AuroraMiddleware/VulkanMemoryAllocator
1
0
Fork 0
mirror of https://github.com/GPUOpen-LibrariesAndSDKs/VulkanMemoryAllocator synced 2024-12-26 10:51:01 +00:00
Code Issues Projects Releases Wiki Activity

Added support for multiple Vulkan memory blocks in custom pools with VMA_POOL_CREATE_LINEAR_ALGORITHM_BIT. Works with free-at-once and stack, doesn't work with double stack or ring buffer.

Browse Source 
Added new structure members VmaPoolStats::blockCount.
This commit is contained in:
Adam Sawicki 2018-08-24 15:36:32 +02:00
parent ee79c63d61
commit 70a683e53f
2 changed files with 324 additions and 120 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

150
src/Tests.cpp
View File

@ -1839,14 +1839,139 @@ static void TestLinearAllocator()
bufInfo.clear();
}
// Try to create pool with maxBlockCount higher than 1. It should fail.
{
VmaPoolCreateInfo altPoolCreateInfo = poolCreateInfo;
altPoolCreateInfo.maxBlockCount = 2;
vmaDestroyPool(g_hAllocator, pool);
}
VmaPool altPool = nullptr;
res = vmaCreatePool(g_hAllocator, &altPoolCreateInfo, &altPool);
assert(res != VK_SUCCESS);
static void TestLinearAllocatorMultiBlock()
{
wprintf(L"Test linear allocator multi block\n");
RandomNumberGenerator rand{345673};
VkBufferCreateInfo sampleBufCreateInfo = { VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO };
sampleBufCreateInfo.size = 1024 * 1024;
sampleBufCreateInfo.usage = VK_BUFFER_USAGE_TRANSFER_SRC_BIT;
VmaAllocationCreateInfo sampleAllocCreateInfo = {};
sampleAllocCreateInfo.usage = VMA_MEMORY_USAGE_CPU_ONLY;
VmaPoolCreateInfo poolCreateInfo = {};
poolCreateInfo.flags = VMA_POOL_CREATE_LINEAR_ALGORITHM_BIT;
VkResult res = vmaFindMemoryTypeIndexForBufferInfo(g_hAllocator, &sampleBufCreateInfo, &sampleAllocCreateInfo, &poolCreateInfo.memoryTypeIndex);
assert(res == VK_SUCCESS);
VmaPool pool = nullptr;
res = vmaCreatePool(g_hAllocator, &poolCreateInfo, &pool);
assert(res == VK_SUCCESS);
VkBufferCreateInfo bufCreateInfo = sampleBufCreateInfo;
VmaAllocationCreateInfo allocCreateInfo = {};
allocCreateInfo.pool = pool;
std::vector<BufferInfo> bufInfo;
VmaAllocationInfo allocInfo;
// Test one-time free.
{
// Allocate buffers until we move to a second block.
VkDeviceMemory lastMem = VK_NULL_HANDLE;
for(uint32_t i = 0; ; ++i)
{
BufferInfo newBufInfo;
res = vmaCreateBuffer(g_hAllocator, &bufCreateInfo, &allocCreateInfo,
&newBufInfo.Buffer, &newBufInfo.Allocation, &allocInfo);
assert(res == VK_SUCCESS);
bufInfo.push_back(newBufInfo);
if(lastMem && allocInfo.deviceMemory != lastMem)
{
break;
}
lastMem = allocInfo.deviceMemory;
}
assert(bufInfo.size() > 2);
// Make sure that pool has now two blocks.
VmaPoolStats poolStats = {};
vmaGetPoolStats(g_hAllocator, pool, &poolStats);
assert(poolStats.blockCount == 2);
// Destroy all the buffers in random order.
while(!bufInfo.empty())
{
const size_t indexToDestroy = rand.Generate() % bufInfo.size();
const BufferInfo& currBufInfo = bufInfo[indexToDestroy];
vmaDestroyBuffer(g_hAllocator, currBufInfo.Buffer, currBufInfo.Allocation);
bufInfo.erase(bufInfo.begin() + indexToDestroy);
}
// Make sure that pool has now at most one block.
vmaGetPoolStats(g_hAllocator, pool, &poolStats);
assert(poolStats.blockCount <= 1);
}
// Test stack.
{
// Allocate buffers until we move to a second block.
