early-access version 1986

CMakeLists.txt

@@ -518,10 +518,6 @@ set(FFmpeg_COMPONENTS
     avutil
     swscale)
 
-if (${CMAKE_SYSTEM_NAME} STREQUAL "Linux")
-    Include(FindPkgConfig REQUIRED)
-    pkg_check_modules(LIBVA libva)
-endif()
 if (NOT YUZU_USE_BUNDLED_FFMPEG)
     # Use system installed FFmpeg
     find_package(FFmpeg QUIET COMPONENTS ${FFmpeg_COMPONENTS})
@@ -544,9 +540,6 @@ endif()
 
 if (YUZU_USE_BUNDLED_FFMPEG)
     if (NOT WIN32)
-        # TODO(lat9nq): Move this to externals/ffmpeg/CMakeLists.txt (and move externals/ffmpeg to
-        # externals/ffmpeg/ffmpeg)
-
         # Build FFmpeg from externals
         message(STATUS "Using FFmpeg from externals")
 
@@ -586,23 +579,20 @@ if (YUZU_USE_BUNDLED_FFMPEG)
                 CACHE PATH "Paths to FFmpeg libraries" FORCE)
         endforeach()
 
+        set(FFmpeg_INCLUDE_DIR
+            "${FFmpeg_PREFIX};${FFmpeg_BUILD_DIR}"
+            CACHE PATH "Path to FFmpeg headers" FORCE)
+
+        if (${CMAKE_SYSTEM_NAME} STREQUAL "Linux")
             Include(FindPkgConfig REQUIRED)
             pkg_check_modules(LIBVA libva)
-        pkg_check_modules(CUDA cuda)
+        endif()
-        pkg_check_modules(FFNVCODEC ffnvcodec)
-        pkg_check_modules(VDPAU vdpau)
-
-        set(FFmpeg_HWACCEL_LIBRARIES)
-        set(FFmpeg_HWACCEL_FLAGS)
-        set(FFmpeg_HWACCEL_INCLUDE_DIRS)
-        set(FFmpeg_HWACCEL_LDFLAGS)
-
         if(LIBVA_FOUND)
             pkg_check_modules(LIBDRM libdrm REQUIRED)
             find_package(X11 REQUIRED)
             pkg_check_modules(LIBVA-DRM libva-drm REQUIRED)
             pkg_check_modules(LIBVA-X11 libva-x11 REQUIRED)
-            list(APPEND FFmpeg_HWACCEL_LIBRARIES
+            set(FFmpeg_LIBVA_LIBRARIES
                 ${LIBDRM_LIBRARIES}
                 ${X11_LIBRARIES}
                 ${LIBVA-DRM_LIBRARIES}
@@ -612,57 +602,11 @@ if (YUZU_USE_BUNDLED_FFMPEG)
                 --enable-hwaccel=h264_vaapi
                 --enable-hwaccel=vp9_vaapi
                 --enable-libdrm)
-            list(APPEND FFmpeg_HWACCEL_INCLUDE_DIRS
-                ${LIBDRM_INCLUDE_DIRS}
-                ${X11_INCLUDE_DIRS}
-                ${LIBVA-DRM_INCLUDE_DIRS}
-                ${LIBVA-X11_INCLUDE_DIRS}
-                ${LIBVA_INCLUDE_DIRS}
-            )
             message(STATUS "VA-API found")
         else()
             set(FFmpeg_HWACCEL_FLAGS --disable-vaapi)
         endif()
 
-        if (FFNVCODEC_FOUND AND CUDA_FOUND)
-            list(APPEND FFmpeg_HWACCEL_FLAGS
-                --enable-cuvid
-                --enable-ffnvcodec
-                --enable-nvdec
-                --enable-cuda-llvm
-                --enable-hwaccel=h264_nvdec
-                --enable-hwaccel=vp9_nvdec
-                --extra-cflags=-I${CUDA_INCLUDE_DIRS}
-            )
-            list(APPEND FFmpeg_HWACCEL_LIBRARIES
-                ${FFNVCODEC_LIBRARIES}
-                ${CUDA_LIBRARIES}
-            )
-            list(APPEND FFmpeg_HWACCEL_INCLUDE_DIRS
-                ${FFNVCODEC_INCLUDE_DIRS}
-                ${CUDA_INCLUDE_DIRS}
-            )
-            list(APPEND FFmpeg_HWACCEL_LDFLAGS
-                ${FFNVCODEC_LDFLAGS}
-                ${CUDA_LDFLAGS}
-            )
-            message(STATUS "ffnvcodec libraries version ${FFNVCODEC_VERSION} found")
-        endif()
-
-        if (VDPAU_FOUND)
-            list(APPEND FFmpeg_HWACCEL_FLAGS
-                --enable-vdpau
-                --enable-hwaccel=h264_vdpau
-                --enable-hwaccel=vp9_vdpau
-            )
-            list(APPEND FFmpeg_HWACCEL_LIBRARIES ${VDPAU_LIBRARIES})
-            list(APPEND FFmpeg_HWACCEL_INCLUDE_DIRS ${VDPAU_INCLUDE_DIRS})
-            list(APPEND FFmpeg_HWACCEL_LDFLAGS ${VDPAU_LDFLAGS})
-            message(STATUS "vdpau libraries version ${VDPAU_VERSION} found")
-        else()
-            list(APPEND FFmpeg_HWACCEL_FLAGS --disable-vdpau)
-        endif()
-
         # `configure` parameters builds only exactly what yuzu needs from FFmpeg
         # `--disable-vdpau` is needed to avoid linking issues
         add_custom_command(
@@ -680,6 +624,7 @@ if (YUZU_USE_BUNDLED_FFMPEG)
                     --disable-network
                     --disable-postproc
                     --disable-swresample
+                    --disable-vdpau
                     --enable-decoder=h264
                     --enable-decoder=vp9
                     --cc="${CMAKE_C_COMPILER}"
@@ -708,26 +653,15 @@ if (YUZU_USE_BUNDLED_FFMPEG)
                 ${FFmpeg_BUILD_DIR}
         )
 
-        set(FFmpeg_INCLUDE_DIR
-            "${FFmpeg_PREFIX};${FFmpeg_BUILD_DIR};${FFmpeg_HWACCEL_INCLUDE_DIRS}"
-            CACHE PATH "Path to FFmpeg headers" FORCE)
-
-        set(FFmpeg_LDFLAGS
-            "${FFmpeg_HWACCEL_LDFLAGS}"
-            CACHE STRING "FFmpeg linker flags" FORCE)
-
         # ALL makes this custom target build every time
         # but it won't actually build if the DEPENDS parameter is up to date
         add_custom_target(ffmpeg-configure ALL DEPENDS ${FFmpeg_MAKEFILE})
         add_custom_target(ffmpeg-build ALL DEPENDS ${FFmpeg_BUILD_LIBRARIES} ffmpeg-configure)
         link_libraries(${FFmpeg_LIBVA_LIBRARIES})
-        set(FFmpeg_LIBRARIES ${FFmpeg_BUILD_LIBRARIES} ${FFmpeg_HWACCEL_LIBRARIES}
+        set(FFmpeg_LIBRARIES ${FFmpeg_LIBVA_LIBRARIES} ${FFmpeg_BUILD_LIBRARIES}
             CACHE PATH "Paths to FFmpeg libraries" FORCE)
         unset(FFmpeg_BUILD_LIBRARIES)
-        unset(FFmpeg_HWACCEL_FLAGS)
+        unset(FFmpeg_LIBVA_LIBRARIES)
-        unset(FFmpeg_HWACCEL_INCLUDE_DIRS)
-        unset(FFmpeg_HWACCEL_LDFLAGS)
-        unset(FFmpeg_HWACCEL_LIBRARIES)
 
         if (FFmpeg_FOUND)
             message(STATUS "Found FFmpeg version ${FFmpeg_VERSION}")
@@ -736,13 +670,12 @@ if (YUZU_USE_BUNDLED_FFMPEG)
         endif()
     else() # WIN32
         # Use yuzu FFmpeg binaries
-        set(FFmpeg_EXT_NAME "ffmpeg-4.4")
+        set(FFmpeg_EXT_NAME "ffmpeg-4.3.1")
         set(FFmpeg_PATH "${CMAKE_BINARY_DIR}/externals/${FFmpeg_EXT_NAME}")
         download_bundled_external("ffmpeg/" ${FFmpeg_EXT_NAME} "")
         set(FFmpeg_FOUND YES)
         set(FFmpeg_INCLUDE_DIR "${FFmpeg_PATH}/include" CACHE PATH "Path to FFmpeg headers" FORCE)
         set(FFmpeg_LIBRARY_DIR "${FFmpeg_PATH}/bin" CACHE PATH "Path to FFmpeg library directory" FORCE)
-        set(FFmpeg_LDFLAGS "" CACHE STRING "FFmpeg linker flags" FORCE)
         set(FFmpeg_DLL_DIR "${FFmpeg_PATH}/bin" CACHE PATH "Path to FFmpeg dll's" FORCE)
         set(FFmpeg_LIBRARIES
             ${FFmpeg_LIBRARY_DIR}/swscale.lib

README.md

@@ -1,7 +1,7 @@
 yuzu emulator early access
 =============
 
-This is the source code for early-access 1984.
+This is the source code for early-access 1986.
 
 ## Legal Notice
 

src/common/settings.cpp

@@ -54,7 +54,7 @@ void LogSettings() {
     log_setting("Renderer_GPUAccuracyLevel", values.gpu_accuracy.GetValue());
     log_setting("Renderer_UseAsynchronousGpuEmulation",
                 values.use_asynchronous_gpu_emulation.GetValue());
-    log_setting("Renderer_NvdecEmulation", values.nvdec_emulation.GetValue());
+    log_setting("Renderer_UseNvdecEmulation", values.use_nvdec_emulation.GetValue());
     log_setting("Renderer_AccelerateASTC", values.accelerate_astc.GetValue());
     log_setting("Renderer_UseVsync", values.use_vsync.GetValue());
     log_setting("Renderer_ShaderBackend", values.shader_backend.GetValue());
@@ -137,7 +137,7 @@ void RestoreGlobalState(bool is_powered_on) {
     values.use_disk_shader_cache.SetGlobal(true);
     values.gpu_accuracy.SetGlobal(true);
     values.use_asynchronous_gpu_emulation.SetGlobal(true);
-    values.nvdec_emulation.SetGlobal(true);
+    values.use_nvdec_emulation.SetGlobal(true);
     values.accelerate_astc.SetGlobal(true);
     values.use_vsync.SetGlobal(true);
     values.shader_backend.SetGlobal(true);

src/common/settings.h

@@ -47,12 +47,6 @@ enum class FullscreenMode : u32 {
     Exclusive = 1,
 };
 
