2026-09-25 09:06:19

Signed-off-by: lizzie <lizzie@eden-emu.dev>

cpmfile.json

@@ -333,11 +333,6 @@
         "repo": "herumi/xbyak",
         "version": "v7.40.1"
     },
-    "zbic": {
-        "hash": "fbe2f37986377d7f0d96ae3224c80b5971df6e8b6961f68061f1975ebb2e8cb78f07b90f014b007be4023dbf5513581504e7f7d61a5237cb2a8a7a61ae11e482",
-        "repo": "kinnay/zbic",
-        "version": "11b08f2712264bbed731545085cbd9702096ceb7"
-    },
     "zlib": {
         "hash": "16fea4df307a68cf0035858abe2fd550250618a97590e202037acd18a666f57afc10f8836cbbd472d54a0e76539d0e558cb26f059d53de52ff90634bbf4f47d4",
         "min_version": "1.2",

externals/CMakeLists.txt

@@ -48,11 +48,6 @@ if (NOT TARGET stb::headers)
     add_library(stb::headers ALIAS stb)
 endif()
 
-AddJsonPackage(NAME zbic DOWNLOAD_ONLY)
-set(ZBIC_INCLUDE_DIR
-    "${zbic_SOURCE_DIR}/src"
-    PARENT_SCOPE)
-
 # ItaniumDemangle (Windows only)
 if (WIN32 AND NOT TARGET LLVM::Demangle)
     add_library(demangle demangle/ItaniumDemangle.cpp)

src/android/app/src/main/java/org/yuzu/yuzu_emu/features/settings/model/BooleanSetting.kt

@@ -30,6 +30,7 @@ enum class BooleanSetting(override val key: String) : AbstractBooleanSetting {
     RENDERER_REACTIVE_FLUSHING("use_reactive_flushing"),
     ENABLE_BUFFER_HISTORY("enable_buffer_history"),
     USE_OPTIMIZED_VERTEX_BUFFERS("use_optimized_vertex_buffers"),
+    ENABLE_GPU_BUFFER_READBACK("enable_gpu_buffer_readback"),
     SYNC_MEMORY_OPERATIONS("sync_memory_operations"),
     BUFFER_REORDER_DISABLE("disable_buffer_reorder"),
     RENDERER_DEBUG("debug"),

src/android/app/src/main/java/org/yuzu/yuzu_emu/features/settings/model/view/SettingsItem.kt

@@ -908,6 +908,13 @@ abstract class SettingsItem(
                     descriptionId = R.string.enable_buffer_history_description
                 )
             )
+            put(
+                SwitchSetting(
+                    BooleanSetting.ENABLE_GPU_BUFFER_READBACK,
+                    titleId = R.string.enable_gpu_buffer_readback,
+                    descriptionId = R.string.enable_gpu_buffer_readback_description
+                )
+            )
             put(
                 SwitchSetting(
                     BooleanSetting.USE_OPTIMIZED_VERTEX_BUFFERS,

src/android/app/src/main/java/org/yuzu/yuzu_emu/features/settings/ui/SettingsFragmentPresenter.kt

@@ -549,6 +549,7 @@ class SettingsFragmentPresenter(
             add(BooleanSetting.RENDERER_FORCE_MAX_CLOCK.key)
             add(BooleanSetting.RENDERER_REACTIVE_FLUSHING.key)
             add(BooleanSetting.ENABLE_BUFFER_HISTORY.key)
+            add(BooleanSetting.ENABLE_GPU_BUFFER_READBACK.key)
             add(BooleanSetting.USE_OPTIMIZED_VERTEX_BUFFERS.key)
 
             add(HeaderSetting(R.string.hacks))

src/android/app/src/main/res/values-ar/strings.xml

@@ -580,6 +580,8 @@
     <string name="renderer_reactive_flushing_description">يحسن دقة العرض في بعض الألعاب على حساب الأداء.</string>
     <string name="enable_buffer_history">تمكين سجل التخزين المؤقت</string>
     <string name="enable_buffer_history_description">يُتيح هذا الخيار الوصول إلى حالات التخزين المؤقت السابقة. وقد يُحسّن جودة العرض وثبات الأداء في بعض الألعاب.</string>
+    <string name="enable_gpu_buffer_readback">تفعيل قراءة مخزن وحدة معالجة الرسومات</string>
+    <string name="enable_gpu_buffer_readback_description">يحافظ هذا النظام على بيانات المخزن المؤقت المُعدّلة بواسطة وحدة معالجة الرسومات عن طريق قراءتها مرة أخرى قبل التحميل. تتطلب بعض الألعاب ذلك لعرض بعض التأثيرات بشكل صحيح. قد يُسبب ذلك مشاكل إذا لم يتمكن الجهاز من التعامل مع عبء العمل الإضافي.</string>
     <string name="use_optimized_vertex_buffers">مخازن الرؤوس المُحسّنة</string>
     <string name="use_optimized_vertex_buffers_description">يُتيح ربطًا مُحسَّنًا لمخازن الرؤوس لتحسين الأداء. يتطلب برامج تشغيل Mesa 26.0+ Turnip/ برامج تشغيل QCOM. قد يتعطل على برامج تشغيل Turnip القديمة (25.3 وما دون).</string>
 
@@ -1097,6 +1099,7 @@
     <string name="gpu_fence_behavior_immediate">فوري</string>
     <string name="gpu_fence_behavior_balanced">متوازن</string>
     <string name="gpu_fence_behavior_accurate">دقيق</string>
+    <string name="gpu_fence_behavior_strict">صارم</string>
 
     <string name="vram_usage_conservative">محافظ</string>
     <string name="vram_usage_aggressive">عدواني</string>

src/android/app/src/main/res/values-de/strings.xml

@@ -967,6 +967,7 @@ Wirklich fortfahren?</string>
     <string name="gpu_fence_behavior_immediate">Direkt</string>
     <string name="gpu_fence_behavior_balanced">Ausgewogen</string>
     <string name="gpu_fence_behavior_accurate">Genau</string>
+    <string name="gpu_fence_behavior_strict">Strikt</string>
 
     <string name="vram_usage_conservative">Konservativ</string>
     <string name="vram_usage_aggressive">Aggressiv</string>

src/android/app/src/main/res/values-es/strings.xml

@@ -524,6 +524,8 @@
     <string name="renderer_reactive_flushing_description">Mejora la precisión de renderizado en algunos juegos, pero reduce el rendimiento.</string>
     <string name="enable_buffer_history">Activar el historial del búfer</string>
     <string name="enable_buffer_history_description">Permite el acceso al estado del búfer anterior. Esta opción puede mejorar la calidad de renderizado y la consistencia en el rendimiento de algunos juegos.</string>
+    <string name="enable_gpu_buffer_readback">Activar la lectura del buffer de la GPU</string>
+    <string name="enable_gpu_buffer_readback_description">Conserva los datos del búfer modificados por la GPU leyéndolos antes de subirlos.\nAlgunos juegos requieren esto para renderizar correctamente ciertos efectos.\nPuede causar problemas si el hardware no puede soportar la carga de trabajo adicional.</string>
     <string name="use_optimized_vertex_buffers">Búferes de vértices optimizados</string>
     <string name="use_optimized_vertex_buffers_description">Permite la optimización del enlace del búfer de vértices para un mejor rendimiento. Requiere controladores Mesa 26.0+ Turnip/ controladores QCOM. Fallará con controladores Turnip más antiguos (versión 25.3 o inferior).</string>
 
@@ -1036,6 +1038,7 @@
     <string name="gpu_fence_behavior_immediate">Inmediato</string>
     <string name="gpu_fence_behavior_balanced">Equilibrado</string>
     <string name="gpu_fence_behavior_accurate">Preciso</string>
+    <string name="gpu_fence_behavior_strict">Estricto</string>
 
     <string name="vram_usage_conservative">Conservador</string>
     <string name="vram_usage_aggressive">Agresivo</string>

src/android/app/src/main/res/values-ru/strings.xml

@@ -569,6 +569,8 @@
     <string name="renderer_reactive_flushing_description">Повышение точности рендеринга в некоторых играх за счет снижения производительности.</string>
     <string name="enable_buffer_history">Включить историю буфера</string>
     <string name="enable_buffer_history_description">Позволяет обращаться к предыдущим состояниям буфера. Эта опция может повысить качество рендеринга и стабильность производительности в некоторых играх.</string>
+    <string name="enable_gpu_buffer_readback">Включить обратное чтение буфера ГПУ</string>
+    <string name="enable_gpu_buffer_readback_description">Сохраняет измененные ГПУ данные буфера путем чтения их обратно перед выгрузками. Некоторые игры требуют этого, чтобы рендерить определенные эффекты правильно. Может вызывать проблемы если оборудование не может обработать дополнительную рабочую нагрузку.</string>
     <string name="use_optimized_vertex_buffers">Оптимизированные вершинные буферы</string>
     <string name="use_optimized_vertex_buffers_description">Включает оптимизированную привязку вершинного буфера для повышения производительности. Требует Mesa Turnip 26.0+ / QCOM. Приводит к вылету на старых версиях драйверов Turnip (25.3 и ниже).</string>
 
@@ -1086,6 +1088,7 @@
     <string name="gpu_fence_behavior_immediate">Мгновенный</string>
     <string name="gpu_fence_behavior_balanced">Сбалансированный</string>
     <string name="gpu_fence_behavior_accurate">Точный</string>
+    <string name="gpu_fence_behavior_strict">Строгий</string>
 
     <string name="vram_usage_conservative">Консервативный</string>
     <string name="vram_usage_aggressive">Агрессивный</string>

src/android/app/src/main/res/values-zh-rCN/strings.xml

@@ -570,6 +570,8 @@
     <string name="renderer_reactive_flushing_description">通过牺牲性能来提升某些游戏的渲染精度。</string>
     <string name="enable_buffer_history">启用缓冲区历史</string>
     <string name="enable_buffer_history_description">启用对先前缓冲区状态的访问。此选项可在某些游戏中提升渲染质量并保持性能的一致性。</string>
+    <string name="enable_gpu_buffer_readback">启用 GPU 缓冲区回读</string>
+    <string name="enable_gpu_buffer_readback_description">在上传前回读经由 GPU 修改过的缓冲区数据，以将其保留。一些游戏会用到这项设定以正确渲染某些效果。如果硬件无法处理额外的工作负载，则可能会导致问题。</string>
     <string name="use_optimized_vertex_buffers">优化顶点缓冲区</string>
     <string name="use_optimized_vertex_buffers_description">启用经过优化的顶点缓冲区绑定以提升性能。需要 Mesa 26.0 及以上版本的 Turnip 或 QCOM 驱动程序。若使用较旧版本的 Turnip 驱动 (25.3 及以下版本) 则会导致崩溃。</string>
 
@@ -1087,6 +1089,7 @@
     <string name="gpu_fence_behavior_immediate">即时</string>
     <string name="gpu_fence_behavior_balanced">均衡</string>
     <string name="gpu_fence_behavior_accurate">精确</string>
+    <string name="gpu_fence_behavior_strict">严格</string>
 
     <string name="vram_usage_conservative">保守式</string>
     <string name="vram_usage_aggressive">主动式</string>

src/android/app/src/main/res/values-zh-rTW/strings.xml

@@ -561,6 +561,8 @@
     <string name="renderer_reactive_flushing_description">犧牲效能，以改善部分遊戲的轉譯準確度</string>
     <string name="enable_buffer_history">啟用緩衝區歷史</string>
     <string name="enable_buffer_history_description">允許存取先前的緩衝區狀態。此選項可能會改善部分遊戲的渲染品質與效能穩定性</string>
+    <string name="enable_gpu_buffer_readback">啟用 GPU 緩衝區讀回</string>
+    <string name="enable_gpu_buffer_readback_description">透過在上傳之前先將 GPU 修改過的緩衝區資料讀回來保存資料，部分遊戲需要啟用此功能才能正常渲染遊戲特效。如果硬體無法負荷可能會導致錯誤</string>
     <string name="use_optimized_vertex_buffers">最佳化頂點緩衝區</string>
     <string name="use_optimized_vertex_buffers_description">啟用最佳化的頂點緩衝區綁定。需要安裝 Mesa 26.0+ Turnip drivers/Qualcomm drivers。使用舊版 Turnip drivers 會導致當機 (25.3版和更低的版本)</string>
 

src/android/app/src/main/res/values/arrays.xml

@@ -558,6 +558,7 @@
         <item>@string/gpu_fence_behavior_immediate</item>
         <item>@string/gpu_fence_behavior_balanced</item>
         <item>@string/gpu_fence_behavior_accurate</item>
+        <item>@string/gpu_fence_behavior_strict</item>
     </string-array>
     <integer-array name="gpuFenceBehaviorValues">
         <item>0</item>

src/android/app/src/main/res/values/strings.xml

@@ -586,6 +586,8 @@
     <string name="renderer_reactive_flushing_description">Improves rendering accuracy in some games at the cost of performance.</string>
     <string name="enable_buffer_history">Enable buffer history</string>
     <string name="enable_buffer_history_description">Enables access to previous buffer states. This option may improve rendering quality and performance consistency in some games.</string>
+    <string name="enable_gpu_buffer_readback">Enable GPU Buffer Readback</string>
+    <string name="enable_gpu_buffer_readback_description">Preserves GPU-modified buffer data by reading it back before uploads. Some games require this to render certain effects properly. May cause issues if the hardware cannot handle the additional workload.</string>
     <string name="use_optimized_vertex_buffers">Optimized Vertex Buffers</string>
     <string name="use_optimized_vertex_buffers_description">Enables optimized vertex buffer binding for improved performance. Requires Mesa 26.0+ Turnip drivers/ QCOM drivers. Will crash on older Turnip drivers (25.3 and below).</string>
 
@@ -1136,6 +1138,7 @@
     <string name="gpu_fence_behavior_immediate">Immediate</string>
     <string name="gpu_fence_behavior_balanced">Balanced</string>
     <string name="gpu_fence_behavior_accurate">Accurate</string>
+    <string name="gpu_fence_behavior_strict">Strict</string>
 
     <!-- ASTC Decoding Method Choices -->
     <string name="accelerate_astc_cpu" translatable="false">CPU</string>

src/audio_core/adsp/adsp.cpp

@@ -1,4 +1,4 @@
-// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
+// SPDX-FileCopyrightText: Copyright 2025 Eden Emulator Project
 // SPDX-License-Identifier: GPL-3.0-or-later
 
 // SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
@@ -19,11 +19,6 @@ ADSP::ADSP(Core::System& system, Sink::Sink& sink) {
     }
 }
 
-void ADSP::Shutdown() {
-    audio_renderer->Stop();
-    opus_decoder->Shutdown();
-}
-
 AudioRenderer::AudioRenderer& ADSP::AudioRenderer() {
     return *audio_renderer;
 }

src/audio_core/adsp/adsp.h

@@ -1,4 +1,4 @@
-// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
+// SPDX-FileCopyrightText: Copyright 2025 Eden Emulator Project
 // SPDX-License-Identifier: GPL-3.0-or-later
 
 // SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
@@ -45,7 +45,6 @@ public:
     explicit ADSP(Core::System& system, Sink::Sink& sink);
     ~ADSP() = default;
 
-    void Shutdown();
     AudioRenderer::AudioRenderer& AudioRenderer();
     OpusDecoder::OpusDecoder& OpusDecoder();
 

src/audio_core/adsp/apps/audio_renderer/audio_renderer.cpp

@@ -34,19 +34,26 @@ void AudioRenderer::Start() {
 
     mailbox.Send(Direction::DSP, Message::InitializeOK);
     if (mailbox.Receive(Direction::Host) != Message::InitializeOK) {
-        LOG_ERROR(Service_Audio, "Host Audio Renderer -- Failed to receive shutdown message response from ADSP!");
+        LOG_ERROR(Service_Audio, "Host Audio Renderer -- Failed to receive shutdown "
+                                 "message response from ADSP!");
+        return;
     }
+    running = true;
 }
 
 void AudioRenderer::Stop() {
-    if (main_thread.joinable()) {
+    if (!running) {
-        main_thread.request_stop();
+        return;
+    }
+
     mailbox.Send(Direction::DSP, Message::Shutdown);
-        if (mailbox.Receive(Direction::Host, main_thread.get_stop_token()) != Message::Shutdown) {
+    if (mailbox.Receive(Direction::Host) != Message::Shutdown) {
-            LOG_ERROR(Service_Audio, "Host Audio Renderer -- Failed to receive shutdown message response from ADSP!");
+        LOG_ERROR(Service_Audio, "Host Audio Renderer -- Failed to receive shutdown "
+                                 "message response from ADSP!");
     }
+    main_thread.request_stop();
     main_thread.join();
-    }
+
     for (auto& stream : streams) {
         if (stream) {
             stream->Stop();
@@ -54,6 +61,7 @@ void AudioRenderer::Stop() {
             stream = nullptr;
         }
     }
+    running = false;
 }
 
 void AudioRenderer::Signal() {
@@ -61,12 +69,13 @@ void AudioRenderer::Signal() {
     Send(Direction::DSP, Message::Render);
 }
 