VkDeviceMemory lastMem = VK_NULL_HANDLE;
for(uint32_t i = 0; ; ++i)
{
BufferInfo newBufInfo;
res = vmaCreateBuffer(g_hAllocator, &bufCreateInfo, &allocCreateInfo,
&newBufInfo.Buffer, &newBufInfo.Allocation, &allocInfo);
assert(res == VK_SUCCESS);
bufInfo.push_back(newBufInfo);
if(lastMem && allocInfo.deviceMemory != lastMem)
{
break;
}
lastMem = allocInfo.deviceMemory;
}
assert(bufInfo.size() > 2);
// Add few more buffers.
for(uint32_t i = 0; i < 5; ++i)
{
BufferInfo newBufInfo;
res = vmaCreateBuffer(g_hAllocator, &bufCreateInfo, &allocCreateInfo,
&newBufInfo.Buffer, &newBufInfo.Allocation, &allocInfo);
assert(res == VK_SUCCESS);
bufInfo.push_back(newBufInfo);
}
// Make sure that pool has now two blocks.
VmaPoolStats poolStats = {};
vmaGetPoolStats(g_hAllocator, pool, &poolStats);
assert(poolStats.blockCount == 2);
// Delete half of buffers, LIFO.
for(size_t i = 0, countToDelete = bufInfo.size() / 2; i < countToDelete; ++i)
{
const BufferInfo& currBufInfo = bufInfo.back();
vmaDestroyBuffer(g_hAllocator, currBufInfo.Buffer, currBufInfo.Allocation);
bufInfo.pop_back();
}
// Add one more buffer.
BufferInfo newBufInfo;
res = vmaCreateBuffer(g_hAllocator, &bufCreateInfo, &allocCreateInfo,
&newBufInfo.Buffer, &newBufInfo.Allocation, &allocInfo);
assert(res == VK_SUCCESS);
bufInfo.push_back(newBufInfo);
// Make sure that pool has now one block.
vmaGetPoolStats(g_hAllocator, pool, &poolStats);
assert(poolStats.blockCount == 1);
// Delete all the remaining buffers, LIFO.
while(!bufInfo.empty())
{
const BufferInfo& currBufInfo = bufInfo.back();
vmaDestroyBuffer(g_hAllocator, currBufInfo.Buffer, currBufInfo.Allocation);
bufInfo.pop_back();
}
}
vmaDestroyPool(g_hAllocator, pool);
@ -3841,6 +3966,16 @@ void Test()
{
wprintf(L"TESTING:\n");
if(false)
{
// # Temporarily insert custom tests here
TestLinearAllocator();
ManuallyTestLinearAllocator();
TestLinearAllocatorMultiBlock();
BenchmarkLinearAllocator();
return;
}
// # Simple tests
TestBasics();
@ -3857,6 +3992,7 @@ void Test()
TestMappingMultithreaded();
TestLinearAllocator();
ManuallyTestLinearAllocator();
TestLinearAllocatorMultiBlock();
BenchmarkLinearAllocator();
TestDefragmentationSimple();
TestDefragmentationFull();

282
src/vk_mem_alloc.h
View File

@ -592,9 +592,6 @@ 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.
Pools with linear algorithm must have only one memory block -
VmaPoolCreateInfo::maxBlockCount must be 1.
\subsection linear_algorithm_free_at_once Free-at-once
In a pool that uses linear algorithm, you still need to free all the allocations
@ -607,6 +604,9 @@ to release all at once.
![Free-at-once](../gfx/Linear_allocator_3_free_at_once.png)
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.
@ -615,6 +615,9 @@ creation (LIFO - Last In First Out), you can achieve behavior of a stack.
![Stack](../gfx/Linear_allocator_4_stack.png)
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
@ -626,12 +629,15 @@ stacks:
To make allocation from upper stack, add flag #VMA_ALLOCATION_CREATE_UPPER_ADDRESS_BIT
to VmaAllocationCreateInfo::flags.
![Double stack](../gfx/Linear_allocator_7_double_stack.png)
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.
![Double stack](../gfx/Linear_allocator_7_double_stack.png)
\subsection linear_algorithm_ring_buffer Ring buffer
When you free some allocations from the beginning and there is not enough free space
@ -649,6 +655,9 @@ succeeds.
