arm_dynarmic_cp15: Implement CNTPCT

arm_dynarmic_cp15: Update CP15
arm_dynarmic_32: InterpreterFallback should never happen
2020-06-17 17:10:24 +01:00 · 2020-06-17 17:10:24 +01:00 · 2020-06-17 17:10:24 +01:00 · 2020-06-17 11:20:52 -04:00 · 2020-06-16 23:44:48 -04:00 · 2020-06-16 22:27:27 -04:00
197 changed files with 7904 additions and 20155 deletions
@@ -1,4 +1,7 @@
 #!/bin/bash -ex

 chmod a+x ./.ci/scripts/format/docker.sh
+# the UID for the container yuzu user is 1027
+sudo chown -R 1027 ./
 docker run -v $(pwd):/yuzu yuzuemu/build-environments:linux-clang-format /bin/bash -ex /yuzu/.ci/scripts/format/docker.sh
+sudo chown -R $UID ./
@@ -2,4 +2,7 @@

 mkdir -p "ccache"  || true
 chmod a+x ./.ci/scripts/linux/docker.sh
+# the UID for the container yuzu user is 1027
+sudo chown -R 1027 ./
 docker run -e ENABLE_COMPATIBILITY_REPORTING -e CCACHE_DIR=/yuzu/ccache -v $(pwd):/yuzu yuzuemu/build-environments:linux-fresh /bin/bash /yuzu/.ci/scripts/linux/docker.sh $1
+sudo chown -R $UID ./
@@ -4,22 +4,10 @@ cd /yuzu

 ccache -s

-# Dirty hack to trick unicorn makefile into believing we are in a MINGW system
-mv /bin/uname /bin/uname1 && echo -e '#!/bin/sh\necho MINGW64' >> /bin/uname
-chmod +x /bin/uname
-
-# Dirty hack to trick unicorn makefile into believing we have cmd
-echo '' >> /bin/cmd
-chmod +x /bin/cmd
-
 mkdir build || true && cd build
 cmake .. -G Ninja -DDISPLAY_VERSION=$1 -DCMAKE_TOOLCHAIN_FILE="$(pwd)/../CMakeModules/MinGWCross.cmake" -DUSE_CCACHE=ON -DYUZU_USE_BUNDLED_UNICORN=ON -DENABLE_COMPATIBILITY_LIST_DOWNLOAD=ON -DCMAKE_BUILD_TYPE=Release
 ninja

-# Clean up the dirty hacks
-rm /bin/uname && mv /bin/uname1 /bin/uname
-rm /bin/cmd
-
 ccache -s

 echo "Tests skipped"
@@ -2,4 +2,7 @@

 mkdir -p "ccache" || true
 chmod a+x ./.ci/scripts/windows/docker.sh
+# the UID for the container yuzu user is 1027
+sudo chown -R 1027 ./
 docker run -e CCACHE_DIR=/yuzu/ccache -v $(pwd):/yuzu yuzuemu/build-environments:linux-mingw /bin/bash -ex /yuzu/.ci/scripts/windows/docker.sh $1
+sudo chown -R $UID ./
@@ -4,7 +4,9 @@ parameters:
  version: ''

 steps:
- script: mkdir build && cd build && cmake -G "Visual Studio 16 2019" -A x64 --config Release -DYUZU_USE_BUNDLED_QT=1 -DYUZU_USE_BUNDLED_SDL2=1 -DYUZU_USE_BUNDLED_UNICORN=1 -DYUZU_USE_QT_WEB_ENGINE=ON -DENABLE_COMPATIBILITY_LIST_DOWNLOAD=ON -DYUZU_ENABLE_COMPATIBILITY_REPORTING=${COMPAT} -DUSE_DISCORD_PRESENCE=ON -DDISPLAY_VERSION=${{ parameters['version'] }} .. && cd ..
+- script: python -m pip install --upgrade pip conan
+  displayName: 'Install conan'
+- script: mkdir build && cd build && cmake -G "Visual Studio 16 2019" -A x64 --config Release -DYUZU_USE_BUNDLED_QT=1 -DYUZU_USE_BUNDLED_UNICORN=1 -DYUZU_USE_QT_WEB_ENGINE=ON -DENABLE_COMPATIBILITY_LIST_DOWNLOAD=ON -DYUZU_ENABLE_COMPATIBILITY_REPORTING=${COMPAT} -DUSE_DISCORD_PRESENCE=ON -DDISPLAY_VERSION=${{ parameters['version'] }} .. && cd ..
  displayName: 'Configure CMake'
 - task: MSBuild@1
  displayName: 'Build'
@@ -5,18 +5,7 @@ variables:
  DisplayVersion: $[counter(variables['DisplayPrefix'], 1)]

 stages:
- stage: format
-  displayName: 'format'
-  jobs:
-  - job: format
-    displayName: 'clang'
-    continueOnError: true
-    pool:
-      vmImage: ubuntu-latest
-    steps:
-    - template: ./templates/format-check.yml
 - stage: build
-  dependsOn: format
  displayName: 'build'
  jobs:
  - job: build
@@ -1,33 +1,15 @@
 [submodule "inih"]
    path = externals/inih/inih
    url = https://github.com/svn2github/inih
-[submodule "boost"]
-    path = externals/boost
-    url = https://github.com/yuzu-emu/ext-boost.git
-[submodule "catch"]
-    path = externals/catch
-    url = https://github.com/philsquared/Catch.git
 [submodule "cubeb"]
    path = externals/cubeb
    url = https://github.com/kinetiknz/cubeb.git
 [submodule "dynarmic"]
    path = externals/dynarmic
    url = https://github.com/MerryMage/dynarmic.git
-[submodule "fmt"]
-    path = externals/fmt
-    url = https://github.com/fmtlib/fmt.git
-[submodule "lz4"]
-    path = externals/lz4
-    url = https://github.com/lz4/lz4.git
 [submodule "unicorn"]
    path = externals/unicorn
    url = https://github.com/yuzu-emu/unicorn
-[submodule "mbedtls"]
-    path = externals/mbedtls
-    url = https://github.com/DarkLordZach/mbedtls
-[submodule "opus"]
-    path = externals/opus/opus
-    url = https://github.com/xiph/opus.git
 [submodule "soundtouch"]
    path = externals/soundtouch
    url = https://github.com/citra-emu/ext-soundtouch.git
@@ -40,15 +22,15 @@
 [submodule "Vulkan-Headers"]
    path = externals/Vulkan-Headers
    url = https://github.com/KhronosGroup/Vulkan-Headers.git
-[submodule "externals/zstd"]
-    path = externals/zstd
-    url = https://github.com/facebook/zstd
 [submodule "sirit"]
    path = externals/sirit
    url = https://github.com/ReinUsesLisp/sirit
+[submodule "mbedtls"]
+    path = externals/mbedtls
+    url = https://github.com/DarkLordZach/mbedtls
 [submodule "libzip"]
    path = externals/libzip/libzip
    url = https://github.com/nih-at/libzip.git
-[submodule "zlib"]
-    path = externals/zlib/zlib
-    url = https://github.com/madler/zlib.git
+[submodule "xbyak"]
+    path = externals/xbyak
+    url = https://github.com/herumi/xbyak.git
@@ -1,7 +1,8 @@
-cmake_minimum_required(VERSION 3.7)
+cmake_minimum_required(VERSION 3.15)

 list(APPEND CMAKE_MODULE_PATH "${CMAKE_CURRENT_SOURCE_DIR}/CMakeModules")
 list(APPEND CMAKE_MODULE_PATH "${CMAKE_CURRENT_SOURCE_DIR}/externals/cmake-modules")
+list(APPEND CMAKE_MODULE_PATH "${CMAKE_CURRENT_SOURCE_DIR}/externals/find-modules")
 include(DownloadExternals)
 include(CMakeDependentOption)

@@ -10,7 +11,6 @@ project(yuzu)
 # Set bundled sdl2/qt as dependent options.
 # OFF by default, but if ENABLE_SDL2 and MSVC are true then ON
 option(ENABLE_SDL2 "Enable the SDL2 frontend" ON)
-CMAKE_DEPENDENT_OPTION(YUZU_USE_BUNDLED_SDL2 "Download bundled SDL2 binaries" ON "ENABLE_SDL2;MSVC" OFF)

 option(ENABLE_QT "Enable the Qt frontend" ON)
 CMAKE_DEPENDENT_OPTION(YUZU_USE_BUNDLED_QT "Download bundled Qt binaries" ON "ENABLE_QT;MSVC" OFF)
@@ -29,6 +29,13 @@ option(ENABLE_VULKAN "Enables Vulkan backend" ON)

 option(USE_DISCORD_PRESENCE "Enables Discord Rich Presence" OFF)

+# Default to a Release build
+get_property(IS_MULTI_CONFIG GLOBAL PROPERTY GENERATOR_IS_MULTI_CONFIG)
+if (NOT IS_MULTI_CONFIG AND NOT CMAKE_BUILD_TYPE)
+    set(CMAKE_BUILD_TYPE "Release" CACHE STRING "Choose the type of build, options are: Debug Release RelWithDebInfo MinSizeRel." FORCE)
+    message(STATUS "Defaulting to a Release build")
+endif()
+
 if(EXISTS ${PROJECT_SOURCE_DIR}/hooks/pre-commit AND NOT EXISTS ${PROJECT_SOURCE_DIR}/.git/hooks/pre-commit)
    message(STATUS "Copying pre-commit hook")
    file(COPY hooks/pre-commit
@@ -53,7 +60,6 @@ endfunction()
 if(EXISTS ${PROJECT_SOURCE_DIR}/.gitmodules)
    check_submodules_present()
 endif()
-
 configure_file(${PROJECT_SOURCE_DIR}/dist/compatibility_list/compatibility_list.qrc
               ${PROJECT_BINARY_DIR}/dist/compatibility_list/compatibility_list.qrc
               COPYONLY)
@@ -115,63 +121,217 @@ message(STATUS "Target architecture: ${ARCHITECTURE}")
 set(CMAKE_CXX_STANDARD 17)
 set(CMAKE_CXX_STANDARD_REQUIRED ON)

-# System imported libraries
-# ======================
-
-find_package(Boost 1.66.0 QUIET)
-if (NOT Boost_FOUND)
-    message(STATUS "Boost 1.66.0 or newer not found, falling back to externals")
-
-    set(BOOST_ROOT "${PROJECT_SOURCE_DIR}/externals/boost")
-    set(Boost_NO_SYSTEM_PATHS OFF)
-    find_package(Boost QUIET REQUIRED)
-endif()
-
 # Output binaries to bin/
 set(CMAKE_RUNTIME_OUTPUT_DIRECTORY ${PROJECT_BINARY_DIR}/bin)