-void AudioRenderer::Wait(std::stop_token stop_token) {
+void AudioRenderer::Wait() {
-    auto msg = mailbox.Receive(Direction::Host, stop_token);
+    auto msg = Receive(Direction::Host);
     if (msg != Message::RenderResponse) {
         LOG_ERROR(Service_Audio,
                   "Did not receive the expected render response from the AudioRenderer! Expected "
-            "{}, got {}", Message::RenderResponse, msg);
+                  "{}, got {}",
+                  Message::RenderResponse, msg);
     }
     PostDSPClearCommandBuffer();
 }
@@ -138,9 +147,7 @@ void AudioRenderer::Main(std::stop_token stop_token) {
     constexpr u64 max_process_time{2'304'000ULL};
 
     while (!stop_token.stop_requested()) {
-        auto msg = mailbox.Receive(Direction::DSP, stop_token);
+        auto msg{mailbox.Receive(Direction::DSP)};
-        if (stop_token.stop_requested())
-            break;
         switch (msg) {
         case Message::Shutdown:
             mailbox.Send(Direction::Host, Message::Shutdown);

src/audio_core/adsp/apps/audio_renderer/audio_renderer.h

@@ -63,15 +63,13 @@ public:
      */
     void Start();
 
-    void NotifyShutdown();
-
     /**
      * Stop the AudioRenderer.
      */
     void Stop();
 
     void Signal();
-    void Wait(std::stop_token stop_token);
+    void Wait();
 
     void Send(Direction dir, u32 message);
     u32 Receive(Direction dir);
@@ -104,6 +102,8 @@ private:
     Mailbox mailbox;
     /// Main thread
     std::jthread main_thread{};
+    /// The current state
+    std::atomic<bool> running{};
     /// Shared memory of input command buffers, set by host, read by DSP
     std::array<CommandBuffer, MaxRendererSessions> command_buffers{};
     /// The command lists to process

src/audio_core/adsp/apps/opus/opus_decoder.cpp

@@ -41,16 +41,19 @@ OpusDecoder::OpusDecoder(Core::System& system_) : system{system_} {
 }
 
 OpusDecoder::~OpusDecoder() {
-    if (main_thread.joinable()) {
+    if (!running) {
+        init_thread.request_stop();
+        return;
+    }
+
     // Shutdown the thread
     Send(Direction::DSP, Message::Shutdown);
+    auto msg = Receive(Direction::Host);
+    ASSERT_MSG(msg == Message::ShutdownOK, "Expected Opus shutdown code {}, got {}",
+               Message::ShutdownOK, msg);
     main_thread.request_stop();
-        auto msg = Receive(Direction::Host, main_thread.get_stop_token());
-        ASSERT_MSG(msg == Message::ShutdownOK, "Expected Opus shutdown code {}, got {}", Message::ShutdownOK, msg);
     main_thread.join();
-    } else {
+    running = false;
-        init_thread.request_stop();
-    }
 }
 
 void OpusDecoder::Send(Direction dir, u32 message) {
@@ -70,6 +73,7 @@ void OpusDecoder::Init(std::stop_token stop_token) {
         return;
     }
     main_thread = std::jthread([this](std::stop_token st) { Main(st); });
+    running = true;
     Send(Direction::Host, Message::StartOK);
 }
 
@@ -78,15 +82,12 @@ void OpusDecoder::Main(std::stop_token stop_token) {
 
     while (!stop_token.stop_requested()) {
         auto msg = Receive(Direction::DSP, stop_token);
-        if (stop_token.stop_requested())
-            break;
-
         switch (msg) {
-        case Message::Shutdown:
+        case Shutdown:
             Send(Direction::Host, Message::ShutdownOK);
             return;
 
-        case Message::GetWorkBufferSize: {
+        case GetWorkBufferSize: {
             auto channel_count = static_cast<s32>(shared_memory->host_send_data[0]);
 
             ASSERT(IsValidChannelCount(channel_count));
@@ -95,7 +96,7 @@ void OpusDecoder::Main(std::stop_token stop_token) {
             Send(Direction::Host, Message::GetWorkBufferSizeOK);
         } break;
 
-        case Message::InitializeDecodeObject: {
+        case InitializeDecodeObject: {
             auto buffer = shared_memory->host_send_data[0];
             auto buffer_size = shared_memory->host_send_data[1];
             auto sample_rate = static_cast<s32>(shared_memory->host_send_data[2]);
@@ -112,7 +113,7 @@ void OpusDecoder::Main(std::stop_token stop_token) {
             Send(Direction::Host, Message::InitializeDecodeObjectOK);
         } break;
 
-        case Message::ShutdownDecodeObject: {
+        case ShutdownDecodeObject: {
             auto buffer = shared_memory->host_send_data[0];
             [[maybe_unused]] auto buffer_size = shared_memory->host_send_data[1];
 
@@ -122,7 +123,7 @@ void OpusDecoder::Main(std::stop_token stop_token) {
             Send(Direction::Host, Message::ShutdownDecodeObjectOK);
         } break;
 
-        case Message::DecodeInterleaved: {
+        case DecodeInterleaved: {
             auto start_time = system.CoreTiming().GetGlobalTimeUs();
 
             auto buffer = shared_memory->host_send_data[0];
@@ -160,19 +161,19 @@ void OpusDecoder::Main(std::stop_token stop_token) {
             Send(Direction::Host, Message::DecodeInterleavedOK);
         } break;
 
-        case Message::MapMemory: {
+        case MapMemory: {
             [[maybe_unused]] auto buffer = shared_memory->host_send_data[0];
             [[maybe_unused]] auto buffer_size = shared_memory->host_send_data[1];
             Send(Direction::Host, Message::MapMemoryOK);
         } break;
 
-        case Message::UnmapMemory: {
+        case UnmapMemory: {
             [[maybe_unused]] auto buffer = shared_memory->host_send_data[0];
             [[maybe_unused]] auto buffer_size = shared_memory->host_send_data[1];
             Send(Direction::Host, Message::UnmapMemoryOK);
         } break;
 
-        case Message::GetWorkBufferSizeForMultiStream: {
+        case GetWorkBufferSizeForMultiStream: {
             auto total_stream_count = static_cast<s32>(shared_memory->host_send_data[0]);
             auto stereo_stream_count = static_cast<s32>(shared_memory->host_send_data[1]);
 
@@ -183,7 +184,7 @@ void OpusDecoder::Main(std::stop_token stop_token) {
             Send(Direction::Host, Message::GetWorkBufferSizeForMultiStreamOK);
         } break;
 
-        case Message::InitializeMultiStreamDecodeObject: {
+        case InitializeMultiStreamDecodeObject: {
             auto buffer = shared_memory->host_send_data[0];
             auto buffer_size = shared_memory->host_send_data[1];
             auto sample_rate = static_cast<s32>(shared_memory->host_send_data[2]);
@@ -210,7 +211,7 @@ void OpusDecoder::Main(std::stop_token stop_token) {
             Send(Direction::Host, Message::InitializeMultiStreamDecodeObjectOK);
         } break;
 
-        case Message::ShutdownMultiStreamDecodeObject: {
+        case ShutdownMultiStreamDecodeObject: {
             auto buffer = shared_memory->host_send_data[0];
             [[maybe_unused]] auto buffer_size = shared_memory->host_send_data[1];
 
@@ -220,7 +221,7 @@ void OpusDecoder::Main(std::stop_token stop_token) {
             Send(Direction::Host, Message::ShutdownMultiStreamDecodeObjectOK);
         } break;
 
-        case Message::DecodeInterleavedForMultiStream: {
+        case DecodeInterleavedForMultiStream: {
             auto start_time = system.CoreTiming().GetGlobalTimeUs();
 
             auto buffer = shared_memory->host_send_data[0];
@@ -265,15 +266,4 @@ void OpusDecoder::Main(std::stop_token stop_token) {
     }
 }
 
-void OpusDecoder::Shutdown() {
-    if (init_thread.joinable()) {
-        init_thread.request_stop();
-        init_thread.join(); //must explicitly wait for init to join
-    }
-    if (main_thread.joinable()) {
-        main_thread.request_stop();
-        Send(Direction::DSP, Message::Shutdown);
-    }
-}
-
 } // namespace AudioCore::ADSP::OpusDecoder

src/audio_core/adsp/apps/opus/opus_decoder.h

@@ -57,7 +57,7 @@ public:
     ~OpusDecoder();
 
     bool IsRunning() const noexcept {
-        return main_thread.joinable();
+        return running;
     }
 
     void Send(Direction dir, u32 message);
@@ -67,8 +67,6 @@ public:
         shared_memory = &shared_memory_;
     }
 
-    void Shutdown();
-
 private:
     /**
      * Initializing thread, launched at audio_core boot to avoid blocking the main emu boot thread.
@@ -87,6 +85,8 @@ private:
     std::jthread init_thread{};
     /// Main thread
     std::jthread main_thread{};
+    /// The current state
+    bool running{};
     /// Structure shared with the host, input data set by the host before sending a mailbox message,
     /// and the responses are written back by the OpusDecoder.
     SharedMemory* shared_memory{};

src/audio_core/audio_core.cpp

@@ -33,7 +33,6 @@ void AudioCore::CreateSinks() {
 
 void AudioCore::Shutdown() {
     audio_manager->Shutdown();
-    adsp->Shutdown();
 }
 
 AudioManager& AudioCore::GetAudioManager() {

src/audio_core/audio_manager.cpp

@@ -17,7 +17,7 @@ AudioManager::AudioManager() {
         std::unique_lock l{events.GetAudioEventLock()};
         events.ClearEvents();
         while (!stop_token.stop_requested()) {
-            const auto timed_out = events.Wait(l, std::chrono::seconds{2});
+            const auto timed_out{events.Wait(l, std::chrono::seconds(2))};
             if (events.CheckAudioEventSet(Event::Type::Max)) {
                 break;
             }
@@ -35,9 +35,9 @@ AudioManager::AudioManager() {
 }
 
 void AudioManager::Shutdown() {
+    events.SetAudioEvent(Event::Type::Max, true);
     if (thread.joinable()) {
         thread.request_stop();
-        events.SetAudioEvent(Event::Type::Max, true);
         thread.join();
     }
 }

src/audio_core/audio_manager.h

@@ -39,7 +39,9 @@ class AudioManager {
 public:
     explicit AudioManager();
 
-    /// @brief Shutdown the audio manager.
+    /**
+     * Shutdown the audio manager.
+     */
     void Shutdown();
 
     /**

src/audio_core/renderer/sink/circular_buffer_sink_info.cpp

@@ -1,6 +1,3 @@
-// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
-// SPDX-License-Identifier: GPL-3.0-or-later
-
 // SPDX-FileCopyrightText: Copyright 2022 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
@@ -38,7 +35,7 @@ void CircularBufferSinkInfo::Update(BehaviorInfo::ErrorInfo& error_info, OutStat
     auto current_params{reinterpret_cast<CircularBufferInParameter*>(parameter.data())};
     auto current_state{reinterpret_cast<CircularBufferState*>(state.data())};
 
-    if (in_use == bool(buffer_params->in_use) && !buffer_unmapped) {
+    if (in_use == buffer_params->in_use && !buffer_unmapped) {
         error_info.error_code = ResultSuccess;
         error_info.address = CpuAddr(0);
         out_status.writeOffset = current_state->last_pos2;

src/audio_core/renderer/sink/sink_info_base.h

@@ -1,4 +1,4 @@
-// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
+// SPDX-FileCopyrightText: Copyright 2025 Eden Emulator Project
 // SPDX-License-Identifier: GPL-3.0-or-later
 
 // SPDX-FileCopyrightText: Copyright 2022 yuzu Emulator Project
@@ -31,10 +31,10 @@ public:
     };
 
     struct DeviceInParameter {
-        /* 0x000 */ u8 name[0x100];
+        /* 0x000 */ char name[0x100];
         /* 0x100 */ u32 input_count;
         /* 0x104 */ std::array<s8, MaxChannels> inputs;
-        /* 0x10A */ u8 unk10A[0x1];
+        /* 0x10A */ char unk10A[0x1];
         /* 0x10B */ bool downmix_enabled;
         /* 0x10C */ std::array<f32, 4> downmix_coeff;
     };
@@ -43,7 +43,7 @@ public:
     struct DeviceState {
         /* 0x00 */ UpsamplerInfo* upsampler_info;
         /* 0x08 */ std::array<Common::FixedPoint<16, 16>, 4> downmix_coeff;
-        /* 0x18 */ u8 unk18[0x18];
+        /* 0x18 */ char unk18[0x18];
     };
     static_assert(sizeof(DeviceState) == 0x30, "DeviceState has the wrong size!");
 
@@ -55,8 +55,8 @@ public:
         /* 0x14 */ u32 previous_pos;
         /* 0x18 */ SampleFormat format;
         /* 0x1C */ std::array<s8, MaxChannels> inputs;
-        /* 0x22 */ u8 in_use;
+        /* 0x22 */ bool in_use;
-        /* 0x23 */ u8 unk23[0x5];
+        /* 0x23 */ char unk23[0x5];
     };
     static_assert(sizeof(CircularBufferInParameter) == 0x28,
                   "CircularBufferInParameter has the wrong size!");
@@ -65,16 +65,16 @@ public:
         /* 0x00 */ u32 last_pos2;
         /* 0x04 */ s32 current_pos;
         /* 0x08 */ u32 last_pos;
-        /* 0x0C */ u8 unk0C[0x4];
+        /* 0x0C */ char unk0C[0x4];
         /* 0x10 */ AddressInfo address_info;
     };
     static_assert(sizeof(CircularBufferState) == 0x30, "CircularBufferState has the wrong size!");
 
     struct InParameter {
         /* 0x000 */ Type type;
-        /* 0x001 */ u8 in_use;
+        /* 0x001 */ bool in_use;
         /* 0x004 */ u32 node_id;
-        /* 0x008 */ u8 unk08[0x18];
+        /* 0x008 */ char unk08[0x18];
         union {
             /* 0x020 */ DeviceInParameter device;
             /* 0x020 */ CircularBufferInParameter circular_buffer;
@@ -84,7 +84,7 @@ public:
 
     struct OutStatus {
         /* 0x00 */ u32 writeOffset;
-        /* 0x04 */ u8 unk04[0x1C];
+        /* 0x04 */ char unk04[0x1C];
     }; // size == 0x20
     static_assert(sizeof(OutStatus) == 0x20, "SinkInfoBase::OutStatus has the wrong size!");
 
@@ -162,18 +162,19 @@ public:
     u8* GetParameter();
 
 protected:
+    /// Type of this sink
+    Type type{Type::Invalid};
+    /// Is this sink in use?
+    bool in_use{};
+    /// Is this sink's buffer unmapped? Circular only
+    bool buffer_unmapped{};
+    /// Node id for this sink
+    u32 node_id{};
     /// State buffer for this sink
     std::array<u8, (std::max)(sizeof(DeviceState), sizeof(CircularBufferState))> state{};
     /// Parameter buffer for this sink
-    std::array<u8, (std::max)(sizeof(DeviceInParameter), sizeof(CircularBufferInParameter))> parameter{};
+    std::array<u8, (std::max)(sizeof(DeviceInParameter), sizeof(CircularBufferInParameter))>
-    /// Type of this sink
+        parameter{};
-    Type type{Type::Invalid};
-    /// Node id for this sink
-    u32 node_id{};
-    /// Is this sink in use?
-    bool in_use : 1 = false;
-    /// Is this sink's buffer unmapped? Circular only
-    bool buffer_unmapped : 1 = false;
 };
 
 } // namespace AudioCore::Renderer

src/audio_core/renderer/system.cpp

@@ -100,9 +100,8 @@ u64 System::GetWorkBufferSize(const AudioRendererParameterInternal& params) {
 }
 
 System::System(Core::System& core_, Kernel::KEvent* adsp_rendered_event_)
-    : core{core_}
+    : core{core_}, audio_renderer{core.AudioCore().ADSP().AudioRenderer()},
-    , adsp_rendered_event{adsp_rendered_event_}
+      adsp_rendered_event{adsp_rendered_event_} {}
-{}
 
 Result System::Initialize(const AudioRendererParameterInternal& params,
                           Kernel::KTransferMemory* transfer_memory, u64 transfer_memory_size,
@@ -407,7 +406,7 @@ void System::Finalize() {
         return;
     }
 
-    if (IsActive()) {
+    if (active) {
         Stop();
     }
 
@@ -434,12 +433,14 @@ void System::Start() {
     std::scoped_lock l{lock};
     frames_elapsed = 0;
     state = State::Started;
+    active = true;
 }
 
 void System::Stop() {
     {
         std::scoped_lock l{lock};
         state = State::Stopped;
+        active = false;
     }
 
     if (execution_mode == ExecutionMode::Auto) {
@@ -479,7 +480,7 @@ Result System::Update(std::span<const u8> input, std::span<u8> performance, std:
         return result;
     }
 