![Ring buffer with lost allocations](../gfx/Linear_allocator_6_ring_buffer_lost.png)
Ring buffer is available only in pools with one memory block -
VmaPoolCreateInfo::maxBlockCount must be 1. Otherwise behavior is undefined.
\page defragmentation Defragmentation
@ -1968,7 +1977,6 @@ typedef struct VmaPoolCreateInfo {
/** \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.
When #VMA_POOL_CREATE_LINEAR_ALGORITHM_BIT is used, default is 1.
Set to same value as VmaPoolCreateInfo::minBlockCount to have fixed amount of memory allocated
throughout whole lifetime of this pool.
@ -2005,13 +2013,16 @@ typedef struct VmaPoolStats {
/** \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.
/** \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.
@ -4506,6 +4517,7 @@ public:
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
@ -5014,6 +5026,18 @@ private:
// after this call.
void IncrementallySortBlocks();
// To be used only without CAN_MAKE_OTHER_LOST flag.
VkResult AllocateFromBlock(
VmaDeviceMemoryBlock* pBlock,
VmaPool hCurrentPool,
uint32_t currentFrameIndex,
VkDeviceSize size,
VkDeviceSize alignment,
VmaAllocationCreateFlags allocFlags,
void* pUserData,
VmaSuballocationType suballocType,
VmaAllocation* pAllocation);
VkResult CreateBlock(VkDeviceSize blockSize, size_t* pNewBlockIndex);
};
@ -9374,15 +9398,18 @@ VkResult VmaBlockVector::CreateMinBlocks()
void VmaBlockVector::GetPoolStats(VmaPoolStats* pStats)
{
VmaMutexLock lock(m_Mutex, m_hAllocator->m_UseMutex);
const size_t blockCount = m_Blocks.size();
pStats->size = 0;
pStats->unusedSize = 0;
pStats->allocationCount = 0;
pStats->unusedRangeCount = 0;
pStats->unusedRangeSizeMax = 0;
pStats->blockCount = blockCount;
VmaMutexLock lock(m_Mutex, m_hAllocator->m_UseMutex);
for(uint32_t blockIndex = 0; blockIndex < m_Blocks.size(); ++blockIndex)
for(uint32_t blockIndex = 0; blockIndex < blockCount; ++blockIndex)
{
const VmaDeviceMemoryBlock* const pBlock = m_Blocks[blockIndex];
VMA_ASSERT(pBlock);
@ -9411,15 +9438,23 @@ VkResult VmaBlockVector::Allocate(
VmaAllocation* pAllocation)
{
const bool isUpperAddress = (createInfo.flags & VMA_ALLOCATION_CREATE_UPPER_ADDRESS_BIT) != 0;
const bool canMakeOtherLost = (createInfo.flags & VMA_ALLOCATION_CREATE_CAN_MAKE_OTHER_LOST_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;
const bool canCreateNewBlock =
((createInfo.flags & VMA_ALLOCATION_CREATE_NEVER_ALLOCATE_BIT) == 0) &&
(m_Blocks.size() < m_MaxBlockCount);
// Upper address can only be used with linear allocator.
if(isUpperAddress && !m_LinearAlgorithm)
// If linearAlgorithm is used, canMakeOtherLost is available only when used as ring buffer.
// Which in turn is available only when maxBlockCount = 1.
if(m_LinearAlgorithm && m_MaxBlockCount > 1)
{
canMakeOtherLost = false;
}
// Upper address can only be used with linear allocator and within single memory block.
if(isUpperAddress &&
(!m_LinearAlgorithm || m_MaxBlockCount > 1))
{
return VK_ERROR_FEATURE_NOT_PRESENT;
}
@ -9440,67 +9475,57 @@ VkResult VmaBlockVector::Allocate(
if(!canMakeOtherLost || canCreateNewBlock)
{
// 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_LinearAlgorithm)
{
// Use only last block.
if(!m_Blocks.empty())
{
VmaDeviceMemoryBlock* const pCurrBlock = m_Blocks.back();
VMA_ASSERT(pCurrBlock);
VkResult res = AllocateFromBlock(
pCurrBlock,
hCurrentPool,
currentFrameIndex,
size,
alignment,
allocFlagsCopy,
createInfo.pUserData,
suballocType,
pAllocation);
if(res == VK_SUCCESS)
{
VMA_DEBUG_LOG(" Returned from last block #%u", (uint32_t)(m_Blocks.size() - 1));
return VK_SUCCESS;
}
}
}
else
{
// 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);
VmaAllocationRequest currRequest = {};
if(pCurrBlock->m_pMetadata->CreateAllocationRequest(
currentFrameIndex,
m_FrameInUseCount,
m_BufferImageGranularity,
size,
alignment,
isUpperAddress,
suballocType,
false, // canMakeOtherLost
&currRequest))
{
// Allocate from pCurrBlock.