-enum class NvdecEmulation : u32 {
-    Off = 0,
-    CPU = 1,
-    GPU = 2,
-};
-
 /** The BasicSetting class is a simple resource manager. It defines a label and default value
  * alongside the actual value of the setting for simpler and less-error prone use with frontend
  * configurations. Setting a default value and label is required, though subclasses may deviate from
@@ -348,7 +342,7 @@ struct Values {
     Setting<bool> use_disk_shader_cache{true, "use_disk_shader_cache"};
     Setting<GPUAccuracy> gpu_accuracy{GPUAccuracy::High, "gpu_accuracy"};
     Setting<bool> use_asynchronous_gpu_emulation{true, "use_asynchronous_gpu_emulation"};
-    Setting<NvdecEmulation> nvdec_emulation{NvdecEmulation::GPU, "nvdec_emulation"};
+    Setting<bool> use_nvdec_emulation{true, "use_nvdec_emulation"};
     Setting<bool> accelerate_astc{true, "accelerate_astc"};
     Setting<bool> use_vsync{true, "use_vsync"};
     BasicSetting<u16> fps_cap{1000, "fps_cap"};

src/core/telemetry_session.cpp

@@ -72,18 +72,6 @@ static const char* TranslateGPUAccuracyLevel(Settings::GPUAccuracy backend) {
     return "Unknown";
 }
 
-static const char* TranslateNvdecEmulation(Settings::NvdecEmulation backend) {
-    switch (backend) {
-    case Settings::NvdecEmulation::Off:
-        return "Off";
-    case Settings::NvdecEmulation::CPU:
-        return "CPU";
-    case Settings::NvdecEmulation::GPU:
-        return "GPU";
-    }
-    return "Unknown";
-}
-
 u64 GetTelemetryId() {
     u64 telemetry_id{};
     const auto filename = Common::FS::GetYuzuPath(Common::FS::YuzuPath::ConfigDir) / "telemetry_id";
@@ -241,8 +229,8 @@ void TelemetrySession::AddInitialInfo(Loader::AppLoader& app_loader,
              TranslateGPUAccuracyLevel(Settings::values.gpu_accuracy.GetValue()));
     AddField(field_type, "Renderer_UseAsynchronousGpuEmulation",
              Settings::values.use_asynchronous_gpu_emulation.GetValue());
-    AddField(field_type, "Renderer_NvdecEmulation",
+    AddField(field_type, "Renderer_UseNvdecEmulation",
-             TranslateNvdecEmulation(Settings::values.nvdec_emulation.GetValue()));
+             Settings::values.use_nvdec_emulation.GetValue());
     AddField(field_type, "Renderer_AccelerateASTC", Settings::values.accelerate_astc.GetValue());
     AddField(field_type, "Renderer_UseVsync", Settings::values.use_vsync.GetValue());
     AddField(field_type, "Renderer_ShaderBackend",

src/video_core/CMakeLists.txt

@@ -97,6 +97,7 @@ add_library(video_core STATIC
     renderer_opengl/gl_stream_buffer.h
     renderer_opengl/gl_texture_cache.cpp
     renderer_opengl/gl_texture_cache.h
+    renderer_opengl/gl_texture_cache_base.cpp
     renderer_opengl/gl_query_cache.cpp
     renderer_opengl/gl_query_cache.h
     renderer_opengl/maxwell_to_gl.h
@@ -155,6 +156,7 @@ add_library(video_core STATIC
     renderer_vulkan/vk_swapchain.h
     renderer_vulkan/vk_texture_cache.cpp
     renderer_vulkan/vk_texture_cache.h
+    renderer_vulkan/vk_texture_cache_base.cpp
     renderer_vulkan/vk_update_descriptor.cpp
     renderer_vulkan/vk_update_descriptor.h
     shader_cache.cpp
@@ -186,6 +188,7 @@ add_library(video_core STATIC
     texture_cache/samples_helper.h
     texture_cache/slot_vector.h
     texture_cache/texture_cache.h
+    texture_cache/texture_cache_base.h
     texture_cache/types.h
     texture_cache/util.cpp
     texture_cache/util.h
@@ -228,7 +231,6 @@ endif()
 
 target_include_directories(video_core PRIVATE ${FFmpeg_INCLUDE_DIR})
 target_link_libraries(video_core PRIVATE ${FFmpeg_LIBRARIES})
-target_link_options(video_core PRIVATE ${FFmpeg_LDFLAGS})
 
 add_dependencies(video_core host_shaders)
 target_include_directories(video_core PRIVATE ${HOST_SHADERS_INCLUDE})

src/video_core/command_classes/codecs/codec.cpp

@@ -5,7 +5,6 @@
 #include <fstream>
 #include <vector>
 #include "common/assert.h"
-#include "common/settings.h"
 #include "video_core/command_classes/codecs/codec.h"
 #include "video_core/command_classes/codecs/h264.h"
 #include "video_core/command_classes/codecs/vp9.h"
@@ -17,24 +16,44 @@ extern "C" {
 }
 
 namespace Tegra {
+#if defined(LIBVA_FOUND)
+// Hardware acceleration code from FFmpeg/doc/examples/hw_decode.c originally under MIT license
 namespace {
-void AVPacketDeleter(AVPacket* ptr) {
+constexpr std::array<const char*, 2> VAAPI_DRIVERS = {
-    av_packet_free(&ptr);
+    "i915",
-}
+    "amdgpu",
+};
 
-using AVPacketPtr = std::unique_ptr<AVPacket, decltype(&AVPacketDeleter)>;
+AVPixelFormat GetHwFormat(AVCodecContext*, const AVPixelFormat* pix_fmts) {
-
-AVPixelFormat GetGpuFormat(AVCodecContext* av_codec_ctx, const AVPixelFormat* pix_fmts) {
     for (const AVPixelFormat* p = pix_fmts; *p != AV_PIX_FMT_NONE; ++p) {
-        if (*p == av_codec_ctx->pix_fmt) {
+        if (*p == AV_PIX_FMT_VAAPI) {
-            return av_codec_ctx->pix_fmt;
+            return AV_PIX_FMT_VAAPI;
         }
     }
     LOG_INFO(Service_NVDRV, "Could not find compatible GPU AV format, falling back to CPU");
-    av_codec_ctx->pix_fmt = AV_PIX_FMT_NONE;
+    return *pix_fmts;
-    return AV_PIX_FMT_NONE;
+}
+
+bool CreateVaapiHwdevice(AVBufferRef** av_hw_device) {
+    AVDictionary* hwdevice_options = nullptr;
+    av_dict_set(&hwdevice_options, "connection_type", "drm", 0);
+    for (const auto& driver : VAAPI_DRIVERS) {
+        av_dict_set(&hwdevice_options, "kernel_driver", driver, 0);
+        const int hwdevice_error = av_hwdevice_ctx_create(av_hw_device, AV_HWDEVICE_TYPE_VAAPI,
+                                                          nullptr, hwdevice_options, 0);
+        if (hwdevice_error >= 0) {
+            LOG_INFO(Service_NVDRV, "Using VA-API with {}", driver);
+            av_dict_free(&hwdevice_options);
+            return true;
+        }
+        LOG_DEBUG(Service_NVDRV, "VA-API av_hwdevice_ctx_create failed {}", hwdevice_error);
+    }
+    LOG_DEBUG(Service_NVDRV, "VA-API av_hwdevice_ctx_create failed for all drivers");
+    av_dict_free(&hwdevice_options);
+    return false;
 }
 } // namespace
+#endif
 
 void AVFrameDeleter(AVFrame* ptr) {
     av_frame_free(&ptr);
@@ -49,158 +68,56 @@ Codec::~Codec() {
         return;
     }
     // Free libav memory
-    avcodec_free_context(&av_codec_ctx);
+    avcodec_send_packet(av_codec_ctx, nullptr);
-    av_buffer_unref(&av_gpu_decoder);
+    AVFrame* av_frame = av_frame_alloc();
+    avcodec_receive_frame(av_codec_ctx, av_frame);
+    avcodec_flush_buffers(av_codec_ctx);
+    av_frame_free(&av_frame);
+    avcodec_close(av_codec_ctx);
+    av_buffer_unref(&av_hw_device);
 }
 
-bool Codec::CreateGpuAvDevice() {
+void Codec::InitializeHwdec() {
+    // Prioritize integrated GPU to mitigate bandwidth bottlenecks
 #if defined(LIBVA_FOUND)
-    static constexpr std::array<const char*, 2> VAAPI_DRIVERS = {
+    if (CreateVaapiHwdevice(&av_hw_device)) {
-        "i915",
+        const auto hw_device_ctx = av_buffer_ref(av_hw_device);
-        "iHD",
-    };
-    AVDictionary* hwdevice_options = nullptr;
-    av_dict_set(&hwdevice_options, "connection_type", "drm", 0);
-    for (const auto& driver : VAAPI_DRIVERS) {
-        av_dict_set(&hwdevice_options, "kernel_driver", driver, 0);
-        const int hwdevice_error = av_hwdevice_ctx_create(&av_gpu_decoder, AV_HWDEVICE_TYPE_VAAPI,
-                                                          nullptr, hwdevice_options, 0);
-        if (hwdevice_error >= 0) {
-            LOG_INFO(Service_NVDRV, "Using VA-API with {}", driver);
-            av_dict_free(&hwdevice_options);
-            av_codec_ctx->pix_fmt = AV_PIX_FMT_VAAPI;
-            return true;
-        }
-        LOG_DEBUG(Service_NVDRV, "VA-API av_hwdevice_ctx_create failed {}", hwdevice_error);
-    }
-    LOG_DEBUG(Service_NVDRV, "VA-API av_hwdevice_ctx_create failed for all drivers");
-    av_dict_free(&hwdevice_options);
-#endif
-    static constexpr auto HW_CONFIG_METHOD = AV_CODEC_HW_CONFIG_METHOD_HW_DEVICE_CTX;
-    static constexpr std::array GPU_DECODER_TYPES{
-        AV_HWDEVICE_TYPE_CUDA,
-        AV_HWDEVICE_TYPE_D3D11VA,
-        AV_HWDEVICE_TYPE_VDPAU,
-    };
-    for (const auto& type : GPU_DECODER_TYPES) {
-        const int hwdevice_res = av_hwdevice_ctx_create(&av_gpu_decoder, type, nullptr, nullptr, 0);
-        if (hwdevice_res < 0) {
-            LOG_DEBUG(Service_NVDRV, "{} av_hwdevice_ctx_create failed {}",
-                      av_hwdevice_get_type_name(type), hwdevice_res);
-            continue;
-        }
-        for (int i = 0;; i++) {
-            const AVCodecHWConfig* config = avcodec_get_hw_config(av_codec, i);
-            if (!config) {
-                LOG_DEBUG(Service_NVDRV, "{} decoder does not support device type {}.",
-                          av_codec->name, av_hwdevice_get_type_name(type));
-                break;
-            }
-            if (config->methods & HW_CONFIG_METHOD && config->device_type == type) {
-                av_codec_ctx->pix_fmt = config->pix_fmt;
-                LOG_INFO(Service_NVDRV, "Using {} GPU decoder", av_hwdevice_get_type_name(type));
-                return true;
-            }
-        }
-    }
-    return false;
-}
-
-void Codec::InitializeAvCodecContext() {
-    av_codec_ctx = avcodec_alloc_context3(av_codec);
-    av_opt_set(av_codec_ctx->priv_data, "tune", "zerolatency", 0);
-}
-
-void Codec::InitializeGpuDecoder() {
-    if (!CreateGpuAvDevice()) {
-        av_buffer_unref(&av_gpu_decoder);
-        return;
-    }
-    auto* hw_device_ctx = av_buffer_ref(av_gpu_decoder);
         ASSERT_MSG(hw_device_ctx, "av_buffer_ref failed");
         av_codec_ctx->hw_device_ctx = hw_device_ctx;
-    av_codec_ctx->get_format = GetGpuFormat;
+        av_codec_ctx->get_format = GetHwFormat;
-}
-
-void Codec::TestGpuDecoder() {
-    static constexpr std::array<u8, 48> vp9_test{
-        0x92, 0x49, 0x83, 0x42, 0x00, 0x09, 0xf8, 0x05, 0x9b, 0x09, 0x1c, 0x12,
-        0x0e, 0x0c, 0x32, 0x00, 0x02, 0x08, 0x7f, 0xcd, 0xec, 0x3f, 0x3b, 0x77,
-        0x81, 0xf6, 0x47, 0xe2, 0xf0, 0x7e, 0x8b, 0x41, 0xfe, 0x3f, 0x87, 0x89,
-        0x7d, 0x37, 0xa2, 0xd1, 0x3e, 0x9c, 0x59, 0x5f, 0x75, 0xee, 0xbb, 0x97,
-    };
-    static constexpr std::array<u8, 48> h264_test{
-        0x00, 0x00, 0x00, 0x01, 0x67, 0x64, 0x00, 0x0D, 0xAC, 0x34, 0xE5, 0x05,
-        0x06, 0x7E, 0x78, 0x40, 0x00, 0x00, 0x19, 0x00, 0x00, 0x05, 0xDA, 0xA3,
-        0xC5, 0x0A, 0x45, 0x80, 0x00, 0x00, 0x00, 0x01, 0x68, 0xEE, 0xB2, 0xC8,
-        0xB0, 0x00, 0x00, 0x01, 0x65, 0x88, 0x80, 0x20, 0x01, 0xFF, 0xF3, 0x7E,
-    };
-    auto test_data = [&] {
-        switch (current_codec) {
-        case NvdecCommon::VideoCodec::H264:
-            return h264_test;
-        case NvdecCommon::VideoCodec::Vp9:
-            return vp9_test;
-        default:
-            UNIMPLEMENTED_MSG("Unknown codec {}", current_codec);
-            return std::array<u8, 48>{};
-        }
-    }();
-    AVPacketPtr packet{av_packet_alloc(), AVPacketDeleter};
-    if (!packet) {
-        LOG_ERROR(Service_NVDRV, "av_packet_alloc failed");
         return;
     }
-    // Temporarily disable logging, the test frames are incomplete and will log errors that can be
+#endif
-    // ignored for our purposes.
+    // TODO more GPU accelerated decoders
-    av_log_set_level(AV_LOG_QUIET);
-    packet->data = test_data.data();
-    packet->size = static_cast<s32>(test_data.size());
-    avcodec_send_packet(av_codec_ctx, packet.get());
-    av_log_set_level(AV_LOG_WARNING);
-
-    // GetGpuFormat is invoked after the avcodec_send_packet call.
-    // Fallback to CPU decoding if no compatible GPU format was found.
-    if (av_codec_ctx->pix_fmt != AV_PIX_FMT_NONE) {
-        avcodec_flush_buffers(av_codec_ctx);
-        using_gpu_decode = true;
-        return;
-    }
-    avcodec_close(av_codec_ctx);
-    av_buffer_unref(&av_gpu_decoder);
-    InitializeAvCodecContext();
-    avcodec_open2(av_codec_ctx, av_codec, nullptr);
 }
 