-    result = info_updater.UpdateEffects(effect_context, IsActive(), memory_pool_workbuffer,
+    result = info_updater.UpdateEffects(effect_context, active, memory_pool_workbuffer,
                                         memory_pool_count);
     if (result.IsError()) {
         LOG_ERROR(Service_Audio, "Failed to update Effects!");
@@ -581,16 +582,16 @@ u32 System::GetRenderingDevice() const {
 }
 
 bool System::IsActive() const {
-    return state == State::Started;
+    return active;
 }
 
 void System::SendCommandToDsp() {
     std::scoped_lock l{lock};
 
     if (initialized) {
-        if (IsActive()) {
+        if (active) {
             terminate_event.Reset();
-            const auto remaining_command_count = core.AudioCore().ADSP().AudioRenderer().GetRemainCommandCount(session_id);
+            const auto remaining_command_count{audio_renderer.GetRemainCommandCount(session_id)};
             u64 command_size{0};
 
             if (remaining_command_count) {
@@ -617,17 +618,16 @@ void System::SendCommandToDsp() {
             auto time_limit{
                 static_cast<u64>((time_limit_percent / 100) * 2'880'000.0 *
                                  (static_cast<f32>(render_time_limit_percent) / 100.0f))};
-            core.AudioCore().ADSP().AudioRenderer().SetCommandBuffer(session_id,
+            audio_renderer.SetCommandBuffer(session_id, translated_addr, command_size, time_limit,
-                translated_addr, command_size,
+                                            applet_resource_user_id, process_handle,
-                time_limit, applet_resource_user_id,
+                                            reset_command_buffers);
-                process_handle, reset_command_buffers);
             reset_command_buffers = false;
             command_buffer_size = command_size;
             if (remaining_command_count == 0) {
                 adsp_rendered_event->Signal(core.Kernel());
             }
         } else {
-            core.AudioCore().ADSP().AudioRenderer().ClearRemainCommandCount(session_id);
+            audio_renderer.ClearRemainCommandCount(session_id);
             terminate_event.Set();
         }
     }
@@ -738,7 +738,7 @@ u64 System::GenerateCommand(std::span<u8> in_command_buffer,
     const auto end_time{core.CoreTiming().GetGlobalTimeNs().count()};
     total_ticks_elapsed += end_time - start_time;
     num_command_lists_generated++;
-    render_start_tick = core.AudioCore().ADSP().AudioRenderer().GetRenderingStartTick(session_id);
+    render_start_tick = audio_renderer.GetRenderingStartTick(session_id);
     frames_elapsed++;
 
     return command_buffer.size;

src/audio_core/renderer/system.h

@@ -1,6 +1,3 @@
-// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
-// SPDX-License-Identifier: GPL-3.0-or-later
-
 // SPDX-FileCopyrightText: Copyright 2022 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
@@ -222,6 +219,8 @@ public:
 private:
     /// Core system
     Core::System& core;
+    /// Reference to the ADSP's AudioRenderer for communication
+    ::AudioCore::ADSP::AudioRenderer::AudioRenderer& audio_renderer;
     /// Is this system initialized?
     bool initialized{};
     /// Is this system currently active?

src/audio_core/renderer/system_manager.cpp

@@ -1,4 +1,4 @@
-// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
+// SPDX-FileCopyrightText: Copyright 2025 Eden Emulator Project
 // SPDX-License-Identifier: GPL-3.0-or-later
 
 // SPDX-FileCopyrightText: Copyright 2022 yuzu Emulator Project
@@ -11,70 +11,73 @@
 #include "audio_core/renderer/system_manager.h"
 #include "common/thread.h"
 #include "core/core.h"
-#include "core/hle/kernel/kernel.h"
-#include "core/hle/kernel/k_scheduler.h"
 #include "core/core_timing.h"
 
 namespace AudioCore::Renderer {
 
 SystemManager::SystemManager(Core::System& core_)
-    : core{core_}
+    : core{core_}, audio_renderer{core.AudioCore().ADSP().AudioRenderer()} {}
-{
-}
 
 SystemManager::~SystemManager() {
     Stop();
 }
 
 void SystemManager::InitializeUnsafe() {
-    if (!thread.joinable()) {
+    if (!active) {
-        core.AudioCore().ADSP().AudioRenderer().Start();
+        active = true;
-        thread = core.Kernel().RunOnHostCoreThread("AudioRenderSystemManager", [this]() {
+        audio_renderer.Start();
-            auto& audio_renderer = core.AudioCore().ADSP().AudioRenderer();
+        thread = std::jthread([this](std::stop_token stop_token) {
-            auto const stop_token = thread.get_stop_token();
+            Common::SetCurrentThreadName("AudioRenderSystemManager");
-            while (!stop_token.stop_requested()) {
+            Common::SetCurrentThreadPriority(Common::ThreadPriority::High);
+            while (active && !stop_token.stop_requested()) {
                 {
-                    std::scoped_lock lk{mutex};
+                    std::scoped_lock l{mutex1};
                     for (auto system : systems)
                         system->SendCommandToDsp();
                 }
                 audio_renderer.Signal();
-                audio_renderer.Wait(stop_token);
+                audio_renderer.Wait();
             }
         });
     }
 }
 
 void SystemManager::Stop() {
-    core.AudioCore().ADSP().AudioRenderer().Stop();
+    if (active) {
-    if (thread.joinable()) {
+        active = false;
         thread.request_stop();
         thread.join();
+        audio_renderer.Stop();
     }
 }
 
 bool SystemManager::Add(System& system_) {
-    std::scoped_lock lk{mutex};
+    std::scoped_lock l2{mutex2};
-    if (systems.size() >= MaxRendererSessions) {
+    if (systems.size() + 1 > MaxRendererSessions) {
         LOG_ERROR(Service_Audio, "Maximum AudioRenderer Systems active, cannot add more!");
         return false;
     }
-    systems.push_back(&system_);
+    {
-    if (!systems.empty()) {
+        std::scoped_lock l{mutex1};
+        if (systems.empty())
             InitializeUnsafe();
     }
+    systems.push_back(&system_);
     return true;
 }
 
 bool SystemManager::Remove(System& system_) {
-    std::scoped_lock lk{mutex};
+    std::scoped_lock l2{mutex2};
+    {
+        std::scoped_lock l{mutex1};
         if (systems.remove(&system_) == 0) {
             LOG_ERROR(Service_Audio, "Failed to remove a render system, it was not found in the list!");
             return false;
         }
-    if (systems.empty()) {
-        Stop();
     }
+
+    if (systems.empty())
+        Stop();
     return true;
 }
 

src/audio_core/renderer/system_manager.h

@@ -1,4 +1,4 @@
-// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
+// SPDX-FileCopyrightText: Copyright 2025 Eden Emulator Project
 // SPDX-License-Identifier: GPL-3.0-or-later
 
 // SPDX-FileCopyrightText: Copyright 2022 yuzu Emulator Project
@@ -75,8 +75,14 @@ private:
     std::list<System*> systems{};
     /// Main worker thread for generating command lists
     std::jthread thread;
-    /// Mutex for the reading and adding/removing systems
+    /// Mutex for the systems
-    std::mutex mutex{};
+    std::mutex mutex1{};
+    /// Mutex for adding/removing systems
+    std::mutex mutex2{};
+    /// Is the system manager thread active?
+    std::atomic<bool> active{};
+    /// Reference to the ADSP's AudioRenderer for communication
+    ::AudioCore::ADSP::AudioRenderer::AudioRenderer& audio_renderer;
 };
 
 } // namespace AudioCore::Renderer

src/common/CMakeLists.txt

@@ -137,8 +137,6 @@ add_library(
     uuid.cpp
     uuid.h
     vector_math.h
-    zbic_compression.cpp
-    zbic_compression.h
     zstd_compression.cpp
     zstd_compression.h
     fs/ryujinx_compat.h fs/ryujinx_compat.cpp
@@ -149,10 +147,6 @@ add_library(
     net/net.h net/net.cpp
     container/unordered_map.h container/unordered_set.h)
 
-set_source_files_properties(zbic_compression.cpp PROPERTIES
-    INCLUDE_DIRECTORIES "${ZBIC_INCLUDE_DIR}"
-    COMPILE_OPTIONS "$<$<CXX_COMPILER_ID:Clang,GNU>:-Wno-unused-function;-Wno-missing-declarations;-Wno-shadow>")
-
 if(WIN32)
     target_sources(common PRIVATE windows/timer_resolution.cpp
         windows/timer_resolution.h)

src/common/bounded_threadsafe_queue.h

@@ -1,4 +1,4 @@
-// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
+// SPDX-FileCopyrightText: Copyright 2025 Eden Emulator Project
 // SPDX-License-Identifier: GPL-3.0-or-later
 // SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
@@ -109,6 +109,12 @@ private:
             if (read_index == producer.index.load(std::memory_order::acquire)) {
                 return false;
             }
+        } else if constexpr (Mode == PopMode::Wait) {
+            // Wait until the queue is not empty.
+            std::unique_lock lock{consumer.cv_mutex};
+            consumer.cv.wait(lock, [this, read_index] {
+                return read_index != producer.index.load(std::memory_order::acquire);
+            });
         } else if constexpr (Mode == PopMode::WaitWithStopToken) {
             // Wait until the queue is not empty.
             std::unique_lock lock{consumer.cv_mutex};
@@ -118,12 +124,6 @@ private:
             if (stop_token.stop_requested()) {
                 return false;
             }
-        } else if constexpr (Mode == PopMode::Wait) {
-            // Wait until the queue is not empty.
-            std::unique_lock lock{consumer.cv_mutex};
-            consumer.cv.wait(lock, [this, read_index] {
-                return read_index != producer.index.load(std::memory_order::acquire);
-            });
         } else {
             static_assert(Mode < PopMode::Count, "Invalid PopMode.");
         }

src/common/settings.cpp

@@ -178,6 +178,10 @@ bool IsGPUFenceBehaviorAccurate() {
     return values.gpu_fence_behavior.GetValue() == GpuFenceBehavior::Accurate;
 }
 
+bool IsGPUFenceBehaviorStrict() {
+    return values.gpu_fence_behavior.GetValue() == GpuFenceBehavior::Strict;
+}
+
 bool IsFastmemEnabled() {
     if (values.cpu_accuracy.GetValue() == Settings::CpuAccuracy::Debugging)
         return bool(values.cpuopt_fastmem);

src/common/settings.h

@@ -525,7 +525,7 @@ struct Values {
     SwitchableSetting<GpuFenceBehavior, true> gpu_fence_behavior{linkage,
                                                                  GpuFenceBehavior::Default,
                                                                  GpuFenceBehavior::Default,
-                                                                 GpuFenceBehavior::Accurate,
+                                                                 GpuFenceBehavior::Strict,
                                                                  "gpu_fence_behavior",
                                                                  Category::RendererAdvanced,
                                                                  Specialization::Default,
@@ -655,6 +655,13 @@ struct Values {
 
     SwitchableSetting<bool> rescale_hack{linkage, false, "rescale_hack",
                                          Category::RendererHacks};
+    SwitchableSetting<bool> enable_gpu_buffer_readback{linkage,
+                                                       false,
+                                                       "enable_gpu_buffer_readback",
+                                                       Category::RendererAdvanced,
+                                                       Specialization::Default,
+                                                       true,
+                                                       true};
 
     SwitchableSetting<bool> use_asynchronous_shaders{linkage, false, "use_asynchronous_shaders",
                                                      Category::RendererHacks};
@@ -972,6 +979,7 @@ bool IsDMALevelSafe();
 bool IsGPUFenceBehaviorDefault();
 bool IsGPUFenceBehaviorBalanced();
 bool IsGPUFenceBehaviorAccurate();
+bool IsGPUFenceBehaviorStrict();
 
 bool IsFastmemEnabled();
 void SetNceEnabled(bool is_64bit);

src/common/settings_enums.h

@@ -137,7 +137,7 @@ ENUM(VramUsageMode, Conservative, Aggressive);
 ENUM(RendererBackend, OpenGL_GLSL, Vulkan, Null, OpenGL_GLASM, OpenGL_SPIRV);
 ENUM(GpuAccuracy, Low, High);
 ENUM(DmaAccuracy, Default, Unsafe, Safe);
-ENUM(GpuFenceBehavior, Default, Immediate, Balanced, Accurate);
+ENUM(GpuFenceBehavior, Default, Immediate, Balanced, Accurate, Strict);
 ENUM(CpuBackend, Dynarmic, Nce);
 ENUM(CpuAccuracy, Auto, Accurate, Unsafe, Paranoid, Debugging);
 ENUM(CpuClock, Normal, Boost, Overclock)

src/common/sparse_large_vector.cpp

@@ -6,7 +6,6 @@
 // SPDX-License-Identifier: GPL-2.0-or-later
 
 #ifdef _WIN32
-#include <algorithm>
 #include <windows.h>
 #include <mutex>
 #else
@@ -21,22 +20,19 @@ namespace Common {
 
 #ifdef _WIN32
 static std::vector<std::pair<u64, u64>> vector_regions {};
-static std::mutex vector_regions_mutex {};
 
 // Workaround for handling non-commited memory accessed by Dynarmic; usually result of an error
 static LONG WINAPI FakePageFaultHandler(PEXCEPTION_POINTERS info) {
     DWORD code = info->ExceptionRecord->ExceptionCode;
-    u64 exception_addr = reinterpret_cast<u64>(info->ExceptionRecord->ExceptionInformation[1]);
+    u64 exception_addr = reinterpret_cast<u64>(info->ExceptionRecord->ExceptionAddress);
 
-    if (code != EXCEPTION_ACCESS_VIOLATION || info->ExceptionRecord->ExceptionInformation[0] == 1) {
+    if (code != EXCEPTION_ACCESS_VIOLATION) {
         // Not our problem
         return EXCEPTION_CONTINUE_SEARCH;
     }
 
     u64 addr = 0, addr2 = 0;
 
-    {
-        std::lock_guard lock(vector_regions_mutex);
     for (auto region: vector_regions) {
         auto addr_shifted = exception_addr >> HostPageBits;
         if (region.first <= addr_shifted && addr_shifted <= region.second) {
@@ -52,7 +48,6 @@ static LONG WINAPI FakePageFaultHandler(PEXCEPTION_POINTERS info) {
             break;
         }
     }
-    }
 
     if (addr == 0 && addr2 == 0) {
         // Not our problem
@@ -82,16 +77,8 @@ bool CommitVectorPage(uintptr_t addr, bool write) noexcept {
     auto res = VirtualQuery(reinterpret_cast<void*>(addr), &info, sizeof(info));
     if (res == 0) {
         LOG_CRITICAL(HW_Memory, "Failed to query large buffer region at {:#x} with error {}, will try committing anyway", addr, GetLastError());
-    } else if (info.State == MEM_COMMIT) {
-        DWORD old_protect {};
-        auto perm = write ? PAGE_READWRITE : PAGE_READONLY;
-        if (!VirtualProtect(reinterpret_cast<void*>(addr), HostPageSize, perm, &old_protect)) {
-            LOG_ERROR(HW_Memory, "Failed to change permissions of large buffer region at {:#x}, error {}", addr, GetLastError());
-            return false;
-        }
-        return true;
     } else if (info.State != MEM_RESERVE) {
-        LOG_ERROR(HW_Memory, "Tried to commit an unreserved large buffer region at {:#x} that is not mapped (state {:#x})", addr, info.State);
+        LOG_ERROR(HW_Memory, "Tried to commit an unreserved large buffer region at {:#x} that is not mapped or is already committed (state {:#x})", addr, info.State);
         return false;
     }
 
@@ -136,8 +123,7 @@ void* AllocateMemoryPages(std::size_t size) noexcept {
     void* base = VirtualAlloc(nullptr, size, MEM_RESERVE, PAGE_READWRITE);
 
     if (base != nullptr) {
-        std::lock_guard lock(vector_regions_mutex);
+        vector_regions.emplace_back(reinterpret_cast<u64>(base), reinterpret_cast<u64>(base) + size);
-        vector_regions.emplace_back(reinterpret_cast<u64>(base) >> HostPageBits, (reinterpret_cast<u64>(base) + size) >> HostPageBits);
 
         static std::once_flag flag;
         std::call_once(flag, []() { AddVectoredExceptionHandler(1, FakePageFaultHandler); });
@@ -163,8 +149,6 @@ void FreeMemoryPages(void* base, [[maybe_unused]] std::size_t size) noexcept {
     if (!base)
         return;
 #ifdef _WIN32
-    std::lock_guard lock(vector_regions_mutex);
-    std::erase_if(vector_regions, [base](const auto& r) {return r.first == reinterpret_cast<u64>(base); });
     ASSERT(VirtualFree(base, 0, MEM_RELEASE));
 #else
     ASSERT(munmap(base, size) == 0);

src/common/sparse_large_vector.h

@@ -28,9 +28,9 @@ constexpr u64 HostPageBits = 12;
 constexpr u64 HostPageMask = ~(HostPageSize - 1);
 bool CommitVectorPage(uintptr_t addr, bool write) noexcept;
 #else
-inline const u64 HostPageSize = sysconf(_SC_PAGESIZE);
+const u64 HostPageSize = sysconf(_SC_PAGESIZE);
-inline const u64 HostPageBits = std::countr_zero(HostPageSize);
+const u64 HostPageBits = std::countr_zero(HostPageSize);
-inline const u64 HostPageMask = ~(HostPageSize - 1);
+const u64 HostPageMask = ~(HostPageSize - 1);
 #endif
 
 void* AllocateMemoryPages(std::size_t size) noexcept;
@@ -81,8 +81,8 @@ public:
             UNREACHABLE_MSG("Out of bounds RW access on SparseLargeVector @ {}", index);
         }
 