VMA_ASSERT(currRequest.itemsToMakeLostCount == 0);
if(mapped)
{
VkResult res = pCurrBlock->Map(m_hAllocator, 1, VMA_NULL);
if(res != VK_SUCCESS)
{
return res;
}
}
// We no longer have an empty Allocation.
if(pCurrBlock->m_pMetadata->IsEmpty())
{
m_HasEmptyBlock = false;
}
*pAllocation = vma_new(m_hAllocator, VmaAllocation_T)(currentFrameIndex, isUserDataString);
pCurrBlock->m_pMetadata->Alloc(currRequest, suballocType, size, isUpperAddress, *pAllocation);
(*pAllocation)->InitBlockAllocation(
hCurrentPool,
VkResult res = AllocateFromBlock(
pCurrBlock,
currRequest.offset,
alignment,
hCurrentPool,
currentFrameIndex,
size,
alignment,
allocFlagsCopy,
createInfo.pUserData,
suballocType,
mapped,
(createInfo.flags & VMA_ALLOCATION_CREATE_CAN_BECOME_LOST_BIT) != 0);
VMA_HEAVY_ASSERT(pCurrBlock->Validate());
VMA_DEBUG_LOG(" Returned from existing allocation #%u", (uint32_t)blockIndex);
(*pAllocation)->SetUserData(m_hAllocator, createInfo.pUserData);
if(VMA_DEBUG_INITIALIZE_ALLOCATIONS)
pAllocation);
if(res == VK_SUCCESS)
{
m_hAllocator->FillAllocation(*pAllocation, VMA_ALLOCATION_FILL_PATTERN_CREATED);
}
if(IsCorruptionDetectionEnabled())
{
VkResult res = pCurrBlock->WriteMagicValueAroundAllocation(m_hAllocator, currRequest.offset, size);
VMA_ASSERT(res == VK_SUCCESS && "Couldn't map block memory to write magic value.");
}
VMA_DEBUG_LOG(" Returned from existing block #%u", (uint32_t)blockIndex);
return VK_SUCCESS;
}
}
}
// 2. Try to create new block.
if(canCreateNewBlock)
@ -9555,56 +9580,24 @@ VkResult VmaBlockVector::Allocate(
VmaDeviceMemoryBlock* const pBlock = m_Blocks[newBlockIndex];
VMA_ASSERT(pBlock->m_pMetadata->GetSize() >= size);
if(mapped)
{
res = pBlock->Map(m_hAllocator, 1, VMA_NULL);
if(res != VK_SUCCESS)
{
return res;
}
}
// Allocate from pBlock. Because it is empty, dstAllocRequest can be trivially filled.
VmaAllocationRequest allocRequest;
if(pBlock->m_pMetadata->CreateAllocationRequest(
currentFrameIndex,
m_FrameInUseCount,
m_BufferImageGranularity,
size,
alignment,
isUpperAddress,
suballocType,
false, // canMakeOtherLost
&allocRequest))
{
*pAllocation = vma_new(m_hAllocator, VmaAllocation_T)(currentFrameIndex, isUserDataString);
pBlock->m_pMetadata->Alloc(allocRequest, suballocType, size, isUpperAddress, *pAllocation);
(*pAllocation)->InitBlockAllocation(
hCurrentPool,
res = AllocateFromBlock(
pBlock,
allocRequest.offset,
alignment,
hCurrentPool,
currentFrameIndex,
size,
alignment,
allocFlagsCopy,
createInfo.pUserData,
suballocType,
mapped,
(createInfo.flags & VMA_ALLOCATION_CREATE_CAN_BECOME_LOST_BIT) != 0);
VMA_HEAVY_ASSERT(pBlock->Validate());
VMA_DEBUG_LOG(" Created new allocation Size=%llu", allocInfo.allocationSize);
(*pAllocation)->SetUserData(m_hAllocator, createInfo.pUserData);
if(VMA_DEBUG_INITIALIZE_ALLOCATIONS)
pAllocation);
if(res == VK_SUCCESS)
{
m_hAllocator->FillAllocation(*pAllocation, VMA_ALLOCATION_FILL_PATTERN_CREATED);
}
if(IsCorruptionDetectionEnabled())
{
res = pBlock->WriteMagicValueAroundAllocation(m_hAllocator, allocRequest.offset, size);
VMA_ASSERT(res == VK_SUCCESS && "Couldn't map block memory to write magic value.");
}
VMA_DEBUG_LOG(" Created new block Size=%llu", newBlockSize);
return VK_SUCCESS;
}
else
{
// Allocation from empty block failed, possibly due to VMA_DEBUG_MARGIN or alignment.
