yuzu-emu
/
yuzu-mainline
Archived
1
0
Fork 0

vk_stream_buffer: Fix out of memory on boot on recent Nvidia drivers

Nvidia recently introduced a new memory type for data streaming
(awesome!), but yuzu was assuming that all heaps had enough memory
for the assumed stream buffer size (256 MiB).

This worked fine on AMD but Nvidia's new memory heap was smaller than
256 MiB. This commit changes this assumption and allocates a bit less
than the size of the preferred heap, with a maximum of 256 MiB (to avoid
allocating all system memory on integrated devices).

- Fixes a crash on NVIDIA 450.82.0.0
This commit is contained in:
ReinUsesLisp 2020-04-17 05:13:35 -03:00
parent 2133482a17
commit a7b6bd56d7
2 changed files with 47 additions and 32 deletions

View File

@ -3,6 +3,7 @@
// Refer to the license.txt file included. // Refer to the license.txt file included.
#include <algorithm> #include <algorithm>
#include <limits>
#include <optional> #include <optional>
#include <tuple> #include <tuple>
#include <vector> #include <vector>
@ -22,22 +23,38 @@ namespace {
constexpr u64 WATCHES_INITIAL_RESERVE = 0x4000; constexpr u64 WATCHES_INITIAL_RESERVE = 0x4000;
constexpr u64 WATCHES_RESERVE_CHUNK = 0x1000; constexpr u64 WATCHES_RESERVE_CHUNK = 0x1000;
constexpr u64 STREAM_BUFFER_SIZE = 256 * 1024 * 1024; constexpr u64 PREFERRED_STREAM_BUFFER_SIZE = 256 * 1024 * 1024;
std::optional<u32> FindMemoryType(const VKDevice& device, u32 filter, /// Find a memory type with the passed requirements
VkMemoryPropertyFlags wanted) { std::optional<u32> FindMemoryType(const VkPhysicalDeviceMemoryProperties& properties,
const auto properties = device.GetPhysical().GetMemoryProperties(); VkMemoryPropertyFlags wanted,
for (u32 i = 0; i < properties.memoryTypeCount; i++) { u32 filter = std::numeric_limits<u32>::max()) {
if (!(filter & (1 << i))) { for (u32 i = 0; i < properties.memoryTypeCount; ++i) {
continue; const auto flags = properties.memoryTypes[i].propertyFlags;
} if ((flags & wanted) == wanted && (filter & (1U << i)) != 0) {
if ((properties.memoryTypes[i].propertyFlags & wanted) == wanted) {
return i; return i;
} }
} }
return std::nullopt; return std::nullopt;
} }
/// Get the preferred host visible memory type.
u32 GetMemoryType(const VkPhysicalDeviceMemoryProperties& properties,
u32 filter = std::numeric_limits<u32>::max()) {
// Prefer device local host visible allocations. Both AMD and Nvidia now provide one.
// Otherwise search for a host visible allocation.
static constexpr auto HOST_MEMORY =
VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT | VK_MEMORY_PROPERTY_HOST_COHERENT_BIT;
static constexpr auto DYNAMIC_MEMORY = HOST_MEMORY | VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT;
std::optional preferred_type = FindMemoryType(properties, DYNAMIC_MEMORY);
if (!preferred_type) {
preferred_type = FindMemoryType(properties, HOST_MEMORY);
ASSERT_MSG(preferred_type, "No host visible and coherent memory type found");
}
return preferred_type.value_or(0);
}
} // Anonymous namespace } // Anonymous namespace
VKStreamBuffer::VKStreamBuffer(const VKDevice& device, VKScheduler& scheduler, VKStreamBuffer::VKStreamBuffer(const VKDevice& device, VKScheduler& scheduler,
@ -51,7 +68,7 @@ VKStreamBuffer::VKStreamBuffer(const VKDevice& device, VKScheduler& scheduler,
VKStreamBuffer::~VKStreamBuffer() = default; VKStreamBuffer::~VKStreamBuffer() = default;
std::tuple<u8*, u64, bool> VKStreamBuffer::Map(u64 size, u64 alignment) { std::tuple<u8*, u64, bool> VKStreamBuffer::Map(u64 size, u64 alignment) {
ASSERT(size <= STREAM_BUFFER_SIZE); ASSERT(size <= stream_buffer_size);
mapped_size = size; mapped_size = size;
if (alignment > 0) { if (alignment > 0) {
@ -61,7 +78,7 @@ std::tuple<u8*, u64, bool> VKStreamBuffer::Map(u64 size, u64 alignment) {
WaitPendingOperations(offset); WaitPendingOperations(offset);
bool invalidated = false; bool invalidated = false;
if (offset + size > STREAM_BUFFER_SIZE) { if (offset + size > stream_buffer_size) {
