Merge pull request #8987 from liamwhite/another-name-for-reinforcement-steel

vulkan: automatically use larger staging buffer sizes when possible
merge-requests/60/head
Fernando S 2022-10-05 08:54:22 +07:00 committed by GitHub
commit 71fe9fd0f2
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
2 changed files with 60 additions and 27 deletions

@ -26,20 +26,39 @@ using namespace Common::Literals;
constexpr VkDeviceSize MAX_ALIGNMENT = 256;
// Maximum size to put elements in the stream buffer
constexpr VkDeviceSize MAX_STREAM_BUFFER_REQUEST_SIZE = 8_MiB;
// Stream buffer size in bytes
constexpr VkDeviceSize STREAM_BUFFER_SIZE = 128_MiB;
constexpr VkDeviceSize REGION_SIZE = STREAM_BUFFER_SIZE / StagingBufferPool::NUM_SYNCS;
constexpr VkMemoryPropertyFlags HOST_FLAGS =
VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT | VK_MEMORY_PROPERTY_HOST_COHERENT_BIT;
constexpr VkMemoryPropertyFlags STREAM_FLAGS = VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT | HOST_FLAGS;
bool IsStreamHeap(VkMemoryHeap heap) noexcept {
return STREAM_BUFFER_SIZE < (heap.size * 2) / 3;
static bool IsStreamHeap(VkMemoryHeap heap, size_t staging_buffer_size) noexcept {
return staging_buffer_size < (heap.size * 2) / 3;
}
static bool HasLargeDeviceLocalHostVisibleMemory(const VkPhysicalDeviceMemoryProperties& props) {
const auto flags{VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT | VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT};
for (u32 type_index = 0; type_index < props.memoryTypeCount; ++type_index) {
const auto& memory_type{props.memoryTypes[type_index]};
if ((memory_type.propertyFlags & flags) != flags) {
// Memory must be device local and host visible
continue;
}
const auto& heap{props.memoryHeaps[memory_type.heapIndex]};
if (heap.size >= 7168_MiB) {
// This is the right type of memory
return true;
}
}
return false;
}
std::optional<u32> FindMemoryTypeIndex(const VkPhysicalDeviceMemoryProperties& props, u32 type_mask,
VkMemoryPropertyFlags flags) noexcept {
VkMemoryPropertyFlags flags,
size_t staging_buffer_size) noexcept {
for (u32 type_index = 0; type_index < props.memoryTypeCount; ++type_index) {
if (((type_mask >> type_index) & 1) == 0) {
// Memory type is incompatible
@ -50,7 +69,7 @@ std::optional<u32> FindMemoryTypeIndex(const VkPhysicalDeviceMemoryProperties& p
// Memory type doesn't have the flags we want
continue;
}
if (!IsStreamHeap(props.memoryHeaps[memory_type.heapIndex])) {
if (!IsStreamHeap(props.memoryHeaps[memory_type.heapIndex], staging_buffer_size)) {
// Memory heap is not suitable for streaming
continue;
}
@ -61,17 +80,17 @@ std::optional<u32> FindMemoryTypeIndex(const VkPhysicalDeviceMemoryProperties& p
}
u32 FindMemoryTypeIndex(const VkPhysicalDeviceMemoryProperties& props, u32 type_mask,
bool try_device_local) {
bool try_device_local, size_t staging_buffer_size) {
std::optional<u32> type;
if (try_device_local) {
// Try to find a DEVICE_LOCAL_BIT type, Nvidia and AMD have a dedicated heap for this
type = FindMemoryTypeIndex(props, type_mask, STREAM_FLAGS);
type = FindMemoryTypeIndex(props, type_mask, STREAM_FLAGS, staging_buffer_size);
if (type) {
return *type;
}
}
// Otherwise try without the DEVICE_LOCAL_BIT
type = FindMemoryTypeIndex(props, type_mask, HOST_FLAGS);
type = FindMemoryTypeIndex(props, type_mask, HOST_FLAGS, staging_buffer_size);
if (type) {
return *type;
}
@ -79,20 +98,32 @@ u32 FindMemoryTypeIndex(const VkPhysicalDeviceMemoryProperties& props, u32 type_
throw vk::Exception(VK_ERROR_OUT_OF_DEVICE_MEMORY);
}
size_t Region(size_t iterator) noexcept {
return iterator / REGION_SIZE;
size_t Region(size_t iterator, size_t region_size) noexcept {
return iterator / region_size;
}
} // Anonymous namespace
StagingBufferPool::StagingBufferPool(const Device& device_, MemoryAllocator& memory_allocator_,
Scheduler& scheduler_)
: device{device_}, memory_allocator{memory_allocator_}, scheduler{scheduler_} {
const auto memory_properties{device.GetPhysical().GetMemoryProperties().memoryProperties};
