renderer_vulkan: Add single-thread record ability to the scheduler

* Async is pretty nice but games that do a lot of flushes might have worse performance due to thread synchronization overhead

* I haven't noticed any cases of this yet but it doesn't hurt making this a UI option

src/citra_qt/configuration/config.cpp

@@ -485,6 +485,7 @@ void Config::ReadRendererValues() {
         ReadSetting(QStringLiteral("graphics_api"), static_cast<u32>(Settings::GraphicsAPI::OpenGL))
             .toUInt());
     Settings::values.physical_device = ReadSetting(QStringLiteral("physical_device"), 0).toUInt();
+    Settings::values.async_command_recording = ReadSetting(QStringLiteral("async_command_recording"), true).toBool();
     Settings::values.use_hw_renderer =
         ReadSetting(QStringLiteral("use_hw_renderer"), true).toBool();
     Settings::values.use_hw_shader = ReadSetting(QStringLiteral("use_hw_shader"), true).toBool();
@@ -1004,6 +1005,7 @@ void Config::SaveRendererValues() {
     WriteSetting(QStringLiteral("graphics_api"), static_cast<u32>(Settings::values.graphics_api),
                  static_cast<u32>(Settings::GraphicsAPI::OpenGL));
     WriteSetting(QStringLiteral("physical_device"), Settings::values.physical_device, 0);
+    WriteSetting(QStringLiteral("async_command_recording"), Settings::values.async_command_recording, true);
     WriteSetting(QStringLiteral("use_hw_renderer"), Settings::values.use_hw_renderer, true);
     WriteSetting(QStringLiteral("use_hw_shader"), Settings::values.use_hw_shader, true);
 #ifdef __APPLE__

src/citra_qt/configuration/configure_graphics.cpp

@@ -83,6 +83,7 @@ void ConfigureGraphics::SetConfiguration() {
     ui->toggle_vsync_new->setChecked(Settings::values.use_vsync_new);
     ui->graphics_api_combo->setCurrentIndex(static_cast<int>(Settings::values.graphics_api));
     ui->physical_device_combo->setCurrentIndex(static_cast<int>(Settings::values.physical_device));
+    ui->toggle_async_recording->setChecked(Settings::values.async_command_recording);
 }
 
 void ConfigureGraphics::ApplyConfiguration() {
@@ -96,6 +97,7 @@ void ConfigureGraphics::ApplyConfiguration() {
     Settings::values.graphics_api =
         static_cast<Settings::GraphicsAPI>(ui->graphics_api_combo->currentIndex());
     Settings::values.physical_device = static_cast<u16>(ui->physical_device_combo->currentIndex());
+    Settings::values.async_command_recording = ui->toggle_async_recording->isChecked();
 }
 
 void ConfigureGraphics::RetranslateUI() {

src/citra_qt/configuration/configure_graphics.ui

@@ -171,6 +171,16 @@
       <string>Advanced</string>
      </property>
      <layout class="QVBoxLayout" name="verticalLayout_2">
+      <item>
+       <widget class="QCheckBox" name="toggle_async_recording">
+        <property name="toolTip">
+         <string>&lt;html&gt;&lt;head/&gt;&lt;body&gt;&lt;p&gt;Offloads command buffer recording and fragment shader generation to a worker thread. Can improve performance especially on weaker systems. Disable if you notice better performance. If unsure leave it enabled,&lt;/p&gt;&lt;/body&gt;&lt;/html&gt;</string>
+        </property>
+        <property name="text">
+         <string>Async Command Recording</string>
+        </property>
+       </widget>
+      </item>
       <item>
        <widget class="QCheckBox" name="toggle_disk_shader_cache">
         <property name="toolTip">

src/core/settings.cpp

@@ -90,6 +90,7 @@ void LogSettings() {
     LogSetting("Core_UseCpuJit", values.use_cpu_jit);
     LogSetting("Core_CPUClockPercentage", values.cpu_clock_percentage);
     LogSetting("Renderer_GraphicsAPI", GetAPIName(values.graphics_api));
+    LogSetting("Renderer_AsyncRecording", values.async_command_recording);
     LogSetting("Renderer_UseHwRenderer", values.use_hw_renderer);
     LogSetting("Renderer_UseHwShader", values.use_hw_shader);
     LogSetting("Renderer_SeparableShader", values.separable_shader);