+# System imported libraries
+# If not found, download any missing through Conan
+# =======================================================================
+set(CONAN_CMAKE_SILENT_OUTPUT TRUE)
+set(CMAKE_FIND_PACKAGE_PREFER_CONFIG TRUE)
+if (YUZU_CONAN_INSTALLED)
+    if (IS_MULTI_CONFIG)
+        include(${CMAKE_BINARY_DIR}/conanbuildinfo_multi.cmake)
+    else()
+        include(${CMAKE_BINARY_DIR}/conanbuildinfo.cmake)
+    endif()
+    list(APPEND CMAKE_MODULE_PATH "${CMAKE_BINARY_DIR}")
+    list(APPEND CMAKE_PREFIX_PATH "${CMAKE_BINARY_DIR}")
+    conan_basic_setup()
+    message(STATUS "Adding conan installed libraries to the search path")
+endif()
+
+macro(yuzu_find_packages)
+    set(options FORCE_REQUIRED)
+    cmake_parse_arguments(FN "${options}" "" "" ${ARGN})
+
+    # Cmake has a *serious* lack of 2D array or associative array...
+    # Capitalization matters here. We need the naming to match the generated paths from Conan
+    set(REQUIRED_LIBS
+    #    Cmake Pkg Prefix  Version     Conan Pkg
+        "Boost             1.71        boost/1.72.0"
+        "Catch2            2.11        catch2/2.11.0"
+        "fmt               6.2         fmt/6.2.0"
+    # can't use until https://github.com/bincrafters/community/issues/1173
+        #"libzip            1.5         libzip/1.5.2@bincrafters/stable"
+        "lz4               1.8         lz4/1.9.2"
+        "nlohmann_json     3.7         nlohmann_json/3.7.3"
+    # we need to be careful as the version check might be broken https://github.com/xiph/opus/issues/110
+        "opus              1.3         opus/1.3.1"
+        "ZLIB              1.2         zlib/1.2.11"
+        "zstd              1.4         zstd/1.4.4"
+    )
+
+    foreach(PACKAGE ${REQUIRED_LIBS})
+        string(REGEX REPLACE "[ \t\r\n]+" ";" PACKAGE_SPLIT ${PACKAGE})
+        list(GET PACKAGE_SPLIT 0 PACKAGE_PREFIX)
+        list(GET PACKAGE_SPLIT 1 PACKAGE_VERSION)
+        list(GET PACKAGE_SPLIT 2 PACKAGE_CONAN)
+        # This function is called twice, once to check if the packages exist on the system already
+        # and a second time to check if conan installed them properly. The second check passes in FORCE_REQUIRED
+        if (NOT ${PACKAGE_PREFIX}_FOUND)
+            if (FN_FORCE_REQUIRED)
+                find_package(${PACKAGE_PREFIX} ${PACKAGE_VERSION} REQUIRED)
+            else()
+                find_package(${PACKAGE_PREFIX} ${PACKAGE_VERSION})
+            endif()
+        endif()
+        if (NOT ${PACKAGE_PREFIX}_FOUND)
+            list(APPEND CONAN_REQUIRED_LIBS ${PACKAGE_CONAN})
+        else()
+            # Set a legacy findPackage.cmake style PACKAGE_LIBRARIES variable for subprojects that rely on this
+            set(${PACKAGE_PREFIX}_LIBRARIES "${PACKAGE_PREFIX}::${PACKAGE_PREFIX}")
+        endif()
+    endforeach()
+    unset(FN_FORCE_REQUIRED)
+endmacro()
+
+# Attempt to locate any packages that are required and report the missing ones in CONAN_REQUIRED_LIBS
+yuzu_find_packages()
+
+# Qt5 requires that we find components, so it doesn't fit our pretty little find package function
+if(ENABLE_QT)
+    # We want to load the generated conan qt config so that we get the QT_ROOT var so that we can use the official
+    # Qt5Config inside the root folder instead of the conan generated one.
+    if(EXISTS ${CMAKE_BINARY_DIR}/qtConfig.cmake)
+        include(${CMAKE_BINARY_DIR}/qtConfig.cmake)
+        list(APPEND CMAKE_MODULE_PATH "${CONAN_QT_ROOT_RELEASE}")
+        list(APPEND CMAKE_PREFIX_PATH "${CONAN_QT_ROOT_RELEASE}")
+    endif()
+    # Workaround for an issue where conan tries to build Qt from scratch instead of download prebuilt binaries
+    set(QT_PREFIX_HINT)
+    if(YUZU_USE_BUNDLED_QT)
+        if ((MSVC_VERSION GREATER_EQUAL 1910 AND MSVC_VERSION LESS 1930) AND ARCHITECTURE_x86_64)
+            set(QT_VER qt-5.12.0-msvc2017_64)
+        else()
+            message(FATAL_ERROR "No bundled Qt binaries for your toolchain. Disable YUZU_USE_BUNDLED_QT and provide your own.")
+        endif()
+
+        if (DEFINED QT_VER)
+            download_bundled_external("qt/" ${QT_VER} QT_PREFIX)
+        endif()
+
+        set(QT_PREFIX_HINT HINTS "${QT_PREFIX}")
+    endif()
+    find_package(Qt5 5.9 COMPONENTS Widgets OpenGL ${QT_PREFIX_HINT})
+    if (NOT Qt5_FOUND)
+        list(APPEND CONAN_REQUIRED_LIBS "qt/5.14.1@bincrafters/stable")
+    endif()
+endif()
+# find SDL2 exports a bunch of variables that are needed, so its easier to do this outside of the yuzu_find_package
+if(ENABLE_SDL2)
+    if(EXISTS ${CMAKE_BINARY_DIR}/sdl2Config.cmake)
+        include(${CMAKE_BINARY_DIR}/sdl2Config.cmake)
+        list(APPEND CMAKE_MODULE_PATH "${CONAN_SDL2_ROOT_RELEASE}")
+        list(APPEND CMAKE_PREFIX_PATH "${CONAN_SDL2_ROOT_RELEASE}")
+    endif()
+    find_package(SDL2)
+    if (NOT SDL2_FOUND)
+        # otherwise add this to the list of libraries to install
+        list(APPEND CONAN_REQUIRED_LIBS "sdl2/2.0.12@bincrafters/stable")
+    endif()
+endif()
+
+# Install any missing dependencies with conan install
+if (CONAN_REQUIRED_LIBS)
+    message(STATUS "Packages ${CONAN_REQUIRED_LIBS} not found!")
+    # Use Conan to fetch the libraries that aren't found
+    # Download conan.cmake automatically, you can also just copy the conan.cmake file
+    if(NOT EXISTS "${CMAKE_BINARY_DIR}/conan.cmake")
+        message(STATUS "Downloading conan.cmake from https://github.com/conan-io/cmake-conan")
+        file(DOWNLOAD "https://github.com/conan-io/cmake-conan/raw/v0.15/conan.cmake"
+                        "${CMAKE_BINARY_DIR}/conan.cmake")
+    endif()
+    include(${CMAKE_BINARY_DIR}/conan.cmake)
+
+    set(CONAN_LIB_OPTIONS
+        libzip:with_openssl=False
+        libzip:enable_windows_crypto=False
+    )
+    conan_check(VERSION 1.24.0 REQUIRED)
+    # Add the bincrafters remote
+    conan_add_remote(NAME bincrafters
+                    URL https://api.bintray.com/conan/bincrafters/public-conan)
+
+    # Manually add iconv to fix a dep conflict between qt and sdl2
+    # We don't need to add it through find_package or anything since the other two can find it just fine
+    if ("${CONAN_REQUIRED_LIBS}" MATCHES "qt" AND "${CONAN_REQUIRED_LIBS}" MATCHES "sdl")
+        list(APPEND CONAN_REQUIRED_LIBS "libiconv/1.16")
+    endif()
+    if (IS_MULTI_CONFIG)
+        conan_cmake_run(REQUIRES ${CONAN_REQUIRED_LIBS}
+                        OPTIONS ${CONAN_LIB_OPTIONS}
+                        BUILD missing
+                        CONFIGURATION_TYPES "Release;Debug"
+                        GENERATORS cmake_multi cmake_find_package_multi)
+        include(${CMAKE_BINARY_DIR}/conanbuildinfo_multi.cmake)
+    else()
+        conan_cmake_run(REQUIRES ${CONAN_REQUIRED_LIBS}
+                        OPTIONS ${CONAN_LIB_OPTIONS}
+                        BUILD missing
+                        GENERATORS cmake cmake_find_package_multi)
+        include(${CMAKE_BINARY_DIR}/conanbuildinfo.cmake)
+    endif()
+    list(APPEND CMAKE_MODULE_PATH "${CMAKE_BINARY_DIR}")
+    list(APPEND CMAKE_PREFIX_PATH "${CMAKE_BINARY_DIR}")
+    conan_basic_setup()
+
+    set(YUZU_CONAN_INSTALLED TRUE CACHE BOOL "If true, the following builds will add conan to the lib search path" FORCE)
+
+    # Now that we've installed what we are missing, try to locate them again,
+    # this time with required, so we bail if its not found.
+    yuzu_find_packages(FORCE_REQUIRED)
+
+    # Due to issues with variable scopes in functions, we need to also find_package(qt5) outside of the function
+    if(ENABLE_QT)
+        list(APPEND CMAKE_MODULE_PATH "${CONAN_QT_ROOT_RELEASE}")
+        list(APPEND CMAKE_PREFIX_PATH "${CONAN_QT_ROOT_RELEASE}")
+        find_package(Qt5 5.9 REQUIRED COMPONENTS Widgets OpenGL)
+        if (YUZU_USE_QT_WEB_ENGINE)
+            find_package(Qt5 REQUIRED COMPONENTS WebEngineCore WebEngineWidgets)
+        endif()
+    endif()
+    if(ENABLE_SDL2)
+        list(APPEND CMAKE_MODULE_PATH "${CONAN_SDL2_ROOT_RELEASE}")
+        list(APPEND CMAKE_PREFIX_PATH "${CONAN_SDL2_ROOT_RELEASE}")
+        find_package(SDL2 REQUIRED)
+    endif()
+
+endif()
+
+# Reexport some targets that are named differently when using the upstream CmakeConfig vs the generated Conan config
+# In order to ALIAS targets to a new name, they first need to be IMPORTED_GLOBAL
+# Dynarmic checks for target `boost` and so we want to make sure it can find it through our system instead of using their external
+if (TARGET Boost::Boost)
+    set_target_properties(Boost::Boost PROPERTIES IMPORTED_GLOBAL TRUE)
+    add_library(Boost::boost ALIAS Boost::Boost)
+    add_library(boost ALIAS Boost::Boost)
+elseif (TARGET Boost::boost)
+    set_target_properties(Boost::boost PROPERTIES IMPORTED_GLOBAL TRUE)
+    add_library(boost ALIAS Boost::boost)
+endif()
+
+if (TARGET sdl2::sdl2)
+    # imported from the conan generated sdl2Config.cmake
+    set_target_properties(sdl2::sdl2 PROPERTIES IMPORTED_GLOBAL TRUE)
+    add_library(SDL2 ALIAS sdl2::sdl2)
+elseif(SDL2_FOUND)
+    # found through the system package manager
+    # Some installations don't set SDL2_LIBRARIES
+    if("${SDL2_LIBRARIES}" STREQUAL "")
+        message(WARNING "SDL2_LIBRARIES wasn't set, manually setting to SDL2::SDL2")
+        set(SDL2_LIBRARIES "SDL2::SDL2")
+    endif()
+
+    include_directories(SYSTEM ${SDL2_INCLUDE_DIRS})
+    add_library(SDL2 INTERFACE)
+    target_link_libraries(SDL2 INTERFACE "${SDL2_LIBRARIES}")
+endif()
+
 # Prefer the -pthread flag on Linux.
 set(THREADS_PREFER_PTHREAD_FLAG ON)
 find_package(Threads REQUIRED)

-if (ENABLE_SDL2)
-    if (YUZU_USE_BUNDLED_SDL2)
-        # Detect toolchain and platform
-        if ((MSVC_VERSION GREATER_EQUAL 1910 AND MSVC_VERSION LESS 1930) AND ARCHITECTURE_x86_64)
-            set(SDL2_VER "SDL2-2.0.8")
-        else()
-            message(FATAL_ERROR "No bundled SDL2 binaries for your toolchain. Disable YUZU_USE_BUNDLED_SDL2 and provide your own.")
-        endif()
-
-        if (DEFINED SDL2_VER)
-            download_bundled_external("sdl2/" ${SDL2_VER} SDL2_PREFIX)
-        endif()
-
-        set(SDL2_FOUND YES)
-        set(SDL2_INCLUDE_DIR "${SDL2_PREFIX}/include" CACHE PATH "Path to SDL2 headers")
-        set(SDL2_LIBRARY "${SDL2_PREFIX}/lib/x64/SDL2.lib" CACHE PATH "Path to SDL2 library")
-        set(SDL2_DLL_DIR "${SDL2_PREFIX}/lib/x64/" CACHE PATH "Path to SDL2.dll")
-
-        add_library(SDL2 INTERFACE)
-        target_link_libraries(SDL2 INTERFACE "${SDL2_LIBRARY}")
-        target_include_directories(SDL2 INTERFACE "${SDL2_INCLUDE_DIR}")
-    else()
-        find_package(SDL2 REQUIRED)
-
-        # Some installations don't set SDL2_LIBRARIES
-        if("${SDL2_LIBRARIES}" STREQUAL "")
-            message(WARNING "SDL2_LIBRARIES wasn't set, manually setting to SDL2::SDL2")
-            set(SDL2_LIBRARIES "SDL2::SDL2")
-        endif()
-
-        include_directories(SYSTEM ${SDL2_INCLUDE_DIRS})
-        add_library(SDL2 INTERFACE)
-        target_link_libraries(SDL2 INTERFACE "${SDL2_LIBRARIES}")
-    endif()
-else()
-    set(SDL2_FOUND NO)
-endif()
-
 # If unicorn isn't found, msvc -> download bundled unicorn; everyone else -> build external
 if (YUZU_USE_BUNDLED_UNICORN)
    if (MSVC)
@@ -212,8 +372,12 @@ if (YUZU_USE_BUNDLED_UNICORN)
        find_package(PythonInterp 2.7 REQUIRED)

        if (MINGW)
+            # Intentionally call the unicorn makefile directly instead of using make.sh so that we can override the
+            # UNAME_S makefile variable to MINGW. This way we don't have to hack at the uname binary to build
+            # Additionally, overriding DO_WINDOWS_EXPORT prevents unicorn from patching the static unicorn.a by using msvc and cmd,
+            # which are both things we don't have in a mingw cross compiling environment.
            add_custom_command(OUTPUT ${LIBUNICORN_LIBRARY}
-                COMMAND ${CMAKE_COMMAND} -E env UNICORN_ARCHS="aarch64" PYTHON="${PYTHON_EXECUTABLE}" /bin/sh make.sh cross-win64
+                COMMAND ${CMAKE_COMMAND} -E env UNICORN_ARCHS="aarch64" PYTHON="${PYTHON_EXECUTABLE}" CC=x86_64-w64-mingw32-gcc AR=x86_64-w64-mingw32-gcc-ar RANLIB=x86_64-w64-mingw32-gcc-ranlib make UNAME_S=MINGW DO_WINDOWS_EXPORT=0
                WORKING_DIRECTORY ${UNICORN_PREFIX}
            )
        else()
@@ -243,32 +407,6 @@ else()
    message(FATAL_ERROR "Could not find or build unicorn which is required.")
 endif()

-if (ENABLE_QT)
-    if (YUZU_USE_BUNDLED_QT)
-        if ((MSVC_VERSION GREATER_EQUAL 1910 AND MSVC_VERSION LESS 1930) AND ARCHITECTURE_x86_64)
-            set(QT_VER qt-5.12.0-msvc2017_64)
-        else()
-            message(FATAL_ERROR "No bundled Qt binaries for your toolchain. Disable YUZU_USE_BUNDLED_QT and provide your own.")
-        endif()
-
-        if (DEFINED QT_VER)
-            download_bundled_external("qt/" ${QT_VER} QT_PREFIX)
-        endif()
-
-        set(QT_PREFIX_HINT HINTS "${QT_PREFIX}")
-    else()
-        # Passing an empty HINTS seems to cause default system paths to get ignored in CMake 2.8 so
-        # make sure to not pass anything if we don't have one.
-        set(QT_PREFIX_HINT)
-    endif()
-
-    find_package(Qt5 REQUIRED COMPONENTS Widgets OpenGL ${QT_PREFIX_HINT})
-
-    if (YUZU_USE_QT_WEB_ENGINE)
-        find_package(Qt5 REQUIRED COMPONENTS WebEngineCore WebEngineWidgets ${QT_PREFIX_HINT})
-    endif ()
-endif()
-
 # Platform-specific library requirements
 # ======================================

@@ -299,7 +437,7 @@ find_program(CLANG_FORMAT
    PATHS ${PROJECT_BINARY_DIR}/externals)
 # if find_program doesn't find it, try to download from externals
 if (NOT CLANG_FORMAT)
-    if (WIN32)
+    if (WIN32 AND NOT CMAKE_CROSSCOMPILING)
        message(STATUS "Clang format not found! Downloading...")
        set(CLANG_FORMAT "${PROJECT_BINARY_DIR}/externals/clang-format${CLANG_FORMAT_POSTFIX}.exe")
        file(DOWNLOAD
@@ -51,6 +51,8 @@ endif()
 # The variable SRC_DIR must be passed into the script (since it uses the current build directory for all values of CMAKE_*_DIR)
 set(VIDEO_CORE "${SRC_DIR}/src/video_core")
 set(HASH_FILES
+    "${VIDEO_CORE}/renderer_opengl/gl_arb_decompiler.cpp"
+    "${VIDEO_CORE}/renderer_opengl/gl_arb_decompiler.h"
    "${VIDEO_CORE}/renderer_opengl/gl_shader_cache.cpp"
    "${VIDEO_CORE}/renderer_opengl/gl_shader_cache.h"
    "${VIDEO_CORE}/renderer_opengl/gl_shader_decompiler.cpp"
@@ -673,10 +673,6 @@ QTabWidget::pane {
    border-bottom-left-radius: 2px;
 }

-QTabWidget::tab-bar {
-    overflow: visible;
-}
-
 QTabBar {
    qproperty-drawBase: 0;
    border-radius: 3px;
@@ -1,30 +1,20 @@
 # Definitions for all external bundled libraries

-set(CMAKE_MODULE_PATH ${CMAKE_MODULE_PATH} ${PROJECT_SOURCE_DIR}/CMakeModules)
+list(APPEND CMAKE_MODULE_PATH "${PROJECT_SOURCE_DIR}/CMakeModules")
+list(APPEND CMAKE_MODULE_PATH "${PROJECT_SOURCE_DIR}/externals/find-modules")
 include(DownloadExternals)

-# pkgconfig -- it is used to find shared libraries without cmake modules on linux systems
-find_package(PkgConfig)
-if (NOT PkgConfig_FOUND)
-    function(pkg_check_modules)
-        # STUB
-    endfunction()
+# xbyak
+if (ARCHITECTURE_x86 OR ARCHITECTURE_x86_64)
+    add_library(xbyak INTERFACE)
+    target_include_directories(xbyak SYSTEM INTERFACE ./xbyak/xbyak)
+    target_compile_definitions(xbyak INTERFACE XBYAK_NO_OP_NAMES)
 endif()

 # Catch
 add_library(catch-single-include INTERFACE)
 target_include_directories(catch-single-include INTERFACE catch/single_include)

-# libfmt
-pkg_check_modules(FMT IMPORTED_TARGET GLOBAL fmt>=6.2.0)
-if (FMT_FOUND)
-    add_library(fmt::fmt ALIAS PkgConfig::FMT)
-else()
-    message(STATUS "fmt 6.2.0 or newer not found, falling back to externals")
-    add_subdirectory(fmt)
-    add_library(fmt::fmt ALIAS fmt)
-endif()
-
 # Dynarmic
 if (ARCHITECTURE_x86_64)
    set(DYNARMIC_TESTS OFF)
@@ -43,17 +33,6 @@ add_subdirectory(glad)
 # inih
 add_subdirectory(inih)

-# lz4
-pkg_check_modules(LIBLZ4 IMPORTED_TARGET GLOBAL liblz4>=1.8.0)
-if (LIBLZ4_FOUND)
-    add_library(lz4_static ALIAS PkgConfig::LIBLZ4)
-else()
-    message(STATUS "liblz4 1.8.0 or newer not found, falling back to externals")
-    set(LZ4_BUNDLED_MODE ON)
-    add_subdirectory(lz4/contrib/cmake_unofficial EXCLUDE_FROM_ALL)
-    target_include_directories(lz4_static INTERFACE ./lz4/lib)
-endif()
-
 # mbedtls
 add_subdirectory(mbedtls EXCLUDE_FROM_ALL)
 target_include_directories(mbedtls PUBLIC ./mbedtls/include)
@@ -66,29 +45,9 @@ target_include_directories(microprofile INTERFACE ./microprofile)
 add_library(unicorn-headers INTERFACE)
 target_include_directories(unicorn-headers INTERFACE ./unicorn/include)

-# Zstandard
-pkg_check_modules(LIBZSTD IMPORTED_TARGET GLOBAL libzstd>=1.3.8)
-if (LIBZSTD_FOUND)
-    add_library(libzstd_static ALIAS PkgConfig::LIBZSTD)
-else()
-    message(STATUS "libzstd 1.3.8 or newer not found, falling back to externals")
-    add_subdirectory(zstd/build/cmake EXCLUDE_FROM_ALL)
-    target_include_directories(libzstd_static INTERFACE ./zstd/lib)
-endif()
-
 # SoundTouch
 add_subdirectory(soundtouch)

