hdkv/yuzu
1
Fork 0
Code Issues Pull Requests Releases 1 Activity

early-access version 3301

Browse Source
This commit is contained in:
pineappleEA
2023-01-08 21:39:56 +01:00
parent 987c0be419
commit 94223944e1
19 changed files with 57 additions and 36 deletions
Download Patch File Download Diff File Expand all files Collapse all files

13
src/video_core/buffer_cache/buffer_cache.h
View File

@@ -1938,21 +1938,14 @@ typename BufferCache<P>::Binding BufferCache<P>::StorageBufferBinding(GPUVAddr s
bool is_written) const {
const GPUVAddr gpu_addr = gpu_memory->Read<u64>(ssbo_addr);
const u32 size = gpu_memory->Read<u32>(ssbo_addr + 8);
const u32 alignment = runtime.GetStorageBufferAlignment();
const GPUVAddr aligned_gpu_addr = Common::AlignDown(gpu_addr, alignment);
const u32 aligned_size =
Common::AlignUp(static_cast<u32>(gpu_addr - aligned_gpu_addr) + size, alignment);
const std::optional<VAddr> cpu_addr = gpu_memory->GpuToCpuAddress(aligned_gpu_addr);
const std::optional<VAddr> cpu_addr = gpu_memory->GpuToCpuAddress(gpu_addr);
if (!cpu_addr || size == 0) {
return NULL_BINDING;
}
const VAddr cpu_end = Common::AlignUp(*cpu_addr + aligned_size, Core::Memory::YUZU_PAGESIZE);
const VAddr cpu_end = Common::AlignUp(*cpu_addr + size, Core::Memory::YUZU_PAGESIZE);
const Binding binding{
.cpu_addr = *cpu_addr,
.size = is_written ? aligned_size : static_cast<u32>(cpu_end - *cpu_addr),
.size = is_written ? size : static_cast<u32>(cpu_end - *cpu_addr),
.buffer_id = BufferId{},
};
return binding;

4
src/video_core/renderer_opengl/gl_buffer_cache.h
View File

@@ -160,10 +160,6 @@ public:
return device.CanReportMemoryUsage();
}
u32 GetStorageBufferAlignment() const {
return static_cast<u32>(device.GetShaderStorageBufferAlignment());
}
private:
static constexpr std::array PABO_LUT{
GL_VERTEX_PROGRAM_PARAMETER_BUFFER_NV, GL_TESS_CONTROL_PROGRAM_PARAMETER_BUFFER_NV,

4
src/video_core/renderer_opengl/gl_rasterizer.cpp
View File

@@ -140,6 +140,7 @@ void RasterizerOpenGL::LoadDiskResources(u64 title_id, std::stop_token stop_load
void RasterizerOpenGL::Clear(u32 layer_count) {
MICROPROFILE_SCOPE(OpenGL_Clears);
gpu_memory->FlushCaching();
const auto& regs = maxwell3d->regs;
bool use_color{};
bool use_depth{};
@@ -208,6 +209,7 @@ void RasterizerOpenGL::PrepareDraw(bool is_indexed, Func&& draw_func) {
MICROPROFILE_SCOPE(OpenGL_Drawing);
SCOPE_EXIT({ gpu.TickWork(); });
gpu_memory->FlushCaching();
query_cache.UpdateCounters();
GraphicsPipeline* const pipeline{shader_cache.CurrentGraphicsPipeline()};
@@ -361,6 +363,7 @@ void RasterizerOpenGL::DrawTexture() {
}
void RasterizerOpenGL::DispatchCompute() {
gpu_memory->FlushCaching();
ComputePipeline* const pipeline{shader_cache.CurrentComputePipeline()};
if (!pipeline) {
return;
@@ -568,6 +571,7 @@ void RasterizerOpenGL::TickFrame() {
}
bool RasterizerOpenGL::AccelerateConditionalRendering() {
gpu_memory->FlushCaching();
if (Settings::IsGPULevelHigh()) {
// Reimplement Host conditional rendering.
return false;

1
src/video_core/renderer_opengl/gl_shader_cache.cpp
View File

@@ -236,7 +236,6 @@ ShaderCache::ShaderCache(RasterizerOpenGL& rasterizer_, Core::Frontend::EmuWindo
.needs_demote_reorder = device.IsAmd(),
.support_snorm_render_buffer = false,
.support_viewport_index_layer = device.HasVertexViewportLayer(),
.min_ssbo_alignment = static_cast<u32>(device.GetShaderStorageBufferAlignment()),
.support_geometry_shader_passthrough = device.HasGeometryShaderPassthrough(),
} {
if (use_asynchronous_shaders) {

2
src/video_core/renderer_vulkan/renderer_vulkan.cpp
View File

@@ -99,6 +99,7 @@ RendererVulkan::RendererVulkan(Core::TelemetrySession& telemetry_session_,
state_tracker, scheduler) {
if (Settings::values.renderer_force_max_clock.GetValue() && device.ShouldBoostClocks()) {
turbo_mode.emplace(instance, dld);
scheduler.RegisterOnSubmit([this] { turbo_mode->QueueSubmitted(); });
}
Report();
} catch (const vk::Exception& exception) {
@@ -107,6 +108,7 @@ RendererVulkan::RendererVulkan(Core::TelemetrySession& telemetry_session_,
}
RendererVulkan::~RendererVulkan() {
scheduler.RegisterOnSubmit([] {});
void(device.GetLogical().WaitIdle());
}