 void Codec::Initialize() {
-    const AVCodecID codec = [&] {
+    AVCodecID codec;
     switch (current_codec) {
     case NvdecCommon::VideoCodec::H264:
-            return AV_CODEC_ID_H264;
+        codec = AV_CODEC_ID_H264;
+        break;
     case NvdecCommon::VideoCodec::Vp9:
-            return AV_CODEC_ID_VP9;
+        codec = AV_CODEC_ID_VP9;
+        break;
     default:
         UNIMPLEMENTED_MSG("Unknown codec {}", current_codec);
-            return AV_CODEC_ID_NONE;
-        }
-    }();
-    av_codec = avcodec_find_decoder(codec);
-
-    InitializeAvCodecContext();
-    if (Settings::values.nvdec_emulation.GetValue() == Settings::NvdecEmulation::GPU) {
-        InitializeGpuDecoder();
-    }
-    if (const int res = avcodec_open2(av_codec_ctx, av_codec, nullptr); res < 0) {
-        LOG_ERROR(Service_NVDRV, "avcodec_open2() Failed with result {}", res);
-        avcodec_free_context(&av_codec_ctx);
-        av_buffer_unref(&av_gpu_decoder);
         return;
     }
-    if (av_codec_ctx->hw_device_ctx) {
+    av_codec = avcodec_find_decoder(codec);
-        TestGpuDecoder();
+    av_codec_ctx = avcodec_alloc_context3(av_codec);
-    } else {
+    av_opt_set(av_codec_ctx->priv_data, "tune", "zerolatency", 0);
+    InitializeHwdec();
+    if (!av_codec_ctx->hw_device_ctx) {
         LOG_INFO(Service_NVDRV, "Using FFmpeg software decoding");
     }
+    const auto av_error = avcodec_open2(av_codec_ctx, av_codec, nullptr);
+    if (av_error < 0) {
+        LOG_ERROR(Service_NVDRV, "avcodec_open2() Failed.");
+        avcodec_close(av_codec_ctx);
+        av_buffer_unref(&av_hw_device);
+        return;
+    }
     initialized = true;
 }
 
@@ -216,9 +133,6 @@ void Codec::Decode() {
     if (is_first_frame) {
         Initialize();
     }
-    if (!initialized) {
-        return;
-    }
     bool vp9_hidden_frame = false;
     std::vector<u8> frame_data;
     if (current_codec == NvdecCommon::VideoCodec::H264) {
@@ -227,48 +141,50 @@ void Codec::Decode() {
         frame_data = vp9_decoder->ComposeFrameHeader(state);
         vp9_hidden_frame = vp9_decoder->WasFrameHidden();
     }
-    AVPacketPtr packet{av_packet_alloc(), AVPacketDeleter};
+    AVPacket packet{};
-    if (!packet) {
+    av_init_packet(&packet);
-        LOG_ERROR(Service_NVDRV, "av_packet_alloc failed");
+    packet.data = frame_data.data();
-        return;
+    packet.size = static_cast<s32>(frame_data.size());
-    }
+    if (const int ret = avcodec_send_packet(av_codec_ctx, &packet); ret) {
-    packet->data = frame_data.data();
+        LOG_DEBUG(Service_NVDRV, "avcodec_send_packet error {}", ret);
-    packet->size = static_cast<s32>(frame_data.size());
-    if (const int res = avcodec_send_packet(av_codec_ctx, packet.get()); res != 0) {
-        LOG_DEBUG(Service_NVDRV, "avcodec_send_packet error {}", res);
         return;
     }
     // Only receive/store visible frames
     if (vp9_hidden_frame) {
         return;
     }
-    AVFramePtr initial_frame{av_frame_alloc(), AVFrameDeleter};
+    AVFrame* hw_frame = av_frame_alloc();
-    AVFramePtr final_frame{nullptr, AVFrameDeleter};
+    AVFrame* sw_frame = hw_frame;
-    ASSERT_MSG(initial_frame, "av_frame_alloc initial_frame failed");
+    ASSERT_MSG(hw_frame, "av_frame_alloc hw_frame failed");
-    if (const int ret = avcodec_receive_frame(av_codec_ctx, initial_frame.get()); ret) {
+    if (const int ret = avcodec_receive_frame(av_codec_ctx, hw_frame); ret) {
         LOG_DEBUG(Service_NVDRV, "avcodec_receive_frame error {}", ret);
+        av_frame_free(&hw_frame);
         return;
     }
-    if (initial_frame->width == 0 || initial_frame->height == 0) {
+    if (!hw_frame->width || !hw_frame->height) {
         LOG_WARNING(Service_NVDRV, "Zero width or height in frame");
+        av_frame_free(&hw_frame);
         return;
     }
-    if (using_gpu_decode) {
+#if defined(LIBVA_FOUND)
-        final_frame = AVFramePtr{av_frame_alloc(), AVFrameDeleter};
+    // Hardware acceleration code from FFmpeg/doc/examples/hw_decode.c under MIT license
-        ASSERT_MSG(final_frame, "av_frame_alloc final_frame failed");
+    if (hw_frame->format == AV_PIX_FMT_VAAPI) {
+        sw_frame = av_frame_alloc();
+        ASSERT_MSG(sw_frame, "av_frame_alloc sw_frame failed");
         // Can't use AV_PIX_FMT_YUV420P and share code with software decoding in vic.cpp
         // because Intel drivers crash unless using AV_PIX_FMT_NV12
-        final_frame->format = AV_PIX_FMT_NV12;
+        sw_frame->format = AV_PIX_FMT_NV12;
-        const int ret = av_hwframe_transfer_data(final_frame.get(), initial_frame.get(), 0);
+        const int transfer_data_ret = av_hwframe_transfer_data(sw_frame, hw_frame, 0);
-        ASSERT_MSG(!ret, "av_hwframe_transfer_data error {}", ret);
+        ASSERT_MSG(!transfer_data_ret, "av_hwframe_transfer_data error {}", transfer_data_ret);
-    } else {
+        av_frame_free(&hw_frame);
-        final_frame = std::move(initial_frame);
     }
-    if (final_frame->format != AV_PIX_FMT_YUV420P && final_frame->format != AV_PIX_FMT_NV12) {
+#endif
-        UNIMPLEMENTED_MSG("Unexpected video format: {}", final_frame->format);
+    if (sw_frame->format != AV_PIX_FMT_YUV420P && sw_frame->format != AV_PIX_FMT_NV12) {
+        UNIMPLEMENTED_MSG("Unexpected video format from host graphics: {}", sw_frame->format);
+        av_frame_free(&sw_frame);
         return;
     }
-    av_frames.push(std::move(final_frame));
+    av_frames.push(AVFramePtr{sw_frame, AVFrameDeleter});
     if (av_frames.size() > 10) {
         LOG_TRACE(Service_NVDRV, "av_frames.push overflow dropped frame");
         av_frames.pop();

src/video_core/command_classes/codecs/codec.h

@@ -50,26 +50,18 @@ public:
 
     /// Returns the value of current_codec
     [[nodiscard]] NvdecCommon::VideoCodec GetCurrentCodec() const;
-
     /// Return name of the current codec
     [[nodiscard]] std::string_view GetCurrentCodecName() const;
 
 private:
-    void InitializeAvCodecContext();
+    void InitializeHwdec();
-
-    void InitializeGpuDecoder();
-
-    void TestGpuDecoder();
-
-    bool CreateGpuAvDevice();
 
     bool initialized{};
-    bool using_gpu_decode{};
     NvdecCommon::VideoCodec current_codec{NvdecCommon::VideoCodec::None};
 
     AVCodec* av_codec{nullptr};
+    AVBufferRef* av_hw_device{nullptr};
     AVCodecContext* av_codec_ctx{nullptr};
-    AVBufferRef* av_gpu_decoder{nullptr};
 