-# Opus
-pkg_check_modules(OPUS IMPORTED_TARGET GLOBAL opus>=1.3.1)
-if (OPUS_FOUND)
-    add_library(opus ALIAS PkgConfig::OPUS)
-else()
-    message(STATUS "opus 1.3.1 or newer not found, falling back to externals")
-    add_subdirectory(opus)
-    target_include_directories(opus INTERFACE ./opus/include)
-endif()
-
 # Cubeb
 if(ENABLE_CUBEB)
    set(BUILD_TESTS OFF CACHE BOOL "")
@@ -106,36 +65,22 @@ if (ENABLE_VULKAN)
    add_subdirectory(sirit)
 endif()

-# zlib
-find_package(ZLIB 1.2.11)
-if (NOT ZLIB_FOUND)
-    message(STATUS "zlib 1.2.11 or newer not found, falling back to externals")
-    add_subdirectory(zlib EXCLUDE_FROM_ALL)
-    set(ZLIB_LIBRARIES z)
-endif()
-
 # libzip
-pkg_check_modules(LIBZIP IMPORTED_TARGET GLOBAL libzip>=1.5.3)
-if (LIBZIP_FOUND)
-    add_library(zip ALIAS PkgConfig::LIBZIP)
-else()
-    message(STATUS "libzip 1.5.3 or newer not found, falling back to externals")
+find_package(Libzip 1.5)
+if (NOT LIBZIP_FOUND)
+    message(STATUS "libzip 1.5 or newer not found, falling back to externals")
    add_subdirectory(libzip EXCLUDE_FROM_ALL)
 endif()

 if (ENABLE_WEB_SERVICE)
    # LibreSSL
-    find_package(OpenSSL COMPONENTS Crypto SSL)
-    if (NOT OpenSSL_FOUND)
-        message(STATUS "OpenSSL not found, falling back to externals")
-        set(LIBRESSL_SKIP_INSTALL ON CACHE BOOL "")
-        add_subdirectory(libressl EXCLUDE_FROM_ALL)
-        target_include_directories(ssl INTERFACE ./libressl/include)
-        target_compile_definitions(ssl PRIVATE -DHAVE_INET_NTOP)
-        get_directory_property(OPENSSL_LIBRARIES
-            DIRECTORY libressl
-            DEFINITION OPENSSL_LIBS)
-    endif()
+    set(LIBRESSL_SKIP_INSTALL ON CACHE BOOL "")
+    add_subdirectory(libressl EXCLUDE_FROM_ALL)
+    target_include_directories(ssl INTERFACE ./libressl/include)
+    target_compile_definitions(ssl PRIVATE -DHAVE_INET_NTOP)
+    get_directory_property(OPENSSL_LIBRARIES
+        DIRECTORY libressl
+        DEFINITION OPENSSL_LIBS)

