fix linking issues

remove latency func
[audio] Nuke cubeb and use SDL3 exclusively
2026-09-02 19:07:34 +00:00 · 2026-09-02 11:37:51 +00:00 · 2026-09-02 10:58:33 +00:00 · 2026-09-02 10:55:54 +00:00 · 2026-09-02 04:07:24 +02:00
43 changed files with 277 additions and 1726 deletions
@@ -52,7 +52,6 @@ if (YUZU_STATIC_ROOM)
    # disable e v e r y t h i n g
    set(ENABLE_QT OFF)
    set(YUZU_CMD OFF)
-    set(ENABLE_CUBEB OFF)
    set(ENABLE_UPDATE_CHECKER OFF)
    set(USE_DISCORD_PRESENCE OFF)
    set(BUILD_TESTING OFF)
@@ -214,8 +213,6 @@ if(MSVC)
 endif()

 # TODO(crueter): Cleanup, each dep that has a bundled option should allow to choose between bundled, external, system
-option(ENABLE_CUBEB "Enables the cubeb audio backend" ON)
-
 set(EXT_DEFAULT OFF)
 if (MSVC OR ANDROID)
    set(EXT_DEFAULT ON)
@@ -238,7 +235,7 @@ option(YUZU_USE_BUNDLED_SIRIT "Download bundled sirit" ${BUNDLED_SIRIT_DEFAULT})
 # FreeBSD 15+ has libusb, versions below should disable it
 cmake_dependent_option(ENABLE_LIBUSB "Enable the use of LibUSB" ON "WIN32 OR LINUX OR FREEBSD OR APPLE" OFF)

-cmake_dependent_option(ENABLE_OPENGL "Enable OpenGL" ON "NOT (WIN32 AND ARCHITECTURE_arm64) AND NOT APPLE" OFF)
+cmake_dependent_option(ENABLE_OPENGL "Enable OpenGL" ON "NOT (WIN32 AND ARCHITECTURE_arm64) AND NOT APPLE AND NOT ANDROID" OFF)
 mark_as_advanced(FORCE ENABLE_OPENGL)

 option(ENABLE_WEB_SERVICE "Enable web services (telemetry, etc.)" ON)
@@ -574,10 +571,6 @@ if (NOT YUZU_STATIC_ROOM)
        find_package(DiscordRPC)
    endif()

-    if (ENABLE_CUBEB)
-        find_package(cubeb)
-    endif()
-
    if (YUZU_TESTS OR DYNARMIC_TESTS)
        find_package(Catch2)
    endif()
@@ -54,18 +54,6 @@
        "repo": "arun11299/cpp-jwt",
        "version": "7f24eb4c32"
    },
-    "cubeb": {
-        "find_args": "CONFIG",
-        "hash": "8a4bcb2f83ba590f52c66626e895304a73eb61928dbc57777e1822e55378e3568366f17f9da4b80036cc2ef4ea9723c32abf6e7d9bbe00fb03654f0991596ab0",
-        "options": [
-            "USE_SANITIZERS OFF",
-            "BUILD_TESTS OFF",
-            "BUILD_TOOLS OFF",
-            "BUNDLE_SPEEX ON"
-        ],
-        "repo": "mozilla/cubeb",
-        "version": "fa02160712"
-    },
    "discord-rpc": {
        "find_args": "MODULE",
        "hash": "8213c43dcb0f7d479f5861091d111ed12fbdec1e62e6d729d65a4bc181d82f48a35d5fd3cd5c291f2393ac7c9681eabc6b76609755f55376284c8a8d67e148f3",
@@ -100,8 +100,6 @@ For reasons unberknownst to any human being, `glslangValidator` will crash upon

 For this reason this patch is NOT applied to default on all platforms (for obvious reasons) - instead this is a HaikuOS specific patch, apply with `git apply <absolute path to patch>` after cloning SPIRV-Tools then `make -C build` and add the resulting binary (in `build/StandAlone/glslang`) into PATH.

-`cubeb_devel` will also not work, either disable cubeb or uninstall it.
-
 Still will not run flawlessly until `mesa-24` is available. Modify CMakeCache.txt with the `.so` of libGL and libGLESv2 by doing the incredibly difficult task of copy pasting them (`cp /boot/system/lib/libGL.so .`)

 If you have `quazip1_qt6_devel`, uninstall it. It may call `Core5Compat` on CMake which is wrongly packaged.
@@ -73,7 +73,6 @@ Certain other dependencies will be fetched by CPM regardless. System packages *c

 * [SimpleIni](https://github.com/brofield/simpleini)
 * [DiscordRPC](https://github.com/eden-emulator/discord-rpc)
-* [cubeb](https://github.com/mozilla/cubeb)
 * [libusb](https://github.com/libusb/libusb)
 * [VulkanMemoryAllocator](https://github.com/GPUOpen-LibrariesAndSDKs/VulkanMemoryAllocator)
 * [sirit](https://github.com/eden-emulator/sirit)
@@ -122,7 +121,7 @@ sudo emerge -a \
    dev-util/spirv-tools dev-util/spirv-headers dev-util/vulkan-headers \
    dev-util/vulkan-utility-libraries dev-util/glslang \
    media-gfx/renderdoc media-libs/libva media-libs/opus media-video/ffmpeg \
-    media-libs/VulkanMemoryAllocator media-libs/libsdl3 media-libs/cubeb \
+    media-libs/VulkanMemoryAllocator media-libs/libsdl3 \
    net-libs/enet \
    sys-libs/zlib \
    dev-cpp/nlohmann_json dev-cpp/simpleini dev-cpp/cpp-httplib dev-cpp/cpp-jwt \
@@ -356,7 +355,7 @@ pacman -Syuu --needed --noconfirm $packages
 <summary>HaikuOS</summary>

 ```sh
-pkgman install git cmake patch libfmt_devel nlohmann_json lz4_devel opus_devel boost1.90_devel vulkan_devel qt6_base_devel qt6_declarative_devel libsdl3_devel ffmpeg7_devel libx11_devel enet_devel catch2_devel quazip1_qt5_devel qt6_5compat_devel glslang qt6_devel qt6_charts_devel cubeb_devel simpleini quazip_qt6_devel
+pkgman install git cmake patch libfmt_devel nlohmann_json lz4_devel opus_devel boost1.90_devel vulkan_devel qt6_base_devel qt6_declarative_devel libsdl3_devel ffmpeg7_devel libx11_devel enet_devel catch2_devel quazip1_qt5_devel qt6_5compat_devel glslang qt6_devel qt6_charts_devel simpleini quazip_qt6_devel
 ```

 [Caveats](./Caveats.md#haikuos).
@@ -4,7 +4,7 @@ Modern game consoles require heavy power to be emulated appropriately. This is w

 For example, take a disk write, instead of emulating a proper SD card we instead use the C++ standard library for I/O. Additionally we use the abstractions provided by the `fs` service to "lie" to programs about certain SD card properties. Notably this includes making up sizes for the fake SD card, giving "realistic" values or expected outputs for a given card, and so on. And instead of writing to an actual SD card, the emulator simply writes to a file.

-This also means grand part of the emulator consists of just re-implementing firmware but using HLE primitives; for example audio doesn't go to an emulated audio device, but rather gets processed on the fly by a dedicated service and then passed to SDL3/cubeb/etc.
+This also means grand part of the emulator consists of just re-implementing firmware but using HLE primitives; for example audio doesn't go to an emulated audio device, but rather gets processed on the fly by a dedicated service and then passed to SDL3 or null backend.

 As such, many of the systems implemented are not 100% accurate to the original software, but they're "good enough" to pass as being so. While we do strive to maintain high compatibility (especially with homebrew), there are realistic limitations to these approaches.

@@ -35,8 +35,6 @@ These options control dependencies.

 - `ENABLE_WEB_SERVICE` (ON) Enable multiplayer service
 - `ENABLE_WIFI_SCAN` (OFF) Enable WiFi scanning (requires iw on Linux) - experimental
- `ENABLE_CUBEB` (ON) Enables the cubeb audio backend
-  - This option is subject for removal.
 - `YUZU_TESTS` (ON) Compile tests - requires Catch2
 - `ENABLE_LTO` (OFF) Enable link-time optimization
  - Not recommended on Windows
@@ -76,7 +74,7 @@ The following options are desktop only.

 - `ENABLE_LIBUSB` (ON) Enable the use of the libusb input backend (HIGHLY RECOMMENDED)
 - `ENABLE_OPENGL` (ON) Enable the OpenGL graphics backend
-  - Unavailable on Windows/ARM64
+  - Unavailable on Windows/ARM64 and on Android
  - You probably shouldn't turn this off.

 ### Qt
@@ -106,36 +106,6 @@ endif()
 # SimpleIni
 AddJsonPackage(simpleini)

-# Most linux distros don't package cubeb, so enable regardless of cpm settings
-if(ENABLE_CUBEB)
-    AddJsonPackage(cubeb)
-
-    if (cubeb_ADDED)
-        if (NOT MSVC)
-            if (TARGET speex)
-                target_compile_options(speex PRIVATE $<$<COMPILE_LANGUAGE:C,CXX>:-Wno-sign-compare>)
-            endif()
-            set_target_properties(cubeb PROPERTIES COMPILE_OPTIONS "")
-            target_compile_options(cubeb INTERFACE
-                $<$<COMPILE_LANGUAGE:C,CXX>:-Wno-implicit-const-int-float-conversion>
-                $<$<COMPILE_LANGUAGE:C,CXX>:-Wno-shadow>
-                $<$<COMPILE_LANGUAGE:C,CXX>:-Wno-missing-declarations>
-                $<$<COMPILE_LANGUAGE:C,CXX>:-Wno-return-type>
-                $<$<COMPILE_LANGUAGE:C,CXX>:-Wno-uninitialized>
-            )
-        else()
-            target_compile_options(cubeb PRIVATE
-                /wd4456
-                /wd4458
-            )
-        endif()
-    endif()
-
-    if (NOT TARGET cubeb::cubeb)
-        add_library(cubeb::cubeb ALIAS cubeb)
-    endif()
-endif()
-
 if (NOT YUZU_USE_BUNDLED_SDL3)
    if (NOT WIN32)
        # Yuzu itself needs: Atomic Audio Events Joystick Haptic Sensor Threads Timers
@@ -15,7 +15,7 @@ pkgs.mkShellNoCC {
    enet libopus vulkan-headers vulkan-utility-libraries
    spirv-tools spirv-headers vulkan-loader unzip
    glslang python3 httplib cpp-jwt ffmpeg-headless
-    libusb1 cubeb
+    libusb1
    # eden
    qt6.qtbase qt6.qtmultimedia qt6.qtwayland qt6.qttools
    qt6.qtwebengine qt6.qt5compat
@@ -58,13 +58,6 @@ class LicensesFragment : Fragment() {
                R.string.license_fidelityfx_fsr_copyright,
                R.string.license_fidelityfx_fsr_text
            ),
-            License(
-                R.string.license_cubeb,
-                R.string.license_cubeb_description,
-                R.string.license_cubeb_link,
-                R.string.license_cubeb_copyright,
-                R.string.license_cubeb_text
-            ),
            License(
                R.string.license_dynarmic,
                R.string.license_dynarmic_description,
@@ -799,9 +799,6 @@
    <string name="theme_mode_light">روشن</string>
    <string name="theme_mode_dark">تاریک</string>

-    <!-- Audio output engines -->
-    <string name="cubeb">cubeb</string>
-
    <!-- Anisotropic filtering options -->
    <string name="multiplier_x2">x2</string>
    <string name="multiplier_x4">x4</string>
@@ -466,13 +466,11 @@
    <string-array name="outputEngineEntries">
        <item>@string/auto</item>
        <item>@string/sdl3</item>
-        <item>@string/cubeb</item>
        <item>@string/string_null</item>
    </string-array>
    <integer-array name="outputEngineValues">
        <item>0</item>
        <item>2</item>
-        <item>1</item>
        <item>3</item>
    </integer-array>

@@ -1241,7 +1241,6 @@

    <!-- Audio output engines -->
    <string name="sdl3" translatable="false">SDL3</string>
-    <string name="cubeb" translatable="false">cubeb</string>

    <!-- Anisotropic filtering options -->
    <string name="multiplier_x2" translatable="false">x2</string>
@@ -1381,23 +1380,6 @@ AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
 LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
 OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
 THE SOFTWARE.
-    </string>
-    <string name="license_cubeb" translatable="false">cubeb</string>
-    <string name="license_cubeb_description" translatable="false">Cross platform audio library</string>
-    <string name="license_cubeb_link" translatable="false">https://github.com/mozilla/cubeb</string>
-    <string name="license_cubeb_copyright" translatable="false">Copyright © 2011 Mozilla Foundation</string>
-    <string name="license_cubeb_text" translatable="false">
-Permission to use, copy, modify, and distribute this software for any
-purpose with or without fee is hereby granted, provided that the above
-copyright notice and this permission notice appear in all copies.\n\n
-
-THE SOFTWARE IS PROVIDED \"AS IS\" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
-WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
-MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
-ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
-WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
-ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
-OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
    </string>
    <string name="license_dynarmic" translatable="false">Dynarmic</string>
    <string name="license_dynarmic_description" translatable="false">An ARM dynamic recompiler</string>
@@ -229,15 +229,6 @@ endif()
 target_include_directories(audio_core PRIVATE ${OPUS_INCLUDE_DIRS})
 target_link_libraries(audio_core PUBLIC common core Opus::opus)

-if (ENABLE_CUBEB)
-    target_sources(audio_core PRIVATE
-        sink/cubeb_sink.cpp
-        sink/cubeb_sink.h)
-
-    target_link_libraries(audio_core PRIVATE cubeb::cubeb)
-    target_compile_definitions(audio_core PRIVATE HAVE_CUBEB=1)
-endif()
-
 target_sources(audio_core PRIVATE
    sink/sdl3_sink.cpp
    sink/sdl3_sink.h)
@@ -245,4 +236,8 @@ target_sources(audio_core PRIVATE
 target_link_libraries(audio_core PRIVATE SDL3::SDL3)
 target_compile_definitions(audio_core PRIVATE HAVE_SDL3)