-        if (!IsCommittedPage(index) && !CommitPage(index)) {
+        if (!IsCommittedPage(index)) {
-            UNREACHABLE_MSG("Cannot access SparseLargeVector index {} with RW permission", index);
+            CommitPage(index);
         }
         return base_ptr[index];
     }
@@ -103,7 +103,8 @@ public:
             LOG_CRITICAL(Common_Memory, "Out of bounds write on SparseLargeVector @ {}", index);
             return;
         }
-        if (IsCommittedPage(index) || CommitPage(index))
+        if (!IsCommittedPage(index))
+            CommitPage(index);
         base_ptr[index] = value;
     }
 
@@ -176,22 +177,16 @@ private:
         return (val >> (page & 63)) & 1;
     }
 
-    constexpr bool CommitPage(std::size_t index) noexcept {
+    constexpr void CommitPage(std::size_t index) noexcept {
         auto page_index = (index * sizeof(T)) >> HostPageBits;
         auto page = reinterpret_cast<uintptr_t>(base_ptr + index) & HostPageMask;
 #if defined(_WIN32)
-        if (!CommitVectorPage(page, true)) {
+        CommitVectorPage(page, true);
-            return false;
-        }
 #else
-        if (mprotect(reinterpret_cast<void*>(page), HostPageSize, PROT_READ | PROT_WRITE) != 0) {
+        mprotect(reinterpret_cast<void*>(page), HostPageSize, PROT_READ | PROT_WRITE);
-            LOG_ERROR(Common_Memory, "Failed to commit large buffer region at index {}, error {}", index, strerror(errno));
-            return false;
-        }
 #endif
 
         committed_pages[page_index >> 6].fetch_or(1ULL << (page_index & 63), std::memory_order_release);
-        return true;
     }
 
     constexpr void DecommitPage(std::size_t index) noexcept {

src/common/zbic_compression.cpp

@@ -1,46 +0,0 @@
-// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
-// SPDX-License-Identifier: GPL-3.0-or-later
-
-#include <cstring>
-
-#include "common/zbic_compression.h"
-
-#define ZSTD_ZBIC_SUPPORT 1
-#define ZSTDLIB_VISIBLE static
-#define ZSTDLIB_HIDDEN static
-#define ZSTDERRORLIB_VISIBLE static
-#define ZSTDERRORLIB_HIDDEN static
-#undef ZSTD_MULTITHREAD
-
-#if defined(__ANDROID__)
-#undef _GNU_SOURCE
-#endif
-
-#include "zstd.h"
-#define g_ZSTD_threading_useless_symbol g_ZSTD_zbic_threading_useless_symbol
-#include "zstd.c"
-#undef g_ZSTD_threading_useless_symbol
-
-namespace Common::Compression {
-
-bool IsZBIC(std::span<const u8> src) {
-    if (src.size() < sizeof(u32)) {
-        return false;
-    }
-    u32 magic = 0;
-    std::memcpy(&magic, src.data(), sizeof(u32));
-    return magic == ZSTD_MAGICNUMBER; // 0x4349425A ("ZBIC")
-}
-
-int DecompressDataZBIC(std::span<u8> dst, std::span<const u8> src) {
-    if (dst.empty() || src.empty()) {
-        return -1;
-    }
-    const size_t res = ZSTD_decompress(dst.data(), dst.size(), src.data(), src.size());
-    if (ZSTD_isError(res)) {
-        return -1;
-    }
-    return static_cast<int>(res);
-}
-
-} // namespace Common::Compression

src/common/zbic_compression.h

@@ -1,15 +0,0 @@
-// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
-// SPDX-License-Identifier: GPL-3.0-or-later
-
-#pragma once
-
-#include <span>
-#include "common/common_types.h"
-
-namespace Common::Compression {
-
-[[nodiscard]] bool IsZBIC(std::span<const u8> src);
-
-[[nodiscard]] int DecompressDataZBIC(std::span<u8> dst, std::span<const u8> src);
-
-} // namespace Common::Compression

src/core/core.cpp

@@ -110,13 +110,8 @@ FileSys::VirtualFile GetGameFileFromPath(const FileSys::VirtualFilesystem& vfs,
 
 struct System::Impl {
     explicit Impl(System& system)
-        : kernel{system}
+        : kernel{system}, fs_controller{system}, hid_core{kernel}, cpu_manager{system},
-        , fs_controller{system}
+          reporter{system}, applet_manager{system}, frontend_applets{system}, profile_manager{} {}
-        , hid_core{kernel}
-        , cpu_manager{system}
-        , reporter{system}
-        , profile_manager{}
-    {}
 
     u64 program_id;
 
@@ -129,12 +124,6 @@ struct System::Impl {
         core_timing.SetMulticore(is_multicore);
         core_timing.Initialize([&system]() { system.RegisterHostThread(); });
 
-        applet_manager.emplace(system);
-        frontend_applets.emplace(system);
-        apm_controller.emplace(core_timing);
-        arp_manager.emplace();
-        profile_manager.emplace();
-
         // Create a default fs if one doesn't already exist.
         if (virtual_filesystem == nullptr) {
             virtual_filesystem = std::make_shared<FileSys::RealVfsFilesystem>();
@@ -144,7 +133,7 @@ struct System::Impl {
         }
 
         // Create default implementations of applets if one is not provided.
-        frontend_applets->SetDefaultAppletsIfMissing();
+        frontend_applets.SetDefaultAppletsIfMissing();
 
         auto const is_async_gpu = Settings::values.use_asynchronous_gpu_emulation.GetValue();
 
@@ -387,7 +376,7 @@ struct System::Impl {
 
         // Register with applet manager
         // All threads are started, begin main process execution, now that we're in the clear
-        applet_manager->CreateAndInsertByFrontendAppletParameters(std::move(process), params);
+        applet_manager.CreateAndInsertByFrontendAppletParameters(std::move(process), params);
 
         if (Settings::values.gamecard_inserted) {
             if (Settings::values.gamecard_current_game) {
@@ -439,27 +428,13 @@ struct System::Impl {
         core_timing.SyncPause(false);
         Network::CancelPendingSocketOperations();
         kernel.SuspendEmulation(true);
-        kernel.ShutdownCores();
-
-        // Notify services helpers of shutdown
-        audio_core->Shutdown();
-
-        // Wait for threads/services to join
         kernel.CloseServices();
-
+        kernel.ShutdownCores();
-        // service shutdown
         services.reset();
         service_manager.reset();
-        frontend_applets.reset();
-        applet_manager.reset();
-        apm_controller.reset();
-        arp_manager.reset();
-        profile_manager.reset();
-
         fs_controller.Reset();
         cheat_engine.reset();
         core_timing.ClearPendingEvents();
-        core_timing.Reset();
         app_loader.reset();
         audio_core.reset();
         gpu_core.reset();
@@ -476,6 +451,9 @@ struct System::Impl {
             room_member->SendGameInfo(game_info);
         }
 
+        // Reset all glue registrations
+        arp_manager.ResetAll();
+
         LOG_DEBUG(Core, "Shutdown OK");
     }
 
@@ -504,13 +482,13 @@ struct System::Impl {
     CpuManager cpu_manager;
     Reporter reporter;
     /// Applets
-    std::optional<Service::AM::AppletManager> applet_manager;
+    Service::AM::AppletManager applet_manager;
-    std::optional<Service::AM::Frontend::FrontendAppletHolder> frontend_applets;
+    Service::AM::Frontend::FrontendAppletHolder frontend_applets;
     /// APM (Performance) services
-    std::optional<Service::APM::Controller> apm_controller;
+    Service::APM::Controller apm_controller{core_timing};
     /// Service State
-    std::optional<Service::Glue::ARPManager> arp_manager;
+    Service::Glue::ARPManager arp_manager;
-    std::optional<Service::Account::ProfileManager> profile_manager;
+    Service::Account::ProfileManager profile_manager;
     /// Network instance
     Network::NetworkInstance network_instance;
     Core::SpeedLimiter speed_limiter;
@@ -837,19 +815,19 @@ void System::RegisterCheatList(const std::vector<Memory::CheatEntry>& list,
 }
 
 void System::SetFrontendAppletSet(Service::AM::Frontend::FrontendAppletSet&& set) {
-    impl->frontend_applets->SetFrontendAppletSet(std::move(set));
+    impl->frontend_applets.SetFrontendAppletSet(std::move(set));
 }
 
 Service::AM::Frontend::FrontendAppletHolder& System::GetFrontendAppletHolder() {
-    return *impl->frontend_applets;
+    return impl->frontend_applets;
 }
 
 const Service::AM::Frontend::FrontendAppletHolder& System::GetFrontendAppletHolder() const {
-    return *impl->frontend_applets;
+    return impl->frontend_applets;
 }
 
 Service::AM::AppletManager& System::GetAppletManager() {
-    return *impl->applet_manager;
+    return impl->applet_manager;
 }
 
 void System::SetContentProvider(std::unique_ptr<FileSys::ContentProviderUnion> provider) {
@@ -890,27 +868,27 @@ const Reporter& System::GetReporter() const {
 }
 
 Service::Glue::ARPManager& System::GetARPManager() {
-    return *impl->arp_manager;
+    return impl->arp_manager;
 }
 
 const Service::Glue::ARPManager& System::GetARPManager() const {
-    return *impl->arp_manager;
+    return impl->arp_manager;
 }
 
 Service::APM::Controller& System::GetAPMController() {
-    return *impl->apm_controller;
+    return impl->apm_controller;
 }
 
 const Service::APM::Controller& System::GetAPMController() const {
-    return *impl->apm_controller;
+    return impl->apm_controller;
 }
 
 Service::Account::ProfileManager& System::GetProfileManager() {
-    return *impl->profile_manager;
+    return impl->profile_manager;
 }
 
 const Service::Account::ProfileManager& System::GetProfileManager() const {
-    return *impl->profile_manager;
+    return impl->profile_manager;
 }
 
 void System::SetExitLocked(bool locked) {

src/core/core_timing.cpp

@@ -304,10 +304,10 @@ std::optional<s64> CoreTiming::Advance() {
 
 void CoreTiming::Reset() {
     paused = true;
-    if (timer_thread.joinable()) {
-        timer_thread.request_stop();
     pause_event.Set();
     event.Set();
+    if (timer_thread.joinable()) {
+        timer_thread.request_stop();
         timer_thread.join();
     }
     has_started = false;

src/core/hle/kernel/global_scheduler_context.cpp

@@ -72,9 +72,11 @@ void GlobalSchedulerContext::UnregisterDummyThreadForWakeup(KThread* thread) noe
 
 void GlobalSchedulerContext::WakeupWaitingDummyThreads(KernelCore& kernel) noexcept {
     ASSERT(this->IsLocked());
+    if (m_woken_dummy_threads.size() > 0) {
         for (auto* thread : m_woken_dummy_threads)
             thread->DummyThreadEndWait(kernel);
         m_woken_dummy_threads.clear();
+    }
 }
 
 } // namespace Kernel

src/core/hle/kernel/k_process.cpp

@@ -59,6 +59,7 @@ Result TerminateChildren(KernelCore& kernel, KProcess* process, const KThread* t
         KThread* cur_child = nullptr;
         {
             KScopedLightLock proc_lk(process->GetListLock());
+
             auto& thread_list = process->GetThreadList();
             for (auto it = thread_list.begin(); it != thread_list.end(); ++it) {
                 if (KThread* thread = std::addressof(*it); thread != thread_to_not_terminate) {

src/core/hle/kernel/k_thread.cpp

@@ -1236,13 +1236,18 @@ namespace Kernel {
         KScopedSchedulerLock sl{kernel};
 
         // Determine if this is the first termination request.
+        const bool first_request = [&]() -> bool {
             // Perform an atomic compare-and-swap from false to true.
             bool expected = false;
+            return m_termination_requested.compare_exchange_strong(expected, true);
+        }();
+
         // If this is the first request, start termination procedure.
-        if (m_termination_requested.compare_exchange_strong(expected, true)) {
+        if (first_request) {
             // If the thread is in initialized state, just change state to terminated.
             if (this->GetState() == ThreadState::Initialized) {
-                return m_thread_state = ThreadState::Terminated;
+                m_thread_state = ThreadState::Terminated;
+                return ThreadState::Terminated;
             }
 
             // Register the terminating dpc.
@@ -1276,6 +1281,7 @@ namespace Kernel {
                 m_wait_queue->CancelWait(kernel, this, ResultTerminationRequested, true);
             }
         }
+
         return this->GetState();
     }
 

src/core/hle/kernel/k_worker_task_manager.cpp

@@ -4,9 +4,7 @@
 // SPDX-FileCopyrightText: Copyright 2022 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
-#include <thread>
 #include "common/assert.h"
-#include "common/thread.h"
 #include "core/hle/kernel/k_process.h"
 #include "core/hle/kernel/k_thread.h"
 #include "core/hle/kernel/k_worker_task.h"
@@ -28,33 +26,7 @@ void KWorkerTask::DoWorkerTask(KernelCore& kernel) {
     }
 }
 
-KWorkerTaskManager::KWorkerTaskManager(KernelCore& kernel) {
+KWorkerTaskManager::KWorkerTaskManager() : m_waiting_thread(1, "KWorkerTaskManager") {}
-    m_waiting_thread = std::jthread([&kernel, this](std::stop_token stop_token) {
-        Common::SetCurrentThreadName("KWorkerManager");
-        while (!stop_token.stop_requested()) {
-            KWorkerTask* t;
-            {
-                std::unique_lock lk{m_task_mutex};
-                m_task_cv.wait(lk, m_waiting_thread.get_stop_token(), [&]() {
-                    return !m_task_queue.empty();
-                });
-                if (stop_token.stop_requested())
-                    break;
-                t = m_task_queue.back();
-                m_task_queue.pop_back();
-            }
-            t->DoWorkerTask(kernel);
-        }
-    });
-}
-
-KWorkerTaskManager::~KWorkerTaskManager() {
-    if (m_waiting_thread.joinable()) {
-        m_waiting_thread.request_stop();
-        m_task_cv.notify_one();
-        m_waiting_thread.join();
-    }
-}
 
 void KWorkerTaskManager::AddTask(KernelCore& kernel, WorkerType type, KWorkerTask* task) {
     ASSERT(type <= WorkerType::Count);
@@ -63,11 +35,10 @@ void KWorkerTaskManager::AddTask(KernelCore& kernel, WorkerType type, KWorkerTas
 
 void KWorkerTaskManager::AddTask(KernelCore& kernel, KWorkerTask* task) {
     KScopedSchedulerLock sl(kernel);
-    {
+    m_waiting_thread.QueueWork([&kernel, task]() {
-        std::scoped_lock lk{m_task_mutex};
+        // Do the task.
-        m_task_queue.emplace_back(task);
+        task->DoWorkerTask(kernel);
-    }
+    });
-    m_task_cv.notify_one();
 }
 
 } // namespace Kernel

src/core/hle/kernel/k_worker_task_manager.h

@@ -1,13 +1,8 @@
-// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
-// SPDX-License-Identifier: GPL-3.0-or-later
-
 // SPDX-FileCopyrightText: Copyright 2022 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
 #pragma once
 
-#include <condition_variable>
-#include <thread>
 #include "common/common_types.h"
 #include "common/thread_worker.h"
 
@@ -23,17 +18,15 @@ public:
         Count,
     };
 
-    KWorkerTaskManager(KernelCore& kernel);
+    KWorkerTaskManager();
-    ~KWorkerTaskManager();
 
     static void AddTask(KernelCore& kernel, WorkerType type, KWorkerTask* task);
+
 private:
     void AddTask(KernelCore& kernel, KWorkerTask* task);
-    std::jthread m_waiting_thread;
 
-    std::mutex m_task_mutex;
+private:
-    std::condition_variable_any m_task_cv;
+    Common::ThreadWorker m_waiting_thread;
-    std::vector<KWorkerTask*> m_task_queue;
 };
 
 } // namespace Kernel

src/core/hle/kernel/kernel.cpp

@@ -80,10 +80,7 @@ struct KernelCore::Impl {
     // so it will be statically given a TLS slot anyways.
     static inline thread_local ThreadLocalData tls_data = {};
 
-    explicit Impl(Core::System& system_, KernelCore& kernel_)
+    explicit Impl(Core::System& system_, KernelCore& kernel_) : system{system_} {
-        : worker_task_manager{kernel_}
-        , system{system_}
-    {
         tls_data.lock = true;
     }
 
@@ -970,16 +967,12 @@ Kernel::KHardwareTimer& KernelCore::HardwareTimer() {
     return *impl->hardware_timer;
 }
 
-KAutoObjectWithListContainer* KernelCore::ObjectListContainer() {
+KAutoObjectWithListContainer& KernelCore::ObjectListContainer() {
-    if (!impl->global_object_list_container)
+    return *impl->global_object_list_container;
-        return nullptr;
-    return std::addressof(*impl->global_object_list_container);
 }
 
