dxvk/src/dxvk/dxvk_device.cpp

319 lines
10 KiB
C++

#include "dxvk_device.h"
#include "dxvk_instance.h"
namespace dxvk {
DxvkDevice::DxvkDevice(
const Rc<DxvkInstance>& instance,
const Rc<DxvkAdapter>& adapter,
const Rc<vk::DeviceFn>& vkd,
const DxvkDeviceFeatures& features)
: m_options (instance->options()),
m_instance (instance),
m_adapter (adapter),
m_vkd (vkd),
m_features (features),
m_properties (adapter->devicePropertiesExt()),
m_perfHints (getPerfHints()),
m_objects (this),
m_submissionQueue (this) {
auto queueFamilies = m_adapter->findQueueFamilies();
m_queues.graphics = getQueue(queueFamilies.graphics, 0);
m_queues.transfer = getQueue(queueFamilies.transfer, 0);
m_queues.sparse = getQueue(queueFamilies.sparse, 0);
}
DxvkDevice::~DxvkDevice() {
// If we are being destroyed during/after DLL process detachment
// from TerminateProcess, etc, our CS threads are already destroyed
// and we cannot synchronize against them.
// The best we can do is just wait for the Vulkan device to be idle.
if (this_thread::isInModuleDetachment())
return;
// Wait for all pending Vulkan commands to be
// executed before we destroy any resources.
this->waitForIdle();
// Stop workers explicitly in order to prevent
// access to structures that are being destroyed.
m_objects.pipelineManager().stopWorkerThreads();
}
bool DxvkDevice::isUnifiedMemoryArchitecture() const {
return m_adapter->isUnifiedMemoryArchitecture();
}
bool DxvkDevice::canUseGraphicsPipelineLibrary() const {
// Without graphicsPipelineLibraryIndependentInterpolationDecoration, we
// cannot use this effectively in many games since no client API provides
// interpoation qualifiers in vertex shaders.
return m_features.extGraphicsPipelineLibrary.graphicsPipelineLibrary
&& m_properties.extGraphicsPipelineLibrary.graphicsPipelineLibraryIndependentInterpolationDecoration
&& m_options.enableGraphicsPipelineLibrary != Tristate::False;
}
bool DxvkDevice::canUsePipelineCacheControl() const {
// Don't bother with this unless the device also supports shader module
// identifiers, since decoding and hashing the shaders is slow otherwise
// and likely provides no benefit over linking pipeline libraries.
return m_features.vk13.pipelineCreationCacheControl
&& m_features.extShaderModuleIdentifier.shaderModuleIdentifier
&& m_options.enableGraphicsPipelineLibrary != Tristate::True;
}
bool DxvkDevice::mustTrackPipelineLifetime() const {
bool result = env::is32BitHostPlatform();
applyTristate(result, m_options.trackPipelineLifetime);
return result && canUseGraphicsPipelineLibrary();
}
DxvkFramebufferSize DxvkDevice::getDefaultFramebufferSize() const {
return DxvkFramebufferSize {
m_properties.core.properties.limits.maxFramebufferWidth,
m_properties.core.properties.limits.maxFramebufferHeight,
m_properties.core.properties.limits.maxFramebufferLayers };
}
VkPipelineStageFlags DxvkDevice::getShaderPipelineStages() const {
VkPipelineStageFlags result = VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT
| VK_PIPELINE_STAGE_VERTEX_SHADER_BIT
| VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT;