+if (ANDROID)
+    target_link_libraries(audio_core PRIVATE OpenSLES)
+endif()
+
 create_target_directory_groups(audio_core)
@@ -1,450 +0,0 @@
-// 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
-
-#include <span>
-#include <vector>
-
-#include "audio_core/common/common.h"
-#include "audio_core/sink/cubeb_sink.h"
-#include "audio_core/sink/sink_stream.h"
-#include "common/logging.h"
-#include "common/scope_exit.h"
-#include "core/core.h"
-
-#ifdef _WIN32
-#include <objbase.h>
-#undef CreateEvent
-#endif
-
-namespace AudioCore::Sink {
-/**
- * Cubeb sink stream, responsible for sinking samples to hardware.
- */
-class CubebSinkStream final : public SinkStream {
-public:
-    /**
-     * Create a new sink stream.
-     *
-     * @param ctx_             - Cubeb context to create this stream with.
-     * @param device_channels_ - Number of channels supported by the hardware.
-     * @param system_channels_ - Number of channels the audio systems expect.
-     * @param output_device    - Cubeb output device id.
-     * @param input_device     - Cubeb input device id.
-     * @param name_            - Name of this stream.
-     * @param type_            - Type of this stream.
-     * @param system_          - Core system.
-     * @param event            - Event used only for audio renderer, signalled on buffer consume.
-     */
-    CubebSinkStream(cubeb* ctx_, u32 device_channels_, u32 system_channels_,
-                    cubeb_devid output_device, cubeb_devid input_device, const std::string& name_,
-                    StreamType type_, Core::System& system_)
-        : SinkStream(system_, type_), ctx{ctx_} {
-#ifdef _WIN32
-        CoInitializeEx(nullptr, COINIT_MULTITHREADED);
-#endif
-        name = name_;
-        device_channels = device_channels_;
-        system_channels = system_channels_;
-
-        cubeb_stream_params params{};
-        params.rate = TargetSampleRate;
-        params.channels = device_channels;
-        params.format = CUBEB_SAMPLE_S16LE;
-        params.prefs = CUBEB_STREAM_PREF_NONE;
-        switch (params.channels) {
-        case 1:
-            params.layout = CUBEB_LAYOUT_MONO;
-            break;
-        case 2:
-            params.layout = CUBEB_LAYOUT_STEREO;
-            break;
-        case 6:
-            params.layout = CUBEB_LAYOUT_3F2_LFE;
-            break;
-        }
-
-        u32 minimum_latency{0};
-        const auto latency_error = cubeb_get_min_latency(ctx, &params, &minimum_latency);
-        if (latency_error != CUBEB_OK) {
-            LOG_CRITICAL(Audio_Sink, "Error getting minimum latency, error: {}", latency_error);
-            minimum_latency = TargetSampleCount * 2;
-        }
-
-        minimum_latency = (std::max)(minimum_latency, TargetSampleCount * 2);
-
-        LOG_INFO(Service_Audio,
-                 "Opening cubeb stream {} type {} with: rate {} channels {} (system channels {}) "
-                 "latency {}",
-                 name, type, params.rate, params.channels, system_channels, minimum_latency);
-
-        auto init_error{0};
-        if (type == StreamType::In) {
-            init_error = cubeb_stream_init(ctx, &stream_backend, name.c_str(), input_device,
-                                           &params, output_device, nullptr, minimum_latency,
-                                           &CubebSinkStream::DataCallback,
-                                           &CubebSinkStream::StateCallback, this);
-        } else {
-            init_error = cubeb_stream_init(ctx, &stream_backend, name.c_str(), input_device,
-                                           nullptr, output_device, &params, minimum_latency,
-                                           &CubebSinkStream::DataCallback,
-                                           &CubebSinkStream::StateCallback, this);
-        }
-
-        if (init_error != CUBEB_OK) {
-            LOG_CRITICAL(Audio_Sink, "Error initializing cubeb stream, error: {}", init_error);
-            return;
-        }
-    }
-
-    /**
-     * Destroy the sink stream.
-     */
-    ~CubebSinkStream() override {
-        LOG_DEBUG(Service_Audio, "Destructing cubeb stream {}", name);
-
-        if (!ctx) {
-            return;
-        }
-
-        Finalize();
-
-#ifdef _WIN32
-        CoUninitialize();
-#endif
-    }
-
-    /**
-     * Finalize the sink stream.
-     */
-    void Finalize() override {
-        Stop();
-        cubeb_stream_destroy(stream_backend);
-    }
-
-    /**
-     * Start the sink stream.
-     *
-     * @param resume - Set to true if this is resuming the stream a previously-active stream.
-     *                 Default false.
-     */
-    void Start(bool resume = false) override {
-        if (!ctx || !paused) {
-            return;
-        }
-
-        paused = false;
-        if (cubeb_stream_start(stream_backend) != CUBEB_OK) {
-            LOG_CRITICAL(Audio_Sink, "Error starting cubeb stream");
-        }
-    }
-
-    /**
-     * Stop the sink stream.
-     */
-    void Stop() override {
-        if (!ctx || paused) {
-            return;
-        }
-
-        SignalPause();
-        if (cubeb_stream_stop(stream_backend) != CUBEB_OK) {
-            LOG_CRITICAL(Audio_Sink, "Error stopping cubeb stream");
-        }
-    }
-
-private:
-    /**
-     * Main callback from Cubeb. Either expects samples from us (audio render/audio out), or will
-     * provide samples to be copied (audio in).
-     *
-     * @param stream      - Cubeb-specific data about the stream.
-     * @param user_data   - Custom data pointer passed along, points to a CubebSinkStream.
-     * @param in_buff     - Input buffer to be used if the stream is an input type.
-     * @param out_buff    - Output buffer to be used if the stream is an output type.
-     * @param num_frames_ - Number of frames of audio in the buffers. Note: Not number of samples.
-     */
-    static long DataCallback([[maybe_unused]] cubeb_stream* stream, void* user_data,
-                             [[maybe_unused]] const void* in_buff, void* out_buff,
-                             long num_frames_) {
-        auto* impl = static_cast<CubebSinkStream*>(user_data);
-        if (!impl) {
-            return -1;
-        }
-
-        const std::size_t num_channels = impl->GetDeviceChannels();
-        const std::size_t frame_size = num_channels;
-        const std::size_t num_frames{static_cast<size_t>(num_frames_)};
-
-        if (impl->type == StreamType::In) {
-            std::span<const s16> input_buffer{reinterpret_cast<const s16*>(in_buff),
-                                              num_frames * frame_size};
-            impl->ProcessAudioIn(input_buffer, num_frames);
-        } else {
-            std::span<s16> output_buffer{reinterpret_cast<s16*>(out_buff), num_frames * frame_size};
-            impl->ProcessAudioOutAndRender(output_buffer, num_frames);
-        }
-
-        return num_frames_;
-    }
-
-    /**
-     * Cubeb callback for if a device state changes. Unused currently.
-     *
-     * @param stream      - Cubeb-specific data about the stream.
-     * @param user_data   - Custom data pointer passed along, points to a CubebSinkStream.
-     * @param state       - New state of the device.
-     */
-    static void StateCallback(cubeb_stream*, void*, cubeb_state) {}
-
-    /// Main Cubeb context
-    cubeb* ctx{};
-    /// Cubeb stream backend
-    cubeb_stream* stream_backend{};
-};
-
-CubebSink::CubebSink(std::string_view target_device_name) {
-    // Cubeb requires COM to be initialized on the thread calling cubeb_init on Windows
-#ifdef _WIN32
-    com_init_result = CoInitializeEx(nullptr, COINIT_MULTITHREADED);
-#endif
-
-    if (cubeb_init(&ctx, "Eden", nullptr) != CUBEB_OK) {
-        LOG_CRITICAL(Audio_Sink, "cubeb_init failed");
-        return;
-    }
-
-    if (target_device_name != auto_device_name && !target_device_name.empty()) {
-        cubeb_device_collection collection;
-        if (cubeb_enumerate_devices(ctx, CUBEB_DEVICE_TYPE_OUTPUT, &collection) != CUBEB_OK) {
-            LOG_WARNING(Audio_Sink, "Audio output device enumeration not supported");
-        } else {
-            const auto collection_end{collection.device + collection.count};
-            const auto device{
-                std::find_if(collection.device, collection_end, [&](const cubeb_device_info& info) {
-                    return info.friendly_name != nullptr &&
-                           target_device_name == std::string(info.friendly_name);
-                })};
-            if (device != collection_end) {
-                output_device = device->devid;
-            }
-            cubeb_device_collection_destroy(ctx, &collection);
-        }
-    }
-
-    cubeb_get_max_channel_count(ctx, &device_channels);
-    device_channels = device_channels >= 6U ? 6U : 2U;
-}
-
-CubebSink::~CubebSink() {
-    if (!ctx) {
-        return;
-    }
-
-    for (auto& sink_stream : sink_streams) {
-        sink_stream.reset();
-    }
-
-    cubeb_destroy(ctx);
-
-#ifdef _WIN32
-    if (SUCCEEDED(com_init_result)) {
-        CoUninitialize();
-    }
-#endif
-}
-
-SinkStream* CubebSink::AcquireSinkStream(Core::System& system, u32 system_channels_,
-                                         const std::string& name, StreamType type) {
-    system_channels = system_channels_;
-    SinkStreamPtr& stream = sink_streams.emplace_back(std::make_unique<CubebSinkStream>(
-        ctx, device_channels, system_channels, output_device, input_device, name, type, system));
-
-    return stream.get();
-}
-
-void CubebSink::CloseStream(SinkStream* stream) {
-    for (size_t i = 0; i < sink_streams.size(); i++) {
-        if (sink_streams[i].get() == stream) {
-            sink_streams[i].reset();
-            sink_streams.erase(sink_streams.begin() + i);
-            break;
-        }
-    }
-}
-
-void CubebSink::CloseStreams() {
-    sink_streams.clear();
-}
-
-f32 CubebSink::GetDeviceVolume() const {
-    if (sink_streams.empty()) {
-        return 1.0f;
-    }
-
-    return sink_streams[0]->GetDeviceVolume();
-}
-
-void CubebSink::SetDeviceVolume(f32 volume) {
-    for (auto& stream : sink_streams) {
-        stream->SetDeviceVolume(volume);
-    }
-}
-
-void CubebSink::SetSystemVolume(f32 volume) {
-    for (auto& stream : sink_streams) {
-        stream->SetSystemVolume(volume);
-    }
-}
-
-std::vector<std::string> ListCubebSinkDevices(bool capture) {
-    std::vector<std::string> device_list;
-    cubeb* ctx;
-
-#ifdef _WIN32
-    auto com_init_result = CoInitializeEx(nullptr, COINIT_MULTITHREADED);
-#endif
-
-    if (cubeb_init(&ctx, "Eden Device Enumerator", nullptr) != CUBEB_OK) {
-        LOG_CRITICAL(Audio_Sink, "cubeb_init failed");
-        return {};
-    }
-
-#ifdef _WIN32
-    if (SUCCEEDED(com_init_result)) {
-        CoUninitialize();
-    }
-#endif
-
-    auto type{capture ? CUBEB_DEVICE_TYPE_INPUT : CUBEB_DEVICE_TYPE_OUTPUT};
-    cubeb_device_collection collection;
-    if (cubeb_enumerate_devices(ctx, type, &collection) != CUBEB_OK) {
-        LOG_WARNING(Audio_Sink, "Audio output device enumeration not supported");
-    } else {
-        for (std::size_t i = 0; i < collection.count; i++) {
-            const cubeb_device_info& device = collection.device[i];
-            if (device.friendly_name && device.friendly_name[0] != '\0' &&
-                device.state == CUBEB_DEVICE_STATE_ENABLED) {
-                device_list.emplace_back(device.friendly_name);
-            }
-        }
-        cubeb_device_collection_destroy(ctx, &collection);
-    }
-
-    cubeb_destroy(ctx);
-    return device_list;
-}
-
-/* REVERSION TO 3833 - function GetCubebLatency REINTRODUCED FROM 3833 - DIABLO 3 FIX */
-u32 GetCubebLatency() {
-    cubeb* ctx;
-
-#ifdef _WIN32
-    auto com_init_result = CoInitializeEx(nullptr, COINIT_MULTITHREADED);
-#endif
-
-    // Init cubeb
-    if (cubeb_init(&ctx, "yuzu Latency Getter", nullptr) != CUBEB_OK) {
-        LOG_CRITICAL(Audio_Sink, "cubeb_init failed");
-        // Return a large latency so we choose SDL instead.
-        return 10000u;
-    }
-
-#ifdef _WIN32
-    if (SUCCEEDED(com_init_result)) {
-        CoUninitialize();
-    }
-#endif
-
-    // Get min latency
-    cubeb_stream_params params{};
-    params.rate = TargetSampleRate;
-    params.channels = 2;
-    params.format = CUBEB_SAMPLE_S16LE;
-    params.prefs = CUBEB_STREAM_PREF_NONE;
-    params.layout = CUBEB_LAYOUT_STEREO;
-
-    u32 latency{0};
-    const auto latency_error = cubeb_get_min_latency(ctx, &params, &latency);
-    if (latency_error != CUBEB_OK) {
-        LOG_CRITICAL(Audio_Sink, "Error getting minimum latency, error: {}", latency_error);
-        latency = TargetSampleCount * 2;
-    }
-    latency = (std::max)(latency, TargetSampleCount * 2);
-    cubeb_destroy(ctx);
-    return latency;
-}
-
-// REVERTED back to 3833 - Below namespace section and function IsCubebSuitable() removed, reverting to GetCubebLatency() above. - DIABLO 3 FIX
-/*
-namespace {
-static long TmpDataCallback(cubeb_stream*, void*, const void*, void*, long) {
-    return TargetSampleCount;
-}
-static void TmpStateCallback(cubeb_stream*, void*, cubeb_state) {}
-} // namespace
-
-bool IsCubebSuitable() {
-#if !defined(HAVE_CUBEB)
-    return false;
-#else
-    cubeb* ctx{nullptr};
-
-#ifdef _WIN32
-    auto com_init_result = CoInitializeEx(nullptr, COINIT_MULTITHREADED);
-#endif
-
-    // Init cubeb
-    if (cubeb_init(&ctx, "Eden Latency Getter", nullptr) != CUBEB_OK) {
-        LOG_ERROR(Audio_Sink, "Cubeb failed to init, it is not suitable.");
-        return false;
-    }
-
-    SCOPE_EXIT {
-        cubeb_destroy(ctx);
-    };
-
-#ifdef _WIN32
-    if (SUCCEEDED(com_init_result)) {
-        CoUninitialize();
-    }
-#endif
-
-    // Get min latency
-    cubeb_stream_params params{};
-    params.rate = TargetSampleRate;
-    params.channels = 2;
-    params.format = CUBEB_SAMPLE_S16LE;
-    params.prefs = CUBEB_STREAM_PREF_NONE;
-    params.layout = CUBEB_LAYOUT_STEREO;
-
-    u32 latency{0};
-    const auto latency_error = cubeb_get_min_latency(ctx, &params, &latency);
-    if (latency_error != CUBEB_OK) {
-        LOG_ERROR(Audio_Sink, "Cubeb could not get min latency, it is not suitable.");
-        return false;
-    }
-    latency = (std::max)(latency, TargetSampleCount * 2);
-
-    // Test opening a device with standard parameters
-    cubeb_devid output_device{0};
-    cubeb_devid input_device{0};
-    std::string name{"Eden test"};
-    cubeb_stream* stream{nullptr};
-
-    if (cubeb_stream_init(ctx, &stream, name.c_str(), input_device, nullptr, output_device, &params,
-                          latency, &TmpDataCallback, &TmpStateCallback, nullptr) != CUBEB_OK) {
-        LOG_CRITICAL(Audio_Sink, "Cubeb could not open a device, it is not suitable.");
-        return false;
-    }
-
-    cubeb_stream_stop(stream);
-    cubeb_stream_destroy(stream);
-    return true;
-#endif
-}
-*/
-
-} // namespace AudioCore::Sink
@@ -1,118 +0,0 @@
-// SPDX-FileCopyrightText: Copyright 2025 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
-
-#pragma once
-
-#include <string>
-#include <vector>
-
-#include <cubeb/cubeb.h>
-
-#include "audio_core/sink/sink.h"
-
-namespace Core {
-class System;
-}
-
-namespace AudioCore::Sink {
-class SinkStream;
-
-/**
- * Cubeb backend sink, holds multiple output streams and is responsible for sinking samples to
- * hardware. Used by Audio Render, Audio In and Audio Out.
- */
-class CubebSink final : public Sink {
-public:
-    explicit CubebSink(std::string_view device_id);
-    ~CubebSink() override;
-
-    /**
-     * Create a new sink stream.
-     *
-     * @param system          - Core system.
-     * @param system_channels - Number of channels the audio system expects.
-     *                          May differ from the device's channel count.
-     * @param name            - Name of this stream.
-     * @param type            - Type of this stream, render/in/out.
-     *
-     * @return A pointer to the created SinkStream
-     */
-    SinkStream* AcquireSinkStream(Core::System& system, u32 system_channels,
-                                  const std::string& name, StreamType type) override;
-
-    /**
-     * Close a given stream.
-     *
-     * @param stream - The stream to close.
-     */
-    void CloseStream(SinkStream* stream) override;
-
-    /**
-     * Close all streams.
-     */
-    void CloseStreams() override;
-
-    /**
-     * Get the device volume. Set from calls to the IAudioDevice service.
-     *
-     * @return Volume of the device.
-     */
-    f32 GetDeviceVolume() const override;
-
-    /**
-     * Set the device volume. Set from calls to the IAudioDevice service.
-     *
-     * @param volume - New volume of the device.
-     */
-    void SetDeviceVolume(f32 volume) override;
-
-    /**
-     * Set the system volume. Comes from the audio system using this stream.
-     *
-     * @param volume - New volume of the system.
-     */
-    void SetSystemVolume(f32 volume) override;
-
-private:
-    /// Backend Cubeb context
-    cubeb* ctx{};
-    /// Cubeb id of the actual hardware output device
-    cubeb_devid output_device{};
-    /// Cubeb id of the actual hardware input device
-    cubeb_devid input_device{};
-    /// Vector of streams managed by this sink
-    std::vector<SinkStreamPtr> sink_streams{};
-
-#ifdef _WIN32
-    /// Cubeb required COM to be initialized multi-threaded on Windows
-    u32 com_init_result = 0;
-#endif
-};
-
-/**
- * Get a list of connected devices from Cubeb.
- *
- * @param capture - Return input (capture) devices if true, otherwise output devices.
- */
-std::vector<std::string> ListCubebSinkDevices(bool capture);
-
-// REVERSION - function GetCubebLatency() reintroduced from EA-3833 - DIABLO 3 FIX
-/**
- * Get the reported latency for this sink.
- *
- * @return Minimum latency for this sink.
- */
-u32 GetCubebLatency();
-
-/**
- * Check if this backend is suitable for use.
- * Checks if enabled, its latency, whether it opens successfully, etc.
- *
- * @return True is this backend is suitable, false otherwise.
- */
-// bool IsCubebSuitable(); // REVERTED BACK TO GetCubebLatency() FROM 3833
-
-} // namespace AudioCore::Sink
@@ -25,9 +25,10 @@ namespace {
        // See https://github.com/PCSX2/pcsx2/pull/12312
        // "SDL and cubeb backends previously resulted in different names for the output which
        // caused them be identified as different applications by the OS."
-        //
-        // Keep in sync with cubeb_sink.cpp name.
        SDL_SetHint("SDL_AUDIO_DEVICE_APP_NAME", "yuzu Latency Getter");
+        SDL_SetHint("SDL_HINT_AUDIO_DEVICE_STREAM_ROLE", "Game");
+        // We do our own processing, so just have SDL copy our audio
+        SDL_SetHint("SDL_HINT_AUDIO_DEVICE_RAW_STREAM", "1");
        if (!SDL_InitSubSystem(SDL_INIT_AUDIO)) {
            LOG_CRITICAL(Audio_Sink, "SDL_InitSubSystem audio failed: {}", SDL_GetError());
            return false;
@@ -82,23 +83,18 @@ public:

        SDL_AudioSpec spec{};
        spec.freq = TargetSampleRate;
-        spec.channels = static_cast<u8>(device_channels);
-        spec.format = SDL_AUDIO_S16;
+        spec.channels = u8(device_channels);
+        spec.format = SDL_AUDIO_S16LE;

-        std::string device_name{output_device};
-        bool capture{false};
-        if (type == StreamType::In) {
-            device_name = input_device;
-            capture = true;
-        }
+        auto const is_capture = (type == StreamType::In);
+        std::string device_name = is_capture ? input_device : output_device;
+        const SDL_AudioDeviceID audio_device = device_name.empty()
+            ? (is_capture
+                ? SDL_AUDIO_DEVICE_DEFAULT_RECORDING
+                : SDL_AUDIO_DEVICE_DEFAULT_PLAYBACK)
+            : FindAudioDeviceByName(device_name, is_capture);

-        const SDL_AudioDeviceID audio_device =
-            device_name.empty() ? (capture ? SDL_AUDIO_DEVICE_DEFAULT_RECORDING
-                                           : SDL_AUDIO_DEVICE_DEFAULT_PLAYBACK)
-                                : FindAudioDeviceByName(device_name, capture);
-
-        stream = SDL_OpenAudioDeviceStream(audio_device, &spec, &SDLSinkStream::DataCallback,
-                                           this);
+        stream = SDL_OpenAudioDeviceStream(audio_device, &spec, &SDLSinkStream::DataCallback, this);

        if (stream == nullptr) {
            LOG_CRITICAL(Audio_Sink, "Error opening SDL audio device: {}", SDL_GetError());
@@ -107,13 +103,12 @@ public:

        SDL_AudioSpec stream_in{};
        SDL_AudioSpec stream_out{};
-        static_cast<void>(SDL_GetAudioStreamFormat(stream, &stream_in, &stream_out));
-
+        void(SDL_GetAudioStreamFormat(stream, &stream_in, &stream_out));
        LOG_INFO(Service_Audio,
                 "Opening SDL stream {} with: rate {} channels {} (system channels {}) "
                 " format {}",
                 static_cast<const void*>(stream), stream_out.freq, stream_out.channels,
-                 system_channels, static_cast<int>(stream_out.format));
+                 system_channels, int(stream_out.format));
    }

    /**
@@ -128,14 +123,12 @@ public:
     * Finalize the sink stream.
     */
    void Finalize() override {
-        if (stream == nullptr) {
-            return;
+        if (stream != nullptr) {
+            Stop();
+            SDL_ClearAudioStream(stream);
+            SDL_DestroyAudioStream(stream);
+            stream = nullptr;
        }
-
-        Stop();
-        SDL_ClearAudioStream(stream);
-        SDL_DestroyAudioStream(stream);
-        stream = nullptr;
    }

    /**
@@ -148,9 +141,8 @@ public:
        if (stream == nullptr || !paused) {
            return;
        }
-
        paused = false;
-        static_cast<void>(SDL_ResumeAudioStreamDevice(stream));
+        void(SDL_ResumeAudioStreamDevice(stream));
    }

    /**
@@ -161,7 +153,7 @@ public:
            return;
        }
        SignalPause();
-        static_cast<void>(SDL_PauseAudioStreamDevice(stream));
+        void(SDL_PauseAudioStreamDevice(stream));
    }

 private:
@@ -173,51 +165,34 @@ private:
     * @param stream   - Buffer of samples to be filled or read.
     * @param len      - Length of the stream in bytes.
     */
-    static void DataCallback(void* userdata, SDL_AudioStream* stream, int additional_amount,
-                             int total_amount) {
-        auto* impl = static_cast<SDLSinkStream*>(userdata);
-
-        if (!impl) {
-            return;
-        }
-
-        const std::size_t num_channels = impl->GetDeviceChannels();
-        const std::size_t frame_size = num_channels;
-
-        if (impl->type == StreamType::In) {
-            const int bytes_available = SDL_GetAudioStreamAvailable(stream);
-            if (bytes_available <= 0) {
-                return;
+    static void DataCallback(void* userdata, SDL_AudioStream* stream, int additional_amount, int total_amount) {
+        if (auto* impl = static_cast<SDLSinkStream*>(userdata); impl) {
+            auto const num_channels = impl->GetDeviceChannels();
+            auto const frame_size = num_channels;
+            if (impl->type == StreamType::In) {
+                auto const bytes_available = SDL_GetAudioStreamAvailable(stream);
+                if (bytes_available > 0) {
+                    std::vector<s16> input(bytes_available / int(sizeof(s16)));
+                    auto const bytes_read = SDL_GetAudioStreamData(stream, input.data(), bytes_available);
+                    if (bytes_read > 0) {
+                        auto const num_frames = std::size_t(bytes_read) / sizeof(s16) / frame_size;
+                        std::span<const s16> input_buffer{input.data(), std::size_t(bytes_read) / sizeof(s16)};
+                        impl->ProcessAudioIn(input_buffer, num_frames);
+                    }
+                }
+            } else if (additional_amount > 0 || total_amount > 0) {
+                auto const bytes_requested = additional_amount > 0 ? additional_amount : total_amount;
+                std::vector<s16> output(bytes_requested / int(sizeof(s16)));
+                auto const num_frames = std::size_t(bytes_requested) / sizeof(s16) / frame_size;
+                std::span<s16> output_buffer{output.data(), output.size()};
+                impl->ProcessAudioOutAndRender(output_buffer, num_frames);
+                void(SDL_PutAudioStreamData(stream, output.data(), bytes_requested));
            }
-
-            std::vector<s16> input(bytes_available / static_cast<int>(sizeof(s16)));
-            const int bytes_read = SDL_GetAudioStreamData(stream, input.data(), bytes_available);
-            if (bytes_read <= 0) {
-                return;
-            }
-
-            const std::size_t num_frames =
-                static_cast<std::size_t>(bytes_read) / sizeof(s16) / frame_size;
-            std::span<const s16> input_buffer{input.data(),
-                                              static_cast<std::size_t>(bytes_read) / sizeof(s16)};
-            impl->ProcessAudioIn(input_buffer, num_frames);
-        } else {
-            if (additional_amount <= 0 && total_amount <= 0) {
-                return;
-            }
-
-            const int bytes_requested = additional_amount > 0 ? additional_amount : total_amount;
-            std::vector<s16> output(bytes_requested / static_cast<int>(sizeof(s16)));
-            const std::size_t num_frames =
-                static_cast<std::size_t>(bytes_requested) / sizeof(s16) / frame_size;
-            std::span<s16> output_buffer{output.data(), output.size()};
-            impl->ProcessAudioOutAndRender(output_buffer, num_frames);
-            static_cast<void>(SDL_PutAudioStreamData(stream, output.data(), bytes_requested));
        }
    }

    /// SDL stream attached to an opened input/output device
-    SDL_AudioStream* stream{};
+    SDL_AudioStream* stream = nullptr;
 };

 SDLSink::SDLSink(std::string_view target_device_name) {
@@ -233,11 +208,9 @@ SDLSink::SDLSink(std::string_view target_device_name) {

 SDLSink::~SDLSink() = default;

-SinkStream* SDLSink::AcquireSinkStream(Core::System& system, u32 system_channels_,
-                                       const std::string&, StreamType type) {
+SinkStream* SDLSink::AcquireSinkStream(Core::System& system, u32 system_channels_, const std::string&, StreamType type) {
    system_channels = system_channels_;
-    SinkStreamPtr& stream = sink_streams.emplace_back(std::make_unique<SDLSinkStream>(
-        device_channels, system_channels, output_device, input_device, type, system));
+    SinkStreamPtr& stream = sink_streams.emplace_back(std::make_unique<SDLSinkStream>(device_channels, system_channels, output_device, input_device, type, system));
    return stream.get();
 }

@@ -256,11 +229,7 @@ void SDLSink::CloseStreams() {
 }

 f32 SDLSink::GetDeviceVolume() const {
-    if (sink_streams.empty()) {
-        return 1.0f;
-    }
-
-    return sink_streams[0]->GetDeviceVolume();
+    return sink_streams.empty() ? 1.0f : sink_streams[0]->GetDeviceVolume();
 }

 void SDLSink::SetDeviceVolume(f32 volume) {
@@ -275,68 +244,21 @@ void SDLSink::SetSystemVolume(f32 volume) {
    }
 }

-std::vector<std::string> ListSDLSinkDevices(bool capture) {
-    if (!InitializeAudio())
-        return {}; //no devices
-
-    std::vector<std::string> device_list;
-    int device_count = 0;
-    SDL_AudioDeviceID* devices =
-        capture ? SDL_GetAudioRecordingDevices(&device_count)
-                : SDL_GetAudioPlaybackDevices(&device_count);
-    if (devices == nullptr) {
-        return device_list;
-    }
-
-    for (int i = 0; i < device_count; ++i) {
-        if (const char* name = SDL_GetAudioDeviceName(devices[i])) {
-            device_list.emplace_back(name);
+std::vector<std::string> ListSDLSinkDevices(bool is_capture) {
+    std::vector<std::string> device_list{};
+    if (InitializeAudio()) {
+        int device_count = 0;
+        SDL_AudioDeviceID* devices = is_capture ? SDL_GetAudioRecordingDevices(&device_count) : SDL_GetAudioPlaybackDevices(&device_count);
+        if (devices != nullptr) {
+            for (int i = 0; i < device_count; ++i) {
+                if (const char* name = SDL_GetAudioDeviceName(devices[i]); name) {
+                    device_list.emplace_back(name);
+                }
+            }
+            SDL_free(devices);
        }
    }
-    SDL_free(devices);
-
    return device_list;
 }

-/* REVERSION to 3833 - function GetSDLLatency() REINTRODUCED FROM 3833 - DIABLO 3 FIX */
-u32 GetSDLLatency() {
-    return TargetSampleCount * 2;
-}
-
-// REVERTED back to 3833 - Below function IsSDLSuitable() removed, reverting to GetSDLLatency() above. - DIABLO 3 FIX
-/*
-bool IsSDLSuitable() {
-#if !defined(HAVE_SDL3)
-    return false;
-#else
-    // Check SDL can init
-    if (!InitializeAudio()!
-        return false;
-
-    // We can set any latency frequency we want with SDL, so no need to check that.
-
-    // Check we can open a device with standard parameters
-    SDL_AudioSpec spec;
-    spec.freq = TargetSampleRate;
-    spec.channels = 2u;
-    spec.format = AUDIO_S16SYS;
-    spec.samples = TargetSampleCount * 2;
-    spec.callback = nullptr;
-    spec.userdata = nullptr;
-
-    SDL_AudioSpec obtained;
-    auto device = SDL_OpenAudioDevice(nullptr, false, &spec, &obtained, false);
-
-    if (device == 0) {
-        LOG_ERROR(Audio_Sink, "SDL failed to open a device, it is not suitable. Error: {}",
-                  SDL_GetError());
-        return false;
-    }
-
-    SDL_CloseAudioDevice(device);
-    return true;
-#endif
-}
-*/
-
 } // namespace AudioCore::Sink
@@ -90,20 +90,4 @@ private:
 */
 std::vector<std::string> ListSDLSinkDevices(bool capture);

-// REVERSION - function GetSDLLatency() reintroduced from EA-3833 - DIABLO 3 FIX
-/**
- * Get the reported latency for this sink.
- *
- * @return Minimum latency for this sink.
- */
-u32 GetSDLLatency();
-
-/** REVERTED back to 3833 - Below function IsSDLSuitable() removed, reverting to GetSDLLatency() above. - DIABLO 3 FIX
- * Check if this backend is suitable for use.
- * Checks if enabled, its latency, whether it opens successfully, etc.
- *
- * @return True is this backend is suitable, false otherwise.
- */
-//bool IsSDLSuitable(); // REVERTED for GetSDLLatency() from EA-3833
-
 } // namespace AudioCore::Sink
@@ -10,9 +10,6 @@
 #include <vector>

 #include "audio_core/sink/sink_details.h"
-#ifdef HAVE_CUBEB
-#include "audio_core/sink/cubeb_sink.h"
-#endif
 #ifdef HAVE_SDL3
 #include "audio_core/sink/sdl3_sink.h"
 #endif
@@ -34,24 +31,10 @@ struct SinkDetails {
    FactoryFn factory;
    /// A method to call to list available devices.
    ListDevicesFn list_devices;
-    /// Method to get the latency of this backend - REINTRODUCED FROM 3833 - DIABLO 3 FIX
-    LatencyFn latency;
-    /// Check whether this backend is suitable to be used.
-    /// SuitableFn is_suitable; // REVERTED FOR LatencyFn latency ABOVE - DIABLO 3 FIX
 };

 // sink_details is ordered in terms of desirability, with the best choice at the top.
-constexpr SinkDetails sink_details[] = {
-#ifdef HAVE_CUBEB
-    SinkDetails{
-        Settings::AudioEngine::Cubeb,
-        [](std::string_view device_id) -> std::unique_ptr<Sink> {
-            return std::make_unique<CubebSink>(device_id);
-        },
-        &ListCubebSinkDevices,
-        &GetCubebLatency,
-    },
-#endif
+static constexpr SinkDetails sink_details[] = {
 #ifdef HAVE_SDL3
    SinkDetails{
        Settings::AudioEngine::Sdl3,
@@ -59,7 +42,6 @@ constexpr SinkDetails sink_details[] = {
            return std::make_unique<SDLSink>(device_id);
        },
        &ListSDLSinkDevices,
-        &GetSDLLatency,
    },
 #endif
    SinkDetails{
@@ -68,69 +50,32 @@ constexpr SinkDetails sink_details[] = {
            return std::make_unique<NullSink>(device_id);
        },
        [](bool capture) { return std::vector<std::string>{"null"}; },
-        []() { return 0u; },
    },
 };

 const SinkDetails& GetOutputSinkDetails(Settings::AudioEngine sink_id) {
    const auto find_backend{[](Settings::AudioEngine id) {
-        return std::find_if(std::begin(sink_details), std::end(sink_details),
-                            [&id](const auto& sink_detail) { return sink_detail.id == id; });
+        return std::ranges::find_if(std::begin(sink_details), std::end(sink_details), [&id](const auto& e) {
+            return e.id == id;
+        });
    }};

    auto iter = find_backend(sink_id);
-
-    if (sink_id == Settings::AudioEngine::Auto) {
-        // REVERTED TO 3833 BELOW - DIABLO 3 FIX
-        /*
-        // Auto-select a backend. Use the sink details ordering, preferring cubeb first, checking
-        // that the backend is available and suitable to use.
-        for (auto& details : sink_details) {
-            if (details.is_suitable()) {
-                iter = &details;
-                break;
-            }
-        }
-        */ // END REVERTED CODE - DIABLO 3 FIX
-
-        // BEGIN REINTRODUCED FROM 3833 - REPLACED CODE BLOCK ABOVE - DIABLO 3 FIX
-        // Auto-select a backend. Prefer CubeB, but it may report a large minimum latency which
-        // causes audio issues, in that case go with SDL.
-#if defined(HAVE_CUBEB) && defined(HAVE_SDL3)
-        iter = find_backend(Settings::AudioEngine::Cubeb);
-        if (iter->latency() > TargetSampleCount * 3) {
+    if (sink_id == Settings::AudioEngine::Cubeb || sink_id == Settings::AudioEngine::Auto) {
        iter = find_backend(Settings::AudioEngine::Sdl3);
-        }
-#else
-        iter = std::begin(sink_details);
-#endif
-        // END REINTRODUCED SECTION FROM 3833 - DIABLO 3 FIX
-        LOG_INFO(Service_Audio, "Auto-selecting the {} backend",
-                 Settings::CanonicalizeEnum(iter->id));
-    /* BEGIN REMOVED - REVERTING BACK TO 3833, this didn't exist at all. - DIABLO 3 FIX
-    } else {
-        if (iter != std::end(sink_details) && !iter->is_suitable()) {
-            LOG_ERROR(Service_Audio, "Selected backend {} is not suitable, falling back to null",
-                      Settings::CanonicalizeEnum(iter->id));
-            iter = find_backend(Settings::AudioEngine::Null);
-        } */ // END REMOVED REVERT - DIABLO 3 FIX
+        LOG_INFO(Service_Audio, "Auto-selecting the {} backend", Settings::CanonicalizeEnum(iter->id));
    }
-
    if (iter == std::end(sink_details)) {
        LOG_ERROR(Audio, "Invalid sink_id {}", Settings::CanonicalizeEnum(sink_id));
        iter = find_backend(Settings::AudioEngine::Null);
    }
-
    return *iter;
 }
 } // Anonymous namespace

 std::vector<Settings::AudioEngine> GetSinkIDs() {
    std::vector<Settings::AudioEngine> sink_ids(std::size(sink_details));
-
-    std::transform(std::begin(sink_details), std::end(sink_details), std::begin(sink_ids),
-                   [](const auto& sink) { return sink.id; });
-
+    std::transform(std::begin(sink_details), std::end(sink_details), std::begin(sink_ids), [](const auto& sink) { return sink.id; });
    return sink_ids;
 }

@@ -20,6 +20,8 @@ namespace VideoCommon {

 enum class BufferFlagBits {
    Picked = 1 << 0,
+    CachedWrites = 1 << 1,
+    PreemtiveDownload = 1 << 2,
 };
 DECLARE_ENUM_FLAG_OPERATORS(BufferFlagBits)