-const KAutoObjectWithListContainer* KernelCore::ObjectListContainer() const {
+const KAutoObjectWithListContainer& KernelCore::ObjectListContainer() const {
-    if (!impl->global_object_list_container)
+    return *impl->global_object_list_container;
-        return nullptr;
-    return std::addressof(*impl->global_object_list_container);
 }
 
 void KernelCore::PrepareReschedule(std::size_t id) {
@@ -1008,10 +1001,16 @@ void KernelCore::UnregisterInUseObject(KAutoObject* object) {
 
 void KernelCore::RunServer(std::unique_ptr<Service::ServerManager>&& server_manager) {
     auto* manager = server_manager.get();
-    if (!impl->is_shutting_down) {
+
+    {
         std::scoped_lock lk{impl->server_lock};
+        if (impl->is_shutting_down) {
+            return;
+        }
+
         impl->server_managers.emplace_back(std::move(server_manager));
     }
+
     manager->LoopProcess();
 }
 
@@ -1256,10 +1255,6 @@ void KernelCore::SuspendEmulation(bool suspended) {
 }
 
 void KernelCore::ShutdownCores() {
-    // Notify shutdown (pre-emptively)
-    for (auto& sm : impl->server_managers) {
-        sm->NotifyShutdown();
-    }
     impl->TerminateAllProcesses();
     KScopedSchedulerLock lk{*this};
     for (auto* thread : impl->shutdown_threads)

src/core/hle/kernel/kernel.h

@@ -175,8 +175,9 @@ public:
     /// Stops execution of 'id' core, in order to reschedule a new thread.
     void PrepareReschedule(std::size_t id);
 
-    KAutoObjectWithListContainer* ObjectListContainer();
+    KAutoObjectWithListContainer& ObjectListContainer();
-    const KAutoObjectWithListContainer* ObjectListContainer() const;
+
+    const KAutoObjectWithListContainer& ObjectListContainer() const;
 
     /// Registers all kernel objects with the global emulation state, this is purely for tracking
     /// leaks after emulation has been shutdown.

src/core/hle/kernel/slab_helpers.h

@@ -151,8 +151,7 @@ public:
         const bool is_initialized = this->IsInitialized();
         uintptr_t arg = 0;
         if (is_initialized) {
-            if (auto const olc = kernel.ObjectListContainer(); olc)
+            kernel.ObjectListContainer().Unregister(this);
-                olc->Unregister(this);
             arg = this->GetPostDestroyArgument();
             this->Finalize(kernel);
         }
@@ -176,7 +175,7 @@ public:
 public:
     static void InitializeSlabHeap(KernelCore& kernel, void* memory, size_t memory_size) {
         kernel.SlabHeap<Derived>().Initialize(memory, memory_size);
-        kernel.ObjectListContainer()->Initialize();
+        kernel.ObjectListContainer().Initialize();
     }
 
     static Derived* Create(KernelCore& kernel) {
@@ -187,7 +186,7 @@ public:
     }
 
     static void Register(KernelCore& kernel, Derived* obj) {
-        return kernel.ObjectListContainer()->Register(obj);
+        return kernel.ObjectListContainer().Register(obj);
     }
 
     static size_t GetObjectSize(KernelCore& kernel) {

src/core/hle/service/am/am.cpp

@@ -1,6 +1,3 @@
-// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
-// SPDX-License-Identifier: GPL-3.0-or-later
-
 // SPDX-FileCopyrightText: Copyright 2018 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
@@ -18,9 +15,13 @@ void LoopProcess(Core::System& system) {
     WindowSystem window_system(system);
     ButtonPoller button_poller(system, window_system);
     EventObserver event_observer(system, window_system);
+
     auto server_manager = std::make_unique<ServerManager>(system);
-    server_manager->RegisterNamedService("appletAE", std::make_shared<IAllSystemAppletProxiesService>(system, window_system));
+
-    server_manager->RegisterNamedService("appletOE", std::make_shared<IApplicationProxyService>(system, window_system));
+    server_manager->RegisterNamedService(
+        "appletAE", std::make_shared<IAllSystemAppletProxiesService>(system, window_system));
+    server_manager->RegisterNamedService(
+        "appletOE", std::make_shared<IApplicationProxyService>(system, window_system));
     ServerManager::RunServer(std::move(server_manager));
 }
 

src/core/hle/service/am/event_observer.cpp

@@ -26,25 +26,24 @@ EventObserver::EventObserver(Core::System& system, WindowSystem& window_system)
     m_window_system.SetEventObserver(this);
     m_wakeup_holder.SetUserData(static_cast<uintptr_t>(UserDataTag::WakeupEvent));
     m_wakeup_holder.LinkToMultiWait(std::addressof(m_multi_wait));
-    system.Kernel().RunOnGuestCoreProcess("am:EventObserver", [this]() {
+    m_thread = std::jthread([this](std::stop_token stop_token) {
-        auto const stop_token = m_stop_source.get_token();
+        Common::SetCurrentThreadName("am:EventObserver");
         while (!stop_token.stop_requested()) {
             auto* signaled_holder = this->WaitSignaled(stop_token);
-            if (stop_token.stop_requested() || !signaled_holder)
+            if (!signaled_holder)
                 break;
             this->Process(signaled_holder);
         }
-        m_process_cv.notify_one();
     });
 }
 
 EventObserver::~EventObserver() {
     // Signal thread and wait for processing to finish.
-    m_stop_source.request_stop();
+    if (m_thread.joinable()) {
+        // Signal thread and wait for processing to finish.
+        m_thread.request_stop();
         m_wakeup_event.Signal(m_system.Kernel());
-    {
+        m_thread.join();
-        std::unique_lock lk{m_process_mutex};
-        m_process_cv.wait(lk);
     }
 
     // Free remaining owned sessions.

src/core/hle/service/am/event_observer.h

@@ -6,8 +6,6 @@
 
 #pragma once
 
-#include <condition_variable>
-#include <stop_token>
 #include "common/polyfill_thread.h"
 #include "common/thread.h"
 #include "core/hle/service/kernel_helpers.h"
@@ -64,9 +62,7 @@ private:
     MultiWait m_deferred_wait_list;
 
     // Processing thread.
-    std::stop_source m_stop_source{};
+    std::jthread m_thread{};
-    std::mutex m_process_mutex;
-    std::condition_variable m_process_cv{};
 };
 
 } // namespace Service::AM

src/core/hle/service/glue/time/manager.cpp

@@ -1,4 +1,4 @@
-// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
+// SPDX-FileCopyrightText: Copyright 2025 Eden Emulator Project
 // SPDX-License-Identifier: GPL-3.0-or-later
 
 // SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
@@ -90,9 +90,10 @@ static Service::PSC::Time::LocationName GetTimeZoneString(
 }
 
 TimeManager::TimeManager(Core::System& system)
-    : m_steady_clock_resource{system}, m_time_zone_binary{system}
+    : m_steady_clock_resource{system}, m_time_zone_binary{system}, m_worker{
-    , m_worker{system, m_steady_clock_resource, m_file_timestamp_worker}
+                                                                       system,
-{
+                                                                       m_steady_clock_resource,
+                                                                       m_file_timestamp_worker} {
     m_time_m =
         system.ServiceManager().GetService<Service::PSC::Time::ServiceManager>("time:m", true);
 

src/core/hle/service/glue/time/worker.cpp

@@ -20,17 +20,15 @@
 
 namespace Service::Glue::Time {
 
-TimeWorker::TimeWorker(Core::System& system, StandardSteadyClockResource& steady_clock_resource, FileTimestampWorker& file_timestamp_worker)
+TimeWorker::TimeWorker(Core::System& system, StandardSteadyClockResource& steady_clock_resource,
-    : m_system{system}
+                       FileTimestampWorker& file_timestamp_worker)
-    , m_ctx{m_system, "Glue:TimeWorker"}
+    : m_system{system}, m_ctx{m_system, "Glue:TimeWorker"}, m_event{m_ctx.CreateEvent(
-    , m_event{m_ctx.CreateEvent("Glue:TimeWorker:Event")}
+                                                                "Glue:TimeWorker:Event")},
-    , m_steady_clock_resource{steady_clock_resource}
+      m_steady_clock_resource{steady_clock_resource},
-    , m_file_timestamp_worker{file_timestamp_worker}
+      m_file_timestamp_worker{file_timestamp_worker}, m_timer_steady_clock{m_ctx.CreateEvent(
-    , m_timer_steady_clock{m_ctx.CreateEvent("Glue:TimeWorker:SteadyClockTimerEvent")}
+                                                          "Glue:TimeWorker:SteadyClockTimerEvent")},
-    , m_timer_file_system{m_ctx.CreateEvent("Glue:TimeWorker:FileTimeTimerEvent")}
+      m_timer_file_system{m_ctx.CreateEvent("Glue:TimeWorker:FileTimeTimerEvent")},
-    , m_alarm_worker{m_system, m_steady_clock_resource}
+      m_alarm_worker{m_system, m_steady_clock_resource}, m_pm_state_change_handler{m_alarm_worker} {
-    , m_pm_state_change_handler{m_alarm_worker}
-{
     m_timer_steady_clock_timing_event = Core::Timing::CreateEvent(
         "Time::SteadyClockEvent",
         [this](s64 time,
@@ -49,16 +47,14 @@ TimeWorker::TimeWorker(Core::System& system, StandardSteadyClockResource& steady
 }
 
 TimeWorker::~TimeWorker() {
-    // Wait for processing to stop
-    m_stop_source.request_stop();
     m_local_clock_event->Signal(m_system.Kernel());
     m_network_clock_event->Signal(m_system.Kernel());
     m_ephemeral_clock_event->Signal(m_system.Kernel());
+    std::this_thread::sleep_for(std::chrono::milliseconds(16));
+
+    m_thread.request_stop();
     m_event->Signal(m_system.Kernel());
-    {
+    m_thread.join();
-        std::unique_lock lk{m_process_mutex};
-        m_process_cv.wait(lk);
-    }
 
     m_ctx.CloseEvent(m_event);
     m_system.CoreTiming().UnscheduleEvent(m_timer_steady_clock_timing_event);
@@ -126,8 +122,13 @@ void TimeWorker::Initialize(std::shared_ptr<Service::PSC::Time::StaticService> t
 }
 
 void TimeWorker::StartThread() {
-    m_system.Kernel().RunOnGuestCoreProcess("TimeWorker", [this]() {
+    m_thread = std::jthread(std::bind_front(&TimeWorker::ThreadFunc, this));
-        auto const stop_token = m_stop_source.get_token();
+}
+
+void TimeWorker::ThreadFunc(std::stop_token stop_token) {
+    Common::SetCurrentThreadName("TimeWorker");
+    Common::SetCurrentThreadPriority(Common::ThreadPriority::Low);
+
     while (!stop_token.stop_requested()) {
         enum class EventType : s32 {
             Exit = 0,
@@ -140,18 +141,18 @@ void TimeWorker::StartThread() {
             UpdateFileTimestamp = 7,
             AutoCorrect = 8,
         };
+
         s32 index{};
+
         if (m_pm_state_change_handler.m_priority != 0) {
             // TODO: gIPmModuleService::GetEvent() 1
-                index = WaitAny(
+            index = WaitAny(m_system.Kernel(),
-                    m_system.Kernel(),
                             &m_event->GetReadableEvent(), // 0
                             &m_alarm_worker.GetEvent()    // 1
             );
         } else {
             // TODO: gIPmModuleService::GetEvent() 1
-                index = WaitAny(
+            index = WaitAny(m_system.Kernel(),
-                    m_system.Kernel(),
                             &m_event->GetReadableEvent(),                       // 0
                             &m_alarm_worker.GetEvent(),                         // 1
                             &m_alarm_worker.GetTimerEvent().GetReadableEvent(), // 2
@@ -164,10 +165,7 @@ void TimeWorker::StartThread() {
             );
         }
 
-            if (stop_token.stop_requested())
+        switch (static_cast<EventType>(index)) {
-                break;
-
-            switch (EventType(index)) {
         case EventType::Exit:
             return;
 
@@ -260,6 +258,7 @@ void TimeWorker::StartThread() {
 
         case EventType::UpdateFileTimestamp:
             m_timer_file_system->Clear(m_system.Kernel());
+
             m_file_timestamp_worker.SetFilesystemPosixTime();
             break;
 
@@ -267,21 +266,22 @@ void TimeWorker::StartThread() {
             m_standard_user_auto_correct_clock_event->Clear(m_system.Kernel());
 
             bool automatic_correction{};
-                R_ASSERT(m_time_sm->IsStandardUserSystemClockAutomaticCorrectionEnabled(&automatic_correction));
+            R_ASSERT(m_time_sm->IsStandardUserSystemClockAutomaticCorrectionEnabled(
+                &automatic_correction));
 
             Service::PSC::Time::SteadyClockTimePoint time_point{};
-                R_ASSERT(m_time_sm->GetStandardUserSystemClockAutomaticCorrectionUpdatedTime(&time_point));
+            R_ASSERT(
+                m_time_sm->GetStandardUserSystemClockAutomaticCorrectionUpdatedTime(&time_point));
 
             m_set_sys->SetUserSystemClockAutomaticCorrectionEnabled(automatic_correction);
             m_set_sys->SetUserSystemClockAutomaticCorrectionUpdatedTime(time_point);
             break;
         }
+
         default:
             UNREACHABLE();
         }
     }
-        m_process_cv.notify_one();
-    });
 }
 
 } // namespace Service::Glue::Time

src/core/hle/service/glue/time/worker.h

@@ -1,12 +1,8 @@
-// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
-// SPDX-License-Identifier: GPL-3.0-or-later
-
 // SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
 #pragma once
 
-#include <condition_variable>
 #include "common/common_types.h"
 #include "core/hle/kernel/k_event.h"
 #include "core/hle/service/glue/time/alarm_worker.h"
@@ -37,17 +33,17 @@ public:
 
     void StartThread();
 
+private:
     template <typename T>
     T GetSettingsItemValue(const std::string& category, const std::string& name);
 
+    void ThreadFunc(std::stop_token stop_token);
+
     Core::System& m_system;
     KernelHelpers::ServiceContext m_ctx;
     std::shared_ptr<Service::Set::ISystemSettingsServer> m_set_sys;
 
-    std::mutex m_process_mutex;
+    std::jthread m_thread;
-    std::condition_variable m_process_cv;
-    std::stop_source m_stop_source;
-
     Kernel::KEvent* m_event{};
     std::shared_ptr<Service::PSC::Time::ServiceManager> m_time_m;
     std::shared_ptr<Service::PSC::Time::StaticService> m_time_sm;

src/core/hle/service/server_manager.cpp

@@ -98,10 +98,8 @@ ServerManager::~ServerManager() {
     // Signal stop.
     m_stop_source.request_stop();
     m_wakeup_event->Signal(m_system.Kernel());
-    if (m_deferral_event) m_deferral_event->Signal(m_system.Kernel()); //for sm: and ports
 
-
+    // Wait for processing to stop.
-    // Wait for processing to stop. Schedule
     m_stopped.Wait();
     m_threads.clear();
 
@@ -255,14 +253,6 @@ void ServerManager::StartAdditionalHostThreads(const char* name, size_t num_thre
     }
 }
 
-/// @brief Notifies that the system is shutting down (pre-emptively terminate threads)
-void ServerManager::NotifyShutdown() {
-    m_stop_source.request_stop();
-    // Wake them up regardless
-    m_wakeup_event->Signal(m_system.Kernel());
-    if (m_deferral_event) m_deferral_event->Signal(m_system.Kernel()); //for sm: and ports
-}
-
 Result ServerManager::LoopProcess() {
     SCOPE_EXIT {
         m_stopped.Set();
@@ -295,8 +285,9 @@ MultiWaitHolder* ServerManager::WaitSignaled() {
         this->LinkDeferred();
 
         // If we're done, return before we start waiting.
-        if (m_stop_source.stop_requested())
+        if (m_stop_source.stop_requested()) {
             return nullptr;
+        }
 
         auto* selected = m_multi_wait.WaitAny(m_system.Kernel());
         if (selected == std::addressof(*m_wakeup_holder)) {

src/core/hle/service/server_manager.h

@@ -1,6 +1,3 @@
-// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
-// SPDX-License-Identifier: GPL-3.0-or-later
-
 // SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
@@ -52,7 +49,6 @@ public:
 
