress: + { + VMA_ASSERT(m_2ndVectorMode != SECOND_VECTOR_RING_BUFFER && + "CRITICAL ERROR: Trying to use linear allocator as double stack while it was already used as ring buffer."); + SuballocationVectorType& suballocations2nd = AccessSuballocations2nd(); + suballocations2nd.push_back(newSuballoc); + m_2ndVectorMode = SECOND_VECTOR_DOUBLE_STACK; } + break; + case VmaAllocationRequestType::EndOf1st: + { + SuballocationVectorType& suballocations1st = AccessSuballocations1st(); - // Try to allocate before 2nd.back(), or end of block if 2nd.empty(). - if(allocSize > size) + VMA_ASSERT(suballocations1st.empty() || + offset >= suballocations1st.back().offset + suballocations1st.back().size); + // Check if it fits before the end of the block. + VMA_ASSERT(offset + request.size <= GetSize()); + + suballocations1st.push_back(newSuballoc); + } + break; + case VmaAllocationRequestType::EndOf2nd: { - return false; + SuballocationVectorType& suballocations1st = AccessSuballocations1st(); + // New allocation at the end of 2-part ring buffer, so before first allocation from 1st vector. + VMA_ASSERT(!suballocations1st.empty() && + offset + request.size <= suballocations1st[m_1stNullItemsBeginCount].offset); + SuballocationVectorType& suballocations2nd = AccessSuballocations2nd(); + + switch (m_2ndVectorMode) + { + case SECOND_VECTOR_EMPTY: + // First allocation from second part ring buffer. + VMA_ASSERT(suballocations2nd.empty()); + m_2ndVectorMode = SECOND_VECTOR_RING_BUFFER; + break; + case SECOND_VECTOR_RING_BUFFER: + // 2-part ring buffer is already started. + VMA_ASSERT(!suballocations2nd.empty()); + break; + case SECOND_VECTOR_DOUBLE_STACK: + VMA_ASSERT(0 && "CRITICAL ERROR: Trying to use linear allocator as ring buffer while it was already used as double stack."); + break; + default: + VMA_ASSERT(0); + } + + suballocations2nd.push_back(newSuballoc); } - VkDeviceSize resultBaseOffset = size - allocSize; - if(!suballocations2nd.empty()) + break; + default: + VMA_ASSERT(0 && "CRITICAL INTERNAL ERROR."); + } + + m_SumFreeSize -= newSuballoc.size; +} + +void VmaBlockMetadata_Linear::Free(VmaAllocHandle allocHandle) +{ + SuballocationVectorType& suballocations1st = AccessSuballocations1st(); + SuballocationVectorType& suballocations2nd = AccessSuballocations2nd(); + VkDeviceSize offset = (VkDeviceSize)allocHandle - 1; + + if (!suballocations1st.empty()) { - const VmaSuballocation& lastSuballoc = suballocations2nd.back(); - resultBaseOffset = lastSuballoc.offset - allocSize; - if(allocSize > lastSuballoc.offset) + // First allocation: Mark it as next empty at the beginning. + VmaSuballocation& firstSuballoc = suballocations1st[m_1stNullItemsBeginCount]; + if (firstSuballoc.offset == offset) { - return false; + firstSuballoc.type = VMA_SUBALLOCATION_TYPE_FREE; + firstSuballoc.userData = VMA_NULL; + m_SumFreeSize += firstSuballoc.size; + ++m_1stNullItemsBeginCount; + CleanupAfterFree(); + return; } } - // Start from offset equal to end of free space. - VkDeviceSize resultOffset = resultBaseOffset; + // Last allocation in 2-part ring buffer or top of upper stack (same logic). + if (m_2ndVectorMode == SECOND_VECTOR_RING_BUFFER || + m_2ndVectorMode == SECOND_VECTOR_DOUBLE_STACK) + { + VmaSuballocation& lastSuballoc = suballocations2nd.back(); + if (lastSuballoc.offset == offset) + { + m_SumFreeSize += lastSuballoc.size; + suballocations2nd.pop_back(); + CleanupAfterFree(); + return; + } + } + // Last allocation in 1st vector. + else if (m_2ndVectorMode == SECOND_VECTOR_EMPTY) + { + VmaSuballocation& lastSuballoc = suballocations1st.back(); + if (lastSuballoc.offset == offset) + { + m_SumFreeSize += lastSuballoc.size; + suballocations1st.pop_back(); + CleanupAfterFree(); + return; + } + } + + VmaSuballocation refSuballoc; + refSuballoc.offset = offset; + // Rest of members stays uninitialized intentionally for better performance. - // Apply VMA_DEBUG_MARGIN at the end. - if(VMA_DEBUG_MARGIN > 0) + // Item from the middle of 1st vector. { - if(resultOffset < VMA_DEBUG_MARGIN) + const SuballocationVectorType::iterator it = VmaBinaryFindSorted( + suballocations1st.begin() + m_1stNullItemsBeginCount, + suballocations1st.end(), + refSuballoc, + VmaSuballocationOffsetLess()); + if (it != suballocations1st.end()) { - return false; + it->type = VMA_SUBALLOCATION_TYPE_FREE; + it->userData = VMA_NULL; + ++m_1stNullItemsMiddleCount; + m_SumFreeSize += it->size; + CleanupAfterFree(); + return; } - resultOffset -= VMA_DEBUG_MARGIN; } - // Apply alignment. - resultOffset = VmaAlignDown(resultOffset, allocAlignment); + if (m_2ndVectorMode != SECOND_VECTOR_EMPTY) + { + // Item from the middle of 2nd vector. + const SuballocationVectorType::iterator it = m_2ndVectorMode == SECOND_VECTOR_RING_BUFFER ? + VmaBinaryFindSorted(suballocations2nd.begin(), suballocations2nd.end(), refSuballoc, VmaSuballocationOffsetLess()) : + VmaBinaryFindSorted(suballocations2nd.begin(), suballocations2nd.end(), refSuballoc, VmaSuballocationOffsetGreater()); + if (it != suballocations2nd.end()) + { + it->type = VMA_SUBALLOCATION_TYPE_FREE; + it->userData = VMA_NULL; + ++m_2ndNullItemsCount; + m_SumFreeSize += it->size; + CleanupAfterFree(); + return; + } + } - // Check next suballocations from 2nd for BufferImageGranularity conflicts. - // Make bigger alignment if necessary. - if(bufferImageGranularity > 1 && bufferImageGranularity != allocAlignment && !suballocations2nd.empty()) + VMA_ASSERT(0 && "Allocation to free not found in linear allocator!"); +} + +void VmaBlockMetadata_Linear::GetAllocationInfo(VmaAllocHandle allocHandle, VmaVirtualAllocationInfo& outInfo) +{ + outInfo.offset = (VkDeviceSize)allocHandle - 1; + VmaSuballocation& suballoc = FindSuballocation(outInfo.offset); + outInfo.size = suballoc.size; + outInfo.pUserData = suballoc.userData; +} + +void VmaBlockMetadata_Linear::Clear() +{ + m_SumFreeSize = GetSize(); + m_Suballocations0.clear(); + m_Suballocations1.clear(); + // Leaving m_1stVectorIndex unchanged - it doesn't matter. + m_2ndVectorMode = SECOND_VECTOR_EMPTY; + m_1stNullItemsBeginCount = 0; + m_1stNullItemsMiddleCount = 0; + m_2ndNullItemsCount = 0; +} + +void VmaBlockMetadata_Linear::SetAllocationUserData(VmaAllocHandle allocHandle, void* userData) +{ + VmaSuballocation& suballoc = FindSuballocation((VkDeviceSize)allocHandle - 1); + suballoc.userData = userData; +} + +void VmaBlockMetadata_Linear::DebugLogAllAllocations() const +{ + const SuballocationVectorType& suballocations1st = AccessSuballocations1st(); + for (auto it = suballocations1st.begin() + m_1stNullItemsBeginCount; it != suballocations1st.end(); ++it) + if (it->type != VMA_SUBALLOCATION_TYPE_FREE) + DebugLogAllocation(it->offset, it->size, it->userData); + + const SuballocationVectorType& suballocations2nd = AccessSuballocations2nd(); + for (auto it = suballocations2nd.begin(); it != suballocations2nd.end(); ++it) + if (it->type != VMA_SUBALLOCATION_TYPE_FREE) + DebugLogAllocation(it->offset, it->size, it->userData); +} + +VmaSuballocation& VmaBlockMetadata_Linear::FindSuballocation(VkDeviceSize offset) +{ + SuballocationVectorType& suballocations1st = AccessSuballocations1st(); + SuballocationVectorType& suballocations2nd = AccessSuballocations2nd(); + + VmaSuballocation refSuballoc; + refSuballoc.offset = offset; + // Rest of members stays uninitialized intentionally for better performance. + + // Item from the 1st vector. { - bool bufferImageGranularityConflict = false; - for(size_t nextSuballocIndex = suballocations2nd.size(); nextSuballocIndex--; ) + const SuballocationVectorType::iterator it = VmaBinaryFindSorted( + suballocations1st.begin() + m_1stNullItemsBeginCount, + suballocations1st.end(), + refSuballoc, + VmaSuballocationOffsetLess()); + if (it != suballocations1st.end()) { - const VmaSuballocation& nextSuballoc = suballocations2nd[nextSuballocIndex]; - if(VmaBlocksOnSamePage(resultOffset, allocSize, nextSuballoc.offset, bufferImageGranularity)) - { - if(VmaIsBufferImageGranularityConflict(nextSuballoc.type, allocType)) - { - bufferImageGranularityConflict = true; - break; - } - } - else - // Already on previous page. - break; + return *it; } - if(bufferImageGranularityConflict) + } + + if (m_2ndVectorMode != SECOND_VECTOR_EMPTY) + { + // Rest of members stays uninitialized intentionally for better performance. + const SuballocationVectorType::iterator it = m_2ndVectorMode == SECOND_VECTOR_RING_BUFFER ? + VmaBinaryFindSorted(suballocations2nd.begin(), suballocations2nd.end(), refSuballoc, VmaSuballocationOffsetLess()) : + VmaBinaryFindSorted(suballocations2nd.begin(), suballocations2nd.end(), refSuballoc, VmaSuballocationOffsetGreater()); + if (it != suballocations2nd.end()) { - resultOffset = VmaAlignDown(resultOffset, bufferImageGranularity); + return *it; } } - // There is enough free space. - const VkDeviceSize endOf1st = !suballocations1st.empty() ? - suballocations1st.back().offset + suballocations1st.back().size : - 0; - if(endOf1st + VMA_DEBUG_MARGIN <= resultOffset) + VMA_ASSERT(0 && "Allocation not found in linear allocator!"); + return suballocations1st.back(); // Should never occur. +} + +bool VmaBlockMetadata_Linear::ShouldCompact1st() const +{ + const size_t nullItemCount = m_1stNullItemsBeginCount + m_1stNullItemsMiddleCount; + const size_t suballocCount = AccessSuballocations1st().size(); + return suballocCount > 32 && nullItemCount * 2 >= (suballocCount - nullItemCount) * 3; +} + +void VmaBlockMetadata_Linear::CleanupAfterFree() +{ + SuballocationVectorType& suballocations1st = AccessSuballocations1st(); + SuballocationVectorType& suballocations2nd = AccessSuballocations2nd(); + + if (IsEmpty()) { - // Check previous suballocations for BufferImageGranularity conflicts. - // If conflict exists, allocation cannot be made here. - if(bufferImageGranularity > 1) + suballocations1st.clear(); + suballocations2nd.clear(); + m_1stNullItemsBeginCount = 0; + m_1stNullItemsMiddleCount = 0; + m_2ndNullItemsCount = 0; + m_2ndVectorMode = SECOND_VECTOR_EMPTY; + } + else + { + const size_t suballoc1stCount = suballocations1st.size(); + const size_t nullItem1stCount = m_1stNullItemsBeginCount + m_1stNullItemsMiddleCount; + VMA_ASSERT(nullItem1stCount <= suballoc1stCount); + + // Find more null items at the beginning of 1st vector. + while (m_1stNullItemsBeginCount < suballoc1stCount && + suballocations1st[m_1stNullItemsBeginCount].type == VMA_SUBALLOCATION_TYPE_FREE) { - for(size_t prevSuballocIndex = suballocations1st.size(); prevSuballocIndex--; ) + ++m_1stNullItemsBeginCount; + --m_1stNullItemsMiddleCount; + } + + // Find more null items at the end of 1st vector. + while (m_1stNullItemsMiddleCount > 0 && + suballocations1st.back().type == VMA_SUBALLOCATION_TYPE_FREE) + { + --m_1stNullItemsMiddleCount; + suballocations1st.pop_back(); + } + + // Find more null items at the end of 2nd vector. + while (m_2ndNullItemsCount > 0 && + suballocations2nd.back().type == VMA_SUBALLOCATION_TYPE_FREE) + { + --m_2ndNullItemsCount; + suballocations2nd.pop_back(); + } + + // Find more null items at the beginning of 2nd vector. + while (m_2ndNullItemsCount > 0 && + suballocations2nd[0].type == VMA_SUBALLOCATION_TYPE_FREE) + { + --m_2ndNullItemsCount; + VmaVectorRemove(suballocations2nd, 0); + } + + if (ShouldCompact1st()) + { + const size_t nonNullItemCount = suballoc1stCount - nullItem1stCount; + size_t srcIndex = m_1stNullItemsBeginCount; + for (size_t dstIndex = 0; dstIndex < nonNullItemCount; ++dstIndex) { - const VmaSuballocation& prevSuballoc = suballocations1st[prevSuballocIndex]; - if(VmaBlocksOnSamePage(prevSuballoc.offset, prevSuballoc.size, resultOffset, bufferImageGranularity)) + while (suballocations1st[srcIndex].type == VMA_SUBALLOCATION_TYPE_FREE) { - if(VmaIsBufferImageGranularityConflict(allocType, prevSuballoc.type)) - { - return false; - } + ++srcIndex; } - else + if (dstIndex != srcIndex) { - // Already on next page. - break; + suballocations1st[dstIndex] = suballocations1st[srcIndex]; } + ++srcIndex; } + suballocations1st.resize(nonNullItemCount); + m_1stNullItemsBeginCount = 0; + m_1stNullItemsMiddleCount = 0; } - // All tests passed: Success. - pAllocationRequest->offset = resultOffset; - pAllocationRequest->sumFreeSize = resultBaseOffset + allocSize - endOf1st; - pAllocationRequest->sumItemSize = 0; - // pAllocationRequest->item unused. - pAllocationRequest->itemsToMakeLostCount = 0; - pAllocationRequest->type = VmaAllocationRequestType::UpperAddress; - return true; + // 2nd vector became empty. + if (suballocations2nd.empty()) + { + m_2ndVectorMode = SECOND_VECTOR_EMPTY; + } + + // 1st vector became empty. + if (suballocations1st.size() - m_1stNullItemsBeginCount == 0) + { + suballocations1st.clear(); + m_1stNullItemsBeginCount = 0; + + if (!suballocations2nd.empty() && m_2ndVectorMode == SECOND_VECTOR_RING_BUFFER) + { + // Swap 1st with 2nd. Now 2nd is empty. + m_2ndVectorMode = SECOND_VECTOR_EMPTY; + m_1stNullItemsMiddleCount = m_2ndNullItemsCount; + while (m_1stNullItemsBeginCount < suballocations2nd.size() && + suballocations2nd[m_1stNullItemsBeginCount].type == VMA_SUBALLOCATION_TYPE_FREE) + { + ++m_1stNullItemsBeginCount; + --m_1stNullItemsMiddleCount; + } + m_2ndNullItemsCount = 0; + m_1stVectorIndex ^= 1; + } + } } - return false; + VMA_HEAVY_ASSERT(Validate()); } bool VmaBlockMetadata_Linear::CreateAllocationRequest_LowerAddress( - uint32_t currentFrameIndex, - uint32_t frameInUseCount, - VkDeviceSize bufferImageGranularity, VkDeviceSize allocSize, VkDeviceSize allocAlignment, VmaSuballocationType allocType, - bool canMakeOtherLost, uint32_t strategy, VmaAllocationRequest* pAllocationRequest) { - const VkDeviceSize size = GetSize(); + const VkDeviceSize blockSize = GetSize(); + const VkDeviceSize debugMargin = GetDebugMargin(); + const VkDeviceSize bufferImageGranularity = GetBufferImageGranularity(); SuballocationVectorType& suballocations1st = AccessSuballocations1st(); SuballocationVectorType& suballocations2nd = AccessSuballocations2nd(); - if(m_2ndVectorMode == SECOND_VECTOR_EMPTY || m_2ndVectorMode == SECOND_VECTOR_DOUBLE_STACK) + if (m_2ndVectorMode == SECOND_VECTOR_EMPTY || m_2ndVectorMode == SECOND_VECTOR_DOUBLE_STACK) { // Try to allocate at the end of 1st vector. VkDeviceSize resultBaseOffset = 0; - if(!suballocations1st.empty()) + if (!suballocations1st.empty()) { const VmaSuballocation& lastSuballoc = suballocations1st.back(); - resultBaseOffset = lastSuballoc.offset + lastSuballoc.size; + resultBaseOffset = lastSuballoc.offset + lastSuballoc.size + debugMargin; } // Start from offset equal to beginning of free space. VkDeviceSize resultOffset = resultBaseOffset; - // Apply VMA_DEBUG_MARGIN at the beginning. - if(VMA_DEBUG_MARGIN > 0) - { - resultOffset += VMA_DEBUG_MARGIN; - } - // Apply alignment. resultOffset = VmaAlignUp(resultOffset, allocAlignment); // Check previous suballocations for BufferImageGranularity conflicts. // Make bigger alignment if necessary. - if(bufferImageGranularity > 1 && bufferImageGranularity != allocAlignment && !suballocations1st.empty()) + if (bufferImageGranularity > 1 && bufferImageGranularity != allocAlignment && !suballocations1st.empty()) { bool bufferImageGranularityConflict = false; - for(size_t prevSuballocIndex = suballocations1st.size(); prevSuballocIndex--; ) + for (size_t prevSuballocIndex = suballocations1st.size(); prevSuballocIndex--; ) { const VmaSuballocation& prevSuballoc = suballocations1st[prevSuballocIndex]; - if(VmaBlocksOnSamePage(prevSuballoc.offset, prevSuballoc.size, resultOffset, bufferImageGranularity)) + if (VmaBlocksOnSamePage(prevSuballoc.offset, prevSuballoc.size, resultOffset, bufferImageGranularity)) { - if(VmaIsBufferImageGranularityConflict(prevSuballoc.type, allocType)) + if (VmaIsBufferImageGranularityConflict(prevSuballoc.type, allocType)) { bufferImageGranularityConflict = true; break; @@ -11457,28 +8636,28 @@ bool VmaBlockMetadata_Linear::CreateAllocationRequest_LowerAddress( // Already on previous page. break; } - if(bufferImageGranularityConflict) + if (bufferImageGranularityConflict) { resultOffset = VmaAlignUp(resultOffset, bufferImageGranularity); } } const VkDeviceSize freeSpaceEnd = m_2ndVectorMode == SECOND_VECTOR_DOUBLE_STACK ? - suballocations2nd.back().offset : size; + suballocations2nd.back().offset : blockSize; // There is enough free space at the end after alignment. - if(resultOffset + allocSize + VMA_DEBUG_MARGIN <= freeSpaceEnd) + if (resultOffset + allocSize + debugMargin <= freeSpaceEnd) { // Check next suballocations for BufferImageGranularity conflicts. // If conflict exists, allocation cannot be made here. - if((allocSize % bufferImageGranularity || resultOffset % bufferImageGranularity) && m_2ndVectorMode == SECOND_VECTOR_DOUBLE_STACK) + if ((allocSize % bufferImageGranularity || resultOffset % bufferImageGranularity) && m_2ndVectorMode == SECOND_VECTOR_DOUBLE_STACK) { - for(size_t nextSuballocIndex = suballocations2nd.size(); nextSuballocIndex--; ) + for (size_t nextSuballocIndex = suballocations2nd.size(); nextSuballocIndex--; ) { const VmaSuballocation& nextSuballoc = suballocations2nd[nextSuballocIndex]; - if(VmaBlocksOnSamePage(resultOffset, allocSize, nextSuballoc.offset, bufferImageGranularity)) + if (VmaBlocksOnSamePage(resultOffset, allocSize, nextSuballoc.offset, bufferImageGranularity)) { - if(VmaIsBufferImageGranularityConflict(allocType, nextSuballoc.type)) + if (VmaIsBufferImageGranularityConflict(allocType, nextSuballoc.type)) { return false; } @@ -11492,52 +8671,43 @@ bool VmaBlockMetadata_Linear::CreateAllocationRequest_LowerAddress( } // All tests passed: Success. - pAllocationRequest->offset = resultOffset; - pAllocationRequest->sumFreeSize = freeSpaceEnd - resultBaseOffset; - pAllocationRequest->sumItemSize = 0; + pAllocationRequest->allocHandle = (VmaAllocHandle)(resultOffset + 1); // pAllocationRequest->item, customData unused. pAllocationRequest->type = VmaAllocationRequestType::EndOf1st; - pAllocationRequest->itemsToMakeLostCount = 0; return true; } } // Wrap-around to end of 2nd vector. Try to allocate there, watching for the // beginning of 1st vector as the end of free space. - if(m_2ndVectorMode == SECOND_VECTOR_EMPTY || m_2ndVectorMode == SECOND_VECTOR_RING_BUFFER) + if (m_2ndVectorMode == SECOND_VECTOR_EMPTY || m_2ndVectorMode == SECOND_VECTOR_RING_BUFFER) { VMA_ASSERT(!suballocations1st.empty()); VkDeviceSize resultBaseOffset = 0; - if(!suballocations2nd.empty()) + if (!suballocations2nd.empty()) { const VmaSuballocation& lastSuballoc = suballocations2nd.back(); - resultBaseOffset = lastSuballoc.offset + lastSuballoc.size; + resultBaseOffset = lastSuballoc.offset + lastSuballoc.size + debugMargin; } // Start from offset equal to beginning of free space. VkDeviceSize resultOffset = resultBaseOffset; - // Apply VMA_DEBUG_MARGIN at the beginning. - if(VMA_DEBUG_MARGIN > 0) - { - resultOffset += VMA_DEBUG_MARGIN; - } - // Apply alignment. resultOffset = VmaAlignUp(resultOffset, allocAlignment); // Check previous suballocations for BufferImageGranularity conflicts. // Make bigger alignment if necessary. - if(bufferImageGranularity > 1 && bufferImageGranularity != allocAlignment && !suballocations2nd.empty()) + if (bufferImageGranularity > 1 && bufferImageGranularity != allocAlignment && !suballocations2nd.empty()) { bool bufferImageGranularityConflict = false; - for(size_t prevSuballocIndex = suballocations2nd.size(); prevSuballocIndex--; ) + for (size_t prevSuballocIndex = suballocations2nd.size(); prevSuballocIndex--; ) { const VmaSuballocation& prevSuballoc = suballocations2nd[prevSuballocIndex]; - if(VmaBlocksOnSamePage(prevSuballoc.offset, prevSuballoc.size, resultOffset, bufferImageGranularity)) + if (VmaBlocksOnSamePage(prevSuballoc.offset, prevSuballoc.size, resultOffset, bufferImageGranularity)) { - if(VmaIsBufferImageGranularityConflict(prevSuballoc.type, allocType)) + if (VmaIsBufferImageGranularityConflict(prevSuballoc.type, allocType)) { bufferImageGranularityConflict = true; break; @@ -11547,102 +8717,30 @@ bool VmaBlockMetadata_Linear::CreateAllocationRequest_LowerAddress( // Already on previous page. break; } - if(bufferImageGranularityConflict) + if (bufferImageGranularityConflict) { resultOffset = VmaAlignUp(resultOffset, bufferImageGranularity); } } - pAllocationRequest->itemsToMakeLostCount = 0; - pAllocationRequest->sumItemSize = 0; size_t index1st = m_1stNullItemsBeginCount; - if(canMakeOtherLost) - { - while(index1st < suballocations1st.size() && - resultOffset + allocSize + VMA_DEBUG_MARGIN > suballocations1st[index1st].offset) - { - // Next colliding allocation at the beginning of 1st vector found. Try to make it lost. - const VmaSuballocation& suballoc = suballocations1st[index1st]; - if(suballoc.type == VMA_SUBALLOCATION_TYPE_FREE) - { - // No problem. - } - else - { - VMA_ASSERT(suballoc.hAllocation != VK_NULL_HANDLE); - if(suballoc.hAllocation->CanBecomeLost() && - suballoc.hAllocation->GetLastUseFrameIndex() + frameInUseCount < currentFrameIndex) - { - ++pAllocationRequest->itemsToMakeLostCount; - pAllocationRequest->sumItemSize += suballoc.size; - } - else - { - return false; - } - } - ++index1st; - } - - // Check next suballocations for BufferImageGranularity conflicts. - // If conflict exists, we must mark more allocations lost or fail. - if(allocSize % bufferImageGranularity || resultOffset % bufferImageGranularity) - { - while(index1st < suballocations1st.size()) - { - const VmaSuballocation& suballoc = suballocations1st[index1st]; - if(VmaBlocksOnSamePage(resultOffset, allocSize, suballoc.offset, bufferImageGranularity)) - { - if(suballoc.hAllocation != VK_NULL_HANDLE) - { - // Not checking actual VmaIsBufferImageGranularityConflict(allocType, suballoc.type). - if(suballoc.hAllocation->CanBecomeLost() && - suballoc.hAllocation->GetLastUseFrameIndex() + frameInUseCount < currentFrameIndex) - { - ++pAllocationRequest->itemsToMakeLostCount; - pAllocationRequest->sumItemSize += suballoc.size; - } - else - { - return false; - } - } - } - else - { - // Already on next page. - break; - } - ++index1st; - } - } - - // Special case: There is not enough room at the end for this allocation, even after making all from the 1st lost. - if(index1st == suballocations1st.size() && - resultOffset + allocSize + VMA_DEBUG_MARGIN > size) - { - // TODO: This is a known bug that it's not yet implemented and the allocation is failing. - VMA_DEBUG_LOG("Unsupported special case in custom pool with linear allocation algorithm used as ring buffer with allocations that can be lost."); - } - } - // There is enough free space at the end after alignment. - if((index1st == suballocations1st.size() && resultOffset + allocSize + VMA_DEBUG_MARGIN <= size) || - (index1st < suballocations1st.size() && resultOffset + allocSize + VMA_DEBUG_MARGIN <= suballocations1st[index1st].offset)) + if ((index1st == suballocations1st.size() && resultOffset + allocSize + debugMargin <= blockSize) || + (index1st < suballocations1st.size() && resultOffset + allocSize + debugMargin <= suballocations1st[index1st].offset)) { // Check next suballocations for BufferImageGranularity conflicts. // If conflict exists, allocation cannot be made here. - if(allocSize % bufferImageGranularity || resultOffset % bufferImageGranularity) + if (allocSize % bufferImageGranularity || resultOffset % bufferImageGranularity) { - for(size_t nextSuballocIndex = index1st; + for (size_t nextSuballocIndex = index1st; nextSuballocIndex < suballocations1st.size(); nextSuballocIndex++) { const VmaSuballocation& nextSuballoc = suballocations1st[nextSuballocIndex]; - if(VmaBlocksOnSamePage(resultOffset, allocSize, nextSuballoc.offset, bufferImageGranularity)) + if (VmaBlocksOnSamePage(resultOffset, allocSize, nextSuballoc.offset, bufferImageGranularity)) { - if(VmaIsBufferImageGranularityConflict(allocType, nextSuballoc.type)) + if (VmaIsBufferImageGranularityConflict(allocType, nextSuballoc.type)) { return false; } @@ -11656,11 +8754,7 @@ bool VmaBlockMetadata_Linear::CreateAllocationRequest_LowerAddress( } // All tests passed: Success. - pAllocationRequest->offset = resultOffset; - pAllocationRequest->sumFreeSize = - (index1st < suballocations1st.size() ? suballocations1st[index1st].offset : size) - - resultBaseOffset - - pAllocationRequest->sumItemSize; + pAllocationRequest->allocHandle = (VmaAllocHandle)(resultOffset + 1); pAllocationRequest->type = VmaAllocationRequestType::EndOf2nd; // pAllocationRequest->item, customData unused. return true; @@ -11670,437 +8764,274 @@ bool VmaBlockMetadata_Linear::CreateAllocationRequest_LowerAddress( return false; } -bool VmaBlockMetadata_Linear::MakeRequestedAllocationsLost( - uint32_t currentFrameIndex, - uint32_t frameInUseCount, +bool VmaBlockMetadata_Linear::CreateAllocationRequest_UpperAddress( + VkDeviceSize allocSize, + VkDeviceSize allocAlignment, + VmaSuballocationType allocType, + uint32_t strategy, VmaAllocationRequest* pAllocationRequest) { - if(pAllocationRequest->itemsToMakeLostCount == 0) - { - return true; - } - - VMA_ASSERT(m_2ndVectorMode == SECOND_VECTOR_EMPTY || m_2ndVectorMode == SECOND_VECTOR_RING_BUFFER); + const VkDeviceSize blockSize = GetSize(); + const VkDeviceSize bufferImageGranularity = GetBufferImageGranularity(); + SuballocationVectorType& suballocations1st = AccessSuballocations1st(); + SuballocationVectorType& suballocations2nd = AccessSuballocations2nd(); - // We always start from 1st. - SuballocationVectorType* suballocations = &AccessSuballocations1st(); - size_t index = m_1stNullItemsBeginCount; - size_t madeLostCount = 0; - while(madeLostCount < pAllocationRequest->itemsToMakeLostCount) + if (m_2ndVectorMode == SECOND_VECTOR_RING_BUFFER) { - if(index == suballocations->size()) - { - index = 0; - // If we get to the end of 1st, we wrap around to beginning of 2nd of 1st. - if(m_2ndVectorMode == SECOND_VECTOR_RING_BUFFER) - { - suballocations = &AccessSuballocations2nd(); - } - // else: m_2ndVectorMode == SECOND_VECTOR_EMPTY: - // suballocations continues pointing at AccessSuballocations1st(). - VMA_ASSERT(!suballocations->empty()); - } - VmaSuballocation& suballoc = (*suballocations)[index]; - if(suballoc.type != VMA_SUBALLOCATION_TYPE_FREE) - { - VMA_ASSERT(suballoc.hAllocation != VK_NULL_HANDLE); - VMA_ASSERT(suballoc.hAllocation->CanBecomeLost()); - if(suballoc.hAllocation->MakeLost(currentFrameIndex, frameInUseCount)) - { - suballoc.type = VMA_SUBALLOCATION_TYPE_FREE; - suballoc.hAllocation = VK_NULL_HANDLE; - m_SumFreeSize += suballoc.size; - if(suballocations == &AccessSuballocations1st()) - { - ++m_1stNullItemsMiddleCount; - } - else - { - ++m_2ndNullItemsCount; - } - ++madeLostCount; - } - else - { - return false; - } - } - ++index; + VMA_ASSERT(0 && "Trying to use pool with linear algorithm as double stack, while it is already being used as ring buffer."); + return false; } - CleanupAfterFree(); - //VMA_HEAVY_ASSERT(Validate()); // Already called by CleanupAfterFree(). - - return true; -} - -uint32_t VmaBlockMetadata_Linear::MakeAllocationsLost(uint32_t currentFrameIndex, uint32_t frameInUseCount) -{ - uint32_t lostAllocationCount = 0; - - SuballocationVectorType& suballocations1st = AccessSuballocations1st(); - for(size_t i = m_1stNullItemsBeginCount, count = suballocations1st.size(); i < count; ++i) + // Try to allocate before 2nd.back(), or end of block if 2nd.empty(). + if (allocSize > blockSize) { - VmaSuballocation& suballoc = suballocations1st[i]; - if(suballoc.type != VMA_SUBALLOCATION_TYPE_FREE && - suballoc.hAllocation->CanBecomeLost() && - suballoc.hAllocation->MakeLost(currentFrameIndex, frameInUseCount)) - { - suballoc.type = VMA_SUBALLOCATION_TYPE_FREE; - suballoc.hAllocation = VK_NULL_HANDLE; - ++m_1stNullItemsMiddleCount; - m_SumFreeSize += suballoc.size; - ++lostAllocationCount; - } + return false; } - - SuballocationVectorType& suballocations2nd = AccessSuballocations2nd(); - for(size_t i = 0, count = suballocations2nd.size(); i < count; ++i) + VkDeviceSize resultBaseOffset = blockSize - allocSize; + if (!suballocations2nd.empty()) { - VmaSuballocation& suballoc = suballocations2nd[i]; - if(suballoc.type != VMA_SUBALLOCATION_TYPE_FREE && - suballoc.hAllocation->CanBecomeLost() && - suballoc.hAllocation->MakeLost(currentFrameIndex, frameInUseCount)) + const VmaSuballocation& lastSuballoc = suballocations2nd.back(); + resultBaseOffset = lastSuballoc.offset - allocSize; + if (allocSize > lastSuballoc.offset) { - suballoc.type = VMA_SUBALLOCATION_TYPE_FREE; - suballoc.hAllocation = VK_NULL_HANDLE; - ++m_2ndNullItemsCount; - m_SumFreeSize += suballoc.size; - ++lostAllocationCount; + return false; } } - if(lostAllocationCount) - { - CleanupAfterFree(); - } + // Start from offset equal to end of free space. + VkDeviceSize resultOffset = resultBaseOffset; - return lostAllocationCount; -} + const VkDeviceSize debugMargin = GetDebugMargin(); -VkResult VmaBlockMetadata_Linear::CheckCorruption(const void* pBlockData) -{ - SuballocationVectorType& suballocations1st = AccessSuballocations1st(); - for(size_t i = m_1stNullItemsBeginCount, count = suballocations1st.size(); i < count; ++i) + // Apply debugMargin at the end. + if (debugMargin > 0) { - const VmaSuballocation& suballoc = suballocations1st[i]; - if(suballoc.type != VMA_SUBALLOCATION_TYPE_FREE) + if (resultOffset < debugMargin) { - if(!VmaValidateMagicValue(pBlockData, suballoc.offset - VMA_DEBUG_MARGIN)) - { - VMA_ASSERT(0 && "MEMORY CORRUPTION DETECTED BEFORE VALIDATED ALLOCATION!"); - return VK_ERROR_VALIDATION_FAILED_EXT; - } - if(!VmaValidateMagicValue(pBlockData, suballoc.offset + suballoc.size)) - { - VMA_ASSERT(0 && "MEMORY CORRUPTION DETECTED AFTER VALIDATED ALLOCATION!"); - return VK_ERROR_VALIDATION_FAILED_EXT; - } + return false; } + resultOffset -= debugMargin; } - SuballocationVectorType& suballocations2nd = AccessSuballocations2nd(); - for(size_t i = 0, count = suballocations2nd.size(); i < count; ++i) + // Apply alignment. + resultOffset = VmaAlignDown(resultOffset, allocAlignment); + + // Check next suballocations from 2nd for BufferImageGranularity conflicts. + // Make bigger alignment if necessary. + if (bufferImageGranularity > 1 && bufferImageGranularity != allocAlignment && !suballocations2nd.empty()) { - const VmaSuballocation& suballoc = suballocations2nd[i]; - if(suballoc.type != VMA_SUBALLOCATION_TYPE_FREE) + bool bufferImageGranularityConflict = false; + for (size_t nextSuballocIndex = suballocations2nd.size(); nextSuballocIndex--; ) { - if(!VmaValidateMagicValue(pBlockData, suballoc.offset - VMA_DEBUG_MARGIN)) - { - VMA_ASSERT(0 && "MEMORY CORRUPTION DETECTED BEFORE VALIDATED ALLOCATION!"); - return VK_ERROR_VALIDATION_FAILED_EXT; - } - if(!VmaValidateMagicValue(pBlockData, suballoc.offset + suballoc.size)) + const VmaSuballocation& nextSuballoc = suballocations2nd[nextSuballocIndex]; + if (VmaBlocksOnSamePage(resultOffset, allocSize, nextSuballoc.offset, bufferImageGranularity)) { - VMA_ASSERT(0 && "MEMORY CORRUPTION DETECTED AFTER VALIDATED ALLOCATION!"); - return VK_ERROR_VALIDATION_FAILED_EXT; + if (VmaIsBufferImageGranularityConflict(nextSuballoc.type, allocType)) + { + bufferImageGranularityConflict = true; + break; + } } + else + // Already on previous page. + break; } - } - - return VK_SUCCESS; -} - -void VmaBlockMetadata_Linear::Alloc( - const VmaAllocationRequest& request, - VmaSuballocationType type, - VkDeviceSize allocSize, - VmaAllocation hAllocation) -{ - const VmaSuballocation newSuballoc = { request.offset, allocSize, hAllocation, type }; - - switch(request.type) - { - case VmaAllocationRequestType::UpperAddress: + if (bufferImageGranularityConflict) { - VMA_ASSERT(m_2ndVectorMode != SECOND_VECTOR_RING_BUFFER && - "CRITICAL ERROR: Trying to use linear allocator as double stack while it was already used as ring buffer."); - SuballocationVectorType& suballocations2nd = AccessSuballocations2nd(); - suballocations2nd.push_back(newSuballoc); - m_2ndVectorMode = SECOND_VECTOR_DOUBLE_STACK; + resultOffset = VmaAlignDown(resultOffset, bufferImageGranularity); } - break; - case VmaAllocationRequestType::EndOf1st: - { - SuballocationVectorType& suballocations1st = AccessSuballocations1st(); - - VMA_ASSERT(suballocations1st.empty() || - request.offset >= suballocations1st.back().offset + suballocations1st.back().size); - // Check if it fits before the end of the block. - VMA_ASSERT(request.offset + allocSize <= GetSize()); + } - suballocations1st.push_back(newSuballoc); - } - break; - case VmaAllocationRequestType::EndOf2nd: + // There is enough free space. + const VkDeviceSize endOf1st = !suballocations1st.empty() ? + suballocations1st.back().offset + suballocations1st.back().size : + 0; + if (endOf1st + debugMargin <= resultOffset) + { + // Check previous suballocations for BufferImageGranularity conflicts. + // If conflict exists, allocation cannot be made here. + if (bufferImageGranularity > 1) { - SuballocationVectorType& suballocations1st = AccessSuballocations1st(); - // New allocation at the end of 2-part ring buffer, so before first allocation from 1st vector. - VMA_ASSERT(!suballocations1st.empty() && - request.offset + allocSize <= suballocations1st[m_1stNullItemsBeginCount].offset); - SuballocationVectorType& suballocations2nd = AccessSuballocations2nd(); - - switch(m_2ndVectorMode) + for (size_t prevSuballocIndex = suballocations1st.size(); prevSuballocIndex--; ) { - case SECOND_VECTOR_EMPTY: - // First allocation from second part ring buffer. - VMA_ASSERT(suballocations2nd.empty()); - m_2ndVectorMode = SECOND_VECTOR_RING_BUFFER; - break; - case SECOND_VECTOR_RING_BUFFER: - // 2-part ring buffer is already started. - VMA_ASSERT(!suballocations2nd.empty()); - break; - case SECOND_VECTOR_DOUBLE_STACK: - VMA_ASSERT(0 && "CRITICAL ERROR: Trying to use linear allocator as ring buffer while it was already used as double stack."); - break; - default: - VMA_ASSERT(0); + const VmaSuballocation& prevSuballoc = suballocations1st[prevSuballocIndex]; + if (VmaBlocksOnSamePage(prevSuballoc.offset, prevSuballoc.size, resultOffset, bufferImageGranularity)) + { + if (VmaIsBufferImageGranularityConflict(allocType, prevSuballoc.type)) + { + return false; + } + } + else + { + // Already on next page. + break; + } } - - suballocations2nd.push_back(newSuballoc); } - break; - default: - VMA_ASSERT(0 && "CRITICAL INTERNAL ERROR."); + + // All tests passed: Success. + pAllocationRequest->allocHandle = (VmaAllocHandle)(resultOffset + 1); + // pAllocationRequest->item unused. + pAllocationRequest->type = VmaAllocationRequestType::UpperAddress; + return true; } - m_SumFreeSize -= newSuballoc.size; + return false; } +#endif // _VMA_BLOCK_METADATA_LINEAR_FUNCTIONS +#endif // _VMA_BLOCK_METADATA_LINEAR -void VmaBlockMetadata_Linear::Free(const VmaAllocation allocation) -{ - FreeAtOffset(allocation->GetOffset()); -} +#ifndef _VMA_BLOCK_METADATA_BUDDY +/* +- GetSize() is the original size of allocated memory block. +- m_UsableSize is this size aligned down to a power of two. + All allocations and calculations happen relative to m_UsableSize. +- GetUnusableSize() is the difference between them. + It is reported as separate, unused range, not available for allocations. -void VmaBlockMetadata_Linear::FreeAtOffset(VkDeviceSize offset) +Node at level 0 has size = m_UsableSize. +Each next level contains nodes with size 2 times smaller than current level. +m_LevelCount is the maximum number of levels to use in the current object. +*/ +class VmaBlockMetadata_Buddy : public VmaBlockMetadata { - SuballocationVectorType& suballocations1st = AccessSuballocations1st(); - SuballocationVectorType& suballocations2nd = AccessSuballocations2nd(); + VMA_CLASS_NO_COPY(VmaBlockMetadata_Buddy) +public: + VmaBlockMetadata_Buddy(const VkAllocationCallbacks* pAllocationCallbacks, + VkDeviceSize bufferImageGranularity, bool isVirtual); + virtual ~VmaBlockMetadata_Buddy(); - if(!suballocations1st.empty()) - { - // First allocation: Mark it as next empty at the beginning. - VmaSuballocation& firstSuballoc = suballocations1st[m_1stNullItemsBeginCount]; - if(firstSuballoc.offset == offset) - { - firstSuballoc.type = VMA_SUBALLOCATION_TYPE_FREE; - firstSuballoc.hAllocation = VK_NULL_HANDLE; - m_SumFreeSize += firstSuballoc.size; - ++m_1stNullItemsBeginCount; - CleanupAfterFree(); - return; - } - } + size_t GetAllocationCount() const override { return m_AllocationCount; } + VkDeviceSize GetSumFreeSize() const override { return m_SumFreeSize + GetUnusableSize(); } + bool IsEmpty() const override { return m_Root->type == Node::TYPE_FREE; } + VkResult CheckCorruption(const void* pBlockData) override { return VK_ERROR_FEATURE_NOT_PRESENT; } + VkDeviceSize GetAllocationOffset(VmaAllocHandle allocHandle) const override { return (VkDeviceSize)allocHandle - 1; }; + void DebugLogAllAllocations() const override { DebugLogAllAllocationNode(m_Root, 0); } - // Last allocation in 2-part ring buffer or top of upper stack (same logic). - if(m_2ndVectorMode == SECOND_VECTOR_RING_BUFFER || - m_2ndVectorMode == SECOND_VECTOR_DOUBLE_STACK) - { - VmaSuballocation& lastSuballoc = suballocations2nd.back(); - if(lastSuballoc.offset == offset) - { - m_SumFreeSize += lastSuballoc.size; - suballocations2nd.pop_back(); - CleanupAfterFree(); - return; - } - } - // Last allocation in 1st vector. - else if(m_2ndVectorMode == SECOND_VECTOR_EMPTY) - { - VmaSuballocation& lastSuballoc = suballocations1st.back(); - if(lastSuballoc.offset == offset) - { - m_SumFreeSize += lastSuballoc.size; - suballocations1st.pop_back(); - CleanupAfterFree(); - return; - } - } + void Init(VkDeviceSize size) override; + bool Validate() const override; - // Item from the middle of 1st vector. - { - VmaSuballocation refSuballoc; - refSuballoc.offset = offset; - // Rest of members stays uninitialized intentionally for better performance. - SuballocationVectorType::iterator it = VmaBinaryFindSorted( - suballocations1st.begin() + m_1stNullItemsBeginCount, - suballocations1st.end(), - refSuballoc, - VmaSuballocationOffsetLess()); - if(it != suballocations1st.end()) - { - it->type = VMA_SUBALLOCATION_TYPE_FREE; - it->hAllocation = VK_NULL_HANDLE; - ++m_1stNullItemsMiddleCount; - m_SumFreeSize += it->size; - CleanupAfterFree(); - return; - } - } + void CalcAllocationStatInfo(VmaStatInfo& outInfo) const override; + void AddPoolStats(VmaPoolStats& inoutStats) const override; - if(m_2ndVectorMode != SECOND_VECTOR_EMPTY) - { - // Item from the middle of 2nd vector. - VmaSuballocation refSuballoc; - refSuballoc.offset = offset; - // Rest of members stays uninitialized intentionally for better performance. - SuballocationVectorType::iterator it = m_2ndVectorMode == SECOND_VECTOR_RING_BUFFER ? - VmaBinaryFindSorted(suballocations2nd.begin(), suballocations2nd.end(), refSuballoc, VmaSuballocationOffsetLess()) : - VmaBinaryFindSorted(suballocations2nd.begin(), suballocations2nd.end(), refSuballoc, VmaSuballocationOffsetGreater()); - if(it != suballocations2nd.end()) - { - it->type = VMA_SUBALLOCATION_TYPE_FREE; - it->hAllocation = VK_NULL_HANDLE; - ++m_2ndNullItemsCount; - m_SumFreeSize += it->size; - CleanupAfterFree(); - return; - } - } +#if VMA_STATS_STRING_ENABLED + void PrintDetailedMap(class VmaJsonWriter& json) const override; +#endif - VMA_ASSERT(0 && "Allocation to free not found in linear allocator!"); -} + bool CreateAllocationRequest( + VkDeviceSize allocSize, + VkDeviceSize allocAlignment, + bool upperAddress, + VmaSuballocationType allocType, + uint32_t strategy, + VmaAllocationRequest* pAllocationRequest) override; -bool VmaBlockMetadata_Linear::ShouldCompact1st() const -{ - const size_t nullItemCount = m_1stNullItemsBeginCount + m_1stNullItemsMiddleCount; - const size_t suballocCount = AccessSuballocations1st().size(); - return suballocCount > 32 && nullItemCount * 2 >= (suballocCount - nullItemCount) * 3; -} + void Alloc( + const VmaAllocationRequest& request, + VmaSuballocationType type, + void* userData) override; -void VmaBlockMetadata_Linear::CleanupAfterFree() -{ - SuballocationVectorType& suballocations1st = AccessSuballocations1st(); - SuballocationVectorType& suballocations2nd = AccessSuballocations2nd(); + void Free(VmaAllocHandle allocHandle) override; + void GetAllocationInfo(VmaAllocHandle allocHandle, VmaVirtualAllocationInfo& outInfo) override; + void Clear() override; + void SetAllocationUserData(VmaAllocHandle allocHandle, void* userData) override; - if(IsEmpty()) +private: + static const size_t MAX_LEVELS = 48; + + struct ValidationContext { - suballocations1st.clear(); - suballocations2nd.clear(); - m_1stNullItemsBeginCount = 0; - m_1stNullItemsMiddleCount = 0; - m_2ndNullItemsCount = 0; - m_2ndVectorMode = SECOND_VECTOR_EMPTY; - } - else + size_t calculatedAllocationCount = 0; + size_t calculatedFreeCount = 0; + VkDeviceSize calculatedSumFreeSize = 0; + }; + struct Node { - const size_t suballoc1stCount = suballocations1st.size(); - const size_t nullItem1stCount = m_1stNullItemsBeginCount + m_1stNullItemsMiddleCount; - VMA_ASSERT(nullItem1stCount <= suballoc1stCount); - - // Find more null items at the beginning of 1st vector. - while(m_1stNullItemsBeginCount < suballoc1stCount && - suballocations1st[m_1stNullItemsBeginCount].hAllocation == VK_NULL_HANDLE) - { - ++m_1stNullItemsBeginCount; - --m_1stNullItemsMiddleCount; - } - - // Find more null items at the end of 1st vector. - while(m_1stNullItemsMiddleCount > 0 && - suballocations1st.back().hAllocation == VK_NULL_HANDLE) + VkDeviceSize offset; + enum TYPE { - --m_1stNullItemsMiddleCount; - suballocations1st.pop_back(); - } + TYPE_FREE, + TYPE_ALLOCATION, + TYPE_SPLIT, + TYPE_COUNT + } type; + Node* parent; + Node* buddy; - // Find more null items at the end of 2nd vector. - while(m_2ndNullItemsCount > 0 && - suballocations2nd.back().hAllocation == VK_NULL_HANDLE) + union { - --m_2ndNullItemsCount; - suballocations2nd.pop_back(); - } + struct + { + Node* prev; + Node* next; + } free; + struct + { + void* userData; + } allocation; + struct + { + Node* leftChild; + } split; + }; + }; - // Find more null items at the beginning of 2nd vector. - while(m_2ndNullItemsCount > 0 && - suballocations2nd[0].hAllocation == VK_NULL_HANDLE) - { - --m_2ndNullItemsCount; - VmaVectorRemove(suballocations2nd, 0); - } + // Size of the memory block aligned down to a power of two. + VkDeviceSize m_UsableSize; + uint32_t m_LevelCount; + VmaPoolAllocator<Node> m_NodeAllocator; + Node* m_Root; + struct + { + Node* front; + Node* back; + } m_FreeList[MAX_LEVELS]; - if(ShouldCompact1st()) - { - const size_t nonNullItemCount = suballoc1stCount - nullItem1stCount; - size_t srcIndex = m_1stNullItemsBeginCount; - for(size_t dstIndex = 0; dstIndex < nonNullItemCount; ++dstIndex) - { - while(suballocations1st[srcIndex].hAllocation == VK_NULL_HANDLE) - { - ++srcIndex; - } - if(dstIndex != srcIndex) - { - suballocations1st[dstIndex] = suballocations1st[srcIndex]; - } - ++srcIndex; - } - suballocations1st.resize(nonNullItemCount); - m_1stNullItemsBeginCount = 0; - m_1stNullItemsMiddleCount = 0; - } + // Number of nodes in the tree with type == TYPE_ALLOCATION. + size_t m_AllocationCount; + // Number of nodes in the tree with type == TYPE_FREE. + size_t m_FreeCount; + // Doesn't include space wasted due to internal fragmentation - allocation sizes are just aligned up to node sizes. + // Doesn't include unusable size. + VkDeviceSize m_SumFreeSize; - // 2nd vector became empty. - if(suballocations2nd.empty()) - { - m_2ndVectorMode = SECOND_VECTOR_EMPTY; - } + VkDeviceSize GetUnusableSize() const { return GetSize() - m_UsableSize; } + VkDeviceSize LevelToNodeSize(uint32_t level) const { return m_UsableSize >> level; } - // 1st vector became empty. - if(suballocations1st.size() - m_1stNullItemsBeginCount == 0) + VkDeviceSize AlignAllocationSize(VkDeviceSize size) const + { + if (!IsVirtual()) { - suballocations1st.clear(); - m_1stNullItemsBeginCount = 0; - - if(!suballocations2nd.empty() && m_2ndVectorMode == SECOND_VECTOR_RING_BUFFER) - { - // Swap 1st with 2nd. Now 2nd is empty. - m_2ndVectorMode = SECOND_VECTOR_EMPTY; - m_1stNullItemsMiddleCount = m_2ndNullItemsCount; - while(m_1stNullItemsBeginCount < suballocations2nd.size() && - suballocations2nd[m_1stNullItemsBeginCount].hAllocation == VK_NULL_HANDLE) - { - ++m_1stNullItemsBeginCount; - --m_1stNullItemsMiddleCount; - } - m_2ndNullItemsCount = 0; - m_1stVectorIndex ^= 1; - } + size = VmaAlignUp(size, (VkDeviceSize)16); } + return VmaNextPow2(size); } + Node* FindAllocationNode(VkDeviceSize offset, uint32_t& outLevel); + void DeleteNodeChildren(Node* node); + bool ValidateNode(ValidationContext& ctx, const Node* parent, const Node* curr, uint32_t level, VkDeviceSize levelNodeSize) const; + uint32_t AllocSizeToLevel(VkDeviceSize allocSize) const; + void CalcAllocationStatInfoNode(VmaStatInfo& inoutInfo, const Node* node, VkDeviceSize levelNodeSize) const; + // Adds node to the front of FreeList at given level. + // node->type must be FREE. + // node->free.prev, next can be undefined. + void AddToFreeListFront(uint32_t level, Node* node); + // Removes node from FreeList at given level. + // node->type must be FREE. + // node->free.prev, next stay untouched. + void RemoveFromFreeList(uint32_t level, Node* node); + void DebugLogAllAllocationNode(Node* node, uint32_t level) const; - VMA_HEAVY_ASSERT(Validate()); -} - - -//////////////////////////////////////////////////////////////////////////////// -// class VmaBlockMetadata_Buddy +#if VMA_STATS_STRING_ENABLED + void PrintDetailedMapNode(class VmaJsonWriter& json, const Node* node, VkDeviceSize levelNodeSize) const; +#endif +}; -VmaBlockMetadata_Buddy::VmaBlockMetadata_Buddy(VmaAllocator hAllocator) : - VmaBlockMetadata(hAllocator), +#ifndef _VMA_BLOCK_METADATA_BUDDY_FUNCTIONS +VmaBlockMetadata_Buddy::VmaBlockMetadata_Buddy(const VkAllocationCallbacks* pAllocationCallbacks, + VkDeviceSize bufferImageGranularity, bool isVirtual) + : VmaBlockMetadata(pAllocationCallbacks, bufferImageGranularity, isVirtual), + m_NodeAllocator(pAllocationCallbacks, 32), // firstBlockCapacity m_Root(VMA_NULL), m_AllocationCount(0), m_FreeCount(1), @@ -12111,7 +9042,8 @@ VmaBlockMetadata_Buddy::VmaBlockMetadata_Buddy(VmaAllocator hAllocator) : VmaBlockMetadata_Buddy::~VmaBlockMetadata_Buddy() { - DeleteNode(m_Root); + DeleteNodeChildren(m_Root); + m_NodeAllocator.Free(m_Root); } void VmaBlockMetadata_Buddy::Init(VkDeviceSize size) @@ -12122,14 +9054,15 @@ void VmaBlockMetadata_Buddy::Init(VkDeviceSize size) m_SumFreeSize = m_UsableSize; // Calculate m_LevelCount. + const VkDeviceSize minNodeSize = IsVirtual() ? 