@@ -56,6 +58,15 @@ public:
        flags |= BufferFlagBits::Picked;
    }

+    void MarkPreemtiveDownload() noexcept {
+        flags |= BufferFlagBits::PreemtiveDownload;
+    }
+
+    /// Unmark buffer as picked
+    void Unpick() noexcept {
+        flags &= ~BufferFlagBits::Picked;
+    }
+
    /// Increases the likeliness of this being a stream buffer
    void IncreaseStreamScore(int score) noexcept {
        stream_score += score;
@@ -76,6 +87,15 @@ public:
        return True(flags & BufferFlagBits::Picked);
    }

+    /// Returns true when the buffer has pending cached writes
+    [[nodiscard]] bool HasCachedWrites() const noexcept {
+        return True(flags & BufferFlagBits::CachedWrites);
+    }
+
+    bool IsPreemtiveDownload() const noexcept {
+        return True(flags & BufferFlagBits::PreemtiveDownload);
+    }
+
    /// Returns the base CPU address of the buffer
    [[nodiscard]] VAddr CpuAddr() const noexcept {
        return cpu_addr;
@@ -7,7 +7,6 @@
 #pragma once

 #include <algorithm>
-#include <limits>
 #include <memory>
 #include <numeric>

@@ -122,6 +121,25 @@ void BufferCache<P>::WriteMemory(DAddr device_addr, u64 size) {
    memory_tracker.MarkRegionAsCpuModified(device_addr, size);
 }

+template <class P>
+void BufferCache<P>::CachedWriteMemory(DAddr device_addr, u64 size) {
+    const bool is_dirty = IsRegionRegistered(device_addr, size);
+    if (!is_dirty) {
+        return;
+    }
+    DAddr aligned_start = Common::AlignDown(device_addr, DEVICE_PAGESIZE);
+    DAddr aligned_end = Common::AlignUp(device_addr + size, DEVICE_PAGESIZE);
+    if (!IsRegionGpuModified(aligned_start, aligned_end - aligned_start)) {
+        WriteMemory(device_addr, size);
+        return;
+    }
+
+    tmp_buffer.resize_destructive(size);
+    device_memory.ReadBlockUnsafe(device_addr, tmp_buffer.data(), size);
+
+    InlineMemoryImplementation(device_addr, size, tmp_buffer);
+}
+
 template <class P>
 bool BufferCache<P>::OnCPUWrite(DAddr device_addr, u64 size) {
    const bool is_dirty = IsRegionRegistered(device_addr, size);
@@ -404,7 +422,7 @@ void BufferCache<P>::UnbindGraphicsStorageBuffers(size_t stage) {
 }

 template <class P>
-void BufferCache<P>::BindGraphicsStorageBuffer(size_t stage, size_t ssbo_index, u32 cbuf_index,
+bool BufferCache<P>::BindGraphicsStorageBuffer(size_t stage, size_t ssbo_index, u32 cbuf_index,
                                               u32 cbuf_offset, bool is_written) {
    const bool already_enabled =
        ((channel_state->enabled_storage_buffers[stage] >> ssbo_index) & 1U) != 0;
@@ -415,7 +433,7 @@ void BufferCache<P>::BindGraphicsStorageBuffer(size_t stage, size_t ssbo_index,
                LOG_WARNING(HW_GPU,
                            "Skipping graphics storage buffer {} due to driver limit {}",
                            ssbo_index, max_bindings);
-                return;
+                return false;
            }
        }
    }
@@ -431,6 +449,7 @@ void BufferCache<P>::BindGraphicsStorageBuffer(size_t stage, size_t ssbo_index,
    const GPUVAddr ssbo_addr = cbufs.const_buffers[cbuf_index].address + cbuf_offset;
    channel_state->storage_buffers[stage][ssbo_index] =
        StorageBufferBinding(ssbo_addr, cbuf_index, is_written);
+    return (channel_state->storage_buffers[stage][ssbo_index].buffer_id != NULL_BUFFER_ID);
 }

 template <class P>
@@ -743,6 +762,16 @@ void BufferCache<P>::BindHostIndexBuffer() {
    }
 }

+template <class P>
+void BufferCache<P>::BindHostVertexBuffer(u32 index, Buffer& buffer, u32 offset, u32 size,
+                                          u32 stride) {
+    if constexpr (IS_OPENGL) {
+        runtime.BindVertexBuffer(index, buffer, offset, size, stride);
+    } else {
+        runtime.BindVertexBuffer(index, buffer.Handle(), offset, size, stride);
+    }
+}
+
 template <class P>
 Binding& BufferCache<P>::VertexBufferSlot(u32 index) {
    ASSERT(index < NUM_VERTEX_BUFFERS);
@@ -1168,7 +1197,7 @@ void BufferCache<P>::DoUpdateGraphicsBuffers(bool is_indexed) {
        if (is_indexed) {
            UpdateIndexBuffer();
        }
-        UpdateVertexBuffers(is_indexed);
+        UpdateVertexBuffers();
        UpdateTransformFeedbackBuffers();
        for (size_t stage = 0; stage < NUM_STAGES; ++stage) {
            UpdateUniformBuffers(stage);
@@ -1222,14 +1251,9 @@ void BufferCache<P>::UpdateIndexBuffer() {
    const GPUVAddr gpu_addr_begin = index_buffer_ref.StartAddress();
    const GPUVAddr gpu_addr_end = index_buffer_ref.EndAddress();
    const std::optional<DAddr> device_addr = gpu_memory->GpuToCpuAddress(gpu_addr_begin);
-    u64 address_size = 0;
-    if (gpu_addr_end > gpu_addr_begin) {
-        address_size = (std::min)(gpu_addr_end - gpu_addr_begin,
-                                  u64{(std::numeric_limits<u32>::max)()});
-    }
-    const u64 draw_size = (u64{index_buffer_ref.count} + u64{index_buffer_ref.first}) *
-                          u64{index_buffer_ref.FormatSizeInBytes()};
-    const u32 size = static_cast<u32>((std::min)(address_size, draw_size));
+    const u32 address_size = static_cast<u32>(gpu_addr_end - gpu_addr_begin);
+    const u32 draw_size = (index_buffer_ref.count + index_buffer_ref.first) * u32(index_buffer_ref.FormatSizeInBytes());
+    const u32 size = (std::min)(address_size, draw_size);
    if (size == 0 || !device_addr) {
        channel_state->index_buffer = NULL_BINDING;
        return;
@@ -1242,142 +1266,20 @@ void BufferCache<P>::UpdateIndexBuffer() {
 }

 template <class P>
-u64 BufferCache<P>::DrawMaxIndex() {
-    if (max_index_scanned) {
-        return cached_max_index;
-    }
-    max_index_scanned = true;
-    cached_max_index = 0;
-    const auto& index_buffer_ref = maxwell3d->draw_manager.draw_state.index_buffer;
-    const u32 count = index_buffer_ref.count;
-    if (count == 0) {
-        return 0;
-    }
-    index_scan_buffer.resize_destructive(count);
-    gpu_memory->ReadBlockUnsafe(index_buffer_ref.IndexStart(), index_scan_buffer.data(),
-                                size_t{count} * sizeof(u32));
-    u32 restart_index = (std::numeric_limits<u32>::max)();
-    if (maxwell3d->regs.primitive_restart.enabled != 0) {
-        restart_index = maxwell3d->regs.primitive_restart.index;
-    }
-    u32 max_index = 0;
-    for (u32 i = 0; i < count; ++i) {
-        const u32 value = index_scan_buffer[i];
-        if (value == restart_index) {
-            continue;
-        }
-        max_index = (std::max)(max_index, value);
-    }
-    cached_max_index = max_index;
-    return cached_max_index;
-}
-
-template <class P>
-u64 BufferCache<P>::StreamAttributeExtent(u32 index) {
-    using VertexAttribute = typename Maxwell::VertexAttribute;
-    if (stream_extents_valid) {
-        return stream_extents[index];
-    }
-    stream_extents_valid = true;
-    stream_extents.fill(0);
-    for (size_t i = 0; i < Maxwell::NumVertexAttributes; ++i) {
-        const auto& attribute = maxwell3d->regs.vertex_attrib_format[i];
-        if (attribute.constant != 0 || attribute.size == VertexAttribute::Size::Invalid) {
-            continue;
-        }
-        const u32 buffer = attribute.buffer.Value();
-        if (buffer >= NUM_VERTEX_BUFFERS) {
-            continue;
-        }
-        const u64 end = static_cast<u64>(attribute.offset.Value()) +
-                        static_cast<u64>(attribute.SizeInBytes());
-        stream_extents[buffer] = (std::max)(stream_extents[buffer], end);
-    }
-    return stream_extents[index];
-}
-
-template <class P>
-u64 BufferCache<P>::DrawVertexBound(u32 index, bool is_indexed) {
-    const auto& array = maxwell3d->regs.vertex_streams[index];
-    if (array.enable == 0) {
-        return 0;
-    }
-    const u64 extent = StreamAttributeExtent(index);
-    if (extent == 0) {
-        return 0;
-    }
-    const u64 stride = static_cast<u64>(array.stride);
-    if (stride == 0) {
-        return extent;
-    }
-    const auto& draw_state = maxwell3d->draw_manager.draw_state;
-    u64 elements = 0;
-    if (maxwell3d->regs.vertex_stream_instances.IsInstancingEnabled(index)) {
-        if (draw_instance_count == 0) {
-            return 0;
-        }
-        const u64 base_instance = static_cast<u64>(draw_state.base_instance);
-        elements = base_instance + 1;
-        if (array.frequency != 0) {
-            elements = (base_instance + static_cast<u64>(draw_instance_count) - 1) /
-                           static_cast<u64>(array.frequency) +
-                       1;
-        }
-    } else if (!is_indexed) {
-        elements = static_cast<u64>(draw_state.vertex_buffer.first) +
-                   static_cast<u64>(draw_state.vertex_buffer.count);
-    } else {
-        const auto format = draw_state.index_buffer.format;
-        u64 max_index = 0xFF;
-        if (format == Maxwell::IndexFormat::UnsignedShort) {
-            max_index = 0xFFFF;
-        } else if (format != Maxwell::IndexFormat::UnsignedByte) {
-            const auto& limit = maxwell3d->regs.vertex_stream_limits[index];
-            const GPUVAddr gpu_addr_begin = array.Address();
-            const GPUVAddr gpu_addr_end = limit.Address() + 1;
-            if (gpu_addr_end <= gpu_addr_begin) {
-                return 0;
-            }
-            const bool walks = gpu_addr_end - gpu_addr_begin >= IMPLAUSIBLE_VERTEX_SIZE ||
-                               !gpu_memory->IsWithinGPUAddressRange(gpu_addr_end);
-            if (!walks) {
-                return 0;
-            }
-            max_index = DrawMaxIndex();
-        }
-        elements = static_cast<u64>(draw_state.base_index) + max_index + 1;
-    }
-    if (elements == 0) {
-        return extent;
-    }
-    return (elements - 1) * stride + extent;
-}
-
-template <class P>
-void BufferCache<P>::UpdateVertexBuffers(bool is_indexed) {
+void BufferCache<P>::UpdateVertexBuffers() {
    auto& flags = maxwell3d->dirty.flags;
-    max_index_scanned = false;
-    stream_extents_valid = false;
-    for (u32 index = 0; index < NUM_VERTEX_BUFFERS; ++index) {
-        const u64 bound = DrawVertexBound(index, is_indexed);
-        if (bound <= last_draw_bounds[index]) {
-            continue;
-        }
-        flags[Dirty::VertexBuffer0 + index] = true;
-        flags[Dirty::VertexBuffers] = true;
-    }
    if (!maxwell3d->dirty.flags[Dirty::VertexBuffers]) {
        return;
    }
    flags[Dirty::VertexBuffers] = false;

    for (u32 index = 0; index < NUM_VERTEX_BUFFERS; ++index) {
-        UpdateVertexBuffer(index, is_indexed);
+        UpdateVertexBuffer(index);
    }
 }

 template <class P>
-void BufferCache<P>::UpdateVertexBuffer(u32 index, bool is_indexed) {
+void BufferCache<P>::UpdateVertexBuffer(u32 index) {
    if (!maxwell3d->dirty.flags[Dirty::VertexBuffer0 + index]) {
        return;
    }
@@ -1386,31 +1288,15 @@ void BufferCache<P>::UpdateVertexBuffer(u32 index, bool is_indexed) {
    const GPUVAddr gpu_addr_begin = array.Address();
    const GPUVAddr gpu_addr_end = limit.Address() + 1;
    const std::optional<DAddr> device_addr = gpu_memory->GpuToCpuAddress(gpu_addr_begin);
-    if (array.enable == 0 || !device_addr || gpu_addr_end <= gpu_addr_begin) {
+    const u32 address_size = static_cast<u32>(gpu_addr_end - gpu_addr_begin);
+    u32 size = address_size; // TODO: Analyze stride and number of vertices
+    if (array.enable == 0 || size == 0 || !device_addr) {
        channel_state->vertex_buffers[index] = NULL_BINDING;
        UpdateVertexBufferSlot(index, NULL_BINDING);
        return;
    }
-    // TODO: Analyze stride and number of vertices
-    constexpr u64 implausible_size = IMPLAUSIBLE_VERTEX_SIZE;
-    u64 address_size = gpu_addr_end - gpu_addr_begin;
-    if (address_size > u64{(std::numeric_limits<u32>::max)()}) {
-        address_size = implausible_size;
-    }
-    const u64 draw_bound = DrawVertexBound(index, is_indexed);
-    last_draw_bounds[index] = (std::numeric_limits<u64>::max)();
-    if (draw_bound != 0) {
-        last_draw_bounds[index] = draw_bound;
-        address_size = (std::min)(address_size, draw_bound);
-    }
-    if (!gpu_memory->IsWithinGPUAddressRange(gpu_addr_end) || address_size >= implausible_size) {
-        address_size = gpu_memory->MaxContinuousRange(gpu_addr_begin, address_size);
-    }
-    const u32 size = static_cast<u32>(address_size);
-    if (size == 0) {
-        channel_state->vertex_buffers[index] = NULL_BINDING;
-        UpdateVertexBufferSlot(index, NULL_BINDING);
-        return;
+    if (!gpu_memory->IsWithinGPUAddressRange(gpu_addr_end) || size >= 64_MiB) {
+        size = static_cast<u32>(gpu_memory->MaxContinuousRange(gpu_addr_begin, size));
    }
    const BufferId buffer_id = FindBuffer(*device_addr, size);
    const Binding binding{
@@ -1692,10 +1578,9 @@ template <class P>
 BufferId BufferCache<P>::CreateBuffer(DAddr device_addr, u32 wanted_size) {
    DAddr device_addr_end = Common::AlignUp(device_addr + wanted_size, CACHING_PAGESIZE);
    device_addr = Common::AlignDown(device_addr, CACHING_PAGESIZE);
-    constexpr u64 max_buffer_size = u64{(std::numeric_limits<u32>::max)()};
-    wanted_size = static_cast<u32>((std::min)(device_addr_end - device_addr, max_buffer_size));
+    wanted_size = static_cast<u32>(device_addr_end - device_addr);
    const OverlapResult overlap = ResolveOverlaps(device_addr, wanted_size);
-    const u32 size = static_cast<u32>((std::min)(overlap.end - overlap.begin, max_buffer_size));
+    const u32 size = static_cast<u32>(overlap.end - overlap.begin);
    const BufferId new_buffer_id = slot_buffers.insert(runtime, overlap.begin, size);
    auto& new_buffer = slot_buffers[new_buffer_id];
    const size_t size_bytes = new_buffer.SizeBytes();
@@ -51,7 +51,6 @@ constexpr u32 NUM_VERTEX_BUFFERS = 16;
 #else
 constexpr u32 NUM_VERTEX_BUFFERS = 32;
 #endif
-constexpr u64 IMPLAUSIBLE_VERTEX_SIZE = 64_MiB;
 constexpr u32 NUM_TRANSFORM_FEEDBACK_BUFFERS = 4;
 constexpr u32 NUM_GRAPHICS_UNIFORM_BUFFERS = 18;
 constexpr u32 NUM_COMPUTE_UNIFORM_BUFFERS = 8;
@@ -218,6 +217,8 @@ public:

    void WriteMemory(DAddr device_addr, u64 size);

+    void CachedWriteMemory(DAddr device_addr, u64 size);
+
    bool OnCPUWrite(DAddr device_addr, u64 size);

    void DownloadMemory(DAddr device_addr, u64 size);
@@ -247,7 +248,7 @@ public:

    void UnbindGraphicsStorageBuffers(size_t stage);

-    void BindGraphicsStorageBuffer(size_t stage, size_t ssbo_index, u32 cbuf_index, u32 cbuf_offset,
+    bool BindGraphicsStorageBuffer(size_t stage, size_t ssbo_index, u32 cbuf_index, u32 cbuf_offset,
                                   bool is_written);

    void UnbindGraphicsTextureBuffers(size_t stage);
@@ -308,10 +309,6 @@ public:
        current_draw_indirect = current_draw_indirect_;
    }