// Allocation from new block failed, possibly due to VMA_DEBUG_MARGIN or alignment.
return VK_ERROR_OUT_OF_DEVICE_MEMORY;
}
}
@ -9819,6 +9812,8 @@ void VmaBlockVector::Remove(VmaDeviceMemoryBlock* pBlock)
void VmaBlockVector::IncrementallySortBlocks()
{
if(!m_LinearAlgorithm)
{
// Bubble sort only until first swap.
for(size_t i = 1; i < m_Blocks.size(); ++i)
{
@ -9828,6 +9823,80 @@ void VmaBlockVector::IncrementallySortBlocks()
return;
}
}
}
}
VkResult VmaBlockVector::AllocateFromBlock(
VmaDeviceMemoryBlock* pBlock,
VmaPool hCurrentPool,
uint32_t currentFrameIndex,
VkDeviceSize size,
VkDeviceSize alignment,
VmaAllocationCreateFlags allocFlags,
void* pUserData,
VmaSuballocationType suballocType,
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,
size,
alignment,
isUpperAddress,
suballocType,
false, // canMakeOtherLost
&currRequest))
{
// Allocate from pCurrBlock.
VMA_ASSERT(currRequest.itemsToMakeLostCount == 0);
if(mapped)
{
VkResult res = pBlock->Map(m_hAllocator, 1, VMA_NULL);
if(res != VK_SUCCESS)
{
return res;
}
}
// We no longer have an empty Allocation.
if(pBlock->m_pMetadata->IsEmpty())
{
m_HasEmptyBlock = false;
}
*pAllocation = vma_new(m_hAllocator, VmaAllocation_T)(currentFrameIndex, isUserDataString);
pBlock->m_pMetadata->Alloc(currRequest, suballocType, size, isUpperAddress, *pAllocation);
(*pAllocation)->InitBlockAllocation(
hCurrentPool,
pBlock,
currRequest.offset,
alignment,
size,
suballocType,
mapped,
(allocFlags & VMA_ALLOCATION_CREATE_CAN_BECOME_LOST_BIT) != 0);
VMA_HEAVY_ASSERT(pBlock->Validate());
(*pAllocation)->SetUserData(m_hAllocator, pUserData);
if(VMA_DEBUG_INITIALIZE_ALLOCATIONS)
{
m_hAllocator->FillAllocation(*pAllocation, VMA_ALLOCATION_FILL_PATTERN_CREATED);
}
if(IsCorruptionDetectionEnabled())
{
VkResult res = pBlock->WriteMagicValueAroundAllocation(m_hAllocator, currRequest.offset, size);
VMA_ASSERT(res == VK_SUCCESS && "Couldn't map block memory to write magic value.");
}
return VK_SUCCESS;
}
return VK_ERROR_OUT_OF_DEVICE_MEMORY;
}
VkResult VmaBlockVector::CreateBlock(VkDeviceSize blockSize, size_t* pNewBlockIndex)
@ -11812,10 +11881,9 @@ VkResult VmaAllocator_T::CreatePool(const VmaPoolCreateInfo* pCreateInfo, VmaPoo
if(newCreateInfo.maxBlockCount == 0)
{
newCreateInfo.maxBlockCount = isLinearAlgorithm ? 1 : SIZE_MAX;
newCreateInfo.maxBlockCount = SIZE_MAX;
}
if(newCreateInfo.minBlockCount > newCreateInfo.maxBlockCount ||
isLinearAlgorithm && newCreateInfo.maxBlockCount > 1)
if(newCreateInfo.minBlockCount > newCreateInfo.maxBlockCount)
{
return VK_ERROR_INITIALIZATION_FAILED;
}