// The buffer would overflow, save the amount of used watches and reset the state. // The buffer would overflow, save the amount of used watches and reset the state.
invalidation_mark = current_watch_cursor; invalidation_mark = current_watch_cursor;
current_watch_cursor = 0; current_watch_cursor = 0;
@ -98,40 +115,37 @@ void VKStreamBuffer::Unmap(u64 size) {
} }
void VKStreamBuffer::CreateBuffers(VkBufferUsageFlags usage) { void VKStreamBuffer::CreateBuffers(VkBufferUsageFlags usage) {
const auto memory_properties = device.GetPhysical().GetMemoryProperties();
const u32 preferred_type = GetMemoryType(memory_properties);
const u32 preferred_heap = memory_properties.memoryTypes[preferred_type].heapIndex;
// Substract from the preferred heap size some bytes to avoid getting out of memory.
const VkDeviceSize heap_size = memory_properties.memoryHeaps[preferred_heap].size;
const VkDeviceSize allocable_size = heap_size - 4 * 1024 * 1024;
VkBufferCreateInfo buffer_ci; VkBufferCreateInfo buffer_ci;
buffer_ci.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO; buffer_ci.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO;
buffer_ci.pNext = nullptr; buffer_ci.pNext = nullptr;
buffer_ci.flags = 0; buffer_ci.flags = 0;
buffer_ci.size = STREAM_BUFFER_SIZE; buffer_ci.size = std::min(PREFERRED_STREAM_BUFFER_SIZE, allocable_size);
buffer_ci.usage = usage; buffer_ci.usage = usage;
buffer_ci.sharingMode = VK_SHARING_MODE_EXCLUSIVE; buffer_ci.sharingMode = VK_SHARING_MODE_EXCLUSIVE;
buffer_ci.queueFamilyIndexCount = 0; buffer_ci.queueFamilyIndexCount = 0;
buffer_ci.pQueueFamilyIndices = nullptr; buffer_ci.pQueueFamilyIndices = nullptr;
const auto& dev = device.GetLogical(); buffer = device.GetLogical().CreateBuffer(buffer_ci);
buffer = dev.CreateBuffer(buffer_ci);
const auto requirements = device.GetLogical().GetBufferMemoryRequirements(*buffer);
const u32 required_flags = requirements.memoryTypeBits;
stream_buffer_size = static_cast<u64>(requirements.size);
const auto& dld = device.GetDispatchLoader();
const auto requirements = dev.GetBufferMemoryRequirements(*buffer);
// Prefer device local host visible allocations (this should hit AMD's pinned memory).
auto type =
FindMemoryType(device, requirements.memoryTypeBits,
VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT | VK_MEMORY_PROPERTY_HOST_COHERENT_BIT |
VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT);
if (!type) {
// Otherwise search for a host visible allocation.
type = FindMemoryType(device, requirements.memoryTypeBits,
VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT |
VK_MEMORY_PROPERTY_HOST_COHERENT_BIT);
ASSERT_MSG(type, "No host visible and coherent memory type found");
}
VkMemoryAllocateInfo memory_ai; VkMemoryAllocateInfo memory_ai;
memory_ai.sType = VK_STRUCTURE_TYPE_MEMORY_ALLOCATE_INFO; memory_ai.sType = VK_STRUCTURE_TYPE_MEMORY_ALLOCATE_INFO;
memory_ai.pNext = nullptr; memory_ai.pNext = nullptr;
memory_ai.allocationSize = requirements.size; memory_ai.allocationSize = requirements.size;
memory_ai.memoryTypeIndex = *type; memory_ai.memoryTypeIndex = GetMemoryType(memory_properties, required_flags);
memory = dev.AllocateMemory(memory_ai); memory = device.GetLogical().AllocateMemory(memory_ai);
buffer.BindMemory(*memory, 0); buffer.BindMemory(*memory, 0);
} }

View File

@ -56,8 +56,9 @@ private:
const VKDevice& device; ///< Vulkan device manager. const VKDevice& device; ///< Vulkan device manager.
VKScheduler& scheduler; ///< Command scheduler. VKScheduler& scheduler; ///< Command scheduler.
vk::Buffer buffer; ///< Mapped buffer. vk::Buffer buffer; ///< Mapped buffer.
vk::DeviceMemory memory; ///< Memory allocation. vk::DeviceMemory memory; ///< Memory allocation.
u64 stream_buffer_size{}; ///< Stream buffer size.
u64 offset{}; ///< Buffer iterator. u64 offset{}; ///< Buffer iterator.
u64 mapped_size{}; ///< Size reserved for the current copy. u64 mapped_size{}; ///< Size reserved for the current copy.