     Result LoopProcess();
     void StartAdditionalHostThreads(const char* name, size_t num_threads);
-    void NotifyShutdown();
 
     static void RunServer(std::unique_ptr<ServerManager>&& server);
 

src/core/hle/service/services.cpp

@@ -117,6 +117,7 @@ Services::Services(std::shared_ptr<SM::ServiceManager>& sm, Core::System& system
         {"friends",    &Friend::LoopProcess},
         {"settings",   &Set::LoopProcess},
         {"psc",        &PSC::LoopProcess},
+        {"glue",       &Glue::LoopProcess},
         {"grc",        &GRC::LoopProcess},
         {"hid",        &HID::LoopProcess},
         {"jit",        &JIT::LoopProcess},
@@ -152,7 +153,6 @@ Services::Services(std::shared_ptr<SM::ServiceManager>& sm, Core::System& system
         {"usb",        &USB::LoopProcess},
         {"i2c",        &I2C::LoopProcess},
         {"gpio",        &GPIO::LoopProcess},
-        {"glue",       &Glue::LoopProcess},
     })
         kernel.RunOnGuestCoreProcess(std::string(e.first), [&system, f = e.second] { f(system); });
 }

src/core/hle/service/sm/sm.cpp

@@ -305,14 +305,16 @@ SM::SM(ServiceManager& service_manager_, Core::System& system_)
 SM::~SM() = default;
 
 void LoopProcess(Core::System& system) {
+    auto& service_manager = system.ServiceManager();
     auto server_manager = std::make_unique<ServerManager>(system);
+
     Kernel::KEvent* deferral_event{};
     server_manager->ManageDeferral(&deferral_event);
-    system.ServiceManager().SetDeferralEvent(deferral_event);
+    service_manager.SetDeferralEvent(deferral_event);
+
     auto sm_service = std::make_shared<SM>(system.ServiceManager(), system);
-    server_manager->ManageNamedPort("sm:", [sm_service] {
+    server_manager->ManageNamedPort("sm:", [sm_service] { return sm_service; });
-        return sm_service;
+
-    });
     ServerManager::RunServer(std::move(server_manager));
 }
 

src/core/hle/service/vi/conductor.cpp

@@ -11,7 +11,6 @@
 #include "common/thread.h"
 #include "core/core.h"
 #include "core/core_timing.h"
-#include "core/hle/kernel/kernel.h"
 #include "core/hle/service/vi/conductor.h"
 #include "core/hle/service/vi/container.h"
 #include "core/hle/service/vi/display_list.h"
@@ -30,26 +29,21 @@ Conductor::Conductor(Core::System& system, Container& container, DisplayList& di
     });
 
     if (system.IsMulticore()) {
-        m_event = Core::Timing::CreateEvent("ScreenComposition", [this](s64 time, std::chrono::nanoseconds ns_late) -> std::optional<std::chrono::nanoseconds> {
+        m_event = Core::Timing::CreateEvent(
+            "ScreenComposition",
+            [this](s64 time,
+                   std::chrono::nanoseconds ns_late) -> std::optional<std::chrono::nanoseconds> {
                 m_signal.Set();
                 return std::chrono::nanoseconds(this->GetNextTicks());
             });
 
         system.CoreTiming().ScheduleLoopingEvent(FrameNs, FrameNs, m_event);
-        m_thread = system.Kernel().RunOnHostCoreThread("VSyncThread", [this]() {
+        m_thread = std::jthread([this](std::stop_token token) { this->VsyncThread(token); });
-            auto const stop_token = m_thread.get_stop_token();
-            Common::SetCurrentThreadPriority(Common::ThreadPriority::VeryHigh);
-            Common::SetCurrentThreadToPerformanceCores();
-            while (!stop_token.stop_requested()) {
-                m_signal.Wait();
-                if (stop_token.stop_requested() || m_system.IsShuttingDown()) {
-                    break;
-                }
-                this->ProcessVsync();
-            }
-        });
     } else {
-        m_event = Core::Timing::CreateEvent("ScreenComposition", [this](s64 time, std::chrono::nanoseconds ns_late) -> std::optional<std::chrono::nanoseconds> {
+        m_event = Core::Timing::CreateEvent(
+            "ScreenComposition",
+            [this](s64 time,
+                   std::chrono::nanoseconds ns_late) -> std::optional<std::chrono::nanoseconds> {
                 this->ProcessVsync();
                 return std::chrono::nanoseconds(this->GetNextTicks());
             });
@@ -60,10 +54,10 @@ Conductor::Conductor(Core::System& system, Container& container, DisplayList& di
 
 Conductor::~Conductor() {
     m_system.CoreTiming().UnscheduleEvent(m_event);
-    if (m_thread.joinable()) {
+
+    if (m_system.IsMulticore()) {
         m_thread.request_stop();
         m_signal.Set();
-        m_thread.join();
     }
 }
 
@@ -89,6 +83,22 @@ void Conductor::ProcessVsync() {
     }
 }
 
+void Conductor::VsyncThread(std::stop_token token) {
+    Common::SetCurrentThreadName("VSyncThread");
+    Common::SetCurrentThreadPriority(Common::ThreadPriority::VeryHigh);
+    Common::SetCurrentThreadToPerformanceCores();
+
+    while (!token.stop_requested()) {
+        m_signal.Wait();
+
+        if (m_system.IsShuttingDown()) {
+            return;
+        }
+
+        this->ProcessVsync();
+    }
+}
+
 s64 Conductor::GetNextTicks() const {
     const auto& settings = Settings::values;
     auto speed_scale = 1.f;

src/core/loader/nso.cpp

@@ -7,14 +7,12 @@
 #include <algorithm>
 #include <cinttypes>
 #include <cstring>
-#include <span>
 #include <vector>
 
 #include "common/common_funcs.h"
 #include "common/hex_util.h"
 #include "common/logging.h"
 #include "common/lz4_compression.h"
-#include "common/zbic_compression.h"
 #include "common/settings.h"
 #include "common/swap.h"
 #include "core/core.h"
@@ -106,36 +104,11 @@ std::optional<VAddr> AppLoader_NSO::LoadModule(Kernel::KProcess& process, Core::
         for (std::size_t i = 0; i < nso_header.segments.size(); ++i) {
             nso_file.Read(compressed_data.data(), nso_header.segments_compressed_size[i], nso_header.segments[i].offset);
             if (nso_header.IsSegmentCompressed(i)) {
-                if (nso_header.IsZBICCompressed()) {
+                int r = Common::Compression::DecompressDataLZ4(decompressed_size.data(), nso_header.segments[i].size, compressed_data.data(), nso_header.segments_compressed_size[i]);
-                    // ZBIC compression
-                    const int r = Common::Compression::DecompressDataZBIC(
-                        std::span<u8>{decompressed_size}.first(nso_header.segments[i].size),
-                        std::span<const u8>{compressed_data}.first(nso_header.segments_compressed_size[i])
-                    );
-                    ASSERT(r > 0);
-                } else {
-                    // LZ4 compression
-                    int r = Common::Compression::DecompressDataLZ4(
-                        decompressed_size.data(),
-                        nso_header.segments[i].size,
-                        compressed_data.data(),
-                        nso_header.segments_compressed_size[i]
-                    );
                 ASSERT(r == int(nso_header.segments[i].size));
-                }
+                std::memcpy(codeset.memory.data() + module_start + nso_header.segments[i].location, decompressed_size.data(), nso_header.segments[i].size);
-
-                std::memcpy(
-                    codeset.memory.data() + module_start + nso_header.segments[i].location,
-                    decompressed_size.data(),
-                    nso_header.segments[i].size
-                );
             } else {
-                // Not compressed
+                std::memcpy(codeset.memory.data() + module_start + nso_header.segments[i].location, compressed_data.data(), nso_header.segments[i].size);
-                std::memcpy(
-                    codeset.memory.data() + module_start + nso_header.segments[i].location,
-                    compressed_data.data(),
-                    nso_header.segments[i].size
-                );
             }
             codeset.segments[i].addr = module_start + nso_header.segments[i].location;
             codeset.segments[i].offset = module_start + nso_header.segments[i].location;

src/core/loader/nso.h

@@ -1,6 +1,3 @@
-// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
-// SPDX-License-Identifier: GPL-3.0-or-later
-
 // SPDX-FileCopyrightText: Copyright 2018 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
@@ -61,9 +58,6 @@ struct NSOHeader {
     std::array<SHA256Hash, 3> segment_hashes;
 
     bool IsSegmentCompressed(size_t segment_num) const;
-    bool IsZBICCompressed() const {
-        return ((flags >> 7) & 1) != 0;
-    }
 };
 static_assert(sizeof(NSOHeader) == 0x100, "NSOHeader has incorrect size.");
 static_assert(std::is_trivially_copyable_v<NSOHeader>, "NSOHeader must be trivially copyable.");

src/core/memory/dmnt_cheat_vm.cpp

@@ -1,15 +1,6 @@
-// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
-// SPDX-License-Identifier: GPL-3.0-or-later
-
 // SPDX-FileCopyrightText: Copyright 2019 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
-// gcc14 bug
-#ifdef __GNUC__
-#pragma GCC diagnostic push
-#pragma GCC diagnostic ignored "-Wmaybe-uninitialized"
-#endif
-
 #include "common/assert.h"
 #include "common/scope_exit.h"
 #include "core/memory/dmnt_cheat_types.h"
@@ -1275,8 +1266,3 @@ void DmntCheatVm::Execute(const CheatProcessMetadata& metadata) {
 }
 
 } // namespace Core::Memory
-
-
-#ifdef __GNUC__
-#pragma GCC diagnostic pop
-#endif

src/core/memory/dmnt_cheat_vm.h

@@ -1,4 +1,4 @@
-// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
+// SPDX-FileCopyrightText: Copyright 2025 Eden Emulator Project
 // SPDX-License-Identifier: GPL-3.0-or-later
 
 // SPDX-FileCopyrightText: Copyright 2019 yuzu Emulator Project
@@ -254,16 +254,14 @@ struct UnrecognizedInstruction {
 
 struct CheatVmOpcode {
     bool begin_conditional_block{};
-    std::variant<
+    std::variant<StoreStaticOpcode, BeginConditionalOpcode, EndConditionalOpcode, ControlLoopOpcode,
-        std::monostate,
-        StoreStaticOpcode, BeginConditionalOpcode, EndConditionalOpcode, ControlLoopOpcode,
                  LoadRegisterStaticOpcode, LoadRegisterMemoryOpcode, StoreStaticToAddressOpcode,
                  PerformArithmeticStaticOpcode, BeginKeypressConditionalOpcode,
                  PerformArithmeticRegisterOpcode, StoreRegisterToAddressOpcode,
                  BeginRegisterConditionalOpcode, SaveRestoreRegisterOpcode,
                  SaveRestoreRegisterMaskOpcode, ReadWriteStaticRegisterOpcode, PauseProcessOpcode,
-        ResumeProcessOpcode, DebugLogOpcode, UnrecognizedInstruction
+                 ResumeProcessOpcode, DebugLogOpcode, UnrecognizedInstruction>
-    > opcode{};
+        opcode{};
 };
 
 class DmntCheatVm {

src/dynarmic/src/dynarmic/backend/x64/emit_x64_memory.h

@@ -170,19 +170,13 @@ template<>
 
     // check for marked bit, use as unmapped if marked
     if (ctx.conf.page_table_marked_bit) {
-        auto const marked_bit = *ctx.conf.page_table_marked_bit;
+        code.bt(page, *ctx.conf.page_table_marked_bit);
-        if (s64(s32(1 << marked_bit)) == s64(1 << marked_bit)) {
-            code.test(page, s32(1 << marked_bit));
-            code.jnz(abort, code.T_NEAR);
-        } else {
-            code.bt(page, marked_bit);
         code.jc(abort, code.T_NEAR);
     }
-    }
     // mask away attributes
     if (ctx.conf.page_table_pointer_mask == 0) {
         code.test(page, page);
-    } else if (s64(s32(ctx.conf.page_table_pointer_mask)) == s64(ctx.conf.page_table_pointer_mask)) {
+    } else if (std::in_range<s32>(ctx.conf.page_table_pointer_mask)) {
         code.and_(page, ctx.conf.page_table_pointer_mask);
     } else {
         code.mov(tmp, ctx.conf.page_table_pointer_mask);

src/hid_core/frontend/emulated_controller.cpp

@@ -761,6 +761,9 @@ void EmulatedController::StartMotionCalibration() {
 }
 
 void EmulatedController::SetButton(const Common::Input::CallbackStatus& callback, std::size_t index, Common::UUID uuid) {
+    const auto player_index = Service::HID::NpadIdTypeToIndex(npad_id_type);
+    const auto& player = Settings::values.players.GetValue()[player_index];
+
     if (index >= controller.button_values.size()) {
         return;
     }
@@ -913,10 +916,21 @@ void EmulatedController::SetButton(const Common::Input::CallbackStatus& callback
         break;
     }
 
-    const auto player_index = Service::HID::NpadIdTypeToIndex(npad_id_type);
+    if (!is_connected) {
-    const auto& player = Settings::values.players.GetValue()[player_index];
+        if (npad_type == NpadStyleIndex::Handheld) {
-    if (player.connected) {
+            if (npad_id_type == NpadIdType::Handheld) {
                 Connect();
+                controller_connected[player_index] = true;
+            }
+        } else if (npad_type != NpadStyleIndex::Handheld) {
+            if (npad_id_type == NpadIdType::Player1) {
+                Connect();
+                controller_connected[player_index] = true;
+            } else if (player.connected && !controller_connected[player_index]) {
+                Connect();
+                controller_connected[player_index] = true;
+            }
+        }
     }
 
     TriggerOnChange(ControllerTriggerType::Button, true);

src/hid_core/frontend/emulated_controller.h

@@ -22,6 +22,7 @@
 #include "common/settings.h"
 #include "common/vector_math.h"
 #include "hid_core/frontend/motion_input.h"
+#include "hid_core/hid_core.h"
 #include "hid_core/hid_types.h"
 #include "hid_core/irsensor/irs_types.h"
 
@@ -584,6 +585,7 @@ private:
     std::array<VibrationValue, 2> last_vibration_value{DEFAULT_VIBRATION_VALUE,
                                                        DEFAULT_VIBRATION_VALUE};
     std::array<std::chrono::steady_clock::time_point, 2> last_vibration_timepoint{};
+    std::array<bool, HIDCore::available_controllers> controller_connected{};
 
     // Atomically synched values
     std::atomic<HID::NpadStyleIndex> npad_type{HID::NpadStyleIndex::None};

src/hid_core/resources/abstracted_pad/abstract_button_handler.cpp

@@ -79,30 +79,28 @@ void NpadAbstractButtonHandler::UpdateAllButtonLifo() {
     Core::HID::NpadIdType npad_id = properties_handler->GetNpadId();
     for (std::size_t i = 0; i < AruidIndexMax; i++) {
         auto* data = applet_resource_holder->applet_resource->GetAruidDataByIndex(i);
-        if (auto const shfmt = data->shared_memory_format; shfmt) {
+        auto& npad_entry = data->shared_memory_format->npad.npad_entry[NpadIdTypeToIndex(npad_id)];
-            auto& npad_entry = shfmt->npad.npad_entry[NpadIdTypeToIndex(npad_id)];
         UpdateButtonLifo(npad_entry, data->aruid);
     }
-    }
 }
 
 void NpadAbstractButtonHandler::UpdateCoreBatteryState() {
     Core::HID::NpadIdType npad_id = properties_handler->GetNpadId();
     for (std::size_t i = 0; i < AruidIndexMax; i++) {
         auto* data = applet_resource_holder->applet_resource->GetAruidDataByIndex(i);
-        if (auto const shfmt = data->shared_memory_format; shfmt) {
+        auto& npad_entry = data->shared_memory_format->npad.npad_entry[NpadIdTypeToIndex(npad_id)];
-            auto& npad_entry = shfmt->npad.npad_entry[NpadIdTypeToIndex(npad_id)];
         UpdateButtonLifo(npad_entry, data->aruid);
     }
-    }
 }
 
 void NpadAbstractButtonHandler::UpdateButtonState(u64 aruid) {
     Core::HID::NpadIdType npad_id = properties_handler->GetNpadId();
-    if (auto data = applet_resource_holder->applet_resource->GetAruidData(aruid); data) {
+    auto* data = applet_resource_holder->applet_resource->GetAruidData(aruid);
+    if (data == nullptr) {
+        return;
+    }
     auto& npad_entry = data->shared_memory_format->npad.npad_entry[NpadIdTypeToIndex(npad_id)];
     UpdateButtonLifo(npad_entry, aruid);
-    }
 }
 
 Result NpadAbstractButtonHandler::SetHomeProtection(bool is_enabled, u64 aruid) {

src/qt_common/config/shared_translation.cpp

@@ -226,7 +226,9 @@ std::unique_ptr<TranslationMap> InitializeTranslations(QObject* parent) {
     INSERT(Settings, dma_accuracy, tr("DMA Accuracy:"),
            tr("Controls the DMA read mode.\nUnsafe is faster, while Safe is more stable and can fix issues in some games.\nDefault follows the GPU Accuracy setting."));
     INSERT(Settings, gpu_fence_behavior, tr("GPU Fence Behavior:"),
-           tr("Controls the GPU fence synchronization behavior.\nImmediate is the fastest option, but can introduce some issues.\nBalanced offers better compatibility and may fix issues in some games.\nAccurate further improves compatibility at the cost of some performance.\nDefault follows the GPU Mode setting."));
+           tr("Controls the GPU fence synchronization behavior.\nImmediate is the fastest option, but can introduce some issues.\nBalanced offers better compatibility and may fix issues in some games.\nAccurate further improves compatibility at the cost of some performance.\nStrict is the slowest option, but can fix issues that require stricter synchronization.\nDefault follows the GPU Accuracy setting."));
+    INSERT(Settings, enable_gpu_buffer_readback, tr("Enable GPU buffer readback"),
+           tr("Preserves GPU-modified data by reading it back before uploading.\nSome games require this to render certain effects properly."));
     INSERT(Settings, use_asynchronous_shaders, tr("Enable asynchronous shader compilation"),
            tr("May reduce shader stutter."));
     INSERT(Settings, gpu_clock, tr("GPU Clocks"),
@@ -440,6 +442,7 @@ std::unique_ptr<ComboboxTranslationMap> ComboboxEnumeration(QObject* parent) {
                               PAIR(GpuFenceBehavior, Immediate, tr("Immediate")),
                               PAIR(GpuFenceBehavior, Balanced, tr("Balanced")),
                               PAIR(GpuFenceBehavior, Accurate, tr("Accurate")),
+                              PAIR(GpuFenceBehavior, Strict, tr("Strict")),
                           }});
     translations->insert(
         {Settings::EnumMetadata<Settings::CpuAccuracy>::Index(),

src/qt_common/render/emu_thread.cpp

@@ -66,17 +66,14 @@ void EmuThread::run() {
         } else {
             QtCommon::system->Pause();
             m_stopped.Set();
+
             EmulationPaused(lk);
-            m_should_run_cv.wait(lk, stop_token, [&] {
+            m_should_run_cv.wait(lk, stop_token, [&] { return m_should_run; });
-                return m_should_run;
-            });
             EmulationResumed(lk);
         }
     }
 