1 : 16; m_LevelCount = 1; - while(m_LevelCount < MAX_LEVELS && - LevelToNodeSize(m_LevelCount) >= MIN_NODE_SIZE) + while (m_LevelCount < MAX_LEVELS && + LevelToNodeSize(m_LevelCount) >= minNodeSize) { ++m_LevelCount; } - Node* rootNode = vma_new(GetAllocationCallbacks(), Node)(); + Node* rootNode = m_NodeAllocator.Alloc(); rootNode->offset = 0; rootNode->type = Node::TYPE_FREE; rootNode->parent = VMA_NULL; @@ -12143,7 +9076,7 @@ bool VmaBlockMetadata_Buddy::Validate() const { // Validate tree. ValidationContext ctx; - if(!ValidateNode(ctx, VMA_NULL, m_Root, 0, LevelToNodeSize(0))) + if (!ValidateNode(ctx, VMA_NULL, m_Root, 0, LevelToNodeSize(0))) { VMA_VALIDATE(false && "ValidateNode failed."); } @@ -12151,18 +9084,18 @@ bool VmaBlockMetadata_Buddy::Validate() const VMA_VALIDATE(m_SumFreeSize == ctx.calculatedSumFreeSize); // Validate free node lists. - for(uint32_t level = 0; level < m_LevelCount; ++level) + for (uint32_t level = 0; level < m_LevelCount; ++level) { VMA_VALIDATE(m_FreeList[level].front == VMA_NULL || m_FreeList[level].front->free.prev == VMA_NULL); - for(Node* node = m_FreeList[level].front; + for (Node* node = m_FreeList[level].front; node != VMA_NULL; node = node->free.next) { VMA_VALIDATE(node->type == Node::TYPE_FREE); - if(node->free.next == VMA_NULL) + if (node->free.next == VMA_NULL) { VMA_VALIDATE(m_FreeList[level].back == node); } @@ -12174,7 +9107,7 @@ bool VmaBlockMetadata_Buddy::Validate() const } // Validate that free lists ar higher levels are empty. - for(uint32_t level = m_LevelCount; level < MAX_LEVELS; ++level) + for (uint32_t level = m_LevelCount; level < MAX_LEVELS; ++level) { VMA_VALIDATE(m_FreeList[level].front == VMA_NULL && m_FreeList[level].back == VMA_NULL); } @@ -12182,39 +9115,17 @@ bool VmaBlockMetadata_Buddy::Validate() const return true; } -VkDeviceSize VmaBlockMetadata_Buddy::GetUnusedRangeSizeMax() const -{ - for(uint32_t level = 0; level < m_LevelCount; ++level) - { - if(m_FreeList[level].front != VMA_NULL) - { - return LevelToNodeSize(level); - } - } - return 0; -} - void VmaBlockMetadata_Buddy::CalcAllocationStatInfo(VmaStatInfo& outInfo) const { - const VkDeviceSize unusableSize = GetUnusableSize(); - + VmaInitStatInfo(outInfo); outInfo.blockCount = 1; - outInfo.allocationCount = outInfo.unusedRangeCount = 0; - outInfo.usedBytes = outInfo.unusedBytes = 0; - - outInfo.allocationSizeMax = outInfo.unusedRangeSizeMax = 0; - outInfo.allocationSizeMin = outInfo.unusedRangeSizeMin = UINT64_MAX; - outInfo.allocationSizeAvg = outInfo.unusedRangeSizeAvg = 0; // Unused. - CalcAllocationStatInfoNode(outInfo, m_Root, LevelToNodeSize(0)); - if(unusableSize > 0) + const VkDeviceSize unusableSize = GetUnusableSize(); + if (unusableSize > 0) { - ++outInfo.unusedRangeCount; - outInfo.unusedBytes += unusableSize; - outInfo.unusedRangeSizeMax = VMA_MAX(outInfo.unusedRangeSizeMax, unusableSize); - outInfo.unusedRangeSizeMin = VMA_MIN(outInfo.unusedRangeSizeMin, unusableSize); + VmaAddStatInfoUnusedRange(outInfo, unusableSize); } } @@ -12226,20 +9137,16 @@ void VmaBlockMetadata_Buddy::AddPoolStats(VmaPoolStats& inoutStats) const inoutStats.unusedSize += m_SumFreeSize + unusableSize; inoutStats.allocationCount += m_AllocationCount; inoutStats.unusedRangeCount += m_FreeCount; - inoutStats.unusedRangeSizeMax = VMA_MAX(inoutStats.unusedRangeSizeMax, GetUnusedRangeSizeMax()); - if(unusableSize > 0) + if (unusableSize > 0) { ++inoutStats.unusedRangeCount; - // Not updating inoutStats.unusedRangeSizeMax with unusableSize because this space is not available for allocations. } } #if VMA_STATS_STRING_ENABLED - void VmaBlockMetadata_Buddy::PrintDetailedMap(class VmaJsonWriter& json) const { - // TODO optimize VmaStatInfo stat; CalcAllocationStatInfo(stat); @@ -12252,7 +9159,7 @@ void VmaBlockMetadata_Buddy::PrintDetailedMap(class VmaJsonWriter& json) const PrintDetailedMapNode(json, m_Root, LevelToNodeSize(0)); const VkDeviceSize unusableSize = GetUnusableSize(); - if(unusableSize > 0) + if (unusableSize > 0) { PrintDetailedMap_UnusedRange(json, m_UsableSize, // offset @@ -12261,52 +9168,47 @@ void VmaBlockMetadata_Buddy::PrintDetailedMap(class VmaJsonWriter& json) const PrintDetailedMap_End(json); } - -#endif // #if VMA_STATS_STRING_ENABLED +#endif // VMA_STATS_STRING_ENABLED bool VmaBlockMetadata_Buddy::CreateAllocationRequest( - uint32_t currentFrameIndex, - uint32_t frameInUseCount, - VkDeviceSize bufferImageGranularity, VkDeviceSize allocSize, VkDeviceSize allocAlignment, bool upperAddress, VmaSuballocationType allocType, - bool canMakeOtherLost, uint32_t strategy, VmaAllocationRequest* pAllocationRequest) { VMA_ASSERT(!upperAddress && "VMA_ALLOCATION_CREATE_UPPER_ADDRESS_BIT can be used only with linear algorithm."); + allocSize = AlignAllocationSize(allocSize); + // Simple way to respect bufferImageGranularity. May be optimized some day. // Whenever it might be an OPTIMAL image... - if(allocType == VMA_SUBALLOCATION_TYPE_UNKNOWN || + if (allocType == VMA_SUBALLOCATION_TYPE_UNKNOWN || allocType == VMA_SUBALLOCATION_TYPE_IMAGE_UNKNOWN || allocType == VMA_SUBALLOCATION_TYPE_IMAGE_OPTIMAL) { - allocAlignment = VMA_MAX(allocAlignment, bufferImageGranularity); - allocSize = VMA_MAX(allocSize, bufferImageGranularity); + allocAlignment = VMA_MAX(allocAlignment, GetBufferImageGranularity()); + allocSize = VmaAlignUp(allocSize, GetBufferImageGranularity()); } - if(allocSize > m_UsableSize) + if (allocSize > m_UsableSize) { return false; } const uint32_t targetLevel = AllocSizeToLevel(allocSize); - for(uint32_t level = targetLevel + 1; level--; ) + for (uint32_t level = targetLevel; level--; ) { - for(Node* freeNode = m_FreeList[level].front; + for (Node* freeNode = m_FreeList[level].front; freeNode != VMA_NULL; freeNode = freeNode->free.next) { - if(freeNode->offset % allocAlignment == 0) + if (freeNode->offset % allocAlignment == 0) { pAllocationRequest->type = VmaAllocationRequestType::Normal; - pAllocationRequest->offset = freeNode->offset; - pAllocationRequest->sumFreeSize = LevelToNodeSize(level); - pAllocationRequest->sumItemSize = 0; - pAllocationRequest->itemsToMakeLostCount = 0; + pAllocationRequest->allocHandle = (VmaAllocHandle)(freeNode->offset + 1); + pAllocationRequest->size = allocSize; pAllocationRequest->customData = (void*)(uintptr_t)level; return true; } @@ -12316,48 +9218,27 @@ bool VmaBlockMetadata_Buddy::CreateAllocationRequest( return false; } -bool VmaBlockMetadata_Buddy::MakeRequestedAllocationsLost( - uint32_t currentFrameIndex, - uint32_t frameInUseCount, - VmaAllocationRequest* pAllocationRequest) -{ - /* - Lost allocations are not supported in buddy allocator at the moment. - Support might be added in the future. - */ - return pAllocationRequest->itemsToMakeLostCount == 0; -} - -uint32_t VmaBlockMetadata_Buddy::MakeAllocationsLost(uint32_t currentFrameIndex, uint32_t frameInUseCount) -{ - /* - Lost allocations are not supported in buddy allocator at the moment. - Support might be added in the future. - */ - return 0; -} - void VmaBlockMetadata_Buddy::Alloc( const VmaAllocationRequest& request, VmaSuballocationType type, - VkDeviceSize allocSize, - VmaAllocation hAllocation) + void* userData) { VMA_ASSERT(request.type == VmaAllocationRequestType::Normal); - const uint32_t targetLevel = AllocSizeToLevel(allocSize); + const uint32_t targetLevel = AllocSizeToLevel(request.size); uint32_t currLevel = (uint32_t)(uintptr_t)request.customData; Node* currNode = m_FreeList[currLevel].front; VMA_ASSERT(currNode != VMA_NULL && currNode->type == Node::TYPE_FREE); - while(currNode->offset != request.offset) + const VkDeviceSize offset = (VkDeviceSize)request.allocHandle - 1; + while (currNode->offset != offset) { currNode = currNode->free.next; VMA_ASSERT(currNode != VMA_NULL && currNode->type == Node::TYPE_FREE); } // Go down, splitting free nodes. - while(currLevel < targetLevel) + while (currLevel < targetLevel) { // currNode is already first free node at currLevel. // Remove it from list of free nodes at this currLevel. @@ -12366,8 +9247,8 @@ void VmaBlockMetadata_Buddy::Alloc( const uint32_t childrenLevel = currLevel + 1; // Create two free sub-nodes. - Node* leftChild = vma_new(GetAllocationCallbacks(), Node)(); - Node* rightChild = vma_new(GetAllocationCallbacks(), Node)(); + Node* leftChild = m_NodeAllocator.Alloc(); + Node* rightChild = m_NodeAllocator.Alloc(); leftChild->offset = currNode->offset; leftChild->type = Node::TYPE_FREE; @@ -12388,13 +9269,12 @@ void VmaBlockMetadata_Buddy::Alloc( AddToFreeListFront(childrenLevel, leftChild); ++m_FreeCount; - //m_SumFreeSize -= LevelToNodeSize(currLevel) % 2; // Useful only when level node sizes can be non power of 2. ++currLevel; currNode = m_FreeList[currLevel].front; /* We can be sure that currNode, as left child of node previously split, - also fullfills the alignment requirement. + also fulfills the alignment requirement. */ } @@ -12406,22 +9286,74 @@ void VmaBlockMetadata_Buddy::Alloc( // Convert to allocation node. currNode->type = Node::TYPE_ALLOCATION; - currNode->allocation.alloc = hAllocation; + currNode->allocation.userData = userData; ++m_AllocationCount; --m_FreeCount; - m_SumFreeSize -= allocSize; + m_SumFreeSize -= request.size; +} + +void VmaBlockMetadata_Buddy::GetAllocationInfo(VmaAllocHandle allocHandle, VmaVirtualAllocationInfo& outInfo) +{ + uint32_t level = 0; + outInfo.offset = (VkDeviceSize)allocHandle - 1; + const Node* const node = FindAllocationNode(outInfo.offset, level); + outInfo.size = LevelToNodeSize(level); + outInfo.pUserData = node->allocation.userData; +} + +void VmaBlockMetadata_Buddy::DeleteNodeChildren(Node* node) +{ + if (node->type == Node::TYPE_SPLIT) + { + DeleteNodeChildren(node->split.leftChild->buddy); + DeleteNodeChildren(node->split.leftChild); + const VkAllocationCallbacks* allocationCallbacks = GetAllocationCallbacks(); + m_NodeAllocator.Free(node->split.leftChild->buddy); + m_NodeAllocator.Free(node->split.leftChild); + } +} + +void VmaBlockMetadata_Buddy::Clear() +{ + DeleteNodeChildren(m_Root); + m_Root->type = Node::TYPE_FREE; + m_AllocationCount = 0; + m_FreeCount = 1; + m_SumFreeSize = m_UsableSize; +} + +void VmaBlockMetadata_Buddy::SetAllocationUserData(VmaAllocHandle allocHandle, void* userData) +{ + uint32_t level = 0; + Node* const node = FindAllocationNode((VkDeviceSize)allocHandle - 1, level); + node->allocation.userData = userData; } -void VmaBlockMetadata_Buddy::DeleteNode(Node* node) +VmaBlockMetadata_Buddy::Node* VmaBlockMetadata_Buddy::FindAllocationNode(VkDeviceSize offset, uint32_t& outLevel) { - if(node->type == Node::TYPE_SPLIT) + Node* node = m_Root; + VkDeviceSize nodeOffset = 0; + outLevel = 0; + VkDeviceSize levelNodeSize = LevelToNodeSize(0); + while (node->type == Node::TYPE_SPLIT) { - DeleteNode(node->split.leftChild->buddy); - DeleteNode(node->split.leftChild); + const VkDeviceSize nextLevelNodeSize = levelNodeSize >> 1; + if (offset < nodeOffset + nextLevelNodeSize) + { + node = node->split.leftChild; + } + else + { + node = node->split.leftChild->buddy; + nodeOffset += nextLevelNodeSize; + } + ++outLevel; + levelNodeSize = nextLevelNodeSize; } - vma_delete(GetAllocationCallbacks(), node); + VMA_ASSERT(node != VMA_NULL && node->type == Node::TYPE_ALLOCATION); + return node; } bool VmaBlockMetadata_Buddy::ValidateNode(ValidationContext& ctx, const Node* parent, const Node* curr, uint32_t level, VkDeviceSize levelNodeSize) const @@ -12430,7 +9362,7 @@ bool VmaBlockMetadata_Buddy::ValidateNode(ValidationContext& ctx, const Node* pa VMA_VALIDATE(curr->parent == parent); VMA_VALIDATE((curr->buddy == VMA_NULL) == (parent == VMA_NULL)); VMA_VALIDATE(curr->buddy == VMA_NULL || curr->buddy->buddy == curr); - switch(curr->type) + switch (curr->type) { case Node::TYPE_FREE: // curr->free.prev, next are validated separately. @@ -12439,28 +9371,30 @@ bool VmaBlockMetadata_Buddy::ValidateNode(ValidationContext& ctx, const Node* pa break; case Node::TYPE_ALLOCATION: ++ctx.calculatedAllocationCount; - ctx.calculatedSumFreeSize += levelNodeSize - curr->allocation.alloc->GetSize(); - VMA_VALIDATE(curr->allocation.alloc != VK_NULL_HANDLE); + if (!IsVirtual()) + { + VMA_VALIDATE(curr->allocation.userData != VMA_NULL); + } break; case Node::TYPE_SPLIT: + { + const uint32_t childrenLevel = level + 1; + const VkDeviceSize childrenLevelNodeSize = levelNodeSize >> 1; + const Node* const leftChild = curr->split.leftChild; + VMA_VALIDATE(leftChild != VMA_NULL); + VMA_VALIDATE(leftChild->offset == curr->offset); + if (!ValidateNode(ctx, curr, leftChild, childrenLevel, childrenLevelNodeSize)) { - const uint32_t childrenLevel = level + 1; - const VkDeviceSize childrenLevelNodeSize = levelNodeSize / 2; - const Node* const leftChild = curr->split.leftChild; - VMA_VALIDATE(leftChild != VMA_NULL); - VMA_VALIDATE(leftChild->offset == curr->offset); - if(!ValidateNode(ctx, curr, leftChild, childrenLevel, childrenLevelNodeSize)) - { - VMA_VALIDATE(false && "ValidateNode for left child failed."); - } - const Node* const rightChild = leftChild->buddy; - VMA_VALIDATE(rightChild->offset == curr->offset + childrenLevelNodeSize); - if(!ValidateNode(ctx, curr, rightChild, childrenLevel, childrenLevelNodeSize)) - { - VMA_VALIDATE(false && "ValidateNode for right child failed."); - } + VMA_VALIDATE(false && "ValidateNode for left child failed."); } - break; + const Node* const rightChild = leftChild->buddy; + VMA_VALIDATE(rightChild->offset == curr->offset + childrenLevelNodeSize); + if (!ValidateNode(ctx, curr, rightChild, childrenLevel, childrenLevelNodeSize)) + { + VMA_VALIDATE(false && "ValidateNode for right child failed."); + } + } + break; default: return false; } @@ -12474,103 +9408,63 @@ uint32_t VmaBlockMetadata_Buddy::AllocSizeToLevel(VkDeviceSize allocSize) const uint32_t level = 0; VkDeviceSize currLevelNodeSize = m_UsableSize; VkDeviceSize nextLevelNodeSize = currLevelNodeSize >> 1; - while(allocSize <= nextLevelNodeSize && level + 1 < m_LevelCount) + while (allocSize <= nextLevelNodeSize && level + 1 < m_LevelCount) { ++level; - currLevelNodeSize = nextLevelNodeSize; - nextLevelNodeSize = currLevelNodeSize >> 1; + currLevelNodeSize >>= 1; + nextLevelNodeSize >>= 1; } return level; } -void VmaBlockMetadata_Buddy::FreeAtOffset(VmaAllocation alloc, VkDeviceSize offset) +void VmaBlockMetadata_Buddy::Free(VmaAllocHandle allocHandle) { - // Find node and level. - Node* node = m_Root; - VkDeviceSize nodeOffset = 0; uint32_t level = 0; - VkDeviceSize levelNodeSize = LevelToNodeSize(0); - while(node->type == Node::TYPE_SPLIT) - { - const VkDeviceSize nextLevelSize = levelNodeSize >> 1; - if(offset < nodeOffset + nextLevelSize) - { - node = node->split.leftChild; - } - else - { - node = node->split.leftChild->buddy; - nodeOffset += nextLevelSize; - } - ++level; - levelNodeSize = nextLevelSize; - } - - VMA_ASSERT(node != VMA_NULL && node->type == Node::TYPE_ALLOCATION); - VMA_ASSERT(alloc == VK_NULL_HANDLE || node->allocation.alloc == alloc); + Node* node = FindAllocationNode((VkDeviceSize)allocHandle - 1, level); ++m_FreeCount; --m_AllocationCount; - m_SumFreeSize += alloc->GetSize(); + m_SumFreeSize += LevelToNodeSize(level); node->type = Node::TYPE_FREE; // Join free nodes if possible. - while(level > 0 && node->buddy->type == Node::TYPE_FREE) + while (level > 0 && node->buddy->type == Node::TYPE_FREE) { RemoveFromFreeList(level, node->buddy); Node* const parent = node->parent; - vma_delete(GetAllocationCallbacks(), node->buddy); - vma_delete(GetAllocationCallbacks(), node); + m_NodeAllocator.Free(node->buddy); + m_NodeAllocator.Free(node); parent->type = Node::TYPE_FREE; node = parent; --level; - //m_SumFreeSize += LevelToNodeSize(level) % 2; // Useful only when level node sizes can be non power of 2. --m_FreeCount; } AddToFreeListFront(level, node); } -void VmaBlockMetadata_Buddy::CalcAllocationStatInfoNode(VmaStatInfo& outInfo, const Node* node, VkDeviceSize levelNodeSize) const +void VmaBlockMetadata_Buddy::CalcAllocationStatInfoNode(VmaStatInfo& inoutInfo, const Node* node, VkDeviceSize levelNodeSize) const { - switch(node->type) + switch (node->type) { case Node::TYPE_FREE: - ++outInfo.unusedRangeCount; - outInfo.unusedBytes += levelNodeSize; - outInfo.unusedRangeSizeMax = VMA_MAX(outInfo.unusedRangeSizeMax, levelNodeSize); - outInfo.unusedRangeSizeMin = VMA_MAX(outInfo.unusedRangeSizeMin, levelNodeSize); + VmaAddStatInfoUnusedRange(inoutInfo, levelNodeSize); break; case Node::TYPE_ALLOCATION: - { - const VkDeviceSize allocSize = node->allocation.alloc->GetSize(); - ++outInfo.allocationCount; - outInfo.usedBytes += allocSize; - outInfo.allocationSizeMax = VMA_MAX(outInfo.allocationSizeMax, allocSize); - outInfo.allocationSizeMin = VMA_MAX(outInfo.allocationSizeMin, allocSize); - - const VkDeviceSize unusedRangeSize = levelNodeSize - allocSize; - if(unusedRangeSize > 0) - { - ++outInfo.unusedRangeCount; - outInfo.unusedBytes += unusedRangeSize; - outInfo.unusedRangeSizeMax = VMA_MAX(outInfo.unusedRangeSizeMax, unusedRangeSize); - outInfo.unusedRangeSizeMin = VMA_MAX(outInfo.unusedRangeSizeMin, unusedRangeSize); - } - } + VmaAddStatInfoAllocation(inoutInfo, levelNodeSize); break; case Node::TYPE_SPLIT: - { - const VkDeviceSize childrenNodeSize = levelNodeSize / 2; - const Node* const leftChild = node->split.leftChild; - CalcAllocationStatInfoNode(outInfo, leftChild, childrenNodeSize); - const Node* const rightChild = leftChild->buddy; - CalcAllocationStatInfoNode(outInfo, rightChild, childrenNodeSize); - } - break; + { + const VkDeviceSize childrenNodeSize = levelNodeSize / 2; + const Node* const leftChild = node->split.leftChild; + CalcAllocationStatInfoNode(inoutInfo, leftChild, childrenNodeSize); + const Node* const rightChild = leftChild->buddy; + CalcAllocationStatInfoNode(inoutInfo, rightChild, childrenNodeSize); + } + break; default: VMA_ASSERT(0); } @@ -12582,7 +9476,7 @@ void VmaBlockMetadata_Buddy::AddToFreeListFront(uint32_t level, Node* node) // List is empty. Node* const frontNode = m_FreeList[level].front; - if(frontNode == VMA_NULL) + if (frontNode == VMA_NULL) { VMA_ASSERT(m_FreeList[level].back == VMA_NULL); node->free.prev = node->free.next = VMA_NULL; @@ -12603,7 +9497,7 @@ void VmaBlockMetadata_Buddy::RemoveFromFreeList(uint32_t level, Node* node) VMA_ASSERT(m_FreeList[level].front != VMA_NULL); // It is at the front. - if(node->free.prev == VMA_NULL) + if (node->free.prev == VMA_NULL) { VMA_ASSERT(m_FreeList[level].front == node); m_FreeList[level].front = node->free.next; @@ -12616,7 +9510,7 @@ void VmaBlockMetadata_Buddy::RemoveFromFreeList(uint32_t level, Node* node) } // It is at the back. - if(node->free.next == VMA_NULL) + if (node->free.next == VMA_NULL) { VMA_ASSERT(m_FreeList[level].back == node); m_FreeList[level].back = node->free.prev; @@ -12629,85 +9523,2067 @@ void VmaBlockMetadata_Buddy::RemoveFromFreeList(uint32_t level, Node* node) } } +void VmaBlockMetadata_Buddy::DebugLogAllAllocationNode(Node* node, uint32_t level) const +{ + switch (node->type) + { + case Node::TYPE_ALLOCATION: + DebugLogAllocation(node->offset, LevelToNodeSize(level), node->allocation.userData); + break; + case Node::TYPE_SPLIT: + { + ++level; + DebugLogAllAllocationNode(node->split.leftChild, level); + DebugLogAllAllocationNode(node->split.leftChild->buddy, level); + } + break; + default: + VMA_ASSERT(0); + } +} + #if VMA_STATS_STRING_ENABLED void VmaBlockMetadata_Buddy::PrintDetailedMapNode(class VmaJsonWriter& json, const Node* node, VkDeviceSize levelNodeSize) const { - switch(node->type) + switch (node->type) { case Node::TYPE_FREE: PrintDetailedMap_UnusedRange(json, node->offset, levelNodeSize); break; case Node::TYPE_ALLOCATION: + PrintDetailedMap_Allocation(json, node->offset, levelNodeSize, node->allocation.userData); + break; + case Node::TYPE_SPLIT: + { + const VkDeviceSize childrenNodeSize = levelNodeSize / 2; + const Node* const leftChild = node->split.leftChild; + PrintDetailedMapNode(json, leftChild, childrenNodeSize); + const Node* const rightChild = leftChild->buddy; + PrintDetailedMapNode(json, rightChild, childrenNodeSize); + } + break; + default: + VMA_ASSERT(0); + } +} +#endif // VMA_STATS_STRING_ENABLED +#endif // _VMA_BLOCK_METADATA_BUDDY_FUNCTIONS +#endif // _VMA_BLOCK_METADATA_BUDDY + +#ifndef _VMA_BLOCK_METADATA_TLSF +// To not search current larger region if first allocation won't succeed and skip to smaller range +// use with VMA_ALLOCATION_CREATE_STRATEGY_MIN_MEMORY_BIT as strategy in CreateAllocationRequest(). +// When fragmentation and reusal of previous blocks doesn't matter then use with +// VMA_ALLOCATION_CREATE_STRATEGY_MIN_TIME_BIT for fastest alloc time possible. +class VmaBlockMetadata_TLSF : public VmaBlockMetadata +{ + VMA_CLASS_NO_COPY(VmaBlockMetadata_TLSF) +public: + VmaBlockMetadata_TLSF(const VkAllocationCallbacks* pAllocationCallbacks, + VkDeviceSize bufferImageGranularity, bool isVirtual); + virtual ~VmaBlockMetadata_TLSF(); + + size_t GetAllocationCount() const override { return m_AllocCount; } + VkDeviceSize GetSumFreeSize() const override { return m_BlocksFreeSize + m_NullBlock->size; } + bool IsEmpty() const override { return m_NullBlock->offset == 0; } + VkDeviceSize GetAllocationOffset(VmaAllocHandle allocHandle) const override { return ((Block*)allocHandle)->offset; }; + + void Init(VkDeviceSize size) override; + bool Validate() const override; + + void CalcAllocationStatInfo(VmaStatInfo& outInfo) const override; + void AddPoolStats(VmaPoolStats& inoutStats) const override; + +#if VMA_STATS_STRING_ENABLED + void PrintDetailedMap(class VmaJsonWriter& json) const override; +#endif + + bool CreateAllocationRequest( + VkDeviceSize allocSize, + VkDeviceSize allocAlignment, + bool upperAddress, + VmaSuballocationType allocType, + uint32_t strategy, + VmaAllocationRequest* pAllocationRequest) override; + + VkResult CheckCorruption(const void* pBlockData) override; + void Alloc( + const VmaAllocationRequest& request, + VmaSuballocationType type, + void* userData) override; + + void Free(VmaAllocHandle allocHandle) override; + void GetAllocationInfo(VmaAllocHandle allocHandle, VmaVirtualAllocationInfo& outInfo) override; + void Clear() override; + void SetAllocationUserData(VmaAllocHandle allocHandle, void* userData) override; + void DebugLogAllAllocations() const override; + +private: + // According to original paper it should be preferable 4 or 5: + // M. Masmano, I. Ripoll, A. Crespo, and J. Real "TLSF: a New Dynamic Memory Allocator for Real-Time Systems" + // http://www.gii.upv.es/tlsf/files/ecrts04_tlsf.pdf + static const uint8_t SECOND_LEVEL_INDEX = 5; + static const uint16_t SMALL_BUFFER_SIZE = 256; + static const uint32_t INITIAL_BLOCK_ALLOC_COUNT = 16; + static const uint8_t MEMORY_CLASS_SHIFT = 7; + static const uint8_t MAX_MEMORY_CLASSES = 65 - MEMORY_CLASS_SHIFT; + + class Block + { + public: + VkDeviceSize offset; + VkDeviceSize size; + Block* prevPhysical; + Block* nextPhysical; + + void MarkFree() { prevFree = VMA_NULL; } + void MarkTaken() { prevFree = this; } + bool IsFree() const { return prevFree != this; } + void*& UserData() { VMA_HEAVY_ASSERT(!IsFree()); return userData; } + Block*& PrevFree() { return prevFree; } + Block*& NextFree() { VMA_HEAVY_ASSERT(IsFree()); return nextFree; } + + private: + Block* prevFree; // Address of the same block here indicates that block is taken + union { - PrintDetailedMap_Allocation(json, node->offset, node->allocation.alloc); - const VkDeviceSize allocSize = node->allocation.alloc->GetSize(); - if(allocSize < levelNodeSize) + Block* nextFree; + void* userData; + }; + }; + + size_t m_AllocCount; + // Total number of free blocks besides null block + size_t m_BlocksFreeCount; + // Total size of free blocks excluding null block + VkDeviceSize m_BlocksFreeSize; + uint32_t m_IsFreeBitmap; + uint8_t m_MemoryClasses; + uint32_t m_InnerIsFreeBitmap[MAX_MEMORY_CLASSES]; + uint32_t m_ListsCount; + /* + * 0: 0-3 lists for small buffers + * 1+: 0-(2^SLI-1) lists for normal buffers + */ + Block** m_FreeList; + VmaPoolAllocator<Block> m_BlockAllocator; + Block* m_NullBlock; + VmaBlockBufferImageGranularity m_GranularityHandler; + + uint8_t SizeToMemoryClass(VkDeviceSize size) const; + uint16_t SizeToSecondIndex(VkDeviceSize size, uint8_t memoryClass) const; + uint32_t GetListIndex(uint8_t memoryClass, uint16_t secondIndex) const; + uint32_t GetListIndex(VkDeviceSize size) const; + + void RemoveFreeBlock(Block* block); + void InsertFreeBlock(Block* block); + void MergeBlock(Block* block, Block* prev); + + Block* FindFreeBlock(VkDeviceSize size, uint32_t& listIndex) const; + bool CheckBlock( + Block& block, + uint32_t listIndex, + VkDeviceSize allocSize, + VkDeviceSize allocAlignment, + VmaSuballocationType allocType, + VmaAllocationRequest* pAllocationRequest); +}; + +#ifndef _VMA_BLOCK_METADATA_TLSF_FUNCTIONS +VmaBlockMetadata_TLSF::VmaBlockMetadata_TLSF(const VkAllocationCallbacks* pAllocationCallbacks, + VkDeviceSize bufferImageGranularity, bool isVirtual) + : VmaBlockMetadata(pAllocationCallbacks, bufferImageGranularity, isVirtual), + m_AllocCount(0), + m_BlocksFreeCount(0), + m_BlocksFreeSize(0), + m_IsFreeBitmap(0), + m_MemoryClasses(0), + m_ListsCount(0), + m_FreeList(VMA_NULL), + m_BlockAllocator(pAllocationCallbacks, INITIAL_BLOCK_ALLOC_COUNT), + m_NullBlock(VMA_NULL), + m_GranularityHandler(bufferImageGranularity) {} + +VmaBlockMetadata_TLSF::~VmaBlockMetadata_TLSF() +{ + if (m_FreeList) + vma_delete_array(GetAllocationCallbacks(), m_FreeList, m_ListsCount); + m_GranularityHandler.Destroy(GetAllocationCallbacks()); +} + +void VmaBlockMetadata_TLSF::Init(VkDeviceSize size) +{ + VmaBlockMetadata::Init(size); + + if (!IsVirtual()) + m_GranularityHandler.Init(GetAllocationCallbacks(), size); + + m_NullBlock = m_BlockAllocator.Alloc(); + m_NullBlock->size = size; + m_NullBlock->offset = 0; + m_NullBlock->prevPhysical = VMA_NULL; + m_NullBlock->nextPhysical = VMA_NULL; + m_NullBlock->MarkFree(); + m_NullBlock->NextFree() = VMA_NULL; + m_NullBlock->PrevFree() = VMA_NULL; + uint8_t memoryClass = SizeToMemoryClass(size); + uint16_t sli = SizeToSecondIndex(size, memoryClass); + m_ListsCount = (memoryClass == 0 ? 0 : (memoryClass - 1) * (1UL << SECOND_LEVEL_INDEX) + sli) + 1; + if (IsVirtual()) + m_ListsCount += 1UL << SECOND_LEVEL_INDEX; + else + m_ListsCount += 4; + + m_MemoryClasses = memoryClass + 2; + memset(m_InnerIsFreeBitmap, 0, MAX_MEMORY_CLASSES * sizeof(uint32_t)); + + m_FreeList = vma_new_array(GetAllocationCallbacks(), Block*, m_ListsCount); + memset(m_FreeList, 0, m_ListsCount * sizeof(Block*)); +} + +bool VmaBlockMetadata_TLSF::Validate() const +{ + VMA_VALIDATE(GetSumFreeSize() <= GetSize()); + + VkDeviceSize calculatedSize = m_NullBlock->size; + VkDeviceSize calculatedFreeSize = m_NullBlock->size; + size_t allocCount = 0; + size_t freeCount = 0; + + // Check integrity of free lists + for (uint32_t list = 0; list < m_ListsCount; ++list) + { + Block* block = m_FreeList[list]; + if (block != VMA_NULL) + { + VMA_VALIDATE(block->IsFree()); + VMA_VALIDATE(block->PrevFree() == VMA_NULL); + while (block->NextFree()) { - PrintDetailedMap_UnusedRange(json, node->offset + allocSize, levelNodeSize - allocSize); + VMA_VALIDATE(block->NextFree()->IsFree()); + VMA_VALIDATE(block->NextFree()->PrevFree() == block); + block = block->NextFree(); } } - break; - case Node::TYPE_SPLIT: + } + + VkDeviceSize nextOffset = m_NullBlock->offset; + auto validateCtx = m_GranularityHandler.StartValidation(GetAllocationCallbacks(), IsVirtual()); + + VMA_VALIDATE(m_NullBlock->nextPhysical == VMA_NULL); + if (m_NullBlock->prevPhysical) + { + VMA_VALIDATE(m_NullBlock->prevPhysical->nextPhysical == m_NullBlock); + } + // Check all blocks + for (Block* prev = m_NullBlock->prevPhysical; prev != VMA_NULL; prev = prev->prevPhysical) + { + VMA_VALIDATE(prev->offset + prev->size == nextOffset); + nextOffset = prev->offset; + calculatedSize += prev->size; + + uint32_t listIndex = GetListIndex(prev->size); + if (prev->IsFree()) + { + ++freeCount; + // Check if free block belongs to free list + Block* freeBlock = m_FreeList[listIndex]; + VMA_VALIDATE(freeBlock != VMA_NULL); + + bool found = false; + do + { + if (freeBlock == prev) + found = true; + + freeBlock = freeBlock->NextFree(); + } while (!found && freeBlock != VMA_NULL); + + VMA_VALIDATE(found); + calculatedFreeSize += prev->size; + } + else + { + ++allocCount; + // Check if taken block is not on a free list + Block* freeBlock = m_FreeList[listIndex]; + while (freeBlock) + { + VMA_VALIDATE(freeBlock != prev); + freeBlock = freeBlock->NextFree(); + } + + if (!IsVirtual()) + { + VMA_VALIDATE(m_GranularityHandler.Validate(validateCtx, prev->offset, prev->size)); + } + } + + if (prev->prevPhysical) { - const VkDeviceSize childrenNodeSize = levelNodeSize / 2; - const Node* const leftChild = node->split.leftChild; - PrintDetailedMapNode(json, leftChild, childrenNodeSize); - const Node* const rightChild = leftChild->buddy; - PrintDetailedMapNode(json, rightChild, childrenNodeSize); + VMA_VALIDATE(prev->prevPhysical->nextPhysical == prev); } + } + + if (!IsVirtual()) + { + VMA_VALIDATE(m_GranularityHandler.FinishValidation(validateCtx)); + } + + VMA_VALIDATE(nextOffset == 0); + VMA_VALIDATE(calculatedSize == GetSize()); + VMA_VALIDATE(calculatedFreeSize == GetSumFreeSize()); + VMA_VALIDATE(allocCount == m_AllocCount); + VMA_VALIDATE(freeCount == m_BlocksFreeCount); + + return true; +} + +void VmaBlockMetadata_TLSF::CalcAllocationStatInfo(VmaStatInfo& outInfo) const +{ + VmaInitStatInfo(outInfo); + outInfo.blockCount = 1; + if (m_NullBlock->size > 0) + VmaAddStatInfoUnusedRange(outInfo, m_NullBlock->size); + + for (Block* block = m_NullBlock->prevPhysical; block != VMA_NULL; block = block->prevPhysical) + { + if (block->IsFree()) + VmaAddStatInfoUnusedRange(outInfo, block->size); + else + VmaAddStatInfoAllocation(outInfo, block->size); + } +} + +void VmaBlockMetadata_TLSF::AddPoolStats(VmaPoolStats& inoutStats) const +{ + inoutStats.size += GetSize(); + inoutStats.unusedSize += GetSumFreeSize(); + inoutStats.allocationCount += m_AllocCount; + inoutStats.unusedRangeCount += m_BlocksFreeCount; + if(m_NullBlock->size > 0) + ++inoutStats.unusedRangeCount; +} + +#if VMA_STATS_STRING_ENABLED +void VmaBlockMetadata_TLSF::PrintDetailedMap(class VmaJsonWriter& json) const +{ + size_t blockCount = m_AllocCount + m_BlocksFreeCount; + VmaStlAllocator<Block*> allocator(GetAllocationCallbacks()); + VmaVector<Block*, VmaStlAllocator<Block*>> blockList(blockCount, allocator); + + size_t i = blockCount; + for (Block* block = m_NullBlock->prevPhysical; block != VMA_NULL; block = block->prevPhysical) + { + blockList[--i] = block; + } + VMA_ASSERT(i == 0); + + VmaStatInfo stat; + CalcAllocationStatInfo(stat); + + PrintDetailedMap_Begin(json, + stat.unusedBytes, + stat.allocationCount, + stat.unusedRangeCount); + + for (; i < blockCount; ++i) + { + Block* block = blockList[i]; + if (block->IsFree()) + PrintDetailedMap_UnusedRange(json, block->offset, block->size); + else + PrintDetailedMap_Allocation(json, block->offset, block->size, block->UserData()); + } + if (m_NullBlock->size > 0) + PrintDetailedMap_UnusedRange(json, m_NullBlock->offset, m_NullBlock->size); + + PrintDetailedMap_End(json); +} +#endif + +bool VmaBlockMetadata_TLSF::CreateAllocationRequest( + VkDeviceSize allocSize, + VkDeviceSize allocAlignment, + bool upperAddress, + VmaSuballocationType allocType, + uint32_t strategy, + VmaAllocationRequest* pAllocationRequest) +{ + VMA_ASSERT(allocSize > 0 && "Cannot allocate empty block!"); + VMA_ASSERT(!upperAddress && "VMA_ALLOCATION_CREATE_UPPER_ADDRESS_BIT can be used only with linear algorithm."); + + // For small granularity round up + if (!IsVirtual()) + m_GranularityHandler.RoundupAllocRequest(allocType, allocSize, allocAlignment); + + allocSize += GetDebugMargin(); + // Quick check for too small pool + if (allocSize > GetSumFreeSize()) + return false; + + // If no free blocks in pool then check only null block + if (m_BlocksFreeCount == 0) + return CheckBlock(*m_NullBlock, m_ListsCount, allocSize, allocAlignment, allocType, pAllocationRequest); + + // Round up to the next block + VkDeviceSize sizeForNextList = allocSize; + VkDeviceSize smallSizeStep = SMALL_BUFFER_SIZE / (IsVirtual() ? 1 << SECOND_LEVEL_INDEX : 4); + if (allocSize > SMALL_BUFFER_SIZE) + { + sizeForNextList += (1ULL << (VMA_BITSCAN_MSB(allocSize) - SECOND_LEVEL_INDEX)); + } + else if (allocSize > SMALL_BUFFER_SIZE - smallSizeStep) + sizeForNextList = SMALL_BUFFER_SIZE + 1; + else + sizeForNextList += smallSizeStep; + + uint32_t nextListIndex = 0; + uint32_t prevListIndex = 0; + Block* nextListBlock = VMA_NULL; + Block* prevListBlock = VMA_NULL; + + // Check blocks according to strategies + if (strategy & VMA_ALLOCATION_CREATE_STRATEGY_MIN_TIME_BIT) + { + // Quick check for larger block first + nextListBlock = FindFreeBlock(sizeForNextList, nextListIndex); + if (nextListBlock != VMA_NULL && CheckBlock(*nextListBlock, nextListIndex, allocSize, allocAlignment, allocType, pAllocationRequest)) + return true; + + // If not fitted then null block + if (CheckBlock(*m_NullBlock, m_ListsCount, allocSize, allocAlignment, allocType, pAllocationRequest)) + return true; + + // Null block failed, search larger bucket + while (nextListBlock) + { + if (CheckBlock(*nextListBlock, nextListIndex, allocSize, allocAlignment, allocType, pAllocationRequest)) + return true; + nextListBlock = nextListBlock->NextFree(); + } + + // Failed again, check best fit bucket + prevListBlock = FindFreeBlock(allocSize, prevListIndex); + while (prevListBlock) + { + if (CheckBlock(*prevListBlock, prevListIndex, allocSize, allocAlignment, allocType, pAllocationRequest)) + return true; + prevListBlock = prevListBlock->NextFree(); + } + } + else if (strategy & VMA_ALLOCATION_CREATE_STRATEGY_MIN_MEMORY_BIT) + { + // Check best fit bucket + prevListBlock = FindFreeBlock(allocSize, prevListIndex); + while (prevListBlock) + { + if (CheckBlock(*prevListBlock, prevListIndex, allocSize, allocAlignment, allocType, pAllocationRequest)) + return true; + prevListBlock = prevListBlock->NextFree(); + } + + // If failed check null block + if (CheckBlock(*m_NullBlock, m_ListsCount, allocSize, allocAlignment, allocType, pAllocationRequest)) + return true; + + // Check larger bucket + nextListBlock = FindFreeBlock(sizeForNextList, nextListIndex); + while (nextListBlock) + { + if (CheckBlock(*nextListBlock, nextListIndex, allocSize, allocAlignment, allocType, pAllocationRequest)) + return true; + nextListBlock = nextListBlock->NextFree(); + } + } + else + { + // Check larger bucket + nextListBlock = FindFreeBlock(sizeForNextList, nextListIndex); + while (nextListBlock) + { + if (CheckBlock(*nextListBlock, nextListIndex, allocSize, allocAlignment, allocType, pAllocationRequest)) + return true; + nextListBlock = nextListBlock->NextFree(); + } + + // If failed check null block + if (CheckBlock(*m_NullBlock, m_ListsCount, allocSize, allocAlignment, allocType, pAllocationRequest)) + return true; + + // Check best fit bucket + prevListBlock = FindFreeBlock(allocSize, prevListIndex); + while (prevListBlock) + { + if (CheckBlock(*prevListBlock, prevListIndex, allocSize, allocAlignment, allocType, pAllocationRequest)) + return true; + prevListBlock = prevListBlock->NextFree(); + } + } + + // Worst case, full search has to be done + while (++nextListIndex < m_ListsCount) + { + nextListBlock = m_FreeList[nextListIndex]; + while (nextListBlock) + { + if (CheckBlock(*nextListBlock, nextListIndex, allocSize, allocAlignment, allocType, pAllocationRequest)) + return true; + nextListBlock = nextListBlock->NextFree(); + } + } + + // No more memory sadly + return false; +} + +VkResult VmaBlockMetadata_TLSF::CheckCorruption(const void* pBlockData) +{ + for (Block* block = m_NullBlock->prevPhysical; block != VMA_NULL; block = block->prevPhysical) + { + if (!block->IsFree()) + { + if (!VmaValidateMagicValue(pBlockData, block->offset + block->size)) + { + VMA_ASSERT(0 && "MEMORY CORRUPTION DETECTED AFTER VALIDATED ALLOCATION!"); + return VK_ERROR_UNKNOWN; + } + } + } + + return VK_SUCCESS; +} + +void VmaBlockMetadata_TLSF::Alloc( + const VmaAllocationRequest& request, + VmaSuballocationType type, + void* userData) +{ + VMA_ASSERT(request.type == VmaAllocationRequestType::TLSF); + + // Get block and pop it from the free list + Block* currentBlock = (Block*)request.allocHandle; + VkDeviceSize offset = request.algorithmData; + VMA_ASSERT(currentBlock != VMA_NULL); + VMA_ASSERT(currentBlock->offset <= offset); + + if (currentBlock != m_NullBlock) + RemoveFreeBlock(currentBlock); + + VkDeviceSize debugMargin = GetDebugMargin(); + VkDeviceSize misssingAlignment = offset - currentBlock->offset; + + // Append missing alignment to prev block or create new one + if (misssingAlignment) + { + Block* prevBlock = currentBlock->prevPhysical; + VMA_ASSERT(prevBlock != VMA_NULL && "There should be no missing alignment at offset 0!"); + + if (prevBlock->IsFree() && prevBlock->size != debugMargin) + { + uint32_t oldList = GetListIndex(prevBlock->size); + prevBlock->size += misssingAlignment; + // Check if new size crosses list bucket + if (oldList != GetListIndex(prevBlock->size)) + { + prevBlock->size -= misssingAlignment; + RemoveFreeBlock(prevBlock); + prevBlock->size += misssingAlignment; + InsertFreeBlock(prevBlock); + } + else + m_BlocksFreeSize += misssingAlignment; + } + else + { + Block* newBlock = m_BlockAllocator.Alloc(); + currentBlock->prevPhysical = newBlock; + prevBlock->nextPhysical = newBlock; + newBlock->prevPhysical = prevBlock; + newBlock->nextPhysical = currentBlock; + newBlock->size = misssingAlignment; + newBlock->offset = currentBlock->offset; + newBlock->MarkTaken(); + + InsertFreeBlock(newBlock); + } + + currentBlock->size -= misssingAlignment; + currentBlock->offset += misssingAlignment; + } + + VkDeviceSize size = request.size + debugMargin; + if (currentBlock->size == size) + { + if (currentBlock == m_NullBlock) + { + // Setup new null block + m_NullBlock = m_BlockAllocator.Alloc(); + m_NullBlock->size = 0; + m_NullBlock->offset = currentBlock->offset + size; + m_NullBlock->prevPhysical = currentBlock; + m_NullBlock->nextPhysical = VMA_NULL; + m_NullBlock->MarkFree(); + m_NullBlock->PrevFree() = VMA_NULL; + m_NullBlock->NextFree() = VMA_NULL; + currentBlock->nextPhysical = m_NullBlock; + currentBlock->MarkTaken(); + } + } + else + { + VMA_ASSERT(currentBlock->size > size && "Proper block already found, shouldn't find smaller one!"); + + // Create new free block + Block* newBlock = m_BlockAllocator.Alloc(); + newBlock->size = currentBlock->size - size; + newBlock->offset = currentBlock->offset + size; + newBlock->prevPhysical = currentBlock; + newBlock->nextPhysical = currentBlock->nextPhysical; + currentBlock->nextPhysical = newBlock; + currentBlock->size = size; + + if (currentBlock == m_NullBlock) + { + m_NullBlock = newBlock; + m_NullBlock->MarkFree(); + m_NullBlock->NextFree() = VMA_NULL; + m_NullBlock->PrevFree() = VMA_NULL; + currentBlock->MarkTaken(); + } + else + { + newBlock->nextPhysical->prevPhysical = newBlock; + newBlock->MarkTaken(); + InsertFreeBlock(newBlock); + } + } + currentBlock->UserData() = userData; + + if (debugMargin > 0) + { + currentBlock->size -= debugMargin; + Block* newBlock = m_BlockAllocator.Alloc(); + newBlock->size = debugMargin; + newBlock->offset = currentBlock->offset + currentBlock->size; + newBlock->prevPhysical = currentBlock; + newBlock->nextPhysical = currentBlock->nextPhysical; + newBlock->MarkTaken(); + currentBlock->nextPhysical->prevPhysical = newBlock; + currentBlock->nextPhysical = newBlock; + InsertFreeBlock(newBlock); + } + + if (!IsVirtual()) + m_GranularityHandler.AllocPages((uint8_t)(uintptr_t)request.customData, + currentBlock->offset, currentBlock->size); + ++m_AllocCount; +} + +void VmaBlockMetadata_TLSF::Free(VmaAllocHandle allocHandle) +{ + Block* block = (Block*)allocHandle; + Block* next = block->nextPhysical; + VMA_ASSERT(!block->IsFree() && "Block is already free!"); + + if (!IsVirtual()) + m_GranularityHandler.FreePages(block->offset, block->size); + --m_AllocCount; + + VkDeviceSize debugMargin = GetDebugMargin(); + if (debugMargin > 0) + { + RemoveFreeBlock(next); + MergeBlock(next, block); + block = next; + next = next->nextPhysical; + } + + // Try merging + Block* prev = block->prevPhysical; + if (prev != VMA_NULL && prev->IsFree() && prev->size != debugMargin) + { + RemoveFreeBlock(prev); + MergeBlock(block, prev); + } + + if (!next->IsFree()) + InsertFreeBlock(block); + else if (next == m_NullBlock) + MergeBlock(m_NullBlock, block); + else + { + RemoveFreeBlock(next); + MergeBlock(next, block); + InsertFreeBlock(next); + } +} + +void VmaBlockMetadata_TLSF::GetAllocationInfo(VmaAllocHandle allocHandle, VmaVirtualAllocationInfo& outInfo) +{ + Block* block = (Block*)allocHandle; + VMA_ASSERT(!block->IsFree() && "Cannot get allocation info for free block!"); + outInfo.offset = block->offset; + outInfo.size = block->size; + outInfo.pUserData = block->UserData(); +} + +void VmaBlockMetadata_TLSF::Clear() +{ + m_AllocCount = 0; + m_BlocksFreeCount = 0; + m_BlocksFreeSize = 0; + m_IsFreeBitmap = 0; + m_NullBlock->offset = 0; + m_NullBlock->size = GetSize(); + Block* block = m_NullBlock->prevPhysical; + m_NullBlock->prevPhysical = VMA_NULL; + while (block) + { + Block* prev = block->prevPhysical; + m_BlockAllocator.Free(block); + block = prev; + } + memset(m_FreeList, 0, m_ListsCount * sizeof(Block*)); + memset(m_InnerIsFreeBitmap, 0, m_MemoryClasses * sizeof(uint32_t)); + m_GranularityHandler.Clear(); +} + +void VmaBlockMetadata_TLSF::SetAllocationUserData(VmaAllocHandle allocHandle, void* userData) +{ + Block* block = (Block*)allocHandle; + VMA_ASSERT(!block->IsFree() && "Trying to set user data for not allocated block!"); + block->UserData() = userData; +} + +void VmaBlockMetadata_TLSF::DebugLogAllAllocations() const +{ + for (Block* block = m_NullBlock->prevPhysical; block != VMA_NULL; block = block->prevPhysical) + if (!block->IsFree()) + DebugLogAllocation(block->offset, block->size, block->UserData()); +} + +uint8_t VmaBlockMetadata_TLSF::SizeToMemoryClass(VkDeviceSize size) const +{ + if (size > SMALL_BUFFER_SIZE) + return VMA_BITSCAN_MSB(size) - MEMORY_CLASS_SHIFT; + return 0; +} + +uint16_t VmaBlockMetadata_TLSF::SizeToSecondIndex(VkDeviceSize size, uint8_t memoryClass) const +{ + if (memoryClass == 0) + { + if (IsVirtual()) + return static_cast<uint16_t>((size - 1) / 8); + else + return static_cast<uint16_t>((size - 1) / 64); + } + return static_cast<uint16_t>((size >> (memoryClass + MEMORY_CLASS_SHIFT - SECOND_LEVEL_INDEX)) ^ (1U << SECOND_LEVEL_INDEX)); +} + +uint32_t VmaBlockMetadata_TLSF::GetListIndex(uint8_t memoryClass, uint16_t secondIndex) const +{ + if (memoryClass == 0) + return secondIndex; + + const uint32_t index = static_cast<uint32_t>(memoryClass - 1) * (1 << SECOND_LEVEL_INDEX) + secondIndex; + if (IsVirtual()) + return index + (1 << SECOND_LEVEL_INDEX); + else + return index + 4; +} + +uint32_t VmaBlockMetadata_TLSF::GetListIndex(VkDeviceSize size) const +{ + uint8_t memoryClass = SizeToMemoryClass(size); + return GetListIndex(memoryClass, SizeToSecondIndex(size, memoryClass)); +} + +void VmaBlockMetadata_TLSF::RemoveFreeBlock(Block* block) +{ + VMA_ASSERT(block != m_NullBlock); + VMA_ASSERT(block->IsFree()); + + if (block->NextFree() != VMA_NULL) + block->NextFree()->PrevFree() = block->PrevFree(); + if (block->PrevFree() != VMA_NULL) + block->PrevFree()->NextFree() = block->NextFree(); + else + { + uint8_t memClass = SizeToMemoryClass(block->size); + uint16_t secondIndex = SizeToSecondIndex(block->size, memClass); + uint32_t index = GetListIndex(memClass, secondIndex); + VMA_ASSERT(m_FreeList[index] == block); + m_FreeList[index] = block->NextFree(); + if (block->NextFree() == VMA_NULL) + { + m_InnerIsFreeBitmap[memClass] &= ~(1U << secondIndex); + if (m_InnerIsFreeBitmap[memClass] == 0) + m_IsFreeBitmap &= ~(1UL << memClass); + } + } + block->MarkTaken(); + block->UserData() = VMA_NULL; + --m_BlocksFreeCount; + m_BlocksFreeSize -= block->size; +} + +void VmaBlockMetadata_TLSF::InsertFreeBlock(Block* block) +{ + VMA_ASSERT(block != m_NullBlock); + VMA_ASSERT(!block->IsFree() && "Cannot insert block twice!"); + + uint8_t memClass = SizeToMemoryClass(block->size); + uint16_t secondIndex = SizeToSecondIndex(block->size, memClass); + uint32_t index = GetListIndex(memClass, secondIndex); + VMA_ASSERT(index < m_ListsCount); + block->PrevFree() = VMA_NULL; + block->NextFree() = m_FreeList[index]; + m_FreeList[index] = block; + if (block->NextFree() != VMA_NULL) + block->NextFree()->PrevFree() = block; + else + { + m_InnerIsFreeBitmap[memClass] |= 1U << secondIndex; + m_IsFreeBitmap |= 1UL << memClass; + } + ++m_BlocksFreeCount; + m_BlocksFreeSize += block->size; +} + +void VmaBlockMetadata_TLSF::MergeBlock(Block* block, Block* prev) +{ + VMA_ASSERT(block->prevPhysical == prev && "Cannot merge seperate physical regions!"); + VMA_ASSERT(!prev->IsFree() && "Cannot merge block that belongs to free list!"); + + block->offset = prev->offset; + block->size += prev->size; + block->prevPhysical = prev->prevPhysical; + if (block->prevPhysical) + block->prevPhysical->nextPhysical = block; + m_BlockAllocator.Free(prev); +} + +VmaBlockMetadata_TLSF::Block* VmaBlockMetadata_TLSF::FindFreeBlock(VkDeviceSize size, uint32_t& listIndex) const +{ + uint8_t memoryClass = SizeToMemoryClass(size); + uint32_t innerFreeMap = m_InnerIsFreeBitmap[memoryClass] & (~0U << SizeToSecondIndex(size, memoryClass)); + if (!innerFreeMap) + { + // Check higher levels for avaiable blocks + uint32_t freeMap = m_IsFreeBitmap & (~0UL << (memoryClass + 1)); + if (!freeMap) + return VMA_NULL; // No more memory avaible + + // Find lowest free region + memoryClass = VMA_BITSCAN_LSB(freeMap); + innerFreeMap = m_InnerIsFreeBitmap[memoryClass]; + VMA_ASSERT(innerFreeMap != 0); + } + // Find lowest free subregion + listIndex = GetListIndex(memoryClass, VMA_BITSCAN_LSB(innerFreeMap)); + VMA_ASSERT(m_FreeList[listIndex]); + return m_FreeList[listIndex]; +} + +bool VmaBlockMetadata_TLSF::CheckBlock( + Block& block, + uint32_t listIndex, + VkDeviceSize allocSize, + VkDeviceSize allocAlignment, + VmaSuballocationType allocType, + VmaAllocationRequest* pAllocationRequest) +{ + VMA_ASSERT(block.IsFree() && "Block is already taken!"); + + VkDeviceSize alignedOffset = VmaAlignUp(block.offset, allocAlignment); + if (block.size < allocSize + alignedOffset - block.offset) + return false; + + // Check for granularity conflicts + if (!IsVirtual() && + m_GranularityHandler.CheckConflictAndAlignUp(alignedOffset, allocSize, block.offset, block.size, allocType)) + return false; + + // Alloc successful + pAllocationRequest->type = VmaAllocationRequestType::TLSF; + pAllocationRequest->allocHandle = (VmaAllocHandle)█ + pAllocationRequest->size = allocSize - GetDebugMargin(); + pAllocationRequest->customData = (void*)allocType; + pAllocationRequest->algorithmData = alignedOffset; + + // Place block at the start of list if it's normal block + if (listIndex != m_ListsCount && block.PrevFree()) + { + block.PrevFree()->NextFree() = block.NextFree(); + if (block.NextFree()) + block.NextFree()->PrevFree() = block.PrevFree(); + block.PrevFree() = VMA_NULL; + block.NextFree() = m_FreeList[listIndex]; + m_FreeList[listIndex] = █ + if (block.NextFree()) + block.NextFree()->PrevFree() = █ + } + + return true; +} +#endif // _VMA_BLOCK_METADATA_TLSF_FUNCTIONS +#endif // _VMA_BLOCK_METADATA_TLSF + +#ifndef _VMA_BLOCK_VECTOR +/* +Sequence of VmaDeviceMemoryBlock. Represents memory blocks allocated for a specific +Vulkan memory type. + +Synchronized internally with a mutex. +*/ +class VmaBlockVector +{ + friend class VmaDefragmentationAlgorithm_Generic; + VMA_CLASS_NO_COPY(VmaBlockVector) +public: + VmaBlockVector( + VmaAllocator hAllocator, + VmaPool hParentPool, + uint32_t memoryTypeIndex, + VkDeviceSize preferredBlockSize, + size_t minBlockCount, + size_t maxBlockCount, + VkDeviceSize bufferImageGranularity, + bool explicitBlockSize, + uint32_t algorithm, + float priority, + VkDeviceSize minAllocationAlignment, + void* pMemoryAllocateNext); + ~VmaBlockVector(); + + VmaAllocator GetAllocator() const { return m_hAllocator; } + VmaPool GetParentPool() const { return m_hParentPool; } + bool IsCustomPool() const { return m_hParentPool != VMA_NULL; } + uint32_t GetMemoryTypeIndex() const { return m_MemoryTypeIndex; } + VkDeviceSize GetPreferredBlockSize() const { return m_PreferredBlockSize; } + VkDeviceSize GetBufferImageGranularity() const { return m_BufferImageGranularity; } + uint32_t GetAlgorithm() const { return m_Algorithm; } + bool HasExplicitBlockSize() const { return m_ExplicitBlockSize; } + float GetPriority() const { return m_Priority; } + void* const GetAllocationNextPtr() const { return m_pMemoryAllocateNext; } + + VkResult CreateMinBlocks(); + void AddPoolStats(VmaPoolStats* pStats); + bool IsEmpty(); + bool IsCorruptionDetectionEnabled() const; + + VkResult Allocate( + VkDeviceSize size, + VkDeviceSize alignment, + const VmaAllocationCreateInfo& createInfo, + VmaSuballocationType suballocType, + size_t allocationCount, + VmaAllocation* pAllocations); + + void Free(const VmaAllocation hAllocation); + // Adds statistics of this BlockVector to pStats. + void AddStats(VmaStats* pStats); + +#if VMA_STATS_STRING_ENABLED + void PrintDetailedMap(class VmaJsonWriter& json); +#endif + + VkResult CheckCorruption(); + + // Saves results in pCtx->res. + void Defragment( + class VmaBlockVectorDefragmentationContext* pCtx, + VmaDefragmentationStats* pStats, VmaDefragmentationFlags flags, + VkDeviceSize& maxCpuBytesToMove, uint32_t& maxCpuAllocationsToMove, + VkDeviceSize& maxGpuBytesToMove, uint32_t& maxGpuAllocationsToMove, + VkCommandBuffer commandBuffer); + void DefragmentationEnd( + class VmaBlockVectorDefragmentationContext* pCtx, + uint32_t flags, + VmaDefragmentationStats* pStats); + + uint32_t ProcessDefragmentations( + class VmaBlockVectorDefragmentationContext* pCtx, + VmaDefragmentationPassMoveInfo* pMove, uint32_t maxMoves); + + void CommitDefragmentations( + class VmaBlockVectorDefragmentationContext* pCtx, + VmaDefragmentationStats* pStats); + + //////////////////////////////////////////////////////////////////////////////// + // To be used only while the m_Mutex is locked. Used during defragmentation. + + size_t GetBlockCount() const { return m_Blocks.size(); } + VmaDeviceMemoryBlock* GetBlock(size_t index) const { return m_Blocks[index]; } + size_t CalcAllocationCount() const; + bool IsBufferImageGranularityConflictPossible() const; + +private: + const VmaAllocator m_hAllocator; + const VmaPool m_hParentPool; + const uint32_t m_MemoryTypeIndex; + const VkDeviceSize m_PreferredBlockSize; + const size_t m_MinBlockCount; + const size_t m_MaxBlockCount; + const VkDeviceSize m_BufferImageGranularity; + const bool m_ExplicitBlockSize; + const uint32_t m_Algorithm; + const float m_Priority; + const VkDeviceSize m_MinAllocationAlignment; + + void* const m_pMemoryAllocateNext; + VMA_RW_MUTEX m_Mutex; + /* There can be at most one allocation that is completely empty (except when minBlockCount > 0) - + a hysteresis to avoid pessimistic case of alternating creation and destruction of a VkDeviceMemory. */ + bool m_HasEmptyBlock; + // Incrementally sorted by sumFreeSize, ascending. + VmaVector<VmaDeviceMemoryBlock*, VmaStlAllocator<VmaDeviceMemoryBlock*>> m_Blocks; + uint32_t m_NextBlockId; + + VkDeviceSize CalcMaxBlockSize() const; + // Finds and removes given block from vector. + void Remove(VmaDeviceMemoryBlock* pBlock); + // Performs single step in sorting m_Blocks. They may not be fully sorted + // after this call. + void IncrementallySortBlocks(); + + VkResult AllocatePage( + VkDeviceSize size, + VkDeviceSize alignment, + const VmaAllocationCreateInfo& createInfo, + VmaSuballocationType suballocType, + VmaAllocation* pAllocation); + + VkResult AllocateFromBlock( + VmaDeviceMemoryBlock* pBlock, + VkDeviceSize size, + VkDeviceSize alignment, + VmaAllocationCreateFlags allocFlags, + void* pUserData, + VmaSuballocationType suballocType, + uint32_t strategy, + VmaAllocation* pAllocation); + + VkResult CreateBlock(VkDeviceSize blockSize, size_t* pNewBlockIndex); + // Saves result to pCtx->res. + void ApplyDefragmentationMovesCpu( + VmaBlockVectorDefragmentationContext* pDefragCtx, + const VmaVector<VmaDefragmentationMove, VmaStlAllocator<VmaDefragmentationMove>>& moves); + // Saves result to pCtx->res. + void ApplyDefragmentationMovesGpu( + VmaBlockVectorDefragmentationContext* pDefragCtx, + VmaVector<VmaDefragmentationMove, VmaStlAllocator<VmaDefragmentationMove>>& moves, + VkCommandBuffer commandBuffer); + + /* + Used during defragmentation. pDefragmentationStats is optional. It is in/out + - updated with new data. + */ + void FreeEmptyBlocks(VmaDefragmentationStats* pDefragmentationStats); + void UpdateHasEmptyBlock(); +}; +#endif // _VMA_BLOCK_VECTOR + +#ifndef _VMA_DEFRAGMENTATION_ALGORITHM +struct VmaDefragmentationMove +{ + size_t srcBlockIndex; + size_t dstBlockIndex; + VkDeviceSize srcOffset; + VkDeviceSize dstOffset; + VmaAllocHandle dstHandle; + VkDeviceSize size; + VmaAllocation