-    void SetDrawInstanceCount(u32 draw_instance_count_) {
-        draw_instance_count = draw_instance_count_;
-    }
-
    [[nodiscard]] std::pair<Buffer*, u32> GetDrawIndirectCount();

    [[nodiscard]] std::pair<Buffer*, u32> GetDrawIndirectBuffer();
@@ -379,6 +376,8 @@ private:

    void BindHostTransformFeedbackBuffers();

+    void BindHostVertexBuffer(u32 index, Buffer& buffer, u32 offset, u32 size, u32 stride);
+
    void BindHostComputeUniformBuffers();

    void BindHostComputeStorageBuffers();
@@ -391,15 +390,9 @@ private:

    void UpdateIndexBuffer();

-    void UpdateVertexBuffers(bool is_indexed);
+    void UpdateVertexBuffers();

-    void UpdateVertexBuffer(u32 index, bool is_indexed);
-
-    [[nodiscard]] u64 DrawVertexBound(u32 index, bool is_indexed);
-
-    [[nodiscard]] u64 DrawMaxIndex();
-
-    [[nodiscard]] u64 StreamAttributeExtent(u32 index);
+    void UpdateVertexBuffer(u32 index);

    void UpdateDrawIndirect();

@@ -491,14 +484,6 @@ private:

    const Tegra::Engines::Maxwell3D::DrawManager::IndirectParams* current_draw_indirect{};

-    u32 draw_instance_count = 0;
-    std::array<u64, NUM_VERTEX_BUFFERS> last_draw_bounds{};
-    Common::ScratchBuffer<u32> index_scan_buffer;
-    u64 cached_max_index = 0;
-    bool max_index_scanned = false;
-    std::array<u64, NUM_VERTEX_BUFFERS> stream_extents{};
-    bool stream_extents_valid = false;
-
    u32 last_index_count = 0;

    u32 enabled_vertex_buffers_mask = 0;
@@ -206,40 +206,6 @@ foreach(VARIANT IN ITEMS ${SHADER_TYPE_VARIANTS})
    set(SHADER_HEADERS ${SHADER_HEADERS} ${VARIANT_HEADER_FILE})
 endforeach()

-set(SHADER_DEFINE_VARIANTS
-    "block_linear_unswizzle_2d.comp|nonarrow|HAS_EXTENDED_TYPES=0"
-    "pitch_unswizzle.comp|nonarrow|HAS_EXTENDED_TYPES=0"
-    "block_linear_unswizzle_3d.comp|nonarrow|HAS_EXTENDED_TYPES=0"
-)
-
-foreach(VARIANT IN ITEMS ${SHADER_DEFINE_VARIANTS})
-    string(REPLACE "|" ";" VARIANT_PARTS ${VARIANT})
-    list(GET VARIANT_PARTS 0 VARIANT_FILENAME)
-    list(GET VARIANT_PARTS 1 VARIANT_SUFFIX)
-    list(GET VARIANT_PARTS 2 VARIANT_DEFINE)
-
-    set(VARIANT_SOURCE ${CMAKE_CURRENT_SOURCE_DIR}/${VARIANT_FILENAME})
-    get_filename_component(VARIANT_STEM ${VARIANT_FILENAME} NAME_WE)
-    get_filename_component(VARIANT_EXT ${VARIANT_FILENAME} EXT)
-    string(REPLACE "." "" VARIANT_EXT ${VARIANT_EXT})
-    set(VARIANT_NAME ${VARIANT_STEM}_${VARIANT_SUFFIX}_${VARIANT_EXT})
-
-    string(TOUPPER ${VARIANT_NAME}_SPV VARIANT_VARIABLE_NAME)
-    set(VARIANT_HEADER_FILE ${SHADER_DIR}/${VARIANT_NAME}_spv.h)
-    add_custom_command(
-        OUTPUT
-            ${VARIANT_HEADER_FILE}
-        COMMAND
-            ${GLSLANGVALIDATOR} -V ${QUIET_FLAG} -I"${FIDELITYFX_INCLUDE_DIR}" ${GLSL_FLAGS}
-            -D${VARIANT_DEFINE}
-            --variable-name ${VARIANT_VARIABLE_NAME} -o ${VARIANT_HEADER_FILE} ${VARIANT_SOURCE}
-            --target-env ${SPIR_V_VERSION}
-        MAIN_DEPENDENCY
-            ${VARIANT_SOURCE}
-    )
-    set(SHADER_HEADERS ${SHADER_HEADERS} ${VARIANT_HEADER_FILE})
-endforeach()
-
 foreach(FILEPATH IN ITEMS ${FIDELITYFX_FILES})
    get_filename_component(FILENAME ${FILEPATH} NAME)
    string(REPLACE "." "_" HEADER_NAME ${FILENAME})
@@ -5,13 +5,9 @@

 #ifdef VULKAN

-#ifndef HAS_EXTENDED_TYPES
-#define HAS_EXTENDED_TYPES 1
-#endif
-#if HAS_EXTENDED_TYPES
 #extension GL_EXT_shader_16bit_storage : require
 #extension GL_EXT_shader_8bit_storage : require
-#endif
+#define HAS_EXTENDED_TYPES 1
 #define BEGIN_PUSH_CONSTANTS layout(push_constant) uniform PushConstants {
 #define END_PUSH_CONSTANTS };
 #define UNIFORM(n)
@@ -5,13 +5,9 @@

 #ifdef VULKAN

-#ifndef HAS_EXTENDED_TYPES
-#define HAS_EXTENDED_TYPES 1
-#endif
-#if HAS_EXTENDED_TYPES
 #extension GL_EXT_shader_16bit_storage : require
 #extension GL_EXT_shader_8bit_storage : require
-#endif
+#define HAS_EXTENDED_TYPES 1
 #define BEGIN_PUSH_CONSTANTS layout(push_constant) uniform PushConstants {
 #define END_PUSH_CONSTANTS };
 #define UNIFORM(n)
@@ -5,13 +5,9 @@

 #ifdef VULKAN

-#ifndef HAS_EXTENDED_TYPES
-#define HAS_EXTENDED_TYPES 1
-#endif
-#if HAS_EXTENDED_TYPES
 #extension GL_EXT_shader_16bit_storage : require
 #extension GL_EXT_shader_8bit_storage : require
-#endif
+#define HAS_EXTENDED_TYPES 1
 #define BEGIN_PUSH_CONSTANTS layout(push_constant) uniform PushConstants {
 #define END_PUSH_CONSTANTS };
 #define UNIFORM(n)
@@ -226,6 +226,22 @@ void BufferCacheRuntime::BindIndexBuffer(Buffer& buffer, u32 offset, u32 size) {
    }
 }

+void BufferCacheRuntime::BindVertexBuffer(u32 index, Buffer& buffer, u32 offset, u32 size,
+                                          u32 stride) {
+    if (index >= max_attributes) {
+        return;
+    }
+    if (has_unified_vertex_buffers) {
+        buffer.MakeResident(GL_READ_ONLY);
+        glBindVertexBuffer(index, 0, 0, static_cast<GLsizei>(stride));
+        glBufferAddressRangeNV(GL_VERTEX_ATTRIB_ARRAY_ADDRESS_NV, index,
+                               buffer.HostGpuAddr() + offset, static_cast<GLsizeiptr>(size));
+    } else {
+        glBindVertexBuffer(index, buffer.Handle(), static_cast<GLintptr>(offset),
+                           static_cast<GLsizei>(stride));
+    }
+}
+
 void BufferCacheRuntime::BindVertexBuffers(VideoCommon::HostBindings<Buffer>& bindings) {
    // TODO: Should HostBindings provide the correct runtime types to avoid these transforms?
    std::array<GLuint, 32> buffer_handles;
@@ -99,6 +99,8 @@ public:

    void BindIndexBuffer(Buffer& buffer, u32 offset, u32 size);

+    void BindVertexBuffer(u32 index, Buffer& buffer, u32 offset, u32 size, u32 stride);
+
    void BindVertexBuffers(VideoCommon::HostBindings<Buffer>& bindings);

    void BindUniformBuffer(size_t stage, u32 binding_index, Buffer& buffer, u32 offset, u32 size);
@@ -259,7 +259,6 @@ void RasterizerOpenGL::PrepareDraw(bool is_indexed, Func&& draw_func) {
 }

 void RasterizerOpenGL::Draw(bool is_indexed, u32 instance_count) {
-    buffer_cache.SetDrawInstanceCount(instance_count);
    PrepareDraw(is_indexed, [this, is_indexed, instance_count](GLenum primitive_mode) {
        const auto& draw_state = maxwell3d->draw_manager.draw_state;
        const GLuint base_instance = GLuint(draw_state.base_instance);
@@ -305,7 +304,6 @@ void RasterizerOpenGL::Draw(bool is_indexed, u32 instance_count) {
 void RasterizerOpenGL::DrawIndirect() {
    const auto& params = maxwell3d->draw_manager.indirect_state;
    buffer_cache.SetDrawIndirect(&params);
-    buffer_cache.SetDrawInstanceCount(0);
    PrepareDraw(params.is_indexed, [this, &params](GLenum primitive_mode) {
        if (params.is_byte_count) {
            const GPUVAddr tfb_object_base_addr = params.indirect_start_address - 4U;
@@ -585,6 +585,29 @@ void BufferCacheRuntime::BindQuadIndexBuffer(PrimitiveTopology topology, u32 fir
    }
 }

+void BufferCacheRuntime::BindVertexBuffer(u32 index, VkBuffer buffer, u32 offset, u32 size, u32 stride) {
+    if (index >= device.GetMaxVertexInputBindings()) {
+        return;
+    }
+    if (device.IsExtExtendedDynamicStateSupported()) {
+        scheduler.Record([index, buffer, offset, size, stride](vk::CommandBuffer cmdbuf) {
+            const VkDeviceSize vk_offset = buffer != VK_NULL_HANDLE ? offset : 0;
+            const VkDeviceSize vk_size = buffer != VK_NULL_HANDLE ? size : VK_WHOLE_SIZE;
+            const VkDeviceSize vk_stride = stride;
+            cmdbuf.BindVertexBuffers2EXT(index, 1, &buffer, &vk_offset, &vk_size, &vk_stride);
+        });
+    } else {
+        if (!device.HasNullDescriptor() && buffer == VK_NULL_HANDLE) {
+            ReserveNullBuffer();
+            buffer = *null_buffer;
+            offset = 0;
+        }
+        scheduler.Record([index, buffer, offset](vk::CommandBuffer cmdbuf) {
+            cmdbuf.BindVertexBuffer(index, buffer, offset);
+        });
+    }
+}
+
 void BufferCacheRuntime::BindVertexBuffers(VideoCommon::HostBindings<Buffer>& bindings) {
    boost::container::static_vector<VkBuffer, VideoCommon::NUM_VERTEX_BUFFERS> buffer_handles(bindings.buffers.size());
    for (u32 i = 0; i < bindings.buffers.size(); ++i) {
@@ -138,6 +138,8 @@ public:

    void BindQuadIndexBuffer(PrimitiveTopology topology, u32 first, u32 count);

+    void BindVertexBuffer(u32 index, VkBuffer buffer, u32 offset, u32 size, u32 stride);
+
    void BindVertexBuffers(VideoCommon::HostBindings<Buffer>& bindings);

    void BindTransformFeedbackBuffer(u32 index, VkBuffer buffer, u32 offset, u32 size);
@@ -22,13 +22,7 @@
 #include "video_core/host_shaders/resolve_conditional_render_comp_spv.h"
 #include "video_core/host_shaders/vulkan_quad_indexed_comp_spv.h"
 #include "video_core/host_shaders/vulkan_uint8_comp_spv.h"
-#include "video_core/host_shaders/block_linear_unswizzle_2d_comp_spv.h"
-#include "video_core/host_shaders/block_linear_unswizzle_2d_nonarrow_comp_spv.h"
 #include "video_core/host_shaders/block_linear_unswizzle_3d_bcn_comp_spv.h"
-#include "video_core/host_shaders/block_linear_unswizzle_3d_comp_spv.h"
-#include "video_core/host_shaders/block_linear_unswizzle_3d_nonarrow_comp_spv.h"
-#include "video_core/host_shaders/pitch_unswizzle_comp_spv.h"
-#include "video_core/host_shaders/pitch_unswizzle_nonarrow_comp_spv.h"
 #include "video_core/renderer_vulkan/vk_compute_pass.h"
 #include "video_core/surface.h"
 #include "video_core/renderer_vulkan/vk_descriptor_pool.h"
@@ -878,291 +872,4 @@ void BlockLinearUnswizzle3DPass::UnswizzleChunk(
    });
 }