if (HasLargeDeviceLocalHostVisibleMemory(memory_properties)) {
// Possible on many integrated and newer discrete cards
staging_buffer_size = 1_GiB;
} else {
// Well-supported default size used by most Vulkan PC games
staging_buffer_size = 256_MiB;
}
region_size = staging_buffer_size / StagingBufferPool::NUM_SYNCS;
const vk::Device& dev = device.GetLogical();
stream_buffer = dev.CreateBuffer(VkBufferCreateInfo{
.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO,
.pNext = nullptr,
.flags = 0,
.size = STREAM_BUFFER_SIZE,
.size = staging_buffer_size,
.usage = VK_BUFFER_USAGE_TRANSFER_SRC_BIT | VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT |
VK_BUFFER_USAGE_INDEX_BUFFER_BIT | VK_BUFFER_USAGE_STORAGE_BUFFER_BIT,
.sharingMode = VK_SHARING_MODE_EXCLUSIVE,
@ -117,19 +148,18 @@ StagingBufferPool::StagingBufferPool(const Device& device_, MemoryAllocator& mem
.image = nullptr,
.buffer = *stream_buffer,
};
const auto memory_properties = device.GetPhysical().GetMemoryProperties().memoryProperties;
VkMemoryAllocateInfo stream_memory_info{
.sType = VK_STRUCTURE_TYPE_MEMORY_ALLOCATE_INFO,
.pNext = make_dedicated ? &dedicated_info : nullptr,
.allocationSize = requirements.size,
.memoryTypeIndex =
FindMemoryTypeIndex(memory_properties, requirements.memoryTypeBits, true),
.memoryTypeIndex = FindMemoryTypeIndex(memory_properties, requirements.memoryTypeBits, true,
staging_buffer_size),
};
stream_memory = dev.TryAllocateMemory(stream_memory_info);
if (!stream_memory) {
LOG_INFO(Render_Vulkan, "Dynamic memory allocation failed, trying with system memory");
stream_memory_info.memoryTypeIndex =
FindMemoryTypeIndex(memory_properties, requirements.memoryTypeBits, false);
stream_memory_info.memoryTypeIndex = FindMemoryTypeIndex(
memory_properties, requirements.memoryTypeBits, false, staging_buffer_size);
stream_memory = dev.AllocateMemory(stream_memory_info);
}
@ -137,7 +167,7 @@ StagingBufferPool::StagingBufferPool(const Device& device_, MemoryAllocator& mem
stream_memory.SetObjectNameEXT("Stream Buffer Memory");
}
stream_buffer.BindMemory(*stream_memory, 0);
stream_pointer = stream_memory.Map(0, STREAM_BUFFER_SIZE);
stream_pointer = stream_memory.Map(0, staging_buffer_size);
}
StagingBufferPool::~StagingBufferPool() = default;
@ -158,25 +188,25 @@ void StagingBufferPool::TickFrame() {
}
StagingBufferRef StagingBufferPool::GetStreamBuffer(size_t size) {
if (AreRegionsActive(Region(free_iterator) + 1,
std::min(Region(iterator + size) + 1, NUM_SYNCS))) {
if (AreRegionsActive(Region(free_iterator, region_size) + 1,
std::min(Region(iterator + size, region_size) + 1, NUM_SYNCS))) {
// Avoid waiting for the previous usages to be free
return GetStagingBuffer(size, MemoryUsage::Upload);
}
const u64 current_tick = scheduler.CurrentTick();
std::fill(sync_ticks.begin() + Region(used_iterator), sync_ticks.begin() + Region(iterator),
current_tick);
std::fill(sync_ticks.begin() + Region(used_iterator, region_size),
sync_ticks.begin() + Region(iterator, region_size), current_tick);
used_iterator = iterator;
free_iterator = std::max(free_iterator, iterator + size);
if (iterator + size >= STREAM_BUFFER_SIZE) {
std::fill(sync_ticks.begin() + Region(used_iterator), sync_ticks.begin() + NUM_SYNCS,
current_tick);
if (iterator + size >= staging_buffer_size) {
std::fill(sync_ticks.begin() + Region(used_iterator, region_size),
sync_ticks.begin() + NUM_SYNCS, current_tick);
used_iterator = 0;
iterator = 0;
free_iterator = size;
if (AreRegionsActive(0, Region(size) + 1)) {
if (AreRegionsActive(0, Region(size, region_size) + 1)) {
// Avoid waiting for the previous usages to be free
return GetStagingBuffer(size, MemoryUsage::Upload);
}

@ -93,6 +93,9 @@ private:
size_t free_iterator = 0;
std::array<u64, NUM_SYNCS> sync_ticks{};
size_t staging_buffer_size = 0;
size_t region_size = 0;
StagingBuffersCache device_local_cache;
StagingBuffersCache upload_cache;
StagingBuffersCache download_cache;