    # lurlparser
    add_subdirectory(lurlparser EXCLUDE_FROM_ALL)
@@ -145,8 +90,4 @@ if (ENABLE_WEB_SERVICE)
    target_include_directories(httplib INTERFACE ./httplib)
    target_compile_definitions(httplib INTERFACE -DCPPHTTPLIB_OPENSSL_SUPPORT)
    target_link_libraries(httplib INTERFACE ${OPENSSL_LIBRARIES})
-
-    # JSON
-    add_library(json-headers INTERFACE)
-    target_include_directories(json-headers INTERFACE ./json)
 endif()
@@ -0,0 +1,49 @@
+
+find_package(PkgConfig QUIET)
+pkg_check_modules(PC_Catch2 QUIET Catch2)
+
+find_path(Catch2_INCLUDE_DIR
+  NAMES catch.hpp
+  PATHS ${PC_Catch2_INCLUDE_DIRS} ${CONAN_CATCH2_ROOT}
+  PATH_SUFFIXES catch2
+)
+
+if(Catch2_INCLUDE_DIR)
+  file(STRINGS "${Catch2_INCLUDE_DIR}/catch.hpp" _Catch2_version_lines
+    REGEX "#define[ \t]+CATCH_VERSION_(MAJOR|MINOR|PATCH)")
+  string(REGEX REPLACE ".*CATCH_VERSION_MAJOR +\([0-9]+\).*" "\\1" _Catch2_version_major "${_Catch2_version_lines}")
+  string(REGEX REPLACE ".*CATCH_VERSION_MINOR +\([0-9]+\).*" "\\1" _Catch2_version_minor "${_Catch2_version_lines}")
+  string(REGEX REPLACE ".*CATCH_VERSION_PATCH +\([0-9]+\).*" "\\1" _Catch2_version_patch "${_Catch2_version_lines}")
+  set(Catch2_VERSION "${_Catch2_version_major}.${_Catch2_version_minor}.${_Catch2_version_patch}")
+  unset(_Catch2_version_major)
+  unset(_Catch2_version_minor)
+  unset(_Catch2_version_patch)
+  unset(_Catch2_version_lines)
+endif()
+
+include(FindPackageHandleStandardArgs)
+find_package_handle_standard_args(Catch2
+  FOUND_VAR Catch2_FOUND
+  REQUIRED_VARS
+    Catch2_INCLUDE_DIR
+    Catch2_VERSION
+  VERSION_VAR Catch2_VERSION
+)
+
+if(Catch2_FOUND)
+  set(Catch2_INCLUDE_DIRS ${Catch2_INCLUDE_DIR})
+  set(Catch2_DEFINITIONS ${PC_Catch2_CFLAGS_OTHER})
+endif()
+
+if(Catch2_FOUND AND NOT TARGET Catch2::Catch2)
+  add_library(Catch2::Catch2 UNKNOWN IMPORTED)
+  set_target_properties(Catch2::Catch2 PROPERTIES
+    IMPORTED_LOCATION "${Catch2_LIBRARY}"
+    INTERFACE_COMPILE_OPTIONS "${PC_Catch2_CFLAGS_OTHER}"
+    INTERFACE_INCLUDE_DIRECTORIES "${Catch2_INCLUDE_DIR}"
+  )
+endif()
+
+mark_as_advanced(
+    Catch2_INCLUDE_DIR
+)
@@ -0,0 +1,72 @@
+
+find_package(PkgConfig QUIET)
+pkg_check_modules(PC_LIBZIP QUIET libzip)
+
+find_path(LIBZIP_INCLUDE_DIR
+  NAMES zip.h
+  PATHS ${PC_LIBZIP_INCLUDE_DIRS}
+  "$ENV{LIB_DIR}/include"
+  "$ENV{INCLUDE}"
+  /usr/local/include
+  /usr/include
+)
+find_path(LIBZIP_INCLUDE_DIR_ZIPCONF
+  NAMES zipconf.h
+  HINTS ${PC_LIBZIP_INCLUDE_DIRS}
+  "$ENV{LIB_DIR}/include"
+  "$ENV{LIB_DIR}/lib/libzip/include"
+  "$ENV{LIB}/lib/libzip/include"
+  /usr/local/lib/libzip/include
+  /usr/lib/libzip/include
+  /usr/local/include
+  /usr/include
+  "$ENV{INCLUDE}"
+)
+find_library(LIBZIP_LIBRARY
+  NAMES zip
+  PATHS ${PC_LIBZIP_LIBRARY_DIRS}
+  "$ENV{LIB_DIR}/lib" "$ENV{LIB}" /usr/local/lib /usr/lib
+)
+
+if (LIBZIP_INCLUDE_DIR_ZIPCONF)
+  FILE(READ "${LIBZIP_INCLUDE_DIR_ZIPCONF}/zipconf.h" _LIBZIP_VERSION_CONTENTS)
+  if (_LIBZIP_VERSION_CONTENTS)
+    STRING(REGEX REPLACE ".*#define LIBZIP_VERSION \"([0-9.]+)\".*" "\\1" LIBZIP_VERSION "${_LIBZIP_VERSION_CONTENTS}")
+  endif()
+  unset(_LIBZIP_VERSION_CONTENTS)
+endif()
+
+set(LIBZIP_VERSION ${LIBZIP_VERSION} CACHE STRING "Version number of libzip")
+
+include(FindPackageHandleStandardArgs)
+find_package_handle_standard_args(Libzip
+  FOUND_VAR LIBZIP_FOUND
+  REQUIRED_VARS
+    LIBZIP_LIBRARY
+    LIBZIP_INCLUDE_DIR
+    LIBZIP_INCLUDE_DIR_ZIPCONF
+    LIBZIP_VERSION
+  VERSION_VAR LIBZIP_VERSION
+)
+
+if(LIBZIP_FOUND)
+  set(LIBZIP_LIBRARIES ${LIBZIP_LIBRARY})
+  set(LIBZIP_INCLUDE_DIRS ${LIBZIP_INCLUDE_DIR})
+  set(LIBZIP_DEFINITIONS ${PC_LIBZIP_CFLAGS_OTHER})
+endif()
+
+if(LIBZIP_FOUND AND NOT TARGET libzip::libzip)
+  add_library(libzip::libzip UNKNOWN IMPORTED)
+  set_target_properties(libzip::libzip PROPERTIES
+    IMPORTED_LOCATION "${LIBZIP_LIBRARY}"
+    INTERFACE_COMPILE_OPTIONS "${PC_LIBZIP_CFLAGS_OTHER}"
+    INTERFACE_INCLUDE_DIRECTORIES "${LIBZIP_INCLUDE_DIR}"
+  )
+endif()
+
+mark_as_advanced(
+    LIBZIP_INCLUDE_DIR
+    LIBZIP_INCLUDE_DIR_ZIPCONF
+    LIBZIP_LIBRARY
+    LIBZIP_VERSION
+)
@@ -0,0 +1,69 @@
+
+find_package(PkgConfig QUIET)
+pkg_check_modules(PC_fmt QUIET fmt)
+
+find_path(fmt_INCLUDE_DIR
+  NAMES format.h
+  PATHS ${PC_fmt_INCLUDE_DIRS} ${CONAN_INCLUDE_DIRS_fmt}
+  PATH_SUFFIXES fmt
+)
+
+find_library(fmt_LIBRARY
+  NAMES fmt
+  PATHS ${PC_fmt_LIBRARY_DIRS} ${CONAN_LIB_DIRS_fmt}
+)
+
+if(fmt_INCLUDE_DIR)
+  set(_fmt_version_file "${fmt_INCLUDE_DIR}/core.h")
+  if(NOT EXISTS "${_fmt_version_file}")
+    set(_fmt_version_file "${fmt_INCLUDE_DIR}/format.h")
+  endif()
+  if(EXISTS "${_fmt_version_file}")
+    # parse "#define FMT_VERSION 60200" to 6.2.0
+    file(STRINGS "${_fmt_version_file}" fmt_VERSION_LINE
+      REGEX "^#define[ \t]+FMT_VERSION[ \t]+[0-9]+$")
+    string(REGEX REPLACE "^#define[ \t]+FMT_VERSION[ \t]+([0-9]+)$"
+      "\\1" fmt_VERSION "${fmt_VERSION_LINE}")
+    foreach(ver "fmt_VERSION_PATCH" "fmt_VERSION_MINOR" "fmt_VERSION_MAJOR")
+      math(EXPR ${ver} "${fmt_VERSION} % 100")
+      math(EXPR fmt_VERSION "(${fmt_VERSION} - ${${ver}}) / 100")
+    endforeach()
+    set(fmt_VERSION
+      "${fmt_VERSION_MAJOR}.${fmt_VERSION_MINOR}.${fmt_VERSION_PATCH}")
+  endif()
+  unset(_fmt_version_file)
+  unset(fmt_VERSION_LINE)
+  unset(fmt_VERSION_MAJOR)
+  unset(fmt_VERSION_MINOR)
+  unset(fmt_VERSION_PATCH)
+endif()
+
+include(FindPackageHandleStandardArgs)
+find_package_handle_standard_args(fmt
+  FOUND_VAR fmt_FOUND
+  REQUIRED_VARS
+    fmt_LIBRARY
+    fmt_INCLUDE_DIR
+    fmt_VERSION
+  VERSION_VAR fmt_VERSION
+)
+
+if(fmt_FOUND)
+  set(fmt_LIBRARIES ${fmt_LIBRARY})
+  set(fmt_INCLUDE_DIRS ${fmt_INCLUDE_DIR})
+  set(fmt_DEFINITIONS ${PC_fmt_CFLAGS_OTHER})
+endif()
+
+if(fmt_FOUND AND NOT TARGET fmt::fmt)
+  add_library(fmt::fmt UNKNOWN IMPORTED)
+  set_target_properties(fmt::fmt PROPERTIES
+    IMPORTED_LOCATION "${fmt_LIBRARY}"
+    INTERFACE_COMPILE_OPTIONS "${PC_fmt_CFLAGS_OTHER}"
+    INTERFACE_INCLUDE_DIRECTORIES "${fmt_INCLUDE_DIR}"
+  )
+endif()
+
+mark_as_advanced(
+    fmt_INCLUDE_DIR
+    fmt_LIBRARY
+)
@@ -0,0 +1,54 @@
+
+find_package(PkgConfig QUIET)
+pkg_check_modules(PC_lz4 QUIET lz4)
+
+find_path(lz4_INCLUDE_DIR
+  NAMES lz4.h
+  PATHS ${PC_lz4_INCLUDE_DIRS}
+)
+find_library(lz4_LIBRARY
+  NAMES lz4
+  PATHS ${PC_lz4_LIBRARY_DIRS}
+)
+
+if(lz4_INCLUDE_DIR)
+  file(STRINGS "${lz4_INCLUDE_DIR}/lz4.h" _lz4_version_lines
+    REGEX "#define[ \t]+LZ4_VERSION_(MAJOR|MINOR|RELEASE)")
+  string(REGEX REPLACE ".*LZ4_VERSION_MAJOR *\([0-9]*\).*" "\\1" _lz4_version_major "${_lz4_version_lines}")
+  string(REGEX REPLACE ".*LZ4_VERSION_MINOR *\([0-9]*\).*" "\\1" _lz4_version_minor "${_lz4_version_lines}")
+  string(REGEX REPLACE ".*LZ4_VERSION_RELEASE *\([0-9]*\).*" "\\1" _lz4_version_release "${_lz4_version_lines}")
+  set(lz4_VERSION "${_lz4_version_major}.${_lz4_version_minor}.${_lz4_version_release}")
+  unset(_lz4_version_major)
+  unset(_lz4_version_minor)
+  unset(_lz4_version_release)
+  unset(_lz4_version_lines)
+endif()
+
+include(FindPackageHandleStandardArgs)
+find_package_handle_standard_args(lz4
+  FOUND_VAR lz4_FOUND
+  REQUIRED_VARS
+    lz4_LIBRARY
+    lz4_INCLUDE_DIR
+  VERSION_VAR lz4_VERSION
+)
+
+if(lz4_FOUND)
+  set(lz4_LIBRARIES ${lz4_LIBRARY})
+  set(lz4_INCLUDE_DIRS ${lz4_INCLUDE_DIR})
+  set(lz4_DEFINITIONS ${PC_lz4_CFLAGS_OTHER})
+endif()
+
+if(lz4_FOUND AND NOT TARGET lz4::lz4)
+  add_library(lz4::lz4 UNKNOWN IMPORTED)
+  set_target_properties(lz4::lz4 PROPERTIES
+    IMPORTED_LOCATION "${lz4_LIBRARY}"
+    INTERFACE_COMPILE_OPTIONS "${PC_lz4_CFLAGS_OTHER}"
+    INTERFACE_INCLUDE_DIRECTORIES "${lz4_INCLUDE_DIR}"
+  )
+endif()
+
+mark_as_advanced(
+    lz4_INCLUDE_DIR
+    lz4_LIBRARY
+)
@@ -0,0 +1,49 @@
+
+find_package(PkgConfig QUIET)
+pkg_check_modules(PC_nlohmann_json QUIET nlohmann_json)
+
+find_path(nlohmann_json_INCLUDE_DIR
+  NAMES json.hpp
+  PATHS ${PC_nlohmann_json_INCLUDE_DIRS}
+  PATH_SUFFIXES nlohmann
+)
+
+if(nlohmann_json_INCLUDE_DIR)
+  file(STRINGS "${nlohmann_json_INCLUDE_DIR}/json.hpp" _nlohmann_json_version_lines
+    REGEX "#define[ \t]+NLOHMANN_JSON_VERSION_(MAJOR|MINOR|PATCH)")
+  string(REGEX REPLACE ".*NLOHMANN_JSON_VERSION_MAJOR +\([0-9]+\).*" "\\1" _nlohmann_json_version_major "${_nlohmann_json_version_lines}")
+  string(REGEX REPLACE ".*NLOHMANN_JSON_VERSION_MINOR +\([0-9]+\).*" "\\1" _nlohmann_json_version_minor "${_nlohmann_json_version_lines}")
+  string(REGEX REPLACE ".*NLOHMANN_JSON_VERSION_PATCH +\([0-9]+\).*" "\\1" _nlohmann_json_version_patch "${_nlohmann_json_version_lines}")
+  set(nlohmann_json_VERSION "${_nlohmann_json_version_major}.${_nlohmann_json_version_minor}.${_nlohmann_json_version_patch}")
+  unset(_nlohmann_json_version_major)
+  unset(_nlohmann_json_version_minor)
+  unset(_nlohmann_json_version_patch)
+  unset(_nlohmann_json_version_lines)
+endif()
+
+include(FindPackageHandleStandardArgs)
+find_package_handle_standard_args(nlohmann_json
+  FOUND_VAR nlohmann_json_FOUND
+  REQUIRED_VARS
+    nlohmann_json_INCLUDE_DIR
+    nlohmann_json_VERSION
+  VERSION_VAR nlohmann_json_VERSION
+)
+
+if(nlohmann_json_FOUND)
+  set(nlohmann_json_INCLUDE_DIRS ${nlohmann_json_INCLUDE_DIR})
+  set(nlohmann_json_DEFINITIONS ${PC_nlohmann_json_CFLAGS_OTHER})
+endif()
+
+if(nlohmann_json_FOUND AND NOT TARGET nlohmann_json::nlohmann_json)
+  add_library(nlohmann_json::nlohmann_json UNKNOWN IMPORTED)
+  set_target_properties(nlohmann_json::nlohmann_json PROPERTIES
+    IMPORTED_LOCATION "${nlohmann_json_LIBRARY}"
+    INTERFACE_COMPILE_OPTIONS "${PC_nlohmann_json_CFLAGS_OTHER}"
+    INTERFACE_INCLUDE_DIRECTORIES "${nlohmann_json_INCLUDE_DIR}"
+  )
+endif()
+
+mark_as_advanced(
+    nlohmann_json_INCLUDE_DIR
+)
@@ -0,0 +1,42 @@
+
+find_package(PkgConfig QUIET)
+pkg_check_modules(PC_opus QUIET opus)
+
+find_path(opus_INCLUDE_DIR
+  NAMES opus.h
+  PATHS ${PC_opus_INCLUDE_DIRS}
+  PATH_SUFFIXES opus
+)
+find_library(opus_LIBRARY
+  NAMES opus
+  PATHS ${PC_opus_LIBRARY_DIRS}
+)
+
+include(FindPackageHandleStandardArgs)
+find_package_handle_standard_args(opus
+  FOUND_VAR opus_FOUND
+  REQUIRED_VARS
+    opus_LIBRARY
+    opus_INCLUDE_DIR
+  VERSION_VAR opus_VERSION
+)
+
+if(opus_FOUND)
+  set(Opus_LIBRARIES ${opus_LIBRARY})
+  set(Opus_INCLUDE_DIRS ${opus_INCLUDE_DIR})
+  set(Opus_DEFINITIONS ${PC_opus_CFLAGS_OTHER})
+endif()
+
+if(opus_FOUND AND NOT TARGET Opus::Opus)
+  add_library(Opus::Opus UNKNOWN IMPORTED)
+  set_target_properties(Opus::Opus PROPERTIES
+    IMPORTED_LOCATION "${opus_LIBRARY}"
+    INTERFACE_COMPILE_OPTIONS "${PC_opus_CFLAGS_OTHER}"
+    INTERFACE_INCLUDE_DIRECTORIES "${opus_INCLUDE_DIR}"
+  )
+endif()
+
+mark_as_advanced(
+    opus_INCLUDE_DIR
+    opus_LIBRARY
+)
@@ -0,0 +1,55 @@
+
+find_package(PkgConfig QUIET)
+pkg_check_modules(PC_zstd QUIET libzstd)
+
+find_path(zstd_INCLUDE_DIR
+  NAMES zstd.h
+  PATHS ${PC_zstd_INCLUDE_DIRS}
+)
+find_library(zstd_LIBRARY
+  NAMES zstd
+  PATHS ${PC_zstd_LIBRARY_DIRS}
+)
+
+if(zstd_INCLUDE_DIR)
+  file(STRINGS "${zstd_INCLUDE_DIR}/zstd.h" _zstd_version_lines
+    REGEX "#define[ \t]+ZSTD_VERSION_(MAJOR|MINOR|RELEASE)")
+  string(REGEX REPLACE ".*ZSTD_VERSION_MAJOR *\([0-9]*\).*" "\\1" _zstd_version_major "${_zstd_version_lines}")
+  string(REGEX REPLACE ".*ZSTD_VERSION_MINOR *\([0-9]*\).*" "\\1" _zstd_version_minor "${_zstd_version_lines}")
+  string(REGEX REPLACE ".*ZSTD_VERSION_RELEASE *\([0-9]*\).*" "\\1" _zstd_version_release "${_zstd_version_lines}")
+  set(zstd_VERSION "${_zstd_version_major}.${_zstd_version_minor}.${_zstd_version_release}")
+  unset(_zstd_version_major)
+  unset(_zstd_version_minor)
+  unset(_zstd_version_release)
+  unset(_zstd_version_lines)
+endif()
+
+include(FindPackageHandleStandardArgs)
+find_package_handle_standard_args(zstd
+  FOUND_VAR zstd_FOUND
+  REQUIRED_VARS
+    zstd_LIBRARY
+    zstd_INCLUDE_DIR
+    zstd_VERSION
+  VERSION_VAR zstd_VERSION
+)
+
+if(zstd_FOUND)
+  set(zstd_LIBRARIES ${zstd_LIBRARY})
+  set(zstd_INCLUDE_DIRS ${zstd_INCLUDE_DIR})
+  set(zstd_DEFINITIONS ${PC_zstd_CFLAGS_OTHER})
+endif()
+
+if(zstd_FOUND AND NOT TARGET zstd::zstd)
+  add_library(zstd::zstd UNKNOWN IMPORTED)
+  set_target_properties(zstd::zstd PROPERTIES
+    IMPORTED_LOCATION "${zstd_LIBRARY}"
+    INTERFACE_COMPILE_OPTIONS "${PC_zstd_CFLAGS_OTHER}"
+    INTERFACE_INCLUDE_DIRECTORIES "${zstd_INCLUDE_DIR}"
+  )
+endif()
+
+mark_as_advanced(
+    zstd_INCLUDE_DIR
+    zstd_LIBRARY
+)
@@ -1,9 +0,0 @@
-JSON for Modern C++
-===================
-
-v3.1.2
-
-This is a mirror providing the single required header file.
-
-The original repository can be found at:
-https://github.com/nlohmann/json/commit/d2dd27dc3b8472dbaa7d66f83619b3ebcd9185fe
@@ -1,254 +0,0 @@
-cmake_minimum_required(VERSION 3.8)
-
-project(opus)
-
-option(OPUS_STACK_PROTECTOR "Use stack protection" OFF)
-option(OPUS_USE_ALLOCA "Use alloca for stack arrays (on non-C99 compilers)" OFF)
-option(OPUS_CUSTOM_MODES "Enable non-Opus modes, e.g. 44.1 kHz & 2^n frames" OFF)
-option(OPUS_FIXED_POINT "Compile as fixed-point (for machines without a fast enough FPU)" OFF)
-option(OPUS_ENABLE_FLOAT_API "Compile with the floating point API (for machines with float library" ON)
-
-include(opus/opus_functions.cmake)
-
-if(OPUS_STACK_PROTECTOR)
-    if(NOT MSVC) # GC on by default on MSVC
-        check_and_set_flag(STACK_PROTECTION_STRONG -fstack-protector-strong)
-    endif()
-else()
-    if(MSVC)
-        check_and_set_flag(BUFFER_SECURITY_CHECK /GS-)
-    endif()
-endif()
-
-add_library(opus STATIC
-    # CELT sources
-    opus/celt/bands.c
-    opus/celt/celt.c
-    opus/celt/celt_decoder.c
-    opus/celt/celt_encoder.c
-    opus/celt/celt_lpc.c
-    opus/celt/cwrs.c
-    opus/celt/entcode.c
-    opus/celt/entdec.c
-    opus/celt/entenc.c
-    opus/celt/kiss_fft.c
-    opus/celt/laplace.c
-    opus/celt/mathops.c
-    opus/celt/mdct.c
-    opus/celt/modes.c
-    opus/celt/pitch.c
-    opus/celt/quant_bands.c
-    opus/celt/rate.c
-    opus/celt/vq.c
-
-    # SILK sources
-    opus/silk/A2NLSF.c
-    opus/silk/CNG.c
-    opus/silk/HP_variable_cutoff.c
-    opus/silk/LPC_analysis_filter.c
-    opus/silk/LPC_fit.c
-    opus/silk/LPC_inv_pred_gain.c
-    opus/silk/LP_variable_cutoff.c
-    opus/silk/NLSF2A.c
-    opus/silk/NLSF_VQ.c
-    opus/silk/NLSF_VQ_weights_laroia.c
-    opus/silk/NLSF_decode.c
-    opus/silk/NLSF_del_dec_quant.c
-    opus/silk/NLSF_encode.c
-    opus/silk/NLSF_stabilize.c
-    opus/silk/NLSF_unpack.c
-    opus/silk/NSQ.c
-    opus/silk/NSQ_del_dec.c
-    opus/silk/PLC.c
-    opus/silk/VAD.c
-    opus/silk/VQ_WMat_EC.c
-    opus/silk/ana_filt_bank_1.c
-    opus/silk/biquad_alt.c
-    opus/silk/bwexpander.c
-    opus/silk/bwexpander_32.c
-    opus/silk/check_control_input.c
-    opus/silk/code_signs.c
-    opus/silk/control_SNR.c
-    opus/silk/control_audio_bandwidth.c
-    opus/silk/control_codec.c
-    opus/silk/dec_API.c
-    opus/silk/decode_core.c
-    opus/silk/decode_frame.c
-    opus/silk/decode_indices.c
-    opus/silk/decode_parameters.c
-    opus/silk/decode_pitch.c
-    opus/silk/decode_pulses.c
-    opus/silk/decoder_set_fs.c
-    opus/silk/enc_API.c
-    opus/silk/encode_indices.c
-    opus/silk/encode_pulses.c
-    opus/silk/gain_quant.c
-    opus/silk/init_decoder.c
-    opus/silk/init_encoder.c
-    opus/silk/inner_prod_aligned.c
-    opus/silk/interpolate.c
-    opus/silk/lin2log.c
-    opus/silk/log2lin.c
-    opus/silk/pitch_est_tables.c
-    opus/silk/process_NLSFs.c
-    opus/silk/quant_LTP_gains.c
-    opus/silk/resampler.c
-    opus/silk/resampler_down2.c
-    opus/silk/resampler_down2_3.c
-    opus/silk/resampler_private_AR2.c
-    opus/silk/resampler_private_IIR_FIR.c
-    opus/silk/resampler_private_down_FIR.c
-    opus/silk/resampler_private_up2_HQ.c
-    opus/silk/resampler_rom.c
-    opus/silk/shell_coder.c
-    opus/silk/sigm_Q15.c
-    opus/silk/sort.c
-    opus/silk/stereo_LR_to_MS.c
-    opus/silk/stereo_MS_to_LR.c
-    opus/silk/stereo_decode_pred.c
-    opus/silk/stereo_encode_pred.c
-    opus/silk/stereo_find_predictor.c
-    opus/silk/stereo_quant_pred.c
-    opus/silk/sum_sqr_shift.c
-    opus/silk/table_LSF_cos.c
-    opus/silk/tables_LTP.c
-    opus/silk/tables_NLSF_CB_NB_MB.c
-    opus/silk/tables_NLSF_CB_WB.c
-    opus/silk/tables_gain.c
-    opus/silk/tables_other.c
-    opus/silk/tables_pitch_lag.c
-    opus/silk/tables_pulses_per_block.c
-
-    # Opus sources
-    opus/src/analysis.c
-    opus/src/mapping_matrix.c
-    opus/src/mlp.c
-    opus/src/mlp_data.c
-    opus/src/opus.c
-    opus/src/opus_decoder.c
-    opus/src/opus_encoder.c
-    opus/src/opus_multistream.c
-    opus/src/opus_multistream_decoder.c
-    opus/src/opus_multistream_encoder.c
-    opus/src/opus_projection_decoder.c
-    opus/src/opus_projection_encoder.c
-    opus/src/repacketizer.c
-)
-
-if (DEBUG)
-    target_sources(opus PRIVATE opus/silk/debug.c)
-endif()
-
-if (OPUS_FIXED_POINT)
-    target_sources(opus PRIVATE
-        opus/silk/fixed/LTP_analysis_filter_FIX.c
-        opus/silk/fixed/LTP_scale_ctrl_FIX.c
-        opus/silk/fixed/apply_sine_window_FIX.c
-        opus/silk/fixed/autocorr_FIX.c
-        opus/silk/fixed/burg_modified_FIX.c
-        opus/silk/fixed/corrMatrix_FIX.c
-        opus/silk/fixed/encode_frame_FIX.c
-        opus/silk/fixed/find_LPC_FIX.c
-        opus/silk/fixed/find_LTP_FIX.c
-        opus/silk/fixed/find_pitch_lags_FIX.c
-        opus/silk/fixed/find_pred_coefs_FIX.c
-        opus/silk/fixed/k2a_FIX.c
-        opus/silk/fixed/k2a_Q16_FIX.c
-        opus/silk/fixed/noise_shape_analysis_FIX.c
-        opus/silk/fixed/pitch_analysis_core_FIX.c
-        opus/silk/fixed/prefilter_FIX.c
-        opus/silk/fixed/process_gains_FIX.c
-        opus/silk/fixed/regularize_correlations_FIX.c
-        opus/silk/fixed/residual_energy16_FIX.c
-        opus/silk/fixed/residual_energy_FIX.c
-        opus/silk/fixed/schur64_FIX.c
-        opus/silk/fixed/schur_FIX.c
-        opus/silk/fixed/solve_LS_FIX.c
-        opus/silk/fixed/vector_ops_FIX.c
-        opus/silk/fixed/warped_autocorrelation_FIX.c
-    )
-else()
-    target_sources(opus PRIVATE
-        opus/silk/float/LPC_analysis_filter_FLP.c
-        opus/silk/float/LPC_inv_pred_gain_FLP.c
-        opus/silk/float/LTP_analysis_filter_FLP.c
-        opus/silk/float/LTP_scale_ctrl_FLP.c
-        opus/silk/float/apply_sine_window_FLP.c
-        opus/silk/float/autocorrelation_FLP.c
-        opus/silk/float/burg_modified_FLP.c
-        opus/silk/float/bwexpander_FLP.c
-        opus/silk/float/corrMatrix_FLP.c
-        opus/silk/float/encode_frame_FLP.c
-        opus/silk/float/energy_FLP.c
-        opus/silk/float/find_LPC_FLP.c
-        opus/silk/float/find_LTP_FLP.c
-        opus/silk/float/find_pitch_lags_FLP.c
-        opus/silk/float/find_pred_coefs_FLP.c
-        opus/silk/float/inner_product_FLP.c
-        opus/silk/float/k2a_FLP.c
-        opus/silk/float/noise_shape_analysis_FLP.c
-        opus/silk/float/pitch_analysis_core_FLP.c
-        opus/silk/float/process_gains_FLP.c
-        opus/silk/float/regularize_correlations_FLP.c
-        opus/silk/float/residual_energy_FLP.c
-        opus/silk/float/scale_copy_vector_FLP.c
-        opus/silk/float/scale_vector_FLP.c
-        opus/silk/float/schur_FLP.c
-        opus/silk/float/sort_FLP.c
-        opus/silk/float/warped_autocorrelation_FLP.c
-        opus/silk/float/wrappers_FLP.c
-    )
-endif()
-
-target_compile_definitions(opus PRIVATE OPUS_BUILD ENABLE_HARDENING)
-
-if(NOT MSVC)
-    if(MINGW)
-        target_compile_definitions(opus PRIVATE _FORTIFY_SOURCE=0)
-    else()
-        target_compile_definitions(opus PRIVATE _FORTIFY_SOURCE=2)
-    endif()
-endif()
-
-# It is strongly recommended to uncomment one of these VAR_ARRAYS: Use C99
-# variable-length arrays for stack allocation USE_ALLOCA: Use alloca() for stack
-# allocation If none is defined, then the fallback is a non-threadsafe global
-# array
-if(OPUS_USE_ALLOCA OR MSVC)
-    target_compile_definitions(opus PRIVATE USE_ALLOCA)
-else()
-    target_compile_definitions(opus PRIVATE VAR_ARRAYS)
-endif()
-
-if(OPUS_CUSTOM_MODES)
-    target_compile_definitions(opus PRIVATE CUSTOM_MODES)
-endif()
-
-if(NOT OPUS_ENABLE_FLOAT_API)
-    target_compile_definitions(opus PRIVATE DISABLE_FLOAT_API)
-endif()
-
-target_compile_definitions(opus
-PUBLIC
-    -DOPUS_VERSION="\\"1.3.1\\""
-
-PRIVATE
-    # Use C99 intrinsics to speed up float-to-int conversion
-    HAVE_LRINTF
-)
-
-if (FIXED_POINT)
-    target_compile_definitions(opus PRIVATE -DFIXED_POINT=1 -DDISABLE_FLOAT_API)
-endif()
-
-target_include_directories(opus
-PUBLIC
-    opus/include
-
-PRIVATE
-    opus/celt
-    opus/silk
-    opus/silk/fixed
-    opus/silk/float
-    opus/src
-)
@@ -1,81 +0,0 @@
-project(zlib C)
-
-include(CheckTypeSize)
-include(CheckFunctionExists)
-include(CheckIncludeFile)
-
-check_include_file(sys/types.h HAVE_SYS_TYPES_H)
-check_include_file(stdint.h    HAVE_STDINT_H)
-check_include_file(stddef.h    HAVE_STDDEF_H)
-
-# Check to see if we have large file support
-set(CMAKE_REQUIRED_DEFINITIONS -D_LARGEFILE64_SOURCE=1)
-# We add these other definitions here because CheckTypeSize.cmake
-# in CMake 2.4.x does not automatically do so and we want
-# compatibility with CMake 2.4.x.
-if(HAVE_SYS_TYPES_H)
-    list(APPEND CMAKE_REQUIRED_DEFINITIONS -DHAVE_SYS_TYPES_H)
-endif()
-if(HAVE_STDINT_H)
-    list(APPEND CMAKE_REQUIRED_DEFINITIONS -DHAVE_STDINT_H)
-endif()
-if(HAVE_STDDEF_H)
-    list(APPEND CMAKE_REQUIRED_DEFINITIONS -DHAVE_STDDEF_H)
-endif()
-check_type_size(off64_t OFF64_T)
-if(HAVE_OFF64_T)
-   add_definitions(-D_LARGEFILE64_SOURCE=1)
-endif()
-set(CMAKE_REQUIRED_DEFINITIONS) # clear variable
-
-# Check for fseeko
-check_function_exists(fseeko HAVE_FSEEKO)
-if(NOT HAVE_FSEEKO)
-    add_definitions(-DNO_FSEEKO)
-endif()
-
-# Check for unistd.h
-check_include_file(unistd.h HAVE_UNISTD_H)
-if(HAVE_UNISTD_H)
-    add_definitions(-DHAVE_UNISTD_H)
-endif()
-
-if(MSVC)
-    add_definitions(-D_CRT_SECURE_NO_DEPRECATE)
-    add_definitions(-D_CRT_NONSTDC_NO_DEPRECATE)
-endif()
-
-add_library(z STATIC
-    zlib/adler32.c
-    zlib/compress.c
-    zlib/crc32.c
-    zlib/crc32.h
-    zlib/deflate.c
-    zlib/deflate.h
-    zlib/gzclose.c
-    zlib/gzguts.h
-    zlib/gzlib.c
-    zlib/gzread.c
-    zlib/gzwrite.c
-    zlib/inffast.h
-    zlib/inffixed.h
-    zlib/inflate.c
-    zlib/inflate.h
-    zlib/infback.c
-    zlib/inftrees.c
-    zlib/inftrees.h
-    zlib/inffast.c
-    zlib/trees.c
-    zlib/trees.h
-    zlib/uncompr.c
-    zlib/zconf.h
-    zlib/zlib.h
-    zlib/zutil.c
-    zlib/zutil.h
-)
-add_library(ZLIB::ZLIB ALIAS z)
-
-target_include_directories(z
-PUBLIC
-    zlib/
-)
@@ -17,7 +17,7 @@ namespace AudioCore {