hAllocation; + VmaDeviceMemoryBlock* pSrcBlock; + VmaDeviceMemoryBlock* pDstBlock; +}; + +/* +Performs defragmentation: + +- Updates `pBlockVector->m_pMetadata`. +- Updates allocations by calling ChangeBlockAllocation() or ChangeOffset(). +- Does not move actual data, only returns requested moves as `moves`. +*/ +class VmaDefragmentationAlgorithm +{ + VMA_CLASS_NO_COPY(VmaDefragmentationAlgorithm) +public: + VmaDefragmentationAlgorithm( + VmaAllocator hAllocator, + VmaBlockVector* pBlockVector) + : m_hAllocator(hAllocator), + m_pBlockVector(pBlockVector) {} + virtual ~VmaDefragmentationAlgorithm() = default; + + virtual void AddAllocation(VmaAllocation hAlloc, VkBool32* pChanged) = 0; + virtual void AddAll() = 0; + + virtual VkResult Defragment( + VmaVector<VmaDefragmentationMove, VmaStlAllocator<VmaDefragmentationMove>>& moves, + VkDeviceSize maxBytesToMove, + uint32_t maxAllocationsToMove, + VmaDefragmentationFlags flags) = 0; + + virtual VkDeviceSize GetBytesMoved() const = 0; + virtual uint32_t GetAllocationsMoved() const = 0; + +protected: + struct AllocationInfo + { + VmaAllocation m_hAllocation; + VkBool32* m_pChanged; + + AllocationInfo() : m_hAllocation(VK_NULL_HANDLE), m_pChanged(VMA_NULL) {} + AllocationInfo(VmaAllocation hAlloc, VkBool32* pChanged) : m_hAllocation(hAlloc), m_pChanged(pChanged) {} + }; + + VmaAllocator const m_hAllocator; + VmaBlockVector* const m_pBlockVector; +}; + +#endif // _VMA_DEFRAGMENTATION_ALGORITHM + +#ifndef _VMA_DEFRAGMENTATION_ALGORITHM_GENERIC +class VmaDefragmentationAlgorithm_Generic : public VmaDefragmentationAlgorithm +{ + VMA_CLASS_NO_COPY(VmaDefragmentationAlgorithm_Generic) +public: + VmaDefragmentationAlgorithm_Generic( + VmaAllocator hAllocator, + VmaBlockVector* pBlockVector, + bool overlappingMoveSupported); + virtual ~VmaDefragmentationAlgorithm_Generic(); + + virtual void AddAll() { m_AllAllocations = true; } + virtual VkDeviceSize GetBytesMoved() const { return m_BytesMoved; } + virtual uint32_t GetAllocationsMoved() const { return m_AllocationsMoved; } + + virtual void AddAllocation(VmaAllocation hAlloc, VkBool32* pChanged); + virtual VkResult Defragment( + VmaVector<VmaDefragmentationMove, VmaStlAllocator<VmaDefragmentationMove>>& moves, + VkDeviceSize maxBytesToMove, + uint32_t maxAllocationsToMove, + VmaDefragmentationFlags flags); + +private: + struct AllocationInfoSizeGreater + { + bool operator()(const AllocationInfo& lhs, const AllocationInfo& rhs) const; + }; + struct AllocationInfoOffsetGreater + { + bool operator()(const AllocationInfo& lhs, const AllocationInfo& rhs) const; + }; + struct BlockInfo + { + size_t m_OriginalBlockIndex; + VmaDeviceMemoryBlock* m_pBlock; + bool m_HasNonMovableAllocations; + VmaVector<AllocationInfo, VmaStlAllocator<AllocationInfo>> m_Allocations; + + BlockInfo(const VkAllocationCallbacks* pAllocationCallbacks); + + void CalcHasNonMovableAllocations(); + void SortAllocationsBySizeDescending(); + void SortAllocationsByOffsetDescending(); + }; + struct BlockPointerLess + { + bool operator()(const BlockInfo* pLhsBlockInfo, const VmaDeviceMemoryBlock* pRhsBlock) const; + bool operator()(const BlockInfo* pLhsBlockInfo, const BlockInfo* pRhsBlockInfo) const; + }; + // 1. Blocks with some non-movable allocations go first. + // 2. Blocks with smaller sumFreeSize go first. + struct BlockInfoCompareMoveDestination + { + bool operator()(const BlockInfo* pLhsBlockInfo, const BlockInfo* pRhsBlockInfo) const; + }; + typedef VmaVector<BlockInfo*, VmaStlAllocator<BlockInfo*>> BlockInfoVector; + + BlockInfoVector m_Blocks; + uint32_t m_AllocationCount; + bool m_AllAllocations; + VkDeviceSize m_BytesMoved; + uint32_t m_AllocationsMoved; + + static bool MoveMakesSense( + size_t dstBlockIndex, VkDeviceSize dstOffset, + size_t srcBlockIndex, VkDeviceSize srcOffset); + + size_t CalcBlocksWithNonMovableCount() const; + VkResult DefragmentRound( + VmaVector<VmaDefragmentationMove, VmaStlAllocator<VmaDefragmentationMove>>& moves, + VkDeviceSize maxBytesToMove, + uint32_t maxAllocationsToMove, + bool freeOldAllocations); +}; +#endif // _VMA_DEFRAGMENTATION_ALGORITHM_GENERIC + +#ifndef _VMA_DEFRAGMENTATION_ALGORITHM_FAST +class VmaDefragmentationAlgorithm_Fast : public VmaDefragmentationAlgorithm +{ + VMA_CLASS_NO_COPY(VmaDefragmentationAlgorithm_Fast) +public: + VmaDefragmentationAlgorithm_Fast( + VmaAllocator hAllocator, + VmaBlockVector* pBlockVector, + bool overlappingMoveSupported); + virtual ~VmaDefragmentationAlgorithm_Fast() = default; + + virtual void AddAll() { m_AllAllocations = true; } + virtual VkDeviceSize GetBytesMoved() const { return m_BytesMoved; } + virtual uint32_t GetAllocationsMoved() const { return m_AllocationsMoved; } + virtual void AddAllocation(VmaAllocation hAlloc, VkBool32* pChanged) { ++m_AllocationCount; } + + virtual VkResult Defragment( + VmaVector< VmaDefragmentationMove, VmaStlAllocator<VmaDefragmentationMove> >& moves, + VkDeviceSize maxBytesToMove, + uint32_t maxAllocationsToMove, + VmaDefragmentationFlags flags); + +private: + struct BlockInfo + { + size_t origBlockIndex; + }; + class FreeSpaceDatabase + { + public: + FreeSpaceDatabase(); + + void Register(size_t blockInfoIndex, VkDeviceSize offset, VkDeviceSize size); + bool Fetch(VkDeviceSize alignment, VkDeviceSize size, + size_t& outBlockInfoIndex, VkDeviceSize& outDstOffset); + + private: + static const size_t MAX_COUNT = 4; + + struct FreeSpace + { + size_t blockInfoIndex; // SIZE_MAX means this structure is invalid. + VkDeviceSize offset; + VkDeviceSize size; + } m_FreeSpaces[MAX_COUNT]; + }; + + const bool m_OverlappingMoveSupported; + + uint32_t m_AllocationCount; + bool m_AllAllocations; + VkDeviceSize m_BytesMoved; + uint32_t m_AllocationsMoved; + + VmaVector<BlockInfo, VmaStlAllocator<BlockInfo>> m_BlockInfos; + + void PreprocessMetadata(); + void PostprocessMetadata(); + void InsertSuballoc(VmaBlockMetadata_Generic* pMetadata, const VmaSuballocation& suballoc); +}; +#endif // _VMA_DEFRAGMENTATION_ALGORITHM_FAST + +#ifndef _VMA_BLOCK_VECTOR_DEFRAGMENTATION_CONTEXT +struct VmaBlockDefragmentationContext +{ + enum BLOCK_FLAG + { + BLOCK_FLAG_USED = 0x00000001, + }; + uint32_t flags; + VkBuffer hBuffer; +}; + +class VmaBlockVectorDefragmentationContext +{ + VMA_CLASS_NO_COPY(VmaBlockVectorDefragmentationContext) +public: + VkResult res; + bool mutexLocked; + VmaVector<VmaBlockDefragmentationContext, VmaStlAllocator<VmaBlockDefragmentationContext>> blockContexts; + VmaVector<VmaDefragmentationMove, VmaStlAllocator<VmaDefragmentationMove>> defragmentationMoves; + uint32_t defragmentationMovesProcessed; + uint32_t defragmentationMovesCommitted; + bool hasDefragmentationPlan; + + VmaBlockVectorDefragmentationContext( + VmaAllocator hAllocator, + VmaPool hCustomPool, // Optional. + VmaBlockVector* pBlockVector); + ~VmaBlockVectorDefragmentationContext(); + + VmaPool GetCustomPool() const { return m_hCustomPool; } + VmaBlockVector* GetBlockVector() const { return m_pBlockVector; } + VmaDefragmentationAlgorithm* GetAlgorithm() const { return m_pAlgorithm; } + void AddAll() { m_AllAllocations = true; } + + void AddAllocation(VmaAllocation hAlloc, VkBool32* pChanged); + void Begin(bool overlappingMoveSupported, VmaDefragmentationFlags flags); + +private: + struct AllocInfo + { + VmaAllocation hAlloc; + VkBool32* pChanged; + }; + + const VmaAllocator m_hAllocator; + // Null if not from custom pool. + const VmaPool m_hCustomPool; + // Redundant, for convenience not to fetch from m_hCustomPool->m_BlockVector or m_hAllocator->m_pBlockVectors. + VmaBlockVector* const m_pBlockVector; + // Owner of this object. + VmaDefragmentationAlgorithm* m_pAlgorithm; + // Used between constructor and Begin. + VmaVector<AllocInfo, VmaStlAllocator<AllocInfo>> m_Allocations; + bool m_AllAllocations; +}; +#endif // _VMA_BLOCK_VECTOR_DEFRAGMENTATION_CONTEXT + +#ifndef _VMA_DEFRAGMENTATION_CONTEXT +struct VmaDefragmentationContext_T +{ +private: + VMA_CLASS_NO_COPY(VmaDefragmentationContext_T) +public: + VmaDefragmentationContext_T( + VmaAllocator hAllocator, + uint32_t flags, + VmaDefragmentationStats* pStats); + ~VmaDefragmentationContext_T(); + + void AddPools(uint32_t poolCount, const VmaPool* pPools); + void AddAllocations( + uint32_t allocationCount, + const VmaAllocation* pAllocations, + VkBool32* pAllocationsChanged); + + /* + Returns: + - `VK_SUCCESS` if succeeded and object can be destroyed immediately. + - `VK_NOT_READY` if succeeded but the object must remain alive until vmaDefragmentationEnd(). + - Negative value if error occurred and object can be destroyed immediately. + */ + VkResult Defragment( + VkDeviceSize maxCpuBytesToMove, uint32_t maxCpuAllocationsToMove, + VkDeviceSize maxGpuBytesToMove, uint32_t maxGpuAllocationsToMove, + VkCommandBuffer commandBuffer, VmaDefragmentationStats* pStats, VmaDefragmentationFlags flags); + + VkResult DefragmentPassBegin(VmaDefragmentationPassInfo* pInfo); + VkResult DefragmentPassEnd(); + +private: + const VmaAllocator m_hAllocator; + const uint32_t m_Flags; + VmaDefragmentationStats* const m_pStats; + + VkDeviceSize m_MaxCpuBytesToMove; + uint32_t m_MaxCpuAllocationsToMove; + VkDeviceSize m_MaxGpuBytesToMove; + uint32_t m_MaxGpuAllocationsToMove; + + // Owner of these objects. + VmaBlockVectorDefragmentationContext* m_DefaultPoolContexts[VK_MAX_MEMORY_TYPES]; + // Owner of these objects. + VmaVector<VmaBlockVectorDefragmentationContext*, VmaStlAllocator<VmaBlockVectorDefragmentationContext*>> m_CustomPoolContexts; +}; +#endif // _VMA_DEFRAGMENTATION_CONTEXT + +#ifndef _VMA_POOL_T +struct VmaPool_T +{ + friend struct VmaPoolListItemTraits; + VMA_CLASS_NO_COPY(VmaPool_T) +public: + VmaBlockVector* m_pBlockVectors[VK_MAX_MEMORY_TYPES]; + VmaDedicatedAllocationList m_DedicatedAllocations[VK_MAX_MEMORY_TYPES]; + + VmaPool_T( + VmaAllocator hAllocator, + const VmaPoolCreateInfo& createInfo); + ~VmaPool_T(); + + uint32_t GetId() const { return m_Id; } + void SetId(uint32_t id) { VMA_ASSERT(m_Id == 0); m_Id = id; } + + const char* GetName() const { return m_Name; } + void SetName(const char* pName); + +#if VMA_STATS_STRING_ENABLED + //void PrintDetailedMap(class VmaStringBuilder& sb); +#endif + +private: + const VmaAllocator m_hAllocator; + uint32_t m_Id; + char* m_Name; + VmaPool_T* m_PrevPool = VMA_NULL; + VmaPool_T* m_NextPool = VMA_NULL; +}; + +struct VmaPoolListItemTraits +{ + typedef VmaPool_T ItemType; + + static ItemType* GetPrev(const ItemType* item) { return item->m_PrevPool; } + static ItemType* GetNext(const ItemType* item) { return item->m_NextPool; } + static ItemType*& AccessPrev(ItemType* item) { return item->m_PrevPool; } + static ItemType*& AccessNext(ItemType* item) { return item->m_NextPool; } +}; +#endif // _VMA_POOL_T + +#ifndef _VMA_CURRENT_BUDGET_DATA +struct VmaCurrentBudgetData +{ + VMA_ATOMIC_UINT64 m_BlockBytes[VK_MAX_MEMORY_HEAPS]; + VMA_ATOMIC_UINT64 m_AllocationBytes[VK_MAX_MEMORY_HEAPS]; + +#if VMA_MEMORY_BUDGET + VMA_ATOMIC_UINT32 m_OperationsSinceBudgetFetch; + VMA_RW_MUTEX m_BudgetMutex; + uint64_t m_VulkanUsage[VK_MAX_MEMORY_HEAPS]; + uint64_t m_VulkanBudget[VK_MAX_MEMORY_HEAPS]; + uint64_t m_BlockBytesAtBudgetFetch[VK_MAX_MEMORY_HEAPS]; +#endif // VMA_MEMORY_BUDGET + + VmaCurrentBudgetData(); + + void AddAllocation(uint32_t heapIndex, VkDeviceSize allocationSize); + void RemoveAllocation(uint32_t heapIndex, VkDeviceSize allocationSize); +}; + +#ifndef _VMA_CURRENT_BUDGET_DATA_FUNCTIONS +VmaCurrentBudgetData::VmaCurrentBudgetData() +{ + for (uint32_t heapIndex = 0; heapIndex < VK_MAX_MEMORY_HEAPS; ++heapIndex) + { + m_BlockBytes[heapIndex] = 0; + m_AllocationBytes[heapIndex] = 0; +#if VMA_MEMORY_BUDGET + m_VulkanUsage[heapIndex] = 0; + m_VulkanBudget[heapIndex] = 0; + m_BlockBytesAtBudgetFetch[heapIndex] = 0; +#endif + } + +#if VMA_MEMORY_BUDGET + m_OperationsSinceBudgetFetch = 0; +#endif +} + +void VmaCurrentBudgetData::AddAllocation(uint32_t heapIndex, VkDeviceSize allocationSize) +{ + m_AllocationBytes[heapIndex] += allocationSize; +#if VMA_MEMORY_BUDGET + ++m_OperationsSinceBudgetFetch; +#endif +} + +void VmaCurrentBudgetData::RemoveAllocation(uint32_t heapIndex, VkDeviceSize allocationSize) +{ + VMA_ASSERT(m_AllocationBytes[heapIndex] >= allocationSize); + m_AllocationBytes[heapIndex] -= allocationSize; +#if VMA_MEMORY_BUDGET + ++m_OperationsSinceBudgetFetch; +#endif +} +#endif // _VMA_CURRENT_BUDGET_DATA_FUNCTIONS +#endif // _VMA_CURRENT_BUDGET_DATA + +#ifndef _VMA_ALLOCATION_OBJECT_ALLOCATOR +/* +Thread-safe wrapper over VmaPoolAllocator free list, for allocation of VmaAllocation_T objects. +*/ +class VmaAllocationObjectAllocator +{ + VMA_CLASS_NO_COPY(VmaAllocationObjectAllocator) +public: + VmaAllocationObjectAllocator(const VkAllocationCallbacks* pAllocationCallbacks) + : m_Allocator(pAllocationCallbacks, 1024) {} + + template<typename... Types> VmaAllocation Allocate(Types&&... args); + void Free(VmaAllocation hAlloc); + +private: + VMA_MUTEX m_Mutex; + VmaPoolAllocator<VmaAllocation_T> m_Allocator; +}; + +template<typename... Types> +VmaAllocation VmaAllocationObjectAllocator::Allocate(Types&&... args) +{ + VmaMutexLock mutexLock(m_Mutex); + return m_Allocator.Alloc<Types...>(std::forward<Types>(args)...); +} + +void VmaAllocationObjectAllocator::Free(VmaAllocation hAlloc) +{ + VmaMutexLock mutexLock(m_Mutex); + m_Allocator.Free(hAlloc); +} +#endif // _VMA_ALLOCATION_OBJECT_ALLOCATOR + +#ifndef _VMA_VIRTUAL_BLOCK_T +struct VmaVirtualBlock_T +{ + VMA_CLASS_NO_COPY(VmaVirtualBlock_T) +public: + const bool m_AllocationCallbacksSpecified; + const VkAllocationCallbacks m_AllocationCallbacks; + + VmaVirtualBlock_T(const VmaVirtualBlockCreateInfo& createInfo); + ~VmaVirtualBlock_T(); + + VkResult Init() { return VK_SUCCESS; } + bool IsEmpty() const { return m_Metadata->IsEmpty(); } + void Free(VmaVirtualAllocation allocation) { m_Metadata->Free((VmaAllocHandle)allocation); } + void SetAllocationUserData(VmaVirtualAllocation allocation, void* userData) { m_Metadata->SetAllocationUserData((VmaAllocHandle)allocation, userData); } + void Clear() { m_Metadata->Clear(); } + + const VkAllocationCallbacks* GetAllocationCallbacks() const; + void GetAllocationInfo(VmaVirtualAllocation allocation, VmaVirtualAllocationInfo& outInfo); + VkResult Allocate(const VmaVirtualAllocationCreateInfo& createInfo, VmaVirtualAllocation& outAllocation, + VkDeviceSize* outOffset); + void CalculateStats(VmaStatInfo& outStatInfo) const; +#if VMA_STATS_STRING_ENABLED + void BuildStatsString(bool detailedMap, VmaStringBuilder& sb) const; +#endif + +private: + VmaBlockMetadata* m_Metadata; +}; + +#ifndef _VMA_VIRTUAL_BLOCK_T_FUNCTIONS +VmaVirtualBlock_T::VmaVirtualBlock_T(const VmaVirtualBlockCreateInfo& createInfo) + : m_AllocationCallbacksSpecified(createInfo.pAllocationCallbacks != VMA_NULL), + m_AllocationCallbacks(createInfo.pAllocationCallbacks != VMA_NULL ? *createInfo.pAllocationCallbacks : VmaEmptyAllocationCallbacks) +{ + const uint32_t algorithm = createInfo.flags & VMA_VIRTUAL_BLOCK_CREATE_ALGORITHM_MASK; + switch (algorithm) + { + case 0: + m_Metadata = vma_new(GetAllocationCallbacks(), VmaBlockMetadata_Generic)(VK_NULL_HANDLE, 1, true); + break; + case VMA_VIRTUAL_BLOCK_CREATE_BUDDY_ALGORITHM_BIT: + m_Metadata = vma_new(GetAllocationCallbacks(), VmaBlockMetadata_Buddy)(VK_NULL_HANDLE, 1, true); + break; + case VMA_VIRTUAL_BLOCK_CREATE_LINEAR_ALGORITHM_BIT: + m_Metadata = vma_new(GetAllocationCallbacks(), VmaBlockMetadata_Linear)(VK_NULL_HANDLE, 1, true); + break; + case VMA_VIRTUAL_BLOCK_CREATE_TLSF_ALGORITHM_BIT: + m_Metadata = vma_new(GetAllocationCallbacks(), VmaBlockMetadata_TLSF)(VK_NULL_HANDLE, 1, true); break; default: VMA_ASSERT(0); } + + m_Metadata->Init(createInfo.size); } -#endif // #if VMA_STATS_STRING_ENABLED +VmaVirtualBlock_T::~VmaVirtualBlock_T() +{ + // Define macro VMA_DEBUG_LOG to receive the list of the unfreed allocations + if (!m_Metadata->IsEmpty()) + m_Metadata->DebugLogAllAllocations(); + // This is the most important assert in the entire library. + // Hitting it means you have some memory leak - unreleased virtual allocations. + VMA_ASSERT(m_Metadata->IsEmpty() && "Some virtual allocations were not freed before destruction of this virtual block!"); + + vma_delete(GetAllocationCallbacks(), m_Metadata); +} -//////////////////////////////////////////////////////////////////////////////// -// class VmaDeviceMemoryBlock +const VkAllocationCallbacks* VmaVirtualBlock_T::GetAllocationCallbacks() const +{ + return m_AllocationCallbacksSpecified ? &m_AllocationCallbacks : VMA_NULL; +} +void VmaVirtualBlock_T::GetAllocationInfo(VmaVirtualAllocation allocation, VmaVirtualAllocationInfo& outInfo) +{ + m_Metadata->GetAllocationInfo((VmaAllocHandle)allocation, outInfo); +} + +VkResult VmaVirtualBlock_T::Allocate(const VmaVirtualAllocationCreateInfo& createInfo, VmaVirtualAllocation& outAllocation, + VkDeviceSize* outOffset) +{ + VmaAllocationRequest request = {}; + if (m_Metadata->CreateAllocationRequest( + createInfo.size, // allocSize + VMA_MAX(createInfo.alignment, (VkDeviceSize)1), // allocAlignment + (createInfo.flags & VMA_VIRTUAL_ALLOCATION_CREATE_UPPER_ADDRESS_BIT) != 0, // upperAddress + VMA_SUBALLOCATION_TYPE_UNKNOWN, // allocType - unimportant + createInfo.flags & VMA_VIRTUAL_ALLOCATION_CREATE_STRATEGY_MASK, // strategy + &request)) + { + m_Metadata->Alloc(request, + VMA_SUBALLOCATION_TYPE_UNKNOWN, // type - unimportant + createInfo.pUserData); + outAllocation = (VmaVirtualAllocation)request.allocHandle; + if(outOffset) + *outOffset = m_Metadata->GetAllocationOffset(request.allocHandle); + return VK_SUCCESS; + } + outAllocation = (VmaVirtualAllocation)VK_NULL_HANDLE; + if (outOffset) + *outOffset = UINT64_MAX; + return VK_ERROR_OUT_OF_DEVICE_MEMORY; +} + +void VmaVirtualBlock_T::CalculateStats(VmaStatInfo& outStatInfo) const +{ + m_Metadata->CalcAllocationStatInfo(outStatInfo); + VmaPostprocessCalcStatInfo(outStatInfo); +} + +#if VMA_STATS_STRING_ENABLED +void VmaVirtualBlock_T::BuildStatsString(bool detailedMap, VmaStringBuilder& sb) const +{ + VmaJsonWriter json(GetAllocationCallbacks(), sb); + json.BeginObject(); + + VmaStatInfo stat = {}; + CalculateStats(stat); + + json.WriteString("Stats"); + VmaPrintStatInfo(json, stat); + + if (detailedMap) + { + json.WriteString("Details"); + m_Metadata->PrintDetailedMap(json); + } + + json.EndObject(); +} +#endif // VMA_STATS_STRING_ENABLED +#endif // _VMA_VIRTUAL_BLOCK_T_FUNCTIONS +#endif // _VMA_VIRTUAL_BLOCK_T + +// Main allocator object. +struct VmaAllocator_T +{ + VMA_CLASS_NO_COPY(VmaAllocator_T) +public: + bool m_UseMutex; + uint32_t m_VulkanApiVersion; + bool m_UseKhrDedicatedAllocation; // Can be set only if m_VulkanApiVersion < VK_MAKE_VERSION(1, 1, 0). + bool m_UseKhrBindMemory2; // Can be set only if m_VulkanApiVersion < VK_MAKE_VERSION(1, 1, 0). + bool m_UseExtMemoryBudget; + bool m_UseAmdDeviceCoherentMemory; + bool m_UseKhrBufferDeviceAddress; + bool m_UseExtMemoryPriority; + VkDevice m_hDevice; + VkInstance m_hInstance; + bool m_AllocationCallbacksSpecified; + VkAllocationCallbacks m_AllocationCallbacks; + VmaDeviceMemoryCallbacks m_DeviceMemoryCallbacks; + VmaAllocationObjectAllocator m_AllocationObjectAllocator; + + // Each bit (1 << i) is set if HeapSizeLimit is enabled for that heap, so cannot allocate more than the heap size. + uint32_t m_HeapSizeLimitMask; + + VkPhysicalDeviceProperties m_PhysicalDeviceProperties; + VkPhysicalDeviceMemoryProperties m_MemProps; + + // Default pools. + VmaBlockVector* m_pBlockVectors[VK_MAX_MEMORY_TYPES]; + VmaDedicatedAllocationList m_DedicatedAllocations[VK_MAX_MEMORY_TYPES]; + + VmaCurrentBudgetData m_Budget; + VMA_ATOMIC_UINT32 m_DeviceMemoryCount; // Total number of VkDeviceMemory objects. + + VmaAllocator_T(const VmaAllocatorCreateInfo* pCreateInfo); + VkResult Init(const VmaAllocatorCreateInfo* pCreateInfo); + ~VmaAllocator_T(); + + const VkAllocationCallbacks* GetAllocationCallbacks() const + { + return m_AllocationCallbacksSpecified ? &m_AllocationCallbacks : VMA_NULL; + } + const VmaVulkanFunctions& GetVulkanFunctions() const + { + return m_VulkanFunctions; + } + + VkPhysicalDevice GetPhysicalDevice() const { return m_PhysicalDevice; } + + VkDeviceSize GetBufferImageGranularity() const + { + return VMA_MAX( + static_cast<VkDeviceSize>(VMA_DEBUG_MIN_BUFFER_IMAGE_GRANULARITY), + m_PhysicalDeviceProperties.limits.bufferImageGranularity); + } + + uint32_t GetMemoryHeapCount() const { return m_MemProps.memoryHeapCount; } + uint32_t GetMemoryTypeCount() const { return m_MemProps.memoryTypeCount; } + + uint32_t MemoryTypeIndexToHeapIndex(uint32_t memTypeIndex) const + { + VMA_ASSERT(memTypeIndex < m_MemProps.memoryTypeCount); + return m_MemProps.memoryTypes[memTypeIndex].heapIndex; + } + // True when specific memory type is HOST_VISIBLE but not HOST_COHERENT. + bool IsMemoryTypeNonCoherent(uint32_t memTypeIndex) const + { + return (m_MemProps.memoryTypes[memTypeIndex].propertyFlags & (VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT | VK_MEMORY_PROPERTY_HOST_COHERENT_BIT)) == + VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT; + } + // Minimum alignment for all allocations in specific memory type. + VkDeviceSize GetMemoryTypeMinAlignment(uint32_t memTypeIndex) const + { + return IsMemoryTypeNonCoherent(memTypeIndex) ? + VMA_MAX((VkDeviceSize)VMA_MIN_ALIGNMENT, m_PhysicalDeviceProperties.limits.nonCoherentAtomSize) : + (VkDeviceSize)VMA_MIN_ALIGNMENT; + } + + bool IsIntegratedGpu() const + { + return m_PhysicalDeviceProperties.deviceType == VK_PHYSICAL_DEVICE_TYPE_INTEGRATED_GPU; + } + + uint32_t GetGlobalMemoryTypeBits() const { return m_GlobalMemoryTypeBits; } + + void GetBufferMemoryRequirements( + VkBuffer hBuffer, + VkMemoryRequirements& memReq, + bool& requiresDedicatedAllocation, + bool& prefersDedicatedAllocation) const; + void GetImageMemoryRequirements( + VkImage hImage, + VkMemoryRequirements& memReq, + bool& requiresDedicatedAllocation, + bool& prefersDedicatedAllocation) const; + + // Main allocation function. + VkResult AllocateMemory( + const VkMemoryRequirements& vkMemReq, + bool requiresDedicatedAllocation, + bool prefersDedicatedAllocation, + VkBuffer dedicatedBuffer, + VkBufferUsageFlags dedicatedBufferUsage, // UINT32_MAX when unknown. + VkImage dedicatedImage, + const VmaAllocationCreateInfo& createInfo, + VmaSuballocationType suballocType, + size_t allocationCount, + VmaAllocation* pAllocations); + + // Main deallocation function. + void FreeMemory( + size_t allocationCount, + const VmaAllocation* pAllocations); + + void CalculateStats(VmaStats* pStats); + + void GetHeapBudgets( + VmaBudget* outBudgets, uint32_t firstHeap, uint32_t heapCount); + +#if VMA_STATS_STRING_ENABLED + void PrintDetailedMap(class VmaJsonWriter& json); +#endif + + VkResult DefragmentationBegin( + const VmaDefragmentationInfo2& info, + VmaDefragmentationStats* pStats, + VmaDefragmentationContext* pContext); + VkResult DefragmentationEnd( + VmaDefragmentationContext context); + + VkResult DefragmentationPassBegin( + VmaDefragmentationPassInfo* pInfo, + VmaDefragmentationContext context); + VkResult DefragmentationPassEnd( + VmaDefragmentationContext context); + + void GetAllocationInfo(VmaAllocation hAllocation, VmaAllocationInfo* pAllocationInfo); + + VkResult CreatePool(const VmaPoolCreateInfo* pCreateInfo, VmaPool* pPool); + void DestroyPool(VmaPool pool); + void GetPoolStats(VmaPool pool, VmaPoolStats* pPoolStats); + + void SetCurrentFrameIndex(uint32_t frameIndex); + uint32_t GetCurrentFrameIndex() const { return m_CurrentFrameIndex.load(); } + + VkResult CheckPoolCorruption(VmaPool hPool); + VkResult CheckCorruption(uint32_t memoryTypeBits); + + // Call to Vulkan function vkAllocateMemory with accompanying bookkeeping. + VkResult AllocateVulkanMemory(const VkMemoryAllocateInfo* pAllocateInfo, VkDeviceMemory* pMemory); + // Call to Vulkan function vkFreeMemory with accompanying bookkeeping. + void FreeVulkanMemory(uint32_t memoryType, VkDeviceSize size, VkDeviceMemory hMemory); + // Call to Vulkan function vkBindBufferMemory or vkBindBufferMemory2KHR. + VkResult BindVulkanBuffer( + VkDeviceMemory memory, + VkDeviceSize memoryOffset, + VkBuffer buffer, + const void* pNext); + // Call to Vulkan function vkBindImageMemory or vkBindImageMemory2KHR. + VkResult BindVulkanImage( + VkDeviceMemory memory, + VkDeviceSize memoryOffset, + VkImage image, + const void* pNext); + + VkResult Map(VmaAllocation hAllocation, void** ppData); + void Unmap(VmaAllocation hAllocation); + + VkResult BindBufferMemory( + VmaAllocation hAllocation, + VkDeviceSize allocationLocalOffset, + VkBuffer hBuffer, + const void* pNext); + VkResult BindImageMemory( + VmaAllocation hAllocation, + VkDeviceSize allocationLocalOffset, + VkImage hImage, + const void* pNext); + + VkResult FlushOrInvalidateAllocation( + VmaAllocation hAllocation, + VkDeviceSize offset, VkDeviceSize size, + VMA_CACHE_OPERATION op); + VkResult FlushOrInvalidateAllocations( + uint32_t allocationCount, + const VmaAllocation* allocations, + const VkDeviceSize* offsets, const VkDeviceSize* sizes, + VMA_CACHE_OPERATION op); + + void FillAllocation(const VmaAllocation hAllocation, uint8_t pattern); + + /* + Returns bit mask of memory types that can support defragmentation on GPU as + they support creation of required buffer for copy operations. + */ + uint32_t GetGpuDefragmentationMemoryTypeBits(); + +#if VMA_EXTERNAL_MEMORY + VkExternalMemoryHandleTypeFlagsKHR GetExternalMemoryHandleTypeFlags(uint32_t memTypeIndex) const + { + return m_TypeExternalMemoryHandleTypes[memTypeIndex]; + } +#endif // #if VMA_EXTERNAL_MEMORY + +private: + VkDeviceSize m_PreferredLargeHeapBlockSize; + + VkPhysicalDevice m_PhysicalDevice; + VMA_ATOMIC_UINT32 m_CurrentFrameIndex; + VMA_ATOMIC_UINT32 m_GpuDefragmentationMemoryTypeBits; // UINT32_MAX means uninitialized. +#if VMA_EXTERNAL_MEMORY + VkExternalMemoryHandleTypeFlagsKHR m_TypeExternalMemoryHandleTypes[VK_MAX_MEMORY_TYPES]; +#endif // #if VMA_EXTERNAL_MEMORY + + VMA_RW_MUTEX m_PoolsMutex; + typedef VmaIntrusiveLinkedList<VmaPoolListItemTraits> PoolList; + // Protected by m_PoolsMutex. + PoolList m_Pools; + uint32_t m_NextPoolId; + + VmaVulkanFunctions m_VulkanFunctions; + + // Global bit mask AND-ed with any memoryTypeBits to disallow certain memory types. + uint32_t m_GlobalMemoryTypeBits; + + void ImportVulkanFunctions(const VmaVulkanFunctions* pVulkanFunctions); + +#if VMA_STATIC_VULKAN_FUNCTIONS == 1 + void ImportVulkanFunctions_Static(); +#endif + + void ImportVulkanFunctions_Custom(const VmaVulkanFunctions* pVulkanFunctions); + +#if VMA_DYNAMIC_VULKAN_FUNCTIONS == 1 + void ImportVulkanFunctions_Dynamic(); +#endif + + void ValidateVulkanFunctions(); + +public: // I'm sorry + VkDeviceSize CalcPreferredBlockSize(uint32_t memTypeIndex); + +private: + VkResult AllocateMemoryOfType( + VmaPool pool, + VkDeviceSize size, + VkDeviceSize alignment, + bool dedicatedPreferred, + VkBuffer dedicatedBuffer, + VkBufferUsageFlags dedicatedBufferUsage, + VkImage dedicatedImage, + const VmaAllocationCreateInfo& createInfo, + uint32_t memTypeIndex, + VmaSuballocationType suballocType, + VmaDedicatedAllocationList& dedicatedAllocations, + VmaBlockVector& blockVector, + size_t allocationCount, + VmaAllocation* pAllocations); + + // Helper function only to be used inside AllocateDedicatedMemory. + VkResult AllocateDedicatedMemoryPage( + VmaPool pool, + VkDeviceSize size, + VmaSuballocationType suballocType, + uint32_t memTypeIndex, + const VkMemoryAllocateInfo& allocInfo, + bool map, + bool isUserDataString, + void* pUserData, + VmaAllocation* pAllocation); + + // Allocates and registers new VkDeviceMemory specifically for dedicated allocations. + VkResult AllocateDedicatedMemory( + VmaPool pool, + VkDeviceSize size, + VmaSuballocationType suballocType, + VmaDedicatedAllocationList& dedicatedAllocations, + uint32_t memTypeIndex, + bool map, + bool isUserDataString, + bool canAliasMemory, + void* pUserData, + float priority, + VkBuffer dedicatedBuffer, + VkBufferUsageFlags dedicatedBufferUsage, + VkImage dedicatedImage, + size_t allocationCount, + VmaAllocation* pAllocations, + const void* pNextChain = nullptr); + + void FreeDedicatedMemory(const VmaAllocation allocation); + + VkResult CalcMemTypeParams( + VmaAllocationCreateInfo& outCreateInfo, + uint32_t memTypeIndex, + VkDeviceSize size, + size_t allocationCount); + VkResult CalcAllocationParams( + VmaAllocationCreateInfo& outCreateInfo, + bool dedicatedRequired, + bool dedicatedPreferred); + + /* + Calculates and returns bit mask of memory types that can support defragmentation + on GPU as they support creation of required buffer for copy operations. + */ + uint32_t CalculateGpuDefragmentationMemoryTypeBits() const; + uint32_t CalculateGlobalMemoryTypeBits() const; + + bool GetFlushOrInvalidateRange( + VmaAllocation allocation, + VkDeviceSize offset, VkDeviceSize size, + VkMappedMemoryRange& outRange) const; + +#if VMA_MEMORY_BUDGET + void UpdateVulkanBudget(); +#endif // #if VMA_MEMORY_BUDGET +}; + + +#ifndef _VMA_MEMORY_FUNCTIONS +static void* VmaMalloc(VmaAllocator hAllocator, size_t size, size_t alignment) +{ + return VmaMalloc(&hAllocator->m_AllocationCallbacks, size, alignment); +} + +static void VmaFree(VmaAllocator hAllocator, void* ptr) +{ + VmaFree(&hAllocator->m_AllocationCallbacks, ptr); +} + +template<typename T> +static T* VmaAllocate(VmaAllocator hAllocator) +{ + return (T*)VmaMalloc(hAllocator, sizeof(T), VMA_ALIGN_OF(T)); +} + +template<typename T> +static T* VmaAllocateArray(VmaAllocator hAllocator, size_t count) +{ + return (T*)VmaMalloc(hAllocator, sizeof(T) * count, VMA_ALIGN_OF(T)); +} + +template<typename T> +static void vma_delete(VmaAllocator hAllocator, T* ptr) +{ + if(ptr != VMA_NULL) + { + ptr->~T(); + VmaFree(hAllocator, ptr); + } +} + +template<typename T> +static void vma_delete_array(VmaAllocator hAllocator, T* ptr, size_t count) +{ + if(ptr != VMA_NULL) + { + for(size_t i = count; i--; ) + ptr[i].~T(); + VmaFree(hAllocator, ptr); + } +} +#endif // _VMA_MEMORY_FUNCTIONS + +#ifndef _VMA_DEVICE_MEMORY_BLOCK_FUNCTIONS VmaDeviceMemoryBlock::VmaDeviceMemoryBlock(VmaAllocator hAllocator) + : m_pMetadata(VMA_NULL), + m_MemoryTypeIndex(UINT32_MAX), + m_Id(0), + m_hMemory(VK_NULL_HANDLE), + m_MapCount(0), + m_pMappedData(VMA_NULL) {} + +VmaDeviceMemoryBlock::~VmaDeviceMemoryBlock() { + VMA_ASSERT(m_MapCount == 0 && "VkDeviceMemory block is being destroyed while it is still mapped."); + VMA_ASSERT(m_hMemory == VK_NULL_HANDLE); } void VmaDeviceMemoryBlock::Init( VmaAllocator hAllocator, - VmaBlockVector* parentBlockVector, VmaPool hParentPool, uint32_t newMemoryTypeIndex, VkDeviceMemory newMemory, VkDeviceSize newSize, uint32_t id, - uint32_t algorithm) + uint32_t algorithm, + VkDeviceSize bufferImageGranularity) { - VMA_ASSERT(parentBlockVector != VMA_NULL); VMA_ASSERT(m_hMemory == VK_NULL_HANDLE); - m_ParentBlockVector = parentBlockVector; m_hParentPool = hParentPool; m_MemoryTypeIndex = newMemoryTypeIndex; m_Id = id; m_hMemory = newMemory; - switch(algorithm) + switch (algorithm) { case VMA_POOL_CREATE_LINEAR_ALGORITHM_BIT: - m_pMetadata = vma_new(hAllocator, VmaBlockMetadata_Linear)(hAllocator); + m_pMetadata = vma_new(hAllocator, VmaBlockMetadata_Linear)(hAllocator->GetAllocationCallbacks(), + bufferImageGranularity, false); // isVirtual break; case VMA_POOL_CREATE_BUDDY_ALGORITHM_BIT: - m_pMetadata = vma_new(hAllocator, VmaBlockMetadata_Buddy)(hAllocator); + m_pMetadata = vma_new(hAllocator, VmaBlockMetadata_Buddy)(hAllocator->GetAllocationCallbacks(), + bufferImageGranularity, false); // isVirtual + break; + case VMA_POOL_CREATE_TLSF_ALGORITHM_BIT: + m_pMetadata = vma_new(hAllocator, VmaBlockMetadata_TLSF)(hAllocator->GetAllocationCallbacks(), + bufferImageGranularity, false); // isVirtual break; default: VMA_ASSERT(0); // Fall-through. case 0: - m_pMetadata = vma_new(hAllocator, VmaBlockMetadata_Generic)(hAllocator); + m_pMetadata = vma_new(hAllocator, VmaBlockMetadata_Generic)(hAllocator->GetAllocationCallbacks(), + bufferImageGranularity, false); // isVirtual } m_pMetadata->Init(newSize); } void VmaDeviceMemoryBlock::Destroy(VmaAllocator allocator) { + // Define macro VMA_DEBUG_LOG to receive the list of the unfreed allocations + if (!m_pMetadata->IsEmpty()) + m_pMetadata->DebugLogAllAllocations(); // This is the most important assert in the entire library. // Hitting it means you have some memory leak - unreleased VmaAllocation objects. VMA_ASSERT(m_pMetadata->IsEmpty() && "Some allocations were not freed before destruction of this memory block!"); @@ -12732,7 +11608,7 @@ VkResult VmaDeviceMemoryBlock::CheckCorruption(VmaAllocator hAllocator) { void* pData = nullptr; VkResult res = Map(hAllocator, 1, &pData); - if(res != VK_SUCCESS) + if (res != VK_SUCCESS) { return res; } @@ -12746,17 +11622,17 @@ VkResult VmaDeviceMemoryBlock::CheckCorruption(VmaAllocator hAllocator) VkResult VmaDeviceMemoryBlock::Map(VmaAllocator hAllocator, uint32_t count, void** ppData) { - if(count == 0) + if (count == 0) { return VK_SUCCESS; } VmaMutexLock lock(m_Mutex, hAllocator->m_UseMutex); - if(m_MapCount != 0) + if (m_MapCount != 0) { m_MapCount += count; VMA_ASSERT(m_pMappedData != VMA_NULL); - if(ppData != VMA_NULL) + if (ppData != VMA_NULL) { *ppData = m_pMappedData; } @@ -12771,9 +11647,9 @@ VkResult VmaDeviceMemoryBlock::Map(VmaAllocator hAllocator, uint32_t count, void VK_WHOLE_SIZE, 0, // flags &m_pMappedData); - if(result == VK_SUCCESS) + if (result == VK_SUCCESS) { - if(ppData != VMA_NULL) + if (ppData != VMA_NULL) { *ppData = m_pMappedData; } @@ -12785,16 +11661,16 @@ VkResult VmaDeviceMemoryBlock::Map(VmaAllocator hAllocator, uint32_t count, void void VmaDeviceMemoryBlock::Unmap(VmaAllocator hAllocator, uint32_t count) { - if(count == 0) + if (count == 0) { return; } VmaMutexLock lock(m_Mutex, hAllocator->m_UseMutex); - if(m_MapCount >= count) + if (m_MapCount >= count) { m_MapCount -= count; - if(m_MapCount == 0) + if (m_MapCount == 0) { m_pMappedData = VMA_NULL; (*hAllocator->GetVulkanFunctions().vkUnmapMemory)(hAllocator->m_hDevice, m_hMemory); @@ -12806,49 +11682,40 @@ void VmaDeviceMemoryBlock::Unmap(VmaAllocator hAllocator, uint32_t count) } } -VkResult VmaDeviceMemoryBlock::WriteMagicValueAroundAllocation(VmaAllocator hAllocator, VkDeviceSize allocOffset, VkDeviceSize allocSize) +VkResult VmaDeviceMemoryBlock::WriteMagicValueAfterAllocation(VmaAllocator hAllocator, VkDeviceSize allocOffset, VkDeviceSize allocSize) { VMA_ASSERT(VMA_DEBUG_MARGIN > 0 && VMA_DEBUG_MARGIN % 4 == 0 && VMA_DEBUG_DETECT_CORRUPTION); - VMA_ASSERT(allocOffset >= VMA_DEBUG_MARGIN); void* pData; VkResult res = Map(hAllocator, 1, &pData); - if(res != VK_SUCCESS) + if (res != VK_SUCCESS) { return res; } - VmaWriteMagicValue(pData, allocOffset - VMA_DEBUG_MARGIN); VmaWriteMagicValue(pData, allocOffset + allocSize); Unmap(hAllocator, 1); - return VK_SUCCESS; } -VkResult VmaDeviceMemoryBlock::ValidateMagicValueAroundAllocation(VmaAllocator hAllocator, VkDeviceSize allocOffset, VkDeviceSize allocSize) +VkResult VmaDeviceMemoryBlock::ValidateMagicValueAfterAllocation(VmaAllocator hAllocator, VkDeviceSize allocOffset, VkDeviceSize allocSize) { VMA_ASSERT(VMA_DEBUG_MARGIN > 0 && VMA_DEBUG_MARGIN % 4 == 0 && VMA_DEBUG_DETECT_CORRUPTION); - VMA_ASSERT(allocOffset >= VMA_DEBUG_MARGIN); void* pData; VkResult res = Map(hAllocator, 1, &pData); - if(res != VK_SUCCESS) + if (res != VK_SUCCESS) { return res; } - if(!VmaValidateMagicValue(pData, allocOffset - VMA_DEBUG_MARGIN)) - { - VMA_ASSERT(0 && "MEMORY CORRUPTION DETECTED BEFORE FREED ALLOCATION!"); - } - else if(!VmaValidateMagicValue(pData, allocOffset + allocSize)) + if (!VmaValidateMagicValue(pData, allocOffset + allocSize)) { VMA_ASSERT(0 && "MEMORY CORRUPTION DETECTED AFTER FREED ALLOCATION!"); } Unmap(hAllocator, 1); - return VK_SUCCESS; } @@ -12885,83 +11752,350 @@ VkResult VmaDeviceMemoryBlock::BindImageMemory( VmaMutexLock lock(m_Mutex, hAllocator->m_UseMutex); return hAllocator->BindVulkanImage(m_hMemory, memoryOffset, hImage, pNext); } +#endif // _VMA_DEVICE_MEMORY_BLOCK_FUNCTIONS -static void InitStatInfo(VmaStatInfo& outInfo) +#ifndef _VMA_ALLOCATION_T_FUNCTIONS +VmaAllocation_T::VmaAllocation_T(bool userDataString) + : m_Alignment{ 1 }, + m_Size{ 0 }, + m_pUserData{ VMA_NULL }, + m_MemoryTypeIndex{ 0 }, + m_Type{ (uint8_t)ALLOCATION_TYPE_NONE }, + m_SuballocationType{ (uint8_t)VMA_SUBALLOCATION_TYPE_UNKNOWN }, + m_MapCount{ 0 }, + m_Flags{ userDataString ? (uint8_t)FLAG_USER_DATA_STRING : (uint8_t)0 } { - memset(&outInfo, 0, sizeof(outInfo)); - outInfo.allocationSizeMin = UINT64_MAX; - outInfo.unusedRangeSizeMin = UINT64_MAX; +#if VMA_STATS_STRING_ENABLED + m_BufferImageUsage = 0; +#endif } -// Adds statistics srcInfo into inoutInfo, like: inoutInfo += srcInfo. -static void VmaAddStatInfo(VmaStatInfo& inoutInfo, const VmaStatInfo& srcInfo) +VmaAllocation_T::~VmaAllocation_T() { - inoutInfo.blockCount += srcInfo.blockCount; - inoutInfo.allocationCount += srcInfo.allocationCount; - inoutInfo.unusedRangeCount += srcInfo.unusedRangeCount; - inoutInfo.usedBytes += srcInfo.usedBytes; - inoutInfo.unusedBytes += srcInfo.unusedBytes; - inoutInfo.allocationSizeMin = VMA_MIN(inoutInfo.allocationSizeMin, srcInfo.allocationSizeMin); - inoutInfo.allocationSizeMax = VMA_MAX(inoutInfo.allocationSizeMax, srcInfo.allocationSizeMax); - inoutInfo.unusedRangeSizeMin = VMA_MIN(inoutInfo.unusedRangeSizeMin, srcInfo.unusedRangeSizeMin); - inoutInfo.unusedRangeSizeMax = VMA_MAX(inoutInfo.unusedRangeSizeMax, srcInfo.unusedRangeSizeMax); + VMA_ASSERT((m_MapCount & ~MAP_COUNT_FLAG_PERSISTENT_MAP) == 0 && "Allocation was not unmapped before destruction."); + + // Check if owned string was freed. + VMA_ASSERT(m_pUserData == VMA_NULL); } -static void VmaPostprocessCalcStatInfo(VmaStatInfo& inoutInfo) +void VmaAllocation_T::InitBlockAllocation( + VmaDeviceMemoryBlock* block, + VmaAllocHandle allocHandle, + VkDeviceSize alignment, + VkDeviceSize size, + uint32_t memoryTypeIndex, + VmaSuballocationType suballocationType, + bool mapped) { - inoutInfo.allocationSizeAvg = (inoutInfo.allocationCount > 0) ? - VmaRoundDiv<VkDeviceSize>(inoutInfo.usedBytes, inoutInfo.allocationCount) : 0; - inoutInfo.unusedRangeSizeAvg = (inoutInfo.unusedRangeCount > 0) ? - VmaRoundDiv<VkDeviceSize>(inoutInfo.unusedBytes, inoutInfo.unusedRangeCount) : 0; + VMA_ASSERT(m_Type == ALLOCATION_TYPE_NONE); + VMA_ASSERT(block != VMA_NULL); + m_Type = (uint8_t)ALLOCATION_TYPE_BLOCK; + m_Alignment = alignment; + m_Size = size; + m_MemoryTypeIndex = memoryTypeIndex; + m_MapCount = mapped ? MAP_COUNT_FLAG_PERSISTENT_MAP : 0; + m_SuballocationType = (uint8_t)suballocationType; + m_BlockAllocation.m_Block = block; + m_BlockAllocation.m_AllocHandle = allocHandle; } -VmaPool_T::VmaPool_T( +void VmaAllocation_T::InitDedicatedAllocation( + VmaPool hParentPool, + uint32_t memoryTypeIndex, + VkDeviceMemory hMemory, + VmaSuballocationType suballocationType, + void* pMappedData, + VkDeviceSize size) +{ + VMA_ASSERT(m_Type == ALLOCATION_TYPE_NONE); + VMA_ASSERT(hMemory != VK_NULL_HANDLE); + m_Type = (uint8_t)ALLOCATION_TYPE_DEDICATED; + m_Alignment = 0; + m_Size = size; + m_MemoryTypeIndex = memoryTypeIndex; + m_SuballocationType = (uint8_t)suballocationType; + m_MapCount = (pMappedData != VMA_NULL) ? MAP_COUNT_FLAG_PERSISTENT_MAP : 0; + m_DedicatedAllocation.m_hParentPool = hParentPool; + m_DedicatedAllocation.m_hMemory = hMemory; + m_DedicatedAllocation.m_pMappedData = pMappedData; + m_DedicatedAllocation.m_Prev = VMA_NULL; + m_DedicatedAllocation.m_Next = VMA_NULL; +} + +void VmaAllocation_T::SetUserData(VmaAllocator hAllocator, void* pUserData) +{ + if (IsUserDataString()) + { + VMA_ASSERT(pUserData == VMA_NULL || pUserData != m_pUserData); + + FreeUserDataString(hAllocator); + + if (pUserData != VMA_NULL) + { + m_pUserData = VmaCreateStringCopy(hAllocator->GetAllocationCallbacks(), (const char*)pUserData); + } + } + else + { + m_pUserData = pUserData; + } +} + +void VmaAllocation_T::ChangeBlockAllocation( VmaAllocator hAllocator, - const VmaPoolCreateInfo& createInfo, - VkDeviceSize preferredBlockSize) : - m_BlockVector( - hAllocator, - this, // hParentPool - createInfo.memoryTypeIndex, - createInfo.blockSize != 0 ? createInfo.blockSize : preferredBlockSize, - createInfo.minBlockCount, - createInfo.maxBlockCount, - (createInfo.flags & VMA_POOL_CREATE_IGNORE_BUFFER_IMAGE_GRANULARITY_BIT) != 0 ? 1 : hAllocator->GetBufferImageGranularity(), - createInfo.frameInUseCount, - createInfo.blockSize != 0, // explicitBlockSize - createInfo.flags & VMA_POOL_CREATE_ALGORITHM_MASK, // algorithm - createInfo.priority, - VMA_MAX(hAllocator->GetMemoryTypeMinAlignment(createInfo.memoryTypeIndex), createInfo.minAllocationAlignment), - createInfo.pMemoryAllocateNext), - m_Id(0), - m_Name(VMA_NULL) + VmaDeviceMemoryBlock* block, + VmaAllocHandle allocHandle) { + VMA_ASSERT(block != VMA_NULL); + VMA_ASSERT(m_Type == ALLOCATION_TYPE_BLOCK); + + // Move mapping reference counter from old block to new block. + if (block != m_BlockAllocation.m_Block) + { + uint32_t mapRefCount = m_MapCount & ~MAP_COUNT_FLAG_PERSISTENT_MAP; + if (IsPersistentMap()) + ++mapRefCount; + m_BlockAllocation.m_Block->Unmap(hAllocator, mapRefCount); + block->Map(hAllocator, mapRefCount, VMA_NULL); + } + + m_BlockAllocation.m_Block = block; + m_BlockAllocation.m_AllocHandle = allocHandle; } -VmaPool_T::~VmaPool_T() +void VmaAllocation_T::ChangeAllocHandle(VmaAllocHandle newAllocHandle) { - VMA_ASSERT(m_PrevPool == VMA_NULL && m_NextPool == VMA_NULL); + VMA_ASSERT(m_Type == ALLOCATION_TYPE_BLOCK); + m_BlockAllocation.m_AllocHandle = newAllocHandle; } -void VmaPool_T::SetName(const char* pName) +VmaAllocHandle VmaAllocation_T::GetAllocHandle() const +{ + switch (m_Type) + { + case ALLOCATION_TYPE_BLOCK: + return m_BlockAllocation.m_AllocHandle; + case ALLOCATION_TYPE_DEDICATED: + return VK_NULL_HANDLE; + default: + VMA_ASSERT(0); + return VK_NULL_HANDLE; + } +} + +VkDeviceSize VmaAllocation_T::GetOffset() const +{ + switch (m_Type) + { + case ALLOCATION_TYPE_BLOCK: + return m_BlockAllocation.m_Block->m_pMetadata->GetAllocationOffset(m_BlockAllocation.m_AllocHandle); + case ALLOCATION_TYPE_DEDICATED: + return 0; + default: + VMA_ASSERT(0); + return 0; + } +} + +VmaPool VmaAllocation_T::GetParentPool() const +{ + switch (m_Type) + { + case ALLOCATION_TYPE_BLOCK: + return m_BlockAllocation.m_Block->GetParentPool(); + case ALLOCATION_TYPE_DEDICATED: + return m_DedicatedAllocation.m_hParentPool; + default: + VMA_ASSERT(0); + return VK_NULL_HANDLE; + } +} + +VkDeviceMemory VmaAllocation_T::GetMemory() const +{ + switch (m_Type) + { + case ALLOCATION_TYPE_BLOCK: + return m_BlockAllocation.m_Block->GetDeviceMemory(); + case ALLOCATION_TYPE_DEDICATED: + return m_DedicatedAllocation.m_hMemory; + default: + VMA_ASSERT(0); + return VK_NULL_HANDLE; + } +} + +void* VmaAllocation_T::GetMappedData() const { - const VkAllocationCallbacks* allocs = m_BlockVector.GetAllocator()->GetAllocationCallbacks(); - VmaFreeString(allocs, m_Name); + switch (m_Type) + { + case ALLOCATION_TYPE_BLOCK: + if (m_MapCount != 0) + { + void* pBlockData = m_BlockAllocation.m_Block->GetMappedData(); + VMA_ASSERT(pBlockData != VMA_NULL); + return (char*)pBlockData + GetOffset(); + } + else + { + return VMA_NULL; + } + break; + case ALLOCATION_TYPE_DEDICATED: + VMA_ASSERT((m_DedicatedAllocation.m_pMappedData != VMA_NULL) == (m_MapCount != 0)); + return m_DedicatedAllocation.m_pMappedData; + default: + VMA_ASSERT(0); + return VMA_NULL; + } +} + +void VmaAllocation_T::DedicatedAllocCalcStatsInfo(VmaStatInfo& outInfo) +{ + VMA_ASSERT(m_Type == ALLOCATION_TYPE_DEDICATED); + outInfo.blockCount = 1; + outInfo.allocationCount = 1; + outInfo.unusedRangeCount = 0; + outInfo.usedBytes = m_Size; + outInfo.unusedBytes = 0; + outInfo.allocationSizeMin = outInfo.allocationSizeMax = m_Size; + outInfo.unusedRangeSizeMin = UINT64_MAX; + outInfo.unusedRangeSizeMax = 0; +} - if(pName != VMA_NULL) +void VmaAllocation_T::BlockAllocMap() +{ + VMA_ASSERT(GetType() == ALLOCATION_TYPE_BLOCK); + + if ((m_MapCount & ~MAP_COUNT_FLAG_PERSISTENT_MAP) < 0x7F) { - m_Name = VmaCreateStringCopy(allocs, pName); + ++m_MapCount; } else { - m_Name = VMA_NULL; + VMA_ASSERT(0 && "Allocation mapped too many times simultaneously."); + } +} + +void VmaAllocation_T::BlockAllocUnmap() +{ + VMA_ASSERT(GetType() == ALLOCATION_TYPE_BLOCK); + + if ((m_MapCount & ~MAP_COUNT_FLAG_PERSISTENT_MAP) != 0) + { + --m_MapCount; + } + else + { + VMA_ASSERT(0 && "Unmapping allocation not previously mapped."); + } +} + +VkResult VmaAllocation_T::DedicatedAllocMap(VmaAllocator hAllocator, void** ppData) +{ + VMA_ASSERT(GetType() == ALLOCATION_TYPE_DEDICATED); + + if (m_MapCount != 0) + { + if ((m_MapCount & ~MAP_COUNT_FLAG_PERSISTENT_MAP) < 0x7F) + { + VMA_ASSERT(m_DedicatedAllocation.m_pMappedData != VMA_NULL); + *ppData = m_DedicatedAllocation.m_pMappedData; + ++m_MapCount; + return VK_SUCCESS; + } + else + { + VMA_ASSERT(0 && "Dedicated allocation mapped too many times simultaneously."); + return VK_ERROR_MEMORY_MAP_FAILED; + } + } + else + { + VkResult result = (*hAllocator->GetVulkanFunctions().vkMapMemory)( + hAllocator->m_hDevice, + m_DedicatedAllocation.m_hMemory, + 0, // offset + VK_WHOLE_SIZE, + 0, // flags + ppData); + if (result == VK_SUCCESS) + { + m_DedicatedAllocation.m_pMappedData = *ppData; + m_MapCount = 1; + } + return result; + } +} + +void VmaAllocation_T::DedicatedAllocUnmap(VmaAllocator hAllocator) +{ + VMA_ASSERT(GetType() == ALLOCATION_TYPE_DEDICATED); + + if ((m_MapCount & ~MAP_COUNT_FLAG_PERSISTENT_MAP) != 0) + { + --m_MapCount; + if (m_MapCount == 0) + { + m_DedicatedAllocation.m_pMappedData = VMA_NULL; + (*hAllocator->GetVulkanFunctions().vkUnmapMemory)( + hAllocator->m_hDevice, + m_DedicatedAllocation.m_hMemory); + } + } + else + { + VMA_ASSERT(0 && "Unmapping dedicated allocation not previously mapped."); } } #if VMA_STATS_STRING_ENABLED +void VmaAllocation_T::InitBufferImageUsage(uint32_t bufferImageUsage) +{ + VMA_ASSERT(m_BufferImageUsage == 0); + m_BufferImageUsage = bufferImageUsage; +} + +void VmaAllocation_T::PrintParameters(class VmaJsonWriter& json) const +{ + json.WriteString("Type"); + json.WriteString(VMA_SUBALLOCATION_TYPE_NAMES[m_SuballocationType]); -#endif // #if VMA_STATS_STRING_ENABLED + json.WriteString("Size"); + json.WriteNumber(m_Size); + if (m_pUserData != VMA_NULL) + { + json.WriteString("UserData"); + if (IsUserDataString()) + { + json.WriteString((const char*)m_pUserData); + } + else + { + json.BeginString(); + json.ContinueString_Pointer(m_pUserData); + json.EndString(); + } + } + + if (m_BufferImageUsage != 0) + { + json.WriteString("Usage"); + json.WriteNumber(m_BufferImageUsage); + } +} +#endif // VMA_STATS_STRING_ENABLED + +void VmaAllocation_T::FreeUserDataString(VmaAllocator hAllocator) +{ + VMA_ASSERT(IsUserDataString()); + VmaFreeString(hAllocator->GetAllocationCallbacks(), (char*)m_pUserData); + m_pUserData = VMA_NULL; +} +#endif // _VMA_ALLOCATION_T_FUNCTIONS + +#ifndef _VMA_BLOCK_VECTOR_FUNCTIONS VmaBlockVector::VmaBlockVector( VmaAllocator hAllocator, VmaPool hParentPool, @@ -12970,20 +12104,18 @@ VmaBlockVector::VmaBlockVector( size_t minBlockCount, size_t maxBlockCount, VkDeviceSize bufferImageGranularity, - uint32_t frameInUseCount, bool explicitBlockSize, uint32_t algorithm, float priority, VkDeviceSize minAllocationAlignment, - void* pMemoryAllocateNext) : - m_hAllocator(hAllocator), + void* pMemoryAllocateNext) + : m_hAllocator(hAllocator), m_hParentPool(hParentPool), m_MemoryTypeIndex(memoryTypeIndex), m_PreferredBlockSize(preferredBlockSize), m_MinBlockCount(minBlockCount), m_MaxBlockCount(maxBlockCount), m_BufferImageGranularity(bufferImageGranularity), - m_FrameInUseCount(frameInUseCount), m_ExplicitBlockSize(explicitBlockSize), m_Algorithm(algorithm), m_Priority(priority), @@ -12991,13 +12123,11 @@ VmaBlockVector::VmaBlockVector( m_pMemoryAllocateNext(pMemoryAllocateNext), m_HasEmptyBlock(false), m_Blocks(VmaStlAllocator<VmaDeviceMemoryBlock*>(hAllocator->GetAllocationCallbacks())), - m_NextBlockId(0) -{ -} + m_NextBlockId(0) {} VmaBlockVector::~VmaBlockVector() { - for(size_t i = m_Blocks.size(); i--; ) + for (size_t i = m_Blocks.size(); i--; ) { m_Blocks[i]->Destroy(m_hAllocator); vma_delete(m_hAllocator, m_Blocks[i]); @@ -13006,10 +12136,10 @@ VmaBlockVector::~VmaBlockVector() VkResult VmaBlockVector::CreateMinBlocks() { - for(size_t i = 0; i < m_MinBlockCount; ++i) + for (size_t i = 0; i < m_MinBlockCount; ++i) { VkResult res = CreateBlock(m_PreferredBlockSize, VMA_NULL); - if(res != VK_SUCCESS) + if (res != VK_SUCCESS) { return res; } @@ -13017,20 +12147,14 @@ VkResult VmaBlockVector::CreateMinBlocks() return VK_SUCCESS; } -void VmaBlockVector::GetPoolStats(VmaPoolStats* pStats) +void VmaBlockVector::AddPoolStats(VmaPoolStats* pStats) { VmaMutexLockRead lock(m_Mutex, m_hAllocator->m_UseMutex); const size_t blockCount = m_Blocks.size(); + pStats->blockCount += blockCount; - pStats->size = 0; - pStats->unusedSize = 0; - pStats->allocationCount = 0; - pStats->unusedRangeCount = 0; - pStats->unusedRangeSizeMax = 0; - pStats->blockCount = blockCount; - - for(uint32_t blockIndex = 0; blockIndex < blockCount; ++blockIndex) + for (uint32_t blockIndex = 0; blockIndex < blockCount; ++blockIndex) { const VmaDeviceMemoryBlock* const pBlock = m_Blocks[blockIndex]; VMA_ASSERT(pBlock); @@ -13054,10 +12178,7 @@ bool VmaBlockVector::IsCorruptionDetectionEnabled() const (m_hAllocator->m_MemProps.memoryTypes[m_MemoryTypeIndex].propertyFlags & requiredMemFlags) == requiredMemFlags; } -static const uint32_t VMA_ALLOCATION_TRY_COUNT = 32; - VkResult VmaBlockVector::Allocate( - uint32_t currentFrameIndex, VkDeviceSize size, VkDeviceSize alignment, const VmaAllocationCreateInfo& createInfo, @@ -13070,7 +12191,7 @@ VkResult VmaBlockVector::Allocate( alignment = VMA_MAX(alignment, m_MinAllocationAlignment); - if(IsCorruptionDetectionEnabled()) + if (IsCorruptionDetectionEnabled()) { size = VmaAlignUp<VkDeviceSize>(size, sizeof(VMA_CORRUPTION_DETECTION_MAGIC_VALUE)); alignment = VmaAlignUp<VkDeviceSize>(alignment, sizeof(VMA_CORRUPTION_DETECTION_MAGIC_VALUE)); @@ -13078,27 +12199,26 @@ VkResult VmaBlockVector::Allocate( { VmaMutexLockWrite lock(m_Mutex, m_hAllocator->m_UseMutex); - for(allocIndex = 0; allocIndex < allocationCount; ++allocIndex) + for (allocIndex = 0; allocIndex < allocationCount; ++allocIndex) { res = AllocatePage( - currentFrameIndex, size, alignment, createInfo, suballocType, pAllocations + allocIndex); - if(res != VK_SUCCESS) + if (res != VK_SUCCESS) { break; } } } - if(res != VK_SUCCESS) + if (res != VK_SUCCESS) { // Free all already created allocations. const uint32_t heapIndex = m_hAllocator->MemoryTypeIndexToHeapIndex(m_MemoryTypeIndex); - while(allocIndex--) + while (allocIndex--) { VmaAllocation_T* const alloc = pAllocations[allocIndex]; const VkDeviceSize allocSize = alloc->GetSize(); @@ -13112,7 +12232,6 @@ VkResult VmaBlockVector::Allocate( } VkResult VmaBlockVector::AllocatePage( - uint32_t currentFrameIndex, VkDeviceSize size, VkDeviceSize alignment, const VmaAllocationCreateInfo& createInfo, @@ -13120,7 +12239,6 @@ VkResult VmaBlockVector::AllocatePage( VmaAllocation* pAllocation) { const bool isUpperAddress = (createInfo.flags & VMA_ALLOCATION_CREATE_UPPER_ADDRESS_BIT) != 0; - bool canMakeOtherLost = (createInfo.flags & VMA_ALLOCATION_CREATE_CAN_MAKE_OTHER_LOST_BIT) != 0; const bool mapped = (createInfo.flags & VMA_ALLOCATION_CREATE_MAPPED_BIT) != 0; const bool isUserDataString = (createInfo.flags & VMA_ALLOCATION_CREATE_USER_DATA_COPY_STRING_BIT) != 0; @@ -13128,362 +12246,180 @@ VkResult VmaBlockVector::AllocatePage( { const uint32_t heapIndex = m_hAllocator->MemoryTypeIndexToHeapIndex(m_MemoryTypeIndex); VmaBudget heapBudget = {}; - m_hAllocator->GetBudget(&heapBudget, heapIndex, 1); + m_hAllocator->GetHeapBudgets(&heapBudget, heapIndex, 1); freeMemory = (heapBudget.usage < heapBudget.budget) ? (heapBudget.budget - heapBudget.usage) : 0; } - const bool canFallbackToDedicated = !IsCustomPool(); + const bool canFallbackToDedicated = !HasExplicitBlockSize() && + (createInfo.flags & VMA_ALLOCATION_CREATE_NEVER_ALLOCATE_BIT) == 0; const bool canCreateNewBlock = ((createInfo.flags & VMA_ALLOCATION_CREATE_NEVER_ALLOCATE_BIT) == 0) && (m_Blocks.size() < m_MaxBlockCount) && (freeMemory >= size || !canFallbackToDedicated); uint32_t strategy = createInfo.flags & VMA_ALLOCATION_CREATE_STRATEGY_MASK; - // If linearAlgorithm is used, canMakeOtherLost is available only when used as ring buffer. - // Which in turn is available only when maxBlockCount = 1. - if(m_Algorithm == VMA_POOL_CREATE_LINEAR_ALGORITHM_BIT && m_MaxBlockCount > 1) - { - canMakeOtherLost = false; - } - // Upper address can only be used with linear allocator and within single memory block. - if(isUpperAddress && + if (isUpperAddress && (m_Algorithm != VMA_POOL_CREATE_LINEAR_ALGORITHM_BIT || m_MaxBlockCount > 1)) { return VK_ERROR_FEATURE_NOT_PRESENT; } - // Validate strategy. - switch(strategy) - { - case 0: - strategy = VMA_ALLOCATION_CREATE_STRATEGY_BEST_FIT_BIT; - break; - case VMA_ALLOCATION_CREATE_STRATEGY_BEST_FIT_BIT: - case VMA_ALLOCATION_CREATE_STRATEGY_WORST_FIT_BIT: - case VMA_ALLOCATION_CREATE_STRATEGY_FIRST_FIT_BIT: - break; - default: - return VK_ERROR_FEATURE_NOT_PRESENT; - } - // Early reject: requested allocation size is larger that maximum block size for this block vector. - if(size + 2 * VMA_DEBUG_MARGIN > m_PreferredBlockSize) + if (size + VMA_DEBUG_MARGIN > m_PreferredBlockSize) { return VK_ERROR_OUT_OF_DEVICE_MEMORY; } - /* - Under certain condition, this whole section can be skipped for optimization, so - we move on directly to trying to allocate with canMakeOtherLost. That's the case - e.g. for custom pools with linear algorithm. - */ - if(!canMakeOtherLost || canCreateNewBlock) + // 1. Search existing allocations. Try to allocate. + if (m_Algorithm == VMA_POOL_CREATE_LINEAR_ALGORITHM_BIT) { - // 1. Search existing allocations. Try to allocate without making other allocations lost. - VmaAllocationCreateFlags allocFlagsCopy = createInfo.flags; - allocFlagsCopy &= ~VMA_ALLOCATION_CREATE_CAN_MAKE_OTHER_LOST_BIT; - - if(m_Algorithm == VMA_POOL_CREATE_LINEAR_ALGORITHM_BIT) + // Use only last block. + if (!m_Blocks.empty()) { - // Use only last block. - if(!m_Blocks.empty()) + VmaDeviceMemoryBlock* const pCurrBlock = m_Blocks.back(); + VMA_ASSERT(pCurrBlock); + VkResult res = AllocateFromBlock( + pCurrBlock, + size, + alignment, + createInfo.flags, + createInfo.pUserData, + suballocType, + strategy, + pAllocation); + if (res == VK_SUCCESS) + { + VMA_DEBUG_LOG(" Returned from last block #%u", pCurrBlock->GetId()); + return VK_SUCCESS; + } + } + } + else + { + if (strategy != VMA_ALLOCATION_CREATE_STRATEGY_MIN_TIME_BIT) // MIN_MEMORY or default + { + // Forward order in m_Blocks - prefer blocks with smallest amount of free space. + for (size_t blockIndex = 0; blockIndex < m_Blocks.size(); ++blockIndex) { - VmaDeviceMemoryBlock* const pCurrBlock = m_Blocks.back(); + VmaDeviceMemoryBlock* const pCurrBlock = m_Blocks[blockIndex]; VMA_ASSERT(pCurrBlock); VkResult res = AllocateFromBlock( pCurrBlock, - currentFrameIndex, size, alignment, - allocFlagsCopy, + createInfo.flags, createInfo.pUserData, suballocType, strategy, pAllocation); - if(res == VK_SUCCESS) + if (res == VK_SUCCESS) { - VMA_DEBUG_LOG(" Returned from last block #%u", pCurrBlock->GetId()); + VMA_DEBUG_LOG(" Returned from existing block #%u", pCurrBlock->GetId()); return VK_SUCCESS; } } } - else + else // VMA_ALLOCATION_CREATE_STRATEGY_MIN_TIME_BIT { - if(strategy == VMA_ALLOCATION_CREATE_STRATEGY_BEST_FIT_BIT) + // Backward order in m_Blocks - prefer blocks with largest amount of free space. + for (size_t blockIndex = m_Blocks.size(); blockIndex--; ) { - // Forward order in m_Blocks - prefer blocks with smallest amount of free space. - for(size_t blockIndex = 0; blockIndex < m_Blocks.size(); ++blockIndex ) - { - VmaDeviceMemoryBlock* const pCurrBlock = m_Blocks[blockIndex]; - VMA_ASSERT(pCurrBlock); - VkResult res = AllocateFromBlock( - pCurrBlock, - currentFrameIndex, - size, - alignment, - allocFlagsCopy, - createInfo.pUserData, - suballocType, - strategy, - pAllocation); - if(res == VK_SUCCESS) - { - VMA_DEBUG_LOG(" Returned from existing block #%u", pCurrBlock->GetId()); - return VK_SUCCESS; - } - } - } - else // WORST_FIT, FIRST_FIT - { - // Backward order in m_Blocks - prefer blocks with largest amount of free space. - for(size_t blockIndex = m_Blocks.size(); blockIndex--; ) - { - VmaDeviceMemoryBlock* const pCurrBlock = m_Blocks[blockIndex]; - VMA_ASSERT(pCurrBlock); - VkResult res = AllocateFromBlock( - pCurrBlock, - currentFrameIndex, - size, - alignment, - allocFlagsCopy, - createInfo.pUserData, - suballocType, - strategy, - pAllocation); - if(res == VK_SUCCESS) - { - VMA_DEBUG_LOG(" Returned from existing block #%u", pCurrBlock->GetId()); - return VK_SUCCESS; - } - } - } - } - - // 2. Try to create new block. - if(canCreateNewBlock) - { - // Calculate optimal size for new block. - VkDeviceSize newBlockSize = m_PreferredBlockSize; - uint32_t newBlockSizeShift = 0; - const uint32_t NEW_BLOCK_SIZE_SHIFT_MAX = 3; - - if(!m_ExplicitBlockSize) - { - // Allocate 1/8, 1/4, 1/2 as first blocks. - const VkDeviceSize maxExistingBlockSize = CalcMaxBlockSize(); - for(uint32_t i = 0; i < NEW_BLOCK_SIZE_SHIFT_MAX; ++i) - { - const VkDeviceSize smallerNewBlockSize = newBlockSize / 2; - if(smallerNewBlockSize > maxExistingBlockSize && smallerNewBlockSize >= size * 2) - { - newBlockSize = smallerNewBlockSize; - ++newBlockSizeShift; - } - else - { - break; - } - } - } - - size_t newBlockIndex = 0; - VkResult res = (newBlockSize <= freeMemory || !canFallbackToDedicated) ? - CreateBlock(newBlockSize, &newBlockIndex) : VK_ERROR_OUT_OF_DEVICE_MEMORY; - // Allocation of this size failed? Try 1/2, 1/4, 1/8 of m_PreferredBlockSize. - if(!m_ExplicitBlockSize) - { - while(res < 0 && newBlockSizeShift < NEW_BLOCK_SIZE_SHIFT_MAX) - { - const VkDeviceSize smallerNewBlockSize = newBlockSize / 2; - if(smallerNewBlockSize >= size) - { - newBlockSize = smallerNewBlockSize; - ++newBlockSizeShift; - res = (newBlockSize <= freeMemory || !canFallbackToDedicated) ? - CreateBlock(newBlockSize, &newBlockIndex) : VK_ERROR_OUT_OF_DEVICE_MEMORY; - } - else - { - break; - } - } - } - - if(res == VK_SUCCESS) - { - VmaDeviceMemoryBlock* const pBlock = m_Blocks[newBlockIndex]; - VMA_ASSERT(pBlock->m_pMetadata->GetSize() >= size); - - res = AllocateFromBlock( - pBlock, - currentFrameIndex, + VmaDeviceMemoryBlock* const pCurrBlock = m_Blocks[blockIndex]; + VMA_ASSERT(pCurrBlock); + VkResult res = AllocateFromBlock( + pCurrBlock, size, alignment, - allocFlagsCopy, + createInfo.flags, createInfo.pUserData, suballocType, strategy, pAllocation); - if(res == VK_SUCCESS) + if (res == VK_SUCCESS) { - VMA_DEBUG_LOG(" Created new block #%u Size=%llu", pBlock->GetId(), newBlockSize); + VMA_DEBUG_LOG(" Returned from existing block #%u", pCurrBlock->GetId()); return VK_SUCCESS; } - else - { - // Allocation from new block failed, possibly due to VMA_DEBUG_MARGIN or alignment. - return VK_ERROR_OUT_OF_DEVICE_MEMORY; - } } } } - // 3. Try to allocate from existing blocks with making other allocations lost. - if(canMakeOtherLost) + // 2. Try to create new block. + if (canCreateNewBlock) { - uint32_t tryIndex = 0; - for(; tryIndex < VMA_ALLOCATION_TRY_COUNT; ++tryIndex) - { - VmaDeviceMemoryBlock* pBestRequestBlock = VMA_NULL; - VmaAllocationRequest bestRequest = {}; - VkDeviceSize bestRequestCost = VK_WHOLE_SIZE; + // Calculate optimal size for new block. + VkDeviceSize newBlockSize = m_PreferredBlockSize; + uint32_t newBlockSizeShift = 0; + const uint32_t NEW_BLOCK_SIZE_SHIFT_MAX = 3; - // 1. Search existing allocations. - if(strategy == VMA_ALLOCATION_CREATE_STRATEGY_BEST_FIT_BIT) + if (!m_ExplicitBlockSize) + { + // Allocate 1/8, 1/4, 1/2 as first blocks. + const VkDeviceSize maxExistingBlockSize = CalcMaxBlockSize(); + for (uint32_t i = 0; i < NEW_BLOCK_SIZE_SHIFT_MAX; ++i) { - // Forward order in m_Blocks - prefer blocks with smallest amount of free space. - for(size_t blockIndex = 0; blockIndex < m_Blocks.size(); ++blockIndex ) + const VkDeviceSize smallerNewBlockSize = newBlockSize / 2; + if (smallerNewBlockSize > maxExistingBlockSize && smallerNewBlockSize >= size * 2) { - VmaDeviceMemoryBlock* const pCurrBlock = m_Blocks[blockIndex]; - VMA_ASSERT(pCurrBlock); - VmaAllocationRequest currRequest = {}; - if(pCurrBlock->m_pMetadata->CreateAllocationRequest( - currentFrameIndex, - m_FrameInUseCount, - m_BufferImageGranularity, - size, - alignment, - (createInfo.flags & VMA_ALLOCATION_CREATE_UPPER_ADDRESS_BIT) != 0, - suballocType, - canMakeOtherLost, - strategy, - &currRequest)) - { - const VkDeviceSize currRequestCost = currRequest.CalcCost(); - if(pBestRequestBlock == VMA_NULL || - currRequestCost < bestRequestCost) - { - pBestRequestBlock = pCurrBlock; - bestRequest = currRequest; - bestRequestCost = currRequestCost; - - if(bestRequestCost == 0) - { - break; - } - } - } + newBlockSize = smallerNewBlockSize; + ++newBlockSizeShift; } - } - else // WORST_FIT, FIRST_FIT - { - // Backward order in m_Blocks - prefer blocks with largest amount of free space. - for(size_t blockIndex = m_Blocks.size(); blockIndex--; ) + else { - VmaDeviceMemoryBlock* const pCurrBlock = m_Blocks[blockIndex]; - VMA_ASSERT(pCurrBlock); - VmaAllocationRequest currRequest = {}; - if(pCurrBlock->m_pMetadata->CreateAllocationRequest( - currentFrameIndex, - m_FrameInUseCount, - m_BufferImageGranularity, - size, - alignment, - (createInfo.flags & VMA_ALLOCATION_CREATE_UPPER_ADDRESS_BIT) != 0, - suballocType, - canMakeOtherLost, - strategy, - &currRequest)) - { - const VkDeviceSize currRequestCost = currRequest.CalcCost(); - if(pBestRequestBlock == VMA_NULL || - currRequestCost < bestRequestCost || - strategy == VMA_ALLOCATION_CREATE_STRATEGY_FIRST_FIT_BIT) - { - pBestRequestBlock = pCurrBlock; - bestRequest = currRequest; - bestRequestCost = currRequestCost; - - if(bestRequestCost == 0 || - strategy == VMA_ALLOCATION_CREATE_STRATEGY_FIRST_FIT_BIT) - { - break; - } - } - } + break; } } + } - if(pBestRequestBlock != VMA_NULL) + size_t newBlockIndex = 0; + VkResult res = (newBlockSize <= freeMemory || !canFallbackToDedicated) ? + CreateBlock(newBlockSize, &newBlockIndex) : VK_ERROR_OUT_OF_DEVICE_MEMORY; + // Allocation of this size failed? Try 1/2, 1/4, 1/8 of m_PreferredBlockSize. + if (!m_ExplicitBlockSize) + { + while (res < 0 && newBlockSizeShift < NEW_BLOCK_SIZE_SHIFT_MAX) { - if(mapped) + const VkDeviceSize smallerNewBlockSize = newBlockSize / 2; + if (smallerNewBlockSize >= size) { - VkResult res = pBestRequestBlock->Map(m_hAllocator, 1, VMA_NULL); - if(res != VK_SUCCESS) - { - return res; - } + newBlockSize = smallerNewBlockSize; + ++newBlockSizeShift; + res = (newBlockSize <= freeMemory || !canFallbackToDedicated) ? + CreateBlock(newBlockSize, &newBlockIndex) : VK_ERROR_OUT_OF_DEVICE_MEMORY; } - - if(pBestRequestBlock->m_pMetadata->MakeRequestedAllocationsLost( - currentFrameIndex, - m_FrameInUseCount, - &bestRequest)) + else { - // Allocate from this pBlock. - *pAllocation = m_hAllocator->m_AllocationObjectAllocator.Allocate(currentFrameIndex, isUserDataString); - pBestRequestBlock->m_pMetadata->Alloc(bestRequest, suballocType, size, *pAllocation); - UpdateHasEmptyBlock(); - (*pAllocation)->InitBlockAllocation( - pBestRequestBlock, - bestRequest.offset, - alignment, - size, - m_MemoryTypeIndex, - suballocType, - mapped, - (createInfo.flags & VMA_ALLOCATION_CREATE_CAN_BECOME_LOST_BIT) != 0); - VMA_HEAVY_ASSERT(pBestRequestBlock->Validate()); - VMA_DEBUG_LOG(" Returned from existing block"); - (*pAllocation)->SetUserData(m_hAllocator, createInfo.pUserData); - m_hAllocator->m_Budget.AddAllocation(m_hAllocator->MemoryTypeIndexToHeapIndex(m_MemoryTypeIndex), size); - if(VMA_DEBUG_INITIALIZE_ALLOCATIONS) - { - m_hAllocator->FillAllocation(*pAllocation, VMA_ALLOCATION_FILL_PATTERN_CREATED); - } - if(IsCorruptionDetectionEnabled()) - { - VkResult res = pBestRequestBlock->WriteMagicValueAroundAllocation(m_hAllocator, bestRequest.offset, size); - VMA_ASSERT(res == VK_SUCCESS && "Couldn't map block memory to write magic value."); - } - return VK_SUCCESS; + break; } - // else: Some allocations must have been touched while we are here. Next try. + } + } + + if (res == VK_SUCCESS) + { + VmaDeviceMemoryBlock* const pBlock = m_Blocks[newBlockIndex]; + VMA_ASSERT(pBlock->m_pMetadata->GetSize() >= size); + + res = AllocateFromBlock( + pBlock, + size, + alignment, + createInfo.flags, + createInfo.pUserData, + suballocType, + strategy, + pAllocation); + if (res == VK_SUCCESS) + { + VMA_DEBUG_LOG(" Created new block #%u Size=%llu", pBlock->GetId(), newBlockSize); + return VK_SUCCESS; } else { - // Could not find place in any of the blocks - break outer loop. - break; + // Allocation from new block failed, possibly due to VMA_DEBUG_MARGIN or alignment. + return VK_ERROR_OUT_OF_DEVICE_MEMORY; } } - /* Maximum number of tries exceeded - a very unlike event when many other - threads are simultaneously touching allocations making it impossible to make - lost at the same time as we try to allocate. */ - if(tryIndex == VMA_ALLOCATION_TRY_COUNT) - { - return VK_ERROR_TOO_MANY_OBJECTS; - } } return VK_ERROR_OUT_OF_DEVICE_MEMORY; @@ -13492,15 +12428,13 @@ VkResult VmaBlockVector::AllocatePage( void VmaBlockVector::Free( const VmaAllocation hAllocation) { - VMA_ASSERT(hAllocation->GetBlock()->GetParentBlockVector() == this); - VmaDeviceMemoryBlock* pBlockToDelete = VMA_NULL; bool budgetExceeded = false; { const uint32_t heapIndex = m_hAllocator->MemoryTypeIndexToHeapIndex(m_MemoryTypeIndex); VmaBudget heapBudget = {}; - m_hAllocator->GetBudget(&heapBudget, heapIndex, 1); + m_hAllocator->GetHeapBudgets(&heapBudget, heapIndex, 1); budgetExceeded = heapBudget.usage >= heapBudget.budget; } @@ -13510,28 +12444,28 @@ void VmaBlockVector::Free( VmaDeviceMemoryBlock* pBlock = hAllocation->GetBlock(); - if(IsCorruptionDetectionEnabled()) + if (IsCorruptionDetectionEnabled()) { - VkResult res = pBlock->ValidateMagicValueAroundAllocation(m_hAllocator, hAllocation->GetOffset(), hAllocation->GetSize()); + VkResult res = pBlock->ValidateMagicValueAfterAllocation(m_hAllocator, hAllocation->GetOffset(), hAllocation->GetSize()); VMA_ASSERT(res == VK_SUCCESS && "Couldn't map block memory to validate magic value."); } - if(hAllocation->IsPersistentMap()) + if (hAllocation->IsPersistentMap()) { pBlock->Unmap(m_hAllocator, 1); } - pBlock->m_pMetadata->Free(hAllocation); + pBlock->m_pMetadata->Free(hAllocation->GetAllocHandle()); VMA_HEAVY_ASSERT(pBlock->Validate()); VMA_DEBUG_LOG(" Freed from MemoryTypeIndex=%u", m_MemoryTypeIndex); const bool canDeleteBlock = m_Blocks.size() > m_MinBlockCount; // pBlock became empty after this deallocation. - if(pBlock->m_pMetadata->IsEmpty()) + if (pBlock->m_pMetadata->IsEmpty()) { // Already has empty block. We don't want to have two, so delete this one. - if((m_HasEmptyBlock || budgetExceeded) && canDeleteBlock) + if ((m_HasEmptyBlock || budgetExceeded) && canDeleteBlock) { pBlockToDelete = pBlock; Remove(pBlock); @@ -13540,10 +12474,10 @@ void VmaBlockVector::Free( } // pBlock didn't become empty, but we have another empty block - find and free that one. // (This is optional, heuristics.) - else if(m_HasEmptyBlock && canDeleteBlock) + else if (m_HasEmptyBlock && canDeleteBlock) { VmaDeviceMemoryBlock* pLastBlock = m_Blocks.back(); - if(pLastBlock->m_pMetadata->IsEmpty()) + if (pLastBlock->m_pMetadata->IsEmpty()) { pBlockToDelete = pLastBlock; m_Blocks.pop_back(); @@ -13556,9 +12490,9 @@ void VmaBlockVector::Free( // Destruction of a free block. Deferred until this point, outside of mutex // lock, for performance reason. - if(pBlockToDelete != VMA_NULL) + if (pBlockToDelete != VMA_NULL) { - VMA_DEBUG_LOG(" Deleted empty block"); + VMA_DEBUG_LOG(" Deleted empty block #%u", pBlockToDelete->GetId()); pBlockToDelete->Destroy(m_hAllocator); vma_delete(m_hAllocator, pBlockToDelete); } @@ -13567,10 +12501,10 @@ void VmaBlockVector::Free( VkDeviceSize VmaBlockVector::CalcMaxBlockSize() const { VkDeviceSize result = 0; - for(size_t i = m_Blocks.size(); i--; ) + for (size_t i = m_Blocks.size(); i--; ) { result = VMA_MAX(result, m_Blocks[i]->m_pMetadata->GetSize()); - if(result >= m_PreferredBlockSize) + if (result >= m_PreferredBlockSize) { break; } @@ -13580,9 +12514,9 @@ VkDeviceSize VmaBlockVector::CalcMaxBlockSize() const void VmaBlockVector::Remove(VmaDeviceMemoryBlock* pBlock) { - for(uint32_t blockIndex = 0; blockIndex < m_Blocks.size(); ++blockIndex) + for (uint32_t blockIndex = 0; blockIndex < m_Blocks.size(); ++blockIndex) { - if(m_Blocks[blockIndex] == pBlock) + if (m_Blocks[blockIndex] == pBlock) { VmaVectorRemove(m_Blocks, blockIndex); return; @@ -13593,12 +12527,12 @@ void VmaBlockVector::Remove(VmaDeviceMemoryBlock* pBlock) void VmaBlockVector::IncrementallySortBlocks() { - if(m_Algorithm != VMA_POOL_CREATE_LINEAR_ALGORITHM_BIT) + if (m_Algorithm != VMA_POOL_CREATE_LINEAR_ALGORITHM_BIT) { // Bubble sort only until first swap. - for(size_t i = 1; i < m_Blocks.size(); ++i) + for (size_t i = 1; i < m_Blocks.size(); ++i) { - if(m_Blocks[i - 1]->m_pMetadata->GetSumFreeSize() > m_Blocks[i]->m_pMetadata->GetSumFreeSize()) + if (m_Blocks[i - 1]->m_pMetadata->GetSumFreeSize() > m_Blocks[i]->m_pMetadata->GetSumFreeSize()) { VMA_SWAP(m_Blocks[i - 1], m_Blocks[i]); return; @@ -13609,7 +12543,6 @@ void VmaBlockVector::IncrementallySortBlocks() VkResult VmaBlockVector::AllocateFromBlock( VmaDeviceMemoryBlock* pBlock, - uint32_t currentFrameIndex, VkDeviceSize size, VkDeviceSize alignment, VmaAllocationCreateFlags allocFlags, @@ -13618,58 +12551,50 @@ VkResult VmaBlockVector::AllocateFromBlock( uint32_t strategy, VmaAllocation* pAllocation) { - VMA_ASSERT((allocFlags & VMA_ALLOCATION_CREATE_CAN_MAKE_OTHER_LOST_BIT) == 0); const bool isUpperAddress = (allocFlags & VMA_ALLOCATION_CREATE_UPPER_ADDRESS_BIT) != 0; const bool mapped = (allocFlags & VMA_ALLOCATION_CREATE_MAPPED_BIT) != 0; const bool isUserDataString = (allocFlags & VMA_ALLOCATION_CREATE_USER_DATA_COPY_STRING_BIT) != 0; VmaAllocationRequest currRequest = {}; - if(pBlock->m_pMetadata->CreateAllocationRequest( - currentFrameIndex, - m_FrameInUseCount, - m_BufferImageGranularity, + if (pBlock->m_pMetadata->CreateAllocationRequest( size, alignment, isUpperAddress, suballocType, - false, // canMakeOtherLost strategy, &currRequest)) { // Allocate from pCurrBlock. - VMA_ASSERT(currRequest.itemsToMakeLostCount == 0); - - if(mapped) + if (mapped) { VkResult res = pBlock->Map(m_hAllocator, 1, VMA_NULL); - if(res != VK_SUCCESS) + if (res != VK_SUCCESS) { return res; } } - *pAllocation = m_hAllocator->m_AllocationObjectAllocator.Allocate(currentFrameIndex, isUserDataString); - pBlock->m_pMetadata->Alloc(currRequest, suballocType, size, *pAllocation); + *pAllocation = m_hAllocator->m_AllocationObjectAllocator.Allocate(isUserDataString); + pBlock->m_pMetadata->Alloc(currRequest, suballocType, *pAllocation); UpdateHasEmptyBlock(); (*pAllocation)->InitBlockAllocation( pBlock, - currRequest.offset, + currRequest.allocHandle, alignment, - size, + currRequest.size, // Not size, as actual allocation size may be larger than requested! m_MemoryTypeIndex, suballocType, - mapped, - (allocFlags & VMA_ALLOCATION_CREATE_CAN_BECOME_LOST_BIT) != 0); + mapped); VMA_HEAVY_ASSERT(pBlock->Validate()); (*pAllocation)->SetUserData(m_hAllocator, pUserData); - m_hAllocator->m_Budget.AddAllocation(m_hAllocator->MemoryTypeIndexToHeapIndex(m_MemoryTypeIndex), size); - if(VMA_DEBUG_INITIALIZE_ALLOCATIONS) + m_hAllocator->m_Budget.AddAllocation(m_hAllocator->MemoryTypeIndexToHeapIndex(m_MemoryTypeIndex), currRequest.size); + if (VMA_DEBUG_INITIALIZE_ALLOCATIONS) { m_hAllocator->FillAllocation(*pAllocation, VMA_ALLOCATION_FILL_PATTERN_CREATED); } - if(IsCorruptionDetectionEnabled()) + if (IsCorruptionDetectionEnabled()) { - VkResult res = pBlock->WriteMagicValueAroundAllocation(m_hAllocator, currRequest.offset, size); + VkResult res = pBlock->WriteMagicValueAfterAllocation(m_hAllocator, (*pAllocation)->GetOffset(), currRequest.size); VMA_ASSERT(res == VK_SUCCESS && "Couldn't map block memory to write magic value."); } return VK_SUCCESS; @@ -13687,35 +12612,35 @@ VkResult VmaBlockVector::CreateBlock(VkDeviceSize blockSize, size_t* pNewBlockIn #if VMA_BUFFER_DEVICE_ADDRESS // Every standalone block can potentially contain a buffer with VK_BUFFER_USAGE_SHADER_DEVICE_ADDRESS_BIT - always enable the feature. VkMemoryAllocateFlagsInfoKHR allocFlagsInfo = { VK_STRUCTURE_TYPE_MEMORY_ALLOCATE_FLAGS_INFO_KHR }; - if(m_hAllocator->m_UseKhrBufferDeviceAddress) + if (m_hAllocator->m_UseKhrBufferDeviceAddress) { allocFlagsInfo.flags = VK_MEMORY_ALLOCATE_DEVICE_ADDRESS_BIT_KHR; VmaPnextChainPushFront(&allocInfo, &allocFlagsInfo); } -#endif // #if VMA_BUFFER_DEVICE_ADDRESS +#endif // VMA_BUFFER_DEVICE_ADDRESS #if VMA_MEMORY_PRIORITY VkMemoryPriorityAllocateInfoEXT priorityInfo = { VK_STRUCTURE_TYPE_MEMORY_PRIORITY_ALLOCATE_INFO_EXT }; - if(m_hAllocator->m_UseExtMemoryPriority) + if (m_hAllocator->m_UseExtMemoryPriority) { priorityInfo.priority = m_Priority; VmaPnextChainPushFront(&allocInfo, &priorityInfo); } -#endif // #if VMA_MEMORY_PRIORITY +#endif // VMA_MEMORY_PRIORITY #if VMA_EXTERNAL_MEMORY // Attach VkExportMemoryAllocateInfoKHR if necessary. VkExportMemoryAllocateInfoKHR exportMemoryAllocInfo = { VK_STRUCTURE_TYPE_EXPORT_MEMORY_ALLOCATE_INFO_KHR }; exportMemoryAllocInfo.handleTypes = m_hAllocator->GetExternalMemoryHandleTypeFlags(m_MemoryTypeIndex); - if(exportMemoryAllocInfo.handleTypes != 0) + if (exportMemoryAllocInfo.handleTypes != 0) { VmaPnextChainPushFront(&allocInfo, &exportMemoryAllocInfo); } -#endif // #if VMA_EXTERNAL_MEMORY +#endif // VMA_EXTERNAL_MEMORY VkDeviceMemory mem = VK_NULL_HANDLE; VkResult res = m_hAllocator->AllocateVulkanMemory(&allocInfo, &mem); - if(res < 0) + if (res < 0) { return res; } @@ -13726,16 +12651,16 @@ VkResult VmaBlockVector::CreateBlock(VkDeviceSize blockSize, size_t* pNewBlockIn VmaDeviceMemoryBlock* const pBlock = vma_new(m_hAllocator, VmaDeviceMemoryBlock)(m_hAllocator); pBlock->Init( m_hAllocator, - this, // parentBlockVector m_hParentPool, m_MemoryTypeIndex, mem, allocInfo.allocationSize, m_NextBlockId++, - m_Algorithm); + m_Algorithm, + m_BufferImageGranularity); m_Blocks.push_back(pBlock); - if(pNewBlockIndex != VMA_NULL) + if (pNewBlockIndex != VMA_NULL) { *pNewBlockIndex = m_Blocks.size() - 1; } @@ -13744,8 +12669,8 @@ VkResult VmaBlockVector::CreateBlock(VkDeviceSize blockSize, size_t* pNewBlockIn } void VmaBlockVector::ApplyDefragmentationMovesCpu( - class VmaBlockVectorDefragmentationContext* pDefragCtx, - const VmaVector< VmaDefragmentationMove, VmaStlAllocator<VmaDefragmentationMove> >& moves) + VmaBlockVectorDefragmentationContext* pDefragCtx, + const VmaVector<VmaDefragmentationMove, VmaStlAllocator<VmaDefragmentationMove>>& moves) { const size_t blockCount = m_Blocks.size(); const bool isNonCoherent = m_hAllocator->IsMemoryTypeNonCoherent(m_MemoryTypeIndex); @@ -13767,7 +12692,7 @@ void VmaBlockVector::ApplyDefragmentationMovesCpu( // Go over all moves. Mark blocks that are used with BLOCK_FLAG_USED. const size_t moveCount = moves.size(); - for(size_t moveIndex = 0; moveIndex < moveCount; ++moveIndex) + for (size_t moveIndex = 0; moveIndex < moveCount; ++moveIndex) { const VmaDefragmentationMove& move = moves[moveIndex]; blockInfo[move.srcBlockIndex].flags |= BLOCK_FLAG_USED; @@ -13777,18 +12702,18 @@ void VmaBlockVector::ApplyDefragmentationMovesCpu( VMA_ASSERT(pDefragCtx->res == VK_SUCCESS); // Go over all blocks. Get mapped pointer or map if necessary. - for(size_t blockIndex = 0; pDefragCtx->res == VK_SUCCESS && blockIndex < blockCount; ++blockIndex) + for (size_t blockIndex = 0; pDefragCtx->res == VK_SUCCESS && blockIndex < blockCount; ++blockIndex) { BlockInfo& currBlockInfo = blockInfo[blockIndex]; VmaDeviceMemoryBlock* pBlock = m_Blocks[blockIndex]; - if((currBlockInfo.flags & BLOCK_FLAG_USED) != 0) + if ((currBlockInfo.flags & BLOCK_FLAG_USED) != 0) { currBlockInfo.pMappedData = pBlock->GetMappedData(); // It is not originally mapped - map it. - if(currBlockInfo.pMappedData == VMA_NULL) + if (currBlockInfo.pMappedData == VMA_NULL) { pDefragCtx->res = pBlock->Map(m_hAllocator, 1, &currBlockInfo.pMappedData); - if(pDefragCtx->res == VK_SUCCESS) + if (pDefragCtx->res == VK_SUCCESS) { currBlockInfo.flags |= BLOCK_FLAG_MAPPED_FOR_DEFRAGMENTATION; } @@ -13797,12 +12722,12 @@ void VmaBlockVector::ApplyDefragmentationMovesCpu( } // Go over all moves. Do actual data transfer. - if(pDefragCtx->res == VK_SUCCESS) + if (pDefragCtx->res == VK_SUCCESS) { const VkDeviceSize nonCoherentAtomSize = m_hAllocator->m_PhysicalDeviceProperties.limits.nonCoherentAtomSize; VkMappedMemoryRange memRange = { VK_STRUCTURE_TYPE_MAPPED_MEMORY_RANGE }; - for(size_t moveIndex = 0; moveIndex < moveCount; ++moveIndex) + for (size_t moveIndex = 0; moveIndex < moveCount; ++moveIndex) { const VmaDefragmentationMove& move = moves[moveIndex]; @@ -13812,7 +12737,7 @@ void VmaBlockVector::ApplyDefragmentationMovesCpu( VMA_ASSERT(srcBlockInfo.pMappedData && dstBlockInfo.pMappedData); // Invalidate source. - if(isNonCoherent) + if (isNonCoherent) { VmaDeviceMemoryBlock* const pSrcBlock = m_Blocks[move.srcBlockIndex]; memRange.memory = pSrcBlock->GetDeviceMemory(); @@ -13829,14 +12754,13 @@ void VmaBlockVector::ApplyDefragmentationMovesCpu( reinterpret_cast<char*>(srcBlockInfo.pMappedData) + move.srcOffset, static_cast<size_t>(move.size)); - if(IsCorruptionDetectionEnabled()) + if (IsCorruptionDetectionEnabled()) { - VmaWriteMagicValue(dstBlockInfo.pMappedData, move.dstOffset - VMA_DEBUG_MARGIN); VmaWriteMagicValue(dstBlockInfo.pMappedData, move.dstOffset + move.size); } // Flush destination. - if(isNonCoherent) + if (isNonCoherent) { VmaDeviceMemoryBlock* const pDstBlock = m_Blocks[move.dstBlockIndex]; memRange.memory = pDstBlock->GetDeviceMemory(); @@ -13851,10 +12775,10 @@ void VmaBlockVector::ApplyDefragmentationMovesCpu( // Go over all blocks in reverse order. Unmap those that were mapped just for defragmentation. // Regardless of pCtx->res == VK_SUCCESS. - for(size_t blockIndex = blockCount; blockIndex--; ) + for (size_t blockIndex = blockCount; blockIndex--; ) { const BlockInfo& currBlockInfo = blockInfo[blockIndex]; - if((currBlockInfo.flags & BLOCK_FLAG_MAPPED_FOR_DEFRAGMENTATION) != 0) + if ((currBlockInfo.flags & BLOCK_FLAG_MAPPED_FOR_DEFRAGMENTATION) != 0) { VmaDeviceMemoryBlock* pBlock = m_Blocks[blockIndex]; pBlock->Unmap(m_hAllocator, 1); @@ -13863,8 +12787,8 @@ void VmaBlockVector::ApplyDefragmentationMovesCpu( } void VmaBlockVector::ApplyDefragmentationMovesGpu( - class VmaBlockVectorDefragmentationContext* pDefragCtx, - VmaVector< VmaDefragmentationMove, VmaStlAllocator<VmaDefragmentationMove> >& moves, + VmaBlockVectorDefragmentationContext* pDefragCtx, + VmaVector<VmaDefragmentationMove, VmaStlAllocator<VmaDefragmentationMove>>& moves, VkCommandBuffer commandBuffer) { const size_t blockCount = m_Blocks.size(); @@ -13874,7 +12798,7 @@ void VmaBlockVector::ApplyDefragmentationMovesGpu( // Go over all moves. Mark blocks that are used with BLOCK_FLAG_USED. const size_t moveCount = moves.size(); - for(size_t moveIndex = 0; moveIndex < moveCount; ++moveIndex) + for (size_t moveIndex = 0; moveIndex < moveCount; ++moveIndex) { const VmaDefragmentationMove& move = moves[moveIndex]; @@ -13893,16 +12817,16 @@ void VmaBlockVector::ApplyDefragmentationMovesGpu( VkBufferCreateInfo bufCreateInfo; VmaFillGpuDefragmentationBufferCreateInfo(bufCreateInfo); - for(size_t blockIndex = 0; pDefragCtx->res == VK_SUCCESS && blockIndex < blockCount; ++blockIndex) + for (size_t blockIndex = 0; pDefragCtx->res == VK_SUCCESS && blockIndex < blockCount; ++blockIndex) { VmaBlockDefragmentationContext& currBlockCtx = pDefragCtx->blockContexts[blockIndex]; VmaDeviceMemoryBlock* pBlock = m_Blocks[blockIndex]; - if((currBlockCtx.flags & VmaBlockDefragmentationContext::BLOCK_FLAG_USED) != 0) + if ((currBlockCtx.flags & VmaBlockDefragmentationContext::BLOCK_FLAG_USED) != 0) { bufCreateInfo.size = pBlock->m_pMetadata->GetSize(); pDefragCtx->res = (*m_hAllocator->GetVulkanFunctions().vkCreateBuffer)( m_hAllocator->m_hDevice, &bufCreateInfo, m_hAllocator->GetAllocationCallbacks(), &currBlockCtx.hBuffer); - if(pDefragCtx->res == VK_SUCCESS) + if (pDefragCtx->res == VK_SUCCESS) { pDefragCtx->res = (*m_hAllocator->GetVulkanFunctions().vkBindBufferMemory)( m_hAllocator->m_hDevice, currBlockCtx.hBuffer, pBlock->GetDeviceMemory(), 0); @@ -13912,9 +12836,9 @@ void VmaBlockVector::ApplyDefragmentationMovesGpu( } // Go over all moves. Post data transfer commands to command buffer. - if(pDefragCtx->res == VK_SUCCESS) + if (pDefragCtx->res == VK_SUCCESS) { - for(size_t moveIndex = 0; moveIndex < moveCount; ++moveIndex) + for (size_t moveIndex = 0; moveIndex < moveCount; ++moveIndex) { const VmaDefragmentationMove& move = moves[moveIndex]; @@ -13933,7 +12857,7 @@ void VmaBlockVector::ApplyDefragmentationMovesGpu( } // Save buffers to defrag context for later destruction. - if(pDefragCtx->res == VK_SUCCESS && moveCount > 0) + if (pDefragCtx->res == VK_SUCCESS && moveCount > 0) { pDefragCtx->res = VK_NOT_READY; } @@ -13941,14 +12865,14 @@ void VmaBlockVector::ApplyDefragmentationMovesGpu( void VmaBlockVector::FreeEmptyBlocks(VmaDefragmentationStats* pDefragmentationStats) { - for(size_t blockIndex = m_Blocks.size(); blockIndex--; ) + for (size_t blockIndex = m_Blocks.size(); blockIndex--; ) { VmaDeviceMemoryBlock* pBlock = m_Blocks[blockIndex]; - if(pBlock->m_pMetadata->IsEmpty()) + if (pBlock->m_pMetadata->IsEmpty()) { - if(m_Blocks.size() > m_MinBlockCount) + if (m_Blocks.size() > m_MinBlockCount) { - if(pDefragmentationStats != VMA_NULL) + if (pDefragmentationStats != VMA_NULL) { ++pDefragmentationStats->deviceMemoryBlocksFreed; pDefragmentationStats->bytesFreed += pBlock->m_pMetadata->GetSize(); @@ -13970,10 +12894,10 @@ void VmaBlockVector::FreeEmptyBlocks(VmaDefragmentationStats* pDefragmentationSt void VmaBlockVector::UpdateHasEmptyBlock() { m_HasEmptyBlock = false; - for(size_t index = 0, count = m_Blocks.size(); index < count; ++index) + for (size_t index = 0, count = m_Blocks.size(); index < count; ++index) { VmaDeviceMemoryBlock* const pBlock = m_Blocks[index]; - if(pBlock->m_pMetadata->IsEmpty()) + if (pBlock->m_pMetadata->IsEmpty()) { m_HasEmptyBlock = true; break; @@ -13982,17 +12906,14 @@ void VmaBlockVector::UpdateHasEmptyBlock() } #if VMA_STATS_STRING_ENABLED - void VmaBlockVector::PrintDetailedMap(class VmaJsonWriter& json) { VmaMutexLockRead lock(m_Mutex, m_hAllocator->m_UseMutex); - json.BeginObject(); - - if(IsCustomPool()) + if (IsCustomPool()) { const char* poolName = m_hParentPool->GetName(); - if(poolName != VMA_NULL && poolName[0] != '\0') + if (poolName != VMA_NULL && poolName[0] != '\0') { json.WriteString("Name"); json.WriteString(poolName); @@ -14006,12 +12927,12 @@ void VmaBlockVector::PrintDetailedMap(class VmaJsonWriter& json) json.WriteString("BlockCount"); json.BeginObject(true); - if(m_MinBlockCount > 0) + if (m_MinBlockCount > 0) { json.WriteString("Min"); json.WriteNumber((uint64_t)m_MinBlockCount); } - if(m_MaxBlockCount < SIZE_MAX) + if (m_MaxBlockCount < SIZE_MAX) { json.WriteString("Max"); json.WriteNumber((uint64_t)m_MaxBlockCount); @@ -14020,13 +12941,7 @@ void VmaBlockVector::PrintDetailedMap(class VmaJsonWriter& json) json.WriteNumber((uint64_t)m_Blocks.size()); json.EndObject(); - if(m_FrameInUseCount > 0) - { - json.WriteString("FrameInUseCount"); - json.WriteNumber(m_FrameInUseCount); - } - - if(m_Algorithm != 0) + if (m_Algorithm != 0) { json.WriteString("Algorithm"); json.WriteString(VmaAlgorithmToStr(m_Algorithm)); @@ -14040,7 +12955,7 @@ void VmaBlockVector::PrintDetailedMap(class VmaJsonWriter& json) json.WriteString("Blocks"); json.BeginObject(); - for(size_t i = 0; i < m_Blocks.size(); ++i) + for (size_t i = 0; i < m_Blocks.size(); ++i) { json.BeginString(); json.ContinueString(m_Blocks[i]->GetId()); @@ -14049,11 +12964,8 @@ void VmaBlockVector::PrintDetailedMap(class VmaJsonWriter& json) m_Blocks[i]->m_pMetadata->PrintDetailedMap(json); } json.EndObject(); - - json.EndObject(); } - -#endif // #if VMA_STATS_STRING_ENABLED +#endif // VMA_STATS_STRING_ENABLED void VmaBlockVector::Defragment( class VmaBlockVectorDefragmentationContext* pCtx, @@ -14075,11 +12987,11 @@ void VmaBlockVector::Defragment( ((1u << m_MemoryTypeIndex) & m_hAllocator->GetGpuDefragmentationMemoryTypeBits()) != 0; // There are options to defragment this memory type. - if(canDefragmentOnCpu || canDefragmentOnGpu) + if (canDefragmentOnCpu || canDefragmentOnGpu) { bool defragmentOnGpu; // There is only one option to defragment this memory type. - if(canDefragmentOnGpu != canDefragmentOnCpu) + if (canDefragmentOnGpu != canDefragmentOnCpu) { defragmentOnGpu = canDefragmentOnGpu; } @@ -14092,11 +13004,11 @@ void VmaBlockVector::Defragment( bool overlappingMoveSupported = !defragmentOnGpu; - if(m_hAllocator->m_UseMutex) + if (m_hAllocator->m_UseMutex) { - if(flags & VMA_DEFRAGMENTATION_FLAG_INCREMENTAL) + if (flags & VMA_DEFRAGMENTATION_FLAG_INCREMENTAL) { - if(!m_Mutex.TryLockWrite()) + if (!m_Mutex.TryLockWrite()) { pCtx->res = VK_ERROR_INITIALIZATION_FAILED; return; @@ -14115,18 +13027,19 @@ void VmaBlockVector::Defragment( const VkDeviceSize maxBytesToMove = defragmentOnGpu ? maxGpuBytesToMove : maxCpuBytesToMove; const uint32_t maxAllocationsToMove = defragmentOnGpu ? maxGpuAllocationsToMove : maxCpuAllocationsToMove; - pCtx->res = pCtx->GetAlgorithm()->Defragment(pCtx->defragmentationMoves, maxBytesToMove, maxAllocationsToMove, flags); + VmaDefragmentationAlgorithm* algo = pCtx->GetAlgorithm(); + pCtx->res = algo->Defragment(pCtx->defragmentationMoves, maxBytesToMove, maxAllocationsToMove, flags); // Accumulate statistics. - if(pStats != VMA_NULL) + if (pStats != VMA_NULL) { - const VkDeviceSize bytesMoved = pCtx->GetAlgorithm()->GetBytesMoved(); - const uint32_t allocationsMoved = pCtx->GetAlgorithm()->GetAllocationsMoved(); + const VkDeviceSize bytesMoved = algo->GetBytesMoved(); + const uint32_t allocationsMoved = algo->GetAllocationsMoved(); pStats->bytesMoved += bytesMoved; pStats->allocationsMoved += allocationsMoved; VMA_ASSERT(bytesMoved <= maxBytesToMove); VMA_ASSERT(allocationsMoved <= maxAllocationsToMove); - if(defragmentOnGpu) + if (defragmentOnGpu) { maxGpuBytesToMove -= bytesMoved; maxGpuAllocationsToMove -= allocationsMoved; @@ -14138,20 +13051,20 @@ void VmaBlockVector::Defragment( } } - if(flags & VMA_DEFRAGMENTATION_FLAG_INCREMENTAL) + if (flags & VMA_DEFRAGMENTATION_FLAG_INCREMENTAL) { - if(m_hAllocator->m_UseMutex) + if (m_hAllocator->m_UseMutex) m_Mutex.UnlockWrite(); - if(pCtx->res >= VK_SUCCESS && !pCtx->defragmentationMoves.empty()) + if (pCtx->res >= VK_SUCCESS && !pCtx->defragmentationMoves.empty()) pCtx->res = VK_NOT_READY; return; } - if(pCtx->res >= VK_SUCCESS) + if (pCtx->res >= VK_SUCCESS) { - if(defragmentOnGpu) + if (defragmentOnGpu) { ApplyDefragmentationMovesGpu(pCtx, pCtx->defragmentationMoves, commandBuffer); } @@ -14168,7 +13081,7 @@ void VmaBlockVector::DefragmentationEnd( uint32_t flags, VmaDefragmentationStats* pStats) { - if(flags & VMA_DEFRAGMENTATION_FLAG_INCREMENTAL && m_hAllocator->m_UseMutex) + if (flags & VMA_DEFRAGMENTATION_FLAG_INCREMENTAL && m_hAllocator->m_UseMutex) { VMA_ASSERT(pCtx->mutexLocked == false); @@ -14179,25 +13092,25 @@ void VmaBlockVector::DefragmentationEnd( } // If the mutex isn't locked we didn't do any work and there is nothing to delete. - if(pCtx->mutexLocked || !m_hAllocator->m_UseMutex) + if (pCtx->mutexLocked || !m_hAllocator->m_UseMutex) { // Destroy buffers. - for(size_t blockIndex = pCtx->blockContexts.size(); blockIndex--;) + for (size_t blockIndex = pCtx->blockContexts.size(); blockIndex--;) { - VmaBlockDefragmentationContext &blockCtx = pCtx->blockContexts[blockIndex]; - if(blockCtx.hBuffer) + VmaBlockDefragmentationContext& blockCtx = pCtx->blockContexts[blockIndex]; + if (blockCtx.hBuffer) { (*m_hAllocator->GetVulkanFunctions().vkDestroyBuffer)(m_hAllocator->m_hDevice, blockCtx.hBuffer, m_hAllocator->GetAllocationCallbacks()); } } - if(pCtx->res >= VK_SUCCESS) + if (pCtx->res >= VK_SUCCESS) { FreeEmptyBlocks(pStats); } } - if(pCtx->mutexLocked) + if (pCtx->mutexLocked) { VMA_ASSERT(m_hAllocator->m_UseMutex); m_Mutex.UnlockWrite(); @@ -14205,14 +13118,14 @@ void VmaBlockVector::DefragmentationEnd( } uint32_t VmaBlockVector::ProcessDefragmentations( - class VmaBlockVectorDefragmentationContext *pCtx, + class VmaBlockVectorDefragmentationContext* pCtx, VmaDefragmentationPassMoveInfo* pMove, uint32_t maxMoves) { VmaMutexLockWrite lock(m_Mutex, m_hAllocator->m_UseMutex); const uint32_t moveCount = VMA_MIN(uint32_t(pCtx->defragmentationMoves.size()) - pCtx->defragmentationMovesProcessed, maxMoves); - for(uint32_t i = 0; i < moveCount; ++ i) + for (uint32_t i = 0; i < moveCount; ++i) { VmaDefragmentationMove& move = pCtx->defragmentationMoves[pCtx->defragmentationMovesProcessed + i]; @@ -14220,7 +13133,7 @@ uint32_t VmaBlockVector::ProcessDefragmentations( pMove->memory = move.pDstBlock->GetDeviceMemory(); pMove->offset = move.dstOffset; - ++ pMove; + ++pMove; } pCtx->defragmentationMovesProcessed += moveCount; @@ -14229,17 +13142,17 @@ uint32_t VmaBlockVector::ProcessDefragmentations( } void VmaBlockVector::CommitDefragmentations( - class VmaBlockVectorDefragmentationContext *pCtx, + class VmaBlockVectorDefragmentationContext* pCtx, VmaDefragmentationStats* pStats) { VmaMutexLockWrite lock(m_Mutex, m_hAllocator->m_UseMutex); - for(uint32_t i = pCtx->defragmentationMovesCommitted; i < pCtx->defragmentationMovesProcessed; ++ i) + for (uint32_t i = pCtx->defragmentationMovesCommitted; i < pCtx->defragmentationMovesProcessed; ++i) { - const VmaDefragmentationMove &move = pCtx->defragmentationMoves[i]; + const VmaDefragmentationMove& move = pCtx->defragmentationMoves[i]; - move.pSrcBlock->m_pMetadata->FreeAtOffset(move.srcOffset); - move.hAllocation->ChangeBlockAllocation(m_hAllocator, move.pDstBlock, move.dstOffset); + move.pSrcBlock->m_pMetadata->Free(move.hAllocation->GetAllocHandle()); + move.hAllocation->ChangeBlockAllocation(m_hAllocator, move.pDstBlock, move.dstHandle); } pCtx->defragmentationMovesCommitted = pCtx->defragmentationMovesProcessed; @@ -14249,7 +13162,7 @@ void VmaBlockVector::CommitDefragmentations( size_t VmaBlockVector::CalcAllocationCount() const { size_t result = 0; - for(size_t i = 0; i < m_Blocks.size(); ++i) + for (size_t i = 0; i < m_Blocks.size(); ++i) { result += m_Blocks[i]->m_pMetadata->GetAllocationCount(); } @@ -14258,17 +13171,17 @@ size_t VmaBlockVector::CalcAllocationCount() const bool VmaBlockVector::IsBufferImageGranularityConflictPossible() const { - if(m_BufferImageGranularity == 1) + if (m_BufferImageGranularity == 1) { return false; } VmaSuballocationType lastSuballocType = VMA_SUBALLOCATION_TYPE_FREE; - for(size_t i = 0, count = m_Blocks.size(); i < count; ++i) + for (size_t i = 0, count = m_Blocks.size(); i < count; ++i) { VmaDeviceMemoryBlock* const pBlock = m_Blocks[i]; VMA_ASSERT(m_Algorithm == 0); VmaBlockMetadata_Generic* const pMetadata = (VmaBlockMetadata_Generic*)pBlock->m_pMetadata; - if(pMetadata->IsBufferImageGranularityConflictPossible(m_BufferImageGranularity, lastSuballocType)) + if (pMetadata->IsBufferImageGranularityConflictPossible(m_BufferImageGranularity, lastSuballocType)) { return true; } @@ -14276,38 +13189,20 @@ bool VmaBlockVector::IsBufferImageGranularityConflictPossible() const return false; } -void VmaBlockVector::MakePoolAllocationsLost( - uint32_t currentFrameIndex, - size_t* pLostAllocationCount) -{ - VmaMutexLockWrite lock(m_Mutex, m_hAllocator->m_UseMutex); - size_t lostAllocationCount = 0; - for(uint32_t blockIndex = 0; blockIndex < m_Blocks.size(); ++blockIndex) - { - VmaDeviceMemoryBlock* const pBlock = m_Blocks[blockIndex]; - VMA_ASSERT(pBlock); - lostAllocationCount += pBlock->m_pMetadata->MakeAllocationsLost(currentFrameIndex, m_FrameInUseCount); - } - if(pLostAllocationCount != VMA_NULL) - { - *pLostAllocationCount = lostAllocationCount; - } -} - VkResult VmaBlockVector::CheckCorruption() { - if(!IsCorruptionDetectionEnabled()) + if (!IsCorruptionDetectionEnabled()) { return VK_ERROR_FEATURE_NOT_PRESENT; } VmaMutexLockRead lock(m_Mutex, m_hAllocator->m_UseMutex); - for(uint32_t blockIndex = 0; blockIndex < m_Blocks.size(); ++blockIndex) + for (uint32_t blockIndex = 0; blockIndex < m_Blocks.size(); ++blockIndex) { VmaDeviceMemoryBlock* const pBlock = m_Blocks[blockIndex]; VMA_ASSERT(pBlock); VkResult res = pBlock->CheckCorruption(m_hAllocator); - if(res != VK_SUCCESS) + if (res != VK_SUCCESS) { return res; } @@ -14322,7 +13217,7 @@ void VmaBlockVector::AddStats(VmaStats* pStats) VmaMutexLockRead lock(m_Mutex, m_hAllocator->m_UseMutex); - for(uint32_t blockIndex = 0; blockIndex < m_Blocks.size(); ++blockIndex) + for (uint32_t blockIndex = 0; blockIndex < m_Blocks.size(); ++blockIndex) { const VmaDeviceMemoryBlock* const pBlock = m_Blocks[blockIndex]; VMA_ASSERT(pBlock); @@ -14334,16 +13229,14 @@ void VmaBlockVector::AddStats(VmaStats* pStats) VmaAddStatInfo(pStats->memoryHeap[memHeapIndex], allocationStatInfo); } } +#endif // _VMA_BLOCK_VECTOR_FUNCTIONS -//////////////////////////////////////////////////////////////////////////////// -// VmaDefragmentationAlgorithm_Generic members definition - +#ifndef _VMA_DEFRAGMENTATION_ALGORITHM_GENERIC_FUNCTIONS VmaDefragmentationAlgorithm_Generic::VmaDefragmentationAlgorithm_Generic( VmaAllocator hAllocator, VmaBlockVector* pBlockVector, - uint32_t currentFrameIndex, - bool overlappingMoveSupported) : - VmaDefragmentationAlgorithm(hAllocator, pBlockVector, currentFrameIndex), + bool overlappingMoveSupported) + : VmaDefragmentationAlgorithm(hAllocator, pBlockVector), m_AllocationCount(0), m_AllAllocations(false), m_BytesMoved(0), @@ -14352,7 +13245,7 @@ VmaDefragmentationAlgorithm_Generic::VmaDefragmentationAlgorithm_Generic( { // Create block info for each block. const size_t blockCount = m_pBlockVector->m_Blocks.size(); - for(size_t blockIndex = 0; blockIndex < blockCount; ++blockIndex) + for (size_t blockIndex = 0; blockIndex < blockCount; ++blockIndex) { BlockInfo* pBlockInfo = vma_new(m_hAllocator, BlockInfo)(m_hAllocator->GetAllocationCallbacks()); pBlockInfo->m_OriginalBlockIndex = blockIndex; @@ -14366,7 +13259,7 @@ VmaDefragmentationAlgorithm_Generic::VmaDefragmentationAlgorithm_Generic( VmaDefragmentationAlgorithm_Generic::~VmaDefragmentationAlgorithm_Generic() { - for(size_t i = m_Blocks.size(); i--; ) + for (size_t i = m_Blocks.size(); i--; ) { vma_delete(m_hAllocator, m_Blocks[i]); } @@ -14374,23 +13267,19 @@ VmaDefragmentationAlgorithm_Generic::~VmaDefragmentationAlgorithm_Generic() void VmaDefragmentationAlgorithm_Generic::AddAllocation(VmaAllocation hAlloc, VkBool32* pChanged) { - // Now as we are inside VmaBlockVector::m_Mutex, we can make final check if this allocation was not lost. - if(hAlloc->GetLastUseFrameIndex() != VMA_FRAME_INDEX_LOST) + VmaDeviceMemoryBlock* pBlock = hAlloc->GetBlock(); + BlockInfoVector::iterator it = VmaBinaryFindFirstNotLess(m_Blocks.begin(), m_Blocks.end(), pBlock, BlockPointerLess()); + if (it != m_Blocks.end() && (*it)->m_pBlock == pBlock) { - VmaDeviceMemoryBlock* pBlock = hAlloc->GetBlock(); - BlockInfoVector::iterator it = VmaBinaryFindFirstNotLess(m_Blocks.begin(), m_Blocks.end(), pBlock, BlockPointerLess()); - if(it != m_Blocks.end() && (*it)->m_pBlock == pBlock) - { - AllocationInfo allocInfo = AllocationInfo(hAlloc, pChanged); - (*it)->m_Allocations.push_back(allocInfo); - } - else - { - VMA_ASSERT(0); - } - - ++m_AllocationCount; + AllocationInfo allocInfo = AllocationInfo(hAlloc, pChanged); + (*it)->m_Allocations.push_back(allocInfo); } + else + { + VMA_ASSERT(0); + } + + ++m_AllocationCount; } VkResult VmaDefragmentationAlgorithm_Generic::DefragmentRound( @@ -14399,7 +13288,7 @@ VkResult VmaDefragmentationAlgorithm_Generic::DefragmentRound( uint32_t maxAllocationsToMove, bool freeOldAllocations) { - if(m_Blocks.empty()) + if (m_Blocks.empty()) { return VK_SUCCESS; } @@ -14409,8 +13298,6 @@ VkResult VmaDefragmentationAlgorithm_Generic::DefragmentRound( uint32_t strategy = VMA_ALLOCATION_CREATE_STRATEGY_MIN_TIME_BIT; // Option 2: //uint32_t strategy = VMA_ALLOCATION_CREATE_STRATEGY_MIN_MEMORY_BIT; - // Option 3: - //uint32_t strategy = VMA_ALLOCATION_CREATE_STRATEGY_MIN_FRAGMENTATION_BIT; size_t srcBlockMinIndex = 0; // When FAST_ALGORITHM, move allocations from only last out of blocks that contain non-movable allocations. @@ -14427,17 +13314,17 @@ VkResult VmaDefragmentationAlgorithm_Generic::DefragmentRound( size_t srcBlockIndex = m_Blocks.size() - 1; size_t srcAllocIndex = SIZE_MAX; - for(;;) + for (;;) { // 1. Find next allocation to move. // 1.1. Start from last to first m_Blocks - they are sorted from most "destination" to most "source". // 1.2. Then start from last to first m_Allocations. - while(srcAllocIndex >= m_Blocks[srcBlockIndex]->m_Allocations.size()) + while (srcAllocIndex >= m_Blocks[srcBlockIndex]->m_Allocations.size()) { - if(m_Blocks[srcBlockIndex]->m_Allocations.empty()) + if (m_Blocks[srcBlockIndex]->m_Allocations.empty()) { // Finished: no more allocations to process. - if(srcBlockIndex == srcBlockMinIndex) + if (srcBlockIndex == srcBlockMinIndex) { return VK_SUCCESS; } @@ -14462,28 +13349,23 @@ VkResult VmaDefragmentationAlgorithm_Generic::DefragmentRound( const VmaSuballocationType suballocType = allocInfo.m_hAllocation->GetSuballocationType(); // 2. Try to find new place for this allocation in preceding or current block. - for(size_t dstBlockIndex = 0; dstBlockIndex <= srcBlockIndex; ++dstBlockIndex) + for (size_t dstBlockIndex = 0; dstBlockIndex <= srcBlockIndex; ++dstBlockIndex) { BlockInfo* pDstBlockInfo = m_Blocks[dstBlockIndex]; + VmaBlockMetadata* pMetadata = pDstBlockInfo->m_pBlock->m_pMetadata; VmaAllocationRequest dstAllocRequest; - if(pDstBlockInfo->m_pBlock->m_pMetadata->CreateAllocationRequest( - m_CurrentFrameIndex, - m_pBlockVector->GetFrameInUseCount(), - m_pBlockVector->GetBufferImageGranularity(), + if (pMetadata->CreateAllocationRequest( size, alignment, false, // upperAddress suballocType, - false, // canMakeOtherLost strategy, &dstAllocRequest) && - MoveMakesSense( - dstBlockIndex, dstAllocRequest.offset, srcBlockIndex, srcOffset)) + MoveMakesSense( + dstBlockIndex, pMetadata->GetAllocationOffset(dstAllocRequest.allocHandle), srcBlockIndex, srcOffset)) { - VMA_ASSERT(dstAllocRequest.itemsToMakeLostCount == 0); - // Reached limit on number of allocations or bytes to move. - if((m_AllocationsMoved + 1 > maxAllocationsToMove) || + if ((m_AllocationsMoved + 1 > maxAllocationsToMove) || (m_BytesMoved + size > maxBytesToMove)) { return VK_SUCCESS; @@ -14493,27 +13375,24 @@ VkResult VmaDefragmentationAlgorithm_Generic::DefragmentRound( move.srcBlockIndex = pSrcBlockInfo->m_OriginalBlockIndex; move.dstBlockIndex = pDstBlockInfo->m_OriginalBlockIndex; move.srcOffset = srcOffset; - move.dstOffset = dstAllocRequest.offset; + move.dstOffset = pMetadata->GetAllocationOffset(dstAllocRequest.allocHandle); move.size = size; move.hAllocation = allocInfo.m_hAllocation; move.pSrcBlock = pSrcBlockInfo->m_pBlock; move.pDstBlock = pDstBlockInfo->m_pBlock; + move.dstHandle = dstAllocRequest.allocHandle; moves.push_back(move); - pDstBlockInfo->m_pBlock->m_pMetadata->Alloc( - dstAllocRequest, - suballocType, - size, - allocInfo.m_hAllocation); + pDstBlockInfo->m_pBlock->m_pMetadata->Alloc(dstAllocRequest, suballocType, allocInfo.m_hAllocation); - if(freeOldAllocations) + if (freeOldAllocations) { - pSrcBlockInfo->m_pBlock->m_pMetadata->FreeAtOffset(srcOffset); - allocInfo.m_hAllocation->ChangeBlockAllocation(m_hAllocator, pDstBlockInfo->m_pBlock, dstAllocRequest.offset); + pSrcBlockInfo->m_pBlock->m_pMetadata->Free(allocInfo.m_hAllocation->GetAllocHandle()); + allocInfo.m_hAllocation->ChangeBlockAllocation(m_hAllocator, pDstBlockInfo->m_pBlock, dstAllocRequest.allocHandle); } - if(allocInfo.m_pChanged != VMA_NULL) + if (allocInfo.m_pChanged != VMA_NULL) { *allocInfo.m_pChanged = VK_TRUE; } @@ -14529,13 +13408,13 @@ VkResult VmaDefragmentationAlgorithm_Generic::DefragmentRound( // If not processed, this allocInfo remains in pBlockInfo->m_Allocations for next round. - if(srcAllocIndex > 0) + if (srcAllocIndex > 0) { --srcAllocIndex; } else { - if(srcBlockIndex > 0) + if (srcBlockIndex > 0) { --srcBlockIndex; srcAllocIndex = SIZE_MAX; @@ -14548,12 +13427,90 @@ VkResult VmaDefragmentationAlgorithm_Generic::DefragmentRound( } } +bool VmaDefragmentationAlgorithm_Generic::AllocationInfoSizeGreater::operator()(const AllocationInfo& lhs, const AllocationInfo& rhs) const +{ + return lhs.m_hAllocation->GetSize() > rhs.m_hAllocation->GetSize(); +} + +bool VmaDefragmentationAlgorithm_Generic::AllocationInfoOffsetGreater::operator()(const AllocationInfo& lhs, const AllocationInfo& rhs) const +{ + return lhs.m_hAllocation->GetOffset() > rhs.m_hAllocation->GetOffset(); +} + +VmaDefragmentationAlgorithm_Generic::BlockInfo::BlockInfo(const VkAllocationCallbacks* pAllocationCallbacks) + : m_OriginalBlockIndex(SIZE_MAX), + m_pBlock(VMA_NULL), + m_HasNonMovableAllocations(true), + m_Allocations(pAllocationCallbacks) {} + +void VmaDefragmentationAlgorithm_Generic::BlockInfo::CalcHasNonMovableAllocations() +{ + const size_t blockAllocCount = m_pBlock->m_pMetadata->GetAllocationCount(); + const size_t defragmentAllocCount = m_Allocations.size(); + m_HasNonMovableAllocations = blockAllocCount != defragmentAllocCount; +} + +void VmaDefragmentationAlgorithm_Generic::BlockInfo::SortAllocationsBySizeDescending() +{ + VMA_SORT(m_Allocations.begin(), m_Allocations.end(), AllocationInfoSizeGreater()); +} + +void VmaDefragmentationAlgorithm_Generic::BlockInfo::SortAllocationsByOffsetDescending() +{ + VMA_SORT(m_Allocations.begin(), m_Allocations.end(), AllocationInfoOffsetGreater()); +} + +bool VmaDefragmentationAlgorithm_Generic::BlockPointerLess::operator()(const BlockInfo* pLhsBlockInfo, const VmaDeviceMemoryBlock* pRhsBlock) const +{ + return pLhsBlockInfo->m_pBlock < pRhsBlock; +} +bool VmaDefragmentationAlgorithm_Generic::BlockPointerLess::operator()(const BlockInfo* pLhsBlockInfo, const BlockInfo* pRhsBlockInfo) const +{ + return pLhsBlockInfo->m_pBlock < pRhsBlockInfo->m_pBlock; +} + +bool VmaDefragmentationAlgorithm_Generic::BlockInfoCompareMoveDestination::operator()(const BlockInfo* pLhsBlockInfo, const BlockInfo* pRhsBlockInfo) const +{ + if (pLhsBlockInfo->m_HasNonMovableAllocations && !pRhsBlockInfo->m_HasNonMovableAllocations) + { + return true; + } + if (!pLhsBlockInfo->m_HasNonMovableAllocations && pRhsBlockInfo->m_HasNonMovableAllocations) + { + return false; + } + if (pLhsBlockInfo->m_pBlock->m_pMetadata->GetSumFreeSize() < pRhsBlockInfo->m_pBlock->m_pMetadata->GetSumFreeSize()) + { + return true; + } + return false; +} + +bool VmaDefragmentationAlgorithm_Generic::MoveMakesSense( + size_t dstBlockIndex, VkDeviceSize dstOffset, + size_t srcBlockIndex, VkDeviceSize srcOffset) +{ + if (dstBlockIndex < srcBlockIndex) + { + return true; + } + if (dstBlockIndex > srcBlockIndex) + { + return false; + } + if (dstOffset < srcOffset) + { + return true; + } + return false; +} + size_t VmaDefragmentationAlgorithm_Generic::CalcBlocksWithNonMovableCount() const { size_t result = 0; - for(size_t i = 0; i < m_Blocks.size(); ++i) + for (size_t i = 0; i < m_Blocks.size(); ++i) { - if(m_Blocks[i]->m_HasNonMovableAllocations) + if (m_Blocks[i]->m_HasNonMovableAllocations) { ++result; } @@ -14567,26 +13524,27 @@ VkResult VmaDefragmentationAlgorithm_Generic::Defragment( uint32_t maxAllocationsToMove, VmaDefragmentationFlags flags) { - if(!m_AllAllocations && m_AllocationCount == 0) + if (!m_AllAllocations && m_AllocationCount == 0) { return VK_SUCCESS; } const size_t blockCount = m_Blocks.size(); - for(size_t blockIndex = 0; blockIndex < blockCount; ++blockIndex) + for (size_t blockIndex = 0; blockIndex < blockCount; ++blockIndex) { BlockInfo* pBlockInfo = m_Blocks[blockIndex]; - if(m_AllAllocations) + if (m_AllAllocations) { VmaBlockMetadata_Generic* pMetadata = (VmaBlockMetadata_Generic*)pBlockInfo->m_pBlock->m_pMetadata; - for(VmaSuballocationList::const_iterator it = pMetadata->m_Suballocations.begin(); + VMA_ASSERT(!pMetadata->IsVirtual()); + for (VmaSuballocationList::const_iterator it = pMetadata->m_Suballocations.begin(); it != pMetadata->m_Suballocations.end(); ++it) { - if(it->type != VMA_SUBALLOCATION_TYPE_FREE) + if (it->type != VMA_SUBALLOCATION_TYPE_FREE) { - AllocationInfo allocInfo = AllocationInfo(it->hAllocation, VMA_NULL); + AllocationInfo allocInfo = AllocationInfo((VmaAllocation)it->userData, VMA_NULL); pBlockInfo->m_Allocations.push_back(allocInfo); } } @@ -14609,42 +13567,21 @@ VkResult VmaDefragmentationAlgorithm_Generic::Defragment( // Execute defragmentation rounds (the main part). VkResult result = VK_SUCCESS; - for(uint32_t round = 0; (round < roundCount) && (result == VK_SUCCESS); ++round) + for (uint32_t round = 0; (round < roundCount) && (result == VK_SUCCESS); ++round) { result = DefragmentRound(moves, maxBytesToMove, maxAllocationsToMove, !