-namespace {
-
-constexpr u32 UNSWIZZLE_BINDING_INPUT_BUFFER = 0;
-constexpr u32 UNSWIZZLE_BINDING_OUTPUT_IMAGE = 1;
-constexpr size_t UNSWIZZLE_NUM_BINDINGS = 2;
-
-constexpr std::array<VkDescriptorSetLayoutBinding, UNSWIZZLE_NUM_BINDINGS>
-    UNSWIZZLE_DESCRIPTOR_SET_BINDINGS{{
-        {
-            .binding = UNSWIZZLE_BINDING_INPUT_BUFFER,
-            .descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER,
-            .descriptorCount = 1,
-            .stageFlags = VK_SHADER_STAGE_COMPUTE_BIT,
-            .pImmutableSamplers = nullptr,
-        },
-        {
-            .binding = UNSWIZZLE_BINDING_OUTPUT_IMAGE,
-            .descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_IMAGE,
-            .descriptorCount = 1,
-            .stageFlags = VK_SHADER_STAGE_COMPUTE_BIT,
-            .pImmutableSamplers = nullptr,
-        },
-    }};
-
-constexpr std::array<VkDescriptorUpdateTemplateEntry, UNSWIZZLE_NUM_BINDINGS>
-    UNSWIZZLE_DESCRIPTOR_UPDATE_TEMPLATE{{
-        {
-            .dstBinding = UNSWIZZLE_BINDING_INPUT_BUFFER,
-            .dstArrayElement = 0,
-            .descriptorCount = 1,
-            .descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER,
-            .offset = UNSWIZZLE_BINDING_INPUT_BUFFER * sizeof(DescriptorUpdateEntry),
-            .stride = sizeof(DescriptorUpdateEntry),
-        },
-        {
-            .dstBinding = UNSWIZZLE_BINDING_OUTPUT_IMAGE,
-            .dstArrayElement = 0,
-            .descriptorCount = 1,
-            .descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_IMAGE,
-            .offset = UNSWIZZLE_BINDING_OUTPUT_IMAGE * sizeof(DescriptorUpdateEntry),
-            .stride = sizeof(DescriptorUpdateEntry),
-        },
-    }};
-
-constexpr DescriptorBankInfo UNSWIZZLE_BANK_INFO{
-    .uniform_buffers = 0,
-    .storage_buffers = 1,
-    .texture_buffers = 0,
-    .image_buffers = 0,
-    .textures = 0,
-    .images = 1,
-    .score = 2,
-};
-
-[[nodiscard]] std::span<const u32> UnswizzleSpv(const Device& device,
-                                                std::span<const u32> extended,
-                                                std::span<const u32> narrow) {
-    if (device.IsStorageBuffer8BitAccessSupported() &&
-        device.IsStorageBuffer16BitAccessSupported()) {
-        return extended;
-    }
-    return narrow;
-}
-
-struct PitchUnswizzlePushConstants {
-    alignas(8) std::array<u32, 2> origin;
-    alignas(8) std::array<s32, 2> destination;
-    u32 bytes_per_block;
-    u32 pitch;
-};
-
-void RecordUnswizzleEntryBarrier(Scheduler& scheduler, VkPipeline vk_pipeline, VkImage vk_image,
-                                 VkImageAspectFlags aspect_mask, bool is_initialized) {
-    scheduler.Record([vk_pipeline, vk_image, aspect_mask,
-                      is_initialized](vk::CommandBuffer cmdbuf) {
-        VkAccessFlags src_access = VK_ACCESS_NONE;
-        VkImageLayout old_layout = VK_IMAGE_LAYOUT_UNDEFINED;
-        if (is_initialized) {
-            src_access = VK_ACCESS_SHADER_WRITE_BIT | VK_ACCESS_TRANSFER_WRITE_BIT |
-                         VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT;
-            old_layout = VK_IMAGE_LAYOUT_GENERAL;
-        }
-        const VkImageMemoryBarrier image_barrier{
-            .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
-            .pNext = nullptr,
-            .srcAccessMask = src_access,
-            .dstAccessMask = VK_ACCESS_SHADER_WRITE_BIT,
-            .oldLayout = old_layout,
-            .newLayout = VK_IMAGE_LAYOUT_GENERAL,
-            .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
-            .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
-            .image = vk_image,
-            .subresourceRange{
-                .aspectMask = aspect_mask,
-                .baseMipLevel = 0,
-                .levelCount = VK_REMAINING_MIP_LEVELS,
-                .baseArrayLayer = 0,
-                .layerCount = VK_REMAINING_ARRAY_LAYERS,
-            },
-        };
-        VkPipelineStageFlags src_stage = VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT;
-        if (is_initialized) {
-            src_stage = vk::PIPELINE_STAGE_GRAPHICS_COMPUTE_TRANSFER;
-        }
-        cmdbuf.PipelineBarrier(src_stage, VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT, 0, image_barrier);
-        cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_COMPUTE, vk_pipeline);
-    });
-}
-
-void RecordUnswizzleExitBarrier(Scheduler& scheduler, VkImage vk_image,
-                                VkImageAspectFlags aspect_mask) {
-    scheduler.Record([vk_image, aspect_mask](vk::CommandBuffer cmdbuf) {
-        const VkImageMemoryBarrier image_barrier{
-            .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
-            .pNext = nullptr,
-            .srcAccessMask = VK_ACCESS_SHADER_WRITE_BIT,
-            .dstAccessMask = VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_TRANSFER_READ_BIT |
-                             VK_ACCESS_COLOR_ATTACHMENT_READ_BIT,
-            .oldLayout = VK_IMAGE_LAYOUT_GENERAL,
-            .newLayout = VK_IMAGE_LAYOUT_GENERAL,
-            .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
-            .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
-            .image = vk_image,
-            .subresourceRange{
-                .aspectMask = aspect_mask,
-                .baseMipLevel = 0,
-                .levelCount = VK_REMAINING_MIP_LEVELS,
-                .baseArrayLayer = 0,
-                .layerCount = VK_REMAINING_ARRAY_LAYERS,
-            },
-        };
-        cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT,
-                               vk::PIPELINE_STAGE_GRAPHICS_COMPUTE_TRANSFER, 0, image_barrier);
-    });
-}
-
-} // Anonymous namespace
-
-BlockLinearUnswizzle2DPass::BlockLinearUnswizzle2DPass(
-    const Device& device_, Scheduler& scheduler_, DescriptorPool& descriptor_pool_,
-    ComputePassDescriptorQueue& compute_pass_descriptor_queue_)
-    : ComputePass(device_, scheduler_, descriptor_pool_, UNSWIZZLE_DESCRIPTOR_SET_BINDINGS,
-                  UNSWIZZLE_DESCRIPTOR_UPDATE_TEMPLATE, UNSWIZZLE_BANK_INFO,
-                  COMPUTE_PUSH_CONSTANT_RANGE<sizeof(
-                      VideoCommon::Accelerated::BlockLinearSwizzle2DParams)>,
-                  UnswizzleSpv(device_, BLOCK_LINEAR_UNSWIZZLE_2D_COMP_SPV,
-                               BLOCK_LINEAR_UNSWIZZLE_2D_NONARROW_COMP_SPV)),
-      scheduler{scheduler_}, compute_pass_descriptor_queue{compute_pass_descriptor_queue_} {}
-
-BlockLinearUnswizzle2DPass::~BlockLinearUnswizzle2DPass() = default;
-
-void BlockLinearUnswizzle2DPass::Unswizzle(
-    Image& image, const StagingBufferRef& map,
-    std::span<const VideoCommon::SwizzleParameters> swizzles) {
-    using namespace VideoCommon::Accelerated;
-    scheduler.RequestOutsideRenderPassOperationContext();
-    const VkPipeline vk_pipeline = *pipeline;
-    const VkImageAspectFlags aspect_mask = image.AspectMask();
-    const VkImage vk_image = image.Handle();
-    const bool is_initialized = image.ExchangeInitialization();
-    RecordUnswizzleEntryBarrier(scheduler, vk_pipeline, vk_image, aspect_mask, is_initialized);
-
-    const u32 num_layers = static_cast<u32>(image.info.resources.layers);
-    for (const VideoCommon::SwizzleParameters& swizzle : swizzles) {
-        const size_t input_offset = swizzle.buffer_offset + map.offset;
-        const u32 num_dispatches_x = Common::DivCeil(swizzle.num_tiles.width, 32U);
-        const u32 num_dispatches_y = Common::DivCeil(swizzle.num_tiles.height, 32U);
-
-        compute_pass_descriptor_queue.Acquire(scheduler, 2);
-        compute_pass_descriptor_queue.AddBuffer(map.buffer, input_offset,
-                                                image.guest_size_bytes - swizzle.buffer_offset);
-        compute_pass_descriptor_queue.AddImage(image.StorageImageView(swizzle.level));
-        const void* const descriptor_data{compute_pass_descriptor_queue.UpdateData()};
-
-        const auto params = MakeBlockLinearSwizzle2DParams(swizzle, image.info);
-        scheduler.Record([this, num_dispatches_x, num_dispatches_y, num_layers, params,
-                          descriptor_data](vk::CommandBuffer cmdbuf) {
-            const VkDescriptorSet set = descriptor_allocator.Commit();
-            device.GetLogical().UpdateDescriptorSet(set, *descriptor_template, descriptor_data);
-            cmdbuf.BindDescriptorSets(VK_PIPELINE_BIND_POINT_COMPUTE, *layout, 0, set, {});
-            cmdbuf.PushConstants(*layout, VK_SHADER_STAGE_COMPUTE_BIT, params);
-            cmdbuf.Dispatch(num_dispatches_x, num_dispatches_y, num_layers);
-        });
-    }
-    RecordUnswizzleExitBarrier(scheduler, vk_image, aspect_mask);
-}
-
-BlockLinearUnswizzleImage3DPass::BlockLinearUnswizzleImage3DPass(
-    const Device& device_, Scheduler& scheduler_, DescriptorPool& descriptor_pool_,
-    ComputePassDescriptorQueue& compute_pass_descriptor_queue_)
-    : ComputePass(device_, scheduler_, descriptor_pool_, UNSWIZZLE_DESCRIPTOR_SET_BINDINGS,
-                  UNSWIZZLE_DESCRIPTOR_UPDATE_TEMPLATE, UNSWIZZLE_BANK_INFO,
-                  COMPUTE_PUSH_CONSTANT_RANGE<sizeof(BlockLinearSwizzle3DParams)>,
-                  UnswizzleSpv(device_, BLOCK_LINEAR_UNSWIZZLE_3D_COMP_SPV,
-                               BLOCK_LINEAR_UNSWIZZLE_3D_NONARROW_COMP_SPV)),
-      scheduler{scheduler_}, compute_pass_descriptor_queue{compute_pass_descriptor_queue_} {}
-
-BlockLinearUnswizzleImage3DPass::~BlockLinearUnswizzleImage3DPass() = default;
-
-void BlockLinearUnswizzleImage3DPass::Unswizzle(
-    Image& image, const StagingBufferRef& map,
-    std::span<const VideoCommon::SwizzleParameters> swizzles) {
-    using namespace VideoCommon::Accelerated;
-    scheduler.RequestOutsideRenderPassOperationContext();
-    const VkPipeline vk_pipeline = *pipeline;
-    const VkImageAspectFlags aspect_mask = image.AspectMask();
-    const VkImage vk_image = image.Handle();
-    const bool is_initialized = image.ExchangeInitialization();
-    RecordUnswizzleEntryBarrier(scheduler, vk_pipeline, vk_image, aspect_mask, is_initialized);
-
-    for (const VideoCommon::SwizzleParameters& swizzle : swizzles) {
-        const size_t input_offset = swizzle.buffer_offset + map.offset;
-        const u32 num_dispatches_x = Common::DivCeil(swizzle.num_tiles.width, 16U);
-        const u32 num_dispatches_y = Common::DivCeil(swizzle.num_tiles.height, 8U);
-        const u32 num_dispatches_z = Common::DivCeil(swizzle.num_tiles.depth, 8U);
-
-        compute_pass_descriptor_queue.Acquire(scheduler, 2);
-        compute_pass_descriptor_queue.AddBuffer(map.buffer, input_offset,
-                                                image.guest_size_bytes - swizzle.buffer_offset);
-        compute_pass_descriptor_queue.AddImage(image.StorageImageView(swizzle.level));
-        const void* const descriptor_data{compute_pass_descriptor_queue.UpdateData()};
-
-        const auto params = MakeBlockLinearSwizzle3DParams(swizzle, image.info);
-        scheduler.Record([this, num_dispatches_x, num_dispatches_y, num_dispatches_z, params,
-                          descriptor_data](vk::CommandBuffer cmdbuf) {
-            const VkDescriptorSet set = descriptor_allocator.Commit();
-            device.GetLogical().UpdateDescriptorSet(set, *descriptor_template, descriptor_data);
-            cmdbuf.BindDescriptorSets(VK_PIPELINE_BIND_POINT_COMPUTE, *layout, 0, set, {});
-            cmdbuf.PushConstants(*layout, VK_SHADER_STAGE_COMPUTE_BIT, params);
-            cmdbuf.Dispatch(num_dispatches_x, num_dispatches_y, num_dispatches_z);
-        });
-    }
-    RecordUnswizzleExitBarrier(scheduler, vk_image, aspect_mask);
-}
-
-PitchUnswizzlePass::PitchUnswizzlePass(
-    const Device& device_, Scheduler& scheduler_, DescriptorPool& descriptor_pool_,
-    ComputePassDescriptorQueue& compute_pass_descriptor_queue_)
-    : ComputePass(device_, scheduler_, descriptor_pool_, UNSWIZZLE_DESCRIPTOR_SET_BINDINGS,
-                  UNSWIZZLE_DESCRIPTOR_UPDATE_TEMPLATE, UNSWIZZLE_BANK_INFO,
-                  COMPUTE_PUSH_CONSTANT_RANGE<sizeof(PitchUnswizzlePushConstants)>,
-                  UnswizzleSpv(device_, PITCH_UNSWIZZLE_COMP_SPV,
-                               PITCH_UNSWIZZLE_NONARROW_COMP_SPV)),
-      scheduler{scheduler_}, compute_pass_descriptor_queue{compute_pass_descriptor_queue_} {}
-
-PitchUnswizzlePass::~PitchUnswizzlePass() = default;
-
-void PitchUnswizzlePass::Unswizzle(Image& image, const StagingBufferRef& map,
-                                   std::span<const VideoCommon::SwizzleParameters> swizzles) {
-    scheduler.RequestOutsideRenderPassOperationContext();
-    const VkPipeline vk_pipeline = *pipeline;
-    const VkImageAspectFlags aspect_mask = image.AspectMask();
-    const VkImage vk_image = image.Handle();
-    const bool is_initialized = image.ExchangeInitialization();
-    RecordUnswizzleEntryBarrier(scheduler, vk_pipeline, vk_image, aspect_mask, is_initialized);
-
-    const u32 bytes_per_block = VideoCore::Surface::BytesPerBlock(image.info.format);
-    const u32 pitch = image.info.pitch;
-    for (const VideoCommon::SwizzleParameters& swizzle : swizzles) {
-        const size_t input_offset = swizzle.buffer_offset + map.offset;
-        const u32 num_dispatches_x = Common::DivCeil(swizzle.num_tiles.width, 32U);
-        const u32 num_dispatches_y = Common::DivCeil(swizzle.num_tiles.height, 32U);
-
-        compute_pass_descriptor_queue.Acquire(scheduler, 2);
-        compute_pass_descriptor_queue.AddBuffer(map.buffer, input_offset,
-                                                image.guest_size_bytes - swizzle.buffer_offset);
-        compute_pass_descriptor_queue.AddImage(image.StorageImageView(swizzle.level));
-        const void* const descriptor_data{compute_pass_descriptor_queue.UpdateData()};
-
-        const PitchUnswizzlePushConstants params{
-            .origin{0, 0},
-            .destination{0, 0},
-            .bytes_per_block = bytes_per_block,
-            .pitch = pitch,
-        };
-        scheduler.Record([this, num_dispatches_x, num_dispatches_y, params,
-                          descriptor_data](vk::CommandBuffer cmdbuf) {
-            const VkDescriptorSet set = descriptor_allocator.Commit();
-            device.GetLogical().UpdateDescriptorSet(set, *descriptor_template, descriptor_data);
-            cmdbuf.BindDescriptorSets(VK_PIPELINE_BIND_POINT_COMPUTE, *layout, 0, set, {});
-            cmdbuf.PushConstants(*layout, VK_SHADER_STAGE_COMPUTE_BIT, params);
-            cmdbuf.Dispatch(num_dispatches_x, num_dispatches_y, 1);
-        });
-    }
-    RecordUnswizzleExitBarrier(scheduler, vk_image, aspect_mask);
-}
-
 } // namespace Vulkan
@@ -164,49 +164,4 @@ private:
    ComputePassDescriptorQueue& compute_pass_descriptor_queue;
 };

-class BlockLinearUnswizzle2DPass final : public ComputePass {
-public:
-    explicit BlockLinearUnswizzle2DPass(
-        const Device& device_, Scheduler& scheduler_, DescriptorPool& descriptor_pool_,
-        ComputePassDescriptorQueue& compute_pass_descriptor_queue_);
-    ~BlockLinearUnswizzle2DPass();
-
-    void Unswizzle(Image& image, const StagingBufferRef& map,
-                   std::span<const VideoCommon::SwizzleParameters> swizzles);
-
-private:
-    Scheduler& scheduler;
-    ComputePassDescriptorQueue& compute_pass_descriptor_queue;
-};
-
-class BlockLinearUnswizzleImage3DPass final : public ComputePass {
-public:
-    explicit BlockLinearUnswizzleImage3DPass(
-        const Device& device_, Scheduler& scheduler_, DescriptorPool& descriptor_pool_,
-        ComputePassDescriptorQueue& compute_pass_descriptor_queue_);
-    ~BlockLinearUnswizzleImage3DPass();
-
-    void Unswizzle(Image& image, const StagingBufferRef& map,
-                   std::span<const VideoCommon::SwizzleParameters> swizzles);
-
-private:
-    Scheduler& scheduler;
-    ComputePassDescriptorQueue& compute_pass_descriptor_queue;
-};
-
-class PitchUnswizzlePass final : public ComputePass {
-public:
-    explicit PitchUnswizzlePass(const Device& device_, Scheduler& scheduler_,
-                                DescriptorPool& descriptor_pool_,
-                                ComputePassDescriptorQueue& compute_pass_descriptor_queue_);
-    ~PitchUnswizzlePass();
-
-    void Unswizzle(Image& image, const StagingBufferRef& map,
-                   std::span<const VideoCommon::SwizzleParameters> swizzles);
-
-private:
-    Scheduler& scheduler;
-    ComputePassDescriptorQueue& compute_pass_descriptor_queue;
-};
-
 } // namespace Vulkan
@@ -694,11 +694,9 @@ void GraphicsPipeline::MakePipeline(VkRenderPass render_pass) {
        const size_t num_vertex_arrays = (std::min)(
            Maxwell::NumVertexArrays, static_cast<size_t>(device.GetMaxVertexInputBindings()));
        for (size_t index = 0; index < num_vertex_arrays; ++index) {
-            const bool instanced = ((key.state.enabled_divisors >> index) & 1) != 0;
-            auto rate = VK_VERTEX_INPUT_RATE_VERTEX;
-            if (instanced) {
-                rate = VK_VERTEX_INPUT_RATE_INSTANCE;
-            }
+            const bool instanced = key.state.binding_divisors[index] != 0;
+            const auto rate =
+                instanced ? VK_VERTEX_INPUT_RATE_INSTANCE : VK_VERTEX_INPUT_RATE_VERTEX;
            vertex_bindings.push_back({
                .binding = static_cast<u32>(index),
                .stride = key.state.vertex_strides[index],
@@ -707,7 +705,7 @@ void GraphicsPipeline::MakePipeline(VkRenderPass render_pass) {
            if (instanced) {
                vertex_binding_divisors.push_back({
                    .binding = static_cast<u32>(index),
-                    .divisor = device.GetVertexAttribDivisor(key.state.binding_divisors[index]),
+                    .divisor = key.state.binding_divisors[index],
                });
            }
        }
@@ -137,12 +137,10 @@ VkRect2D GetScissorState(const Maxwell& regs, size_t index, u32 up_scale = 1, u3
    max_y = (std::max)(max_y, 0);

    if (src.enable) {
-        const s32 min_x = static_cast<s32>(src.min_x.Value());
-        const s32 max_x = static_cast<s32>(src.max_x.Value());
-        scissor.offset.x = scale_up(min_x);
+        scissor.offset.x = scale_up(src.min_x);
        scissor.offset.y = scale_up(min_y);
-        scissor.extent.width = scale_up((std::max)(max_x - min_x, 0));
-        scissor.extent.height = scale_up((std::max)(max_y - min_y, 0));
+        scissor.extent.width = scale_up(src.max_x - src.min_x);
+        scissor.extent.height = scale_up(max_y - min_y);
    } else {
        scissor.offset.x = 0;
        scissor.offset.y = 0;
@@ -262,7 +260,6 @@ void RasterizerVulkan::PrepareDraw(bool is_indexed, Func&& draw_func) {
 }

 void RasterizerVulkan::Draw(bool is_indexed, u32 instance_count) {
-    buffer_cache.SetDrawInstanceCount(instance_count);
    PrepareDraw(is_indexed, [this, is_indexed, instance_count] {
        const auto& draw_state = maxwell3d->draw_manager.draw_state;
        const u32 num_instances{instance_count};
@@ -298,7 +295,6 @@ void RasterizerVulkan::Draw(bool is_indexed, u32 instance_count) {
 void RasterizerVulkan::DrawIndirect() {
    const auto& params = maxwell3d->draw_manager.indirect_state;
    buffer_cache.SetDrawIndirect(&params);
-    buffer_cache.SetDrawInstanceCount(0);
    PrepareDraw(params.is_indexed, [this, &params] {
        const auto indirect_buffer = buffer_cache.GetDrawIndirectBuffer();
        const auto& buffer = indirect_buffer.first;
@@ -1921,19 +1917,13 @@ void RasterizerVulkan::UpdateVertexInput(Tegra::Engines::Maxwell3D::Regs& regs)
    for (u32 binding = 0; binding < max_bindings; ++binding) {
        const auto& input_binding{regs.vertex_streams[binding]};
        const bool is_instanced{regs.vertex_stream_instances.IsInstancingEnabled(binding)};
-        auto input_rate = VK_VERTEX_INPUT_RATE_VERTEX;
-        u32 divisor = 1;
-        if (is_instanced) {
-            input_rate = VK_VERTEX_INPUT_RATE_INSTANCE;
-            divisor = device.GetVertexAttribDivisor(input_binding.frequency);
-        }
        bindings.push_back({
            .sType = VK_STRUCTURE_TYPE_VERTEX_INPUT_BINDING_DESCRIPTION_2_EXT,
            .pNext = nullptr,
            .binding = binding,
            .stride = input_binding.stride,
-            .inputRate = input_rate,
-            .divisor = divisor,
+            .inputRate = is_instanced ? VK_VERTEX_INPUT_RATE_INSTANCE : VK_VERTEX_INPUT_RATE_VERTEX,
+            .divisor = is_instanced ? input_binding.frequency : 1,
        });
    }