 constexpr u32 STREAM_SAMPLE_RATE{48000};
 constexpr u32 STREAM_NUM_CHANNELS{2};
-
+using VoiceChannelHolder = std::array<VoiceResourceInformation*, 6>;
 class AudioRenderer::VoiceState {
 public:
    bool IsPlaying() const {
@@ -37,9 +37,10 @@ public:
    }

    void SetWaveIndex(std::size_t index);
-    std::vector<s16> DequeueSamples(std::size_t sample_count, Core::Memory::Memory& memory);
+    std::vector<s16> DequeueSamples(std::size_t sample_count, Core::Memory::Memory& memory,
+                                    const VoiceChannelHolder& voice_resources);
    void UpdateState();
-    void RefreshBuffer(Core::Memory::Memory& memory);
+    void RefreshBuffer(Core::Memory::Memory& memory, const VoiceChannelHolder& voice_resources);

 private:
    bool is_in_use{};
@@ -79,7 +80,7 @@ AudioRenderer::AudioRenderer(Core::Timing::CoreTiming& core_timing, Core::Memory
                             std::shared_ptr<Kernel::WritableEvent> buffer_event,
                             std::size_t instance_number)
    : worker_params{params}, buffer_event{buffer_event}, voices(params.voice_count),
-      effects(params.effect_count), memory{memory_} {
+      voice_resources(params.voice_count), effects(params.effect_count), memory{memory_} {
    behavior_info.SetUserRevision(params.revision);
    audio_out = std::make_unique<AudioCore::AudioOut>();
    stream = audio_out->OpenStream(core_timing, STREAM_SAMPLE_RATE, STREAM_NUM_CHANNELS,
@@ -127,6 +128,12 @@ ResultVal<std::vector<u8>> AudioRenderer::UpdateAudioRenderer(const std::vector<
                input_params.data() + sizeof(UpdateDataHeader) + config.behavior_size,
                memory_pool_count * sizeof(MemoryPoolInfo));

+    // Copy voice resources
+    const std::size_t voice_resource_offset{sizeof(UpdateDataHeader) + config.behavior_size +
+                                            config.memory_pools_size};
+    std::memcpy(voice_resources.data(), input_params.data() + voice_resource_offset,
+                sizeof(VoiceResourceInformation) * voice_resources.size());
+
    // Copy VoiceInfo structs
    std::size_t voice_offset{sizeof(UpdateDataHeader) + config.behavior_size +
                             config.memory_pools_size + config.voice_resource_size};
@@ -220,14 +227,15 @@ void AudioRenderer::VoiceState::SetWaveIndex(std::size_t index) {
    is_refresh_pending = true;
 }

-std::vector<s16> AudioRenderer::VoiceState::DequeueSamples(std::size_t sample_count,
-                                                           Core::Memory::Memory& memory) {
+std::vector<s16> AudioRenderer::VoiceState::DequeueSamples(
+    std::size_t sample_count, Core::Memory::Memory& memory,
+    const VoiceChannelHolder& voice_resources) {
    if (!IsPlaying()) {
        return {};
    }

    if (is_refresh_pending) {
-        RefreshBuffer(memory);
+        RefreshBuffer(memory, voice_resources);
    }

    const std::size_t max_size{samples.size() - offset};
@@ -271,7 +279,8 @@ void AudioRenderer::VoiceState::UpdateState() {
    is_in_use = info.is_in_use;
 }

-void AudioRenderer::VoiceState::RefreshBuffer(Core::Memory::Memory& memory) {
+void AudioRenderer::VoiceState::RefreshBuffer(Core::Memory::Memory& memory,
+                                              const VoiceChannelHolder& voice_resources) {
    const auto wave_buffer_address = info.wave_buffer[wave_index].buffer_addr;
    const auto wave_buffer_size = info.wave_buffer[wave_index].buffer_sz;
    std::vector<s16> new_samples(wave_buffer_size / sizeof(s16));
@@ -296,17 +305,77 @@ void AudioRenderer::VoiceState::RefreshBuffer(Core::Memory::Memory& memory) {
    }

    switch (info.channel_count) {
-    case 1:
+    case 1: {
        // 1 channel is upsampled to 2 channel
        samples.resize(new_samples.size() * 2);
+
        for (std::size_t index = 0; index < new_samples.size(); ++index) {
-            samples[index * 2] = new_samples[index];
-            samples[index * 2 + 1] = new_samples[index];
+            auto sample = static_cast<float>(new_samples[index]);
+            if (voice_resources[0]->in_use) {
+                sample *= voice_resources[0]->mix_volumes[0];
+            }
+
+            samples[index * 2] = static_cast<s16>(sample * info.volume);
+            samples[index * 2 + 1] = static_cast<s16>(sample * info.volume);
        }
        break;
+    }
    case 2: {
        // 2 channel is played as is
        samples = std::move(new_samples);
+        const std::size_t sample_count = (samples.size() / 2);
+        for (std::size_t index = 0; index < sample_count; ++index) {
+            const std::size_t index_l = index * 2;
+            const std::size_t index_r = index * 2 + 1;
+
+            auto sample_l = static_cast<float>(samples[index_l]);
+            auto sample_r = static_cast<float>(samples[index_r]);
+
+            if (voice_resources[0]->in_use) {
+                sample_l *= voice_resources[0]->mix_volumes[0];
+            }
+
+            if (voice_resources[1]->in_use) {
+                sample_r *= voice_resources[1]->mix_volumes[1];
+            }
+
+            samples[index_l] = static_cast<s16>(sample_l * info.volume);
+            samples[index_r] = static_cast<s16>(sample_r * info.volume);
+        }
+        break;
+    }
+    case 6: {
+        samples.resize((new_samples.size() / 6) * 2);
+        const std::size_t sample_count = samples.size() / 2;
+
+        for (std::size_t index = 0; index < sample_count; ++index) {
+            auto FL = static_cast<float>(new_samples[index * 6]);
+            auto FR = static_cast<float>(new_samples[index * 6 + 1]);
+            auto FC = static_cast<float>(new_samples[index * 6 + 2]);
+            auto BL = static_cast<float>(new_samples[index * 6 + 4]);
+            auto BR = static_cast<float>(new_samples[index * 6 + 5]);
+
+            if (voice_resources[0]->in_use) {
+                FL *= voice_resources[0]->mix_volumes[0];
+            }
+            if (voice_resources[1]->in_use) {
+                FR *= voice_resources[1]->mix_volumes[1];
+            }
+            if (voice_resources[2]->in_use) {
+                FC *= voice_resources[2]->mix_volumes[2];
+            }
+            if (voice_resources[4]->in_use) {
+                BL *= voice_resources[4]->mix_volumes[4];
+            }
+            if (voice_resources[5]->in_use) {
+                BR *= voice_resources[5]->mix_volumes[5];
+            }
+
+            samples[index * 2] =
+                static_cast<s16>((0.3694f * FL + 0.2612f * FC + 0.3694f * BL) * info.volume);
+            samples[index * 2 + 1] =
+                static_cast<s16>((0.3694f * FR + 0.2612f * FC + 0.3694f * BR) * info.volume);
+        }
        break;
    }
    default:
@@ -352,11 +421,17 @@ void AudioRenderer::QueueMixedBuffer(Buffer::Tag tag) {
        if (!voice.IsPlaying()) {
            continue;
        }
+        VoiceChannelHolder resources{};
+        for (u32 channel = 0; channel < voice.GetInfo().channel_count; channel++) {
+            const auto channel_resource_id = voice.GetInfo().voice_channel_resource_ids[channel];
+            resources[channel] = &voice_resources[channel_resource_id];
+        }

        std::size_t offset{};
        s64 samples_remaining{BUFFER_SIZE};
        while (samples_remaining > 0) {
-            const std::vector<s16> samples{voice.DequeueSamples(samples_remaining, memory)};
+            const std::vector<s16> samples{
+                voice.DequeueSamples(samples_remaining, memory, resources)};

            if (samples.empty()) {
                break;
@@ -9,6 +9,7 @@
 #include <vector>

 #include "audio_core/behavior_info.h"
+#include "audio_core/common.h"
 #include "audio_core/stream.h"
 #include "common/common_funcs.h"
 #include "common/common_types.h"
@@ -116,6 +117,14 @@ struct WaveBuffer {
 };
 static_assert(sizeof(WaveBuffer) == 0x38, "WaveBuffer has wrong size");

+struct VoiceResourceInformation {
+    s32_le id{};
+    std::array<float_le, MAX_MIX_BUFFERS> mix_volumes{};
+    bool in_use{};
+    INSERT_PADDING_BYTES(11);
+};
+static_assert(sizeof(VoiceResourceInformation) == 0x70, "VoiceResourceInformation has wrong size");
+
 struct VoiceInfo {
    u32_le id;
    u32_le node_id;
@@ -244,6 +253,7 @@ private:
    AudioRendererParameter worker_params;
    std::shared_ptr<Kernel::WritableEvent> buffer_event;
    std::vector<VoiceState> voices;
+    std::vector<VoiceResourceInformation> voice_resources;
    std::vector<EffectState> effects;
    std::unique_ptr<AudioOut> audio_out;
    StreamPtr stream;
@@ -14,6 +14,7 @@ constexpr ResultCode ERR_INVALID_PARAMETERS{ErrorModule::Audio, 41};
 }

 constexpr u32_le CURRENT_PROCESS_REVISION = Common::MakeMagic('R', 'E', 'V', '8');
+constexpr std::size_t MAX_MIX_BUFFERS = 24;

 static constexpr u32 VersionFromRevision(u32_le rev) {
    // "REV7" -> 7
@@ -32,6 +32,8 @@ add_custom_command(OUTPUT scm_rev.cpp
    DEPENDS
      # WARNING! It was too much work to try and make a common location for this list,
      # so if you need to change it, please update CMakeModules/GenerateSCMRev.cmake as well
+      "${VIDEO_CORE}/renderer_opengl/gl_arb_decompiler.cpp"
+      "${VIDEO_CORE}/renderer_opengl/gl_arb_decompiler.h"
      "${VIDEO_CORE}/renderer_opengl/gl_shader_cache.cpp"
      "${VIDEO_CORE}/renderer_opengl/gl_shader_cache.h"
      "${VIDEO_CORE}/renderer_opengl/gl_shader_decompiler.cpp"
@@ -123,6 +125,8 @@ add_library(common STATIC
    lz4_compression.cpp
    lz4_compression.h
    math_util.h
+    memory_detect.cpp
+    memory_detect.h
    memory_hook.cpp
    memory_hook.h
    microprofile.cpp
@@ -148,6 +152,8 @@ add_library(common STATIC
    thread.h
    thread_queue_list.h
    threadsafe_queue.h
+    time_zone.cpp
+    time_zone.h
    timer.cpp
    timer.h
    uint128.cpp
@@ -167,10 +173,12 @@ if(ARCHITECTURE_x86_64)
        PRIVATE
            x64/cpu_detect.cpp
            x64/cpu_detect.h
+            x64/xbyak_abi.h
+            x64/xbyak_util.h
    )
 endif()

 create_target_directory_groups(common)