(flags & VMA_DEFRAGMENTATION_FLAG_INCREMENTAL)); } return result; } +#endif // _VMA_DEFRAGMENTATION_ALGORITHM_GENERIC_FUNCTIONS -bool VmaDefragmentationAlgorithm_Generic::MoveMakesSense( - size_t dstBlockIndex, VkDeviceSize dstOffset, - size_t srcBlockIndex, VkDeviceSize srcOffset) -{ - if(dstBlockIndex < srcBlockIndex) - { - return true; - } - if(dstBlockIndex > srcBlockIndex) - { - return false; - } - if(dstOffset < srcOffset) - { - return true; - } - return false; -} - -//////////////////////////////////////////////////////////////////////////////// -// VmaDefragmentationAlgorithm_Fast - +#ifndef _VMA_DEFRAGMENTATION_ALGORITHM_FAST_FUNCTIONS VmaDefragmentationAlgorithm_Fast::VmaDefragmentationAlgorithm_Fast( VmaAllocator hAllocator, VmaBlockVector* pBlockVector, - uint32_t currentFrameIndex, - bool overlappingMoveSupported) : - VmaDefragmentationAlgorithm(hAllocator, pBlockVector, currentFrameIndex), + bool overlappingMoveSupported) + : VmaDefragmentationAlgorithm(hAllocator, pBlockVector), m_OverlappingMoveSupported(overlappingMoveSupported), m_AllocationCount(0), m_AllAllocations(false), @@ -14653,15 +13590,10 @@ VmaDefragmentationAlgorithm_Fast::VmaDefragmentationAlgorithm_Fast( m_BlockInfos(VmaStlAllocator<BlockInfo>(hAllocator->GetAllocationCallbacks())) { VMA_ASSERT(VMA_DEBUG_MARGIN == 0); - -} - -VmaDefragmentationAlgorithm_Fast::~VmaDefragmentationAlgorithm_Fast() -{ } VkResult VmaDefragmentationAlgorithm_Fast::Defragment( - VmaVector< VmaDefragmentationMove, VmaStlAllocator<VmaDefragmentationMove> >& moves, + VmaVector<VmaDefragmentationMove, VmaStlAllocator<VmaDefragmentationMove>>& moves, VkDeviceSize maxBytesToMove, uint32_t maxAllocationsToMove, VmaDefragmentationFlags flags) @@ -14669,7 +13601,7 @@ VkResult VmaDefragmentationAlgorithm_Fast::Defragment( VMA_ASSERT(m_AllAllocations || m_pBlockVector->CalcAllocationCount() == m_AllocationCount); const size_t blockCount = m_pBlockVector->GetBlockCount(); - if(blockCount == 0 || maxBytesToMove == 0 || maxAllocationsToMove == 0) + if (blockCount == 0 || maxBytesToMove == 0 || maxAllocationsToMove == 0) { return VK_SUCCESS; } @@ -14679,7 +13611,7 @@ VkResult VmaDefragmentationAlgorithm_Fast::Defragment( // Sort blocks in order from most destination. m_BlockInfos.resize(blockCount); - for(size_t i = 0; i < blockCount; ++i) + for (size_t i = 0; i < blockCount; ++i) { m_BlockInfos[i].origBlockIndex = i; } @@ -14687,7 +13619,7 @@ VkResult VmaDefragmentationAlgorithm_Fast::Defragment( VMA_SORT(m_BlockInfos.begin(), m_BlockInfos.end(), [this](const BlockInfo& lhs, const BlockInfo& rhs) -> bool { return m_pBlockVector->GetBlock(lhs.origBlockIndex)->m_pMetadata->GetSumFreeSize() < m_pBlockVector->GetBlock(rhs.origBlockIndex)->m_pMetadata->GetSumFreeSize(); - }); + }); // THE MAIN ALGORITHM @@ -14701,18 +13633,18 @@ VkResult VmaDefragmentationAlgorithm_Fast::Defragment( VkDeviceSize dstOffset = 0; bool end = false; - for(size_t srcBlockInfoIndex = 0; !end && srcBlockInfoIndex < blockCount; ++srcBlockInfoIndex) + for (size_t srcBlockInfoIndex = 0; !end && srcBlockInfoIndex < blockCount; ++srcBlockInfoIndex) { const size_t srcOrigBlockIndex = m_BlockInfos[srcBlockInfoIndex].origBlockIndex; VmaDeviceMemoryBlock* const pSrcBlock = m_pBlockVector->GetBlock(srcOrigBlockIndex); VmaBlockMetadata_Generic* const pSrcMetadata = (VmaBlockMetadata_Generic*)pSrcBlock->m_pMetadata; - for(VmaSuballocationList::iterator srcSuballocIt = pSrcMetadata->m_Suballocations.begin(); + for (VmaSuballocationList::iterator srcSuballocIt = pSrcMetadata->m_Suballocations.begin(); !end && srcSuballocIt != pSrcMetadata->m_Suballocations.end(); ) { - VmaAllocation_T* const pAlloc = srcSuballocIt->hAllocation; + VmaAllocation const pAlloc = (VmaAllocation)srcSuballocIt->userData; const VkDeviceSize srcAllocAlignment = pAlloc->GetAlignment(); const VkDeviceSize srcAllocSize = srcSuballocIt->size; - if(m_AllocationsMoved == maxAllocationsToMove || + if (m_AllocationsMoved == maxAllocationsToMove || m_BytesMoved + srcAllocSize > maxBytesToMove) { end = true; @@ -14724,7 +13656,7 @@ VkResult VmaDefragmentationAlgorithm_Fast::Defragment( // Try to place it in one of free spaces from the database. size_t freeSpaceInfoIndex; VkDeviceSize dstAllocOffset; - if(freeSpaceDb.Fetch(srcAllocAlignment, srcAllocSize, + if (freeSpaceDb.Fetch(srcAllocAlignment, srcAllocSize, freeSpaceInfoIndex, dstAllocOffset)) { size_t freeSpaceOrigBlockIndex = m_BlockInfos[freeSpaceInfoIndex].origBlockIndex; @@ -14732,7 +13664,7 @@ VkResult VmaDefragmentationAlgorithm_Fast::Defragment( VmaBlockMetadata_Generic* pFreeSpaceMetadata = (VmaBlockMetadata_Generic*)pFreeSpaceBlock->m_pMetadata; // Same block - if(freeSpaceInfoIndex == srcBlockInfoIndex) + if (freeSpaceInfoIndex == srcBlockInfoIndex) { VMA_ASSERT(dstAllocOffset <= srcAllocOffset); @@ -14740,7 +13672,7 @@ VkResult VmaDefragmentationAlgorithm_Fast::Defragment( VmaSuballocation suballoc = *srcSuballocIt; suballoc.offset = dstAllocOffset; - suballoc.hAllocation->ChangeOffset(dstAllocOffset); + ((VmaAllocation)(suballoc.userData))->ChangeAllocHandle((VmaAllocHandle)(dstAllocOffset + 1)); m_BytesMoved += srcAllocSize; ++m_AllocationsMoved; @@ -14755,6 +13687,7 @@ VkResult VmaDefragmentationAlgorithm_Fast::Defragment( move.dstBlockIndex = freeSpaceOrigBlockIndex; move.srcOffset = srcAllocOffset; move.dstOffset = dstAllocOffset; + move.dstHandle = (VmaAllocHandle)(dstAllocOffset + 1); move.size = srcAllocSize; moves.push_back(move); @@ -14768,7 +13701,7 @@ VkResult VmaDefragmentationAlgorithm_Fast::Defragment( VmaSuballocation suballoc = *srcSuballocIt; suballoc.offset = dstAllocOffset; - suballoc.hAllocation->ChangeBlockAllocation(m_hAllocator, pFreeSpaceBlock, dstAllocOffset); + ((VmaAllocation)(suballoc.userData))->ChangeBlockAllocation(m_hAllocator, pFreeSpaceBlock, (VmaAllocHandle)(dstAllocOffset + 1)); m_BytesMoved += srcAllocSize; ++m_AllocationsMoved; @@ -14783,6 +13716,7 @@ VkResult VmaDefragmentationAlgorithm_Fast::Defragment( move.dstBlockIndex = freeSpaceOrigBlockIndex; move.srcOffset = srcAllocOffset; move.dstOffset = dstAllocOffset; + move.dstHandle = (VmaAllocHandle)(dstAllocOffset + 1); move.size = srcAllocSize; moves.push_back(move); @@ -14793,7 +13727,7 @@ VkResult VmaDefragmentationAlgorithm_Fast::Defragment( dstAllocOffset = VmaAlignUp(dstOffset, srcAllocAlignment); // If the allocation doesn't fit before the end of dstBlock, forward to next block. - while(dstBlockInfoIndex < srcBlockInfoIndex && + while (dstBlockInfoIndex < srcBlockInfoIndex && dstAllocOffset + srcAllocSize > dstBlockSize) { // But before that, register remaining free space at the end of dst block. @@ -14809,21 +13743,21 @@ VkResult VmaDefragmentationAlgorithm_Fast::Defragment( } // Same block - if(dstBlockInfoIndex == srcBlockInfoIndex) + if (dstBlockInfoIndex == srcBlockInfoIndex) { VMA_ASSERT(dstAllocOffset <= srcAllocOffset); const bool overlap = dstAllocOffset + srcAllocSize > srcAllocOffset; bool skipOver = overlap; - if(overlap && m_OverlappingMoveSupported && dstAllocOffset < srcAllocOffset) + if (overlap && m_OverlappingMoveSupported && dstAllocOffset < srcAllocOffset) { // If destination and source place overlap, skip if it would move it // by only < 1/64 of its size. skipOver = (srcAllocOffset - dstAllocOffset) * 64 < srcAllocSize; } - if(skipOver) + if (skipOver) { freeSpaceDb.Register(dstBlockInfoIndex, dstOffset, srcAllocOffset - dstOffset); @@ -14834,7 +13768,7 @@ VkResult VmaDefragmentationAlgorithm_Fast::Defragment( else { srcSuballocIt->offset = dstAllocOffset; - srcSuballocIt->hAllocation->ChangeOffset(dstAllocOffset); + ((VmaAllocation)(srcSuballocIt->userData))->ChangeAllocHandle((VmaAllocHandle)(dstAllocOffset + 1)); dstOffset = dstAllocOffset + srcAllocSize; m_BytesMoved += srcAllocSize; ++m_AllocationsMoved; @@ -14844,6 +13778,7 @@ VkResult VmaDefragmentationAlgorithm_Fast::Defragment( move.dstBlockIndex = dstOrigBlockIndex; move.srcOffset = srcAllocOffset; move.dstOffset = dstAllocOffset; + move.dstHandle = (VmaAllocHandle)(dstAllocOffset + 1); move.size = srcAllocSize; moves.push_back(move); @@ -14859,7 +13794,7 @@ VkResult VmaDefragmentationAlgorithm_Fast::Defragment( VmaSuballocation suballoc = *srcSuballocIt; suballoc.offset = dstAllocOffset; - suballoc.hAllocation->ChangeBlockAllocation(m_hAllocator, pDstBlock, dstAllocOffset); + ((VmaAllocation)(suballoc.userData))->ChangeBlockAllocation(m_hAllocator, pDstBlock, (VmaAllocHandle)(dstAllocOffset + 1)); dstOffset = dstAllocOffset + srcAllocSize; m_BytesMoved += srcAllocSize; ++m_AllocationsMoved; @@ -14875,6 +13810,7 @@ VkResult VmaDefragmentationAlgorithm_Fast::Defragment( move.dstBlockIndex = dstOrigBlockIndex; move.srcOffset = srcAllocOffset; move.dstOffset = dstAllocOffset; + move.dstHandle = (VmaAllocHandle)(dstAllocOffset + 1); move.size = srcAllocSize; moves.push_back(move); @@ -14890,20 +13826,98 @@ VkResult VmaDefragmentationAlgorithm_Fast::Defragment( return VK_SUCCESS; } +VmaDefragmentationAlgorithm_Fast::FreeSpaceDatabase::FreeSpaceDatabase() +{ + FreeSpace s = {}; + s.blockInfoIndex = SIZE_MAX; + for (size_t i = 0; i < MAX_COUNT; ++i) + { + m_FreeSpaces[i] = s; + } +} + +void VmaDefragmentationAlgorithm_Fast::FreeSpaceDatabase::Register(size_t blockInfoIndex, VkDeviceSize offset, VkDeviceSize size) +{ + // Find first invalid or the smallest structure. + size_t bestIndex = SIZE_MAX; + for (size_t i = 0; i < MAX_COUNT; ++i) + { + // Empty structure. + if (m_FreeSpaces[i].blockInfoIndex == SIZE_MAX) + { + bestIndex = i; + break; + } + if (m_FreeSpaces[i].size < size && + (bestIndex == SIZE_MAX || m_FreeSpaces[bestIndex].size > m_FreeSpaces[i].size)) + { + bestIndex = i; + } + } + + if (bestIndex != SIZE_MAX) + { + m_FreeSpaces[bestIndex].blockInfoIndex = blockInfoIndex; + m_FreeSpaces[bestIndex].offset = offset; + m_FreeSpaces[bestIndex].size = size; + } +} + +bool VmaDefragmentationAlgorithm_Fast::FreeSpaceDatabase::Fetch(VkDeviceSize alignment, VkDeviceSize size, + size_t& outBlockInfoIndex, VkDeviceSize& outDstOffset) +{ + size_t bestIndex = SIZE_MAX; + VkDeviceSize bestFreeSpaceAfter = 0; + for (size_t i = 0; i < MAX_COUNT; ++i) + { + // Structure is valid. + if (m_FreeSpaces[i].blockInfoIndex != SIZE_MAX) + { + const VkDeviceSize dstOffset = VmaAlignUp(m_FreeSpaces[i].offset, alignment); + // Allocation fits into this structure. + if (dstOffset + size <= m_FreeSpaces[i].offset + m_FreeSpaces[i].size) + { + const VkDeviceSize freeSpaceAfter = (m_FreeSpaces[i].offset + m_FreeSpaces[i].size) - + (dstOffset + size); + if (bestIndex == SIZE_MAX || freeSpaceAfter > bestFreeSpaceAfter) + { + bestIndex = i; + bestFreeSpaceAfter = freeSpaceAfter; + } + } + } + } + + if (bestIndex != SIZE_MAX) + { + outBlockInfoIndex = m_FreeSpaces[bestIndex].blockInfoIndex; + outDstOffset = VmaAlignUp(m_FreeSpaces[bestIndex].offset, alignment); + + // Leave this structure for remaining empty space. + const VkDeviceSize alignmentPlusSize = (outDstOffset - m_FreeSpaces[bestIndex].offset) + size; + m_FreeSpaces[bestIndex].offset += alignmentPlusSize; + m_FreeSpaces[bestIndex].size -= alignmentPlusSize; + + return true; + } + + return false; +} + void VmaDefragmentationAlgorithm_Fast::PreprocessMetadata() { const size_t blockCount = m_pBlockVector->GetBlockCount(); - for(size_t blockIndex = 0; blockIndex < blockCount; ++blockIndex) + for (size_t blockIndex = 0; blockIndex < blockCount; ++blockIndex) { VmaBlockMetadata_Generic* const pMetadata = (VmaBlockMetadata_Generic*)m_pBlockVector->GetBlock(blockIndex)->m_pMetadata; pMetadata->m_FreeCount = 0; pMetadata->m_SumFreeSize = pMetadata->GetSize(); pMetadata->m_FreeSuballocationsBySize.clear(); - for(VmaSuballocationList::iterator it = pMetadata->m_Suballocations.begin(); + for (VmaSuballocationList::iterator it = pMetadata->m_Suballocations.begin(); it != pMetadata->m_Suballocations.end(); ) { - if(it->type == VMA_SUBALLOCATION_TYPE_FREE) + if (it->type == VMA_SUBALLOCATION_TYPE_FREE) { VmaSuballocationList::iterator nextIt = it; ++nextIt; @@ -14921,14 +13935,14 @@ void VmaDefragmentationAlgorithm_Fast::PreprocessMetadata() void VmaDefragmentationAlgorithm_Fast::PostprocessMetadata() { const size_t blockCount = m_pBlockVector->GetBlockCount(); - for(size_t blockIndex = 0; blockIndex < blockCount; ++blockIndex) + for (size_t blockIndex = 0; blockIndex < blockCount; ++blockIndex) { VmaBlockMetadata_Generic* const pMetadata = (VmaBlockMetadata_Generic*)m_pBlockVector->GetBlock(blockIndex)->m_pMetadata; const VkDeviceSize blockSize = pMetadata->GetSize(); // No allocations in this block - entire area is free. - if(pMetadata->m_Suballocations.empty()) + if (pMetadata->m_Suballocations.empty()) { pMetadata->m_FreeCount = 1; //pMetadata->m_SumFreeSize is already set to blockSize. @@ -14945,7 +13959,7 @@ void VmaDefragmentationAlgorithm_Fast::PostprocessMetadata() { VkDeviceSize offset = 0; VmaSuballocationList::iterator it; - for(it = pMetadata->m_Suballocations.begin(); + for (it = pMetadata->m_Suballocations.begin(); it != pMetadata->m_Suballocations.end(); ++it) { @@ -14953,7 +13967,7 @@ void VmaDefragmentationAlgorithm_Fast::PostprocessMetadata() VMA_ASSERT(it->offset >= offset); // Need to insert preceding free space. - if(it->offset > offset) + if (it->offset > offset) { ++pMetadata->m_FreeCount; const VkDeviceSize freeSize = it->offset - offset; @@ -14963,10 +13977,7 @@ void VmaDefragmentationAlgorithm_Fast::PostprocessMetadata() VMA_NULL, // hAllocation VMA_SUBALLOCATION_TYPE_FREE }; VmaSuballocationList::iterator precedingFreeIt = pMetadata->m_Suballocations.insert(it, suballoc); - if(freeSize >= VMA_MIN_FREE_SUBALLOCATION_SIZE_TO_REGISTER) - { - pMetadata->m_FreeSuballocationsBySize.push_back(precedingFreeIt); - } + pMetadata->m_FreeSuballocationsBySize.push_back(precedingFreeIt); } pMetadata->m_SumFreeSize -= it->size; @@ -14974,7 +13985,7 @@ void VmaDefragmentationAlgorithm_Fast::PostprocessMetadata() } // Need to insert trailing free space. - if(offset < blockSize) + if (offset < blockSize) { ++pMetadata->m_FreeCount; const VkDeviceSize freeSize = blockSize - offset; @@ -14985,10 +13996,7 @@ void VmaDefragmentationAlgorithm_Fast::PostprocessMetadata() VMA_SUBALLOCATION_TYPE_FREE }; VMA_ASSERT(it == pMetadata->m_Suballocations.end()); VmaSuballocationList::iterator trailingFreeIt = pMetadata->m_Suballocations.insert(it, suballoc); - if(freeSize > VMA_MIN_FREE_SUBALLOCATION_SIZE_TO_REGISTER) - { - pMetadata->m_FreeSuballocationsBySize.push_back(trailingFreeIt); - } + pMetadata->m_FreeSuballocationsBySize.push_back(trailingFreeIt); } VMA_SORT( @@ -15003,27 +14011,59 @@ void VmaDefragmentationAlgorithm_Fast::PostprocessMetadata() void VmaDefragmentationAlgorithm_Fast::InsertSuballoc(VmaBlockMetadata_Generic* pMetadata, const VmaSuballocation& suballoc) { - // TODO: Optimize somehow. Remember iterator instead of searching for it linearly. - VmaSuballocationList::iterator it = pMetadata->m_Suballocations.begin(); - while(it != pMetadata->m_Suballocations.end()) + VmaSuballocationList& suballocs = pMetadata->m_Suballocations; + VmaSuballocationList::iterator elementAfter; + const VkDeviceSize last = suballocs.rbegin()->offset; + const VkDeviceSize first = suballocs.begin()->offset; + + if (last <= suballoc.offset) + elementAfter = suballocs.end(); + else if (first >= suballoc.offset) + elementAfter = suballocs.begin(); + else { - if(it->offset < suballoc.offset) + const size_t suballocCount = suballocs.size(); + const VkDeviceSize step = (last - first + suballocs.begin()->size) / suballocCount; + // If offset to be inserted is closer to the end of range, search from the end + if ((suballoc.offset - first) / step > suballocCount / 2) { - ++it; + elementAfter = suballocs.begin(); + for (VmaSuballocationList::reverse_iterator suballocItem = ++suballocs.rbegin(); + suballocItem != suballocs.rend(); + ++suballocItem) + { + if (suballocItem->offset <= suballoc.offset) + { + elementAfter = --suballocItem; + break; + } + } + } + else + { + elementAfter = suballocs.end(); + for (VmaSuballocationList::iterator suballocItem = ++suballocs.begin(); + suballocItem != suballocs.end(); + ++suballocItem) + { + if (suballocItem->offset >= suballoc.offset) + { + elementAfter = suballocItem; + break; + } + } } } - pMetadata->m_Suballocations.insert(it, suballoc); + pMetadata->m_Suballocations.insert(elementAfter, suballoc); } +#endif // _VMA_DEFRAGMENTATION_ALGORITHM_FAST_FUNCTIONS -//////////////////////////////////////////////////////////////////////////////// -// VmaBlockVectorDefragmentationContext - +#ifndef _VMA_BLOCK_VECTOR_DEFRAGMENTATION_CONTEXT_FUNCTIONS VmaBlockVectorDefragmentationContext::VmaBlockVectorDefragmentationContext( VmaAllocator hAllocator, VmaPool hCustomPool, - VmaBlockVector* pBlockVector, - uint32_t currFrameIndex) : - res(VK_SUCCESS), + VmaBlockVector* pBlockVector) + : res(VK_SUCCESS), mutexLocked(false), blockContexts(VmaStlAllocator<VmaBlockDefragmentationContext>(hAllocator->GetAllocationCallbacks())), defragmentationMoves(VmaStlAllocator<VmaDefragmentationMove>(hAllocator->GetAllocationCallbacks())), @@ -15033,12 +14073,9 @@ VmaBlockVectorDefragmentationContext::VmaBlockVectorDefragmentationContext( m_hAllocator(hAllocator), m_hCustomPool(hCustomPool), m_pBlockVector(pBlockVector), - m_CurrFrameIndex(currFrameIndex), m_pAlgorithm(VMA_NULL), m_Allocations(VmaStlAllocator<AllocInfo>(hAllocator->GetAllocationCallbacks())), - m_AllAllocations(false) -{ -} + m_AllAllocations(false) {} VmaBlockVectorDefragmentationContext::~VmaBlockVectorDefragmentationContext() { @@ -15063,47 +14100,44 @@ void VmaBlockVectorDefragmentationContext::Begin(bool overlappingMoveSupported, /* Fast algorithm is supported only when certain criteria are met: - VMA_DEBUG_MARGIN is 0. - - All allocations in this block vector are moveable. + - All allocations in this block vector are movable. - There is no possibility of image/buffer granularity conflict. - The defragmentation is not incremental */ - if(VMA_DEBUG_MARGIN == 0 && + if (VMA_DEBUG_MARGIN == 0 && allAllocations && !m_pBlockVector->IsBufferImageGranularityConflictPossible() && !(flags & VMA_DEFRAGMENTATION_FLAG_INCREMENTAL)) { m_pAlgorithm = vma_new(m_hAllocator, VmaDefragmentationAlgorithm_Fast)( - m_hAllocator, m_pBlockVector, m_CurrFrameIndex, overlappingMoveSupported); + m_hAllocator, m_pBlockVector, overlappingMoveSupported); } else { m_pAlgorithm = vma_new(m_hAllocator, VmaDefragmentationAlgorithm_Generic)( - m_hAllocator, m_pBlockVector, m_CurrFrameIndex, overlappingMoveSupported); + m_hAllocator, m_pBlockVector, overlappingMoveSupported); } - if(allAllocations) + if (allAllocations) { m_pAlgorithm->AddAll(); } else { - for(size_t i = 0, count = m_Allocations.size(); i < count; ++i) + for (size_t i = 0, count = m_Allocations.size(); i < count; ++i) { m_pAlgorithm->AddAllocation(m_Allocations[i].hAlloc, m_Allocations[i].pChanged); } } } +#endif // _VMA_BLOCK_VECTOR_DEFRAGMENTATION_CONTEXT_FUNCTIONS -//////////////////////////////////////////////////////////////////////////////// -// VmaDefragmentationContext - +#ifndef _VMA_DEFRAGMENTATION_CONTEXT_FUNCTIONS VmaDefragmentationContext_T::VmaDefragmentationContext_T( VmaAllocator hAllocator, - uint32_t currFrameIndex, uint32_t flags, - VmaDefragmentationStats* pStats) : - m_hAllocator(hAllocator), - m_CurrFrameIndex(currFrameIndex), + VmaDefragmentationStats* pStats) + : m_hAllocator(hAllocator), m_Flags(flags), m_pStats(pStats), m_CustomPoolContexts(VmaStlAllocator<VmaBlockVectorDefragmentationContext*>(hAllocator->GetAllocationCallbacks())) @@ -15113,16 +14147,16 @@ VmaDefragmentationContext_T::VmaDefragmentationContext_T( VmaDefragmentationContext_T::~VmaDefragmentationContext_T() { - for(size_t i = m_CustomPoolContexts.size(); i--; ) + for (size_t i = m_CustomPoolContexts.size(); i--; ) { VmaBlockVectorDefragmentationContext* pBlockVectorCtx = m_CustomPoolContexts[i]; pBlockVectorCtx->GetBlockVector()->DefragmentationEnd(pBlockVectorCtx, m_Flags, m_pStats); vma_delete(m_hAllocator, pBlockVectorCtx); } - for(size_t i = m_hAllocator->m_MemProps.memoryTypeCount; i--; ) + for (size_t i = m_hAllocator->m_MemProps.memoryTypeCount; i--; ) { VmaBlockVectorDefragmentationContext* pBlockVectorCtx = m_DefaultPoolContexts[i]; - if(pBlockVectorCtx) + if (pBlockVectorCtx) { pBlockVectorCtx->GetBlockVector()->DefragmentationEnd(pBlockVectorCtx, m_Flags, m_pStats); vma_delete(m_hAllocator, pBlockVectorCtx); @@ -15132,35 +14166,40 @@ VmaDefragmentationContext_T::~VmaDefragmentationContext_T() void VmaDefragmentationContext_T::AddPools(uint32_t poolCount, const VmaPool* pPools) { - for(uint32_t poolIndex = 0; poolIndex < poolCount; ++poolIndex) + for (uint32_t poolIndex = 0; poolIndex < poolCount; ++poolIndex) { VmaPool pool = pPools[poolIndex]; VMA_ASSERT(pool); - // Pools with algorithm other than default are not defragmented. - if(pool->m_BlockVector.GetAlgorithm() == 0) + for(uint32_t memTypeIndex = 0; memTypeIndex < m_hAllocator->GetMemoryTypeCount(); ++memTypeIndex) { - VmaBlockVectorDefragmentationContext* pBlockVectorDefragCtx = VMA_NULL; - - for(size_t i = m_CustomPoolContexts.size(); i--; ) + if(pool->m_pBlockVectors[memTypeIndex]) { - if(m_CustomPoolContexts[i]->GetCustomPool() == pool) + // Pools with algorithm other than default are not defragmented. + if (pool->m_pBlockVectors[memTypeIndex]->GetAlgorithm() == 0) { - pBlockVectorDefragCtx = m_CustomPoolContexts[i]; - break; - } - } + VmaBlockVectorDefragmentationContext* pBlockVectorDefragCtx = VMA_NULL; - if(!pBlockVectorDefragCtx) - { - pBlockVectorDefragCtx = vma_new(m_hAllocator, VmaBlockVectorDefragmentationContext)( - m_hAllocator, - pool, - &pool->m_BlockVector, - m_CurrFrameIndex); - m_CustomPoolContexts.push_back(pBlockVectorDefragCtx); - } + for (size_t i = m_CustomPoolContexts.size(); i--; ) + { + if (m_CustomPoolContexts[i]->GetCustomPool() == pool) + { + pBlockVectorDefragCtx = m_CustomPoolContexts[i]; + break; + } + } + + if (!pBlockVectorDefragCtx) + { + pBlockVectorDefragCtx = vma_new(m_hAllocator, VmaBlockVectorDefragmentationContext)( + m_hAllocator, + pool, + pool->m_pBlockVectors[memTypeIndex]); + m_CustomPoolContexts.push_back(pBlockVectorDefragCtx); + } - pBlockVectorDefragCtx->AddAll(); + pBlockVectorDefragCtx->AddAll(); + } + } } } } @@ -15171,39 +14210,37 @@ void VmaDefragmentationContext_T::AddAllocations( VkBool32* pAllocationsChanged) { // Dispatch pAllocations among defragmentators. Create them when necessary. - for(uint32_t allocIndex = 0; allocIndex < allocationCount; ++allocIndex) + for (uint32_t allocIndex = 0; allocIndex < allocationCount; ++allocIndex) { const VmaAllocation hAlloc = pAllocations[allocIndex]; VMA_ASSERT(hAlloc); + const uint32_t memTypeIndex = hAlloc->GetMemoryTypeIndex(); // DedicatedAlloc cannot be defragmented. - if((hAlloc->GetType() == VmaAllocation_T::ALLOCATION_TYPE_BLOCK) && - // Lost allocation cannot be defragmented. - (hAlloc->GetLastUseFrameIndex() != VMA_FRAME_INDEX_LOST)) + if (hAlloc->GetType() == VmaAllocation_T::ALLOCATION_TYPE_BLOCK) { VmaBlockVectorDefragmentationContext* pBlockVectorDefragCtx = VMA_NULL; const VmaPool hAllocPool = hAlloc->GetBlock()->GetParentPool(); // This allocation belongs to custom pool. - if(hAllocPool != VK_NULL_HANDLE) + if (hAllocPool != VK_NULL_HANDLE) { // Pools with algorithm other than default are not defragmented. - if(hAllocPool->m_BlockVector.GetAlgorithm() == 0) + if (hAllocPool->m_pBlockVectors[memTypeIndex]->GetAlgorithm() == 0) { - for(size_t i = m_CustomPoolContexts.size(); i--; ) + for (size_t i = m_CustomPoolContexts.size(); i--; ) { - if(m_CustomPoolContexts[i]->GetCustomPool() == hAllocPool) + if (m_CustomPoolContexts[i]->GetCustomPool() == hAllocPool) { pBlockVectorDefragCtx = m_CustomPoolContexts[i]; break; } } - if(!pBlockVectorDefragCtx) + if (!pBlockVectorDefragCtx) { pBlockVectorDefragCtx = vma_new(m_hAllocator, VmaBlockVectorDefragmentationContext)( m_hAllocator, hAllocPool, - &hAllocPool->m_BlockVector, - m_CurrFrameIndex); + hAllocPool->m_pBlockVectors[memTypeIndex]); m_CustomPoolContexts.push_back(pBlockVectorDefragCtx); } } @@ -15211,20 +14248,20 @@ void VmaDefragmentationContext_T::AddAllocations( // This allocation belongs to default pool. else { - const uint32_t memTypeIndex = hAlloc->GetMemoryTypeIndex(); pBlockVectorDefragCtx = m_DefaultPoolContexts[memTypeIndex]; - if(!pBlockVectorDefragCtx) + if (!pBlockVectorDefragCtx) { + VMA_ASSERT(m_hAllocator->m_pBlockVectors[memTypeIndex] && "Trying to use unsupported memory type!"); + pBlockVectorDefragCtx = vma_new(m_hAllocator, VmaBlockVectorDefragmentationContext)( m_hAllocator, VMA_NULL, // hCustomPool - m_hAllocator->m_pBlockVectors[memTypeIndex], - m_CurrFrameIndex); + m_hAllocator->m_pBlockVectors[memTypeIndex]); m_DefaultPoolContexts[memTypeIndex] = pBlockVectorDefragCtx; } } - if(pBlockVectorDefragCtx) + if (pBlockVectorDefragCtx) { VkBool32* const pChanged = (pAllocationsChanged != VMA_NULL) ? &pAllocationsChanged[allocIndex] : VMA_NULL; @@ -15239,12 +14276,12 @@ VkResult VmaDefragmentationContext_T::Defragment( VkDeviceSize maxGpuBytesToMove, uint32_t maxGpuAllocationsToMove, VkCommandBuffer commandBuffer, VmaDefragmentationStats* pStats, VmaDefragmentationFlags flags) { - if(pStats) + if (pStats) { memset(pStats, 0, sizeof(VmaDefragmentationStats)); } - if(flags & VMA_DEFRAGMENTATION_FLAG_INCREMENTAL) + if (flags & VMA_DEFRAGMENTATION_FLAG_INCREMENTAL) { // For incremental defragmetnations, we just earmark how much we can move // The real meat is in the defragmentation steps @@ -15254,14 +14291,14 @@ VkResult VmaDefragmentationContext_T::Defragment( m_MaxGpuBytesToMove = maxGpuBytesToMove; m_MaxGpuAllocationsToMove = maxGpuAllocationsToMove; - if(m_MaxCpuBytesToMove == 0 && m_MaxCpuAllocationsToMove == 0 && + if (m_MaxCpuBytesToMove == 0 && m_MaxCpuAllocationsToMove == 0 && m_MaxGpuBytesToMove == 0 && m_MaxGpuAllocationsToMove == 0) return VK_SUCCESS; return VK_NOT_READY; } - if(commandBuffer == VK_NULL_HANDLE) + if (commandBuffer == VK_NULL_HANDLE) { maxGpuBytesToMove = 0; maxGpuAllocationsToMove = 0; @@ -15270,12 +14307,12 @@ VkResult VmaDefragmentationContext_T::Defragment( VkResult res = VK_SUCCESS; // Process default pools. - for(uint32_t memTypeIndex = 0; + for (uint32_t memTypeIndex = 0; memTypeIndex < m_hAllocator->GetMemoryTypeCount() && res >= VK_SUCCESS; ++memTypeIndex) { VmaBlockVectorDefragmentationContext* pBlockVectorCtx = m_DefaultPoolContexts[memTypeIndex]; - if(pBlockVectorCtx) + if (pBlockVectorCtx) { VMA_ASSERT(pBlockVectorCtx->GetBlockVector()); pBlockVectorCtx->GetBlockVector()->Defragment( @@ -15284,7 +14321,7 @@ VkResult VmaDefragmentationContext_T::Defragment( maxCpuBytesToMove, maxCpuAllocationsToMove, maxGpuBytesToMove, maxGpuAllocationsToMove, commandBuffer); - if(pBlockVectorCtx->res != VK_SUCCESS) + if (pBlockVectorCtx->res != VK_SUCCESS) { res = pBlockVectorCtx->res; } @@ -15292,7 +14329,7 @@ VkResult VmaDefragmentationContext_T::Defragment( } // Process custom pools. - for(size_t customCtxIndex = 0, customCtxCount = m_CustomPoolContexts.size(); + for (size_t customCtxIndex = 0, customCtxCount = m_CustomPoolContexts.size(); customCtxIndex < customCtxCount && res >= VK_SUCCESS; ++customCtxIndex) { @@ -15304,7 +14341,7 @@ VkResult VmaDefragmentationContext_T::Defragment( maxCpuBytesToMove, maxCpuAllocationsToMove, maxGpuBytesToMove, maxGpuAllocationsToMove, commandBuffer); - if(pBlockVectorCtx->res != VK_SUCCESS) + if (pBlockVectorCtx->res != VK_SUCCESS) { res = pBlockVectorCtx->res; } @@ -15319,16 +14356,16 @@ VkResult VmaDefragmentationContext_T::DefragmentPassBegin(VmaDefragmentationPass uint32_t movesLeft = pInfo->moveCount; // Process default pools. - for(uint32_t memTypeIndex = 0; + for (uint32_t memTypeIndex = 0; memTypeIndex < m_hAllocator->GetMemoryTypeCount(); ++memTypeIndex) { - VmaBlockVectorDefragmentationContext *pBlockVectorCtx = m_DefaultPoolContexts[memTypeIndex]; - if(pBlockVectorCtx) + VmaBlockVectorDefragmentationContext* pBlockVectorCtx = m_DefaultPoolContexts[memTypeIndex]; + if (pBlockVectorCtx) { VMA_ASSERT(pBlockVectorCtx->GetBlockVector()); - if(!pBlockVectorCtx->hasDefragmentationPlan) + if (!pBlockVectorCtx->hasDefragmentationPlan) { pBlockVectorCtx->GetBlockVector()->Defragment( pBlockVectorCtx, @@ -15337,7 +14374,7 @@ VkResult VmaDefragmentationContext_T::DefragmentPassBegin(VmaDefragmentationPass m_MaxGpuBytesToMove, m_MaxGpuAllocationsToMove, VK_NULL_HANDLE); - if(pBlockVectorCtx->res < VK_SUCCESS) + if (pBlockVectorCtx->res < VK_SUCCESS) continue; pBlockVectorCtx->hasDefragmentationPlan = true; @@ -15353,14 +14390,14 @@ VkResult VmaDefragmentationContext_T::DefragmentPassBegin(VmaDefragmentationPass } // Process custom pools. - for(size_t customCtxIndex = 0, customCtxCount = m_CustomPoolContexts.size(); + for (size_t customCtxIndex = 0, customCtxCount = m_CustomPoolContexts.size(); customCtxIndex < customCtxCount; ++customCtxIndex) { - VmaBlockVectorDefragmentationContext *pBlockVectorCtx = m_CustomPoolContexts[customCtxIndex]; + VmaBlockVectorDefragmentationContext* pBlockVectorCtx = m_CustomPoolContexts[customCtxIndex]; VMA_ASSERT(pBlockVectorCtx && pBlockVectorCtx->GetBlockVector()); - if(!pBlockVectorCtx->hasDefragmentationPlan) + if (!pBlockVectorCtx->hasDefragmentationPlan) { pBlockVectorCtx->GetBlockVector()->Defragment( pBlockVectorCtx, @@ -15369,7 +14406,7 @@ VkResult VmaDefragmentationContext_T::DefragmentPassBegin(VmaDefragmentationPass m_MaxGpuBytesToMove, m_MaxGpuAllocationsToMove, VK_NULL_HANDLE); - if(pBlockVectorCtx->res < VK_SUCCESS) + if (pBlockVectorCtx->res < VK_SUCCESS) continue; pBlockVectorCtx->hasDefragmentationPlan = true; @@ -15387,21 +14424,22 @@ VkResult VmaDefragmentationContext_T::DefragmentPassBegin(VmaDefragmentationPass return VK_SUCCESS; } + VkResult VmaDefragmentationContext_T::DefragmentPassEnd() { VkResult res = VK_SUCCESS; // Process default pools. - for(uint32_t memTypeIndex = 0; + for (uint32_t memTypeIndex = 0; memTypeIndex < m_hAllocator->GetMemoryTypeCount(); ++memTypeIndex) { - VmaBlockVectorDefragmentationContext *pBlockVectorCtx = m_DefaultPoolContexts[memTypeIndex]; - if(pBlockVectorCtx) + VmaBlockVectorDefragmentationContext* pBlockVectorCtx = m_DefaultPoolContexts[memTypeIndex]; + if (pBlockVectorCtx) { VMA_ASSERT(pBlockVectorCtx->GetBlockVector()); - if(!pBlockVectorCtx->hasDefragmentationPlan) + if (!pBlockVectorCtx->hasDefragmentationPlan) { res = VK_NOT_READY; continue; @@ -15410,20 +14448,20 @@ VkResult VmaDefragmentationContext_T::DefragmentPassEnd() pBlockVectorCtx->GetBlockVector()->CommitDefragmentations( pBlockVectorCtx, m_pStats); - if(pBlockVectorCtx->defragmentationMoves.size() != pBlockVectorCtx->defragmentationMovesCommitted) + if (pBlockVectorCtx->defragmentationMoves.size() != pBlockVectorCtx->defragmentationMovesCommitted) res = VK_NOT_READY; } } // Process custom pools. - for(size_t customCtxIndex = 0, customCtxCount = m_CustomPoolContexts.size(); + for (size_t customCtxIndex = 0, customCtxCount = m_CustomPoolContexts.size(); customCtxIndex < customCtxCount; ++customCtxIndex) { - VmaBlockVectorDefragmentationContext *pBlockVectorCtx = m_CustomPoolContexts[customCtxIndex]; + VmaBlockVectorDefragmentationContext* pBlockVectorCtx = m_CustomPoolContexts[customCtxIndex]; VMA_ASSERT(pBlockVectorCtx && pBlockVectorCtx->GetBlockVector()); - if(!pBlockVectorCtx->hasDefragmentationPlan) + if (!pBlockVectorCtx->hasDefragmentationPlan) { res = VK_NOT_READY; continue; @@ -15432,638 +14470,77 @@ VkResult VmaDefragmentationContext_T::DefragmentPassEnd() pBlockVectorCtx->GetBlockVector()->CommitDefragmentations( pBlockVectorCtx, m_pStats); - if(pBlockVectorCtx->defragmentationMoves.size() != pBlockVectorCtx->defragmentationMovesCommitted) + if (pBlockVectorCtx->defragmentationMoves.size() != pBlockVectorCtx->defragmentationMovesCommitted) res = VK_NOT_READY; } return res; } +#endif // _VMA_DEFRAGMENTATION_CONTEXT_FUNCTIONS -//////////////////////////////////////////////////////////////////////////////// -// VmaRecorder - -#if VMA_RECORDING_ENABLED - -VmaRecorder::VmaRecorder() : - m_UseMutex(true), - m_Flags(0), - m_File(VMA_NULL), - m_RecordingStartTime(std::chrono::high_resolution_clock::now()) -{ -} - -VkResult VmaRecorder::Init(const VmaRecordSettings& settings, bool useMutex) -{ - m_UseMutex = useMutex; - m_Flags = settings.flags; - -#if defined(_WIN32) - // Open file for writing. - errno_t err = fopen_s(&m_File, settings.pFilePath, "wb"); - - if(err != 0) - { - return VK_ERROR_INITIALIZATION_FAILED; - } -#else - // Open file for writing. - m_File = fopen(settings.pFilePath, "wb"); - - if(m_File == 0) - { - return VK_ERROR_INITIALIZATION_FAILED; - } -#endif - - // Write header. - fprintf(m_File, "%s\n", "Vulkan Memory Allocator,Calls recording"); - fprintf(m_File, "%s\n", "1,8"); - - return VK_SUCCESS; -} - -VmaRecorder::~VmaRecorder() -{ - if(m_File != VMA_NULL) - { - fclose(m_File); - } -} - -void VmaRecorder::RecordCreateAllocator(uint32_t frameIndex) -{ - CallParams callParams; - GetBasicParams(callParams); - - VmaMutexLock lock(m_FileMutex, m_UseMutex); - fprintf(m_File, "%u,%.3f,%u,vmaCreateAllocator\n", callParams.threadId, callParams.time, frameIndex); - Flush(); -} - -void VmaRecorder::RecordDestroyAllocator(uint32_t frameIndex) -{ - CallParams callParams; - GetBasicParams(callParams); - - VmaMutexLock lock(m_FileMutex, m_UseMutex); - fprintf(m_File, "%u,%.3f,%u,vmaDestroyAllocator\n", callParams.threadId, callParams.time, frameIndex); - Flush(); -} - -void VmaRecorder::RecordCreatePool(uint32_t frameIndex, const VmaPoolCreateInfo& createInfo, VmaPool pool) -{ - CallParams callParams; - GetBasicParams(callParams); - - VmaMutexLock lock(m_FileMutex, m_UseMutex); - fprintf(m_File, "%u,%.3f,%u,vmaCreatePool,%u,%u,%llu,%llu,%llu,%u,%p\n", callParams.threadId, callParams.time, frameIndex, - createInfo.memoryTypeIndex, - createInfo.flags, - createInfo.blockSize, - (uint64_t)createInfo.minBlockCount, - (uint64_t)createInfo.maxBlockCount, - createInfo.frameInUseCount, - pool); - Flush(); -} - -void VmaRecorder::RecordDestroyPool(uint32_t frameIndex, VmaPool pool) -{ - CallParams callParams; - GetBasicParams(callParams); - - VmaMutexLock lock(m_FileMutex, m_UseMutex); - fprintf(m_File, "%u,%.3f,%u,vmaDestroyPool,%p\n", callParams.threadId, callParams.time, frameIndex, - pool); - Flush(); -} - -void VmaRecorder::RecordAllocateMemory(uint32_t frameIndex, - const VkMemoryRequirements& vkMemReq, - const VmaAllocationCreateInfo& createInfo, - VmaAllocation allocation) -{ - CallParams callParams; - GetBasicParams(callParams); - - VmaMutexLock lock(m_FileMutex, m_UseMutex); - UserDataString userDataStr(createInfo.flags, createInfo.pUserData); - fprintf(m_File, "%u,%.3f,%u,vmaAllocateMemory,%llu,%llu,%u,%u,%u,%u,%u,%u,%p,%p,%s\n", callParams.threadId, callParams.time, frameIndex, - vkMemReq.size, - vkMemReq.alignment, - vkMemReq.memoryTypeBits, - createInfo.flags, - createInfo.usage, - createInfo.requiredFlags, - createInfo.preferredFlags, - createInfo.memoryTypeBits, - createInfo.pool, - allocation, - userDataStr.GetString()); - Flush(); -} - -void VmaRecorder::RecordAllocateMemoryPages(uint32_t frameIndex, - const VkMemoryRequirements& vkMemReq, - const VmaAllocationCreateInfo& createInfo, - uint64_t allocationCount, - const VmaAllocation* pAllocations) -{ - CallParams callParams; - GetBasicParams(callParams); - - VmaMutexLock lock(m_FileMutex, m_UseMutex); - UserDataString userDataStr(createInfo.flags, createInfo.pUserData); - fprintf(m_File, "%u,%.3f,%u,vmaAllocateMemoryPages,%llu,%llu,%u,%u,%u,%u,%u,%u,%p,", callParams.threadId, callParams.time, frameIndex, - vkMemReq.size, - vkMemReq.alignment, - vkMemReq.memoryTypeBits, - createInfo.flags, - createInfo.usage, - createInfo.requiredFlags, - createInfo.preferredFlags, - createInfo.memoryTypeBits, - createInfo.pool); - PrintPointerList(allocationCount, pAllocations); - fprintf(m_File, ",%s\n", userDataStr.GetString()); - Flush(); -} - -void VmaRecorder::RecordAllocateMemoryForBuffer(uint32_t frameIndex, - const VkMemoryRequirements& vkMemReq, - bool requiresDedicatedAllocation, - bool prefersDedicatedAllocation, - const VmaAllocationCreateInfo& createInfo, - VmaAllocation allocation) -{ - CallParams callParams; - GetBasicParams(callParams); - - VmaMutexLock lock(m_FileMutex, m_UseMutex); - UserDataString userDataStr(createInfo.flags, createInfo.pUserData); - fprintf(m_File, "%u,%.3f,%u,vmaAllocateMemoryForBuffer,%llu,%llu,%u,%u,%u,%u,%u,%u,%u,%u,%p,%p,%s\n", callParams.threadId, callParams.time, frameIndex, - vkMemReq.size, - vkMemReq.alignment, - vkMemReq.memoryTypeBits, - requiresDedicatedAllocation ? 1 : 0, - prefersDedicatedAllocation ? 1 : 0, - createInfo.flags, - createInfo.usage, - createInfo.requiredFlags, - createInfo.preferredFlags, - createInfo.memoryTypeBits, - createInfo.pool, - allocation, - userDataStr.GetString()); - Flush(); -} - -void VmaRecorder::RecordAllocateMemoryForImage(uint32_t frameIndex, - const VkMemoryRequirements& vkMemReq, - bool requiresDedicatedAllocation, - bool prefersDedicatedAllocation, - const VmaAllocationCreateInfo& createInfo, - VmaAllocation allocation) -{ - CallParams callParams; - GetBasicParams(callParams); - - VmaMutexLock lock(m_FileMutex, m_UseMutex); - UserDataString userDataStr(createInfo.flags, createInfo.pUserData); - fprintf(m_File, "%u,%.3f,%u,vmaAllocateMemoryForImage,%llu,%llu,%u,%u,%u,%u,%u,%u,%u,%u,%p,%p,%s\n", callParams.threadId, callParams.time, frameIndex, - vkMemReq.size, - vkMemReq.alignment, - vkMemReq.memoryTypeBits, - requiresDedicatedAllocation ? 1 : 0, - prefersDedicatedAllocation ? 1 : 0, - createInfo.flags, - createInfo.usage, - createInfo.requiredFlags, - createInfo.preferredFlags, - createInfo.memoryTypeBits, - createInfo.pool, - allocation, - userDataStr.GetString()); - Flush(); -} - -void VmaRecorder::RecordFreeMemory(uint32_t frameIndex, - VmaAllocation allocation) -{ - CallParams callParams; - GetBasicParams(callParams); - - VmaMutexLock lock(m_FileMutex, m_UseMutex); - fprintf(m_File, "%u,%.3f,%u,vmaFreeMemory,%p\n", callParams.threadId, callParams.time, frameIndex, - allocation); - Flush(); -} - -void VmaRecorder::RecordFreeMemoryPages(uint32_t frameIndex, - uint64_t allocationCount, - const VmaAllocation* pAllocations) -{ - CallParams callParams; - GetBasicParams(callParams); - - VmaMutexLock lock(m_FileMutex, m_UseMutex); - fprintf(m_File, "%u,%.3f,%u,vmaFreeMemoryPages,", callParams.threadId, callParams.time, frameIndex); - PrintPointerList(allocationCount, pAllocations); - fprintf(m_File, "\n"); - Flush(); -} - -void VmaRecorder::RecordSetAllocationUserData(uint32_t frameIndex, - VmaAllocation allocation, - const void* pUserData) -{ - CallParams callParams; - GetBasicParams(callParams); - - VmaMutexLock lock(m_FileMutex, m_UseMutex); - UserDataString userDataStr( - allocation->IsUserDataString() ? VMA_ALLOCATION_CREATE_USER_DATA_COPY_STRING_BIT : 0, - pUserData); - fprintf(m_File, "%u,%.3f,%u,vmaSetAllocationUserData,%p,%s\n", callParams.threadId, callParams.time, frameIndex, - allocation, - userDataStr.GetString()); - Flush(); -} - -void VmaRecorder::RecordCreateLostAllocation(uint32_t frameIndex, - VmaAllocation allocation) -{ - CallParams callParams; - GetBasicParams(callParams); - - VmaMutexLock lock(m_FileMutex, m_UseMutex); - fprintf(m_File, "%u,%.3f,%u,vmaCreateLostAllocation,%p\n", callParams.threadId, callParams.time, frameIndex, - allocation); - Flush(); -} - -void VmaRecorder::RecordMapMemory(uint32_t frameIndex, - VmaAllocation allocation) -{ - CallParams callParams; - GetBasicParams(callParams); - - VmaMutexLock lock(m_FileMutex, m_UseMutex); - fprintf(m_File, "%u,%.3f,%u,vmaMapMemory,%p\n", callParams.threadId, callParams.time, frameIndex, - allocation); - Flush(); -} - -void VmaRecorder::RecordUnmapMemory(uint32_t frameIndex, - VmaAllocation allocation) -{ - CallParams callParams; - GetBasicParams(callParams); - - VmaMutexLock lock(m_FileMutex, m_UseMutex); - fprintf(m_File, "%u,%.3f,%u,vmaUnmapMemory,%p\n", callParams.threadId, callParams.time, frameIndex, - allocation); - Flush(); -} - -void VmaRecorder::RecordFlushAllocation(uint32_t frameIndex, - VmaAllocation allocation, VkDeviceSize offset, VkDeviceSize size) -{ - CallParams callParams; - GetBasicParams(callParams); - - VmaMutexLock lock(m_FileMutex, m_UseMutex); - fprintf(m_File, "%u,%.3f,%u,vmaFlushAllocation,%p,%llu,%llu\n", callParams.threadId, callParams.time, frameIndex, - allocation, - offset, - size); - Flush(); -} - -void VmaRecorder::RecordInvalidateAllocation(uint32_t frameIndex, - VmaAllocation allocation, VkDeviceSize offset, VkDeviceSize size) -{ - CallParams callParams; - GetBasicParams(callParams); - - VmaMutexLock lock(m_FileMutex, m_UseMutex); - fprintf(m_File, "%u,%.3f,%u,vmaInvalidateAllocation,%p,%llu,%llu\n", callParams.threadId, callParams.time, frameIndex, - allocation, - offset, - size); - Flush(); -} - -void VmaRecorder::RecordCreateBuffer(uint32_t frameIndex, - const VkBufferCreateInfo& bufCreateInfo, - const VmaAllocationCreateInfo& allocCreateInfo, - VmaAllocation allocation) -{ - CallParams callParams; - GetBasicParams(callParams); - - VmaMutexLock lock(m_FileMutex, m_UseMutex); - UserDataString userDataStr(allocCreateInfo.flags, allocCreateInfo.pUserData); - fprintf(m_File, "%u,%.3f,%u,vmaCreateBuffer,%u,%llu,%u,%u,%u,%u,%u,%u,%u,%p,%p,%s\n", callParams.threadId, callParams.time, frameIndex, - bufCreateInfo.flags, - bufCreateInfo.size, - bufCreateInfo.usage, - bufCreateInfo.sharingMode, - allocCreateInfo.flags, - allocCreateInfo.usage, - allocCreateInfo.requiredFlags, - allocCreateInfo.preferredFlags, - allocCreateInfo.memoryTypeBits, - allocCreateInfo.pool, - allocation, - userDataStr.GetString()); - Flush(); -} - -void VmaRecorder::RecordCreateImage(uint32_t frameIndex, - const VkImageCreateInfo& imageCreateInfo, - const VmaAllocationCreateInfo& allocCreateInfo, - VmaAllocation allocation) -{ - CallParams callParams; - GetBasicParams(callParams); - - VmaMutexLock lock(m_FileMutex, m_UseMutex); - UserDataString userDataStr(allocCreateInfo.flags, allocCreateInfo.pUserData); - fprintf(m_File, "%u,%.3f,%u,vmaCreateImage,%u,%u,%u,%u,%u,%u,%u,%u,%u,%u,%u,%u,%u,%u,%u,%u,%u,%u,%p,%p,%s\n", callParams.threadId, callParams.time, frameIndex, - imageCreateInfo.flags, - imageCreateInfo.imageType, - imageCreateInfo.format, - imageCreateInfo.extent.width, - imageCreateInfo.extent.height, - imageCreateInfo.extent.depth, - imageCreateInfo.mipLevels, - imageCreateInfo.arrayLayers, - imageCreateInfo.samples, - imageCreateInfo.tiling, - imageCreateInfo.usage, - imageCreateInfo.sharingMode, - imageCreateInfo.initialLayout, - allocCreateInfo.flags, - allocCreateInfo.usage, - allocCreateInfo.requiredFlags, - allocCreateInfo.preferredFlags, - allocCreateInfo.memoryTypeBits, - allocCreateInfo.pool, - allocation, - userDataStr.GetString()); - Flush(); -} - -void VmaRecorder::RecordDestroyBuffer(uint32_t frameIndex, - VmaAllocation allocation) -{ - CallParams callParams; - GetBasicParams(callParams); - - VmaMutexLock lock(m_FileMutex, m_UseMutex); - fprintf(m_File, "%u,%.3f,%u,vmaDestroyBuffer,%p\n", callParams.threadId, callParams.time, frameIndex, - allocation); - Flush(); -} - -void VmaRecorder::RecordDestroyImage(uint32_t frameIndex, - VmaAllocation allocation) -{ - CallParams callParams; - GetBasicParams(callParams); - - VmaMutexLock lock(m_FileMutex, m_UseMutex); - fprintf(m_File, "%u,%.3f,%u,vmaDestroyImage,%p\n", callParams.threadId, callParams.time, frameIndex, - allocation); - Flush(); -} - -void VmaRecorder::RecordTouchAllocation(uint32_t frameIndex, - VmaAllocation allocation) -{ - CallParams callParams; - GetBasicParams(callParams); - - VmaMutexLock lock(m_FileMutex, m_UseMutex); - fprintf(m_File, "%u,%.3f,%u,vmaTouchAllocation,%p\n", callParams.threadId, callParams.time, frameIndex, - allocation); - Flush(); -} - -void VmaRecorder::RecordGetAllocationInfo(uint32_t frameIndex, - VmaAllocation allocation) -{ - CallParams callParams; - GetBasicParams(callParams); - - VmaMutexLock lock(m_FileMutex, m_UseMutex); - fprintf(m_File, "%u,%.3f,%u,vmaGetAllocationInfo,%p\n", callParams.threadId, callParams.time, frameIndex, - allocation); - Flush(); -} - -void VmaRecorder::RecordMakePoolAllocationsLost(uint32_t frameIndex, - VmaPool pool) -{ - CallParams callParams; - GetBasicParams(callParams); - - VmaMutexLock lock(m_FileMutex, m_UseMutex); - fprintf(m_File, "%u,%.3f,%u,vmaMakePoolAllocationsLost,%p\n", callParams.threadId, callParams.time, frameIndex, - pool); - Flush(); -} - -void VmaRecorder::RecordDefragmentationBegin(uint32_t frameIndex, - const VmaDefragmentationInfo2& info, - VmaDefragmentationContext ctx) -{ - CallParams callParams; - GetBasicParams(callParams); - - VmaMutexLock lock(m_FileMutex, m_UseMutex); - fprintf(m_File, "%u,%.3f,%u,vmaDefragmentationBegin,%u,", callParams.threadId, callParams.time, frameIndex, - info.flags); - PrintPointerList(info.allocationCount, info.pAllocations); - fprintf(m_File, ","); - PrintPointerList(info.poolCount, info.pPools); - fprintf(m_File, ",%llu,%u,%llu,%u,%p,%p\n", - info.maxCpuBytesToMove, - info.maxCpuAllocationsToMove, - info.maxGpuBytesToMove, - info.maxGpuAllocationsToMove, - info.commandBuffer, - ctx); - Flush(); -} - -void VmaRecorder::RecordDefragmentationEnd(uint32_t frameIndex, - VmaDefragmentationContext ctx) -{ - CallParams callParams; - GetBasicParams(callParams); - - VmaMutexLock lock(m_FileMutex, m_UseMutex); - fprintf(m_File, "%u,%.3f,%u,vmaDefragmentationEnd,%p\n", callParams.threadId, callParams.time, frameIndex, - ctx); - Flush(); -} - -void VmaRecorder::RecordSetPoolName(uint32_t frameIndex, - VmaPool pool, - const char* name) -{ - CallParams callParams; - GetBasicParams(callParams); - - VmaMutexLock lock(m_FileMutex, m_UseMutex); - fprintf(m_File, "%u,%.3f,%u,vmaSetPoolName,%p,%s\n", callParams.threadId, callParams.time, frameIndex, - pool, name != VMA_NULL ? name : ""); - Flush(); -} - -VmaRecorder::UserDataString::UserDataString(VmaAllocationCreateFlags allocFlags, const void* pUserData) +#ifndef _VMA_POOL_T_FUNCTIONS +VmaPool_T::VmaPool_T( + VmaAllocator hAllocator, + const VmaPoolCreateInfo& createInfo) : + m_hAllocator(hAllocator), + m_pBlockVectors{}, + m_Id(0), + m_Name(VMA_NULL) { - if(pUserData != VMA_NULL) + for(uint32_t memTypeIndex = 0; memTypeIndex < hAllocator->GetMemoryTypeCount(); ++memTypeIndex) { - if((allocFlags & VMA_ALLOCATION_CREATE_USER_DATA_COPY_STRING_BIT) != 0) - { - m_Str = (const char*)pUserData; - } - else + // Create only supported types + if((hAllocator->GetGlobalMemoryTypeBits() & (1u << memTypeIndex)) != 0) { - // If VMA_ALLOCATION_CREATE_USER_DATA_COPY_STRING_BIT is not specified, convert the string's memory address to a string and store it. - snprintf(m_PtrStr, 17, "%p", pUserData); - m_Str = m_PtrStr; + m_pBlockVectors[memTypeIndex] = vma_new(hAllocator, VmaBlockVector)( + hAllocator, + this, // hParentPool + memTypeIndex, + createInfo.blockSize != 0 ? createInfo.blockSize : hAllocator->CalcPreferredBlockSize(memTypeIndex), + createInfo.minBlockCount, + createInfo.maxBlockCount, + (createInfo.flags& VMA_POOL_CREATE_IGNORE_BUFFER_IMAGE_GRANULARITY_BIT) != 0 ? 