if (m_features.core.features.geometryShader)
result |= VK_PIPELINE_STAGE_GEOMETRY_SHADER_BIT;
if (m_features.core.features.tessellationShader) {
result |= VK_PIPELINE_STAGE_TESSELLATION_CONTROL_SHADER_BIT
| VK_PIPELINE_STAGE_TESSELLATION_EVALUATION_SHADER_BIT;
}
return result;
}
DxvkDeviceOptions DxvkDevice::options() const {
DxvkDeviceOptions options;
options.maxNumDynamicUniformBuffers = m_properties.core.properties.limits.maxDescriptorSetUniformBuffersDynamic;
options.maxNumDynamicStorageBuffers = m_properties.core.properties.limits.maxDescriptorSetStorageBuffersDynamic;
return options;
}
Rc<DxvkCommandList> DxvkDevice::createCommandList() {
Rc<DxvkCommandList> cmdList = m_recycledCommandLists.retrieveObject();
if (cmdList == nullptr)
cmdList = new DxvkCommandList(this);
return cmdList;
}
Rc<DxvkContext> DxvkDevice::createContext(DxvkContextType type) {
return new DxvkContext(this, type);
}
Rc<DxvkGpuEvent> DxvkDevice::createGpuEvent() {
return new DxvkGpuEvent(m_vkd);
}
Rc<DxvkGpuQuery> DxvkDevice::createGpuQuery(
VkQueryType type,
VkQueryControlFlags flags,
uint32_t index) {
return new DxvkGpuQuery(m_vkd, type, flags, index);
}
Rc<DxvkFence> DxvkDevice::createFence(
const DxvkFenceCreateInfo& fenceInfo) {
return new DxvkFence(this, fenceInfo);
}
Rc<DxvkBuffer> DxvkDevice::createBuffer(
const DxvkBufferCreateInfo& createInfo,
VkMemoryPropertyFlags memoryType) {
return new DxvkBuffer(this, createInfo, m_objects.memoryManager(), memoryType);
}
Rc<DxvkBufferView> DxvkDevice::createBufferView(
const Rc<DxvkBuffer>& buffer,
const DxvkBufferViewCreateInfo& createInfo) {
return new DxvkBufferView(m_vkd, buffer, createInfo);
}
Rc<DxvkImage> DxvkDevice::createImage(
const DxvkImageCreateInfo& createInfo,
VkMemoryPropertyFlags memoryType) {
return new DxvkImage(this, createInfo, m_objects.memoryManager(), memoryType);
}
Rc<DxvkImage> DxvkDevice::createImageFromVkImage(
const DxvkImageCreateInfo& createInfo,
VkImage image) {
return new DxvkImage(this, createInfo, image);
}
Rc<DxvkImageView> DxvkDevice::createImageView(
const Rc<DxvkImage>& image,
const DxvkImageViewCreateInfo& createInfo) {
return new DxvkImageView(m_vkd, image, createInfo);
}
Rc<DxvkSampler> DxvkDevice::createSampler(
const DxvkSamplerCreateInfo& createInfo) {
return new DxvkSampler(this, createInfo);
}
Rc<DxvkSparsePageAllocator> DxvkDevice::createSparsePageAllocator() {
return new DxvkSparsePageAllocator(m_objects.memoryManager());
}
DxvkStatCounters DxvkDevice::getStatCounters() {
DxvkPipelineCount pipe = m_objects.pipelineManager().getPipelineCount();
DxvkStatCounters result;
result.setCtr(DxvkStatCounter::PipeCountGraphics, pipe.numGraphicsPipelines);
result.setCtr(DxvkStatCounter::PipeCountLibrary, pipe.numGraphicsLibraries);
result.setCtr(DxvkStatCounter::PipeCountCompute, pipe.numComputePipelines);
result.setCtr(DxvkStatCounter::PipeCompilerBusy, m_objects.pipelineManager().isCompilingShaders());
result.setCtr(DxvkStatCounter::GpuIdleTicks, m_submissionQueue.gpuIdleTicks());
std::lock_guard<sync::Spinlock> lock(m_statLock);
result.merge(m_statCounters);
return result;