4
src/video_core/renderer_vulkan/vk_buffer_cache.cpp
View File

@@ -330,10 +330,6 @@ bool BufferCacheRuntime::CanReportMemoryUsage() const {
return device.CanReportMemoryUsage();
}
u32 BufferCacheRuntime::GetStorageBufferAlignment() const {
return static_cast<u32>(device.GetStorageBufferAlignment());
}
void BufferCacheRuntime::Finish() {
scheduler.Finish();
}

2
src/video_core/renderer_vulkan/vk_buffer_cache.h
View File

@@ -73,8 +73,6 @@ public:
bool CanReportMemoryUsage() const;
u32 GetStorageBufferAlignment() const;
[[nodiscard]] StagingBufferRef UploadStagingBuffer(size_t size);
[[nodiscard]] StagingBufferRef DownloadStagingBuffer(size_t size);

1
src/video_core/renderer_vulkan/vk_pipeline_cache.cpp
View File

@@ -344,7 +344,6 @@ PipelineCache::PipelineCache(RasterizerVulkan& rasterizer_, const Device& device
driver_id == VK_DRIVER_ID_AMD_PROPRIETARY || driver_id == VK_DRIVER_ID_AMD_OPEN_SOURCE,
.support_snorm_render_buffer = true,
.support_viewport_index_layer = device.IsExtShaderViewportIndexLayerSupported(),
.min_ssbo_alignment = static_cast<u32>(device.GetStorageBufferAlignment()),
.support_geometry_shader_passthrough = device.IsNvGeometryShaderPassthroughSupported(),
};

2
src/video_core/renderer_vulkan/vk_rasterizer.cpp
View File

@@ -298,6 +298,7 @@ void RasterizerVulkan::Clear(u32 layer_count) {
MICROPROFILE_SCOPE(Vulkan_Clearing);
FlushWork();
gpu_memory->FlushCaching();
query_cache.UpdateCounters();
@@ -656,6 +657,7 @@ void RasterizerVulkan::TickFrame() {
}
bool RasterizerVulkan::AccelerateConditionalRendering() {
gpu_memory->FlushCaching();
if (Settings::IsGPULevelHigh()) {
// TODO(Blinkhawk): Reimplement Host conditional rendering.
return false;

5
src/video_core/renderer_vulkan/vk_scheduler.cpp
View File

@@ -213,6 +213,11 @@ void Scheduler::SubmitExecution(VkSemaphore signal_semaphore, VkSemaphore wait_s
.signalSemaphoreCount = num_signal_semaphores,
.pSignalSemaphores = signal_semaphores.data(),
};
if (on_submit) {
on_submit();
}
switch (const VkResult result = device.GetGraphicsQueue().Submit(submit_info)) {
case VK_SUCCESS:
break;

7
src/video_core/renderer_vulkan/vk_scheduler.h
View File

@@ -5,6 +5,7 @@
#include <condition_variable>
#include <cstddef>
#include <functional>
#include <memory>
#include <thread>
#include <utility>
@@ -66,6 +67,11 @@ public:
query_cache = &query_cache_;
}
// Registers a callback to perform on queue submission.
void RegisterOnSubmit(std::function<void()>&& func) {
on_submit = std::move(func);
}
/// Send work to a separate thread.
template <typename T>
void Record(T&& command) {
@@ -216,6 +222,7 @@ private:
vk::CommandBuffer current_cmdbuf;
std::unique_ptr<CommandChunk> chunk;
std::function<void()> on_submit;
State state;

17
src/video_core/renderer_vulkan/vk_turbo_mode.cpp
View File

@@ -14,11 +14,21 @@ using namespace Common::Literals;
TurboMode::TurboMode(const vk::Instance& instance, const vk::InstanceDispatch& dld)
: m_device{CreateDevice(instance, dld, VK_NULL_HANDLE)}, m_allocator{m_device, false} {
{
std::scoped_lock lk{m_submission_lock};
m_submission_time = std::chrono::steady_clock::now();
}
m_thread = std::jthread([&](auto stop_token) { Run(stop_token); });
}
TurboMode::~TurboMode() = default;
void TurboMode::QueueSubmitted() {
std::scoped_lock lk{m_submission_lock};
m_submission_time = std::chrono::steady_clock::now();
m_submission_cv.notify_one();
}
void TurboMode::Run(std::stop_token stop_token) {
auto& dld = m_device.GetLogical();
@@ -199,6 +209,13 @@ void TurboMode::Run(std::stop_token stop_token) {
// Wait for completion.
fence.Wait();
// Wait for the next graphics queue submission if necessary.
std::unique_lock lk{m_submission_lock};
Common::CondvarWait(m_submission_cv, lk, stop_token, [this] {
return (std::chrono::steady_clock::now() - m_submission_time) <=
std::chrono::milliseconds{100};
});
}
}

9
src/video_core/renderer_vulkan/vk_turbo_mode.h
View File

@@ -3,6 +3,9 @@
#pragma once
#include <chrono>
#include <mutex>
#include "common/polyfill_thread.h"
#include "video_core/vulkan_common/vulkan_device.h"
#include "video_core/vulkan_common/vulkan_memory_allocator.h"
@@ -15,11 +18,17 @@ public:
explicit TurboMode(const vk::Instance& instance, const vk::InstanceDispatch& dld);
~TurboMode();
void QueueSubmitted();
private:
void Run(std::stop_token stop_token);
Device m_device;
MemoryAllocator m_allocator;
std::mutex m_submission_lock;
std::condition_variable_any m_submission_cv;
std::chrono::time_point<std::chrono::steady_clock> m_submission_time{};
std::jthread m_thread;
};