1 : hAllocator->GetBufferImageGranularity(), + false, // explicitBlockSize + createInfo.flags & VMA_POOL_CREATE_ALGORITHM_MASK, // algorithm + createInfo.priority, + VMA_MAX(hAllocator->GetMemoryTypeMinAlignment(memTypeIndex), createInfo.minAllocationAlignment), + createInfo.pMemoryAllocateNext); } } - else - { - m_Str = ""; - } } -void VmaRecorder::WriteConfiguration( - const VkPhysicalDeviceProperties& devProps, - const VkPhysicalDeviceMemoryProperties& memProps, - uint32_t vulkanApiVersion, - bool dedicatedAllocationExtensionEnabled, - bool bindMemory2ExtensionEnabled, - bool memoryBudgetExtensionEnabled, - bool deviceCoherentMemoryExtensionEnabled) +VmaPool_T::~VmaPool_T() { - fprintf(m_File, "Config,Begin\n"); - - fprintf(m_File, "VulkanApiVersion,%u,%u\n", VK_VERSION_MAJOR(vulkanApiVersion), VK_VERSION_MINOR(vulkanApiVersion)); - - fprintf(m_File, "PhysicalDevice,apiVersion,%u\n", devProps.apiVersion); - fprintf(m_File, "PhysicalDevice,driverVersion,%u\n", devProps.driverVersion); - fprintf(m_File, "PhysicalDevice,vendorID,%u\n", devProps.vendorID); - fprintf(m_File, "PhysicalDevice,deviceID,%u\n", devProps.deviceID); - fprintf(m_File, "PhysicalDevice,deviceType,%u\n", devProps.deviceType); - fprintf(m_File, "PhysicalDevice,deviceName,%s\n", devProps.deviceName); - - fprintf(m_File, "PhysicalDeviceLimits,maxMemoryAllocationCount,%u\n", devProps.limits.maxMemoryAllocationCount); - fprintf(m_File, "PhysicalDeviceLimits,bufferImageGranularity,%llu\n", devProps.limits.bufferImageGranularity); - fprintf(m_File, "PhysicalDeviceLimits,nonCoherentAtomSize,%llu\n", devProps.limits.nonCoherentAtomSize); - - fprintf(m_File, "PhysicalDeviceMemory,HeapCount,%u\n", memProps.memoryHeapCount); - for(uint32_t i = 0; i < memProps.memoryHeapCount; ++i) - { - fprintf(m_File, "PhysicalDeviceMemory,Heap,%u,size,%llu\n", i, memProps.memoryHeaps[i].size); - fprintf(m_File, "PhysicalDeviceMemory,Heap,%u,flags,%u\n", i, memProps.memoryHeaps[i].flags); - } - fprintf(m_File, "PhysicalDeviceMemory,TypeCount,%u\n", memProps.memoryTypeCount); - for(uint32_t i = 0; i < memProps.memoryTypeCount; ++i) + VMA_ASSERT(m_PrevPool == VMA_NULL && m_NextPool == VMA_NULL); + for(uint32_t memTypeIndex = 0; memTypeIndex < m_hAllocator->GetMemoryTypeCount(); ++memTypeIndex) { - fprintf(m_File, "PhysicalDeviceMemory,Type,%u,heapIndex,%u\n", i, memProps.memoryTypes[i].heapIndex); - fprintf(m_File, "PhysicalDeviceMemory,Type,%u,propertyFlags,%u\n", i, memProps.memoryTypes[i].propertyFlags); + vma_delete(m_hAllocator, m_pBlockVectors[memTypeIndex]); } - - fprintf(m_File, "Extension,VK_KHR_dedicated_allocation,%u\n", dedicatedAllocationExtensionEnabled ? 1 : 0); - fprintf(m_File, "Extension,VK_KHR_bind_memory2,%u\n", bindMemory2ExtensionEnabled ? 1 : 0); - fprintf(m_File, "Extension,VK_EXT_memory_budget,%u\n", memoryBudgetExtensionEnabled ? 1 : 0); - fprintf(m_File, "Extension,VK_AMD_device_coherent_memory,%u\n", deviceCoherentMemoryExtensionEnabled ? 1 : 0); - - fprintf(m_File, "Macro,VMA_DEBUG_ALWAYS_DEDICATED_MEMORY,%u\n", VMA_DEBUG_ALWAYS_DEDICATED_MEMORY ? 1 : 0); - fprintf(m_File, "Macro,VMA_MIN_ALIGNMENT,%llu\n", (VkDeviceSize)VMA_MIN_ALIGNMENT); - fprintf(m_File, "Macro,VMA_DEBUG_MARGIN,%llu\n", (VkDeviceSize)VMA_DEBUG_MARGIN); - fprintf(m_File, "Macro,VMA_DEBUG_INITIALIZE_ALLOCATIONS,%u\n", VMA_DEBUG_INITIALIZE_ALLOCATIONS ? 1 : 0); - fprintf(m_File, "Macro,VMA_DEBUG_DETECT_CORRUPTION,%u\n", VMA_DEBUG_DETECT_CORRUPTION ? 1 : 0); - fprintf(m_File, "Macro,VMA_DEBUG_GLOBAL_MUTEX,%u\n", VMA_DEBUG_GLOBAL_MUTEX ? 1 : 0); - fprintf(m_File, "Macro,VMA_DEBUG_MIN_BUFFER_IMAGE_GRANULARITY,%llu\n", (VkDeviceSize)VMA_DEBUG_MIN_BUFFER_IMAGE_GRANULARITY); - fprintf(m_File, "Macro,VMA_SMALL_HEAP_MAX_SIZE,%llu\n", (VkDeviceSize)VMA_SMALL_HEAP_MAX_SIZE); - fprintf(m_File, "Macro,VMA_DEFAULT_LARGE_HEAP_BLOCK_SIZE,%llu\n", (VkDeviceSize)VMA_DEFAULT_LARGE_HEAP_BLOCK_SIZE); - - fprintf(m_File, "Config,End\n"); -} - -void VmaRecorder::GetBasicParams(CallParams& outParams) -{ - #if defined(_WIN32) - outParams.threadId = GetCurrentThreadId(); - #else - // Use C++11 features to get thread id and convert it to uint32_t. - // There is room for optimization since sstream is quite slow. - // Is there a better way to convert std::this_thread::get_id() to uint32_t? - std::thread::id thread_id = std::this_thread::get_id(); - std::stringstream thread_id_to_string_converter; - thread_id_to_string_converter << thread_id; - std::string thread_id_as_string = thread_id_to_string_converter.str(); - outParams.threadId = static_cast<uint32_t>(std::stoi(thread_id_as_string.c_str())); - #endif - - auto current_time = std::chrono::high_resolution_clock::now(); - - outParams.time = std::chrono::duration<double, std::chrono::seconds::period>(current_time - m_RecordingStartTime).count(); } -void VmaRecorder::PrintPointerList(uint64_t count, const VmaAllocation* pItems) +void VmaPool_T::SetName(const char* pName) { - if(count) + for(uint32_t memTypeIndex = 0; memTypeIndex < m_hAllocator->GetMemoryTypeCount(); ++memTypeIndex) { - fprintf(m_File, "%p", pItems[0]); - for(uint64_t i = 1; i < count; ++i) + if(m_pBlockVectors[memTypeIndex]) { - fprintf(m_File, " %p", pItems[i]); - } - } -} + const VkAllocationCallbacks* allocs = m_pBlockVectors[memTypeIndex]->GetAllocator()->GetAllocationCallbacks(); + VmaFreeString(allocs, m_Name); -void VmaRecorder::Flush() -{ - if((m_Flags & VMA_RECORD_FLUSH_AFTER_CALL_BIT) != 0) - { - fflush(m_File); + if (pName != VMA_NULL) + { + m_Name = VmaCreateStringCopy(allocs, pName); + } + else + { + m_Name = VMA_NULL; + } + } } } +#endif // _VMA_POOL_T_FUNCTIONS -#endif // #if VMA_RECORDING_ENABLED - -//////////////////////////////////////////////////////////////////////////////// -// VmaAllocationObjectAllocator - -VmaAllocationObjectAllocator::VmaAllocationObjectAllocator(const VkAllocationCallbacks* pAllocationCallbacks) : - m_Allocator(pAllocationCallbacks, 1024) -{ -} - -template<typename... Types> VmaAllocation VmaAllocationObjectAllocator::Allocate(Types... args) -{ - VmaMutexLock mutexLock(m_Mutex); - return m_Allocator.Alloc<Types...>(std::forward<Types>(args)...); -} - -void VmaAllocationObjectAllocator::Free(VmaAllocation hAlloc) -{ - VmaMutexLock mutexLock(m_Mutex); - m_Allocator.Free(hAlloc); -} - -//////////////////////////////////////////////////////////////////////////////// -// VmaAllocator_T - +#ifndef _VMA_ALLOCATOR_T_FUNCTIONS VmaAllocator_T::VmaAllocator_T(const VmaAllocatorCreateInfo* pCreateInfo) : m_UseMutex((pCreateInfo->flags & VMA_ALLOCATOR_CREATE_EXTERNALLY_SYNCHRONIZED_BIT) == 0), m_VulkanApiVersion(pCreateInfo->vulkanApiVersion != 0 ? pCreateInfo->vulkanApiVersion : VK_API_VERSION_1_0), @@ -16083,13 +14560,9 @@ VmaAllocator_T::VmaAllocator_T(const VmaAllocatorCreateInfo* pCreateInfo) : m_DeviceMemoryCount(0), m_PreferredLargeHeapBlockSize(0), m_PhysicalDevice(pCreateInfo->physicalDevice), - m_CurrentFrameIndex(0), m_GpuDefragmentationMemoryTypeBits(UINT32_MAX), m_NextPoolId(0), m_GlobalMemoryTypeBits(UINT32_MAX) -#if VMA_RECORDING_ENABLED - ,m_pRecorder(VMA_NULL) -#endif { if(m_VulkanApiVersion >= VK_MAKE_VERSION(1, 1, 0)) { @@ -16156,7 +14629,6 @@ VmaAllocator_T::VmaAllocator_T(const VmaAllocatorCreateInfo* pCreateInfo) : memset(&m_MemProps, 0, sizeof(m_MemProps)); memset(&m_pBlockVectors, 0, sizeof(m_pBlockVectors)); - memset(&m_pSmallBufferBlockVectors, 0, sizeof(m_pSmallBufferBlockVectors)); memset(&m_VulkanFunctions, 0, sizeof(m_VulkanFunctions)); #if VMA_EXTERNAL_MEMORY @@ -16211,38 +14683,26 @@ VmaAllocator_T::VmaAllocator_T(const VmaAllocatorCreateInfo* pCreateInfo) : for(uint32_t memTypeIndex = 0; memTypeIndex < GetMemoryTypeCount(); ++memTypeIndex) { - const VkDeviceSize preferredBlockSize = CalcPreferredBlockSize(memTypeIndex); - - m_pBlockVectors[memTypeIndex] = vma_new(this, VmaBlockVector)( - this, - VK_NULL_HANDLE, // hParentPool - memTypeIndex, - preferredBlockSize, - 0, - SIZE_MAX, - GetBufferImageGranularity(), - pCreateInfo->frameInUseCount, - false, // explicitBlockSize - false, // linearAlgorithm - 0.5f, // priority (0.5 is the default per Vulkan spec) - GetMemoryTypeMinAlignment(memTypeIndex), // minAllocationAlignment - VMA_NULL); // // pMemoryAllocateNext - m_pSmallBufferBlockVectors[memTypeIndex] = vma_new(this, VmaBlockVector)( - this, - VK_NULL_HANDLE, // hParentPool - memTypeIndex, - preferredBlockSize, - 0, - SIZE_MAX, - 1, // bufferImageGranularity forced to 1 !!! - pCreateInfo->frameInUseCount, - false, // explicitBlockSize - false, // linearAlgorithm - 0.5f, // priority (0.5 is the default per Vulkan spec) - GetMemoryTypeMinAlignment(memTypeIndex), // minAllocationAlignment - VMA_NULL); // // pMemoryAllocateNext - // No need to call m_pBlockVectors[memTypeIndex][blockVectorTypeIndex]->CreateMinBlocks here, - // becase minBlockCount is 0. + // Create only supported types + if((m_GlobalMemoryTypeBits & (1u << memTypeIndex)) != 0) + { + const VkDeviceSize preferredBlockSize = CalcPreferredBlockSize(memTypeIndex); + m_pBlockVectors[memTypeIndex] = vma_new(this, VmaBlockVector)( + this, + VK_NULL_HANDLE, // hParentPool + memTypeIndex, + preferredBlockSize, + 0, + SIZE_MAX, + GetBufferImageGranularity(), + false, // explicitBlockSize + 0, // algorithm + 0.5f, // priority (0.5 is the default per Vulkan spec) + GetMemoryTypeMinAlignment(memTypeIndex), // minAllocationAlignment + VMA_NULL); // // pMemoryAllocateNext + // No need to call m_pBlockVectors[memTypeIndex][blockVectorTypeIndex]->CreateMinBlocks here, + // becase minBlockCount is 0. + } } } @@ -16250,31 +14710,6 @@ VkResult VmaAllocator_T::Init(const VmaAllocatorCreateInfo* pCreateInfo) { VkResult res = VK_SUCCESS; - if(pCreateInfo->pRecordSettings != VMA_NULL && - !VmaStrIsEmpty(pCreateInfo->pRecordSettings->pFilePath)) - { -#if VMA_RECORDING_ENABLED - m_pRecorder = vma_new(this, VmaRecorder)(); - res = m_pRecorder->Init(*pCreateInfo->pRecordSettings, m_UseMutex); - if(res != VK_SUCCESS) - { - return res; - } - m_pRecorder->WriteConfiguration( - m_PhysicalDeviceProperties, - m_MemProps, - m_VulkanApiVersion, - m_UseKhrDedicatedAllocation, - m_UseKhrBindMemory2, - m_UseExtMemoryBudget, - m_UseAmdDeviceCoherentMemory); - m_pRecorder->RecordCreateAllocator(GetCurrentFrameIndex()); -#else - VMA_ASSERT(0 && "VmaAllocatorCreateInfo::pRecordSettings used, but not supported due to VMA_RECORDING_ENABLED not defined to 1."); - return VK_ERROR_FEATURE_NOT_PRESENT; -#endif - } - #if VMA_MEMORY_BUDGET if(m_UseExtMemoryBudget) { @@ -16287,24 +14722,10 @@ VkResult VmaAllocator_T::Init(const VmaAllocatorCreateInfo* pCreateInfo) VmaAllocator_T::~VmaAllocator_T() { -#if VMA_RECORDING_ENABLED - if(m_pRecorder != VMA_NULL) - { - m_pRecorder->RecordDestroyAllocator(GetCurrentFrameIndex()); - vma_delete(this, m_pRecorder); - } -#endif - VMA_ASSERT(m_Pools.IsEmpty()); for(size_t memTypeIndex = GetMemoryTypeCount(); memTypeIndex--; ) { - if(!m_DedicatedAllocations[memTypeIndex].IsEmpty()) - { - VMA_ASSERT(0 && "Unfreed dedicated allocations found."); - } - - vma_delete(this, m_pSmallBufferBlockVectors[memTypeIndex]); vma_delete(this, m_pBlockVectors[memTypeIndex]); } } @@ -16332,6 +14753,8 @@ void VmaAllocator_T::ImportVulkanFunctions(const VmaVulkanFunctions* pVulkanFunc void VmaAllocator_T::ImportVulkanFunctions_Static() { // Vulkan 1.0 + m_VulkanFunctions.vkGetInstanceProcAddr = (PFN_vkGetInstanceProcAddr)vkGetInstanceProcAddr; + m_VulkanFunctions.vkGetDeviceProcAddr = (PFN_vkGetDeviceProcAddr)vkGetDeviceProcAddr; m_VulkanFunctions.vkGetPhysicalDeviceProperties = (PFN_vkGetPhysicalDeviceProperties)vkGetPhysicalDeviceProperties; m_VulkanFunctions.vkGetPhysicalDeviceMemoryProperties = (PFN_vkGetPhysicalDeviceMemoryProperties)vkGetPhysicalDeviceMemoryProperties; m_VulkanFunctions.vkAllocateMemory = (PFN_vkAllocateMemory)vkAllocateMemory; @@ -16363,7 +14786,7 @@ void VmaAllocator_T::ImportVulkanFunctions_Static() #endif } -#endif // #if VMA_STATIC_VULKAN_FUNCTIONS == 1 +#endif // VMA_STATIC_VULKAN_FUNCTIONS == 1 void VmaAllocator_T::ImportVulkanFunctions_Custom(const VmaVulkanFunctions* pVulkanFunctions) { @@ -16372,6 +14795,8 @@ void VmaAllocator_T::ImportVulkanFunctions_Custom(const VmaVulkanFunctions* pVul #define VMA_COPY_IF_NOT_NULL(funcName) \ if(pVulkanFunctions->funcName != VMA_NULL) m_VulkanFunctions.funcName = pVulkanFunctions->funcName; + VMA_COPY_IF_NOT_NULL(vkGetInstanceProcAddr); + VMA_COPY_IF_NOT_NULL(vkGetDeviceProcAddr); VMA_COPY_IF_NOT_NULL(vkGetPhysicalDeviceProperties); VMA_COPY_IF_NOT_NULL(vkGetPhysicalDeviceMemoryProperties); VMA_COPY_IF_NOT_NULL(vkAllocateMemory); @@ -16411,14 +14836,19 @@ void VmaAllocator_T::ImportVulkanFunctions_Custom(const VmaVulkanFunctions* pVul void VmaAllocator_T::ImportVulkanFunctions_Dynamic() { + VMA_ASSERT(m_VulkanFunctions.vkGetInstanceProcAddr && m_VulkanFunctions.vkGetDeviceProcAddr && + "To use VMA_DYNAMIC_VULKAN_FUNCTIONS in new versions of VMA you now have to pass " + "VmaVulkanFunctions::vkGetInstanceProcAddr and vkGetDeviceProcAddr as VmaAllocatorCreateInfo::pVulkanFunctions. " + "Other members can be null."); + #define VMA_FETCH_INSTANCE_FUNC(memberName, functionPointerType, functionNameString) \ if(m_VulkanFunctions.memberName == VMA_NULL) \ m_VulkanFunctions.memberName = \ - (functionPointerType)vkGetInstanceProcAddr(m_hInstance, functionNameString); + (functionPointerType)m_VulkanFunctions.vkGetInstanceProcAddr(m_hInstance, functionNameString); #define VMA_FETCH_DEVICE_FUNC(memberName, functionPointerType, functionNameString) \ if(m_VulkanFunctions.memberName == VMA_NULL) \ m_VulkanFunctions.memberName = \ - (functionPointerType)vkGetDeviceProcAddr(m_hDevice, functionNameString); + (functionPointerType)m_VulkanFunctions.vkGetDeviceProcAddr(m_hDevice, functionNameString); VMA_FETCH_INSTANCE_FUNC(vkGetPhysicalDeviceProperties, PFN_vkGetPhysicalDeviceProperties, "vkGetPhysicalDeviceProperties"); VMA_FETCH_INSTANCE_FUNC(vkGetPhysicalDeviceMemoryProperties, PFN_vkGetPhysicalDeviceMemoryProperties, "vkGetPhysicalDeviceMemoryProperties"); @@ -16476,7 +14906,7 @@ void VmaAllocator_T::ImportVulkanFunctions_Dynamic() #undef VMA_FETCH_INSTANCE_FUNC } -#endif // #if VMA_DYNAMIC_VULKAN_FUNCTIONS == 1 +#endif // VMA_DYNAMIC_VULKAN_FUNCTIONS == 1 void VmaAllocator_T::ValidateVulkanFunctions() { @@ -16531,15 +14961,18 @@ VkDeviceSize VmaAllocator_T::CalcPreferredBlockSize(uint32_t memTypeIndex) } VkResult VmaAllocator_T::AllocateMemoryOfType( + VmaPool pool, VkDeviceSize size, VkDeviceSize alignment, - bool dedicatedAllocation, + bool dedicatedPreferred, VkBuffer dedicatedBuffer, VkBufferUsageFlags dedicatedBufferUsage, VkImage dedicatedImage, const VmaAllocationCreateInfo& createInfo, uint32_t memTypeIndex, VmaSuballocationType suballocType, + VmaDedicatedAllocationList& dedicatedAllocations, + VmaBlockVector& blockVector, size_t allocationCount, VmaAllocation* pAllocations) { @@ -16547,165 +14980,168 @@ VkResult VmaAllocator_T::AllocateMemoryOfType( VMA_DEBUG_LOG(" AllocateMemory: MemoryTypeIndex=%u, AllocationCount=%zu, Size=%llu", memTypeIndex, allocationCount, size); VmaAllocationCreateInfo finalCreateInfo = createInfo; + VkResult res = CalcMemTypeParams( + finalCreateInfo, + memTypeIndex, + size, + allocationCount); + if(res != VK_SUCCESS) + return res; - // If memory type is not HOST_VISIBLE, disable MAPPED. - if((finalCreateInfo.flags & VMA_ALLOCATION_CREATE_MAPPED_BIT) != 0 && - (m_MemProps.memoryTypes[memTypeIndex].propertyFlags & VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT) == 0) + if((finalCreateInfo.flags & VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT) != 0) { - finalCreateInfo.flags &= ~VMA_ALLOCATION_CREATE_MAPPED_BIT; + return AllocateDedicatedMemory( + pool, + size, + suballocType, + dedicatedAllocations, + memTypeIndex, + (finalCreateInfo.flags & VMA_ALLOCATION_CREATE_MAPPED_BIT) != 0, + (finalCreateInfo.flags & VMA_ALLOCATION_CREATE_USER_DATA_COPY_STRING_BIT) != 0, + (finalCreateInfo.flags & VMA_ALLOCATION_CREATE_CAN_ALIAS_BIT) != 0, + finalCreateInfo.pUserData, + finalCreateInfo.priority, + dedicatedBuffer, + dedicatedBufferUsage, + dedicatedImage, + allocationCount, + pAllocations, + blockVector.GetAllocationNextPtr()); } - // If memory is lazily allocated, it should be always dedicated. - if(finalCreateInfo.usage == VMA_MEMORY_USAGE_GPU_LAZILY_ALLOCATED) + else { - finalCreateInfo.flags |= VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT; - } + const bool canAllocateDedicated = + (finalCreateInfo.flags & VMA_ALLOCATION_CREATE_NEVER_ALLOCATE_BIT) == 0 && + (pool == VK_NULL_HANDLE || !blockVector.HasExplicitBlockSize()); - bool isSmallBuffer = dedicatedBuffer != VK_NULL_HANDLE && size <= 4096; // TODO - VmaBlockVector* const blockVector = isSmallBuffer ? m_pSmallBufferBlockVectors[memTypeIndex] : m_pBlockVectors[memTypeIndex]; - VMA_ASSERT(blockVector); + if(canAllocateDedicated) + { + // Heuristics: Allocate dedicated memory if requested size if greater than half of preferred block size. + if(size > blockVector.GetPreferredBlockSize() / 2) + { + dedicatedPreferred = true; + } + // Protection against creating each allocation as dedicated when we reach or exceed heap size/budget, + // which can quickly deplete maxMemoryAllocationCount: Don't prefer dedicated allocations when above + // 3/4 of the maximum allocation count. + if(m_DeviceMemoryCount.load() > m_PhysicalDeviceProperties.limits.maxMemoryAllocationCount * 3 / 4) + { + dedicatedPreferred = false; + } - const VkDeviceSize preferredBlockSize = blockVector->GetPreferredBlockSize(); - bool preferDedicatedMemory = - VMA_DEBUG_ALWAYS_DEDICATED_MEMORY || - dedicatedAllocation || - // Heuristics: Allocate dedicated memory if requested size if greater than half of preferred block size. - size > preferredBlockSize / 2; + if(dedicatedPreferred) + { + res = AllocateDedicatedMemory( + pool, + size, + suballocType, + dedicatedAllocations, + memTypeIndex, + (finalCreateInfo.flags & VMA_ALLOCATION_CREATE_MAPPED_BIT) != 0, + (finalCreateInfo.flags & VMA_ALLOCATION_CREATE_USER_DATA_COPY_STRING_BIT) != 0, + (finalCreateInfo.flags & VMA_ALLOCATION_CREATE_CAN_ALIAS_BIT) != 0, + finalCreateInfo.pUserData, + finalCreateInfo.priority, + dedicatedBuffer, + dedicatedBufferUsage, + dedicatedImage, + allocationCount, + pAllocations, + blockVector.GetAllocationNextPtr()); + if(res == VK_SUCCESS) + { + // Succeeded: AllocateDedicatedMemory function already filld pMemory, nothing more to do here. + VMA_DEBUG_LOG(" Allocated as DedicatedMemory"); + return VK_SUCCESS; + } + } + } - if(preferDedicatedMemory && - (finalCreateInfo.flags & VMA_ALLOCATION_CREATE_NEVER_ALLOCATE_BIT) == 0 && - finalCreateInfo.pool == VK_NULL_HANDLE) - { - finalCreateInfo.flags |= VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT; - } + res = blockVector.Allocate( + size, + alignment, + finalCreateInfo, + suballocType, + allocationCount, + pAllocations); + if(res == VK_SUCCESS) + return VK_SUCCESS; - if((finalCreateInfo.flags & VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT) != 0) - { - if((finalCreateInfo.flags & VMA_ALLOCATION_CREATE_NEVER_ALLOCATE_BIT) != 0) - { - return VK_ERROR_OUT_OF_DEVICE_MEMORY; - } - else + // Try dedicated memory. + if(canAllocateDedicated && !dedicatedPreferred) { - return AllocateDedicatedMemory( + res = AllocateDedicatedMemory( + pool, size, suballocType, + dedicatedAllocations, memTypeIndex, - (finalCreateInfo.flags & VMA_ALLOCATION_CREATE_WITHIN_BUDGET_BIT) != 0, (finalCreateInfo.flags & VMA_ALLOCATION_CREATE_MAPPED_BIT) != 0, (finalCreateInfo.flags & VMA_ALLOCATION_CREATE_USER_DATA_COPY_STRING_BIT) != 0, + (finalCreateInfo.flags & VMA_ALLOCATION_CREATE_CAN_ALIAS_BIT) != 0, finalCreateInfo.pUserData, finalCreateInfo.priority, dedicatedBuffer, dedicatedBufferUsage, dedicatedImage, allocationCount, - pAllocations); - } - } - else - { - VkResult res = blockVector->Allocate( - m_CurrentFrameIndex.load(), - size, - alignment, - finalCreateInfo, - suballocType, - allocationCount, - pAllocations); - if(res == VK_SUCCESS) - { - return res; - } - - // 5. Try dedicated memory. - if((finalCreateInfo.flags & VMA_ALLOCATION_CREATE_NEVER_ALLOCATE_BIT) != 0) - { - return VK_ERROR_OUT_OF_DEVICE_MEMORY; - } - - // Protection against creating each allocation as dedicated when we reach or exceed heap size/budget, - // which can quickly deplete maxMemoryAllocationCount: Don't try dedicated allocations when above - // 3/4 of the maximum allocation count. - if(m_DeviceMemoryCount.load() > m_PhysicalDeviceProperties.limits.maxMemoryAllocationCount * 3 / 4) - { - return VK_ERROR_OUT_OF_DEVICE_MEMORY; - } - - res = AllocateDedicatedMemory( - size, - suballocType, - memTypeIndex, - (finalCreateInfo.flags & VMA_ALLOCATION_CREATE_WITHIN_BUDGET_BIT) != 0, - (finalCreateInfo.flags & VMA_ALLOCATION_CREATE_MAPPED_BIT) != 0, - (finalCreateInfo.flags & VMA_ALLOCATION_CREATE_USER_DATA_COPY_STRING_BIT) != 0, - finalCreateInfo.pUserData, - finalCreateInfo.priority, - dedicatedBuffer, - dedicatedBufferUsage, - dedicatedImage, - allocationCount, - pAllocations); - if(res == VK_SUCCESS) - { - // Succeeded: AllocateDedicatedMemory function already filld pMemory, nothing more to do here. - VMA_DEBUG_LOG(" Allocated as DedicatedMemory"); - return VK_SUCCESS; - } - else - { - // Everything failed: Return error code. - VMA_DEBUG_LOG(" vkAllocateMemory FAILED"); - return res; + pAllocations, + blockVector.GetAllocationNextPtr()); + if(res == VK_SUCCESS) + { + // Succeeded: AllocateDedicatedMemory function already filld pMemory, nothing more to do here. + VMA_DEBUG_LOG(" Allocated as DedicatedMemory"); + return VK_SUCCESS; + } } + // Everything failed: Return error code. + VMA_DEBUG_LOG(" vkAllocateMemory FAILED"); + return res; } } VkResult VmaAllocator_T::AllocateDedicatedMemory( + VmaPool pool, VkDeviceSize size, VmaSuballocationType suballocType, + VmaDedicatedAllocationList& dedicatedAllocations, uint32_t memTypeIndex, - bool withinBudget, bool map, bool isUserDataString, + bool canAliasMemory, void* pUserData, float priority, VkBuffer dedicatedBuffer, VkBufferUsageFlags dedicatedBufferUsage, VkImage dedicatedImage, size_t allocationCount, - VmaAllocation* pAllocations) + VmaAllocation* pAllocations, + const void* pNextChain) { VMA_ASSERT(allocationCount > 0 && pAllocations); - if(withinBudget) - { - const uint32_t heapIndex = MemoryTypeIndexToHeapIndex(memTypeIndex); - VmaBudget heapBudget = {}; - GetBudget(&heapBudget, heapIndex, 1); - if(heapBudget.usage + size * allocationCount > heapBudget.budget) - { - return VK_ERROR_OUT_OF_DEVICE_MEMORY; - } - } - VkMemoryAllocateInfo allocInfo = { VK_STRUCTURE_TYPE_MEMORY_ALLOCATE_INFO }; allocInfo.memoryTypeIndex = memTypeIndex; allocInfo.allocationSize = size; + allocInfo.pNext = pNextChain; #if VMA_DEDICATED_ALLOCATION || VMA_VULKAN_VERSION >= 1001000 VkMemoryDedicatedAllocateInfoKHR dedicatedAllocInfo = { VK_STRUCTURE_TYPE_MEMORY_DEDICATED_ALLOCATE_INFO_KHR }; - if(m_UseKhrDedicatedAllocation || m_VulkanApiVersion >= VK_MAKE_VERSION(1, 1, 0)) + if(!canAliasMemory) { - if(dedicatedBuffer != VK_NULL_HANDLE) - { - VMA_ASSERT(dedicatedImage == VK_NULL_HANDLE); - dedicatedAllocInfo.buffer = dedicatedBuffer; - VmaPnextChainPushFront(&allocInfo, &dedicatedAllocInfo); - } - else if(dedicatedImage != VK_NULL_HANDLE) + if(m_UseKhrDedicatedAllocation || m_VulkanApiVersion >= VK_MAKE_VERSION(1, 1, 0)) { - dedicatedAllocInfo.image = dedicatedImage; - VmaPnextChainPushFront(&allocInfo, &dedicatedAllocInfo); + if(dedicatedBuffer != VK_NULL_HANDLE) + { + VMA_ASSERT(dedicatedImage == VK_NULL_HANDLE); + dedicatedAllocInfo.buffer = dedicatedBuffer; + VmaPnextChainPushFront(&allocInfo, &dedicatedAllocInfo); + } + else if(dedicatedImage != VK_NULL_HANDLE) + { + dedicatedAllocInfo.image = dedicatedImage; + VmaPnextChainPushFront(&allocInfo, &dedicatedAllocInfo); + } } } #endif // #if VMA_DEDICATED_ALLOCATION || VMA_VULKAN_VERSION >= 1001000 @@ -16756,6 +15192,7 @@ VkResult VmaAllocator_T::AllocateDedicatedMemory( for(allocIndex = 0; allocIndex < allocationCount; ++allocIndex) { res = AllocateDedicatedMemoryPage( + pool, size, suballocType, memTypeIndex, @@ -16772,16 +15209,10 @@ VkResult VmaAllocator_T::AllocateDedicatedMemory( if(res == VK_SUCCESS) { - // Register them in m_DedicatedAllocations. + for (allocIndex = 0; allocIndex < allocationCount; ++allocIndex) { - VmaMutexLockWrite lock(m_DedicatedAllocationsMutex[memTypeIndex], m_UseMutex); - DedicatedAllocationLinkedList& dedicatedAllocations = m_DedicatedAllocations[memTypeIndex]; - for(allocIndex = 0; allocIndex < allocationCount; ++allocIndex) - { - dedicatedAllocations.PushBack(pAllocations[allocIndex]); - } + dedicatedAllocations.Register(pAllocations[allocIndex]); } - VMA_DEBUG_LOG(" Allocated DedicatedMemory Count=%zu, MemoryTypeIndex=#%u", allocationCount, memTypeIndex); } else @@ -16815,6 +15246,7 @@ VkResult VmaAllocator_T::AllocateDedicatedMemory( } VkResult VmaAllocator_T::AllocateDedicatedMemoryPage( + VmaPool pool, VkDeviceSize size, VmaSuballocationType suballocType, uint32_t memTypeIndex, @@ -16850,8 +15282,8 @@ VkResult VmaAllocator_T::AllocateDedicatedMemoryPage( } } - *pAllocation = m_AllocationObjectAllocator.Allocate(m_CurrentFrameIndex.load(), isUserDataString); - (*pAllocation)->InitDedicatedAllocation(memTypeIndex, hMemory, suballocType, pMappedData, size); + *pAllocation = m_AllocationObjectAllocator.Allocate(isUserDataString); + (*pAllocation)->InitDedicatedAllocation(pool, memTypeIndex, hMemory, suballocType, pMappedData, size); (*pAllocation)->SetUserData(this, pUserData); m_Budget.AddAllocation(MemoryTypeIndexToHeapIndex(memTypeIndex), size); if(VMA_DEBUG_INITIALIZE_ALLOCATIONS) @@ -16926,6 +15358,78 @@ void VmaAllocator_T::GetImageMemoryRequirements( } } +VkResult VmaAllocator_T::CalcMemTypeParams( + VmaAllocationCreateInfo& inoutCreateInfo, + uint32_t memTypeIndex, + VkDeviceSize size, + size_t allocationCount) +{ + // If memory type is not HOST_VISIBLE, disable MAPPED. + if((inoutCreateInfo.flags & VMA_ALLOCATION_CREATE_MAPPED_BIT) != 0 && + (m_MemProps.memoryTypes[memTypeIndex].propertyFlags & VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT) == 0) + { + inoutCreateInfo.flags &= ~VMA_ALLOCATION_CREATE_MAPPED_BIT; + } + + if((inoutCreateInfo.flags & VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT) != 0 && + (inoutCreateInfo.flags & VMA_ALLOCATION_CREATE_WITHIN_BUDGET_BIT) != 0) + { + const uint32_t heapIndex = MemoryTypeIndexToHeapIndex(memTypeIndex); + VmaBudget heapBudget = {}; + GetHeapBudgets(&heapBudget, heapIndex, 1); + if(heapBudget.usage + size * allocationCount > heapBudget.budget) + { + return VK_ERROR_OUT_OF_DEVICE_MEMORY; + } + } + return VK_SUCCESS; +} + +VkResult VmaAllocator_T::CalcAllocationParams( + VmaAllocationCreateInfo& inoutCreateInfo, + bool dedicatedRequired, + bool dedicatedPreferred) +{ + if(dedicatedRequired || + // If memory is lazily allocated, it should be always dedicated. + inoutCreateInfo.usage == VMA_MEMORY_USAGE_GPU_LAZILY_ALLOCATED) + { + inoutCreateInfo.flags |= VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT; + } + + if(inoutCreateInfo.pool != VK_NULL_HANDLE && (inoutCreateInfo.flags & VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT) != 0) + { + // Assuming here every block has the same block size and priority. + for(uint32_t memTypeIndex = 0; memTypeIndex < GetMemoryTypeCount(); ++memTypeIndex) + { + if(inoutCreateInfo.pool->m_pBlockVectors[memTypeIndex]) + { + if(inoutCreateInfo.pool->m_pBlockVectors[memTypeIndex]->HasExplicitBlockSize()) + { + VMA_ASSERT(0 && "Specifying VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT while current custom pool doesn't support dedicated allocations."); + return VK_ERROR_FEATURE_NOT_PRESENT; + } + inoutCreateInfo.priority = inoutCreateInfo.pool->m_pBlockVectors[memTypeIndex]->GetPriority(); + break; + } + } + } + + if((inoutCreateInfo.flags & VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT) != 0 && + (inoutCreateInfo.flags & VMA_ALLOCATION_CREATE_NEVER_ALLOCATE_BIT) != 0) + { + VMA_ASSERT(0 && "Specifying VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT together with VMA_ALLOCATION_CREATE_NEVER_ALLOCATE_BIT makes no sense."); + return VK_ERROR_FEATURE_NOT_PRESENT; + } + + if(VMA_DEBUG_ALWAYS_DEDICATED_MEMORY && + (inoutCreateInfo.flags & VMA_ALLOCATION_CREATE_NEVER_ALLOCATE_BIT) != 0) + { + inoutCreateInfo.flags |= VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT; + } + return VK_SUCCESS; +} + VkResult VmaAllocator_T::AllocateMemory( const VkMemoryRequirements& vkMemReq, bool requiresDedicatedAllocation, @@ -16944,127 +15448,54 @@ VkResult VmaAllocator_T::AllocateMemory( if(vkMemReq.size == 0) { - return VK_ERROR_VALIDATION_FAILED_EXT; - } - if((createInfo.flags & VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT) != 0 && - (createInfo.flags & VMA_ALLOCATION_CREATE_NEVER_ALLOCATE_BIT) != 0) - { - VMA_ASSERT(0 && "Specifying VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT together with VMA_ALLOCATION_CREATE_NEVER_ALLOCATE_BIT makes no sense."); - return VK_ERROR_OUT_OF_DEVICE_MEMORY; - } - if((createInfo.flags & VMA_ALLOCATION_CREATE_MAPPED_BIT) != 0 && - (createInfo.flags & VMA_ALLOCATION_CREATE_CAN_BECOME_LOST_BIT) != 0) - { - VMA_ASSERT(0 && "Specifying VMA_ALLOCATION_CREATE_MAPPED_BIT together with VMA_ALLOCATION_CREATE_CAN_BECOME_LOST_BIT is invalid."); - return VK_ERROR_OUT_OF_DEVICE_MEMORY; - } - if(requiresDedicatedAllocation) - { - if((createInfo.flags & VMA_ALLOCATION_CREATE_NEVER_ALLOCATE_BIT) != 0) - { - VMA_ASSERT(0 && "VMA_ALLOCATION_CREATE_NEVER_ALLOCATE_BIT specified while dedicated allocation is required."); - return VK_ERROR_OUT_OF_DEVICE_MEMORY; - } - if(createInfo.pool != VK_NULL_HANDLE) - { - VMA_ASSERT(0 && "Pool specified while dedicated allocation is required."); - return VK_ERROR_OUT_OF_DEVICE_MEMORY; - } - } - if((createInfo.pool != VK_NULL_HANDLE) && - ((createInfo.flags & (VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT)) != 0)) - { - VMA_ASSERT(0 && "Specifying VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT when pool != null is invalid."); - return VK_ERROR_OUT_OF_DEVICE_MEMORY; + return VK_ERROR_INITIALIZATION_FAILED; } - if(createInfo.pool != VK_NULL_HANDLE) - { - VmaAllocationCreateInfo createInfoForPool = createInfo; - // If memory type is not HOST_VISIBLE, disable MAPPED. - if((createInfoForPool.flags & VMA_ALLOCATION_CREATE_MAPPED_BIT) != 0 && - (m_MemProps.memoryTypes[createInfo.pool->m_BlockVector.GetMemoryTypeIndex()].propertyFlags & VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT) == 0) - { - createInfoForPool.flags &= ~VMA_ALLOCATION_CREATE_MAPPED_BIT; - } + VmaAllocationCreateInfo createInfoFinal = createInfo; + VkResult res = CalcAllocationParams(createInfoFinal, requiresDedicatedAllocation, prefersDedicatedAllocation); + if(res != VK_SUCCESS) + return res; - return createInfo.pool->m_BlockVector.Allocate( - m_CurrentFrameIndex.load(), + // Bit mask of memory Vulkan types acceptable for this allocation. + uint32_t memoryTypeBits = vkMemReq.memoryTypeBits; + uint32_t memTypeIndex = UINT32_MAX; + res = vmaFindMemoryTypeIndex(this, memoryTypeBits, &createInfoFinal, &memTypeIndex); + // Can't find any single memory type matching requirements. res is VK_ERROR_FEATURE_NOT_PRESENT. + if(res != VK_SUCCESS) + return res; + do + { + VmaBlockVector* blockVector = createInfoFinal.pool == VK_NULL_HANDLE ? m_pBlockVectors[memTypeIndex] : createInfoFinal.pool->m_pBlockVectors[memTypeIndex]; + VMA_ASSERT(blockVector && "Trying to use unsupported memory type!"); + VmaDedicatedAllocationList& dedicatedAllocations = createInfoFinal.pool == VK_NULL_HANDLE ? m_DedicatedAllocations[memTypeIndex] : createInfoFinal.pool->m_DedicatedAllocations[memTypeIndex]; + res = AllocateMemoryOfType( + createInfoFinal.pool, vkMemReq.size, vkMemReq.alignment, - createInfoForPool, + requiresDedicatedAllocation || prefersDedicatedAllocation, + dedicatedBuffer, + dedicatedBufferUsage, + dedicatedImage, + createInfoFinal, + memTypeIndex, suballocType, + dedicatedAllocations, + *blockVector, allocationCount, pAllocations); - } - else - { - // Bit mask of memory Vulkan types acceptable for this allocation. - uint32_t memoryTypeBits = vkMemReq.memoryTypeBits; - uint32_t memTypeIndex = UINT32_MAX; - VkResult res = vmaFindMemoryTypeIndex(this, memoryTypeBits, &createInfo, &memTypeIndex); + // Allocation succeeded if(res == VK_SUCCESS) - { - res = AllocateMemoryOfType( - vkMemReq.size, - vkMemReq.alignment, - requiresDedicatedAllocation || prefersDedicatedAllocation, - dedicatedBuffer, - dedicatedBufferUsage, - dedicatedImage, - createInfo, - memTypeIndex, - suballocType, - allocationCount, - pAllocations); - // Succeeded on first try. - if(res == VK_SUCCESS) - { - return res; - } - // Allocation from this memory type failed. Try other compatible memory types. - else - { - for(;;) - { - // Remove old memTypeIndex from list of possibilities. - memoryTypeBits &= ~(1u << memTypeIndex); - // Find alternative memTypeIndex. - res = vmaFindMemoryTypeIndex(this, memoryTypeBits, &createInfo, &memTypeIndex); - if(res == VK_SUCCESS) - { - res = AllocateMemoryOfType( - vkMemReq.size, - vkMemReq.alignment, - requiresDedicatedAllocation || prefersDedicatedAllocation, - dedicatedBuffer, - dedicatedBufferUsage, - dedicatedImage, - createInfo, - memTypeIndex, - suballocType, - allocationCount, - pAllocations); - // Allocation from this alternative memory type succeeded. - if(res == VK_SUCCESS) - { - return res; - } - // else: Allocation from this memory type failed. Try next one - next loop iteration. - } - // No other matching memory type index could be found. - else - { - // Not returning res, which is VK_ERROR_FEATURE_NOT_PRESENT, because we already failed to allocate once. - return VK_ERROR_OUT_OF_DEVICE_MEMORY; - } - } - } - } - // Can't find any single memory type maching requirements. res is VK_ERROR_FEATURE_NOT_PRESENT. - else - return res; - } + return VK_SUCCESS; + + // Remove old memTypeIndex from list of possibilities. + memoryTypeBits &= ~(1u << memTypeIndex); + // Find alternative memTypeIndex. + res = vmaFindMemoryTypeIndex(this, memoryTypeBits, &createInfoFinal, &memTypeIndex); + } while(res == VK_SUCCESS); + + // No other matching memory type index could be found. + // Not returning res, which is VK_ERROR_FEATURE_NOT_PRESENT, because we already failed to allocate once. + return VK_ERROR_OUT_OF_DEVICE_MEMORY; } void VmaAllocator_T::FreeMemory( @@ -17079,39 +15510,37 @@ void VmaAllocator_T::FreeMemory( if(allocation != VK_NULL_HANDLE) { - if(TouchAllocation(allocation)) + if(VMA_DEBUG_INITIALIZE_ALLOCATIONS) { - if(VMA_DEBUG_INITIALIZE_ALLOCATIONS) - { - FillAllocation(allocation, VMA_ALLOCATION_FILL_PATTERN_DESTROYED); - } + FillAllocation(allocation, VMA_ALLOCATION_FILL_PATTERN_DESTROYED); + } - switch(allocation->GetType()) + switch(allocation->GetType()) + { + case VmaAllocation_T::ALLOCATION_TYPE_BLOCK: { - case VmaAllocation_T::ALLOCATION_TYPE_BLOCK: + VmaBlockVector* pBlockVector = VMA_NULL; + VmaPool hPool = allocation->GetParentPool(); + const uint32_t memTypeIndex = allocation->GetMemoryTypeIndex(); + if(hPool != VK_NULL_HANDLE) { - VmaBlockVector* pBlockVector = VMA_NULL; - VmaPool hPool = allocation->GetBlock()->GetParentPool(); - if(hPool != VK_NULL_HANDLE) - { - pBlockVector = &hPool->m_BlockVector; - } - else - { - pBlockVector = allocation->GetBlock()->GetParentBlockVector(); - } - pBlockVector->Free(allocation); + pBlockVector = hPool->m_pBlockVectors[memTypeIndex]; } - break; - case VmaAllocation_T::ALLOCATION_TYPE_DEDICATED: - FreeDedicatedMemory(allocation); - break; - default: - VMA_ASSERT(0); + else + { + pBlockVector = m_pBlockVectors[memTypeIndex]; + } + VMA_ASSERT(pBlockVector && "Trying to free memory of unsupported type!"); + pBlockVector->Free(allocation); } + break; + case VmaAllocation_T::ALLOCATION_TYPE_DEDICATED: + FreeDedicatedMemory(allocation); + break; + default: + VMA_ASSERT(0); } - // Do this regardless of whether the allocation is lost. Lost allocations still account to Budget.AllocationBytes. m_Budget.RemoveAllocation(MemoryTypeIndexToHeapIndex(allocation->GetMemoryTypeIndex()), allocation->GetSize()); allocation->SetUserData(this, VMA_NULL); m_AllocationObjectAllocator.Free(allocation); @@ -17122,22 +15551,18 @@ void VmaAllocator_T::FreeMemory( void VmaAllocator_T::CalculateStats(VmaStats* pStats) { // Initialize. - InitStatInfo(pStats->total); + VmaInitStatInfo(pStats->total); for(size_t i = 0; i < VK_MAX_MEMORY_TYPES; ++i) - InitStatInfo(pStats->memoryType[i]); + VmaInitStatInfo(pStats->memoryType[i]); for(size_t i = 0; i < VK_MAX_MEMORY_HEAPS; ++i) - InitStatInfo(pStats->memoryHeap[i]); + VmaInitStatInfo(pStats->memoryHeap[i]); // Process default pools. for(uint32_t memTypeIndex = 0; memTypeIndex < GetMemoryTypeCount(); ++memTypeIndex) { VmaBlockVector* const pBlockVector = m_pBlockVectors[memTypeIndex]; - VMA_ASSERT(pBlockVector); - pBlockVector->AddStats(pStats); - - VmaBlockVector* const pSmallBufferBlockVector = m_pSmallBufferBlockVectors[memTypeIndex]; - VMA_ASSERT(pSmallBufferBlockVector); - pSmallBufferBlockVector->AddStats(pStats); + if (pBlockVector != VMA_NULL) + pBlockVector->AddStats(pStats); } // Process custom pools. @@ -17145,7 +15570,17 @@ void VmaAllocator_T::CalculateStats(VmaStats* pStats) VmaMutexLockRead lock(m_PoolsMutex, m_UseMutex); for(VmaPool pool = m_Pools.Front(); pool != VMA_NULL; pool = m_Pools.GetNext(pool)) { - pool->m_BlockVector.AddStats(pStats); + for(uint32_t memTypeIndex = 0; memTypeIndex < GetMemoryTypeCount(); ++memTypeIndex) + { + if (pool->m_pBlockVectors[memTypeIndex]) + { + VmaBlockVector& blockVector = *pool->m_pBlockVectors[memTypeIndex]; + blockVector.AddStats(pStats); + const uint32_t memTypeIndex = blockVector.GetMemoryTypeIndex(); + const uint32_t memHeapIndex = MemoryTypeIndexToHeapIndex(memTypeIndex); + pool->m_DedicatedAllocations[memTypeIndex].AddStats(pStats, memTypeIndex, memHeapIndex); + } + } } } @@ -17153,17 +15588,7 @@ void VmaAllocator_T::CalculateStats(VmaStats* pStats) for(uint32_t memTypeIndex = 0; memTypeIndex < GetMemoryTypeCount(); ++memTypeIndex) { const uint32_t memHeapIndex = MemoryTypeIndexToHeapIndex(memTypeIndex); - VmaMutexLockRead dedicatedAllocationsLock(m_DedicatedAllocationsMutex[memTypeIndex], m_UseMutex); - DedicatedAllocationLinkedList& dedicatedAllocList = m_DedicatedAllocations[memTypeIndex]; - for(VmaAllocation alloc = dedicatedAllocList.Front(); - alloc != VMA_NULL; alloc = dedicatedAllocList.GetNext(alloc)) - { - VmaStatInfo allocationStatInfo; - alloc->DedicatedAllocCalcStatsInfo(allocationStatInfo); - VmaAddStatInfo(pStats->total, allocationStatInfo); - VmaAddStatInfo(pStats->memoryType[memTypeIndex], allocationStatInfo); - VmaAddStatInfo(pStats->memoryHeap[memHeapIndex], allocationStatInfo); - } + m_DedicatedAllocations[memTypeIndex].AddStats(pStats, memTypeIndex, memHeapIndex); } // Postprocess. @@ -17174,7 +15599,7 @@ void VmaAllocator_T::CalculateStats(VmaStats* pStats) VmaPostprocessCalcStatInfo(pStats->memoryHeap[i]); } -void VmaAllocator_T::GetBudget(VmaBudget* outBudget, uint32_t firstHeap, uint32_t heapCount) +void VmaAllocator_T::GetHeapBudgets(VmaBudget* outBudgets, uint32_t firstHeap, uint32_t heapCount) { #if VMA_MEMORY_BUDGET if(m_UseExtMemoryBudget) @@ -17182,52 +15607,50 @@ void VmaAllocator_T::GetBudget(VmaBudget* outBudget, uint32_t firstHeap, uint32_ if(m_Budget.m_OperationsSinceBudgetFetch < 30) { VmaMutexLockRead lockRead(m_Budget.m_BudgetMutex, m_UseMutex); - for(uint32_t i = 0; i < heapCount; ++i, ++outBudget) + for(uint32_t i = 0; i < heapCount; ++i, ++outBudgets) { const uint32_t heapIndex = firstHeap + i; - outBudget->blockBytes = m_Budget.m_BlockBytes[heapIndex]; - outBudget->allocationBytes = m_Budget.m_AllocationBytes[heapIndex]; + outBudgets->blockBytes = m_Budget.m_BlockBytes[heapIndex]; + outBudgets->allocationBytes = m_Budget.m_AllocationBytes[heapIndex]; - if(m_Budget.m_VulkanUsage[heapIndex] + outBudget->blockBytes > m_Budget.m_BlockBytesAtBudgetFetch[heapIndex]) + if(m_Budget.m_VulkanUsage[heapIndex] + outBudgets->blockBytes > m_Budget.m_BlockBytesAtBudgetFetch[heapIndex]) { - outBudget->usage = m_Budget.m_VulkanUsage[heapIndex] + - outBudget->blockBytes - m_Budget.m_BlockBytesAtBudgetFetch[heapIndex]; + outBudgets->usage = m_Budget.m_VulkanUsage[heapIndex] + + outBudgets->blockBytes - m_Budget.m_BlockBytesAtBudgetFetch[heapIndex]; } else { - outBudget->usage = 0; + outBudgets->usage = 0; } // Have to take MIN with heap size because explicit HeapSizeLimit is included in it. - outBudget->budget = VMA_MIN( + outBudgets->budget = VMA_MIN( m_Budget.m_VulkanBudget[heapIndex], m_MemProps.memoryHeaps[heapIndex].size); } } else { UpdateVulkanBudget(); // Outside of mutex lock - GetBudget(outBudget, firstHeap, heapCount); // Recursion + GetHeapBudgets(outBudgets, firstHeap, heapCount); // Recursion } } else #endif { - for(uint32_t i = 0; i < heapCount; ++i, ++outBudget) + for(uint32_t i = 0; i < heapCount; ++i, ++outBudgets) { const uint32_t heapIndex = firstHeap + i; - outBudget->blockBytes = m_Budget.m_BlockBytes[heapIndex]; - outBudget->allocationBytes = m_Budget.m_AllocationBytes[heapIndex]; + outBudgets->blockBytes = m_Budget.m_BlockBytes[heapIndex]; + outBudgets->allocationBytes = m_Budget.m_AllocationBytes[heapIndex]; - outBudget->usage = outBudget->blockBytes; - outBudget->budget = m_MemProps.memoryHeaps[heapIndex].size * 8 / 10; // 80% heuristics. + outBudgets->usage = outBudgets->blockBytes; + outBudgets->budget = m_MemProps.memoryHeaps[heapIndex].size * 8 / 10; // 80% heuristics. } } } -static const uint32_t VMA_VENDOR_ID_AMD = 4098; - VkResult VmaAllocator_T::DefragmentationBegin( const VmaDefragmentationInfo2& info, VmaDefragmentationStats* pStats, @@ -17239,7 +15662,7 @@ VkResult VmaAllocator_T::DefragmentationBegin( } *pContext = vma_new(this, VmaDefragmentationContext_T)( - this, m_CurrentFrameIndex.load(), info.flags, pStats); + this, info.flags, pStats); (*pContext)->AddPools(info.poolCount, info.pPools); (*pContext)->AddAllocations( @@ -17272,135 +15695,21 @@ VkResult VmaAllocator_T::DefragmentationPassBegin( { return context->DefragmentPassBegin(pInfo); } + VkResult VmaAllocator_T::DefragmentationPassEnd( VmaDefragmentationContext context) { return context->DefragmentPassEnd(); - } void VmaAllocator_T::GetAllocationInfo(VmaAllocation hAllocation, VmaAllocationInfo* pAllocationInfo) { - if(hAllocation->CanBecomeLost()) - { - /* - Warning: This is a carefully designed algorithm. - Do not modify unless you really know what you're doing :) - */ - const uint32_t localCurrFrameIndex = m_CurrentFrameIndex.load(); - uint32_t localLastUseFrameIndex = hAllocation->GetLastUseFrameIndex(); - for(;;) - { - if(localLastUseFrameIndex == VMA_FRAME_INDEX_LOST) - { - pAllocationInfo->memoryType = UINT32_MAX; - pAllocationInfo->deviceMemory = VK_NULL_HANDLE; - pAllocationInfo->offset = 0; - pAllocationInfo->size = hAllocation->GetSize(); - pAllocationInfo->pMappedData = VMA_NULL; - pAllocationInfo->pUserData = hAllocation->GetUserData(); - return; - } - else if(localLastUseFrameIndex == localCurrFrameIndex) - { - pAllocationInfo->memoryType = hAllocation->GetMemoryTypeIndex(); - pAllocationInfo->deviceMemory = hAllocation->GetMemory(); - pAllocationInfo->offset = hAllocation->GetOffset(); - pAllocationInfo->size = hAllocation->GetSize(); - pAllocationInfo->pMappedData = VMA_NULL; - pAllocationInfo->pUserData = hAllocation->GetUserData(); - return; - } - else // Last use time earlier than current time. - { - if(hAllocation->CompareExchangeLastUseFrameIndex(localLastUseFrameIndex, localCurrFrameIndex)) - { - localLastUseFrameIndex = localCurrFrameIndex; - } - } - } - } - else - { -#if VMA_STATS_STRING_ENABLED - uint32_t localCurrFrameIndex = m_CurrentFrameIndex.load(); - uint32_t localLastUseFrameIndex = hAllocation->GetLastUseFrameIndex(); - for(;;) - { - VMA_ASSERT(localLastUseFrameIndex != VMA_FRAME_INDEX_LOST); - if(localLastUseFrameIndex == localCurrFrameIndex) - { - break; - } - else // Last use time earlier than current time. - { - if(hAllocation->CompareExchangeLastUseFrameIndex(localLastUseFrameIndex, localCurrFrameIndex)) - { - localLastUseFrameIndex = localCurrFrameIndex; - } - } - } -#endif - - pAllocationInfo->memoryType = hAllocation->GetMemoryTypeIndex(); - pAllocationInfo->deviceMemory = hAllocation->GetMemory(); - pAllocationInfo->offset = hAllocation->GetOffset(); - pAllocationInfo->size = hAllocation->GetSize(); - pAllocationInfo->pMappedData = hAllocation->GetMappedData(); - pAllocationInfo->pUserData = hAllocation->GetUserData(); - } -} - -bool VmaAllocator_T::TouchAllocation(VmaAllocation hAllocation) -{ - // This is a stripped-down version of VmaAllocator_T::GetAllocationInfo. - if(hAllocation->CanBecomeLost()) - { - uint32_t localCurrFrameIndex = m_CurrentFrameIndex.load(); - uint32_t localLastUseFrameIndex = hAllocation->GetLastUseFrameIndex(); - for(;;) - { - if(localLastUseFrameIndex == VMA_FRAME_INDEX_LOST) - { - return false; - } - else if(localLastUseFrameIndex == localCurrFrameIndex) - { - return true; - } - else // Last use time earlier than current time. - { - if(hAllocation->CompareExchangeLastUseFrameIndex(localLastUseFrameIndex, localCurrFrameIndex)) - { - localLastUseFrameIndex = localCurrFrameIndex; - } - } - } - } - else - { -#if VMA_STATS_STRING_ENABLED - uint32_t localCurrFrameIndex = m_CurrentFrameIndex.load(); - uint32_t localLastUseFrameIndex = hAllocation->GetLastUseFrameIndex(); - for(;;) - { - VMA_ASSERT(localLastUseFrameIndex != VMA_FRAME_INDEX_LOST); - if(localLastUseFrameIndex == localCurrFrameIndex) - { - break; - } - else // Last use time earlier than current time. - { - if(hAllocation->CompareExchangeLastUseFrameIndex(localLastUseFrameIndex, localCurrFrameIndex)) - { - localLastUseFrameIndex = localCurrFrameIndex; - } - } - } -#endif - - return true; - } + pAllocationInfo->memoryType = hAllocation->GetMemoryTypeIndex(); + pAllocationInfo->deviceMemory = hAllocation->GetMemory(); + pAllocationInfo->offset = hAllocation->GetOffset(); + pAllocationInfo->size = hAllocation->GetSize(); + pAllocationInfo->pMappedData = hAllocation->GetMappedData(); + pAllocationInfo->pUserData = hAllocation->GetUserData(); } VkResult VmaAllocator_T::CreatePool(const VmaPoolCreateInfo* pCreateInfo, VmaPool* pPool) @@ -17423,27 +15732,26 @@ VkResult VmaAllocator_T::CreatePool(const VmaPoolCreateInfo* pCreateInfo, VmaPoo { return VK_ERROR_INITIALIZATION_FAILED; } - // Memory type index out of range or forbidden. - if(pCreateInfo->memoryTypeIndex >= GetMemoryTypeCount() || - ((1u << pCreateInfo->memoryTypeIndex) & m_GlobalMemoryTypeBits) == 0) - { - return VK_ERROR_FEATURE_NOT_PRESENT; - } if(newCreateInfo.minAllocationAlignment > 0) { VMA_ASSERT(VmaIsPow2(newCreateInfo.minAllocationAlignment)); } - const VkDeviceSize preferredBlockSize = CalcPreferredBlockSize(newCreateInfo.memoryTypeIndex); - - *pPool = vma_new(this, VmaPool_T)(this, newCreateInfo, preferredBlockSize); + *pPool = vma_new(this, VmaPool_T)(this, newCreateInfo); - VkResult res = (*pPool)->m_BlockVector.CreateMinBlocks(); - if(res != VK_SUCCESS) + for(uint32_t memTypeIndex = 0; memTypeIndex < GetMemoryTypeCount(); ++memTypeIndex) { - vma_delete(this, *pPool); - *pPool = VMA_NULL; - return