}
DxvkMemoryStats DxvkDevice::getMemoryStats(uint32_t heap) {
return m_objects.memoryManager().getMemoryStats(heap);
}
uint32_t DxvkDevice::getCurrentFrameId() const {
return m_statCounters.getCtr(DxvkStatCounter::QueuePresentCount);
}
void DxvkDevice::registerShader(const Rc<DxvkShader>& shader) {
m_objects.pipelineManager().registerShader(shader);
}
void DxvkDevice::requestCompileShader(
const Rc<DxvkShader>& shader) {
m_objects.pipelineManager().requestCompileShader(shader);
}
void DxvkDevice::presentImage(
const Rc<vk::Presenter>& presenter,
DxvkSubmitStatus* status) {
status->result = VK_NOT_READY;
DxvkPresentInfo presentInfo;
presentInfo.presenter = presenter;
m_submissionQueue.present(presentInfo, status);
std::lock_guard<sync::Spinlock> statLock(m_statLock);
m_statCounters.addCtr(DxvkStatCounter::QueuePresentCount, 1);
}
void DxvkDevice::submitCommandList(
const Rc<DxvkCommandList>& commandList) {
DxvkSubmitInfo submitInfo = { };
submitInfo.cmdList = commandList;
m_submissionQueue.submit(submitInfo);
std::lock_guard<sync::Spinlock> statLock(m_statLock);
m_statCounters.merge(commandList->statCounters());
}
VkResult DxvkDevice::waitForSubmission(DxvkSubmitStatus* status) {
VkResult result = status->result.load();
if (result == VK_NOT_READY) {
m_submissionQueue.synchronizeSubmission(status);
result = status->result.load();
}
return result;
}
void DxvkDevice::waitForResource(const Rc<DxvkResource>& resource, DxvkAccess access) {
if (resource->isInUse(access)) {
auto t0 = dxvk::high_resolution_clock::now();
m_submissionQueue.synchronizeUntil([resource, access] {
return !resource->isInUse(access);
});
auto t1 = dxvk::high_resolution_clock::now();
auto us = std::chrono::duration_cast<std::chrono::microseconds>(t1 - t0);
std::lock_guard<sync::Spinlock> lock(m_statLock);
m_statCounters.addCtr(DxvkStatCounter::GpuSyncCount, 1);
m_statCounters.addCtr(DxvkStatCounter::GpuSyncTicks, us.count());
}
}
void DxvkDevice::waitForIdle() {
this->lockSubmission();
if (m_vkd->vkDeviceWaitIdle(m_vkd->device()) != VK_SUCCESS)
Logger::err("DxvkDevice: waitForIdle: Operation failed");
this->unlockSubmission();
}
DxvkDevicePerfHints DxvkDevice::getPerfHints() {
DxvkDevicePerfHints hints;
hints.preferFbDepthStencilCopy = m_features.extShaderStencilExport
&& (m_adapter->matchesDriver(VK_DRIVER_ID_MESA_RADV_KHR, 0, 0)
|| m_adapter->matchesDriver(VK_DRIVER_ID_AMD_OPEN_SOURCE_KHR, 0, 0)
|| m_adapter->matchesDriver(VK_DRIVER_ID_AMD_PROPRIETARY_KHR, 0, 0));
hints.preferFbResolve = m_features.amdShaderFragmentMask
&& (m_adapter->matchesDriver(VK_DRIVER_ID_AMD_OPEN_SOURCE_KHR, 0, 0)
|| m_adapter->matchesDriver(VK_DRIVER_ID_AMD_PROPRIETARY_KHR, 0, 0));
return hints;
}
void DxvkDevice::recycleCommandList(const Rc<DxvkCommandList>& cmdList) {
m_recycledCommandLists.returnObject(cmdList);
}
DxvkDeviceQueue DxvkDevice::getQueue(
uint32_t family,
uint32_t index) const {
VkQueue queue = VK_NULL_HANDLE;
if (family != VK_QUEUE_FAMILY_IGNORED)
m_vkd->vkGetDeviceQueue(m_vkd->device(), family, index, &queue);
return DxvkDeviceQueue { queue, family, index };
}
}