     GPU& gpu;
     const NvdecCommon::NvdecRegisters& state;

src/video_core/renderer_opengl/gl_graphics_pipeline.cpp

@@ -15,7 +15,7 @@
 #include "video_core/renderer_opengl/gl_shader_util.h"
 #include "video_core/renderer_opengl/gl_state_tracker.h"
 #include "video_core/shader_notify.h"
-#include "video_core/texture_cache/texture_cache.h"
+#include "video_core/texture_cache/texture_cache_base.h"
 
 #if defined(_MSC_VER) && defined(NDEBUG)
 #define LAMBDA_FORCEINLINE [[msvc::forceinline]]

src/video_core/renderer_opengl/gl_rasterizer.cpp

@@ -32,7 +32,7 @@
 #include "video_core/renderer_opengl/maxwell_to_gl.h"
 #include "video_core/renderer_opengl/renderer_opengl.h"
 #include "video_core/shader_cache.h"
-#include "video_core/texture_cache/texture_cache.h"
+#include "video_core/texture_cache/texture_cache_base.h"
 
 namespace OpenGL {
 

src/video_core/renderer_opengl/gl_texture_cache.cpp

@@ -18,10 +18,8 @@
 #include "video_core/renderer_opengl/maxwell_to_gl.h"
 #include "video_core/renderer_opengl/util_shaders.h"
 #include "video_core/surface.h"
-#include "video_core/texture_cache/format_lookup_table.h"
+#include "video_core/texture_cache/formatter.h"
 #include "video_core/texture_cache/samples_helper.h"
-#include "video_core/texture_cache/texture_cache.h"
-#include "video_core/textures/decoders.h"
 
 namespace OpenGL {
 namespace {

src/video_core/renderer_opengl/gl_texture_cache.h

@@ -12,7 +12,7 @@
 #include "shader_recompiler/shader_info.h"
 #include "video_core/renderer_opengl/gl_resource_manager.h"
 #include "video_core/renderer_opengl/util_shaders.h"
-#include "video_core/texture_cache/texture_cache.h"
+#include "video_core/texture_cache/texture_cache_base.h"
 
 namespace OpenGL {
 

src/video_core/renderer_opengl/gl_texture_cache_base.cpp

@@ -0,0 +1,10 @@
+// Copyright 2021 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "video_core/renderer_opengl/gl_texture_cache.h"
+#include "video_core/texture_cache/texture_cache.h"
+
+namespace VideoCommon {
+template class VideoCommon::TextureCache<OpenGL::TextureCacheParams>;
+}

src/video_core/renderer_vulkan/vk_rasterizer.cpp

@@ -32,7 +32,7 @@
 #include "video_core/renderer_vulkan/vk_texture_cache.h"
 #include "video_core/renderer_vulkan/vk_update_descriptor.h"
 #include "video_core/shader_cache.h"
-#include "video_core/texture_cache/texture_cache.h"
+#include "video_core/texture_cache/texture_cache_base.h"
 #include "video_core/vulkan_common/vulkan_device.h"
 #include "video_core/vulkan_common/vulkan_wrapper.h"
 

src/video_core/renderer_vulkan/vk_texture_cache.cpp

@@ -19,6 +19,8 @@
 #include "video_core/renderer_vulkan/vk_scheduler.h"
 #include "video_core/renderer_vulkan/vk_staging_buffer_pool.h"
 #include "video_core/renderer_vulkan/vk_texture_cache.h"
+#include "video_core/texture_cache/formatter.h"
+#include "video_core/texture_cache/samples_helper.h"
 #include "video_core/vulkan_common/vulkan_device.h"
 #include "video_core/vulkan_common/vulkan_memory_allocator.h"
 #include "video_core/vulkan_common/vulkan_wrapper.h"

src/video_core/renderer_vulkan/vk_texture_cache.h

@@ -9,7 +9,7 @@
 
 #include "shader_recompiler/shader_info.h"
 #include "video_core/renderer_vulkan/vk_staging_buffer_pool.h"
-#include "video_core/texture_cache/texture_cache.h"
+#include "video_core/texture_cache/texture_cache_base.h"
 #include "video_core/vulkan_common/vulkan_memory_allocator.h"
 #include "video_core/vulkan_common/vulkan_wrapper.h"
 

src/video_core/renderer_vulkan/vk_texture_cache_base.cpp

@@ -0,0 +1,10 @@
+// Copyright 2021 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "video_core/renderer_vulkan/vk_texture_cache.h"
+#include "video_core/texture_cache/texture_cache.h"
+
+namespace VideoCommon {
+template class VideoCommon::TextureCache<Vulkan::TextureCacheParams>;
+}

src/video_core/texture_cache/image_view_info.cpp

@@ -6,7 +6,7 @@
 
 #include "common/assert.h"
 #include "video_core/texture_cache/image_view_info.h"
-#include "video_core/texture_cache/texture_cache.h"
+#include "video_core/texture_cache/texture_cache_base.h"
 #include "video_core/texture_cache/types.h"
 #include "video_core/textures/texture.h"
 
@@ -14,6 +14,8 @@ namespace VideoCommon {
 
 namespace {
 
+using Tegra::Texture::TextureType;
+
 constexpr u8 RENDER_TARGET_SWIZZLE = std::numeric_limits<u8>::max();
 
 [[nodiscard]] u8 CastSwizzle(SwizzleSource source) {

src/video_core/texture_cache/texture_cache.h

@@ -4,48 +4,11 @@
 
 #pragma once
 
-#include <algorithm>
-#include <array>
-#include <bit>
-#include <memory>
-#include <mutex>
-#include <optional>
-#include <span>
-#include <type_traits>
-#include <unordered_map>
-#include <unordered_set>
-#include <utility>
-#include <vector>
-
-#include <boost/container/small_vector.hpp>
-
 #include "common/alignment.h"
-#include "common/common_types.h"
-#include "common/literals.h"
-#include "common/logging/log.h"
 #include "common/settings.h"
-#include "video_core/compatible_formats.h"
-#include "video_core/delayed_destruction_ring.h"
 #include "video_core/dirty_flags.h"
-#include "video_core/engines/fermi_2d.h"
-#include "video_core/engines/kepler_compute.h"
-#include "video_core/engines/maxwell_3d.h"
-#include "video_core/memory_manager.h"
-#include "video_core/rasterizer_interface.h"
-#include "video_core/surface.h"
-#include "video_core/texture_cache/descriptor_table.h"
-#include "video_core/texture_cache/format_lookup_table.h"
-#include "video_core/texture_cache/formatter.h"
-#include "video_core/texture_cache/image_base.h"
-#include "video_core/texture_cache/image_info.h"
-#include "video_core/texture_cache/image_view_base.h"
-#include "video_core/texture_cache/image_view_info.h"
-#include "video_core/texture_cache/render_targets.h"
 #include "video_core/texture_cache/samples_helper.h"
-#include "video_core/texture_cache/slot_vector.h"
+#include "video_core/texture_cache/texture_cache_base.h"
-#include "video_core/texture_cache/types.h"
-#include "video_core/texture_cache/util.h"
-#include "video_core/textures/texture.h"
 
 namespace VideoCommon {
 
@@ -61,352 +24,6 @@ using VideoCore::Surface::PixelFormatFromRenderTargetFormat;
 using VideoCore::Surface::SurfaceType;
 using namespace Common::Literals;
 