res; + // Create only supported types + if((m_GlobalMemoryTypeBits & (1u << memTypeIndex)) != 0) + { + VkResult res = (*pPool)->m_pBlockVectors[memTypeIndex]->CreateMinBlocks(); + if(res != VK_SUCCESS) + { + vma_delete(this, *pPool); + *pPool = VMA_NULL; + return res; + } + } } // Add to m_Pools. @@ -17469,7 +15777,20 @@ void VmaAllocator_T::DestroyPool(VmaPool pool) void VmaAllocator_T::GetPoolStats(VmaPool pool, VmaPoolStats* pPoolStats) { - pool->m_BlockVector.GetPoolStats(pPoolStats); + pPoolStats->size = 0; + pPoolStats->unusedSize = 0; + pPoolStats->allocationCount = 0; + pPoolStats->unusedRangeCount = 0; + pPoolStats->blockCount = 0; + + for(uint32_t memTypeIndex = 0; memTypeIndex < GetMemoryTypeCount(); ++memTypeIndex) + { + if((m_GlobalMemoryTypeBits & (1u << memTypeIndex)) != 0) + { + pool->m_pBlockVectors[memTypeIndex]->AddPoolStats(pPoolStats); + pool->m_DedicatedAllocations[memTypeIndex].AddPoolStats(pPoolStats); + } + } } void VmaAllocator_T::SetCurrentFrameIndex(uint32_t frameIndex) @@ -17484,18 +15805,15 @@ void VmaAllocator_T::SetCurrentFrameIndex(uint32_t frameIndex) #endif // #if VMA_MEMORY_BUDGET } -void VmaAllocator_T::MakePoolAllocationsLost( - VmaPool hPool, - size_t* pLostAllocationCount) -{ - hPool->m_BlockVector.MakePoolAllocationsLost( - m_CurrentFrameIndex.load(), - pLostAllocationCount); -} - VkResult VmaAllocator_T::CheckPoolCorruption(VmaPool hPool) { - return hPool->m_BlockVector.CheckCorruption(); + for(uint32_t memTypeIndex = 0; memTypeIndex < GetMemoryTypeCount(); ++memTypeIndex) + { + if((m_GlobalMemoryTypeBits & (1u << memTypeIndex)) != 0) + { + return hPool->m_pBlockVectors[memTypeIndex]->CheckCorruption(); + } + } } VkResult VmaAllocator_T::CheckCorruption(uint32_t memoryTypeBits) @@ -17505,10 +15823,9 @@ VkResult VmaAllocator_T::CheckCorruption(uint32_t memoryTypeBits) // Process default pools. for(uint32_t memTypeIndex = 0; memTypeIndex < GetMemoryTypeCount(); ++memTypeIndex) { - if(((1u << memTypeIndex) & memoryTypeBits) != 0) + VmaBlockVector* const pBlockVector = m_pBlockVectors[memTypeIndex]; + if(pBlockVector != VMA_NULL) { - VmaBlockVector* const pBlockVector = m_pBlockVectors[memTypeIndex]; - VMA_ASSERT(pBlockVector); VkResult localRes = pBlockVector->CheckCorruption(); switch(localRes) { @@ -17528,18 +15845,21 @@ VkResult VmaAllocator_T::CheckCorruption(uint32_t memoryTypeBits) VmaMutexLockRead lock(m_PoolsMutex, m_UseMutex); for(VmaPool pool = m_Pools.Front(); pool != VMA_NULL; pool = m_Pools.GetNext(pool)) { - if(((1u << pool->m_BlockVector.GetMemoryTypeIndex()) & memoryTypeBits) != 0) + for(uint32_t memTypeIndex = 0; memTypeIndex < GetMemoryTypeCount(); ++memTypeIndex) { - VkResult localRes = pool->m_BlockVector.CheckCorruption(); - switch(localRes) + if(pool->m_pBlockVectors[memTypeIndex] && ((1u << memTypeIndex) & memoryTypeBits) != 0) { - case VK_ERROR_FEATURE_NOT_PRESENT: - break; - case VK_SUCCESS: - finalRes = VK_SUCCESS; - break; - default: - return localRes; + VkResult localRes = pool->m_pBlockVectors[memTypeIndex]->CheckCorruption(); + switch(localRes) + { + case VK_ERROR_FEATURE_NOT_PRESENT: + break; + case VK_SUCCESS: + finalRes = VK_SUCCESS; + break; + default: + return localRes; + } } } } @@ -17548,37 +15868,6 @@ VkResult VmaAllocator_T::CheckCorruption(uint32_t memoryTypeBits) return finalRes; } -void VmaAllocator_T::CreateLostAllocation(VmaAllocation* pAllocation) -{ - *pAllocation = m_AllocationObjectAllocator.Allocate(VMA_FRAME_INDEX_LOST, false); - (*pAllocation)->InitLost(); -} - -// An object that increments given atomic but decrements it back in the destructor unless Commit() is called. -template<typename T> -struct AtomicTransactionalIncrement -{ -public: - typedef std::atomic<T> AtomicT; - ~AtomicTransactionalIncrement() - { - if(m_Atomic) - --(*m_Atomic); - } - T Increment(AtomicT* atomic) - { - m_Atomic = atomic; - return m_Atomic->fetch_add(1); - } - void Commit() - { - m_Atomic = nullptr; - } - -private: - AtomicT* m_Atomic = nullptr; -}; - VkResult VmaAllocator_T::AllocateVulkanMemory(const VkMemoryAllocateInfo* pAllocateInfo, VkDeviceMemory* pMemory) { AtomicTransactionalIncrement<uint32_t> deviceMemoryCountIncrement; @@ -17720,11 +16009,6 @@ VkResult VmaAllocator_T::BindVulkanImage( VkResult VmaAllocator_T::Map(VmaAllocation hAllocation, void** ppData) { - if(hAllocation->CanBecomeLost()) - { - return VK_ERROR_MEMORY_MAP_FAILED; - } - switch(hAllocation->GetType()) { case VmaAllocation_T::ALLOCATION_TYPE_BLOCK: @@ -17781,7 +16065,7 @@ VkResult VmaAllocator_T::BindBufferMemory( case VmaAllocation_T::ALLOCATION_TYPE_BLOCK: { VmaDeviceMemoryBlock* const pBlock = hAllocation->GetBlock(); - VMA_ASSERT(pBlock && "Binding buffer to allocation that doesn't belong to any block. Is the allocation lost?"); + VMA_ASSERT(pBlock && "Binding buffer to allocation that doesn't belong to any block."); res = pBlock->BindBufferMemory(this, hAllocation, allocationLocalOffset, hBuffer, pNext); break; } @@ -17806,7 +16090,7 @@ VkResult VmaAllocator_T::BindImageMemory( case VmaAllocation_T::ALLOCATION_TYPE_BLOCK: { VmaDeviceMemoryBlock* pBlock = hAllocation->GetBlock(); - VMA_ASSERT(pBlock && "Binding image to allocation that doesn't belong to any block. Is the allocation lost?"); + VMA_ASSERT(pBlock && "Binding image to allocation that doesn't belong to any block."); res = pBlock->BindImageMemory(this, hAllocation, allocationLocalOffset, hImage, pNext); break; } @@ -17888,10 +16172,16 @@ void VmaAllocator_T::FreeDedicatedMemory(const VmaAllocation allocation) VMA_ASSERT(allocation && allocation->GetType() == VmaAllocation_T::ALLOCATION_TYPE_DEDICATED); const uint32_t memTypeIndex = allocation->GetMemoryTypeIndex(); + VmaPool parentPool = allocation->GetParentPool(); + if(parentPool == VK_NULL_HANDLE) { - VmaMutexLockWrite lock(m_DedicatedAllocationsMutex[memTypeIndex], m_UseMutex); - DedicatedAllocationLinkedList& dedicatedAllocations = m_DedicatedAllocations[memTypeIndex]; - dedicatedAllocations.Remove(allocation); + // Default pool + m_DedicatedAllocations[memTypeIndex].Unregister(allocation); + } + else + { + // Custom pool + parentPool->m_DedicatedAllocations[memTypeIndex].Unregister(allocation); } VkDeviceMemory hMemory = allocation->GetMemory(); @@ -18022,7 +16312,6 @@ bool VmaAllocator_T::GetFlushOrInvalidateRange( } #if VMA_MEMORY_BUDGET - void VmaAllocator_T::UpdateVulkanBudget() { VMA_ASSERT(m_UseExtMemoryBudget); @@ -18060,13 +16349,11 @@ void VmaAllocator_T::UpdateVulkanBudget() m_Budget.m_OperationsSinceBudgetFetch = 0; } } - -#endif // #if VMA_MEMORY_BUDGET +#endif // VMA_MEMORY_BUDGET void VmaAllocator_T::FillAllocation(const VmaAllocation hAllocation, uint8_t pattern) { if(VMA_DEBUG_INITIALIZE_ALLOCATIONS && - !hAllocation->CanBecomeLost() && (m_MemProps.memoryTypes[hAllocation->GetMemoryTypeIndex()].propertyFlags & VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT) != 0) { void* pData = VMA_NULL; @@ -18096,14 +16383,12 @@ uint32_t VmaAllocator_T::GetGpuDefragmentationMemoryTypeBits() } #if VMA_STATS_STRING_ENABLED - void VmaAllocator_T::PrintDetailedMap(VmaJsonWriter& json) { bool dedicatedAllocationsStarted = false; for(uint32_t memTypeIndex = 0; memTypeIndex < GetMemoryTypeCount(); ++memTypeIndex) { - VmaMutexLockRead dedicatedAllocationsLock(m_DedicatedAllocationsMutex[memTypeIndex], m_UseMutex); - DedicatedAllocationLinkedList& dedicatedAllocList = m_DedicatedAllocations[memTypeIndex]; + VmaDedicatedAllocationList& dedicatedAllocList = m_DedicatedAllocations[memTypeIndex]; if(!dedicatedAllocList.IsEmpty()) { if(dedicatedAllocationsStarted == false) @@ -18117,17 +16402,7 @@ void VmaAllocator_T::PrintDetailedMap(VmaJsonWriter& json) json.ContinueString(memTypeIndex); json.EndString(); - json.BeginArray(); - - for(VmaAllocation alloc = dedicatedAllocList.Front(); - alloc != VMA_NULL; alloc = dedicatedAllocList.GetNext(alloc)) - { - json.BeginObject(true); - alloc->PrintParameters(json); - json.EndObject(); - } - - json.EndArray(); + dedicatedAllocList.BuildStatsString(json); } } if(dedicatedAllocationsStarted) @@ -18135,52 +16410,30 @@ void VmaAllocator_T::PrintDetailedMap(VmaJsonWriter& json) json.EndObject(); } - // Default pools { bool allocationsStarted = false; for(uint32_t memTypeIndex = 0; memTypeIndex < GetMemoryTypeCount(); ++memTypeIndex) { - if(m_pBlockVectors[memTypeIndex]->IsEmpty() == false) + VmaBlockVector* pBlockVector = m_pBlockVectors[memTypeIndex]; + if(pBlockVector != VMA_NULL) { - if(allocationsStarted == false) + if (pBlockVector->IsEmpty() == false) { - allocationsStarted = true; - json.WriteString("DefaultPools"); - json.BeginObject(); - } - - json.BeginString("Type "); - json.ContinueString(memTypeIndex); - json.EndString(); + if (allocationsStarted == false) + { + allocationsStarted = true; + json.WriteString("DefaultPools"); + json.BeginObject(); + } - m_pBlockVectors[memTypeIndex]->PrintDetailedMap(json); - } - } - if(allocationsStarted) - { - json.EndObject(); - } - } + json.BeginString("Type "); + json.ContinueString(memTypeIndex); + json.EndString(); - // Small buffer pools - { - bool allocationsStarted = false; - for(uint32_t memTypeIndex = 0; memTypeIndex < GetMemoryTypeCount(); ++memTypeIndex) - { - if(m_pSmallBufferBlockVectors[memTypeIndex]->IsEmpty() == false) - { - if(allocationsStarted == false) - { - allocationsStarted = true; - json.WriteString("SmallBufferPools"); json.BeginObject(); + pBlockVector->PrintDetailedMap(json); + json.EndObject(); } - - json.BeginString("Type "); - json.ContinueString(memTypeIndex); - json.EndString(); - - m_pSmallBufferBlockVectors[memTypeIndex]->PrintDetailedMap(json); } } if(allocationsStarted) @@ -18202,28 +16455,47 @@ void VmaAllocator_T::PrintDetailedMap(VmaJsonWriter& json) json.ContinueString(pool->GetId()); json.EndString(); - pool->m_BlockVector.PrintDetailedMap(json); + json.BeginObject(); + for(uint32_t memTypeIndex = 0; memTypeIndex < GetMemoryTypeCount(); ++memTypeIndex) + { + if (pool->m_pBlockVectors[memTypeIndex]) + { + pool->m_pBlockVectors[memTypeIndex]->PrintDetailedMap(json); + } + + if (!pool->m_DedicatedAllocations[memTypeIndex].IsEmpty()) + { + json.WriteString("DedicatedAllocations"); + pool->m_DedicatedAllocations->BuildStatsString(json); + } + } + json.EndObject(); } json.EndObject(); } } } +#endif // VMA_STATS_STRING_ENABLED +#endif // _VMA_ALLOCATOR_T_FUNCTIONS -#endif // #if VMA_STATS_STRING_ENABLED - -//////////////////////////////////////////////////////////////////////////////// -// Public interface +#ifndef _VMA_PUBLIC_INTERFACE VMA_CALL_PRE VkResult VMA_CALL_POST vmaCreateAllocator( const VmaAllocatorCreateInfo* pCreateInfo, VmaAllocator* pAllocator) { VMA_ASSERT(pCreateInfo && pAllocator); VMA_ASSERT(pCreateInfo->vulkanApiVersion == 0 || - (VK_VERSION_MAJOR(pCreateInfo->vulkanApiVersion) == 1 && VK_VERSION_MINOR(pCreateInfo->vulkanApiVersion) <= 2)); + (VK_VERSION_MAJOR(pCreateInfo->vulkanApiVersion) == 1 && VK_VERSION_MINOR(pCreateInfo->vulkanApiVersion) <= 3)); VMA_DEBUG_LOG("vmaCreateAllocator"); *pAllocator = vma_new(pCreateInfo->pAllocationCallbacks, VmaAllocator_T)(pCreateInfo); - return (*pAllocator)->Init(pCreateInfo); + VkResult result = (*pAllocator)->Init(pCreateInfo); + if(result < 0) + { + vma_delete(pCreateInfo->pAllocationCallbacks, *pAllocator); + *pAllocator = VK_NULL_HANDLE; + } + return result; } VMA_CALL_PRE void VMA_CALL_POST vmaDestroyAllocator( @@ -18232,7 +16504,7 @@ VMA_CALL_PRE void VMA_CALL_POST vmaDestroyAllocator( if(allocator != VK_NULL_HANDLE) { VMA_DEBUG_LOG("vmaDestroyAllocator"); - VkAllocationCallbacks allocationCallbacks = allocator->m_AllocationCallbacks; + VkAllocationCallbacks allocationCallbacks = allocator->m_AllocationCallbacks; // Have to copy the callbacks when destroying. vma_delete(&allocationCallbacks, allocator); } } @@ -18276,7 +16548,6 @@ VMA_CALL_PRE void VMA_CALL_POST vmaSetCurrentFrameIndex( uint32_t frameIndex) { VMA_ASSERT(allocator); - VMA_ASSERT(frameIndex != VMA_FRAME_INDEX_LOST); VMA_DEBUG_GLOBAL_MUTEX_LOCK @@ -18292,13 +16563,13 @@ VMA_CALL_PRE void VMA_CALL_POST vmaCalculateStats( allocator->CalculateStats(pStats); } -VMA_CALL_PRE void VMA_CALL_POST vmaGetBudget( +VMA_CALL_PRE void VMA_CALL_POST vmaGetHeapBudgets( VmaAllocator allocator, - VmaBudget* pBudget) + VmaBudget* pBudgets) { - VMA_ASSERT(allocator && pBudget); + VMA_ASSERT(allocator && pBudgets); VMA_DEBUG_GLOBAL_MUTEX_LOCK - allocator->GetBudget(pBudget, 0, allocator->GetMemoryHeapCount()); + allocator->GetHeapBudgets(pBudgets, 0, allocator->GetMemoryHeapCount()); } #if VMA_STATS_STRING_ENABLED @@ -18311,13 +16582,13 @@ VMA_CALL_PRE void VMA_CALL_POST vmaBuildStatsString( VMA_ASSERT(allocator && ppStatsString); VMA_DEBUG_GLOBAL_MUTEX_LOCK - VmaStringBuilder sb(allocator); + VmaStringBuilder sb(allocator->GetAllocationCallbacks()); { VmaJsonWriter json(allocator->GetAllocationCallbacks(), sb); json.BeginObject(); - VmaBudget budget[VK_MAX_MEMORY_HEAPS]; - allocator->GetBudget(budget, 0, allocator->GetMemoryHeapCount()); + VmaBudget budgets[VK_MAX_MEMORY_HEAPS]; + allocator->GetHeapBudgets(budgets, 0, allocator->GetMemoryHeapCount()); VmaStats stats; allocator->CalculateStats(&stats); @@ -18347,13 +16618,13 @@ VMA_CALL_PRE void VMA_CALL_POST vmaBuildStatsString( json.BeginObject(); { json.WriteString("BlockBytes"); - json.WriteNumber(budget[heapIndex].blockBytes); + json.WriteNumber(budgets[heapIndex].blockBytes); json.WriteString("AllocationBytes"); - json.WriteNumber(budget[heapIndex].allocationBytes); + json.WriteNumber(budgets[heapIndex].allocationBytes); json.WriteString("Usage"); - json.WriteNumber(budget[heapIndex].usage); + json.WriteNumber(budgets[heapIndex].usage); json.WriteString("Budget"); - json.WriteNumber(budget[heapIndex].budget); + json.WriteNumber(budgets[heapIndex].budget); } json.EndObject(); @@ -18434,14 +16705,7 @@ VMA_CALL_PRE void VMA_CALL_POST vmaBuildStatsString( json.EndObject(); } - const size_t len = sb.GetLength(); - char* const pChars = vma_new_array(allocator, char, len + 1); - if(len > 0) - { - memcpy(pChars, sb.GetData(), len); - } - pChars[len] = '\0'; - *ppStatsString = pChars; + *ppStatsString = VmaCreateStringCopy(allocator->GetAllocationCallbacks(), sb.GetData(), sb.GetLength()); } VMA_CALL_PRE void VMA_CALL_POST vmaFreeStatsString( @@ -18451,12 +16715,11 @@ VMA_CALL_PRE void VMA_CALL_POST vmaFreeStatsString( if(pStatsString != VMA_NULL) { VMA_ASSERT(allocator); - size_t len = strlen(pStatsString); - vma_delete_array(allocator, pStatsString, len + 1); + VmaFreeString(allocator->GetAllocationCallbacks(), pStatsString); } } -#endif // #if VMA_STATS_STRING_ENABLED +#endif // VMA_STATS_STRING_ENABLED /* This function is not protected by any mutex because it just reads immutable data. @@ -18571,12 +16834,13 @@ VMA_CALL_PRE VkResult VMA_CALL_POST vmaFindMemoryTypeIndexForBufferInfo( const VkDevice hDev = allocator->m_hDevice; VkBuffer hBuffer = VK_NULL_HANDLE; - VkResult res = allocator->GetVulkanFunctions().vkCreateBuffer( + const VmaVulkanFunctions* funcs = &allocator->GetVulkanFunctions(); + VkResult res = funcs->vkCreateBuffer( hDev, pBufferCreateInfo, allocator->GetAllocationCallbacks(), &hBuffer); if(res == VK_SUCCESS) { VkMemoryRequirements memReq = {}; - allocator->GetVulkanFunctions().vkGetBufferMemoryRequirements( + funcs->vkGetBufferMemoryRequirements( hDev, hBuffer, &memReq); res = vmaFindMemoryTypeIndex( @@ -18585,7 +16849,7 @@ VMA_CALL_PRE VkResult VMA_CALL_POST vmaFindMemoryTypeIndexForBufferInfo( pAllocationCreateInfo, pMemoryTypeIndex); - allocator->GetVulkanFunctions().vkDestroyBuffer( + funcs->vkDestroyBuffer( hDev, hBuffer, allocator->GetAllocationCallbacks()); } return res; @@ -18604,12 +16868,13 @@ VMA_CALL_PRE VkResult VMA_CALL_POST vmaFindMemoryTypeIndexForImageInfo( const VkDevice hDev = allocator->m_hDevice; VkImage hImage = VK_NULL_HANDLE; - VkResult res = allocator->GetVulkanFunctions().vkCreateImage( + const VmaVulkanFunctions* funcs = &allocator->GetVulkanFunctions(); + VkResult res = funcs->vkCreateImage( hDev, pImageCreateInfo, allocator->GetAllocationCallbacks(), &hImage); if(res == VK_SUCCESS) { VkMemoryRequirements memReq = {}; - allocator->GetVulkanFunctions().vkGetImageMemoryRequirements( + funcs->vkGetImageMemoryRequirements( hDev, hImage, &memReq); res = vmaFindMemoryTypeIndex( @@ -18618,7 +16883,7 @@ VMA_CALL_PRE VkResult VMA_CALL_POST vmaFindMemoryTypeIndexForImageInfo( pAllocationCreateInfo, pMemoryTypeIndex); - allocator->GetVulkanFunctions().vkDestroyImage( + funcs->vkDestroyImage( hDev, hImage, allocator->GetAllocationCallbacks()); } return res; @@ -18635,16 +16900,7 @@ VMA_CALL_PRE VkResult VMA_CALL_POST vmaCreatePool( VMA_DEBUG_GLOBAL_MUTEX_LOCK - VkResult res = allocator->CreatePool(pCreateInfo, pPool); - -#if VMA_RECORDING_ENABLED - if(allocator->GetRecorder() != VMA_NULL) - { - allocator->GetRecorder()->RecordCreatePool(allocator->GetCurrentFrameIndex(), *pCreateInfo, *pPool); - } -#endif - - return res; + return allocator->CreatePool(pCreateInfo, pPool); } VMA_CALL_PRE void VMA_CALL_POST vmaDestroyPool( @@ -18662,13 +16918,6 @@ VMA_CALL_PRE void VMA_CALL_POST vmaDestroyPool( VMA_DEBUG_GLOBAL_MUTEX_LOCK -#if VMA_RECORDING_ENABLED - if(allocator->GetRecorder() != VMA_NULL) - { - allocator->GetRecorder()->RecordDestroyPool(allocator->GetCurrentFrameIndex(), pool); - } -#endif - allocator->DestroyPool(pool); } @@ -18684,25 +16933,6 @@ VMA_CALL_PRE void VMA_CALL_POST vmaGetPoolStats( allocator->GetPoolStats(pool, pPoolStats); } -VMA_CALL_PRE void VMA_CALL_POST vmaMakePoolAllocationsLost( - VmaAllocator allocator, - VmaPool pool, - size_t* pLostAllocationCount) -{ - VMA_ASSERT(allocator && pool); - - VMA_DEBUG_GLOBAL_MUTEX_LOCK - -#if VMA_RECORDING_ENABLED - if(allocator->GetRecorder() != VMA_NULL) - { - allocator->GetRecorder()->RecordMakePoolAllocationsLost(allocator->GetCurrentFrameIndex(), pool); - } -#endif - - allocator->MakePoolAllocationsLost(pool, pLostAllocationCount); -} - VMA_CALL_PRE VkResult VMA_CALL_POST vmaCheckPoolCorruption(VmaAllocator allocator, VmaPool pool) { VMA_ASSERT(allocator && pool); @@ -18740,13 +16970,6 @@ VMA_CALL_PRE void VMA_CALL_POST vmaSetPoolName( VMA_DEBUG_GLOBAL_MUTEX_LOCK pool->SetName(pName); - -#if VMA_RECORDING_ENABLED - if(allocator->GetRecorder() != VMA_NULL) - { - allocator->GetRecorder()->RecordSetPoolName(allocator->GetCurrentFrameIndex(), pool, pName); - } -#endif } VMA_CALL_PRE VkResult VMA_CALL_POST vmaAllocateMemory( @@ -18774,17 +16997,6 @@ VMA_CALL_PRE VkResult VMA_CALL_POST vmaAllocateMemory( 1, // allocationCount pAllocation); -#if VMA_RECORDING_ENABLED - if(allocator->GetRecorder() != VMA_NULL) - { - allocator->GetRecorder()->RecordAllocateMemory( - allocator->GetCurrentFrameIndex(), - *pVkMemoryRequirements, - *pCreateInfo, - *pAllocation); - } -#endif - if(pAllocationInfo != VMA_NULL && result == VK_SUCCESS) { allocator->GetAllocationInfo(*pAllocation, pAllocationInfo); @@ -18824,18 +17036,6 @@ VMA_CALL_PRE VkResult VMA_CALL_POST vmaAllocateMemoryPages( allocationCount, pAllocations); -#if VMA_RECORDING_ENABLED - if(allocator->GetRecorder() != VMA_NULL) - { - allocator->GetRecorder()->RecordAllocateMemoryPages( - allocator->GetCurrentFrameIndex(), - *pVkMemoryRequirements, - *pCreateInfo, - (uint64_t)allocationCount, - pAllocations); - } -#endif - if(pAllocationInfo != VMA_NULL && result == VK_SUCCESS) { for(size_t i = 0; i < allocationCount; ++i) @@ -18879,19 +17079,6 @@ VMA_CALL_PRE VkResult VMA_CALL_POST vmaAllocateMemoryForBuffer( 1, // allocationCount pAllocation); -#if VMA_RECORDING_ENABLED - if(allocator->GetRecorder() != VMA_NULL) - { - allocator->GetRecorder()->RecordAllocateMemoryForBuffer( - allocator->GetCurrentFrameIndex(), - vkMemReq, - requiresDedicatedAllocation, - prefersDedicatedAllocation, - *pCreateInfo, - *pAllocation); - } -#endif - if(pAllocationInfo && result == VK_SUCCESS) { allocator->GetAllocationInfo(*pAllocation, pAllocationInfo); @@ -18931,19 +17118,6 @@ VMA_CALL_PRE VkResult VMA_CALL_POST vmaAllocateMemoryForImage( 1, // allocationCount pAllocation); -#if VMA_RECORDING_ENABLED - if(allocator->GetRecorder() != VMA_NULL) - { - allocator->GetRecorder()->RecordAllocateMemoryForImage( - allocator->GetCurrentFrameIndex(), - vkMemReq, - requiresDedicatedAllocation, - prefersDedicatedAllocation, - *pCreateInfo, - *pAllocation); - } -#endif - if(pAllocationInfo && result == VK_SUCCESS) { allocator->GetAllocationInfo(*pAllocation, pAllocationInfo); @@ -18967,15 +17141,6 @@ VMA_CALL_PRE void VMA_CALL_POST vmaFreeMemory( VMA_DEBUG_GLOBAL_MUTEX_LOCK -#if VMA_RECORDING_ENABLED - if(allocator->GetRecorder() != VMA_NULL) - { - allocator->GetRecorder()->RecordFreeMemory( - allocator->GetCurrentFrameIndex(), - allocation); - } -#endif - allocator->FreeMemory( 1, // allocationCount &allocation); @@ -18997,16 +17162,6 @@ VMA_CALL_PRE void VMA_CALL_POST vmaFreeMemoryPages( VMA_DEBUG_GLOBAL_MUTEX_LOCK -#if VMA_RECORDING_ENABLED - if(allocator->GetRecorder() != VMA_NULL) - { - allocator->GetRecorder()->RecordFreeMemoryPages( - allocator->GetCurrentFrameIndex(), - (uint64_t)allocationCount, - pAllocations); - } -#endif - allocator->FreeMemory(allocationCount, pAllocations); } @@ -19019,38 +17174,9 @@ VMA_CALL_PRE void VMA_CALL_POST vmaGetAllocationInfo( VMA_DEBUG_GLOBAL_MUTEX_LOCK -#if VMA_RECORDING_ENABLED - if(allocator->GetRecorder() != VMA_NULL) - { - allocator->GetRecorder()->RecordGetAllocationInfo( - allocator->GetCurrentFrameIndex(), - allocation); - } -#endif - allocator->GetAllocationInfo(allocation, pAllocationInfo); } -VMA_CALL_PRE VkBool32 VMA_CALL_POST vmaTouchAllocation( - VmaAllocator allocator, - VmaAllocation allocation) -{ - VMA_ASSERT(allocator && allocation); - - VMA_DEBUG_GLOBAL_MUTEX_LOCK - -#if VMA_RECORDING_ENABLED - if(allocator->GetRecorder() != VMA_NULL) - { - allocator->GetRecorder()->RecordTouchAllocation( - allocator->GetCurrentFrameIndex(), - allocation); - } -#endif - - return allocator->TouchAllocation(allocation); -} - VMA_CALL_PRE void VMA_CALL_POST vmaSetAllocationUserData( VmaAllocator allocator, VmaAllocation allocation, @@ -19061,36 +17187,16 @@ VMA_CALL_PRE void VMA_CALL_POST vmaSetAllocationUserData( VMA_DEBUG_GLOBAL_MUTEX_LOCK allocation->SetUserData(allocator, pUserData); - -#if VMA_RECORDING_ENABLED - if(allocator->GetRecorder() != VMA_NULL) - { - allocator->GetRecorder()->RecordSetAllocationUserData( - allocator->GetCurrentFrameIndex(), - allocation, - pUserData); - } -#endif } -VMA_CALL_PRE void VMA_CALL_POST vmaCreateLostAllocation( - VmaAllocator allocator, - VmaAllocation* pAllocation) +VMA_CALL_PRE void VMA_CALL_POST vmaGetAllocationMemoryProperties( + VmaAllocator VMA_NOT_NULL allocator, + VmaAllocation VMA_NOT_NULL allocation, + VkMemoryPropertyFlags* VMA_NOT_NULL pFlags) { - VMA_ASSERT(allocator && pAllocation); - - VMA_DEBUG_GLOBAL_MUTEX_LOCK; - - allocator->CreateLostAllocation(pAllocation); - -#if VMA_RECORDING_ENABLED - if(allocator->GetRecorder() != VMA_NULL) - { - allocator->GetRecorder()->RecordCreateLostAllocation( - allocator->GetCurrentFrameIndex(), - *pAllocation); - } -#endif + VMA_ASSERT(allocator && allocation && pFlags); + const uint32_t memTypeIndex = allocation->GetMemoryTypeIndex(); + *pFlags = allocator->m_MemProps.memoryTypes[memTypeIndex].propertyFlags; } VMA_CALL_PRE VkResult VMA_CALL_POST vmaMapMemory( @@ -19102,18 +17208,7 @@ VMA_CALL_PRE VkResult VMA_CALL_POST vmaMapMemory( VMA_DEBUG_GLOBAL_MUTEX_LOCK - VkResult res = allocator->Map(allocation, ppData); - -#if VMA_RECORDING_ENABLED - if(allocator->GetRecorder() != VMA_NULL) - { - allocator->GetRecorder()->RecordMapMemory( - allocator->GetCurrentFrameIndex(), - allocation); - } -#endif - - return res; + return allocator->Map(allocation, ppData); } VMA_CALL_PRE void VMA_CALL_POST vmaUnmapMemory( @@ -19124,19 +17219,14 @@ VMA_CALL_PRE void VMA_CALL_POST vmaUnmapMemory( VMA_DEBUG_GLOBAL_MUTEX_LOCK -#if VMA_RECORDING_ENABLED - if(allocator->GetRecorder() != VMA_NULL) - { - allocator->GetRecorder()->RecordUnmapMemory( - allocator->GetCurrentFrameIndex(), - allocation); - } -#endif - allocator->Unmap(allocation); } -VMA_CALL_PRE VkResult VMA_CALL_POST vmaFlushAllocation(VmaAllocator allocator, VmaAllocation allocation, VkDeviceSize offset, VkDeviceSize size) +VMA_CALL_PRE VkResult VMA_CALL_POST vmaFlushAllocation( + VmaAllocator allocator, + VmaAllocation allocation, + VkDeviceSize offset, + VkDeviceSize size) { VMA_ASSERT(allocator && allocation); @@ -19146,19 +17236,14 @@ VMA_CALL_PRE VkResult VMA_CALL_POST vmaFlushAllocation(VmaAllocator allocator, V const VkResult res = allocator->FlushOrInvalidateAllocation(allocation, offset, size, VMA_CACHE_FLUSH); -#if VMA_RECORDING_ENABLED - if(allocator->GetRecorder() != VMA_NULL) - { - allocator->GetRecorder()->RecordFlushAllocation( - allocator->GetCurrentFrameIndex(), - allocation, offset, size); - } -#endif - return res; } -VMA_CALL_PRE VkResult VMA_CALL_POST vmaInvalidateAllocation(VmaAllocator allocator, VmaAllocation allocation, VkDeviceSize offset, VkDeviceSize size) +VMA_CALL_PRE VkResult VMA_CALL_POST vmaInvalidateAllocation( + VmaAllocator allocator, + VmaAllocation allocation, + VkDeviceSize offset, + VkDeviceSize size) { VMA_ASSERT(allocator && allocation); @@ -19168,15 +17253,6 @@ VMA_CALL_PRE VkResult VMA_CALL_POST vmaInvalidateAllocation(VmaAllocator allocat const VkResult res = allocator->FlushOrInvalidateAllocation(allocation, offset, size, VMA_CACHE_INVALIDATE); -#if VMA_RECORDING_ENABLED - if(allocator->GetRecorder() != VMA_NULL) - { - allocator->GetRecorder()->RecordInvalidateAllocation( - allocator->GetCurrentFrameIndex(), - allocation, offset, size); - } -#endif - return res; } @@ -19202,13 +17278,6 @@ VMA_CALL_PRE VkResult VMA_CALL_POST vmaFlushAllocations( const VkResult res = allocator->FlushOrInvalidateAllocations(allocationCount, allocations, offsets, sizes, VMA_CACHE_FLUSH); -#if VMA_RECORDING_ENABLED - if(allocator->GetRecorder() != VMA_NULL) - { - //TODO - } -#endif - return res; } @@ -19234,17 +17303,12 @@ VMA_CALL_PRE VkResult VMA_CALL_POST vmaInvalidateAllocations( const VkResult res = allocator->FlushOrInvalidateAllocations(allocationCount, allocations, offsets, sizes, VMA_CACHE_INVALIDATE); -#if VMA_RECORDING_ENABLED - if(allocator->GetRecorder() != VMA_NULL) - { - //TODO - } -#endif - return res; } -VMA_CALL_PRE VkResult VMA_CALL_POST vmaCheckCorruption(VmaAllocator allocator, uint32_t memoryTypeBits) +VMA_CALL_PRE VkResult VMA_CALL_POST vmaCheckCorruption( + VmaAllocator allocator, + uint32_t memoryTypeBits) { VMA_ASSERT(allocator); @@ -19315,14 +17379,6 @@ VMA_CALL_PRE VkResult VMA_CALL_POST vmaDefragmentationBegin( VkResult res = allocator->DefragmentationBegin(*pInfo, pStats, pContext); -#if VMA_RECORDING_ENABLED - if(allocator->GetRecorder() != VMA_NULL) - { - allocator->GetRecorder()->RecordDefragmentationBegin( - allocator->GetCurrentFrameIndex(), *pInfo, *pContext); - } -#endif - return res; } @@ -19337,15 +17393,6 @@ VMA_CALL_PRE VkResult VMA_CALL_POST vmaDefragmentationEnd( if(context != VK_NULL_HANDLE) { VMA_DEBUG_GLOBAL_MUTEX_LOCK - -#if VMA_RECORDING_ENABLED - if(allocator->GetRecorder() != VMA_NULL) - { - allocator->GetRecorder()->RecordDefragmentationEnd( - allocator->GetCurrentFrameIndex(), context); - } -#endif - return allocator->DefragmentationEnd(context); } else @@ -19375,6 +17422,7 @@ VMA_CALL_PRE VkResult VMA_CALL_POST vmaBeginDefragmentationPass( return allocator->DefragmentationPassBegin(pInfo, context); } + VMA_CALL_PRE VkResult VMA_CALL_POST vmaEndDefragmentationPass( VmaAllocator allocator, VmaDefragmentationContext context) @@ -19462,13 +17510,13 @@ VMA_CALL_PRE VkResult VMA_CALL_POST vmaCreateBuffer( if(pBufferCreateInfo->size == 0) { - return VK_ERROR_VALIDATION_FAILED_EXT; + return VK_ERROR_INITIALIZATION_FAILED; } if((pBufferCreateInfo->usage & VK_BUFFER_USAGE_SHADER_DEVICE_ADDRESS_BIT_COPY) != 0 && !allocator->m_UseKhrBufferDeviceAddress) { VMA_ASSERT(0 && "Creating a buffer with VK_BUFFER_USAGE_SHADER_DEVICE_ADDRESS_BIT is not valid if VMA_ALLOCATOR_CREATE_BUFFER_DEVICE_ADDRESS_BIT was not used."); - return VK_ERROR_VALIDATION_FAILED_EXT; + return VK_ERROR_INITIALIZATION_FAILED; } VMA_DEBUG_LOG("vmaCreateBuffer"); @@ -19506,16 +17554,100 @@ VMA_CALL_PRE VkResult VMA_CALL_POST vmaCreateBuffer( 1, // allocationCount pAllocation); -#if VMA_RECORDING_ENABLED - if(allocator->GetRecorder() != VMA_NULL) + if(res >= 0) { - allocator->GetRecorder()->RecordCreateBuffer( - allocator->GetCurrentFrameIndex(), - *pBufferCreateInfo, - *pAllocationCreateInfo, - *pAllocation); + // 3. Bind buffer with memory. + if((pAllocationCreateInfo->flags & VMA_ALLOCATION_CREATE_DONT_BIND_BIT) == 0) + { + res = allocator->BindBufferMemory(*pAllocation, 0, *pBuffer, VMA_NULL); + } + if(res >= 0) + { + // All steps succeeded. + #if VMA_STATS_STRING_ENABLED + (*pAllocation)->InitBufferImageUsage(pBufferCreateInfo->usage); + #endif + if(pAllocationInfo != VMA_NULL) + { + allocator->GetAllocationInfo(*pAllocation, pAllocationInfo); + } + + return VK_SUCCESS; + } + allocator->FreeMemory( + 1, // allocationCount + pAllocation); + *pAllocation = VK_NULL_HANDLE; + (*allocator->GetVulkanFunctions().vkDestroyBuffer)(allocator->m_hDevice, *pBuffer, allocator->GetAllocationCallbacks()); + *pBuffer = VK_NULL_HANDLE; + return res; } -#endif + (*allocator->GetVulkanFunctions().vkDestroyBuffer)(allocator->m_hDevice, *pBuffer, allocator->GetAllocationCallbacks()); + *pBuffer = VK_NULL_HANDLE; + return res; + } + return res; +} + +VMA_CALL_PRE VkResult VMA_CALL_POST vmaCreateBufferWithAlignment( + VmaAllocator allocator, + const VkBufferCreateInfo* pBufferCreateInfo, + const VmaAllocationCreateInfo* pAllocationCreateInfo, + VkDeviceSize minAlignment, + VkBuffer* pBuffer, + VmaAllocation* pAllocation, + VmaAllocationInfo* pAllocationInfo) +{ + VMA_ASSERT(allocator && pBufferCreateInfo && pAllocationCreateInfo && VmaIsPow2(minAlignment) && pBuffer && pAllocation); + + if(pBufferCreateInfo->size == 0) + { + return VK_ERROR_INITIALIZATION_FAILED; + } + if((pBufferCreateInfo->usage & VK_BUFFER_USAGE_SHADER_DEVICE_ADDRESS_BIT_COPY) != 0 && + !allocator->m_UseKhrBufferDeviceAddress) + { + VMA_ASSERT(0 && "Creating a buffer with VK_BUFFER_USAGE_SHADER_DEVICE_ADDRESS_BIT is not valid if VMA_ALLOCATOR_CREATE_BUFFER_DEVICE_ADDRESS_BIT was not used."); + return VK_ERROR_INITIALIZATION_FAILED; + } + + VMA_DEBUG_LOG("vmaCreateBufferWithAlignment"); + + VMA_DEBUG_GLOBAL_MUTEX_LOCK + + *pBuffer = VK_NULL_HANDLE; + *pAllocation = VK_NULL_HANDLE; + + // 1. Create VkBuffer. + VkResult res = (*allocator->GetVulkanFunctions().vkCreateBuffer)( + allocator->m_hDevice, + pBufferCreateInfo, + allocator->GetAllocationCallbacks(), + pBuffer); + if(res >= 0) + { + // 2. vkGetBufferMemoryRequirements. + VkMemoryRequirements vkMemReq = {}; + bool requiresDedicatedAllocation = false; + bool prefersDedicatedAllocation = false; + allocator->GetBufferMemoryRequirements(*pBuffer, vkMemReq, + requiresDedicatedAllocation, prefersDedicatedAllocation); + + // 2a. Include minAlignment + vkMemReq.alignment = VMA_MAX(vkMemReq.alignment, minAlignment); + + // 3. Allocate memory using allocator. + res = allocator->AllocateMemory( + vkMemReq, + requiresDedicatedAllocation, + prefersDedicatedAllocation, + *pBuffer, // dedicatedBuffer + pBufferCreateInfo->usage, // dedicatedBufferUsage + VK_NULL_HANDLE, // dedicatedImage + *pAllocationCreateInfo, + VMA_SUBALLOCATION_TYPE_BUFFER, + 1, // allocationCount + pAllocation); if(res >= 0) { @@ -19568,15 +17700,6 @@ VMA_CALL_PRE void VMA_CALL_POST vmaDestroyBuffer( VMA_DEBUG_GLOBAL_MUTEX_LOCK -#if VMA_RECORDING_ENABLED - if(allocator->GetRecorder() != VMA_NULL) - { - allocator->GetRecorder()->RecordDestroyBuffer( - allocator->GetCurrentFrameIndex(), - allocation); - } -#endif - if(buffer != VK_NULL_HANDLE) { (*allocator->GetVulkanFunctions().vkDestroyBuffer)(allocator->m_hDevice, buffer, allocator->GetAllocationCallbacks()); @@ -19606,7 +17729,7 @@ VMA_CALL_PRE VkResult VMA_CALL_POST vmaCreateImage( pImageCreateInfo->mipLevels == 0 || pImageCreateInfo->arrayLayers == 0) { - return VK_ERROR_VALIDATION_FAILED_EXT; + return VK_ERROR_INITIALIZATION_FAILED; } VMA_DEBUG_LOG("vmaCreateImage"); @@ -19647,17 +17770,6 @@ VMA_CALL_PRE VkResult VMA_CALL_POST vmaCreateImage( 1, // allocationCount pAllocation); -#if VMA_RECORDING_ENABLED - if(allocator->GetRecorder() != VMA_NULL) - { - allocator->GetRecorder()->RecordCreateImage( - allocator->GetCurrentFrameIndex(), - *pImageCreateInfo, - *pAllocationCreateInfo, - *pAllocation); - } -#endif - if(res >= 0) { // 3. Bind image with memory. @@ -19709,15 +17821,6 @@ VMA_CALL_PRE void VMA_CALL_POST vmaDestroyImage( VMA_DEBUG_GLOBAL_MUTEX_LOCK -#if VMA_RECORDING_ENABLED - if(allocator->GetRecorder() != VMA_NULL) - { - allocator->GetRecorder()->RecordDestroyImage( - allocator->GetCurrentFrameIndex(), - allocation); - } -#endif - if(image != VK_NULL_HANDLE) { (*allocator->GetVulkanFunctions().vkDestroyImage)(allocator->m_hDevice, image, allocator->GetAllocationCallbacks()); @@ -19730,4 +17833,2069 @@ VMA_CALL_PRE void VMA_CALL_POST vmaDestroyImage( } } -#endif // #ifdef VMA_IMPLEMENTATION +VMA_CALL_PRE VkResult VMA_CALL_POST vmaCreateVirtualBlock( + const VmaVirtualBlockCreateInfo* VMA_NOT_NULL pCreateInfo, + VmaVirtualBlock VMA_NULLABLE * VMA_NOT_NULL pVirtualBlock) +{ + VMA_ASSERT(pCreateInfo && pVirtualBlock); + VMA_ASSERT(pCreateInfo->size > 0); + VMA_DEBUG_LOG("vmaCreateVirtualBlock"); + VMA_DEBUG_GLOBAL_MUTEX_LOCK; + *pVirtualBlock = vma_new(pCreateInfo->pAllocationCallbacks, VmaVirtualBlock_T)(*pCreateInfo); + VkResult res = (*pVirtualBlock)->Init(); + if(res < 0) + { + vma_delete(pCreateInfo->pAllocationCallbacks, *pVirtualBlock); + *pVirtualBlock = VK_NULL_HANDLE; + } + return res; +} + +VMA_CALL_PRE void VMA_CALL_POST vmaDestroyVirtualBlock(VmaVirtualBlock VMA_NULLABLE virtualBlock) +{ + if(virtualBlock != VK_NULL_HANDLE) + { + VMA_DEBUG_LOG("vmaDestroyVirtualBlock"); + VMA_DEBUG_GLOBAL_MUTEX_LOCK; + VkAllocationCallbacks allocationCallbacks = virtualBlock->m_AllocationCallbacks; // Have to copy the callbacks when destroying. + vma_delete(&allocationCallbacks, virtualBlock); + } +} + +VMA_CALL_PRE VkBool32 VMA_CALL_POST vmaIsVirtualBlockEmpty(VmaVirtualBlock VMA_NOT_NULL virtualBlock) +{ + VMA_ASSERT(virtualBlock != VK_NULL_HANDLE); + VMA_DEBUG_LOG("vmaIsVirtualBlockEmpty"); + VMA_DEBUG_GLOBAL_MUTEX_LOCK; + return virtualBlock->IsEmpty() ? VK_TRUE : VK_FALSE; +} + +VMA_CALL_PRE void VMA_CALL_POST vmaGetVirtualAllocationInfo(VmaVirtualBlock VMA_NOT_NULL virtualBlock, + VmaVirtualAllocation VMA_NOT_NULL_NON_DISPATCHABLE allocation, VmaVirtualAllocationInfo* VMA_NOT_NULL pVirtualAllocInfo) +{ + VMA_ASSERT(virtualBlock != VK_NULL_HANDLE && pVirtualAllocInfo != VMA_NULL); + VMA_DEBUG_LOG("vmaGetVirtualAllocationInfo"); + VMA_DEBUG_GLOBAL_MUTEX_LOCK; + virtualBlock->GetAllocationInfo(allocation, *pVirtualAllocInfo); +} + +VMA_CALL_PRE VkResult VMA_CALL_POST vmaVirtualAllocate(VmaVirtualBlock VMA_NOT_NULL virtualBlock, + const VmaVirtualAllocationCreateInfo* VMA_NOT_NULL pCreateInfo, VmaVirtualAllocation VMA_NULLABLE_NON_DISPATCHABLE* VMA_NOT_NULL pAllocation, + VkDeviceSize* VMA_NULLABLE pOffset) +{ + VMA_ASSERT(virtualBlock != VK_NULL_HANDLE && pCreateInfo != VMA_NULL && pAllocation != VMA_NULL); + VMA_DEBUG_LOG("vmaVirtualAllocate"); + VMA_DEBUG_GLOBAL_MUTEX_LOCK; + return virtualBlock->Allocate(*pCreateInfo, *pAllocation, pOffset); +} + +VMA_CALL_PRE void VMA_CALL_POST vmaVirtualFree(VmaVirtualBlock VMA_NOT_NULL virtualBlock, VmaVirtualAllocation VMA_NULLABLE_NON_DISPATCHABLE allocation) +{ + if(allocation != VK_NULL_HANDLE) + { + VMA_ASSERT(virtualBlock != VK_NULL_HANDLE); + VMA_DEBUG_LOG("vmaVirtualFree"); + VMA_DEBUG_GLOBAL_MUTEX_LOCK; + virtualBlock->Free(allocation); + } +} + +VMA_CALL_PRE void VMA_CALL_POST vmaClearVirtualBlock(VmaVirtualBlock VMA_NOT_NULL virtualBlock) +{ + VMA_ASSERT(virtualBlock != VK_NULL_HANDLE); + VMA_DEBUG_LOG("vmaClearVirtualBlock"); + VMA_DEBUG_GLOBAL_MUTEX_LOCK; + virtualBlock->Clear(); +} + +VMA_CALL_PRE void VMA_CALL_POST vmaSetVirtualAllocationUserData(VmaVirtualBlock VMA_NOT_NULL virtualBlock, + VmaVirtualAllocation VMA_NOT_NULL_NON_DISPATCHABLE allocation, void* VMA_NULLABLE pUserData) +{ + VMA_ASSERT(virtualBlock != VK_NULL_HANDLE); + VMA_DEBUG_LOG("vmaSetVirtualAllocationUserData"); + VMA_DEBUG_GLOBAL_MUTEX_LOCK; + virtualBlock->SetAllocationUserData(allocation, pUserData); +} + +VMA_CALL_PRE void VMA_CALL_POST vmaCalculateVirtualBlockStats(VmaVirtualBlock VMA_NOT_NULL virtualBlock, + VmaStatInfo* VMA_NOT_NULL pStatInfo) +{ + VMA_ASSERT(virtualBlock != VK_NULL_HANDLE && pStatInfo != VMA_NULL); + VMA_DEBUG_LOG("vmaCalculateVirtualBlockStats"); + VMA_DEBUG_GLOBAL_MUTEX_LOCK; + virtualBlock->CalculateStats(*pStatInfo); +} + +#if VMA_STATS_STRING_ENABLED + +VMA_CALL_PRE void VMA_CALL_POST vmaBuildVirtualBlockStatsString(VmaVirtualBlock VMA_NOT_NULL virtualBlock, + char* VMA_NULLABLE * VMA_NOT_NULL ppStatsString, VkBool32 detailedMap) +{ + VMA_ASSERT(virtualBlock != VK_NULL_HANDLE && ppStatsString != VMA_NULL); + VMA_DEBUG_GLOBAL_MUTEX_LOCK; + const VkAllocationCallbacks* allocationCallbacks = virtualBlock->GetAllocationCallbacks(); + VmaStringBuilder sb(allocationCallbacks); + virtualBlock->BuildStatsString(detailedMap != VK_FALSE, sb); + *ppStatsString = VmaCreateStringCopy(allocationCallbacks, sb.GetData(), sb.GetLength()); +} + +VMA_CALL_PRE void VMA_CALL_POST vmaFreeVirtualBlockStatsString(VmaVirtualBlock VMA_NOT_NULL virtualBlock, + char* VMA_NULLABLE pStatsString) +{ + if(pStatsString != VMA_NULL) + { + VMA_ASSERT(virtualBlock != VK_NULL_HANDLE); + VMA_DEBUG_GLOBAL_MUTEX_LOCK; + VmaFreeString(virtualBlock->GetAllocationCallbacks(), pStatsString); + } +} +#endif // VMA_STATS_STRING_ENABLED +#endif // _VMA_PUBLIC_INTERFACE +#endif // VMA_IMPLEMENTATION + +/** +\page quick_start Quick start + +\section quick_start_project_setup Project setup + +Vulkan Memory Allocator comes in form of a "stb-style" 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 exactly 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. + +This library includes header `<vulkan/vulkan.h>`, which in turn +includes `<windows.h>` on Windows. If you need some specific macros defined +before including these headers (like `WIN32_LEAN_AND_MEAN` or +`WINVER` for Windows, `VK_USE_PLATFORM_WIN32_KHR` for Vulkan), you must define +them before every `#include` of this library. + +\note This library is written in C++, but has C-compatible interface. +Thus you can include and use vk_mem_alloc.h in C or C++ code, but full +implementation with `VMA_IMPLEMENTATION` macro must be compiled as C++, NOT as C. + + +\section quick_start_initialization Initialization + +At program startup: + +-# Initialize Vulkan to have `VkPhysicalDevice`, `VkDevice` and `VkInstance` object. +-# Fill VmaAllocatorCreateInfo structure and create #VmaAllocator object by + calling vmaCreateAllocator(). + +Only members `physicalDevice`, `device`, `instance` are required. +However, you should inform the library which Vulkan version do you use by setting +VmaAllocatorCreateInfo::vulkanApiVersion and which extensions did you enable +by setting VmaAllocatorCreateInfo::flags (like #VMA_ALLOCATOR_CREATE_BUFFER_DEVICE_ADDRESS_BIT for VK_KHR_buffer_device_address). +Otherwise, VMA would use only features of Vulkan 1.0 core with no extensions. + +You may need to configure importing Vulkan functions. There are 3 ways to do this: + +-# **If you link with Vulkan static library** (e.g. "vulkan-1.lib" on Windows): + - You don't need to do anything. + - VMA will use these, as macro `VMA_STATIC_VULKAN_FUNCTIONS` is defined to 1 by default. +-# **If you want VMA to fetch pointers to Vulkan functions dynamically** using `vkGetInstanceProcAddr`, + `vkGetDeviceProcAddr` (this is the option presented in the example below): + - Define `VMA_STATIC_VULKAN_FUNCTIONS` to 0, `VMA_DYNAMIC_VULKAN_FUNCTIONS` to 1. + - Provide pointers to these two functions via VmaVulkanFunctions::vkGetInstanceProcAddr, + VmaVulkanFunctions::vkGetDeviceProcAddr. + - The library will fetch pointers to all other functions it needs internally. +-# **If you fetch pointers to all Vulkan functions in a custom way**, e.g. using some loader like + [Volk](https://github.com/zeux/volk): + - Define `VMA_STATIC_VULKAN_FUNCTIONS` and `VMA_DYNAMIC_VULKAN_FUNCTIONS` to 0. + - Pass these pointers via structure #VmaVulkanFunctions. + +\code +VmaVulkanFunctions vulkanFunctions = {}; +vulkanFunctions.vkGetInstanceProcAddr = &vkGetInstanceProcAddr; +vulkanFunctions.vkGetDeviceProcAddr = &vkGetDeviceProcAddr; + +VmaAllocatorCreateInfo allocatorCreateInfo = {}; +allocatorCreateInfo.vulkanApiVersion = VK_API_VERSION_1_2; +allocatorCreateInfo.physicalDevice = physicalDevice; +allocatorCreateInfo.device = device; +allocatorCreateInfo.instance = instance; +allocatorCreateInfo.pVulkanFunctions = &vulkanFunctions; + +VmaAllocator allocator; +vmaCreateAllocator(&allocatorCreateInfo, &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, plus #VmaAllocation objects that represents its underlying memory. + +\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(), vmaFindMemoryTypeIndexForBufferInfo(), + vmaFindMemoryTypeIndexForImageInfo(). +-# 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. + vmaAllocateMemoryPages() function is also provided for creating multiple allocations at once, + which may be useful for sparse binding. +-# 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() + or their extended versions: vmaBindBufferMemory2(), vmaBindImageMemory2(). +-# 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 easiest and 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. + +\section choosing_memory_type_dedicated_allocations Dedicated allocations + +Memory for allocations is reserved out of larger block of `VkDeviceMemory` +allocated from Vulkan internally. That is the main feature of this whole library. +You can still request a separate memory block to be created for an allocation, +just like you would do in a trivial solution without using any allocator. +In that case, a buffer or image is always bound to that memory at offset 0. +This is called a "dedicated allocation". +You can explicitly request it by using flag #VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT. +The library can also internally decide to use dedicated allocation in some cases, e.g.: + +- When the size of the allocation is large. +- When [VK_KHR_dedicated_allocation](@ref vk_khr_dedicated_allocation) extension is enabled + and it reports that dedicated allocation is required or recommended for the resource. +- When allocation of next big memory block fails due to not enough device memory, + but allocation with the exact requested size succeeds. + + +\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. +The way it is 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; +VkBuffer 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 + +When mapping, you may see a warning from Vulkan validation layer similar to this one: + +<i>Mapping an image with layout VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL can result in undefined behavior if this memory is used by the device. Only GENERAL or PREINITIALIZED should be used.