src/core/settings.h

@@ -169,6 +169,7 @@ struct Values {
     u16 physical_device;
     bool renderer_debug;
     bool dump_command_buffers;
+    bool async_command_recording;
     bool use_hw_renderer;
     bool use_hw_shader;
     bool separable_shader;

src/video_core/renderer_vulkan/vk_scheduler.cpp

@@ -4,7 +4,7 @@
 #include <mutex>
 #include <utility>
 #include "common/microprofile.h"
-#include "common/thread.h"
+#include "core/settings.h"
 #include "video_core/renderer_vulkan/vk_scheduler.h"
 #include "video_core/renderer_vulkan/vk_instance.h"
 #include "video_core/renderer_vulkan/renderer_vulkan.h"
@@ -16,7 +16,7 @@ void Scheduler::CommandChunk::ExecuteAll(vk::CommandBuffer render_cmdbuf, vk::Co
     while (command != nullptr) {
         auto next = command->GetNext();
         command->Execute(render_cmdbuf, upload_cmdbuf);
-        command->~Command();
+        std::destroy_at(command);
         command = next;
     }
     submit = false;
@@ -26,10 +26,13 @@ void Scheduler::CommandChunk::ExecuteAll(vk::CommandBuffer render_cmdbuf, vk::Co
 }
 
 Scheduler::Scheduler(const Instance& instance, RendererVulkan& renderer)
-    : instance{instance}, renderer{renderer}, master_semaphore{instance}, command_pool{instance, master_semaphore} {
+    : instance{instance}, renderer{renderer}, master_semaphore{instance}, command_pool{instance, master_semaphore},
-    AcquireNewChunk();
+      use_worker_thread{Settings::values.async_command_recording} {
     AllocateWorkerCommandBuffers();
-    worker_thread = std::jthread([this](std::stop_token token) { WorkerThread(token); });
+    if (use_worker_thread) {
+        AcquireNewChunk();
+        worker_thread = std::jthread([this](std::stop_token token) { WorkerThread(token); });
+    }
 }
 
 Scheduler::~Scheduler() = default;
@@ -47,6 +50,10 @@ void Scheduler::Finish(vk::Semaphore signal, vk::Semaphore wait) {
 
 MICROPROFILE_DEFINE(Vulkan_WaitForWorker, "Vulkan", "Wait for worker", MP_RGB(255, 192, 192));
 void Scheduler::WaitWorker() {
+    if (!use_worker_thread) {
+        return;
+    }
+
     MICROPROFILE_SCOPE(Vulkan_WaitForWorker);
     DispatchWork();
 
@@ -162,8 +169,12 @@ void Scheduler::SubmitExecution(vk::Semaphore signal_semaphore, vk::Semaphore wa
         }
     });
 
-    chunk->MarkSubmit();
+    if (!use_worker_thread) {
-    DispatchWork();
+        AllocateWorkerCommandBuffers();
+    } else {
+        chunk->MarkSubmit();
+        DispatchWork();
+    }
 }
 
 void Scheduler::AcquireNewChunk() {

src/video_core/renderer_vulkan/vk_scheduler.h

@@ -52,6 +52,11 @@ public:
     /// Records the command to the current chunk.
     template <typename T>
     void Record(T&& command) {
+        if (!use_worker_thread) {
+            command(render_cmdbuf, upload_cmdbuf);
+            return;
+        }
+
         if (chunk->Record(command)) {
             return;
         }
@@ -144,7 +149,7 @@ private:
                 return false;
             }
             Command* const current_last = last;
-            last = new (data.data() + command_offset) FuncType(std::move(command));
+            last = std::construct_at(reinterpret_cast<FuncType*>(data.data() + command_offset), std::move(command));
 
             if (current_last) {
                 current_last->SetNext(last);
@@ -202,7 +207,7 @@ private:
     std::condition_variable_any work_cv;
     std::condition_variable wait_cv;
     std::jthread worker_thread;
-    std::jthread prsent_thread;
+    bool use_worker_thread;
 };
 
 } // namespace Vulkan