     // Shutdown the main emulated process
-    // Needs to pause first, we may not have paused before shutting down.
-    QtCommon::system->Pause();
     QtCommon::system->DetachDebugger();
     QtCommon::system->ShutdownMainProcess();
 }

src/shader_recompiler/backend/spirv/emit_spirv.cpp

@@ -511,15 +511,16 @@ void PatchPhiNodes(IR::Program& program, EmitContext& ctx) {
             }
 
             // Start "before" first PHI; advance on phi_arg == 0
-    size_t phi_index = size_t(-1);
+            size_t phi_index = static_cast<size_t>(-1);
-    ctx.PatchDeferredPhi([&ctx, &phi_index, phi_insts = std::move(phi_instructions)](size_t phi_arg, Id parent) -> std::pair<Id, Id> {
+
+            ctx.PatchDeferredPhi([&](size_t phi_arg, Id parent) -> std::pair<Id, Id> {
                 if (phi_arg == 0) {
                     ++phi_index;
                 }
-        IR::Inst* phi = phi_insts[phi_index];
+                IR::Inst* phi = phi_instructions[phi_index];
                 return { ctx.Def(phi->Arg(phi_arg)), parent };
             });
-}
+        }
 } // Anonymous namespace
 
 std::vector<u32> EmitSPIRV(const Profile& profile, const RuntimeInfo& runtime_info, IR::Program& program, Bindings& bindings) {

src/video_core/buffer_cache/buffer_cache.h

@@ -249,10 +249,6 @@ bool BufferCache<P>::DMACopy(GPUVAddr src_address, GPUVAddr dest_address, u64 am
     runtime.CopyBuffer(dest_buffer, src_buffer, copies, true);
     if (has_new_downloads) {
         memory_tracker.MarkRegionAsGpuModified(*cpu_dest_address, amount);
-        const bool should_sync = Settings::IsGPUFenceBehaviorBalanced() || Settings::IsGPUFenceBehaviorAccurate();
-        if (should_sync) {
-            runtime.Finish();
-        }
     }
 
     Tegra::Memory::DeviceGuestMemoryScoped<u8, Tegra::Memory::GuestMemoryFlags::UnsafeReadWrite>
@@ -1234,7 +1230,7 @@ void BufferCache<P>::BindHostComputeStorageBuffers() {
         buffer.MarkUsage(offset, size);
 
         if (is_written) {
-            MarkWrittenBuffer(binding.buffer_id, binding.device_addr, size, true);
+            MarkWrittenBuffer(binding.buffer_id, binding.device_addr, size);
         }
 
         if constexpr (NEEDS_BIND_STORAGE_INDEX) {
@@ -1520,13 +1516,11 @@ void BufferCache<P>::UpdateComputeTextureBuffers() {
 }
 
 template <class P>
-void BufferCache<P>::MarkWrittenBuffer(BufferId buffer_id, DAddr device_addr, u32 size, bool needs_sync) {
+void BufferCache<P>::MarkWrittenBuffer(BufferId buffer_id, DAddr device_addr, u32 size) {
     if constexpr (!IS_OPENGL) {
-        if (needs_sync) {
         Buffer& buffer = slot_buffers[buffer_id];
         buffer.setWriteTick(runtime.CurrentTick());
     }
-    }
     memory_tracker.MarkRegionAsGpuModified(device_addr, size);
     gpu_modified_ranges.Add(device_addr, size);
     uncommitted_gpu_modified_ranges.Add(device_addr, size);
@@ -1541,11 +1535,8 @@ BufferId BufferCache<P>::FindBuffer(DAddr device_addr, u32 size, bool sparse_com
     const BufferId buffer_id = page_table[page];
     if (buffer_id) {
         Buffer& buffer = slot_buffers[buffer_id];
+        WaitForGpuFenceIfNeeded(buffer);
         if (buffer.IsInBounds(device_addr, size)) {
-            const bool should_sync = Settings::IsGPUFenceBehaviorAccurate();
-            if (should_sync) {
-                SynchronizeBufferWrites(buffer);
-            }
             bool usable = true;
             if constexpr (requires { buffer.IsSparseCompatible(); }) {
                 if (sparse_compatible && !buffer.IsSparseCompatible()) {
@@ -1561,13 +1552,19 @@ BufferId BufferCache<P>::FindBuffer(DAddr device_addr, u32 size, bool sparse_com
 }
 
 template <class P>
-void BufferCache<P>::SynchronizeBufferWrites(Buffer& buffer) {
+void BufferCache<P>::WaitForGpuFenceIfNeeded(Buffer& buffer) {
     if constexpr (!IS_OPENGL) {
+        const bool gpu_fence_accurate = Settings::IsGPUFenceBehaviorAccurate();
+        const bool gpu_fence_strict = Settings::IsGPUFenceBehaviorStrict();
+        if (gpu_fence_accurate || gpu_fence_strict) {
+            const u64 gpu_tick_delay = gpu_fence_strict ? 0 : 3;
             const u64 buffer_tick = buffer.getWriteTick();
-        if (!runtime.IsFree(buffer_tick)) {
+            const u64 gpu_tick = runtime.KnownGpuTick();
+            if (buffer_tick > gpu_tick + gpu_tick_delay) {
                 runtime.Wait(buffer_tick);
             }
         }
+    }
 }
 
 template <class P>
@@ -1798,6 +1795,9 @@ void BufferCache<P>::ImmediateUploadMemory([[maybe_unused]] Buffer& buffer,
                 if (immediate_buffer.empty()) {
                     immediate_buffer = ImmediateBuffer(largest_copy);
                 }
+                if (Settings::values.enable_gpu_buffer_readback.GetValue()) {
+                    DownloadBufferMemory(buffer, device_addr, copy.size);
+                }
                 device_memory.ReadBlockUnsafe(device_addr, immediate_buffer.data(), copy.size);
                 upload_span = immediate_buffer.subspan(0, copy.size);
             }
@@ -1816,6 +1816,9 @@ void BufferCache<P>::MappedUploadMemory([[maybe_unused]] Buffer& buffer,
         for (BufferCopy& copy : copies) {
             u8* const src_pointer = staging_pointer.data() + copy.src_offset;
             const DAddr device_addr = buffer.CpuAddr() + copy.dst_offset;
+            if (Settings::values.enable_gpu_buffer_readback.GetValue()) {
+                DownloadBufferMemory(buffer, device_addr, copy.size);
+            }
             device_memory.ReadBlockUnsafe(device_addr, src_pointer, copy.size);
             // Apply the staging offset
             copy.src_offset += upload_staging.offset;

src/video_core/buffer_cache/buffer_cache_base.h

@@ -430,11 +430,11 @@ private:
 
     void UpdateComputeTextureBuffers();
 
-    void MarkWrittenBuffer(BufferId buffer_id, DAddr device_addr, u32 size, bool needs_sync = false);
+    void MarkWrittenBuffer(BufferId buffer_id, DAddr device_addr, u32 size);
 
     [[nodiscard]] BufferId FindBuffer(DAddr device_addr, u32 size, bool sparse_compatible);
 
-    void SynchronizeBufferWrites(Buffer& buffer);
+    void WaitForGpuFenceIfNeeded(Buffer& buffer);
 
     [[nodiscard]] OverlapResult ResolveOverlaps(DAddr device_addr, u32 wanted_size);
 

src/video_core/fence_manager.h

@@ -72,7 +72,7 @@ public:
     }
 
     void SignalFence(std::function<void()>&& func) {
-        const bool delay_fence = Settings::IsGPUFenceBehaviorDefault() ? Settings::IsGPULevelHigh() : Settings::IsGPUFenceBehaviorBalanced() || Settings::IsGPUFenceBehaviorAccurate();
+        const bool delay_fence = Settings::IsGPUFenceBehaviorDefault() ? Settings::IsGPULevelHigh() : Settings::IsGPUFenceBehaviorBalanced() || Settings::IsGPUFenceBehaviorAccurate() || Settings::IsGPUFenceBehaviorStrict();
         const bool should_flush = ShouldFlush();
         if constexpr (!can_async_check) {
             TryReleasePendingFences<false>();

src/video_core/query_cache/query_cache.h

@@ -260,7 +260,7 @@ void QueryCacheBase<Traits>::CounterReport(GPUVAddr addr, QueryType counter_type
     };
     u8* pointer = impl->device_memory.template GetPointer<u8>(cpu_addr);
     u8* pointer_timestamp = impl->device_memory.template GetPointer<u8>(cpu_addr + 8);
-    bool is_synced = (Settings::IsGPUFenceBehaviorDefault() ? !Settings::IsGPULevelHigh() : !Settings::IsGPUFenceBehaviorBalanced() && !Settings::IsGPUFenceBehaviorAccurate()) && is_fence;
+    bool is_synced = (Settings::IsGPUFenceBehaviorDefault() ? !Settings::IsGPULevelHigh() : !Settings::IsGPUFenceBehaviorBalanced() && !Settings::IsGPUFenceBehaviorAccurate() && !Settings::IsGPUFenceBehaviorStrict()) && is_fence;
     std::function<void()> operation([this, is_synced, streamer, query_base = query, query_location,
                                      pointer, pointer_timestamp] {
         if (True(query_base->flags & QueryFlagBits::IsInvalidated)) {

src/video_core/renderer_vulkan/vk_buffer_cache.cpp

@@ -200,7 +200,7 @@ public:
             if (host_visible) {
                 return StagingBufferRef{};
             }
-            return staging_pool.Request(size_bytes, MemoryUsage::Upload);
+            return staging_pool.Request(device, size_bytes, MemoryUsage::Upload);
         }();
 
         u8* staging_data = host_visible ? buffer.Mapped().data() : staging.mapped_span.data();
@@ -375,11 +375,11 @@ BufferCacheRuntime::BufferCacheRuntime(const Device& device_, MemoryAllocator& m
 }
 
 StagingBufferRef BufferCacheRuntime::UploadStagingBuffer(size_t size) {
-    return staging_pool.Request(size, MemoryUsage::Upload);
+    return staging_pool.Request(device, size, MemoryUsage::Upload);
 }
 
 StagingBufferRef BufferCacheRuntime::DownloadStagingBuffer(size_t size, bool deferred) {
-    return staging_pool.Request(size, MemoryUsage::Download, deferred);
+    return staging_pool.Request(device, size, MemoryUsage::Download, deferred);
 }
 
 VkFormat BufferCacheRuntime::TexelBufferFormat(VideoCore::Surface::PixelFormat format) const {
@@ -419,15 +419,15 @@ void BufferCacheRuntime::TickFrame(Common::SlotVector<Buffer>& slot_buffers) noe
 }
 
 u64 BufferCacheRuntime::CurrentTick() {
-    return scheduler.CurrentTick();
+    return scheduler.GetMasterSemaphore().CurrentTick();
 }
 
-bool BufferCacheRuntime::IsFree(u64 tick) {
+u64 BufferCacheRuntime::KnownGpuTick() {
-    return scheduler.IsFree(tick);
+    return scheduler.GetMasterSemaphore().KnownGpuTick();
 }
 
-void BufferCacheRuntime::Wait(u64 tick) {
+void BufferCacheRuntime::Wait(u64 buffer_tick) {
-    scheduler.Wait(tick);
+    scheduler.Wait(buffer_tick);
 }
 
 void BufferCacheRuntime::Finish() {

src/video_core/renderer_vulkan/vk_buffer_cache.h

@@ -112,9 +112,9 @@ public:
 
     u64 CurrentTick();
 
-    bool IsFree(u64 tick);
+    u64 KnownGpuTick();
 
-    void Wait(u64 tick);
+    void Wait(u64 buffer_tick);
 
     void Finish();
 
@@ -162,7 +162,7 @@ public:
     std::span<u8> BindMappedUniformBuffer([[maybe_unused]] size_t stage,
                                           [[maybe_unused]] u32 binding_index,
                                           u32 size) {
-        const StagingBufferRef ref = staging_pool.Request(size, MemoryUsage::Upload);
+        const StagingBufferRef ref = staging_pool.Request(device, size, MemoryUsage::Upload);
         guest_descriptor_queue.AddBuffer(ref.buffer, ref.device_address,
                                          static_cast<u32>(ref.offset), size);
         return ref.mapped_span;

src/video_core/renderer_vulkan/vk_compute_pass.cpp

@@ -287,7 +287,7 @@ Uint8Pass::~Uint8Pass() = default;
 std::pair<VkBuffer, VkDeviceSize> Uint8Pass::Assemble(u32 num_vertices, VkBuffer src_buffer,
                                                       u32 src_offset) {
     const u32 staging_size = static_cast<u32>(num_vertices * sizeof(u16));
-    const auto staging = staging_buffer_pool.Request(staging_size, MemoryUsage::DeviceLocal);
+    const auto staging = staging_buffer_pool.Request(device, staging_size, MemoryUsage::DeviceLocal);
 
     compute_pass_descriptor_queue.Acquire(scheduler, 2);
     compute_pass_descriptor_queue.AddBuffer(src_buffer, src_offset, num_vertices);
@@ -345,7 +345,7 @@ std::pair<VkBuffer, VkDeviceSize> QuadIndexedPass::Assemble(
     const u32 num_tri_vertices = (is_strip ? (num_vertices - 2) / 2 : num_vertices / 4) * 6;
 
     const std::size_t staging_size = num_tri_vertices * sizeof(u32);
-    const auto staging = staging_buffer_pool.Request(staging_size, MemoryUsage::DeviceLocal);
+    const auto staging = staging_buffer_pool.Request(device, staging_size, MemoryUsage::DeviceLocal);
 
     compute_pass_descriptor_queue.Acquire(scheduler, 2);
     compute_pass_descriptor_queue.AddBuffer(src_buffer, src_offset, input_size);

src/video_core/renderer_vulkan/vk_query_cache.cpp

@@ -852,7 +852,7 @@ public:
 
     void PushUnsyncedQueries() override {
         CloseCounter();
-        auto staging_ref = staging_pool.Request(
+        auto staging_ref = staging_pool.Request(device,
             pending_flush_queries.size() * TFBQueryBank::QUERY_SIZE, MemoryUsage::Download, true);
         size_t offset_base = staging_ref.offset;
         for (auto q : pending_flush_queries) {
@@ -1657,7 +1657,7 @@ void QueryCacheRuntime::SyncValues(std::span<SyncValuesType> values, VkBuffer ba
     impl->copies_setup.clear();
     impl->copies_setup.resize(impl->little_cache.size());
     if constexpr (SyncValuesType::GeneratesBaseBuffer) {
-        ref = impl->staging_pool.Request(total_size, MemoryUsage::Upload);
+        ref = impl->staging_pool.Request(impl->device, total_size, MemoryUsage::Upload);
         size_t current_offset = ref.offset;
         size_t accumulated_size = 0;
         for (size_t i = 0; i < values.size(); i++) {

src/video_core/renderer_vulkan/vk_staging_buffer_pool.cpp

@@ -28,55 +28,42 @@ using namespace Common::Literals;
 // Maximum potential alignment of a Vulkan buffer
 constexpr VkDeviceSize MAX_ALIGNMENT = 256;
 
-// Stream buffer size in bytes
+size_t GetStreamBufferSize(const Device& device, size_t max_stream_buffer_size, size_t max_alignment) {
-// *NIX drivers are more sensitive to increased buffers for streaming.
-// Windows ones however, can intake bigger buffers and generally do not OOM.
-// - GTX 960 on Windows will not OOM with 256mib
-// - GT 1030 on ^NIX will OOM with 256mib
-#if defined(__FreeBSD__)
-constexpr VkDeviceSize MAX_STREAM_BUFFER_SIZE = 128_MiB;
-#else
-constexpr VkDeviceSize MAX_STREAM_BUFFER_SIZE = 256_MiB;
-#endif
-
-size_t GetStreamBufferSize(const Device& device) {
     if (!device.HasDebuggingToolAttached()) {
-        return MAX_STREAM_BUFFER_SIZE;
+        return max_stream_buffer_size;
     }
 