-target_link_libraries(common PUBLIC Boost::boost fmt microprofile)
-target_link_libraries(common PRIVATE lz4_static libzstd_static)
+target_link_libraries(common PUBLIC Boost::boost fmt::fmt microprofile)
+target_link_libraries(common PRIVATE lz4::lz4 zstd::zstd xbyak)
@@ -0,0 +1,60 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#ifdef _WIN32
+// clang-format off
+#include <windows.h>
+#include <sysinfoapi.h>
+// clang-format on
+#else
+#include <sys/types.h>
+#ifdef __APPLE__
+#include <sys/sysctl.h>
+#else
+#include <sys/sysinfo.h>
+#endif
+#endif
+
+#include "common/memory_detect.h"
+
+namespace Common {
+
+// Detects the RAM and Swapfile sizes
+static MemoryInfo Detect() {
+    MemoryInfo mem_info{};
+
+#ifdef _WIN32
+    MEMORYSTATUSEX memorystatus;
+    memorystatus.dwLength = sizeof(memorystatus);
+    GlobalMemoryStatusEx(&memorystatus);
+    mem_info.TotalPhysicalMemory = memorystatus.ullTotalPhys;
+    mem_info.TotalSwapMemory = memorystatus.ullTotalPageFile - mem_info.TotalPhysicalMemory;
+#elif defined(__APPLE__)
+    u64 ramsize;
+    struct xsw_usage vmusage;
+    std::size_t sizeof_ramsize = sizeof(ramsize);
+    std::size_t sizeof_vmusage = sizeof(vmusage);
+    // hw and vm are defined in sysctl.h
+    // https://github.com/apple/darwin-xnu/blob/master/bsd/sys/sysctl.h#L471
+    // sysctlbyname(const char *, void *, size_t *, void *, size_t);
+    sysctlbyname("hw.memsize", &ramsize, &sizeof_ramsize, NULL, 0);
+    sysctlbyname("vm.swapusage", &vmusage, &sizeof_vmusage, NULL, 0);
+    mem_info.TotalPhysicalMemory = ramsize;
+    mem_info.TotalSwapMemory = vmusage.xsu_total;
+#else
+    struct sysinfo meminfo;
+    sysinfo(&meminfo);
+    mem_info.TotalPhysicalMemory = meminfo.totalram;
+    mem_info.TotalSwapMemory = meminfo.totalswap;
+#endif
+
+    return mem_info;
+}
+
+const MemoryInfo& GetMemInfo() {
+    static MemoryInfo mem_info = Detect();
+    return mem_info;
+}
+
+} // namespace Common
@@ -0,0 +1,22 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "common/common_types.h"
+
+namespace Common {
+
+struct MemoryInfo {
+    u64 TotalPhysicalMemory{};
+    u64 TotalSwapMemory{};
+};
+
+/**
+ * Gets the memory info of the host system
+ * @return Reference to a MemoryInfo struct with the physical and swap memory sizes in bytes
+ */
+const MemoryInfo& GetMemInfo();
+
+} // namespace Common
@@ -0,0 +1,49 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <chrono>
+#include <iomanip>
+#include <sstream>
+
+#include "common/logging/log.h"
+#include "common/time_zone.h"
+
+namespace Common::TimeZone {
+
+std::string GetDefaultTimeZone() {
+    return "GMT";
+}
+
+static std::string GetOsTimeZoneOffset() {
+    const std::time_t t{std::time(nullptr)};
+    const std::tm tm{*std::localtime(&t)};
+
+    std::stringstream ss;
+    ss << std::put_time(&tm, "%z"); // Get the current timezone offset, e.g. "-400", as a string
+
+    return ss.str();
+}
+
+static int ConvertOsTimeZoneOffsetToInt(const std::string& timezone) {
+    try {
+        return std::stoi(timezone);
+    } catch (const std::invalid_argument&) {
+        LOG_CRITICAL(Common, "invalid_argument with {}!", timezone);
+        return 0;
+    } catch (const std::out_of_range&) {
+        LOG_CRITICAL(Common, "out_of_range with {}!", timezone);
+        return 0;
+    }
+}
+
+std::chrono::seconds GetCurrentOffsetSeconds() {
+    const int offset{ConvertOsTimeZoneOffsetToInt(GetOsTimeZoneOffset())};
+
+    int seconds{(offset / 100) * 60 * 60}; // Convert hour component to seconds
+    seconds += (offset % 100) * 60;        // Convert minute component to seconds
+
+    return std::chrono::seconds{seconds};
+}
+
+} // namespace Common::TimeZone
@@ -0,0 +1,18 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <chrono>
+#include <string>
+
+namespace Common::TimeZone {
+
+/// Gets the default timezone, i.e. "GMT"
+std::string GetDefaultTimeZone();
+
+/// Gets the offset of the current timezone (from the default), in seconds
+std::chrono::seconds GetCurrentOffsetSeconds();
+
+} // namespace Common::TimeZone
@@ -0,0 +1,229 @@
+// Copyright 2016 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <bitset>
+#include <initializer_list>
+#include <xbyak.h>
+#include "common/assert.h"
+
+namespace Common::X64 {
+
+inline std::size_t RegToIndex(const Xbyak::Reg& reg) {
+    using Kind = Xbyak::Reg::Kind;
+    ASSERT_MSG((reg.getKind() & (Kind::REG | Kind::XMM)) != 0,
+               "RegSet only support GPRs and XMM registers.");
+    ASSERT_MSG(reg.getIdx() < 16, "RegSet only supports XXM0-15.");
+    return reg.getIdx() + (reg.getKind() == Kind::REG ? 0 : 16);
+}
+
+inline Xbyak::Reg64 IndexToReg64(std::size_t reg_index) {
+    ASSERT(reg_index < 16);
+    return Xbyak::Reg64(static_cast<int>(reg_index));
+}
+
+inline Xbyak::Xmm IndexToXmm(std::size_t reg_index) {
+    ASSERT(reg_index >= 16 && reg_index < 32);
+    return Xbyak::Xmm(static_cast<int>(reg_index - 16));
+}
+
+inline Xbyak::Reg IndexToReg(std::size_t reg_index) {
+    if (reg_index < 16) {
+        return IndexToReg64(reg_index);
+    } else {
+        return IndexToXmm(reg_index);
+    }
+}
+
+inline std::bitset<32> BuildRegSet(std::initializer_list<Xbyak::Reg> regs) {
+    std::bitset<32> bits;
+    for (const Xbyak::Reg& reg : regs) {
+        bits[RegToIndex(reg)] = true;
+    }
+    return bits;
+}
+
+const std::bitset<32> ABI_ALL_GPRS(0x0000FFFF);
+const std::bitset<32> ABI_ALL_XMMS(0xFFFF0000);
+
+#ifdef _WIN32
+
+// Microsoft x64 ABI
+const Xbyak::Reg ABI_RETURN = Xbyak::util::rax;
+const Xbyak::Reg ABI_PARAM1 = Xbyak::util::rcx;
+const Xbyak::Reg ABI_PARAM2 = Xbyak::util::rdx;
+const Xbyak::Reg ABI_PARAM3 = Xbyak::util::r8;
+const Xbyak::Reg ABI_PARAM4 = Xbyak::util::r9;
+
+const std::bitset<32> ABI_ALL_CALLER_SAVED = BuildRegSet({
+    // GPRs
+    Xbyak::util::rcx,
+    Xbyak::util::rdx,
+    Xbyak::util::r8,
+    Xbyak::util::r9,
+    Xbyak::util::r10,
+    Xbyak::util::r11,
+    // XMMs
+    Xbyak::util::xmm0,
+    Xbyak::util::xmm1,
+    Xbyak::util::xmm2,
+    Xbyak::util::xmm3,
+    Xbyak::util::xmm4,
+    Xbyak::util::xmm5,
+});
+
+const std::bitset<32> ABI_ALL_CALLEE_SAVED = BuildRegSet({
+    // GPRs
+    Xbyak::util::rbx,
+    Xbyak::util::rsi,
+    Xbyak::util::rdi,
+    Xbyak::util::rbp,
+    Xbyak::util::r12,
+    Xbyak::util::r13,
+    Xbyak::util::r14,
+    Xbyak::util::r15,
+    // XMMs
+    Xbyak::util::xmm6,
+    Xbyak::util::xmm7,
+    Xbyak::util::xmm8,
+    Xbyak::util::xmm9,
+    Xbyak::util::xmm10,
+    Xbyak::util::xmm11,
+    Xbyak::util::xmm12,
+    Xbyak::util::xmm13,
+    Xbyak::util::xmm14,
+    Xbyak::util::xmm15,
+});
+
+constexpr size_t ABI_SHADOW_SPACE = 0x20;
+
+#else
+
+// System V x86-64 ABI
+const Xbyak::Reg ABI_RETURN = Xbyak::util::rax;
+const Xbyak::Reg ABI_PARAM1 = Xbyak::util::rdi;
+const Xbyak::Reg ABI_PARAM2 = Xbyak::util::rsi;
+const Xbyak::Reg ABI_PARAM3 = Xbyak::util::rdx;
+const Xbyak::Reg ABI_PARAM4 = Xbyak::util::rcx;
+
+const std::bitset<32> ABI_ALL_CALLER_SAVED = BuildRegSet({
+    // GPRs
+    Xbyak::util::rcx,
+    Xbyak::util::rdx,
+    Xbyak::util::rdi,
+    Xbyak::util::rsi,
+    Xbyak::util::r8,
+    Xbyak::util::r9,
+    Xbyak::util::r10,
+    Xbyak::util::r11,
+    // XMMs
+    Xbyak::util::xmm0,
+    Xbyak::util::xmm1,
+    Xbyak::util::xmm2,
+    Xbyak::util::xmm3,
+    Xbyak::util::xmm4,
+    Xbyak::util::xmm5,
+    Xbyak::util::xmm6,
+    Xbyak::util::xmm7,
+    Xbyak::util::xmm8,
+    Xbyak::util::xmm9,
+    Xbyak::util::xmm10,
+    Xbyak::util::xmm11,
+    Xbyak::util::xmm12,
+    Xbyak::util::xmm13,
+    Xbyak::util::xmm14,
+    Xbyak::util::xmm15,
+});
+
+const std::bitset<32> ABI_ALL_CALLEE_SAVED = BuildRegSet({
+    // GPRs
+    Xbyak::util::rbx,
+    Xbyak::util::rbp,
+    Xbyak::util::r12,
+    Xbyak::util::r13,
+    Xbyak::util::r14,
+    Xbyak::util::r15,
+});
+
+constexpr size_t ABI_SHADOW_SPACE = 0;
+
+#endif
+
+struct ABIFrameInfo {
+    s32 subtraction;
+    s32 xmm_offset;
+};
+
+inline ABIFrameInfo ABI_CalculateFrameSize(std::bitset<32> regs, size_t rsp_alignment,
+                                           size_t needed_frame_size) {
+    const auto count = (regs & ABI_ALL_GPRS).count();
+    rsp_alignment -= count * 8;
+    size_t subtraction = 0;
+    const auto xmm_count = (regs & ABI_ALL_XMMS).count();
+    if (xmm_count) {
+        // If we have any XMMs to save, we must align the stack here.
+        subtraction = rsp_alignment & 0xF;
+    }
+    subtraction += 0x10 * xmm_count;
+    size_t xmm_base_subtraction = subtraction;
+    subtraction += needed_frame_size;
+    subtraction += ABI_SHADOW_SPACE;
+    // Final alignment.
+    rsp_alignment -= subtraction;
+    subtraction += rsp_alignment & 0xF;
+
+    return ABIFrameInfo{static_cast<s32>(subtraction),
+                        static_cast<s32>(subtraction - xmm_base_subtraction)};
+}
+
+inline size_t ABI_PushRegistersAndAdjustStack(Xbyak::CodeGenerator& code, std::bitset<32> regs,
+                                              size_t rsp_alignment, size_t needed_frame_size = 0) {
+    auto frame_info = ABI_CalculateFrameSize(regs, rsp_alignment, needed_frame_size);
+
+    for (std::size_t i = 0; i < regs.size(); ++i) {
+        if (regs[i] && ABI_ALL_GPRS[i]) {
+            code.push(IndexToReg64(i));
+        }
+    }
+
+    if (frame_info.subtraction != 0) {
+        code.sub(code.rsp, frame_info.subtraction);
+    }
+
+    for (std::size_t i = 0; i < regs.size(); ++i) {
+        if (regs[i] && ABI_ALL_XMMS[i]) {
+            code.movaps(code.xword[code.rsp + frame_info.xmm_offset], IndexToXmm(i));
+            frame_info.xmm_offset += 0x10;
+        }
+    }
+
+    return ABI_SHADOW_SPACE;
+}
+
+inline void ABI_PopRegistersAndAdjustStack(Xbyak::CodeGenerator& code, std::bitset<32> regs,
+                                           size_t rsp_alignment, size_t needed_frame_size = 0) {
+    auto frame_info = ABI_CalculateFrameSize(regs, rsp_alignment, needed_frame_size);
+
+    for (std::size_t i = 0; i < regs.size(); ++i) {
+        if (regs[i] && ABI_ALL_XMMS[i]) {
+            code.movaps(IndexToXmm(i), code.xword[code.rsp + frame_info.xmm_offset]);
+            frame_info.xmm_offset += 0x10;
+        }
+    }
+
+    if (frame_info.subtraction != 0) {
+        code.add(code.rsp, frame_info.subtraction);
+    }
+
+    // GPRs need to be popped in reverse order
+    for (std::size_t j = 0; j < regs.size(); ++j) {
+        const std::size_t i = regs.size() - j - 1;
+        if (regs[i] && ABI_ALL_GPRS[i]) {
+            code.pop(IndexToReg64(i));
+        }
+    }
+}
+
+} // namespace Common::X64
@@ -0,0 +1,47 @@
+// Copyright 2016 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <type_traits>
+#include <xbyak.h>
+#include "common/x64/xbyak_abi.h"
+
+namespace Common::X64 {
+
+// Constants for use with cmpps/cmpss
+enum {
+    CMP_EQ = 0,
+    CMP_LT = 1,
+    CMP_LE = 2,
+    CMP_UNORD = 3,
+    CMP_NEQ = 4,
+    CMP_NLT = 5,
+    CMP_NLE = 6,
+    CMP_ORD = 7,
+};
+
+constexpr bool IsWithin2G(uintptr_t ref, uintptr_t target) {
+    const u64 distance = target - (ref + 5);
+    return !(distance >= 0x8000'0000ULL && distance <= ~0x8000'0000ULL);
+}
+
+inline bool IsWithin2G(const Xbyak::CodeGenerator& code, uintptr_t target) {
+    return IsWithin2G(reinterpret_cast<uintptr_t>(code.getCurr()), target);
+}
+
+template <typename T>
+inline void CallFarFunction(Xbyak::CodeGenerator& code, const T f) {
+    static_assert(std::is_pointer_v<T>, "Argument must be a (function) pointer.");
+    size_t addr = reinterpret_cast<size_t>(f);
+    if (IsWithin2G(code, addr)) {
+        code.call(f);
+    } else {
+        // ABI_RETURN is a safe temp register to use before a call
+        code.mov(ABI_RETURN, addr);
+        code.call(ABI_RETURN);
+    }
+}
+
+} // namespace Common::X64
@@ -606,11 +606,11 @@ endif()
 create_target_directory_groups(core)

 target_link_libraries(core PUBLIC common PRIVATE audio_core video_core)
-target_link_libraries(core PUBLIC Boost::boost PRIVATE fmt json-headers mbedtls opus unicorn)
+target_link_libraries(core PUBLIC Boost::boost PRIVATE fmt::fmt nlohmann_json::nlohmann_json mbedtls Opus::Opus unicorn)

 if (YUZU_ENABLE_BOXCAT)
    target_compile_definitions(core PRIVATE -DYUZU_ENABLE_BOXCAT)
-    target_link_libraries(core PRIVATE httplib json-headers zip)
+    target_link_libraries(core PRIVATE httplib nlohmann_json::nlohmann_json zip)
 endif()

 if (ENABLE_WEB_SERVICE)
@@ -50,7 +50,8 @@ public:
    }

    void InterpreterFallback(u32 pc, std::size_t num_instructions) override {
-        UNIMPLEMENTED();
+        UNIMPLEMENTED_MSG("This should never happen, pc = {:08X}, code = {:08X}", pc,
+                          MemoryReadCode(pc));
    }

    void ExceptionRaised(u32 pc, Dynarmic::A32::Exception exception) override {
@@ -89,8 +90,6 @@ public:

    ARM_Dynarmic_32& parent;
    std::size_t num_interpreted_instructions{};
-    u64 tpidrro_el0{};
-    u64 tpidr_el0{};
 };

 std::shared_ptr<Dynarmic::A32::Jit> ARM_Dynarmic_32::MakeJit(Common::PageTable& page_table,
@@ -99,7 +98,7 @@ std::shared_ptr<Dynarmic::A32::Jit> ARM_Dynarmic_32::MakeJit(Common::PageTable&
    config.callbacks = cb.get();
    // TODO(bunnei): Implement page table for 32-bit
    // config.page_table = &page_table.pointers;
-    config.coprocessors[15] = std::make_shared<DynarmicCP15>((u32*)&CP15_regs[0]);
+    config.coprocessors[15] = cp15;
    config.define_unpredictable_behaviour = true;
    return std::make_unique<Dynarmic::A32::Jit>(config);
 }
@@ -112,13 +111,13 @@ void ARM_Dynarmic_32::Run() {
 }

 void ARM_Dynarmic_32::Step() {
-    cb->InterpreterFallback(jit->Regs()[15], 1);
+    jit->Step();
 }

 ARM_Dynarmic_32::ARM_Dynarmic_32(System& system, ExclusiveMonitor& exclusive_monitor,
                                 std::size_t core_index)
-    : ARM_Interface{system},
-      cb(std::make_unique<DynarmicCallbacks32>(*this)), core_index{core_index},
+    : ARM_Interface{system}, cb(std::make_unique<DynarmicCallbacks32>(*this)),
+      cp15(std::make_shared<DynarmicCP15>(*this)), core_index{core_index},
      exclusive_monitor{dynamic_cast<DynarmicExclusiveMonitor&>(exclusive_monitor)} {}

 ARM_Dynarmic_32::~ARM_Dynarmic_32() = default;
@@ -154,19 +153,19 @@ void ARM_Dynarmic_32::SetPSTATE(u32 cpsr) {
 }

 u64 ARM_Dynarmic_32::GetTlsAddress() const {
-    return CP15_regs[static_cast<std::size_t>(CP15Register::CP15_THREAD_URO)];
+    return cp15->uro;
 }

 void ARM_Dynarmic_32::SetTlsAddress(VAddr address) {
-    CP15_regs[static_cast<std::size_t>(CP15Register::CP15_THREAD_URO)] = static_cast<u32>(address);
+    cp15->uro = static_cast<u32>(address);
 }

 u64 ARM_Dynarmic_32::GetTPIDR_EL0() const {
-    return cb->tpidr_el0;
+    return cp15->uprw;
 }

 void ARM_Dynarmic_32::SetTPIDR_EL0(u64 value) {
-    cb->tpidr_el0 = value;
+    cp15->uprw = value;
 }

 void ARM_Dynarmic_32::SaveContext(ThreadContext32& ctx) {
@@ -22,6 +22,7 @@ class Memory;
 namespace Core {

 class DynarmicCallbacks32;
+class DynarmicCP15;
 class DynarmicExclusiveMonitor;
 class System;

@@ -66,12 +67,14 @@ private:
        std::unordered_map<JitCacheKey, std::shared_ptr<Dynarmic::A32::Jit>, Common::PairHash>;

    friend class DynarmicCallbacks32;
+    friend class DynarmicCP15;
+
    std::unique_ptr<DynarmicCallbacks32> cb;
    JitCacheType jit_cache;
    std::shared_ptr<Dynarmic::A32::Jit> jit;
+    std::shared_ptr<DynarmicCP15> cp15;
    std::size_t core_index;
    DynarmicExclusiveMonitor& exclusive_monitor;
-    std::array<u32, 84> CP15_regs{};
 };

 } // namespace Core
@@ -2,79 +2,132 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.

+#include <fmt/format.h>
+#include "common/logging/log.h"
+#include "core/arm/dynarmic/arm_dynarmic_32.h"
 #include "core/arm/dynarmic/arm_dynarmic_cp15.h"
+#include "core/core.h"
+#include "core/core_timing.h"
+#include "core/core_timing_util.h"

 using Callback = Dynarmic::A32::Coprocessor::Callback;
 using CallbackOrAccessOneWord = Dynarmic::A32::Coprocessor::CallbackOrAccessOneWord;
 using CallbackOrAccessTwoWords = Dynarmic::A32::Coprocessor::CallbackOrAccessTwoWords;

+template <>
+struct fmt::formatter<Dynarmic::A32::CoprocReg> {
+    constexpr auto parse(format_parse_context& ctx) {
+        return ctx.begin();
+    }
+    template <typename FormatContext>
+    auto format(const Dynarmic::A32::CoprocReg& reg, FormatContext& ctx) {
+        return format_to(ctx.out(), "cp{}", static_cast<size_t>(reg));
+    }
+};
+
+namespace Core {
+
+static u32 dummy_value;
+
 std::optional<Callback> DynarmicCP15::CompileInternalOperation(bool two, unsigned opc1,
                                                               CoprocReg CRd, CoprocReg CRn,
                                                               CoprocReg CRm, unsigned opc2) {
+    LOG_CRITICAL(Core_ARM, "CP15: cdp{} p15, {}, {}, {}, {}, {}", two ? "2" : "", opc1, CRd, CRn,
+                 CRm, opc2);
    return {};
 }

 CallbackOrAccessOneWord DynarmicCP15::CompileSendOneWord(bool two, unsigned opc1, CoprocReg CRn,
                                                         CoprocReg CRm, unsigned opc2) {
-    // TODO(merry): Privileged CP15 registers
-
    if (!two && CRn == CoprocReg::C7 && opc1 == 0 && CRm == CoprocReg::C5 && opc2 == 4) {
+        // CP15_FLUSH_PREFETCH_BUFFER
        // This is a dummy write, we ignore the value written here.
-        return &CP15[static_cast<std::size_t>(CP15Register::CP15_FLUSH_PREFETCH_BUFFER)];
+        return &dummy_value;
    }

    if (!two && CRn == CoprocReg::C7 && opc1 == 0 && CRm == CoprocReg::C10) {
        switch (opc2) {
        case 4:
+            // CP15_DATA_SYNC_BARRIER
            // This is a dummy write, we ignore the value written here.
-            return &CP15[static_cast<std::size_t>(CP15Register::CP15_DATA_SYNC_BARRIER)];
+            return &dummy_value;
        case 5:
+            // CP15_DATA_MEMORY_BARRIER
            // This is a dummy write, we ignore the value written here.
-            return &CP15[static_cast<std::size_t>(CP15Register::CP15_DATA_MEMORY_BARRIER)];
-        default:
-            return {};
+            return &dummy_value;
        }
    }

    if (!two && CRn == CoprocReg::C13 && opc1 == 0 && CRm == CoprocReg::C0 && opc2 == 2) {
-        return &CP15[static_cast<std::size_t>(CP15Register::CP15_THREAD_UPRW)];
+        // CP15_THREAD_UPRW
+        return &uprw;
    }

+    LOG_CRITICAL(Core_ARM, "CP15: mcr{} p15, {}, <Rt>, {}, {}, {}", two ? "2" : "", opc1, CRn, CRm,
+                 opc2);
    return {};
 }

 CallbackOrAccessTwoWords DynarmicCP15::CompileSendTwoWords(bool two, unsigned opc, CoprocReg CRm) {
+    LOG_CRITICAL(Core_ARM, "CP15: mcrr{} p15, {}, <Rt>, <Rt2>, {}", two ? "2" : "", opc, CRm);
    return {};
 }

 CallbackOrAccessOneWord DynarmicCP15::CompileGetOneWord(bool two, unsigned opc1, CoprocReg CRn,
                                                        CoprocReg CRm, unsigned opc2) {
-    // TODO(merry): Privileged CP15 registers
-
    if (!two && CRn == CoprocReg::C13 && opc1 == 0 && CRm == CoprocReg::C0) {
        switch (opc2) {
        case 2:
-            return &CP15[static_cast<std::size_t>(CP15Register::CP15_THREAD_UPRW)];
+            // CP15_THREAD_UPRW
+            return &uprw;
        case 3:
-            return &CP15[static_cast<std::size_t>(CP15Register::CP15_THREAD_URO)];
-        default:
-            return {};
+            // CP15_THREAD_URO
+            return &uro;
        }
    }

+    LOG_CRITICAL(Core_ARM, "CP15: mrc{} p15, {}, <Rt>, {}, {}, {}", two ? "2" : "", opc1, CRn, CRm,
+                 opc2);
    return {};
 }

 CallbackOrAccessTwoWords DynarmicCP15::CompileGetTwoWords(bool two, unsigned opc, CoprocReg CRm) {
+    if (!two && opc == 0 && CRm == CoprocReg::C14) {
+        // CNTPCT
+        const auto callback = static_cast<u64 (*)(Dynarmic::A32::Jit*, void*, u32, u32)>(
+            [](Dynarmic::A32::Jit*, void* arg, u32, u32) -> u64 {
+                ARM_Dynarmic_32& parent = *(ARM_Dynarmic_32*)arg;
+                return Timing::CpuCyclesToClockCycles(parent.system.CoreTiming().GetTicks());
+            });
+        return Dynarmic::A32::Coprocessor::Callback{callback, (void*)&parent};
+    }
+
+    LOG_CRITICAL(Core_ARM, "CP15: mrrc{} p15, {}, <Rt>, <Rt2>, {}", two ? "2" : "", opc, CRm);
    return {};
 }

 std::optional<Callback> DynarmicCP15::CompileLoadWords(bool two, bool long_transfer, CoprocReg CRd,
                                                       std::optional<u8> option) {
+    if (option) {
+        LOG_CRITICAL(Core_ARM, "CP15: mrrc{}{} p15, {}, [...], {}", two ? "2" : "",
+                     long_transfer ? "l" : "", CRd, *option);
+    } else {
+        LOG_CRITICAL(Core_ARM, "CP15: mrrc{}{} p15, {}, [...]", two ? "2" : "",
+                     long_transfer ? "l" : "", CRd);
+    }
    return {};
 }

 std::optional<Callback> DynarmicCP15::CompileStoreWords(bool two, bool long_transfer, CoprocReg CRd,
                                                        std::optional<u8> option) {
+    if (option) {
+        LOG_CRITICAL(Core_ARM, "CP15: mrrc{}{} p15, {}, [...], {}", two ? "2" : "",
+                     long_transfer ? "l" : "", CRd, *option);
+    } else {
+        LOG_CRITICAL(Core_ARM, "CP15: mrrc{}{} p15, {}, [...]", two ? "2" : "",
+                     long_transfer ? "l" : "", CRd);
+    }
    return {};
 }
+
+} // namespace Core
@@ -10,128 +10,15 @@
 #include <dynarmic/A32/coprocessor.h>
 #include "common/common_types.h"

-enum class CP15Register {
-    // c0 - Information registers
-    CP15_MAIN_ID,
-    CP15_CACHE_TYPE,
-    CP15_TCM_STATUS,
-    CP15_TLB_TYPE,
-    CP15_CPU_ID,
-    CP15_PROCESSOR_FEATURE_0,
-    CP15_PROCESSOR_FEATURE_1,
-    CP15_DEBUG_FEATURE_0,
-    CP15_AUXILIARY_FEATURE_0,
-    CP15_MEMORY_MODEL_FEATURE_0,
-    CP15_MEMORY_MODEL_FEATURE_1,
-    CP15_MEMORY_MODEL_FEATURE_2,
-    CP15_MEMORY_MODEL_FEATURE_3,
-    CP15_ISA_FEATURE_0,
-    CP15_ISA_FEATURE_1,
-    CP15_ISA_FEATURE_2,
-    CP15_ISA_FEATURE_3,
-    CP15_ISA_FEATURE_4,
+namespace Core {

-    // c1 - Control registers
-    CP15_CONTROL,
-    CP15_AUXILIARY_CONTROL,
-    CP15_COPROCESSOR_ACCESS_CONTROL,
-
-    // c2 - Translation table registers
-    CP15_TRANSLATION_BASE_TABLE_0,
-    CP15_TRANSLATION_BASE_TABLE_1,
-    CP15_TRANSLATION_BASE_CONTROL,
-    CP15_DOMAIN_ACCESS_CONTROL,
-    CP15_RESERVED,
-
-    // c5 - Fault status registers
-    CP15_FAULT_STATUS,
-    CP15_INSTR_FAULT_STATUS,
-    CP15_COMBINED_DATA_FSR = CP15_FAULT_STATUS,
-    CP15_INST_FSR,
-
-    // c6 - Fault Address registers
-    CP15_FAULT_ADDRESS,
-    CP15_COMBINED_DATA_FAR = CP15_FAULT_ADDRESS,
-    CP15_WFAR,
-    CP15_IFAR,
-
-    // c7 - Cache operation registers
-    CP15_WAIT_FOR_INTERRUPT,
-    CP15_PHYS_ADDRESS,
-    CP15_INVALIDATE_INSTR_CACHE,
-    CP15_INVALIDATE_INSTR_CACHE_USING_MVA,
-    CP15_INVALIDATE_INSTR_CACHE_USING_INDEX,
-    CP15_FLUSH_PREFETCH_BUFFER,
-    CP15_FLUSH_BRANCH_TARGET_CACHE,
-    CP15_FLUSH_BRANCH_TARGET_CACHE_ENTRY,
-    CP15_INVALIDATE_DATA_CACHE,
-    CP15_INVALIDATE_DATA_CACHE_LINE_USING_MVA,
-    CP15_INVALIDATE_DATA_CACHE_LINE_USING_INDEX,
-    CP15_INVALIDATE_DATA_AND_INSTR_CACHE,
-    CP15_CLEAN_DATA_CACHE,
-    CP15_CLEAN_DATA_CACHE_LINE_USING_MVA,
-    CP15_CLEAN_DATA_CACHE_LINE_USING_INDEX,
-    CP15_DATA_SYNC_BARRIER,
-    CP15_DATA_MEMORY_BARRIER,
-    CP15_CLEAN_AND_INVALIDATE_DATA_CACHE,
-    CP15_CLEAN_AND_INVALIDATE_DATA_CACHE_LINE_USING_MVA,
-    CP15_CLEAN_AND_INVALIDATE_DATA_CACHE_LINE_USING_INDEX,
-
-    // c8 - TLB operations
-    CP15_INVALIDATE_ITLB,
-    CP15_INVALIDATE_ITLB_SINGLE_ENTRY,
-    CP15_INVALIDATE_ITLB_ENTRY_ON_ASID_MATCH,
-    CP15_INVALIDATE_ITLB_ENTRY_ON_MVA,
-    CP15_INVALIDATE_DTLB,
-    CP15_INVALIDATE_DTLB_SINGLE_ENTRY,
-    CP15_INVALIDATE_DTLB_ENTRY_ON_ASID_MATCH,
-    CP15_INVALIDATE_DTLB_ENTRY_ON_MVA,
-    CP15_INVALIDATE_UTLB,
-    CP15_INVALIDATE_UTLB_SINGLE_ENTRY,
-    CP15_INVALIDATE_UTLB_ENTRY_ON_ASID_MATCH,
-    CP15_INVALIDATE_UTLB_ENTRY_ON_MVA,
-
-    // c9 - Data cache lockdown register
-    CP15_DATA_CACHE_LOCKDOWN,
-
-    // c10 - TLB/Memory map registers
-    CP15_TLB_LOCKDOWN,
-    CP15_PRIMARY_REGION_REMAP,
-    CP15_NORMAL_REGION_REMAP,
-
-    // c13 - Thread related registers
-    CP15_PID,
-    CP15_CONTEXT_ID,
-    CP15_THREAD_UPRW, // Thread ID register - User/Privileged Read/Write
-    CP15_THREAD_URO,  // Thread ID register - User Read Only (Privileged R/W)
-    CP15_THREAD_PRW,  // Thread ID register - Privileged R/W only.
-
-    // c15 - Performance and TLB lockdown registers
-    CP15_PERFORMANCE_MONITOR_CONTROL,
-    CP15_CYCLE_COUNTER,
-    CP15_COUNT_0,
-    CP15_COUNT_1,
-    CP15_READ_MAIN_TLB_LOCKDOWN_ENTRY,
-    CP15_WRITE_MAIN_TLB_LOCKDOWN_ENTRY,
-    CP15_MAIN_TLB_LOCKDOWN_VIRT_ADDRESS,
-    CP15_MAIN_TLB_LOCKDOWN_PHYS_ADDRESS,
-    CP15_MAIN_TLB_LOCKDOWN_ATTRIBUTE,
-    CP15_TLB_DEBUG_CONTROL,
-
-    // Skyeye defined
-    CP15_TLB_FAULT_ADDR,
-    CP15_TLB_FAULT_STATUS,
-
-    // Not an actual register.
-    // All registers should be defined above this.
-    CP15_REGISTER_COUNT,
-};
+class ARM_Dynarmic_32;

 class DynarmicCP15 final : public Dynarmic::A32::Coprocessor {
 public:
    using CoprocReg = Dynarmic::A32::CoprocReg;