@@ -449,9 +449,7 @@ void Scheduler::EndRenderPass()
                                         | VK_ACCESS_COLOR_ATTACHMENT_READ_BIT
                                         | VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT
                                         | VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_READ_BIT
-                                         | VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT
-                                         | VK_ACCESS_TRANSFER_READ_BIT
-                                         | VK_ACCESS_TRANSFER_WRITE_BIT,
+                                         | VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT,
                        .oldLayout = VK_IMAGE_LAYOUT_GENERAL,
                        .newLayout = VK_IMAGE_LAYOUT_GENERAL,
                        .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
@@ -462,7 +460,7 @@ void Scheduler::EndRenderPass()
            }
            cmdbuf.EndRenderPass();
            cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_EARLY_FRAGMENT_TESTS_BIT | VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT |
-                                   VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT, vk::PIPELINE_STAGE_GRAPHICS_COMPUTE_TRANSFER,
+                                   VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT, vk::PIPELINE_STAGE_GRAPHICS_COMPUTE,
                                   0, nullptr, nullptr, vk::Span(barriers.data(), num_images));
            if (has_transform_feedback) {
                static constexpr VkMemoryBarrier XFB_OUTPUT_BARRIER{
@@ -160,55 +160,6 @@ constexpr VkBorderColor ConvertBorderColor(const std::array<float, 4>& color) {
          info.size.depth == 1;
 }

-[[nodiscard]] PixelFormat UnswizzleViewFormat(u32 bytes_per_block) {
-    switch (bytes_per_block) {
-    case 1:
-        return PixelFormat::R8_UINT;
-    case 2:
-        return PixelFormat::R16_UINT;
-    case 4:
-        return PixelFormat::R32_UINT;
-    case 8:
-        return PixelFormat::R32G32_UINT;
-    case 16:
-        return PixelFormat::R32G32B32A32_UINT;
-    default:
-        return PixelFormat::Invalid;
-    }
-}
-
-constexpr u32 UNSWIZZLE_WORKGROUP_INVOCATIONS = 32 * 32;
-
-[[nodiscard]] bool SupportsAcceleratedUnswizzleDevice(const Device& device) {
-    return device.IsKhrImageFormatListSupported() &&
-           device.GetMaxComputeWorkGroupInvocations() >= UNSWIZZLE_WORKGROUP_INVOCATIONS;
-}
-
-[[nodiscard]] bool SupportsAcceleratedUnswizzle(const Device& device, const ImageInfo& info) {
-    if (!SupportsAcceleratedUnswizzleDevice(device)) {
-        return false;
-    }
-    if (info.num_samples > 1) {
-        return false;
-    }
-    if (info.type != ImageType::e2D && info.type != ImageType::e3D &&
-        info.type != ImageType::Linear) {
-        return false;
-    }
-    const PixelFormat view_format =
-        UnswizzleViewFormat(VideoCore::Surface::BytesPerBlock(info.format));
-    if (view_format == PixelFormat::Invalid) {
-        return false;
-    }
-    if (!VideoCore::Surface::IsViewCompatible(info.format, view_format, false, true)) {
-        return false;
-    }
-    const auto host_format =
-        MaxwellToVK::SurfaceFormat(device, FormatType::Optimal, false, view_format);
-    return device.IsFormatSupported(host_format.format, VK_FORMAT_FEATURE_STORAGE_IMAGE_BIT,
-                                    FormatType::Optimal);
-}
-
 [[nodiscard]] VkImageCreateInfo MakeImageCreateInfo(const Device& device, const ImageInfo& info,
                                                    std::optional<VkFormat> format_override = {}) {
    auto format_info =
@@ -297,18 +248,8 @@ constexpr u32 UNSWIZZLE_WORKGROUP_INVOCATIONS = 32 * 32;
    return allocator.CreateImage(image_ci);
 }

-[[nodiscard]] VkImageViewType StorageViewType(ImageType type) {
-    if (type == ImageType::e3D) {
-        return VK_IMAGE_VIEW_TYPE_3D;
-    }
-    if (type == ImageType::Linear) {
-        return VK_IMAGE_VIEW_TYPE_2D;
-    }
-    return VK_IMAGE_VIEW_TYPE_2D_ARRAY;
-}
-
 [[nodiscard]] vk::ImageView MakeStorageView(const vk::Device& device, u32 level, VkImage image,
-                                            VkFormat format, VkImageViewType view_type) {
+                                            VkFormat format) {
    static constexpr VkImageViewUsageCreateInfo storage_image_view_usage_create_info{
        .sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_USAGE_CREATE_INFO,
        .pNext = nullptr,
@@ -319,7 +260,7 @@ constexpr u32 UNSWIZZLE_WORKGROUP_INVOCATIONS = 32 * 32;
        .pNext = &storage_image_view_usage_create_info,
        .flags = 0,
        .image = image,
-        .viewType = view_type,
+        .viewType = VK_IMAGE_VIEW_TYPE_2D_ARRAY,
        .format = format,
        .components{
            .r = VK_COMPONENT_SWIZZLE_IDENTITY,
@@ -717,11 +658,18 @@ void CopyBufferToImage(vk::CommandBuffer cmdbuf, VkBuffer src_buffer, VkImage im
            .subresourceRange = subresource_range,
    };

-    cmdbuf.PipelineBarrier(vk::PIPELINE_STAGE_GRAPHICS_COMPUTE, VK_PIPELINE_STAGE_TRANSFER_BIT, 0,
+    cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT |
+                           VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT |
+                           VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT, VK_PIPELINE_STAGE_TRANSFER_BIT, 0,
                           read_barrier);
    cmdbuf.CopyBufferToImage(src_buffer, image, VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, copies);
-    cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_TRANSFER_BIT, vk::PIPELINE_STAGE_GRAPHICS_COMPUTE, 0,
-                           nullptr, nullptr, write_barrier);
+    // TODO: Move this to another API
+    cmdbuf.PipelineBarrier(
+            VK_PIPELINE_STAGE_TRANSFER_BIT,
+            VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT |
+            VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT |
+            VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT,
+            0, nullptr, nullptr, write_barrier);
 }

 [[nodiscard]] VkImageBlit MakeImageBlit(const Region2D& dst_region, const Region2D& src_region,
@@ -1020,14 +968,6 @@ TextureCacheRuntime::TextureCacheRuntime(const Device& device_, Scheduler& sched
        bl3d_unswizzle_pass.emplace(device, scheduler, descriptor_pool,
                                   staging_buffer_pool, compute_pass_descriptor_queue);
    }
-    if (SupportsAcceleratedUnswizzleDevice(device)) {
-        bl_unswizzle_2d_pass.emplace(device, scheduler, descriptor_pool,
-                                     compute_pass_descriptor_queue);
-        bl_unswizzle_image_3d_pass.emplace(device, scheduler, descriptor_pool,
-                                           compute_pass_descriptor_queue);
-        pitch_unswizzle_pass.emplace(device, scheduler, descriptor_pool,
-                                     compute_pass_descriptor_queue);
-    }
 }

 void TextureCacheRuntime::Finish() {
@@ -1962,12 +1902,16 @@ Image::Image(TextureCacheRuntime& runtime_, const ImageInfo& info_, GPUVAddr gpu
    if (runtime->device.HasDebuggingToolAttached()) {
        original_image.SetObjectNameEXT(VideoCommon::Name(*this).c_str());
    }
-    if (False(flags & VideoCommon::ImageFlagBits::Converted) &&
-        SupportsAcceleratedUnswizzle(runtime->device, info)) {
-        flags |= VideoCommon::ImageFlagBits::AcceleratedUpload;
-    }
    current_image = &Image::original_image;
    storage_image_views.resize(info.resources.levels);
+    if (WillUseAcceleratedAstcDecode(runtime->device, info)) {
+        const auto& device = runtime->device.GetLogical();
+        const VkFormat storage_format = VK_FORMAT_A8B8G8R8_UNORM_PACK32;
+        for (s32 level = 0; level < info.resources.levels; ++level) {
+            storage_image_views[level] =
+                MakeStorageView(device, level, *original_image, storage_format);
+        }
+    }
 }

 Image::Image(const VideoCommon::NullImageParams& params) : VideoCommon::ImageBase{params} {}
@@ -2091,7 +2035,7 @@ void Image::UploadMemory(VkBuffer buffer, VkDeviceSize offset,
                                                temp_vk_image, info.format, info.num_samples,
                                                {image_copies.data(), image_copies.size()}, false);
        }
-        InitializationFor(current_image) = true;
+        initialized = true;
        runtime->ReleaseMsaaScratchImage(temp_vk_image);

        if (is_rescaled) {
@@ -2112,7 +2056,7 @@ void Image::UploadMemory(VkBuffer buffer, VkDeviceSize offset,
    const VkBuffer src_buffer = buffer;
    const VkImage vk_image = *original_image;
    const VkImageAspectFlags vk_aspect_mask = aspect_mask;
-    const bool was_initialized = std::exchange(InitializationFor(&Image::original_image), true);
+    const bool was_initialized = std::exchange(initialized, true);

    scheduler->Record([src_buffer, vk_image, vk_aspect_mask, was_initialized,
                       vk_copies](vk::CommandBuffer cmdbuf) {
@@ -2377,46 +2321,16 @@ void Image::DownloadMemory(const StagingBufferRef& map, std::span<const BufferIm
    DownloadMemory(buffers, offsets, copies);
 }

-std::vector<vk::ImageView>& Image::StorageViewsFor(vk::Image Image::*image) {
-    if (image == &Image::scaled_image) {
-        if (scaled_storage_image_views.empty()) {
-            scaled_storage_image_views.resize(info.resources.levels);
-        }
-        return scaled_storage_image_views;
-    }
-    return storage_image_views;
-}
-
-bool& Image::InitializationFor(vk::Image Image::*image) noexcept {
-    if (image == &Image::scaled_image) {
-        return scaled_initialized;
-    }
-    return original_initialized;
-}
-
 VkImageView Image::StorageImageView(s32 level) noexcept {
-    const bool astc_decode = WillUseAcceleratedAstcDecode(runtime->device, info);
-    const bool unswizzle_upload =
-        !astc_decode && True(flags & ImageFlagBits::AcceleratedUpload);
-    vk::Image Image::*target = current_image;
-    if (astc_decode || unswizzle_upload) {
-        target = &Image::original_image;
-    }
-    auto& view = StorageViewsFor(target)[level];
+    auto& view = storage_image_views[level];
    if (!view) {
        auto format_info =
            MaxwellToVK::SurfaceFormat(runtime->device, FormatType::Optimal, true, info.format);
-        if (astc_decode) {
+        if (WillUseAcceleratedAstcDecode(runtime->device, info)) {
            format_info.format = VK_FORMAT_A8B8G8R8_UNORM_PACK32;
        }
-        if (unswizzle_upload) {
-            const PixelFormat view_format =
-                UnswizzleViewFormat(VideoCore::Surface::BytesPerBlock(info.format));
-            format_info = MaxwellToVK::SurfaceFormat(runtime->device, FormatType::Optimal, false,
-                                                     view_format);
-        }
-        view = MakeStorageView(runtime->device.GetLogical(), level, *(this->*target),
-                               format_info.format, StorageViewType(info.type));
+        view = MakeStorageView(runtime->device.GetLogical(), level, *(this->*current_image),
+                               format_info.format);
    }
    return *view;
 }
@@ -2456,7 +2370,6 @@ bool Image::ScaleUp(bool ignore) {
    }
    if (NeedsScaleHelper()) {
        if (!BlitScaleHelper(true)) {
-            flags &= ~ImageFlagBits::Rescaled;
            current_image = &Image::original_image;
            return false;
        }
@@ -2646,10 +2559,6 @@ ImageView::ImageView(TextureCacheRuntime& runtime, const VideoCommon::ImageViewI
    if (device->IsExtAstcDecodeModeSupported() && IsLdrAstcFormat(format_info.format)) {
        view_next = &astc_decode_mode;
    }
-    auto subresource_range = MakeSubresourceRange(aspect_mask, info.range);
-    if (True(flags & VideoCommon::ImageViewFlagBits::Slice)) {
-        subresource_range.levelCount = 1;
-    }
    const VkImageViewCreateInfo create_info{
        .sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO,
        .pNext = view_next,
@@ -2658,7 +2567,7 @@ ImageView::ImageView(TextureCacheRuntime& runtime, const VideoCommon::ImageViewI
        .viewType = VkImageViewType{},
        .format = format_info.format,
        .components = swizzle_mapping,
-        .subresourceRange = subresource_range,
+        .subresourceRange = MakeSubresourceRange(aspect_mask, info.range),
    };
    const auto create = [&](TextureType tex_type, std::optional<u32> num_layers) {
        VkImageViewCreateInfo ci{create_info};
@@ -3232,19 +3141,6 @@ void TextureCacheRuntime::AccelerateImageUpload(
        return astc_decoder_pass->Assemble(image, map, swizzles);
    }

-    if (bl_unswizzle_2d_pass && image.info.type == ImageType::e2D) {
-        return bl_unswizzle_2d_pass->Unswizzle(image, map, swizzles);
-    }
-
-    if (bl_unswizzle_image_3d_pass && image.info.type == ImageType::e3D &&
-        !IsPixelFormatBCn(image.info.format)) {
-        return bl_unswizzle_image_3d_pass->Unswizzle(image, map, swizzles);
-    }
-
-    if (pitch_unswizzle_pass && image.info.type == ImageType::Linear) {
-        return pitch_unswizzle_pass->Unswizzle(image, map, swizzles);
-    }
-
    if (!Settings::values.gpu_unswizzle_enabled.GetValue() || !bl3d_unswizzle_pass) {
        if (IsPixelFormatBCn(image.info.format) && image.info.type == ImageType::e3D) {
            ASSERT(false && "GPU unswizzle is disabled for BCn 3D texture");
@@ -159,9 +159,6 @@ public:
    std::optional<ASTCDecoderPass> astc_decoder_pass;

    std::optional<BlockLinearUnswizzle3DPass> bl3d_unswizzle_pass;
-    std::optional<BlockLinearUnswizzle2DPass> bl_unswizzle_2d_pass;
-    std::optional<BlockLinearUnswizzleImage3DPass> bl_unswizzle_image_3d_pass;
-    std::optional<PitchUnswizzlePass> pitch_unswizzle_pass;
    const Settings::ResolutionScalingInfo& resolution;
    std::array<std::vector<VkFormat>, VideoCore::Surface::MaxPixelFormat> view_formats;

@@ -353,7 +350,7 @@ public:

    /// Returns true when the image is already initialized and mark it as initialized
    [[nodiscard]] bool ExchangeInitialization() noexcept {
-        return std::exchange(InitializationFor(current_image), true);
+        return std::exchange(initialized, true);
    }

    VkImageView StorageImageView(s32 level) noexcept;
@@ -373,10 +370,6 @@ private:

    bool NeedsScaleHelper() const;

-    std::vector<vk::ImageView>& StorageViewsFor(vk::Image Image::*image);
-
-    bool& InitializationFor(vk::Image Image::*image) noexcept;
-
    Scheduler* scheduler{};
    TextureCacheRuntime* runtime{};

@@ -394,10 +387,8 @@ private:
    vk::Image Image::*current_image{};

    std::vector<vk::ImageView> storage_image_views;
-    std::vector<vk::ImageView> scaled_storage_image_views;
    VkImageAspectFlags aspect_mask = 0;
-    bool original_initialized = false;
-    bool scaled_initialized = false;
+    bool initialized = false;

    std::optional<Framebuffer> scale_framebuffer;
    std::optional<Framebuffer> normal_framebuffer;
@@ -1,6 +1,3 @@
-// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
-// SPDX-License-Identifier: GPL-3.0-or-later
-
 // SPDX-FileCopyrightText: Copyright 2020 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later