     VkDeviceSize size{0};
     bool has_device_local_host_visible_heap{};
-    ForEachDeviceLocalHostVisibleHeap(device, [&size, &has_device_local_host_visible_heap](
+    ForEachDeviceLocalHostVisibleHeap(device, [&size, &has_device_local_host_visible_heap](size_t index, VkMemoryHeap& heap) {
-                                                  size_t index, VkMemoryHeap& heap) {
         has_device_local_host_visible_heap = true;
-        size = (std::max)(size, heap.size);
+        size = std::max<size_t>(size, heap.size);
     });
     if (has_device_local_host_visible_heap) {
         // If rebar is not supported, cut the max heap size to 40%. This will allow 2 captures to be
         // loaded at the same time in RenderDoc. If rebar is supported, this shouldn't be an issue
         // as the heap will be much larger.
-        if (size <= MAX_STREAM_BUFFER_SIZE) {
+        if (size <= max_stream_buffer_size) {
             size = size * 40 / 100;
         }
     } else {
-        size = MAX_STREAM_BUFFER_SIZE;
+        size = max_stream_buffer_size;
     }
-    return (std::min)(Common::AlignUp(size, MAX_ALIGNMENT), MAX_STREAM_BUFFER_SIZE);
+    return std::min<size_t>(Common::AlignUp(size, max_alignment), max_stream_buffer_size);
 }
 } // Anonymous namespace
 
-StagingBufferPool::StagingBufferPool(const Device& device_, MemoryAllocator& memory_allocator_,
+StagingBufferPool::StagingBufferPool(const Device& device, MemoryAllocator& memory_allocator_, Scheduler& scheduler_)
-                                     Scheduler& scheduler_)
+    : memory_allocator{memory_allocator_}, scheduler{scheduler_}
-    : device{device_}, memory_allocator{memory_allocator_}, scheduler{scheduler_},
+    , stream_buffer_size{GetStreamBufferSize(device, 256_MiB, MAX_ALIGNMENT)}
-      stream_buffer_size{GetStreamBufferSize(device)}, region_size{stream_buffer_size /
+{
-                                                                   StagingBufferPool::NUM_SYNCS} {
     VkBufferCreateInfo stream_ci = {
         .sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO,
         .pNext = nullptr,
         .flags = 0,
         .size = stream_buffer_size,
-        .usage = VK_BUFFER_USAGE_TRANSFER_SRC_BIT | VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT |
+        .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,
+            | VK_BUFFER_USAGE_INDEX_BUFFER_BIT | VK_BUFFER_USAGE_STORAGE_BUFFER_BIT,
         .sharingMode = VK_SHARING_MODE_EXCLUSIVE,
         .queueFamilyIndexCount = 0,
         .pQueueFamilyIndices = nullptr,
@@ -87,7 +74,16 @@ StagingBufferPool::StagingBufferPool(const Device& device_, MemoryAllocator& mem
     if (device.IsBufferDeviceAddressSupported()) {
         stream_ci.usage |= VK_BUFFER_USAGE_SHADER_DEVICE_ADDRESS_BIT;
     }
+    // Some drivers are more sensitive to increased buffer sizes
+    try {
         stream_buffer = memory_allocator.CreateBuffer(stream_ci, MemoryUsage::Stream);
+    } catch (vk::Exception& e) {
+        LOG_ERROR(Render_Vulkan, "Can't fit {} bytes buffer, halving", stream_ci.size);
+        stream_buffer_size = GetStreamBufferSize(device, 128_MiB, MAX_ALIGNMENT);
+        stream_ci.size = stream_buffer_size;
+        stream_buffer = memory_allocator.CreateBuffer(stream_ci, MemoryUsage::Stream);
+    }
+    region_size = stream_buffer_size / StagingBufferPool::NUM_SYNCS;
     if (device.HasDebuggingToolAttached()) {
         stream_buffer.SetObjectNameEXT("Stream Buffer");
     }
@@ -100,11 +96,10 @@ StagingBufferPool::StagingBufferPool(const Device& device_, MemoryAllocator& mem
 
 StagingBufferPool::~StagingBufferPool() = default;
 
-StagingBufferRef StagingBufferPool::Request(size_t size, MemoryUsage usage, bool deferred) {
+StagingBufferRef StagingBufferPool::Request(const Device& device, size_t size, MemoryUsage usage, bool deferred) {
-    if (!deferred && usage == MemoryUsage::Upload && size <= region_size) {
+    return (!deferred && usage == MemoryUsage::Upload && size <= region_size)
-        return GetStreamBuffer(size);
+        ? GetStreamBuffer(device, size)
-    }
+        : GetStagingBuffer(device, size, usage, deferred);
-    return GetStagingBuffer(size, usage, deferred);
 }
 
 void StagingBufferPool::FreeDeferred(StagingBufferRef& ref) {
@@ -127,11 +122,10 @@ void StagingBufferPool::TickFrame() {
     ReleaseCache(MemoryUsage::Download);
 }
 
-StagingBufferRef StagingBufferPool::GetStreamBuffer(size_t size) {
+StagingBufferRef StagingBufferPool::GetStreamBuffer(const Device& device, size_t size) {
-    if (AreRegionsActive(Region(free_iterator) + 1,
+    if (AreRegionsActive(Region(free_iterator) + 1, (std::min)(Region(iterator + size) + 1, NUM_SYNCS))) {
-                         (std::min)(Region(iterator + size) + 1, NUM_SYNCS))) {
         // Avoid waiting for the previous usages to be free
-        return GetStagingBuffer(size, MemoryUsage::Upload);
+        return GetStagingBuffer(device, size, MemoryUsage::Upload);
     }
     const u64 current_tick = scheduler.CurrentTick();
     std::fill(sync_ticks.begin() + Region(used_iterator), sync_ticks.begin() + Region(iterator),
@@ -140,15 +134,14 @@ StagingBufferRef StagingBufferPool::GetStreamBuffer(size_t size) {
     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,
+        std::fill(sync_ticks.begin() + Region(used_iterator), sync_ticks.begin() + NUM_SYNCS, current_tick);
-                  current_tick);
         used_iterator = 0;
         iterator = 0;
         free_iterator = size;
 
         if (AreRegionsActive(0, Region(size) + 1)) {
             // Avoid waiting for the previous usages to be free
-            return GetStagingBuffer(size, MemoryUsage::Upload);
+            return GetStagingBuffer(device, size, MemoryUsage::Upload);
         }
     }
     const size_t offset = iterator;
@@ -156,7 +149,7 @@ StagingBufferRef StagingBufferPool::GetStreamBuffer(size_t size) {
     return StagingBufferRef{
         .buffer = *stream_buffer,
         .device_address = stream_buffer_address,
-        .offset = static_cast<VkDeviceSize>(offset),
+        .offset = VkDeviceSize(offset),
         .mapped_span = stream_pointer.subspan(offset, size),
         .usage{},
         .log2_level{},
@@ -166,21 +159,18 @@ StagingBufferRef StagingBufferPool::GetStreamBuffer(size_t size) {
 
 bool StagingBufferPool::AreRegionsActive(size_t region_begin, size_t region_end) const {
     const u64 gpu_tick = scheduler.GetMasterSemaphore().KnownGpuTick();
-    return std::any_of(sync_ticks.begin() + region_begin, sync_ticks.begin() + region_end,
+    return std::any_of(sync_ticks.begin() + region_begin, sync_ticks.begin() + region_end, [gpu_tick](u64 sync_tick) {
-                       [gpu_tick](u64 sync_tick) { return gpu_tick < sync_tick; });
+        return gpu_tick < sync_tick;
+    });
 };
 
-StagingBufferRef StagingBufferPool::GetStagingBuffer(size_t size, MemoryUsage usage,
+StagingBufferRef StagingBufferPool::GetStagingBuffer(const Device& device, size_t size, MemoryUsage usage, bool deferred) {
-                                                     bool deferred) {
+    if (const std::optional<StagingBufferRef> ref = TryGetReservedBuffer(size, usage, deferred))
-    if (const std::optional<StagingBufferRef> ref = TryGetReservedBuffer(size, usage, deferred)) {
         return *ref;
-    }
+    return CreateStagingBuffer(device, size, usage, deferred);
-    return CreateStagingBuffer(size, usage, deferred);
 }
 
-std::optional<StagingBufferRef> StagingBufferPool::TryGetReservedBuffer(size_t size,
+std::optional<StagingBufferRef> StagingBufferPool::TryGetReservedBuffer(size_t size, MemoryUsage usage, bool deferred) {
-                                                                        MemoryUsage usage,
-                                                                        bool deferred) {
     StagingBuffers& cache_level = GetCache(usage)[Common::Log2Ceil(size)];
 
     const auto is_free = [this](const StagingBuffer& entry) {
@@ -202,7 +192,7 @@ std::optional<StagingBufferRef> StagingBufferPool::TryGetReservedBuffer(size_t s
     return it->Ref();
 }
 
-StagingBufferRef StagingBufferPool::CreateStagingBuffer(size_t size, MemoryUsage usage, bool deferred) {
+StagingBufferRef StagingBufferPool::CreateStagingBuffer(const Device& device, size_t size, MemoryUsage usage, bool deferred) {
     auto const log2_size = Common::Log2Ceil<u32>(u32(size));
     VkBufferCreateInfo buffer_ci = {
         .sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO,

src/video_core/renderer_vulkan/vk_staging_buffer_pool.h

@@ -33,11 +33,10 @@ class StagingBufferPool {
 public:
     static constexpr size_t NUM_SYNCS = 16;
 
-    explicit StagingBufferPool(const Device& device, MemoryAllocator& memory_allocator,
+    explicit StagingBufferPool(const Device& device, MemoryAllocator& memory_allocator, Scheduler& scheduler);
-                               Scheduler& scheduler);
     ~StagingBufferPool();
 
-    StagingBufferRef Request(size_t size, MemoryUsage usage, bool deferred = false);
+    StagingBufferRef Request(const Device& device, size_t size, MemoryUsage usage, bool deferred = false);
     void FreeDeferred(StagingBufferRef& ref);
 
     [[nodiscard]] VkBuffer StreamBuf() const noexcept {
@@ -84,27 +83,18 @@ private:
     static constexpr size_t NUM_LEVELS = sizeof(size_t) * CHAR_BIT;
     using StagingBuffersCache = std::array<StagingBuffers, NUM_LEVELS>;
 
-    StagingBufferRef GetStreamBuffer(size_t size);
+    StagingBufferRef GetStreamBuffer(const Device& device, size_t size);
-
     bool AreRegionsActive(size_t region_begin, size_t region_end) const;
-
+    StagingBufferRef GetStagingBuffer(const Device& device, size_t size, MemoryUsage usage, bool deferred = false);
-    StagingBufferRef GetStagingBuffer(size_t size, MemoryUsage usage, bool deferred = false);
+    std::optional<StagingBufferRef> TryGetReservedBuffer(size_t size, MemoryUsage usage, bool deferred);
-
+    StagingBufferRef CreateStagingBuffer(const Device& device, size_t size, MemoryUsage usage, bool deferred);
-    std::optional<StagingBufferRef> TryGetReservedBuffer(size_t size, MemoryUsage usage,
-                                                         bool deferred);
-
-    StagingBufferRef CreateStagingBuffer(size_t size, MemoryUsage usage, bool deferred);
-
     StagingBuffersCache& GetCache(MemoryUsage usage);
-
     void ReleaseCache(MemoryUsage usage);
-
     void ReleaseLevel(StagingBuffersCache& cache, size_t log2);
     size_t Region(size_t iter) const noexcept {
         return iter / region_size;
     }
 
-    const Device& device;
     MemoryAllocator& memory_allocator;
     Scheduler& scheduler;
 

src/video_core/renderer_vulkan/vk_texture_cache.cpp

@@ -975,11 +975,11 @@ void TextureCacheRuntime::Finish() {
 }
 
 StagingBufferRef TextureCacheRuntime::UploadStagingBuffer(size_t size, bool deferred) {
-    return staging_buffer_pool.Request(size, MemoryUsage::Upload, deferred);
+    return staging_buffer_pool.Request(device, size, MemoryUsage::Upload, deferred);
 }
 
 StagingBufferRef TextureCacheRuntime::DownloadStagingBuffer(size_t size, bool deferred) {
-    return staging_buffer_pool.Request(size, MemoryUsage::Download, deferred);
+    return staging_buffer_pool.Request(device, size, MemoryUsage::Download, deferred);
 }
 
 void TextureCacheRuntime::FreeDeferredStagingBuffer(StagingBufferRef& ref) {

src/yuzu_headless/main.cpp

@@ -1,122 +0,0 @@
-// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
-// SPDX-License-Identifier: GPL-3.0-or-later
-
-// SPDX-FileCopyrightText: 2014 Citra Emulator Project
-// SPDX-License-Identifier: GPL-2.0-or-later
-
-#include <chrono>
-#include <iostream>
-#include <memory>
-#include <mutex>
-#include <regex>
-#include <string>
-
-#include "common/logging.h"
-#include "common/settings.h"
-#include "core/core.h"
-#include "core/cpu_manager.h"
-#include "core/file_sys/registered_cache.h"
-#include "core/file_sys/vfs/vfs_real.h"
-#include "core/hle/service/am/applet_manager.h"
-#include "core/hle/service/filesystem/filesystem.h"
-#include "core/loader/loader.h"
-#include "frontend_common/config.h"
-#include "input_common/main.h"
-#include "video_core/gpu.h"
-#include "video_core/renderer_base.h"
-#include "yuzu_cmd/emu_window/emu_window_sdl3.h"
-#include "yuzu_cmd/emu_window/emu_window_sdl3_null.h"
-#include "yuzu_cmd/emu_window/emu_window_sdl3_vk.h"
-
-class EmuWindow_Headless : public Core::Frontend::EmuWindow {
-public:
-    explicit EmuWindow_Headless() = default;
-    ~EmuWindow_Headless() = default;
-
-    bool IsShown() const override {
-        return true;
-    }
-    void OnMinimalClientAreaChangeRequest(std::pair<u32, u32> minimal_size) override {
-
-    }
-    std::unique_ptr<Core::Frontend::GraphicsContext> CreateSharedContext() const override {
-        return std::make_unique<DummyContext>();
-    }
-};
-
-int main(int argc, char *argv[]) {
-    struct {
-        Core::System system{};
-        EmuWindow_Headless emu_window{};
-    } state = {};
-
-    Common::Log::Initialize();
-    Common::Log::SetColorConsoleBackendEnabled(true);
-    Common::Log::Start();
-
-    if (argc < 2) {
-        LOG_CRITICAL(Frontend, "Usage: {} [ms-to-run] [file]", argv[0]);
-        return EXIT_FAILURE;
-    }
-
-    std::chrono::milliseconds time_quanta = std::chrono::milliseconds{atoi(argv[1])};
-    auto const time_end = std::chrono::steady_clock::now() + time_quanta;
-    std::string filepath = argv[2];
-
-    // apply the log_filter setting
-    // the logger was initialized before and doesn't pick up the filter on its own
-    Common::Log::Filter filter;
-    filter.ParseFilterString("*:Info");
-    Common::Log::SetGlobalFilter(filter);
-
-    if (filepath.empty()) {
-        LOG_CRITICAL(Frontend, "Failed to load ROM: No ROM specified");
-        return EXIT_FAILURE;
-    }
-
-    state.system.Initialize();
-
-    InputCommon::InputSubsystem input_subsystem{};
-
-    // Apply the command line arguments
-    state.system.ApplySettings();
-    Settings::values.renderer_backend.SetValue(Settings::RendererBackend::Null);
-
-    state.system.SetContentProvider(std::make_unique<FileSys::ContentProviderUnion>());
-    state.system.SetFilesystem(std::make_shared<FileSys::RealVfsFilesystem>());
-    state.system.GetFileSystemController().CreateFactories(*state.system.GetFilesystem());
-    state.system.GetUserChannel().clear();
-
-    Service::AM::FrontendAppletParameters load_parameters{
-        .applet_id = Service::AM::AppletId::Application,
-    };
-    if (auto const load_result = state.system.Load(state.emu_window, filepath, load_parameters); load_result != Core::SystemResultStatus::Success) {
-        LOG_CRITICAL(Frontend, "load result = {}", u32(load_result));
-        // shutdown
-        void(state.system.Pause());
-        state.system.DetachDebugger();
-        state.system.ShutdownMainProcess();
-        return EXIT_FAILURE;
-    }
-    // Core is loaded, start the GPU (makes the GPU contexts current to this thread)
-    state.system.GPU().Start();
-    state.system.GetCpuManager().OnGpuReady();
-    // don't do anything, SDL3 already exists for us :D
-    state.system.RegisterExitCallback([] {});
-    void(state.system.Run());
-    if (state.system.DebuggerEnabled())
-        state.system.InitializeDebugger();
-    while (state.system.IsPoweredOn() && !state.system.GetExitRequested()) {
-        auto const time_now = std::chrono::steady_clock::now();
-        if (time_now > time_end)
-            break;
-    }
-    // shutdown
-    void(state.system.Pause());
-    state.system.DetachDebugger();
-    state.system.ShutdownMainProcess();
-    return EXIT_SUCCESS;
-}
-
-#define VMA_IMPLEMENTATION
-#include "video_core/vulkan_common/vma.h"