-template <class P>
-class TextureCache {
-    /// Address shift for caching images into a hash table
-    static constexpr u64 PAGE_BITS = 20;
-
-    /// Enables debugging features to the texture cache
-    static constexpr bool ENABLE_VALIDATION = P::ENABLE_VALIDATION;
-    /// Implement blits as copies between framebuffers
-    static constexpr bool FRAMEBUFFER_BLITS = P::FRAMEBUFFER_BLITS;
-    /// True when some copies have to be emulated
-    static constexpr bool HAS_EMULATED_COPIES = P::HAS_EMULATED_COPIES;
-    /// True when the API can provide info about the memory of the device.
-    static constexpr bool HAS_DEVICE_MEMORY_INFO = P::HAS_DEVICE_MEMORY_INFO;
-
-    /// Image view ID for null descriptors
-    static constexpr ImageViewId NULL_IMAGE_VIEW_ID{0};
-    /// Sampler ID for bugged sampler ids
-    static constexpr SamplerId NULL_SAMPLER_ID{0};
-
-    static constexpr u64 DEFAULT_EXPECTED_MEMORY = 1_GiB;
-    static constexpr u64 DEFAULT_CRITICAL_MEMORY = 2_GiB;
-
-    using Runtime = typename P::Runtime;
-    using Image = typename P::Image;
-    using ImageAlloc = typename P::ImageAlloc;
-    using ImageView = typename P::ImageView;
-    using Sampler = typename P::Sampler;
-    using Framebuffer = typename P::Framebuffer;
-
-    struct BlitImages {
-        ImageId dst_id;
-        ImageId src_id;
-        PixelFormat dst_format;
-        PixelFormat src_format;
-    };
-
-    template <typename T>
-    struct IdentityHash {
-        [[nodiscard]] size_t operator()(T value) const noexcept {
-            return static_cast<size_t>(value);
-        }
-    };
-
-public:
-    explicit TextureCache(Runtime&, VideoCore::RasterizerInterface&, Tegra::Engines::Maxwell3D&,
-                          Tegra::Engines::KeplerCompute&, Tegra::MemoryManager&);
-
-    /// Notify the cache that a new frame has been queued
-    void TickFrame();
-
-    /// Return a constant reference to the given image view id
-    [[nodiscard]] const ImageView& GetImageView(ImageViewId id) const noexcept;
-
-    /// Return a reference to the given image view id
-    [[nodiscard]] ImageView& GetImageView(ImageViewId id) noexcept;
-
-    /// Mark an image as modified from the GPU
-    void MarkModification(ImageId id) noexcept;
-
-    /// Fill image_view_ids with the graphics images in indices
-    void FillGraphicsImageViews(std::span<const u32> indices,
-                                std::span<ImageViewId> image_view_ids);
-
-    /// Fill image_view_ids with the compute images in indices
-    void FillComputeImageViews(std::span<const u32> indices, std::span<ImageViewId> image_view_ids);
-
-    /// Get the sampler from the graphics descriptor table in the specified index
-    Sampler* GetGraphicsSampler(u32 index);
-
-    /// Get the sampler from the compute descriptor table in the specified index
-    Sampler* GetComputeSampler(u32 index);
-
-    /// Refresh the state for graphics image view and sampler descriptors
-    void SynchronizeGraphicsDescriptors();
-
-    /// Refresh the state for compute image view and sampler descriptors
-    void SynchronizeComputeDescriptors();
-
-    /// Update bound render targets and upload memory if necessary
-    /// @param is_clear True when the render targets are being used for clears
-    void UpdateRenderTargets(bool is_clear);
-
-    /// Find a framebuffer with the currently bound render targets
-    /// UpdateRenderTargets should be called before this
-    Framebuffer* GetFramebuffer();
-
-    /// Mark images in a range as modified from the CPU
-    void WriteMemory(VAddr cpu_addr, size_t size);
-
-    /// Download contents of host images to guest memory in a region
-    void DownloadMemory(VAddr cpu_addr, size_t size);
-
-    /// Remove images in a region
-    void UnmapMemory(VAddr cpu_addr, size_t size);
-
-    /// Remove images in a region
-    void UnmapGPUMemory(GPUVAddr gpu_addr, size_t size);
-
-    /// Blit an image with the given parameters
-    void BlitImage(const Tegra::Engines::Fermi2D::Surface& dst,
-                   const Tegra::Engines::Fermi2D::Surface& src,
-                   const Tegra::Engines::Fermi2D::Config& copy);
-
-    /// Invalidate the contents of the color buffer index
-    /// These contents become unspecified, the cache can assume aggressive optimizations.
-    void InvalidateColorBuffer(size_t index);
-
-    /// Invalidate the contents of the depth buffer
-    /// These contents become unspecified, the cache can assume aggressive optimizations.
-    void InvalidateDepthBuffer();
-
-    /// Try to find a cached image view in the given CPU address
-    [[nodiscard]] ImageView* TryFindFramebufferImageView(VAddr cpu_addr);
-
-    /// Return true when there are uncommitted images to be downloaded
-    [[nodiscard]] bool HasUncommittedFlushes() const noexcept;
-
-    /// Return true when the caller should wait for async downloads
-    [[nodiscard]] bool ShouldWaitAsyncFlushes() const noexcept;
-
-    /// Commit asynchronous downloads
-    void CommitAsyncFlushes();
-
-    /// Pop asynchronous downloads
-    void PopAsyncFlushes();
-
-    /// Return true when a CPU region is modified from the GPU
-    [[nodiscard]] bool IsRegionGpuModified(VAddr addr, size_t size);
-
-    std::mutex mutex;
-
-private:
-    /// Iterate over all page indices in a range
-    template <typename Func>
-    static void ForEachCPUPage(VAddr addr, size_t size, Func&& func) {
-        static constexpr bool RETURNS_BOOL = std::is_same_v<std::invoke_result<Func, u64>, bool>;
-        const u64 page_end = (addr + size - 1) >> PAGE_BITS;
-        for (u64 page = addr >> PAGE_BITS; page <= page_end; ++page) {
-            if constexpr (RETURNS_BOOL) {
-                if (func(page)) {
-                    break;
-                }
-            } else {
-                func(page);
-            }
-        }
-    }
-
-    template <typename Func>
-    static void ForEachGPUPage(GPUVAddr addr, size_t size, Func&& func) {
-        static constexpr bool RETURNS_BOOL = std::is_same_v<std::invoke_result<Func, u64>, bool>;
-        const u64 page_end = (addr + size - 1) >> PAGE_BITS;
-        for (u64 page = addr >> PAGE_BITS; page <= page_end; ++page) {
-            if constexpr (RETURNS_BOOL) {
-                if (func(page)) {
-                    break;
-                }
-            } else {
-                func(page);
-            }
-        }
-    }
-
-    /// Runs the Garbage Collector.
-    void RunGarbageCollector();
-
-    /// Fills image_view_ids in the image views in indices
-    void FillImageViews(DescriptorTable<TICEntry>& table,
-                        std::span<ImageViewId> cached_image_view_ids, std::span<const u32> indices,
-                        std::span<ImageViewId> image_view_ids);
-
-    /// Find or create an image view in the guest descriptor table
-    ImageViewId VisitImageView(DescriptorTable<TICEntry>& table,
-                               std::span<ImageViewId> cached_image_view_ids, u32 index);
-
-    /// Find or create a framebuffer with the given render target parameters
-    FramebufferId GetFramebufferId(const RenderTargets& key);
-
-    /// Refresh the contents (pixel data) of an image
-    void RefreshContents(Image& image, ImageId image_id);
-
-    /// Upload data from guest to an image
-    template <typename StagingBuffer>
-    void UploadImageContents(Image& image, StagingBuffer& staging_buffer);
-
-    /// Find or create an image view from a guest descriptor
-    [[nodiscard]] ImageViewId FindImageView(const TICEntry& config);
-
-    /// Create a new image view from a guest descriptor
-    [[nodiscard]] ImageViewId CreateImageView(const TICEntry& config);
-
-    /// Find or create an image from the given parameters
-    [[nodiscard]] ImageId FindOrInsertImage(const ImageInfo& info, GPUVAddr gpu_addr,
-                                            RelaxedOptions options = RelaxedOptions{});
-
-    /// Find an image from the given parameters
-    [[nodiscard]] ImageId FindImage(const ImageInfo& info, GPUVAddr gpu_addr,
-                                    RelaxedOptions options);
-
-    /// Create an image from the given parameters
-    [[nodiscard]] ImageId InsertImage(const ImageInfo& info, GPUVAddr gpu_addr,
-                                      RelaxedOptions options);
-
-    /// Create a new image and join perfectly matching existing images
-    /// Remove joined images from the cache
-    [[nodiscard]] ImageId JoinImages(const ImageInfo& info, GPUVAddr gpu_addr, VAddr cpu_addr);
-
-    /// Return a blit image pair from the given guest blit parameters
-    [[nodiscard]] BlitImages GetBlitImages(const Tegra::Engines::Fermi2D::Surface& dst,
-                                           const Tegra::Engines::Fermi2D::Surface& src);
-
-    /// Find or create a sampler from a guest descriptor sampler
-    [[nodiscard]] SamplerId FindSampler(const TSCEntry& config);
-
-    /// Find or create an image view for the given color buffer index
-    [[nodiscard]] ImageViewId FindColorBuffer(size_t index, bool is_clear);
-
-    /// Find or create an image view for the depth buffer
-    [[nodiscard]] ImageViewId FindDepthBuffer(bool is_clear);
-
-    /// Find or create a view for a render target with the given image parameters
-    [[nodiscard]] ImageViewId FindRenderTargetView(const ImageInfo& info, GPUVAddr gpu_addr,
-                                                   bool is_clear);
-
-    /// Iterates over all the images in a region calling func
-    template <typename Func>
-    void ForEachImageInRegion(VAddr cpu_addr, size_t size, Func&& func);
-
-    template <typename Func>
-    void ForEachImageInRegionGPU(GPUVAddr gpu_addr, size_t size, Func&& func);
-
-    template <typename Func>
-    void ForEachSparseImageInRegion(GPUVAddr gpu_addr, size_t size, Func&& func);
-
-    /// Iterates over all the images in a region calling func
-    template <typename Func>
-    void ForEachSparseSegment(ImageBase& image, Func&& func);
-
-    /// Find or create an image view in the given image with the passed parameters
-    [[nodiscard]] ImageViewId FindOrEmplaceImageView(ImageId image_id, const ImageViewInfo& info);
-
-    /// Register image in the page table
-    void RegisterImage(ImageId image);
-
-    /// Unregister image from the page table
-    void UnregisterImage(ImageId image);
-
-    /// Track CPU reads and writes for image
-    void TrackImage(ImageBase& image, ImageId image_id);
-
-    /// Stop tracking CPU reads and writes for image
-    void UntrackImage(ImageBase& image, ImageId image_id);
-
-    /// Delete image from the cache
-    void DeleteImage(ImageId image);
-
-    /// Remove image views references from the cache
-    void RemoveImageViewReferences(std::span<const ImageViewId> removed_views);
-
-    /// Remove framebuffers using the given image views from the cache
-    void RemoveFramebuffers(std::span<const ImageViewId> removed_views);
-
-    /// Mark an image as modified from the GPU
-    void MarkModification(ImageBase& image) noexcept;
-
-    /// Synchronize image aliases, copying data if needed
-    void SynchronizeAliases(ImageId image_id);
-
-    /// Prepare an image to be used
-    void PrepareImage(ImageId image_id, bool is_modification, bool invalidate);
-
-    /// Prepare an image view to be used
-    void PrepareImageView(ImageViewId image_view_id, bool is_modification, bool invalidate);
-
-    /// Execute copies from one image to the other, even if they are incompatible
-    void CopyImage(ImageId dst_id, ImageId src_id, std::span<const ImageCopy> copies);
-
-    /// Bind an image view as render target, downloading resources preemtively if needed
-    void BindRenderTarget(ImageViewId* old_id, ImageViewId new_id);
-
-    /// Create a render target from a given image and image view parameters
-    [[nodiscard]] std::pair<FramebufferId, ImageViewId> RenderTargetFromImage(
-        ImageId, const ImageViewInfo& view_info);
-
-    /// Returns true if the current clear parameters clear the whole image of a given image view
-    [[nodiscard]] bool IsFullClear(ImageViewId id);
-
-    Runtime& runtime;
-    VideoCore::RasterizerInterface& rasterizer;
-    Tegra::Engines::Maxwell3D& maxwell3d;
-    Tegra::Engines::KeplerCompute& kepler_compute;
-    Tegra::MemoryManager& gpu_memory;
-
-    DescriptorTable<TICEntry> graphics_image_table{gpu_memory};
-    DescriptorTable<TSCEntry> graphics_sampler_table{gpu_memory};
-    std::vector<SamplerId> graphics_sampler_ids;
-    std::vector<ImageViewId> graphics_image_view_ids;
-
-    DescriptorTable<TICEntry> compute_image_table{gpu_memory};
-    DescriptorTable<TSCEntry> compute_sampler_table{gpu_memory};
-    std::vector<SamplerId> compute_sampler_ids;
-    std::vector<ImageViewId> compute_image_view_ids;
-
-    RenderTargets render_targets;
-
-    std::unordered_map<TICEntry, ImageViewId> image_views;
-    std::unordered_map<TSCEntry, SamplerId> samplers;
-    std::unordered_map<RenderTargets, FramebufferId> framebuffers;
-
-    std::unordered_map<u64, std::vector<ImageMapId>, IdentityHash<u64>> page_table;
-    std::unordered_map<u64, std::vector<ImageId>, IdentityHash<u64>> gpu_page_table;
-    std::unordered_map<u64, std::vector<ImageId>, IdentityHash<u64>> sparse_page_table;
-
-    std::unordered_map<ImageId, std::vector<ImageViewId>> sparse_views;
-
-    VAddr virtual_invalid_space{};
-
-    bool has_deleted_images = false;
-    u64 total_used_memory = 0;
-    u64 minimum_memory;
-    u64 expected_memory;
-    u64 critical_memory;
-
-    SlotVector<Image> slot_images;
-    SlotVector<ImageMapView> slot_map_views;
-    SlotVector<ImageView> slot_image_views;
-    SlotVector<ImageAlloc> slot_image_allocs;
-    SlotVector<Sampler> slot_samplers;
-    SlotVector<Framebuffer> slot_framebuffers;
-
-    // TODO: This data structure is not optimal and it should be reworked
-    std::vector<ImageId> uncommitted_downloads;
-    std::queue<std::vector<ImageId>> committed_downloads;
-
-    static constexpr size_t TICKS_TO_DESTROY = 6;
-    DelayedDestructionRing<Image, TICKS_TO_DESTROY> sentenced_images;
-    DelayedDestructionRing<ImageView, TICKS_TO_DESTROY> sentenced_image_view;
-    DelayedDestructionRing<Framebuffer, TICKS_TO_DESTROY> sentenced_framebuffers;
-
-    std::unordered_map<GPUVAddr, ImageAllocId> image_allocs_table;
-
-    u64 modification_tick = 0;
-    u64 frame_tick = 0;
-    typename SlotVector<Image>::Iterator deletion_iterator;
-};
-
 template <class P>
 TextureCache<P>::TextureCache(Runtime& runtime_, VideoCore::RasterizerInterface& rasterizer_,
                               Tegra::Engines::Maxwell3D& maxwell3d_,
@@ -820,40 +437,6 @@ void TextureCache<P>::BlitImage(const Tegra::Engines::Fermi2D::Surface& dst,
     }
 }
 