</i> + +It happens because the library maps entire `VkDeviceMemory` block, where different +types of images and buffers may end up together, especially on GPUs with unified memory like Intel. +You can safely ignore it if you are sure you access only memory of the intended +object that you wanted to map. + + +\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 flush and invalidate + +Memory in Vulkan doesn't need to be unmapped before using it on GPU, +but unless a memory types has `VK_MEMORY_PROPERTY_HOST_COHERENT_BIT` flag set, +you need to manually **invalidate** cache before reading of mapped pointer +and **flush** cache after writing to mapped pointer. +Map/unmap operations don't do that automatically. +Vulkan provides following functions for this purpose `vkFlushMappedMemoryRanges()`, +`vkInvalidateMappedMemoryRanges()`, but this library provides more convenient +functions that refer to given allocation object: vmaFlushAllocation(), +vmaInvalidateAllocation(), +or multiple objects at once: vmaFlushAllocations(), vmaInvalidateAllocations(). + +Regions of memory specified for flush/invalidate must be aligned to +`VkPhysicalDeviceLimits::nonCoherentAtomSize`. This is automatically ensured by the library. +In any memory type that is `HOST_VISIBLE` but not `HOST_COHERENT`, all allocations +within blocks are aligned to this value, so their offsets are always multiply of +`nonCoherentAtomSize` and two different allocations never share same "line" of this size. + +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 `HOST_COHERENT` flag on all memory types that are +`HOST_VISIBLE`, 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: call vmaGetAllocationMemoryProperties() +and look for `VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT` flag. + +\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.preferredFlags = VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT; + +VkBuffer buf; +VmaAllocation alloc; +vmaCreateBuffer(allocator, &bufCreateInfo, &allocCreateInfo, &buf, &alloc, nullptr); + +VkMemoryPropertyFlags memFlags; +vmaGetAllocationMemoryProperties(allocator, alloc, &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.pMappedData != nullptr) +{ + // Allocation ended up in mappable memory. + // It is 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 staying_within_budget Staying within budget + +When developing a graphics-intensive game or program, it is important to avoid allocating +more GPU memory than it is physically available. When the memory is over-committed, +various bad things can happen, depending on the specific GPU, graphics driver, and +operating system: + +- It may just work without any problems. +- The application may slow down because some memory blocks are moved to system RAM + and the GPU has to access them through PCI Express bus. +- A new allocation may take very long time to complete, even few seconds, and possibly + freeze entire system. +- The new allocation may fail with `VK_ERROR_OUT_OF_DEVICE_MEMORY`. +- It may even result in GPU crash (TDR), observed as `VK_ERROR_DEVICE_LOST` + returned somewhere later. + +\section staying_within_budget_querying_for_budget Querying for budget + +To query for current memory usage and available budget, use function vmaGetHeapBudgets(). +Returned structure #VmaBudget contains quantities expressed in bytes, per Vulkan memory heap. + +Please note that this function returns different information and works faster than +vmaCalculateStats(). vmaGetHeapBudgets() can be called every frame or even before every +allocation, while vmaCalculateStats() is intended to be used rarely, +only to obtain statistical information, e.g. for debugging purposes. + +It is recommended to use <b>VK_EXT_memory_budget</b> device extension to obtain information +about the budget from Vulkan device. VMA is able to use this extension automatically. +When not enabled, the allocator behaves same way, but then it estimates current usage +and available budget based on its internal information and Vulkan memory heap sizes, +which may be less precise. In order to use this extension: + +1. Make sure extensions VK_EXT_memory_budget and VK_KHR_get_physical_device_properties2 + required by it are available and enable them. Please note that the first is a device + extension and the second is instance extension! +2. Use flag #VMA_ALLOCATOR_CREATE_EXT_MEMORY_BUDGET_BIT when creating #VmaAllocator object. +3. Make sure to call vmaSetCurrentFrameIndex() every frame. Budget is queried from + Vulkan inside of it to avoid overhead of querying it with every allocation. + +\section staying_within_budget_controlling_memory_usage Controlling memory usage + +There are many ways in which you can try to stay within the budget. + +First, when making new allocation requires allocating a new memory block, the library +tries not to exceed the budget automatically. If a block with default recommended size +(e.g. 256 MB) would go over budget, a smaller block is allocated, possibly even +dedicated memory for just this resource. + +If the size of the requested resource plus current memory usage is more than the +budget, by default the library still tries to create it, leaving it to the Vulkan +implementation whether the allocation succeeds or fails. You can change this behavior +by using #VMA_ALLOCATION_CREATE_WITHIN_BUDGET_BIT flag. With it, the allocation is +not made if it would exceed the budget or if the budget is already exceeded. +The allocation then fails with `VK_ERROR_OUT_OF_DEVICE_MEMORY`. +Example usage pattern may be to pass the #VMA_ALLOCATION_CREATE_WITHIN_BUDGET_BIT flag +when creating resources that are not essential for the application (e.g. the texture +of a specific object) and not to pass it when creating critically important resources +(e.g. render targets). + +Finally, you can also use #VMA_ALLOCATION_CREATE_NEVER_ALLOCATE_BIT flag to make sure +a new allocation is created only when it fits inside one of the existing memory blocks. +If it would require to allocate a new block, if fails instead with `VK_ERROR_OUT_OF_DEVICE_MEMORY`. +This also ensures that the function call is very fast because it never goes to Vulkan +to obtain a new block. + +Please note that creating \ref custom_memory_pools with VmaPoolCreateInfo::minBlockCount +set to more than 0 will try to allocate memory blocks without checking whether they +fit within budget. + + +\page resource_aliasing Resource aliasing (overlap) + +New explicit graphics APIs (Vulkan and Direct3D 12), thanks to manual memory +management, give an opportunity to alias (overlap) multiple resources in the +same region of memory - a feature not available in the old APIs (Direct3D 11, OpenGL). +It can be useful to save video memory, but it must be used with caution. + +For example, if you know the flow of your whole render frame in advance, you +are going to use some intermediate textures or buffers only during a small range of render passes, +and you know these ranges don't overlap in time, you can bind these resources to +the same place in memory, even if they have completely different parameters (width, height, format etc.). + + + +Such scenario is possible using VMA, but you need to create your images manually. +Then you need to calculate parameters of an allocation to be made using formula: + +- allocation size = max(size of each image) +- allocation alignment = max(alignment of each image) +- allocation memoryTypeBits = bitwise AND(memoryTypeBits of each image) + +Following example shows two different images bound to the same place in memory, +allocated to fit largest of them. + +\code +// A 512x512 texture to be sampled. +VkImageCreateInfo img1CreateInfo = { VK_STRUCTURE_TYPE_IMAGE_CREATE_INFO }; +img1CreateInfo.imageType = VK_IMAGE_TYPE_2D; +img1CreateInfo.extent.width = 512; +img1CreateInfo.extent.height = 512; +img1CreateInfo.extent.depth = 1; +img1CreateInfo.mipLevels = 10; +img1CreateInfo.arrayLayers = 1; +img1CreateInfo.format = VK_FORMAT_R8G8B8A8_SRGB; +img1CreateInfo.tiling = VK_IMAGE_TILING_OPTIMAL; +img1CreateInfo.initialLayout = VK_IMAGE_LAYOUT_UNDEFINED; +img1CreateInfo.usage = VK_IMAGE_USAGE_TRANSFER_DST_BIT | VK_IMAGE_USAGE_SAMPLED_BIT; +img1CreateInfo.samples = VK_SAMPLE_COUNT_1_BIT; + +// A full screen texture to be used as color attachment. +VkImageCreateInfo img2CreateInfo = { VK_STRUCTURE_TYPE_IMAGE_CREATE_INFO }; +img2CreateInfo.imageType = VK_IMAGE_TYPE_2D; +img2CreateInfo.extent.width = 1920; +img2CreateInfo.extent.height = 1080; +img2CreateInfo.extent.depth = 1; +img2CreateInfo.mipLevels = 1; +img2CreateInfo.arrayLayers = 1; +img2CreateInfo.format = VK_FORMAT_R8G8B8A8_UNORM; +img2CreateInfo.tiling = VK_IMAGE_TILING_OPTIMAL; +img2CreateInfo.initialLayout = VK_IMAGE_LAYOUT_UNDEFINED; +img2CreateInfo.usage = VK_IMAGE_USAGE_SAMPLED_BIT | VK_IMAGE_USAGE_COLOR_ATTACHMENT_BIT; +img2CreateInfo.samples = VK_SAMPLE_COUNT_1_BIT; + +VkImage img1; +res = vkCreateImage(device, &img1CreateInfo, nullptr, &img1); +VkImage img2; +res = vkCreateImage(device, &img2CreateInfo, nullptr, &img2); + +VkMemoryRequirements img1MemReq; +vkGetImageMemoryRequirements(device, img1, &img1MemReq); +VkMemoryRequirements img2MemReq; +vkGetImageMemoryRequirements(device, img2, &img2MemReq); + +VkMemoryRequirements finalMemReq = {}; +finalMemReq.size = std::max(img1MemReq.size, img2MemReq.size); +finalMemReq.alignment = std::max(img1MemReq.alignment, img2MemReq.alignment); +finalMemReq.memoryTypeBits = img1MemReq.memoryTypeBits & img2MemReq.memoryTypeBits; +// Validate if(finalMemReq.memoryTypeBits != 0) + +VmaAllocationCreateInfo allocCreateInfo = {}; +allocCreateInfo.usage = VMA_MEMORY_USAGE_GPU_ONLY; + +VmaAllocation alloc; +res = vmaAllocateMemory(allocator, &finalMemReq, &allocCreateInfo, &alloc, nullptr); + +res = vmaBindImageMemory(allocator, alloc, img1); +res = vmaBindImageMemory(allocator, alloc, img2); + +// You can use img1, img2 here, but not at the same time! + +vmaFreeMemory(allocator, alloc); +vkDestroyImage(allocator, img2, nullptr); +vkDestroyImage(allocator, img1, nullptr); +\endcode + +Remember that using resources that alias in memory requires proper synchronization. +You need to issue a memory barrier to make sure commands that use `img1` and `img2` +don't overlap on GPU timeline. +You also need to treat a resource after aliasing as uninitialized - containing garbage data. +For example, if you use `img1` and then want to use `img2`, you need to issue +an image memory barrier for `img2` with `oldLayout` = `VK_IMAGE_LAYOUT_UNDEFINED`. + +Additional considerations: + +- Vulkan also allows to interpret contents of memory between aliasing resources consistently in some cases. +See chapter 11.8. "Memory Aliasing" of Vulkan specification or `VK_IMAGE_CREATE_ALIAS_BIT` flag. +- You can create more complex layout where different images and buffers are bound +at different offsets inside one large allocation. For example, one can imagine +a big texture used in some render passes, aliasing with a set of many small buffers +used between in some further passes. To bind a resource at non-zero offset of an allocation, +use vmaBindBufferMemory2() / vmaBindImageMemory2(). +- Before allocating memory for the resources you want to alias, check `memoryTypeBits` +returned in memory requirements of each resource to make sure the bits overlap. +Some GPUs may expose multiple memory types suitable e.g. only for buffers or +images with `COLOR_ATTACHMENT` usage, so the sets of memory types supported by your +resources may be disjoint. Aliasing them is not possible in that case. + + +\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. +- Use extra parameters for a set of your allocations that are available in #VmaPoolCreateInfo but not in + #VmaAllocationCreateInfo - e.g., custom minimum alignment, custom `pNext` chain. + +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 + +New versions of this library support creating dedicated allocations in custom pools. +It is supported only when VmaPoolCreateInfo::blockSize = 0. +To use this feature, set VmaAllocationCreateInfo::pool to the pointer to your custom pool and +VmaAllocationCreateInfo::flags to #VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT. + +\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. + +\section linear_algorithm Linear allocation algorithm + +Each Vulkan memory block managed by this library has accompanying metadata that +keeps track of used and unused regions. By default, the metadata structure and +algorithm tries to find best place for new allocations among free regions to +optimize memory usage. This way you can allocate and free objects in any order. + + + +Sometimes there is a need to use simpler, linear allocation algorithm. You can +create custom pool that uses such algorithm by adding flag +#VMA_POOL_CREATE_LINEAR_ALGORITHM_BIT to VmaPoolCreateInfo::flags while creating +#VmaPool object. Then an alternative metadata management is used. It always +creates new allocations after last one and doesn't reuse free regions after +allocations freed in the middle. It results in better allocation performance and +less memory consumed by metadata. + + + +With this one flag, you can create a custom pool that can be used in many ways: +free-at-once, stack, double stack, and ring buffer. See below for details. +You don't need to specify explicitly which of these options you are going to use - it is detected automatically. + +\subsection linear_algorithm_free_at_once Free-at-once + +In a pool that uses linear algorithm, you still need to free all the allocations +individually, e.g. by using vmaFreeMemory() or vmaDestroyBuffer(). You can free +them in any order. New allocations are always made after last one - free space +in the middle is not reused. However, when you release all the allocation and +the pool becomes empty, allocation starts from the beginning again. This way you +can use linear algorithm to speed up creation of allocations that you are going +to release all at once. + + + +This mode is also available for pools created with VmaPoolCreateInfo::maxBlockCount +value that allows multiple memory blocks. + +\subsection linear_algorithm_stack Stack + +When you free an allocation that was created last, its space can be reused. +Thanks to this, if you always release allocations in the order opposite to their +creation (LIFO - Last In First Out), you can achieve behavior of a stack. + + + +This mode is also available for pools created with VmaPoolCreateInfo::maxBlockCount +value that allows multiple memory blocks. + +\subsection linear_algorithm_double_stack Double stack + +The space reserved by a custom pool with linear algorithm may be used by two +stacks: + +- First, default one, growing up from offset 0. +- Second, "upper" one, growing down from the end towards lower offsets. + +To make allocation from the upper stack, add flag #VMA_ALLOCATION_CREATE_UPPER_ADDRESS_BIT +to VmaAllocationCreateInfo::flags. + + + +Double stack is available only in pools with one memory block - +VmaPoolCreateInfo::maxBlockCount must be 1. Otherwise behavior is undefined. + +When the two stacks' ends meet so there is not enough space between them for a +new allocation, such allocation fails with usual +`VK_ERROR_OUT_OF_DEVICE_MEMORY` error. + +\subsection linear_algorithm_ring_buffer Ring buffer + +When you free some allocations from the beginning and there is not enough free space +for a new one at the end of a pool, allocator's "cursor" wraps around to the +beginning and starts allocation there. Thanks to this, if you always release +allocations in the same order as you created them (FIFO - First In First Out), +you can achieve behavior of a ring buffer / queue. + + + +Ring buffer is available only in pools with one memory block - +VmaPoolCreateInfo::maxBlockCount must be 1. Otherwise behavior is undefined. + +\section buddy_algorithm Buddy allocation algorithm + +There is another allocation algorithm that can be used with custom pools, called +"buddy". Its internal data structure is based on a binary tree of blocks, each having +size that is a power of two and a half of its parent's size. When you want to +allocate memory of certain size, a free node in the tree is located. If it is too +large, it is recursively split into two halves (called "buddies"). However, if +requested allocation size is not a power of two, the size of the allocation is +aligned up to the nearest power of two and the remaining space is wasted. When +two buddy nodes become free, they are merged back into one larger node. + + + +The advantage of buddy allocation algorithm over default algorithm is faster +allocation and deallocation, as well as smaller external fragmentation. The +disadvantage is more wasted space (internal fragmentation). +For more information, please search the Internet for "Buddy memory allocation" - +sources that describe this concept in general. + +To use buddy allocation algorithm with a custom pool, add flag +#VMA_POOL_CREATE_BUDDY_ALGORITHM_BIT to VmaPoolCreateInfo::flags while creating +#VmaPool object. + +Several limitations apply to pools that use buddy algorithm: + +- It is recommended to use VmaPoolCreateInfo::blockSize that is a power of two. + Otherwise, only largest power of two smaller than the size is used for + allocations. The remaining space always stays unused. +- [Margins](@ref debugging_memory_usage_margins) and + [corruption detection](@ref debugging_memory_usage_corruption_detection) + don't work in such pools. +- [Defragmentation](@ref defragmentation) doesn't work with allocations made from + such pool. + +\page defragmentation Defragmentation + +Interleaved allocations and deallocations of many objects of varying size can +cause fragmentation over time, 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 defragmentation feature: +structure #VmaDefragmentationInfo2, function vmaDefragmentationBegin(), vmaDefragmentationEnd(). +Given set of allocations, +this function can move them to compact used memory, ensure more continuous free +space and possibly also free some `VkDeviceMemory` blocks. + +What the defragmentation does is: + +- Updates #VmaAllocation objects to point to new `VkDeviceMemory` and offset. + After allocation has been moved, its VmaAllocationInfo::deviceMemory and/or + VmaAllocationInfo::offset changes. You must query them again using + vmaGetAllocationInfo() if you need them. +- Moves actual data in memory. + +What it doesn't do, so you need to do it yourself: + +- Recreate buffers and images that were bound to allocations that were defragmented and + bind them with their new places in memory. + You must use `vkDestroyBuffer()`, `vkDestroyImage()`, + `vkCreateBuffer()`, `vkCreateImage()`, vmaBindBufferMemory(), vmaBindImageMemory() + for that purpose and NOT vmaDestroyBuffer(), + vmaDestroyImage(), vmaCreateBuffer(), vmaCreateImage(), because you don't need to + destroy or create allocation objects! +- Recreate views and update descriptors that point to these buffers and images. + +\section defragmentation_cpu Defragmenting CPU memory + +Following example demonstrates how you can run defragmentation on CPU. +Only allocations created in memory types that are `HOST_VISIBLE` can be defragmented. +Others are ignored. + +The way it works is: + +- It temporarily maps entire memory blocks when necessary. +- It moves data using `memmove()` function. + +\code +// Given following variables already initialized: +VkDevice device; +VmaAllocator allocator; +std::vector<VkBuffer> buffers; +std::vector<VmaAllocation> allocations; + + +const uint32_t allocCount = (uint32_t)allocations.size(); +std::vector<VkBool32> allocationsChanged(allocCount); + +VmaDefragmentationInfo2 defragInfo = {}; +defragInfo.allocationCount = allocCount; +defragInfo.pAllocations = allocations.data(); +defragInfo.pAllocationsChanged = allocationsChanged.data(); +defragInfo.maxCpuBytesToMove = VK_WHOLE_SIZE; // No limit. +defragInfo.maxCpuAllocationsToMove = UINT32_MAX; // No limit. + +VmaDefragmentationContext defragCtx; +vmaDefragmentationBegin(allocator, &defragInfo, nullptr, &defragCtx); +vmaDefragmentationEnd(allocator, defragCtx); + +for(uint32_t i = 0; i < allocCount; ++i) +{ + if(allocationsChanged[i]) + { + // Destroy buffer that is immutably bound to memory region which is no longer valid. + vkDestroyBuffer(device, buffers[i], nullptr); + + // Create new buffer with same parameters. + VkBufferCreateInfo bufferInfo = ...; + vkCreateBuffer(device, &bufferInfo, nullptr, &buffers[i]); + + // You can make dummy call to vkGetBufferMemoryRequirements here to silence validation layer warning. + + // Bind new buffer to new memory region. Data contained in it is already moved. + VmaAllocationInfo allocInfo; + vmaGetAllocationInfo(allocator, allocations[i], &allocInfo); + vmaBindBufferMemory(allocator, allocations[i], buffers[i]); + } +} +\endcode + +Setting VmaDefragmentationInfo2::pAllocationsChanged is optional. +This output array tells whether particular allocation in VmaDefragmentationInfo2::pAllocations at the same index +has been modified during defragmentation. +You can pass null, but you then need to query every allocation passed to defragmentation +for new parameters using vmaGetAllocationInfo() if you might need to recreate and rebind a buffer or image associated with it. + +If you use [Custom memory pools](@ref choosing_memory_type_custom_memory_pools), +you can fill VmaDefragmentationInfo2::poolCount and VmaDefragmentationInfo2::pPools +instead of VmaDefragmentationInfo2::allocationCount and VmaDefragmentationInfo2::pAllocations +to defragment all allocations in given pools. +You cannot use VmaDefragmentationInfo2::pAllocationsChanged in that case. +You can also combine both methods. + +\section defragmentation_gpu Defragmenting GPU memory + +It is also possible to defragment allocations created in memory types that are not `HOST_VISIBLE`. +To do that, you need to pass a command buffer that meets requirements as described in +VmaDefragmentationInfo2::commandBuffer. The way it works is: + +- It creates temporary buffers and binds them to entire memory blocks when necessary. +- It issues `vkCmdCopyBuffer()` to passed command buffer. + +Example: + +\code +// Given following variables already initialized: +VkDevice device; +VmaAllocator allocator; +VkCommandBuffer commandBuffer; +std::vector<VkBuffer> buffers; +std::vector<VmaAllocation> allocations; + + +const uint32_t allocCount = (uint32_t)allocations.size(); +std::vector<VkBool32> allocationsChanged(allocCount); + +VkCommandBufferBeginInfo cmdBufBeginInfo = ...; +vkBeginCommandBuffer(commandBuffer, &cmdBufBeginInfo); + +VmaDefragmentationInfo2 defragInfo = {}; +defragInfo.allocationCount = allocCount; +defragInfo.pAllocations = allocations.data(); +defragInfo.pAllocationsChanged = allocationsChanged.data(); +defragInfo.maxGpuBytesToMove = VK_WHOLE_SIZE; // Notice it is "GPU" this time. +defragInfo.maxGpuAllocationsToMove = UINT32_MAX; // Notice it is "GPU" this time. +defragInfo.commandBuffer = commandBuffer; + +VmaDefragmentationContext defragCtx; +vmaDefragmentationBegin(allocator, &defragInfo, nullptr, &defragCtx); + +vkEndCommandBuffer(commandBuffer); + +// Submit commandBuffer. +// Wait for a fence that ensures commandBuffer execution finished. + +vmaDefragmentationEnd(allocator, defragCtx); + +for(uint32_t i = 0; i < allocCount; ++i) +{ + if(allocationsChanged[i]) + { + // Destroy buffer that is immutably bound to memory region which is no longer valid. + vkDestroyBuffer(device, buffers[i], nullptr); + + // Create new buffer with same parameters. + VkBufferCreateInfo bufferInfo = ...; + vkCreateBuffer(device, &bufferInfo, nullptr, &buffers[i]); + + // You can make dummy call to vkGetBufferMemoryRequirements here to silence validation layer warning. + + // Bind new buffer to new memory region. Data contained in it is already moved. + VmaAllocationInfo allocInfo; + vmaGetAllocationInfo(allocator, allocations[i], &allocInfo); + vmaBindBufferMemory(allocator, allocations[i], buffers[i]); + } +} +\endcode + +You can combine these two methods by specifying non-zero `maxGpu*` as well as `maxCpu*` parameters. +The library automatically chooses best method to defragment each memory pool. + +You may try not to block your entire program to wait until defragmentation finishes, +but do it in the background, as long as you carefully fullfill requirements described +in function vmaDefragmentationBegin(). + +\section defragmentation_additional_notes Additional notes + +It is only legal to defragment allocations bound to: + +- buffers +- images created with `VK_IMAGE_CREATE_ALIAS_BIT`, `VK_IMAGE_TILING_LINEAR`, and + being currently in `VK_IMAGE_LAYOUT_GENERAL` or `VK_IMAGE_LAYOUT_PREINITIALIZED`. + +Defragmentation of images created with `VK_IMAGE_TILING_OPTIMAL` or in any other +layout may give undefined results. + +If you defragment allocations bound to images, new images to be bound to new +memory region after defragmentation should be created with `VK_IMAGE_LAYOUT_PREINITIALIZED` +and then transitioned to their original layout from before defragmentation if +needed using an image memory barrier. + +While using defragmentation, you may experience validation layer warnings, which you just need to ignore. +See [Validation layer warnings](@ref general_considerations_validation_layer_warnings). + +Please don't expect memory to be fully compacted after defragmentation. +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 maximize continuous +empty space inside remaining blocks, while minimizing the number and size of allocations that +need to be moved. Some fragmentation may still remain - this is normal. + +\section defragmentation_custom_algorithm Writing custom defragmentation algorithm + +If you want to implement your own, custom defragmentation algorithm, +there is infrastructure prepared for that, +but it is not exposed through the library API - you need to hack its source code. +Here are steps needed to do this: + +-# Main thing you need to do is to define your own class derived from base abstract + class `VmaDefragmentationAlgorithm` and implement your version of its pure virtual methods. + See definition and comments of this class for details. +-# Your code needs to interact with device memory block metadata. + If you need more access to its data than it is provided by its public interface, + declare your new class as a friend class e.g. in class `VmaBlockMetadata_Generic`. +-# If you want to create a flag that would enable your algorithm or pass some additional + flags to configure it, add them to `VmaDefragmentationFlagBits` and use them in + VmaDefragmentationInfo2::flags. +-# Modify function `VmaBlockVectorDefragmentationContext::Begin` to create object + of your new class whenever needed. + + +\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 is 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(). + +\note Passing string name to VMA allocation doesn't automatically set it to the Vulkan buffer or image created with it. +You must do it manually using an extension like VK_EXT_debug_utils, which is independent of this library. + + +\page virtual_allocator Virtual allocator + +As an extra feature, the core allocation algorithm of the library is exposed through a simple and convenient API of "virtual allocator". +It doesn't allocate any real GPU memory. It just keeps track of used and free regions of a "virtual block". +You can use it to allocate your own memory or other objects, even completely unrelated to Vulkan. +A common use case is sub-allocation of pieces of one large GPU buffer. + +\section virtual_allocator_creating_virtual_block Creating virtual block + +To use this functionality, there is no main "allocator" object. +You don't need to have #VmaAllocator object created. +All you need to do is to create a separate #VmaVirtualBlock object for each block of memory you want to be managed by the allocator: + +-# Fill in #VmaVirtualBlockCreateInfo structure. +-# Call vmaCreateVirtualBlock(). Get new #VmaVirtualBlock object. + +Example: + +\code +VmaVirtualBlockCreateInfo blockCreateInfo = {}; +blockCreateInfo.size = 1048576; // 1 MB + +VmaVirtualBlock block; +VkResult res = vmaCreateVirtualBlock(&blockCreateInfo, &block); +\endcode + +\section virtual_allocator_making_virtual_allocations Making virtual allocations + +#VmaVirtualBlock object contains internal data structure that keeps track of free and occupied regions +using the same code as the main Vulkan memory allocator. +Similarly to #VmaAllocation for standard GPU allocations, there is #VmaVirtualAllocation type +that represents an opaque handle to an allocation withing the virtual block. + +In order to make such allocation: + +-# Fill in #VmaVirtualAllocationCreateInfo structure. +-# Call vmaVirtualAllocate(). Get new #VmaVirtualAllocation object that represents the allocation. + You can also receive `VkDeviceSize offset` that was assigned to the allocation. + +Example: + +\code +VmaVirtualAllocationCreateInfo allocCreateInfo = {}; +allocCreateInfo.size = 4096; // 4 KB + +VmaVirtualAllocation alloc; +VkDeviceSize offset; +res = vmaVirtualAllocate(block, &allocCreateInfo, &alloc, &offset); +if(res == VK_SUCCESS) +{ + // Use the 4 KB of your memory starting at offset. +} +else +{ + // Allocation failed - no space for it could be found. Handle this error! +} +\endcode + +\section virtual_allocator_deallocation Deallocation + +When no longer needed, an allocation can be freed by calling vmaVirtualFree(). +You can only pass to this function an allocation that was previously returned by vmaVirtualAllocate() +called for the same #VmaVirtualBlock. + +When whole block is no longer needed, the block object can be released by calling vmaDestroyVirtualBlock(). +All allocations must be freed before the block is destroyed, which is checked internally by an assert. +However, if you don't want to call vmaVirtualFree() for each allocation, you can use vmaClearVirtualBlock() to free them all at once - +a feature not available in normal Vulkan memory allocator. Example: + +\code +vmaVirtualFree(block, alloc); +vmaDestroyVirtualBlock(block); +\endcode + +\section virtual_allocator_allocation_parameters Allocation parameters + +You can attach a custom pointer to each allocation by using vmaSetVirtualAllocationUserData(). +Its default value is null. +It can be used to store any data that needs to be associated with that allocation - e.g. an index, a handle, or a pointer to some +larger data structure containing more information. Example: + +\code +struct CustomAllocData +{ + std::string m_AllocName; +}; +CustomAllocData* allocData = new CustomAllocData(); +allocData->m_AllocName = "My allocation 1"; +vmaSetVirtualAllocationUserData(block, alloc, allocData); +\endcode + +The pointer can later be fetched, along with allocation offset and size, by passing the allocation handle to function +vmaGetVirtualAllocationInfo() and inspecting returned structure #VmaVirtualAllocationInfo. +If you allocated a new object to be used as the custom pointer, don't forget to delete that object before freeing the allocation! +Example: + +\code +VmaVirtualAllocationInfo allocInfo; +vmaGetVirtualAllocationInfo(block, alloc, &allocInfo); +delete (CustomAllocData*)allocInfo.pUserData; + +vmaVirtualFree(block, alloc); +\endcode + +\section virtual_allocator_alignment_and_units Alignment and units + +It feels natural to express sizes and offsets in bytes. +If an offset of an allocation needs to be aligned to a multiply of some number (e.g. 4 bytes), you can fill optional member +VmaVirtualAllocationCreateInfo::alignment to request it. Example: + +\code +VmaVirtualAllocationCreateInfo allocCreateInfo = {}; +allocCreateInfo.size = 4096; // 4 KB +allocCreateInfo.alignment = 4; // Returned offset must be a multiply of 4 B + +VmaVirtualAllocation alloc; +res = vmaVirtualAllocate(block, &allocCreateInfo, &alloc, nullptr); +\endcode + +Alignments of different allocations made from one block may vary. +However, if all alignments and sizes are always multiply of some size e.g. 4 B or `sizeof(MyDataStruct)`, +you can express all sizes, alignments, and offsets in multiples of that size instead of individual bytes. +It might be more convenient, but you need to make sure to use this new unit consistently in all the places: + +- VmaVirtualBlockCreateInfo::size +- VmaVirtualAllocationCreateInfo::size and VmaVirtualAllocationCreateInfo::alignment +- Using offset returned by vmaVirtualAllocate() or in VmaVirtualAllocationInfo::offset + +\section virtual_allocator_statistics Statistics + +You can obtain statistics of a virtual block using vmaCalculateVirtualBlockStats(). +The function fills structure #VmaStatInfo - same as used by the normal Vulkan memory allocator. +Example: + +\code +VmaStatInfo statInfo; +vmaCalculateVirtualBlockStats(block, &statInfo); +printf("My virtual block has %llu bytes used by %u virtual allocations\n", + statInfo.usedBytes, statInfo.allocationCount); +\endcode + +You can also request a full list of allocations and free regions as a string in JSON format by calling +vmaBuildVirtualBlockStatsString(). +Returned string must be later freed using vmaFreeVirtualBlockStatsString(). +The format of this string differs from the one returned by the main Vulkan allocator, but it is similar. + +\section virtual_allocator_additional_considerations Additional considerations + +The "virtual allocator" functionality is implemented on a level of individual memory blocks. +Keeping track of a whole collection of blocks, allocating new ones when out of free space, +deleting empty ones, and deciding which one to try first for a new allocation must be implemented by the user. + +Alternative allocation algorithms are supported, just like in custom pools of the real GPU memory. +See enum #VmaVirtualBlockCreateFlagBits to learn how to specify them (e.g. #VMA_VIRTUAL_BLOCK_CREATE_LINEAR_ALGORITHM_BIT). +You can find their description in chapter \ref custom_memory_pools. +Allocation strategies are also supported. +See enum #VmaVirtualAllocationCreateFlagBits to learn how to specify them (e.g. #VMA_VIRTUAL_ALLOCATION_CREATE_STRATEGY_MIN_TIME_BIT). + +Following features are supported only by the allocator of the real GPU memory and not by virtual allocations: +buffer-image granularity, `VMA_DEBUG_MARGIN`, `VMA_MIN_ALIGNMENT`. + + +\page debugging_memory_usage Debugging incorrect memory usage + +If you suspect a bug with memory usage, like usage of uninitialized memory or +memory being overwritten out of bounds of an allocation, +you can use debug features of this library to verify this. + +\section debugging_memory_usage_initialization Memory initialization + +If you experience a bug with incorrect and nondeterministic data in your program and you suspect uninitialized memory to be used, +you can enable automatic memory initialization to verify this. +To do it, define macro `VMA_DEBUG_INITIALIZE_ALLOCATIONS` to 1. + +\code +#define VMA_DEBUG_INITIALIZE_ALLOCATIONS 1 +#include "vk_mem_alloc.h" +\endcode + +It makes memory of all new allocations initialized to bit pattern `0xDCDCDCDC`. +Before an allocation is destroyed, its memory is filled with bit pattern `0xEFEFEFEF`. +Memory is automatically mapped and unmapped if necessary. + +If you find these values while debugging your program, good chances are that you incorrectly +read Vulkan memory that is allocated but not initialized, or already freed, respectively. + +Memory initialization works only with memory types that are `HOST_VISIBLE`. +It works also with dedicated allocations. + +\section debugging_memory_usage_margins Margins + +By default, allocations are laid out in memory blocks next to each other if possible +(considering required alignment, `bufferImageGranularity`, and `nonCoherentAtomSize`). + + + +Define macro `VMA_DEBUG_MARGIN` to some non-zero value (e.g. 16) to enforce specified +number of bytes as a margin after every allocation. + +\code +#define VMA_DEBUG_MARGIN 16 +#include "vk_mem_alloc.h" +\endcode + + + +If your bug goes away after enabling margins, it means it may be caused by memory +being overwritten outside of allocation boundaries. It is not 100% certain though. +Change in application behavior may also be caused by different order and distribution +of allocations across memory blocks after margins are applied. + +Margins work with all types of memory. + +Margin is applied only to allocations made out of memory blocks and not to dedicated +allocations, which have their own memory block of specific size. +It is thus not applied to allocations made using #VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT flag +or those automatically decided to put into dedicated allocations, e.g. due to its +large size or recommended by VK_KHR_dedicated_allocation extension. +Margins are also not active in custom pools created with #VMA_POOL_CREATE_BUDDY_ALGORITHM_BIT flag. + +Margins appear in [JSON dump](@ref statistics_json_dump) as part of free space. + +Note that enabling margins increases memory usage and fragmentation. + +Margins do not apply to \ref virtual_allocator. + +\section debugging_memory_usage_corruption_detection Corruption detection + +You can additionally define macro `VMA_DEBUG_DETECT_CORRUPTION` to 1 to enable validation +of contents of the margins. + +\code +#define VMA_DEBUG_MARGIN 16 +#define VMA_DEBUG_DETECT_CORRUPTION 1 +#include "vk_mem_alloc.h" +\endcode + +When this feature is enabled, number of bytes specified as `VMA_DEBUG_MARGIN` +(it must be multiply of 4) after every allocation is filled with a magic number. +This idea is also know as "canary". +Memory is automatically mapped and unmapped if necessary. + +This number is validated automatically when the allocation is destroyed. +If it is not equal to the expected value, `VMA_ASSERT()` is executed. +It clearly means that either CPU or GPU overwritten the memory outside of boundaries of the allocation, +which indicates a serious bug. + +You can also explicitly request checking margins of all allocations in all memory blocks +that belong to specified memory types by using function vmaCheckCorruption(), +or in memory blocks that belong to specified custom pool, by using function +vmaCheckPoolCorruption(). + +Margin validation (corruption detection) works only for memory types that are +`HOST_VISIBLE` and `HOST_COHERENT`. + + +\page opengl_interop OpenGL Interop + +VMA provides some features that help with interoperability with OpenGL. + +\section opengl_interop_exporting_memory Exporting memory + +If you want to attach `VkExportMemoryAllocateInfoKHR` structure to `pNext` chain of memory allocations made by the library: + +It is recommended to create \ref custom_memory_pools for such allocations. +Define and fill in your `VkExportMemoryAllocateInfoKHR` structure and attach it to VmaPoolCreateInfo::pMemoryAllocateNext +while creating the custom pool. +Please note that the structure must remain alive and unchanged for the whole lifetime of the #VmaPool, +not only while creating it, as no copy of the structure is made, +but its original pointer is used for each allocation instead. + +If you want to export all memory allocated by the library from certain memory types, +also dedicated allocations or other allocations made from default pools, +an alternative solution is to fill in VmaAllocatorCreateInfo::pTypeExternalMemoryHandleTypes. +It should point to an array with `VkExternalMemoryHandleTypeFlagsKHR` to be automatically passed by the library +through `VkExportMemoryAllocateInfoKHR` on each allocation made from a specific memory type. +Please note that new versions of the library also support dedicated allocations created in custom pools. + +You should not mix these two methods in a way that allows to apply both to the same memory type. +Otherwise, `VkExportMemoryAllocateInfoKHR` structure would be attached twice to the `pNext` chain of `VkMemoryAllocateInfo`. + + +\section opengl_interop_custom_alignment Custom alignment + +Buffers or images exported to a different API like OpenGL may require a different alignment, +higher than the one used by the library automatically, queried from functions like `vkGetBufferMemoryRequirements`. +To impose such alignment: + +It is recommended to create \ref custom_memory_pools for such allocations. +Set VmaPoolCreateInfo::minAllocationAlignment member to the minimum alignment required for each allocation +to be made out of this pool. +The alignment actually used will be the maximum of this member and the alignment returned for the specific buffer or image +from a function like `vkGetBufferMemoryRequirements`, which is called by VMA automatically. + +If you want to create a buffer with a specific minimum alignment out of default pools, +use special function vmaCreateBufferWithAlignment(), which takes additional parameter `minAlignment`. + +Note the problem of alignment affects only resources placed inside bigger `VkDeviceMemory` blocks and not dedicated +allocations, as these, by definition, always have alignment = 0 because the resource is bound to the beginning of its dedicated block. +Contrary to Direct3D 12, Vulkan doesn't have a concept of alignment of the entire memory block passed on its allocation. + + +\page usage_patterns Recommended usage patterns + +See also slides from talk: +[Sawicki, Adam. Advanced Graphics Techniques Tutorial: Memory management in Vulkan and DX12. Game Developers Conference, 2018](https://www.gdcvault.com/play/1025458/Advanced-Graphics-Techniques-Tutorial-New) + + +\section usage_patterns_common_mistakes Common mistakes + +<b>Use of CPU_TO_GPU instead of CPU_ONLY memory</b> + +#VMA_MEMORY_USAGE_CPU_TO_GPU is recommended only for resources that will be +mapped and written by the CPU, as well as read directly by the GPU - like some +buffers or textures updated every frame (dynamic). If you create a staging copy +of a resource to be written by CPU and then used as a source of transfer to +another resource placed in the GPU memory, that staging resource should be +created with #VMA_MEMORY_USAGE_CPU_ONLY. Please read the descriptions of these +enums carefully for details. + +<b>Unnecessary use of custom pools</b> + +\ref custom_memory_pools may be useful for special purposes - when you want to +keep certain type of resources separate e.g. to reserve minimum amount of memory +for them or limit maximum amount of memory they can occupy. For most +resources this is not needed and so it is not recommended to create #VmaPool +objects and allocations out of them. Allocating from the default pool is sufficient. + +\section usage_patterns_simple Simple patterns + +\subsection usage_patterns_simple_render_targets Render targets + +<b>When:</b> +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)"). + +<b>What to do:</b> +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 + +<b>When:</b> +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. + +<b>What to do:</b> +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 + +<b>When:</b> +Any resources that change frequently (aka "dynamic"), e.g. every frame or every draw call, +written on CPU, read on GPU. + +<b>What to do:</b> +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 on CPU, as it may be very slow. +Also note that textures written directly from the host through a mapped pointer need to be in LINEAR not OPTIMAL layout. + +\subsection usage_patterns_readback Readback + +<b>When:</b> +Resources that contain data written by GPU that you want to read back on CPU, +e.g. results of some computations. + +<b>What to do:</b> +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 comparably 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 transfer each time. +-# Create just a single copy using #VMA_MEMORY_USAGE_CPU_TO_GPU, map it and fill it on CPU, + read it directly on GPU. +-# Create just a 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 unnecessary time and memory overhead + related to using a second copy and making transfer. +- 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. + +Note that textures accessed directly from the host through a mapped pointer need to be in LINEAR layout, +which may slow down their usage on the device. +Textures accessed only by the device and transfer operations can use OPTIMAL layout. + +If you don't want to specialize your code for specific types of GPUs, you 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. + +For example, define `VMA_ASSERT(expr)` before including the library to provide +custom implementation of the assertion, compatible with your project. +By default it is defined to standard C `assert(expr)` in `_DEBUG` configuration +and empty otherwise. + +\section config_Vulkan_functions Pointers to Vulkan functions + +There are multiple ways to import pointers to Vulkan functions in the library. +In the simplest case you don't need to do anything. +If the compilation or linking of your program or the initialization of the #VmaAllocator +doesn't work for you, you can try to reconfigure it. + +First, the allocator tries to fetch pointers to Vulkan functions linked statically, +like this: + +\code +m_VulkanFunctions.vkAllocateMemory = (PFN_vkAllocateMemory)vkAllocateMemory; +\endcode + +If you want to disable this feature, set configuration macro: `#define VMA_STATIC_VULKAN_FUNCTIONS 0`. + +Second, you can provide the pointers yourself by setting member VmaAllocatorCreateInfo::pVulkanFunctions. +You can fetch them e.g. using functions `vkGetInstanceProcAddr` and `vkGetDeviceProcAddr` or +by using a helper library like [volk](https://github.com/zeux/volk). + +Third, VMA tries to fetch remaining pointers that are still null by calling +`vkGetInstanceProcAddr` and `vkGetDeviceProcAddr` on its own. +If you want to disable this feature, set configuration macro: `#define VMA_DYNAMIC_VULKAN_FUNCTIONS 0`. + +Finally, all the function pointers required by the library (considering selected +Vulkan version and enabled extensions) are checked with `VMA_ASSERT` if they are not null. + + +\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 + +When device memory of certain heap runs out of free space, new allocations may +fail (returning error code) or they may succeed, silently pushing some existing +memory blocks from GPU VRAM to system RAM (which degrades performance). This +behavior is implementation-dependent - it depends on GPU vendor and graphics +driver. + +On AMD cards it can be controlled while creating Vulkan device object by using +VK_AMD_memory_overallocation_behavior extension, if available. + +Alternatively, 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 #VmaAllocator`to 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 is 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/1.2-extensions/html/chap50.html#VK_KHR_dedicated_allocation) +- [VK_KHR_dedicated_allocation unofficial manual](http://asawicki.info/articles/VK_KHR_dedicated_allocation.php5) + + + +\page vk_amd_device_coherent_memory VK_AMD_device_coherent_memory + +VK_AMD_device_coherent_memory is a device extension that enables access to +additional memory types with `VK_MEMORY_PROPERTY_DEVICE_COHERENT_BIT_AMD` and +`VK_MEMORY_PROPERTY_DEVICE_UNCACHED_BIT_AMD` flag. It is useful mostly for +allocation of buffers intended for writing "breadcrumb markers" in between passes +or draw calls, which in turn are useful for debugging GPU crash/hang/TDR cases. + +When the extension is available but has not been enabled, Vulkan physical device +still exposes those memory types, but their usage is forbidden. VMA automatically +takes care of that - it returns `VK_ERROR_FEATURE_NOT_PRESENT` when an attempt +to allocate memory of such type is made. + +If you want to use this extension in connection with VMA, follow these steps: + +\section vk_amd_device_coherent_memory_initialization Initialization + +1) Call `vkEnumerateDeviceExtensionProperties` for the physical device. +Check if the extension is supported - if returned array of `VkExtensionProperties` contains "VK_AMD_device_coherent_memory". + +2) Call `vkGetPhysicalDeviceFeatures2` for the physical device instead of old `vkGetPhysicalDeviceFeatures`. +Attach additional structure `VkPhysicalDeviceCoherentMemoryFeaturesAMD` to `VkPhysicalDeviceFeatures2::pNext` to be returned. +Check if the device feature is really supported - check if `VkPhysicalDeviceCoherentMemoryFeaturesAMD::deviceCoherentMemory` is true. + +3) While creating device with `vkCreateDevice`, enable this extension - add "VK_AMD_device_coherent_memory" +to the list passed as `VkDeviceCreateInfo::ppEnabledExtensionNames`. + +4) While creating the device, also don't set `VkDeviceCreateInfo::pEnabledFeatures`. +Fill in `VkPhysicalDeviceFeatures2` structure instead and pass it as `VkDeviceCreateInfo::pNext`. +Enable this device feature - attach additional structure `VkPhysicalDeviceCoherentMemoryFeaturesAMD` to +`VkPhysicalDeviceFeatures2::pNext` and set its member `deviceCoherentMemory` to `VK_TRUE`. + +5) While creating #VmaAllocator with vmaCreateAllocator() inform VMA that you +have enabled this extension and feature - add #VMA_ALLOCATOR_CREATE_AMD_DEVICE_COHERENT_MEMORY_BIT +to VmaAllocatorCreateInfo::flags. + +\section vk_amd_device_coherent_memory_usage Usage + +After following steps described above, you can create VMA allocations and custom pools +out of the special `DEVICE_COHERENT` and `DEVICE_UNCACHED` memory types on eligible +devices. There are multiple ways to do it, for example: + +- You can request or prefer to allocate out of such memory types by adding + `VK_MEMORY_PROPERTY_DEVICE_COHERENT_BIT_AMD` to VmaAllocationCreateInfo::requiredFlags + or VmaAllocationCreateInfo::preferredFlags. Those flags can be freely mixed with + other ways of \ref choosing_memory_type, like setting VmaAllocationCreateInfo::usage. +- If you manually found memory type index to use for this purpose, force allocation + from this specific index by setting VmaAllocationCreateInfo::memoryTypeBits `= 1u << index`. + +\section vk_amd_device_coherent_memory_more_information More information + +To learn more about this extension, see [VK_AMD_device_coherent_memory in Vulkan specification](https://www.khronos.org/registry/vulkan/specs/1.2-extensions/man/html/VK_AMD_device_coherent_memory.html) + +Example use of this extension can be found in the code of the sample and test suite +accompanying this library. + + +\page enabling_buffer_device_address Enabling buffer device address + +Device extension VK_KHR_buffer_device_address +allow to fetch raw GPU pointer to a buffer and pass it for usage in a shader code. +It is promoted to core Vulkan 1.2. + +If you want to use this feature in connection with VMA, follow these steps: + +\section enabling_buffer_device_address_initialization Initialization + +1) (For Vulkan version < 1.2) Call `vkEnumerateDeviceExtensionProperties` for the physical device. +Check if the extension is supported - if returned array of `VkExtensionProperties` contains +"VK_KHR_buffer_device_address". + +2) Call `vkGetPhysicalDeviceFeatures2` for the physical device instead of old `vkGetPhysicalDeviceFeatures`. +Attach additional structure `VkPhysicalDeviceBufferDeviceAddressFeatures*` to `VkPhysicalDeviceFeatures2::pNext` to be returned. +Check if the device feature is really supported - check if `VkPhysicalDeviceBufferDeviceAddressFeatures::bufferDeviceAddress` is true. + +3) (For Vulkan version < 1.2) While creating device with `vkCreateDevice`, enable this extension - add +"VK_KHR_buffer_device_address" to the list passed as `VkDeviceCreateInfo::ppEnabledExtensionNames`. + +4) While creating the device, also don't set `VkDeviceCreateInfo::pEnabledFeatures`. +Fill in `VkPhysicalDeviceFeatures2` structure instead and pass it as `VkDeviceCreateInfo::pNext`. +Enable this device feature - attach additional structure `VkPhysicalDeviceBufferDeviceAddressFeatures*` to +`VkPhysicalDeviceFeatures2::pNext` and set its member `bufferDeviceAddress` to `VK_TRUE`. + +5) While creating #VmaAllocator with vmaCreateAllocator() inform VMA that you +have enabled this feature - add #VMA_ALLOCATOR_CREATE_BUFFER_DEVICE_ADDRESS_BIT +to VmaAllocatorCreateInfo::flags. + +\section enabling_buffer_device_address_usage Usage + +After following steps described above, you can create buffers with `VK_BUFFER_USAGE_SHADER_DEVICE_ADDRESS_BIT*` using VMA. +The library automatically adds `VK_MEMORY_ALLOCATE_DEVICE_ADDRESS_BIT*` to +allocated memory blocks wherever it might be needed. + +Please note that the library supports only `VK_BUFFER_USAGE_SHADER_DEVICE_ADDRESS_BIT*`. +The second part of this functionality related to "capture and replay" is not supported, +as it is intended for usage in debugging tools like RenderDoc, not in everyday Vulkan usage. + +\section enabling_buffer_device_address_more_information More information + +To learn more about this extension, see [VK_KHR_buffer_device_address in Vulkan specification](https://www.khronos.org/registry/vulkan/specs/1.2-extensions/html/chap46.html#VK_KHR_buffer_device_address) + +Example use of this extension can be found in the code of the sample and test suite +accompanying this library. + +\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. + This includes allocation and deallocation from default memory pool, as well as custom #VmaPool. +- 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. +- #VmaVirtualBlock is also not safe to be used from multiple threads simultaneously. + +\section general_considerations_validation_layer_warnings Validation layer warnings + +When using this library, you can meet following types of warnings issued by +Vulkan validation layer. They don't necessarily indicate a bug, so you may need +to just ignore them. + +- *vkBindBufferMemory(): Binding memory to buffer 0xeb8e4 but vkGetBufferMemoryRequirements() has not been called on that buffer.* + - It happens when VK_KHR_dedicated_allocation extension is enabled. + `vkGetBufferMemoryRequirements2KHR` function is used instead, while validation layer seems to be unaware of it. +- *Mapping an image with layout VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL can result in undefined behavior if this memory is used by the device. Only GENERAL or PREINITIALIZED should be used.* + - It happens when you map a buffer or image, because the library maps entire + `VkDeviceMemory` block, where different types of images and buffers may end + up together, especially on GPUs with unified memory like Intel. +- *Non-linear image 0xebc91 is aliased with linear buffer 0xeb8e4 which may indicate a bug.* + - It may happen when you use [defragmentation](@ref defragmentation). + +\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, 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.** Uploading (streaming) and downloading data of buffers and images + between CPU and GPU memory and related synchronization is responsibility of the user. + Defining some "texture" object that would automatically stream its data from a + staging copy in CPU memory to GPU memory would rather be a feature of another, + higher-level library implemented on top of VMA. +- **Recreation of buffers and images.** Although the library has functions for + buffer and image creation (vmaCreateBuffer(), vmaCreateImage()), you need to + recreate these objects yourself after defragmentation. That is because the big + structures `VkBufferCreateInfo`, `VkImageCreateInfo` are not stored in + #VmaAllocation object. +- **Handling CPU memory allocation failures.** When dynamically creating small C++ + objects in CPU memory (not Vulkan memory), allocation failures are not checked + and handled gracefully, because that would complicate code significantly and + is usually not needed in desktop PC applications anyway. + Success of an allocation is just checked with an assert. +- **Code free of any compiler warnings.** Maintaining the library to compile and + work correctly on so many different platforms is hard enough. Being free of + any warnings, on any version of any compiler, is simply not feasible. + There are many preprocessor macros that make some variables unused, function parameters unreferenced, + or conditional expressions constant in some configurations. + The code of this library should not be bigger or more complicated just to silence these warnings. + It is recommended to disable such warnings instead. +- This is a C++ library with C interface. **Bindings or ports to any other programming languages** are welcome as external projects but + are not going to be included into this repository. +*/ |