@@ -26,8 +23,8 @@ struct BlockLinearSwizzle2DParams {
 };

 struct BlockLinearSwizzle3DParams {
-    alignas(16) std::array<u32, 3> origin;
-    alignas(16) std::array<s32, 3> destination;
+    std::array<u32, 3> origin;
+    std::array<s32, 3> destination;
    u32 bytes_per_block_log2;
    u32 slice_size;
    u32 block_size;
@@ -20,7 +20,6 @@
 #include "video_core/engines/kepler_compute.h"
 #include "video_core/guest_memory.h"
 #include "video_core/host1x/gpu_device_memory_manager.h"
-#include "video_core/texture_cache/accelerated_swizzle.h"
 #include "video_core/texture_cache/image_view_base.h"
 #include "video_core/texture_cache/samples_helper.h"
 #include "video_core/texture_cache/texture_cache_base.h"
@@ -280,11 +279,11 @@ void TextureCache<P>::CheckFeedbackLoop(std::span<const ImageViewInOut> views) {

            const ImageId view_image_id = slot_image_views[view.id].image_id;
            {
-                bool is_feedback = false;
+                bool is_continue = false;
                for (size_t i = 0; i < 8; ++i)
-                    is_feedback |= (rt_active_mask & (1u << i)) && view_image_id == rt_image_id[i];
-                if (is_feedback)
-                    return true;
+                    is_continue |= (rt_active_mask & (1u << i)) && view_image_id == rt_image_id[i];
+                if (is_continue)
+                    continue;
            }
            if (depth_active && view_image_id == rt_depth_image_id) {
                return true;
@@ -627,25 +626,14 @@ void TextureCache<P>::DownloadMemory(DAddr cpu_addr, size_t size) {
    std::ranges::sort(images, [this](ImageId lhs, ImageId rhs) {
        return slot_images[lhs].modification_tick < slot_images[rhs].modification_tick;
    });
-    size_t total_size_bytes = 0;
-    for (const ImageId image_id : images) {
-        total_size_bytes += slot_images[image_id].unswizzled_size_bytes;
-    }
-    auto download_map = runtime.DownloadStagingBuffer(total_size_bytes);
    for (const ImageId image_id : images) {
        Image& image = slot_images[image_id];
+        auto map = runtime.DownloadStagingBuffer(image.unswizzled_size_bytes);
        const auto copies = FixSmallVectorADL(FullDownloadCopies(image.info));
-        image.DownloadMemory(download_map, copies);
-        download_map.offset += image.unswizzled_size_bytes;
-    }
-    runtime.Finish();
-    std::span<u8> download_span = download_map.mapped_span;
-    for (const ImageId image_id : images) {
-        const ImageBase& image = slot_images[image_id];
-        const auto copies = FixSmallVectorADL(FullDownloadCopies(image.info));
-        SwizzleImage(*gpu_memory, image.gpu_addr, image.info, copies, download_span,
+        image.DownloadMemory(map, copies);
+        runtime.Finish();
+        SwizzleImage(*gpu_memory, image.gpu_addr, image.info, copies, map.mapped_span,
                     swizzle_data_buffer);
-        download_span = download_span.subspan(image.unswizzled_size_bytes);
    }
 }

@@ -1134,12 +1122,6 @@ void TextureCache<P>::RefreshContents(Image& image, ImageId image_id) {

    TrackImage(image, image_id);

-    if (image.info.rescaleable &&
-        IsRegionGpuModified(image.cpu_addr, image.guest_size_bytes)) {
-        runtime.TransitionImageLayout(image);
-        return;
-    }
-
    if (image.info.num_samples > 1 && !runtime.CanUploadMSAA()) {
        LOG_WARNING(HW_GPU, "MSAA image uploads are not implemented");
        runtime.TransitionImageLayout(image);
@@ -1175,7 +1157,8 @@ void TextureCache<P>::UploadImageContents(Image& image, StagingBuffer& staging)
    const GPUVAddr gpu_addr = image.gpu_addr;

    if (True(image.flags & ImageFlagBits::AcceleratedUpload)) {
-        gpu_memory->ReadBlockUnsafe(gpu_addr, mapped_span.data(), image.guest_size_bytes);
+        gpu_memory->ReadBlock(gpu_addr, mapped_span.data(), mapped_span.size_bytes(),
+                              VideoCommon::CacheType::NoTextureCache);
        const auto uploads = FullUploadSwizzles(image.info);
        runtime.AccelerateImageUpload(image, staging, FixSmallVectorADL(uploads), 0, 0);
        return;
@@ -1282,9 +1265,6 @@ ImageId TextureCache<P>::FindImage(const ImageInfo& info, GPUVAddr gpu_addr,

 template <class P>
 bool TextureCache<P>::ImageCanRescale(ImageBase& image) {
-    if (!Settings::values.resolution_info.active) {
-        return false;
-    }
    if (!image.info.rescaleable) {
        return false;
    }
@@ -1372,12 +1352,12 @@ void TextureCache<P>::QueueAsyncDecode(Image& image, ImageId image_id) {
    decode->image_id = image_id;
    async_decodes.push_back(std::move(decode));

-    Common::ScratchBuffer<u8> local_unswizzle_data_buffer(image.unswizzled_size_bytes);
+    std::vector<u8> local_unswizzle_data_buffer(image.unswizzled_size_bytes, 0);
    Tegra::Memory::GpuGuestMemory<u8, Tegra::Memory::GuestMemoryFlags::UnsafeRead> swizzle_data(*gpu_memory, image.gpu_addr, image.guest_size_bytes, &swizzle_data_buffer);
    auto copies = UnswizzleImage(*gpu_memory, image.gpu_addr, image.info, swizzle_data, local_unswizzle_data_buffer);
    const size_t out_size = MapSizeBytes(image);

-    auto func = [out_size, copies = std::move(copies), info = image.info,
+    auto func = [out_size, copies, info = image.info,
                 input = std::move(local_unswizzle_data_buffer),
                 async_decode = decode_ptr]() mutable {
        async_decode->decoded_data.resize_destructive(out_size);
@@ -1446,16 +1426,18 @@ void TextureCache<P>::TickAsyncUnswizzle() {
    Image& image = slot_images[task.image_id];

    if (!task.initialized) {
-        task.total_size = image.guest_size_bytes;
+        task.total_size = MapSizeBytes(image);
        task.staging_buffer = runtime.UploadStagingBuffer(task.total_size, true);

-        const auto layout = FullUploadSwizzles(task.info);
-        const auto params =
-            VideoCommon::Accelerated::MakeBlockLinearSwizzle3DParams(layout.front(), task.info);
-        task.bytes_per_slice = params.slice_size;
-        task.chunked = task.info.block.depth == 0;
+        const auto& info = image.info;
+        const u32 bytes_per_block = BytesPerBlock(info.format);
+        const u32 width_blocks = Common::DivCeil(info.size.width, 4u);
+        const u32 height_blocks = Common::DivCeil(info.size.height, 4u);
+
+        const u32 stride = width_blocks * bytes_per_block;
+        const u32 aligned_height = height_blocks;
+        task.bytes_per_slice = static_cast<size_t>(stride) * aligned_height;
        task.last_submitted_offset = 0;
-        task.slices_submitted = 0;
        task.initialized = true;
    }

@@ -1470,39 +1452,31 @@ void TextureCache<P>::TickAsyncUnswizzle() {
            if (copy_amount == 0) copy_amount = task.bytes_per_slice;
        }

-        gpu_memory->ReadBlockUnsafe(image.gpu_addr + task.current_offset,
-                                    task.staging_buffer.mapped_span.data() + task.current_offset,
-                                    copy_amount);
+        gpu_memory->ReadBlock(image.gpu_addr + task.current_offset,
+                              task.staging_buffer.mapped_span.data() + task.current_offset,
+                              copy_amount);
        task.current_offset += copy_amount;
    }

    const bool is_final_batch = task.current_offset >= task.total_size;
+    const size_t bytes_ready = task.current_offset - task.last_submitted_offset;
+    const u32 complete_slices = static_cast<u32>(bytes_ready / task.bytes_per_slice);

-    if (task.chunked) {
-        const size_t bytes_ready = task.current_offset - task.last_submitted_offset;
-        const u32 complete_slices = static_cast<u32>(bytes_ready / task.bytes_per_slice);
+    if (complete_slices >= swizzle_slices_per_batch || (is_final_batch && complete_slices > 0)) {
+        const u32 z_start = static_cast<u32>(task.last_submitted_offset / task.bytes_per_slice);
+        const u32 slices_to_process = (std::min)(complete_slices, swizzle_slices_per_batch);
+        const u32 z_count = (std::min)(slices_to_process, image.info.size.depth - z_start);

-        if (complete_slices >= swizzle_slices_per_batch || (is_final_batch && complete_slices > 0)) {
-            const u32 z_start = task.slices_submitted;
-            const u32 slices_to_process = (std::min)(complete_slices, swizzle_slices_per_batch);
-            const u32 z_count = (std::min)(slices_to_process, image.info.size.depth - z_start);
-
-            if (z_count > 0) {
-                const auto uploads = FullUploadSwizzles(task.info);
-                runtime.AccelerateImageUpload(image, task.staging_buffer,
-                                              FixSmallVectorADL(uploads), z_start, z_count);
-                task.last_submitted_offset += static_cast<size_t>(z_count) * task.bytes_per_slice;
-                task.slices_submitted += z_count;
-            }
+        if (z_count > 0) {
+            const auto uploads = FullUploadSwizzles(task.info);
+            runtime.AccelerateImageUpload(image, task.staging_buffer, FixSmallVectorADL(uploads), z_start, z_count);
+            task.last_submitted_offset += (static_cast<size_t>(z_count) * task.bytes_per_slice);
        }
-    } else if (is_final_batch && task.slices_submitted == 0) {
-        const auto uploads = FullUploadSwizzles(task.info);
-        runtime.AccelerateImageUpload(image, task.staging_buffer, FixSmallVectorADL(uploads), 0,
-                                      image.info.size.depth);
-        task.slices_submitted = image.info.size.depth;
    }

-    const bool all_slices_submitted = task.slices_submitted >= image.info.size.depth;
+    // Check if complete
+    const u32 slices_submitted = static_cast<u32>(task.last_submitted_offset / task.bytes_per_slice);
+    const bool all_slices_submitted = slices_submitted >= image.info.size.depth;

    if (is_final_batch && all_slices_submitted) {
        runtime.FreeDeferredStagingBuffer(task.staging_buffer);
@@ -139,8 +139,6 @@ class TextureCache : public VideoCommon::ChannelSetupCaches<TextureCacheChannelI
        AsyncBuffer staging_buffer;
        size_t last_submitted_offset = 0;
        size_t bytes_per_slice;
-        u32 slices_submitted = 0;
-        bool chunked = false;
        bool initialized = false;
    };

@@ -1219,11 +1219,6 @@ bool Device::GetSuitability(bool requires_swapchain) {
            VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_TRANSFORM_FEEDBACK_PROPERTIES_EXT;
        SetNext(next, properties.transform_feedback);
    }
-    if (extensions.vertex_attribute_divisor) {
-        properties.vertex_attribute_divisor.sType =
-            VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_VERTEX_ATTRIBUTE_DIVISOR_PROPERTIES_EXT;
-        SetNext(next, properties.vertex_attribute_divisor);
-    }
    if (extensions.maintenance5) {
        properties.maintenance5.sType =
            VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_MAINTENANCE_5_PROPERTIES_KHR;
@@ -70,7 +70,6 @@ VK_DEFINE_HANDLE(VmaAllocator)
    FEATURE(EXT, ProvokingVertex, PROVOKING_VERTEX, provoking_vertex)                              \
    FEATURE(EXT, Robustness2, ROBUSTNESS_2, robustness2)                                           \
    FEATURE(EXT, TransformFeedback, TRANSFORM_FEEDBACK, transform_feedback)                        \
-    FEATURE(EXT, VertexAttributeDivisor, VERTEX_ATTRIBUTE_DIVISOR, vertex_attribute_divisor)       \
    FEATURE(EXT, VertexInputDynamicState, VERTEX_INPUT_DYNAMIC_STATE, vertex_input_dynamic_state)  \
    FEATURE(KHR, Maintenance5, MAINTENANCE_5, maintenance5)                                        \
    FEATURE(KHR, Maintenance6, MAINTENANCE_6, maintenance6)                                        \
@@ -93,6 +92,7 @@ VK_DEFINE_HANDLE(VmaAllocator)
    EXTENSION(EXT, SHADER_STENCIL_EXPORT, shader_stencil_export)                                   \
    EXTENSION(EXT, SHADER_VIEWPORT_INDEX_LAYER, shader_viewport_index_layer)                       \
    EXTENSION(EXT, TOOLING_INFO, tooling_info)                                                     \
+    EXTENSION(EXT, VERTEX_ATTRIBUTE_DIVISOR, vertex_attribute_divisor)                             \
    EXTENSION(KHR, CREATE_RENDERPASS_2, create_renderpass2)                                        \
    EXTENSION(KHR, DEPTH_STENCIL_RESOLVE, depth_stencil_resolve)                                   \
    EXTENSION(KHR, DRAW_INDIRECT_COUNT, draw_indirect_count)                                       \
@@ -359,7 +359,6 @@ public:

 #define FN_MAX_LIMIT_LIST \
    FN_MAX_LIMIT_ELEM(ComputeSharedMemorySize) \
-    FN_MAX_LIMIT_ELEM(ComputeWorkGroupInvocations) \
    FN_MAX_LIMIT_ELEM(PerStageDescriptorSampledImages) \
    FN_MAX_LIMIT_ELEM(PerStageResources) \
    FN_MAX_LIMIT_ELEM(DescriptorSetSamplers) \
@@ -690,32 +689,6 @@ FN_MAX_LIMIT_LIST
        return features.host_query_reset.hostQueryReset != VK_FALSE;
    }

-    u32 GetMaxVertexAttribDivisor() const {
-        const u32 reported = properties.vertex_attribute_divisor.maxVertexAttribDivisor;
-        if (reported == 0) {
-            return 1;
-        }
-        return reported;
-    }
-
-    bool IsVertexAttributeInstanceRateZeroDivisorSupported() const {
-        return features.vertex_attribute_divisor.vertexAttributeInstanceRateZeroDivisor == VK_TRUE;
-    }
-
-    u32 GetVertexAttribDivisor(u32 frequency) const {
-        const u32 max_divisor = GetMaxVertexAttribDivisor();
-        if (frequency == 0) {
-            if (IsVertexAttributeInstanceRateZeroDivisorSupported()) {
-                return 0;
-            }
-            return max_divisor;
-        }
-        if (frequency > max_divisor) {
-            return max_divisor;
-        }
-        return frequency;
-    }
-
    /// Returns true if the device supports VK_EXT_transform_feedback.
    bool IsExtTransformFeedbackSupported() const {
        return extensions.transform_feedback;
@@ -1216,7 +1189,6 @@ private:
        VkPhysicalDeviceDescriptorBufferPropertiesEXT descriptor_buffer{};
        VkPhysicalDeviceSubgroupSizeControlProperties subgroup_size_control{};
        VkPhysicalDeviceTransformFeedbackPropertiesEXT transform_feedback{};
-        VkPhysicalDeviceVertexAttributeDivisorPropertiesEXT vertex_attribute_divisor{};
        VkPhysicalDeviceMaintenance5PropertiesKHR maintenance5{};
        VkPhysicalDeviceDepthStencilResolveProperties depth_stencil_resolve{};
        VkPhysicalDeviceCustomBorderColorPropertiesEXT custom_border_color{};
@@ -191,6 +191,7 @@ static FileSys::VirtualFile VfsDirectoryCreateFileWrapper(const FileSys::Virtual
 #include <windows.h>
 #ifdef _MSC_VER
 #pragma comment(lib, "Dwmapi.lib")
+#pragma comment(lib, "winmm.lib")
 #endif

 static inline void ApplyWindowsTitleBarDarkMode(HWND hwnd, bool enabled) {
Author	SHA1	Message	Date
lizzie	6a7c069824	fix linking issues	2026-09-02 11:37:51 +00:00
lizzie	86df49bbc9	remove latency func	2026-09-02 10:58:33 +00:00
lizzie	4e59b71c59	[audio] Nuke cubeb and use SDL3 exclusively Signed-off-by: lizzie <lizzie@eden-emu.dev>	2026-09-02 10:55:54 +00:00
lizzie	5c20f244c9	[cmake] stop building OpenGL on Android (#4342 ) Signed-off-by: lizzie <lizzie@eden-emu.dev> - [x] I have read and followed the [Contribution Guidelines](https://git.eden-emu.dev/eden-emu/eden/src/branch/master/CONTRIBUTING.md#code-contributions). - [x] I have read and followed the [AI Policy](https://git.eden-emu.dev/eden-emu/eden/src/branch/master/docs/policies/AI.md) - [x] I have read and followed the [Coding Guidelines](https://git.eden-emu.dev/eden-emu/eden/src/branch/master/docs/policies/Coding.md) to the best of my ability. ------------------- we don't even support OpenGL on android Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/4342 Reviewed-by: crueter <crueter@eden-emu.dev> Reviewed-by: CamilleLaVey <camillelavey99@gmail.com>	2026-09-02 04:07:24 +02:00