-template <class P>
-void TextureCache<P>::InvalidateColorBuffer(size_t index) {
-    ImageViewId& color_buffer_id = render_targets.color_buffer_ids[index];
-    color_buffer_id = FindColorBuffer(index, false);
-    if (!color_buffer_id) {
-        LOG_ERROR(HW_GPU, "Invalidating invalid color buffer in index={}", index);
-        return;
-    }
-    // When invalidating a color buffer, the old contents are no longer relevant
-    ImageView& color_buffer = slot_image_views[color_buffer_id];
-    Image& image = slot_images[color_buffer.image_id];
-    image.flags &= ~ImageFlagBits::CpuModified;
-    image.flags &= ~ImageFlagBits::GpuModified;
-
-    runtime.InvalidateColorBuffer(color_buffer, index);
-}
-
-template <class P>
-void TextureCache<P>::InvalidateDepthBuffer() {
-    ImageViewId& depth_buffer_id = render_targets.depth_buffer_id;
-    depth_buffer_id = FindDepthBuffer(false);
-    if (!depth_buffer_id) {
-        LOG_ERROR(HW_GPU, "Invalidating invalid depth buffer");
-        return;
-    }
-    // When invalidating the depth buffer, the old contents are no longer relevant
-    ImageBase& image = slot_images[slot_image_views[depth_buffer_id].image_id];
-    image.flags &= ~ImageFlagBits::CpuModified;
-    image.flags &= ~ImageFlagBits::GpuModified;
-
-    ImageView& depth_buffer = slot_image_views[depth_buffer_id];
-    runtime.InvalidateDepthBuffer(depth_buffer);
-}
-
 template <class P>
 typename P::ImageView* TextureCache<P>::TryFindFramebufferImageView(VAddr cpu_addr) {
     // TODO: Properly implement this

src/video_core/texture_cache/texture_cache_base.h

@@ -0,0 +1,385 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <array>
+#include <mutex>
+#include <span>
+#include <type_traits>
+#include <unordered_map>
+#include <unordered_set>
+#include <vector>
+
+#include "common/common_types.h"
+#include "common/literals.h"
+#include "video_core/compatible_formats.h"
+#include "video_core/delayed_destruction_ring.h"
+#include "video_core/engines/fermi_2d.h"
+#include "video_core/engines/kepler_compute.h"
+#include "video_core/engines/maxwell_3d.h"
+#include "video_core/memory_manager.h"
+#include "video_core/rasterizer_interface.h"
+#include "video_core/surface.h"
+#include "video_core/texture_cache/descriptor_table.h"
+#include "video_core/texture_cache/image_base.h"
+#include "video_core/texture_cache/image_info.h"
+#include "video_core/texture_cache/image_view_info.h"
+#include "video_core/texture_cache/render_targets.h"
+#include "video_core/texture_cache/slot_vector.h"
+#include "video_core/texture_cache/types.h"
+#include "video_core/texture_cache/util.h"
+#include "video_core/textures/texture.h"
+
+namespace VideoCommon {
+
+using Tegra::Texture::SwizzleSource;
+using Tegra::Texture::TICEntry;
+using Tegra::Texture::TSCEntry;
+using VideoCore::Surface::GetFormatType;
+using VideoCore::Surface::IsCopyCompatible;
+using VideoCore::Surface::PixelFormat;
+using VideoCore::Surface::PixelFormatFromDepthFormat;
+using VideoCore::Surface::PixelFormatFromRenderTargetFormat;
+using namespace Common::Literals;
+
+template <class P>
+class TextureCache {
+    /// Address shift for caching images into a hash table
+    static constexpr u64 PAGE_BITS = 20;
+
+    /// Enables debugging features to the texture cache
+    static constexpr bool ENABLE_VALIDATION = P::ENABLE_VALIDATION;
+    /// Implement blits as copies between framebuffers
+    static constexpr bool FRAMEBUFFER_BLITS = P::FRAMEBUFFER_BLITS;
+    /// True when some copies have to be emulated
+    static constexpr bool HAS_EMULATED_COPIES = P::HAS_EMULATED_COPIES;
+    /// True when the API can provide info about the memory of the device.
+    static constexpr bool HAS_DEVICE_MEMORY_INFO = P::HAS_DEVICE_MEMORY_INFO;
+
+    /// Image view ID for null descriptors
+    static constexpr ImageViewId NULL_IMAGE_VIEW_ID{0};
+    /// Sampler ID for bugged sampler ids
+    static constexpr SamplerId NULL_SAMPLER_ID{0};
+
+    static constexpr u64 DEFAULT_EXPECTED_MEMORY = 1_GiB;
+    static constexpr u64 DEFAULT_CRITICAL_MEMORY = 2_GiB;
+
+    using Runtime = typename P::Runtime;
+    using Image = typename P::Image;
+    using ImageAlloc = typename P::ImageAlloc;
+    using ImageView = typename P::ImageView;
+    using Sampler = typename P::Sampler;
+    using Framebuffer = typename P::Framebuffer;
+
+    struct BlitImages {
+        ImageId dst_id;
+        ImageId src_id;
+        PixelFormat dst_format;
+        PixelFormat src_format;
+    };
+
+    template <typename T>
+    struct IdentityHash {
+        [[nodiscard]] size_t operator()(T value) const noexcept {
+            return static_cast<size_t>(value);
+        }
+    };
+
+public:
+    explicit TextureCache(Runtime&, VideoCore::RasterizerInterface&, Tegra::Engines::Maxwell3D&,
+                          Tegra::Engines::KeplerCompute&, Tegra::MemoryManager&);
+
+    /// Notify the cache that a new frame has been queued
+    void TickFrame();
+
+    /// Return a constant reference to the given image view id
+    [[nodiscard]] const ImageView& GetImageView(ImageViewId id) const noexcept;
+
+    /// Return a reference to the given image view id
+    [[nodiscard]] ImageView& GetImageView(ImageViewId id) noexcept;
+
+    /// Mark an image as modified from the GPU
+    void MarkModification(ImageId id) noexcept;
+
+    /// Fill image_view_ids with the graphics images in indices
+    void FillGraphicsImageViews(std::span<const u32> indices,
+                                std::span<ImageViewId> image_view_ids);
+
+    /// Fill image_view_ids with the compute images in indices
+    void FillComputeImageViews(std::span<const u32> indices, std::span<ImageViewId> image_view_ids);
+
+    /// Get the sampler from the graphics descriptor table in the specified index
+    Sampler* GetGraphicsSampler(u32 index);
+
+    /// Get the sampler from the compute descriptor table in the specified index
+    Sampler* GetComputeSampler(u32 index);
+
+    /// Refresh the state for graphics image view and sampler descriptors
+    void SynchronizeGraphicsDescriptors();
+
+    /// Refresh the state for compute image view and sampler descriptors
+    void SynchronizeComputeDescriptors();
+
+    /// Update bound render targets and upload memory if necessary
+    /// @param is_clear True when the render targets are being used for clears
+    void UpdateRenderTargets(bool is_clear);
+
+    /// Find a framebuffer with the currently bound render targets
+    /// UpdateRenderTargets should be called before this
+    Framebuffer* GetFramebuffer();
+
+    /// Mark images in a range as modified from the CPU
+    void WriteMemory(VAddr cpu_addr, size_t size);
+
+    /// Download contents of host images to guest memory in a region
+    void DownloadMemory(VAddr cpu_addr, size_t size);
+
+    /// Remove images in a region
+    void UnmapMemory(VAddr cpu_addr, size_t size);
+
+    /// Remove images in a region
+    void UnmapGPUMemory(GPUVAddr gpu_addr, size_t size);
+
+    /// Blit an image with the given parameters
+    void BlitImage(const Tegra::Engines::Fermi2D::Surface& dst,
+                   const Tegra::Engines::Fermi2D::Surface& src,
+                   const Tegra::Engines::Fermi2D::Config& copy);
+
+    /// Try to find a cached image view in the given CPU address
+    [[nodiscard]] ImageView* TryFindFramebufferImageView(VAddr cpu_addr);
+
+    /// Return true when there are uncommitted images to be downloaded
+    [[nodiscard]] bool HasUncommittedFlushes() const noexcept;
+
+    /// Return true when the caller should wait for async downloads
+    [[nodiscard]] bool ShouldWaitAsyncFlushes() const noexcept;
+
+    /// Commit asynchronous downloads
+    void CommitAsyncFlushes();
+
+    /// Pop asynchronous downloads
+    void PopAsyncFlushes();
+
+    /// Return true when a CPU region is modified from the GPU
+    [[nodiscard]] bool IsRegionGpuModified(VAddr addr, size_t size);
+
+    std::mutex mutex;
+
+private:
+    /// Iterate over all page indices in a range
+    template <typename Func>
+    static void ForEachCPUPage(VAddr addr, size_t size, Func&& func) {
+        static constexpr bool RETURNS_BOOL = std::is_same_v<std::invoke_result<Func, u64>, bool>;
+        const u64 page_end = (addr + size - 1) >> PAGE_BITS;
+        for (u64 page = addr >> PAGE_BITS; page <= page_end; ++page) {
+            if constexpr (RETURNS_BOOL) {
+                if (func(page)) {
+                    break;
+                }
+            } else {
+                func(page);
+            }
+        }
+    }
+
+    template <typename Func>
+    static void ForEachGPUPage(GPUVAddr addr, size_t size, Func&& func) {
+        static constexpr bool RETURNS_BOOL = std::is_same_v<std::invoke_result<Func, u64>, bool>;
+        const u64 page_end = (addr + size - 1) >> PAGE_BITS;
+        for (u64 page = addr >> PAGE_BITS; page <= page_end; ++page) {
+            if constexpr (RETURNS_BOOL) {
+                if (func(page)) {
+                    break;
+                }
+            } else {
+                func(page);
+            }
+        }
+    }
+
+    /// Runs the Garbage Collector.
+    void RunGarbageCollector();
+
+    /// Fills image_view_ids in the image views in indices
+    void FillImageViews(DescriptorTable<TICEntry>& table,
+                        std::span<ImageViewId> cached_image_view_ids, std::span<const u32> indices,
+                        std::span<ImageViewId> image_view_ids);
+
+    /// Find or create an image view in the guest descriptor table
+    ImageViewId VisitImageView(DescriptorTable<TICEntry>& table,
+                               std::span<ImageViewId> cached_image_view_ids, u32 index);
+
+    /// Find or create a framebuffer with the given render target parameters
+    FramebufferId GetFramebufferId(const RenderTargets& key);
+
+    /// Refresh the contents (pixel data) of an image
+    void RefreshContents(Image& image, ImageId image_id);
+
+    /// Upload data from guest to an image
+    template <typename StagingBuffer>
+    void UploadImageContents(Image& image, StagingBuffer& staging_buffer);
+
+    /// Find or create an image view from a guest descriptor
+    [[nodiscard]] ImageViewId FindImageView(const TICEntry& config);
+
+    /// Create a new image view from a guest descriptor
+    [[nodiscard]] ImageViewId CreateImageView(const TICEntry& config);
+
+    /// Find or create an image from the given parameters
+    [[nodiscard]] ImageId FindOrInsertImage(const ImageInfo& info, GPUVAddr gpu_addr,
+                                            RelaxedOptions options = RelaxedOptions{});
+
+    /// Find an image from the given parameters
+    [[nodiscard]] ImageId FindImage(const ImageInfo& info, GPUVAddr gpu_addr,
+                                    RelaxedOptions options);
+
+    /// Create an image from the given parameters
+    [[nodiscard]] ImageId InsertImage(const ImageInfo& info, GPUVAddr gpu_addr,
+                                      RelaxedOptions options);
+
+    /// Create a new image and join perfectly matching existing images
+    /// Remove joined images from the cache
+    [[nodiscard]] ImageId JoinImages(const ImageInfo& info, GPUVAddr gpu_addr, VAddr cpu_addr);
+
+    /// Return a blit image pair from the given guest blit parameters
+    [[nodiscard]] BlitImages GetBlitImages(const Tegra::Engines::Fermi2D::Surface& dst,
+                                           const Tegra::Engines::Fermi2D::Surface& src);
+
+    /// Find or create a sampler from a guest descriptor sampler
+    [[nodiscard]] SamplerId FindSampler(const TSCEntry& config);
+
+    /// Find or create an image view for the given color buffer index
+    [[nodiscard]] ImageViewId FindColorBuffer(size_t index, bool is_clear);
+
+    /// Find or create an image view for the depth buffer
+    [[nodiscard]] ImageViewId FindDepthBuffer(bool is_clear);
+
+    /// Find or create a view for a render target with the given image parameters
+    [[nodiscard]] ImageViewId FindRenderTargetView(const ImageInfo& info, GPUVAddr gpu_addr,
+                                                   bool is_clear);
+
+    /// Iterates over all the images in a region calling func
+    template <typename Func>
+    void ForEachImageInRegion(VAddr cpu_addr, size_t size, Func&& func);
+
+    template <typename Func>
+    void ForEachImageInRegionGPU(GPUVAddr gpu_addr, size_t size, Func&& func);
+
+    template <typename Func>
+    void ForEachSparseImageInRegion(GPUVAddr gpu_addr, size_t size, Func&& func);
+
+    /// Iterates over all the images in a region calling func
+    template <typename Func>
+    void ForEachSparseSegment(ImageBase& image, Func&& func);
+
+    /// Find or create an image view in the given image with the passed parameters
+    [[nodiscard]] ImageViewId FindOrEmplaceImageView(ImageId image_id, const ImageViewInfo& info);
+
+    /// Register image in the page table
+    void RegisterImage(ImageId image);
+
+    /// Unregister image from the page table
+    void UnregisterImage(ImageId image);
+
+    /// Track CPU reads and writes for image
+    void TrackImage(ImageBase& image, ImageId image_id);
+
+    /// Stop tracking CPU reads and writes for image
+    void UntrackImage(ImageBase& image, ImageId image_id);
+
+    /// Delete image from the cache
+    void DeleteImage(ImageId image);
+
+    /// Remove image views references from the cache
+    void RemoveImageViewReferences(std::span<const ImageViewId> removed_views);
+
+    /// Remove framebuffers using the given image views from the cache
+    void RemoveFramebuffers(std::span<const ImageViewId> removed_views);
+
+    /// Mark an image as modified from the GPU
+    void MarkModification(ImageBase& image) noexcept;
+
+    /// Synchronize image aliases, copying data if needed
+    void SynchronizeAliases(ImageId image_id);
+
+    /// Prepare an image to be used
+    void PrepareImage(ImageId image_id, bool is_modification, bool invalidate);
+
+    /// Prepare an image view to be used
+    void PrepareImageView(ImageViewId image_view_id, bool is_modification, bool invalidate);
+
+    /// Execute copies from one image to the other, even if they are incompatible
+    void CopyImage(ImageId dst_id, ImageId src_id, std::span<const ImageCopy> copies);
+
+    /// Bind an image view as render target, downloading resources preemtively if needed
+    void BindRenderTarget(ImageViewId* old_id, ImageViewId new_id);
+
+    /// Create a render target from a given image and image view parameters
+    [[nodiscard]] std::pair<FramebufferId, ImageViewId> RenderTargetFromImage(
+        ImageId, const ImageViewInfo& view_info);
+
+    /// Returns true if the current clear parameters clear the whole image of a given image view
+    [[nodiscard]] bool IsFullClear(ImageViewId id);
+
+    Runtime& runtime;
+    VideoCore::RasterizerInterface& rasterizer;
+    Tegra::Engines::Maxwell3D& maxwell3d;
+    Tegra::Engines::KeplerCompute& kepler_compute;
+    Tegra::MemoryManager& gpu_memory;
+
+    DescriptorTable<TICEntry> graphics_image_table{gpu_memory};
+    DescriptorTable<TSCEntry> graphics_sampler_table{gpu_memory};
+    std::vector<SamplerId> graphics_sampler_ids;
+    std::vector<ImageViewId> graphics_image_view_ids;
+
+    DescriptorTable<TICEntry> compute_image_table{gpu_memory};
+    DescriptorTable<TSCEntry> compute_sampler_table{gpu_memory};
+    std::vector<SamplerId> compute_sampler_ids;
+    std::vector<ImageViewId> compute_image_view_ids;
+
+    RenderTargets render_targets;
+
+    std::unordered_map<TICEntry, ImageViewId> image_views;
+    std::unordered_map<TSCEntry, SamplerId> samplers;
+    std::unordered_map<RenderTargets, FramebufferId> framebuffers;
+
+    std::unordered_map<u64, std::vector<ImageMapId>, IdentityHash<u64>> page_table;
+    std::unordered_map<u64, std::vector<ImageId>, IdentityHash<u64>> gpu_page_table;
+    std::unordered_map<u64, std::vector<ImageId>, IdentityHash<u64>> sparse_page_table;
+
+    std::unordered_map<ImageId, std::vector<ImageViewId>> sparse_views;
+
+    VAddr virtual_invalid_space{};
+
+    bool has_deleted_images = false;
+    u64 total_used_memory = 0;
+    u64 minimum_memory;
+    u64 expected_memory;
+    u64 critical_memory;
+
+    SlotVector<Image> slot_images;
+    SlotVector<ImageMapView> slot_map_views;
+    SlotVector<ImageView> slot_image_views;
+    SlotVector<ImageAlloc> slot_image_allocs;
+    SlotVector<Sampler> slot_samplers;
+    SlotVector<Framebuffer> slot_framebuffers;
+
+    // TODO: This data structure is not optimal and it should be reworked
+    std::vector<ImageId> uncommitted_downloads;
+    std::queue<std::vector<ImageId>> committed_downloads;
+
+    static constexpr size_t TICKS_TO_DESTROY = 6;
+    DelayedDestructionRing<Image, TICKS_TO_DESTROY> sentenced_images;
+    DelayedDestructionRing<ImageView, TICKS_TO_DESTROY> sentenced_image_view;
+    DelayedDestructionRing<Framebuffer, TICKS_TO_DESTROY> sentenced_framebuffers;
+
+    std::unordered_map<GPUVAddr, ImageAllocId> image_allocs_table;
+
+    u64 modification_tick = 0;
+    u64 frame_tick = 0;
+    typename SlotVector<Image>::Iterator deletion_iterator;
+};
+
+} // namespace VideoCommon

src/video_core/video_core.cpp

@@ -37,8 +37,7 @@ std::unique_ptr<VideoCore::RendererBase> CreateRenderer(
 namespace VideoCore {
 
 std::unique_ptr<Tegra::GPU> CreateGPU(Core::Frontend::EmuWindow& emu_window, Core::System& system) {
-    const auto nvdec_value = Settings::values.nvdec_emulation.GetValue();
+    const bool use_nvdec = Settings::values.use_nvdec_emulation.GetValue();
-    const bool use_nvdec = nvdec_value != Settings::NvdecEmulation::Off;
     const bool use_async = Settings::values.use_asynchronous_gpu_emulation.GetValue();
     auto gpu = std::make_unique<Tegra::GPU>(system, use_async, use_nvdec);
     auto context = emu_window.CreateSharedContext();

src/yuzu/configuration/config.cpp

@@ -826,7 +826,7 @@ void Config::ReadRendererValues() {
     ReadGlobalSetting(Settings::values.use_disk_shader_cache);
     ReadGlobalSetting(Settings::values.gpu_accuracy);
     ReadGlobalSetting(Settings::values.use_asynchronous_gpu_emulation);
-    ReadGlobalSetting(Settings::values.nvdec_emulation);
+    ReadGlobalSetting(Settings::values.use_nvdec_emulation);
     ReadGlobalSetting(Settings::values.accelerate_astc);
     ReadGlobalSetting(Settings::values.use_vsync);
     ReadGlobalSetting(Settings::values.shader_backend);
@@ -1372,10 +1372,7 @@ void Config::SaveRendererValues() {
                  static_cast<u32>(Settings::values.gpu_accuracy.GetDefault()),
                  Settings::values.gpu_accuracy.UsingGlobal());
     WriteGlobalSetting(Settings::values.use_asynchronous_gpu_emulation);
-    WriteSetting(QString::fromStdString(Settings::values.nvdec_emulation.GetLabel()),
+    WriteGlobalSetting(Settings::values.use_nvdec_emulation);
-                 static_cast<u32>(Settings::values.nvdec_emulation.GetValue(global)),
-                 static_cast<u32>(Settings::values.nvdec_emulation.GetDefault()),
-                 Settings::values.nvdec_emulation.UsingGlobal());
     WriteGlobalSetting(Settings::values.accelerate_astc);
     WriteGlobalSetting(Settings::values.use_vsync);
     WriteSetting(QString::fromStdString(Settings::values.shader_backend.GetLabel()),

src/yuzu/configuration/config.h

@@ -182,6 +182,5 @@ private:
 Q_DECLARE_METATYPE(Settings::CPUAccuracy);
 Q_DECLARE_METATYPE(Settings::GPUAccuracy);
 Q_DECLARE_METATYPE(Settings::FullscreenMode);
-Q_DECLARE_METATYPE(Settings::NvdecEmulation);
 Q_DECLARE_METATYPE(Settings::RendererBackend);
 Q_DECLARE_METATYPE(Settings::ShaderBackend);

src/yuzu/configuration/configure_graphics.cpp

@@ -88,30 +88,24 @@ void ConfigureGraphics::SetConfiguration() {
     ui->api_widget->setEnabled(runtime_lock);
     ui->use_asynchronous_gpu_emulation->setEnabled(runtime_lock);
     ui->use_disk_shader_cache->setEnabled(runtime_lock);
-    ui->nvdec_emulation_widget->setEnabled(runtime_lock);
+    ui->use_nvdec_emulation->setEnabled(runtime_lock);
     ui->accelerate_astc->setEnabled(runtime_lock);
     ui->use_disk_shader_cache->setChecked(Settings::values.use_disk_shader_cache.GetValue());
     ui->use_asynchronous_gpu_emulation->setChecked(
         Settings::values.use_asynchronous_gpu_emulation.GetValue());
+    ui->use_nvdec_emulation->setChecked(Settings::values.use_nvdec_emulation.GetValue());
     ui->accelerate_astc->setChecked(Settings::values.accelerate_astc.GetValue());
 
     if (Settings::IsConfiguringGlobal()) {
         ui->api->setCurrentIndex(static_cast<int>(Settings::values.renderer_backend.GetValue()));
         ui->fullscreen_mode_combobox->setCurrentIndex(
             static_cast<int>(Settings::values.fullscreen_mode.GetValue()));
-        ui->nvdec_emulation->setCurrentIndex(
-            static_cast<int>(Settings::values.nvdec_emulation.GetValue()));
         ui->aspect_ratio_combobox->setCurrentIndex(Settings::values.aspect_ratio.GetValue());
     } else {
         ConfigurationShared::SetPerGameSetting(ui->api, &Settings::values.renderer_backend);
         ConfigurationShared::SetHighlight(ui->api_widget,
                                           !Settings::values.renderer_backend.UsingGlobal());
 
-        ConfigurationShared::SetPerGameSetting(ui->nvdec_emulation,
-                                               &Settings::values.nvdec_emulation);
-        ConfigurationShared::SetHighlight(ui->nvdec_emulation_widget,
-                                          !Settings::values.nvdec_emulation.UsingGlobal());
-
         ConfigurationShared::SetPerGameSetting(ui->fullscreen_mode_combobox,
                                                &Settings::values.fullscreen_mode);
         ConfigurationShared::SetHighlight(ui->fullscreen_mode_label,
@@ -143,6 +137,8 @@ void ConfigureGraphics::ApplyConfiguration() {
     ConfigurationShared::ApplyPerGameSetting(&Settings::values.use_asynchronous_gpu_emulation,
                                              ui->use_asynchronous_gpu_emulation,
                                              use_asynchronous_gpu_emulation);
+    ConfigurationShared::ApplyPerGameSetting(&Settings::values.use_nvdec_emulation,
+                                             ui->use_nvdec_emulation, use_nvdec_emulation);
     ConfigurationShared::ApplyPerGameSetting(&Settings::values.accelerate_astc, ui->accelerate_astc,
                                              accelerate_astc);
 
@@ -151,9 +147,6 @@ void ConfigureGraphics::ApplyConfiguration() {
         if (Settings::values.renderer_backend.UsingGlobal()) {
             Settings::values.renderer_backend.SetValue(GetCurrentGraphicsBackend());
         }
-        if (Settings::values.nvdec_emulation.UsingGlobal()) {
-            Settings::values.nvdec_emulation.SetValue(GetCurrentNvdecEmulation());
-        }
         if (Settings::values.shader_backend.UsingGlobal()) {
             Settings::values.shader_backend.SetValue(shader_backend);
         }
@@ -187,13 +180,6 @@ void ConfigureGraphics::ApplyConfiguration() {
             }
         }
 
-        if (ui->nvdec_emulation->currentIndex() == ConfigurationShared::USE_GLOBAL_INDEX) {
-            Settings::values.nvdec_emulation.SetGlobal(true);
-        } else {
-            Settings::values.nvdec_emulation.SetGlobal(false);
-            Settings::values.nvdec_emulation.SetValue(GetCurrentNvdecEmulation());
-        }
-
         if (ui->bg_combobox->currentIndex() == ConfigurationShared::USE_GLOBAL_INDEX) {
             Settings::values.bg_red.SetGlobal(true);
             Settings::values.bg_green.SetGlobal(true);
@@ -292,20 +278,6 @@ Settings::RendererBackend ConfigureGraphics::GetCurrentGraphicsBackend() const {
                                                   ConfigurationShared::USE_GLOBAL_OFFSET);
 }
 
-Settings::NvdecEmulation ConfigureGraphics::GetCurrentNvdecEmulation() const {
-    if (Settings::IsConfiguringGlobal()) {
-        return static_cast<Settings::NvdecEmulation>(ui->nvdec_emulation->currentIndex());
-    }
-
-    if (ui->nvdec_emulation->currentIndex() == ConfigurationShared::USE_GLOBAL_INDEX) {
-        Settings::values.nvdec_emulation.SetGlobal(true);
-        return Settings::values.nvdec_emulation.GetValue();
-    }
-    Settings::values.nvdec_emulation.SetGlobal(false);
-    return static_cast<Settings::NvdecEmulation>(ui->nvdec_emulation->currentIndex() -
-                                                 ConfigurationShared::USE_GLOBAL_OFFSET);
-}
-
 void ConfigureGraphics::SetupPerGameUI() {
     if (Settings::IsConfiguringGlobal()) {
         ui->api->setEnabled(Settings::values.renderer_backend.UsingGlobal());
@@ -314,7 +286,7 @@ void ConfigureGraphics::SetupPerGameUI() {
         ui->aspect_ratio_combobox->setEnabled(Settings::values.aspect_ratio.UsingGlobal());
         ui->use_asynchronous_gpu_emulation->setEnabled(
             Settings::values.use_asynchronous_gpu_emulation.UsingGlobal());
-        ui->nvdec_emulation->setEnabled(Settings::values.nvdec_emulation.UsingGlobal());
+        ui->use_nvdec_emulation->setEnabled(Settings::values.use_nvdec_emulation.UsingGlobal());
         ui->accelerate_astc->setEnabled(Settings::values.accelerate_astc.UsingGlobal());
         ui->use_disk_shader_cache->setEnabled(Settings::values.use_disk_shader_cache.UsingGlobal());
         ui->bg_button->setEnabled(Settings::values.bg_red.UsingGlobal());
@@ -329,6 +301,8 @@ void ConfigureGraphics::SetupPerGameUI() {
 
     ConfigurationShared::SetColoredTristate(
         ui->use_disk_shader_cache, Settings::values.use_disk_shader_cache, use_disk_shader_cache);
+    ConfigurationShared::SetColoredTristate(
+        ui->use_nvdec_emulation, Settings::values.use_nvdec_emulation, use_nvdec_emulation);
     ConfigurationShared::SetColoredTristate(ui->accelerate_astc, Settings::values.accelerate_astc,
                                             accelerate_astc);
     ConfigurationShared::SetColoredTristate(ui->use_asynchronous_gpu_emulation,
@@ -342,6 +316,4 @@ void ConfigureGraphics::SetupPerGameUI() {
         static_cast<int>(Settings::values.fullscreen_mode.GetValue(true)));
     ConfigurationShared::InsertGlobalItem(
         ui->api, static_cast<int>(Settings::values.renderer_backend.GetValue(true)));
-    ConfigurationShared::InsertGlobalItem(
-        ui->nvdec_emulation, static_cast<int>(Settings::values.nvdec_emulation.GetValue(true)));
 }

src/yuzu/configuration/configure_graphics.h

@@ -43,7 +43,6 @@ private:
     void SetupPerGameUI();
 
     Settings::RendererBackend GetCurrentGraphicsBackend() const;
-    Settings::NvdecEmulation GetCurrentNvdecEmulation() const;
 
     std::unique_ptr<Ui::ConfigureGraphics> ui;
     QColor bg_color;

src/yuzu/configuration/configure_graphics.ui

@@ -167,6 +167,13 @@
           </property>
          </widget>
         </item>
+        <item>
+         <widget class="QCheckBox" name="use_nvdec_emulation">
+          <property name="text">
+           <string>Use NVDEC emulation</string>
+          </property>
+         </widget>
+        </item>
         <item>
           <widget class="QCheckBox" name="accelerate_astc">
             <property name="text">
@@ -174,50 +181,6 @@
             </property>
           </widget>
         </item>
-        <item>
-         <widget class="QWidget" name="nvdec_emulation_widget" native="true">
-          <layout class="QHBoxLayout" name="nvdec_emulation_layout">
-           <property name="leftMargin">
-            <number>0</number>
-           </property>
-           <property name="topMargin">
-            <number>0</number>
-           </property>
-           <property name="rightMargin">
-            <number>0</number>
-           </property>
-           <property name="bottomMargin">
-            <number>0</number>
-           </property>
-           <item>
-            <widget class="QLabel" name="nvdec_emulation_label">
-             <property name="text">
-              <string>NVDEC emulation:</string>
-             </property>
-            </widget>
-           </item>
-           <item>
-            <widget class="QComboBox" name="nvdec_emulation">
-             <item>
-              <property name="text">
-               <string>Disabled</string>
-              </property>
-             </item>
-             <item>
-              <property name="text">
-               <string>CPU Decoding</string>
-              </property>
-             </item>
-             <item>
-              <property name="text">
-               <string>GPU Decoding</string>
-              </property>
-             </item>
-            </widget>
-           </item>
-          </layout>
-         </widget>
-        </item>
         <item>
          <widget class="QWidget" name="fullscreen_mode_layout" native="true">
           <layout class="QHBoxLayout" name="horizontalLayout_1">

src/yuzu/main.cpp

@@ -2859,8 +2859,6 @@ void GMainWindow::OnToggleFilterBar() {
 }
 
 void GMainWindow::OnCaptureScreenshot() {
-    OnPauseGame();
-
     const u64 title_id = Core::System::GetInstance().CurrentProcess()->GetTitleID();
     const auto screenshot_path =
         QString::fromStdString(Common::FS::GetYuzuPathString(Common::FS::YuzuPath::ScreenshotsDir));
@@ -2872,23 +2870,22 @@ void GMainWindow::OnCaptureScreenshot() {
                            .arg(date);
 
     if (!Common::FS::CreateDir(screenshot_path.toStdString())) {
-        OnStartGame();
         return;
     }
 
 #ifdef _WIN32
     if (UISettings::values.enable_screenshot_save_as) {
+        OnPauseGame();
         filename = QFileDialog::getSaveFileName(this, tr("Capture Screenshot"), filename,
                                                 tr("PNG Image (*.png)"));
-        if (filename.isEmpty()) {
         OnStartGame();
+        if (filename.isEmpty()) {
             return;
         }
     }
 #endif
     render_window->CaptureScreenshot(UISettings::values.screenshot_resolution_factor.GetValue(),
                                      filename);
-    OnStartGame();
 }
 
 // TODO: Written 2020-10-01: Remove per-game config migration code when it is irrelevant

src/yuzu_cmd/config.cpp

@@ -465,7 +465,7 @@ void Config::ReadValues() {
     ReadSetting("Renderer", Settings::values.disable_fps_limit);
     ReadSetting("Renderer", Settings::values.shader_backend);
     ReadSetting("Renderer", Settings::values.use_asynchronous_shaders);
-    ReadSetting("Renderer", Settings::values.nvdec_emulation);
+    ReadSetting("Renderer", Settings::values.use_nvdec_emulation);
     ReadSetting("Renderer", Settings::values.accelerate_astc);
     ReadSetting("Renderer", Settings::values.use_fast_gpu_time);
     ReadSetting("Renderer", Settings::values.use_caches_gc);

src/yuzu_cmd/default_ini.h

@@ -261,9 +261,9 @@ shader_backend =
 # 0 (default): Off, 1: On
 use_asynchronous_shaders =
 
-# NVDEC emulation.
+# Enable NVDEC emulation.
-# 0: Disabled, 1: CPU Decoding, 2 (default): GPU Decoding
+# 0: Off, 1 (default): On
-nvdec_emulation =
+use_nvdec_emulation =
 
 # Accelerate ASTC texture decoding.
 # 0: Off, 1 (default): On