-    explicit DynarmicCP15(u32* cp15) : CP15(cp15){};
+    explicit DynarmicCP15(ARM_Dynarmic_32& parent) : parent(parent) {}

    std::optional<Callback> CompileInternalOperation(bool two, unsigned opc1, CoprocReg CRd,
                                                     CoprocReg CRn, CoprocReg CRm,
@@ -147,6 +34,9 @@ public:
    std::optional<Callback> CompileStoreWords(bool two, bool long_transfer, CoprocReg CRd,
                                              std::optional<u8> option) override;

-private:
-    u32* CP15{};
+    ARM_Dynarmic_32& parent;
+    u32 uprw;
+    u32 uro;
 };
+
+} // namespace Core
@@ -95,6 +95,10 @@ u32 NACP::GetSupportedLanguages() const {
    return raw.supported_languages;
 }

+u64 NACP::GetDeviceSaveDataSize() const {
+    return raw.device_save_data_size;
+}
+
 std::vector<u8> NACP::GetRawBytes() const {
    std::vector<u8> out(sizeof(RawNACP));
    std::memcpy(out.data(), &raw, sizeof(RawNACP));
@@ -113,6 +113,7 @@ public:
    u32 GetSupportedLanguages() const;
    std::vector<u8> GetRawBytes() const;
    bool GetUserAccountSwitchLock() const;
+    u64 GetDeviceSaveDataSize() const;

 private:
    RawNACP raw{};
@@ -10,6 +10,7 @@
 #include "common/file_util.h"
 #include "common/hex_util.h"
 #include "common/logging/log.h"
+#include "common/string_util.h"
 #include "core/core.h"
 #include "core/file_sys/content_archive.h"
 #include "core/file_sys/control_metadata.h"
@@ -48,6 +49,23 @@ std::string FormatTitleVersion(u32 version, TitleVersionFormat format) {
    return fmt::format("v{}.{}.{}", bytes[3], bytes[2], bytes[1]);
 }

+std::shared_ptr<VfsDirectory> FindSubdirectoryCaseless(const std::shared_ptr<VfsDirectory> dir,
+                                                       std::string_view name) {
+#ifdef _WIN32
+    return dir->GetSubdirectory(name);
+#else
+    const auto subdirs = dir->GetSubdirectories();
+    for (const auto& subdir : subdirs) {
+        std::string dir_name = Common::ToLower(subdir->GetName());
+        if (dir_name == name) {
+            return subdir;
+        }
+    }
+
+    return nullptr;
+#endif
+}
+
 PatchManager::PatchManager(u64 title_id) : title_id(title_id) {}

 PatchManager::~PatchManager() = default;
@@ -104,7 +122,7 @@ VirtualDir PatchManager::PatchExeFS(VirtualDir exefs) const {
            if (std::find(disabled.begin(), disabled.end(), subdir->GetName()) != disabled.end())
                continue;

-            auto exefs_dir = subdir->GetSubdirectory("exefs");
+            auto exefs_dir = FindSubdirectoryCaseless(subdir, "exefs");
            if (exefs_dir != nullptr)
                layers.push_back(std::move(exefs_dir));
        }
@@ -130,7 +148,7 @@ std::vector<VirtualFile> PatchManager::CollectPatches(const std::vector<VirtualD
        if (std::find(disabled.cbegin(), disabled.cend(), subdir->GetName()) != disabled.cend())
            continue;

-        auto exefs_dir = subdir->GetSubdirectory("exefs");
+        auto exefs_dir = FindSubdirectoryCaseless(subdir, "exefs");
        if (exefs_dir != nullptr) {
            for (const auto& file : exefs_dir->GetFiles()) {
                if (file->GetExtension() == "ips") {
@@ -295,7 +313,7 @@ std::vector<Core::Memory::CheatEntry> PatchManager::CreateCheatList(
            continue;
        }

-        auto cheats_dir = subdir->GetSubdirectory("cheats");
+        auto cheats_dir = FindSubdirectoryCaseless(subdir, "cheats");
        if (cheats_dir != nullptr) {
            auto res = ReadCheatFileFromFolder(system, title_id, build_id_, cheats_dir, true);
            if (res.has_value()) {
@@ -340,11 +358,11 @@ static void ApplyLayeredFS(VirtualFile& romfs, u64 title_id, ContentRecordType t
            continue;
        }

-        auto romfs_dir = subdir->GetSubdirectory("romfs");
+        auto romfs_dir = FindSubdirectoryCaseless(subdir, "romfs");
        if (romfs_dir != nullptr)
            layers.push_back(std::move(romfs_dir));

-        auto ext_dir = subdir->GetSubdirectory("romfs_ext");
+        auto ext_dir = FindSubdirectoryCaseless(subdir, "romfs_ext");
        if (ext_dir != nullptr)
            layers_ext.push_back(std::move(ext_dir));
    }
@@ -470,7 +488,7 @@ std::map<std::string, std::string, std::less<>> PatchManager::GetPatchVersionNam
        for (const auto& mod : mod_dir->GetSubdirectories()) {
            std::string types;

-            const auto exefs_dir = mod->GetSubdirectory("exefs");
+            const auto exefs_dir = FindSubdirectoryCaseless(mod, "exefs");
            if (IsDirValidAndNonEmpty(exefs_dir)) {
                bool ips = false;
                bool ipswitch = false;
@@ -494,9 +512,9 @@ std::map<std::string, std::string, std::less<>> PatchManager::GetPatchVersionNam
                if (layeredfs)
                    AppendCommaIfNotEmpty(types, "LayeredExeFS");
            }
-            if (IsDirValidAndNonEmpty(mod->GetSubdirectory("romfs")))
+            if (IsDirValidAndNonEmpty(FindSubdirectoryCaseless(mod, "romfs")))
                AppendCommaIfNotEmpty(types, "LayeredFS");
-            if (IsDirValidAndNonEmpty(mod->GetSubdirectory("cheats")))
+            if (IsDirValidAndNonEmpty(FindSubdirectoryCaseless(mod, "cheats")))
                AppendCommaIfNotEmpty(types, "Cheats");

            if (types.empty())
@@ -29,6 +29,11 @@ enum class TitleVersionFormat : u8 {
 std::string FormatTitleVersion(u32 version,
                               TitleVersionFormat format = TitleVersionFormat::ThreeElements);

+// Returns a directory with name matching name case-insensitive. Returns nullptr if directory
+// doesn't have a directory with name.
+std::shared_ptr<VfsDirectory> FindSubdirectoryCaseless(const std::shared_ptr<VfsDirectory> dir,
+                                                       std::string_view name);
+
 // A centralized class to manage patches to games.
 class PatchManager {
 public:
@@ -57,7 +57,8 @@ void PrintSaveDataDescriptorWarnings(SaveDataDescriptor meta) {
 bool ShouldSaveDataBeAutomaticallyCreated(SaveDataSpaceId space, const SaveDataDescriptor& desc) {
    return desc.type == SaveDataType::CacheStorage || desc.type == SaveDataType::TemporaryStorage ||
           (space == SaveDataSpaceId::NandUser && ///< Normal Save Data -- Current Title & User
-            desc.type == SaveDataType::SaveData && desc.title_id == 0 && desc.save_id == 0);
+            (desc.type == SaveDataType::SaveData || desc.type == SaveDataType::DeviceSaveData) &&
+            desc.title_id == 0 && desc.save_id == 0);
 }

 } // Anonymous namespace
@@ -139,8 +140,10 @@ std::string SaveDataFactory::GetFullPath(SaveDataSpaceId space, SaveDataType typ
                                         u128 user_id, u64 save_id) {
    // According to switchbrew, if a save is of type SaveData and the title id field is 0, it should
    // be interpreted as the title id of the current process.
-    if (type == SaveDataType::SaveData && title_id == 0) {
-        title_id = Core::System::GetInstance().CurrentProcess()->GetTitleID();
+    if (type == SaveDataType::SaveData || type == SaveDataType::DeviceSaveData) {
+        if (title_id == 0) {
+            title_id = Core::System::GetInstance().CurrentProcess()->GetTitleID();
+        }
    }

    std::string out = GetSaveDataSpaceIdPath(space);
@@ -12,17 +12,17 @@ namespace SystemVersionData {
 // This section should reflect the best system version to describe yuzu's HLE api.
 // TODO(DarkLordZach): Update when HLE gets better.

-constexpr u8 VERSION_MAJOR = 5;
-constexpr u8 VERSION_MINOR = 1;
-constexpr u8 VERSION_MICRO = 0;
+constexpr u8 VERSION_MAJOR = 10;
+constexpr u8 VERSION_MINOR = 0;
+constexpr u8 VERSION_MICRO = 2;

-constexpr u8 REVISION_MAJOR = 3;
+constexpr u8 REVISION_MAJOR = 1;
 constexpr u8 REVISION_MINOR = 0;

 constexpr char PLATFORM_STRING[] = "NX";
-constexpr char VERSION_HASH[] = "23f9df53e25709d756e0c76effcb2473bd3447dd";
-constexpr char DISPLAY_VERSION[] = "5.1.0";
-constexpr char DISPLAY_TITLE[] = "NintendoSDK Firmware for NX 5.1.0-3.0";
+constexpr char VERSION_HASH[] = "f90143fa8bbc061d4f68c35f95f04f8080c0ecdc";
+constexpr char DISPLAY_VERSION[] = "10.0.2";
+constexpr char DISPLAY_TITLE[] = "NintendoSDK Firmware for NX 10.0.2-1.0";

 } // namespace SystemVersionData

@@ -46,7 +46,7 @@ private:
 EmuWindow::EmuWindow() {
    // TODO: Find a better place to set this.
    config.min_client_area_size =
-        std::make_pair(Layout::ScreenUndocked::Width, Layout::ScreenUndocked::Height);
+        std::make_pair(Layout::MinimumSize::Width, Layout::MinimumSize::Height);
    active_config = config;
    touch_state = std::make_shared<TouchState>();
    Input::RegisterFactory<Input::TouchDevice>("emu_window", touch_state);
@@ -8,6 +8,11 @@

 namespace Layout {

+namespace MinimumSize {
+constexpr u32 Width = 640;
+constexpr u32 Height = 360;
+} // namespace MinimumSize
+
 namespace ScreenUndocked {
 constexpr u32 Width = 1280;
 constexpr u32 Height = 720;
@@ -13,7 +13,6 @@
 #include "common/common_funcs.h"
 #include "common/common_types.h"
 #include "common/logging/log.h"
-#include "core/core.h"
 #include "core/hle/ipc_helpers.h"
 #include "core/hle/kernel/handle_table.h"
 #include "core/hle/kernel/hle_ipc.h"
@@ -57,7 +56,6 @@ std::shared_ptr<WritableEvent> HLERequestContext::SleepClientThread(
            return true;
        });

-    auto& kernel = Core::System::GetInstance().Kernel();
    if (!writable_event) {
        // Create event if not provided
        const auto pair = WritableEvent::CreateEventPair(kernel, "HLE Pause Event: " + reason);
@@ -79,9 +77,11 @@ std::shared_ptr<WritableEvent> HLERequestContext::SleepClientThread(
    return writable_event;
 }

-HLERequestContext::HLERequestContext(std::shared_ptr<Kernel::ServerSession> server_session,
+HLERequestContext::HLERequestContext(KernelCore& kernel, Core::Memory::Memory& memory,
+                                     std::shared_ptr<ServerSession> server_session,
                                     std::shared_ptr<Thread> thread)
-    : server_session(std::move(server_session)), thread(std::move(thread)) {
+    : server_session(std::move(server_session)),
+      thread(std::move(thread)), kernel{kernel}, memory{memory} {
    cmd_buf[0] = 0;
 }

@@ -216,7 +216,6 @@ ResultCode HLERequestContext::PopulateFromIncomingCommandBuffer(const HandleTabl
 ResultCode HLERequestContext::WriteToOutgoingCommandBuffer(Thread& thread) {
    auto& owner_process = *thread.GetOwnerProcess();
    auto& handle_table = owner_process.GetHandleTable();
-    auto& memory = Core::System::GetInstance().Memory();

    std::array<u32, IPC::COMMAND_BUFFER_LENGTH> dst_cmdbuf;
    memory.ReadBlock(owner_process, thread.GetTLSAddress(), dst_cmdbuf.data(),
@@ -286,7 +285,6 @@ std::vector<u8> HLERequestContext::ReadBuffer(std::size_t buffer_index) const {
    std::vector<u8> buffer;
    const bool is_buffer_a{BufferDescriptorA().size() > buffer_index &&
                           BufferDescriptorA()[buffer_index].Size()};
-    auto& memory = Core::System::GetInstance().Memory();

    if (is_buffer_a) {
        ASSERT_MSG(BufferDescriptorA().size() > buffer_index,
@@ -319,7 +317,6 @@ std::size_t HLERequestContext::WriteBuffer(const void* buffer, std::size_t size,
        size = buffer_size; // TODO(bunnei): This needs to be HW tested
    }

-    auto& memory = Core::System::GetInstance().Memory();
    if (is_buffer_b) {
        ASSERT_MSG(BufferDescriptorB().size() > buffer_index,
                   "BufferDescriptorB invalid buffer_index {}", buffer_index);
@@ -19,6 +19,10 @@

 union ResultCode;

+namespace Core::Memory {
+class Memory;
+}
+
 namespace Service {
 class ServiceFrameworkBase;
 }
@@ -28,6 +32,7 @@ namespace Kernel {
 class Domain;
 class HandleTable;
 class HLERequestContext;
+class KernelCore;
 class Process;
 class ServerSession;
 class Thread;
@@ -98,7 +103,8 @@ protected:
 */
 class HLERequestContext {
 public:
-    explicit HLERequestContext(std::shared_ptr<ServerSession> session,
+    explicit HLERequestContext(KernelCore& kernel, Core::Memory::Memory& memory,
+                               std::shared_ptr<ServerSession> session,
                               std::shared_ptr<Thread> thread);
    ~HLERequestContext();

@@ -305,6 +311,9 @@ private:

    std::vector<std::shared_ptr<SessionRequestHandler>> domain_request_handlers;
    bool is_thread_waiting{};
+
+    KernelCore& kernel;
+    Core::Memory::Memory& memory;
 };

 } // namespace Kernel
@@ -132,7 +132,8 @@ std::shared_ptr<ResourceLimit> Process::GetResourceLimit() const {

 u64 Process::GetTotalPhysicalMemoryAvailable() const {
    const u64 capacity{resource_limit->GetCurrentResourceValue(ResourceType::PhysicalMemory) +
-                       page_table->GetTotalHeapSize() + image_size + main_thread_stack_size};
+                       page_table->GetTotalHeapSize() + GetSystemResourceSize() + image_size +
+                       main_thread_stack_size};

    if (capacity < memory_usage_capacity) {
        return capacity;
@@ -146,7 +147,8 @@ u64 Process::GetTotalPhysicalMemoryAvailableWithoutSystemResource() const {
 }

 u64 Process::GetTotalPhysicalMemoryUsed() const {
-    return image_size + main_thread_stack_size + page_table->GetTotalHeapSize();
+    return image_size + main_thread_stack_size + page_table->GetTotalHeapSize() +
+           GetSystemResourceSize();
 }

 u64 Process::GetTotalPhysicalMemoryUsedWithoutSystemResource() const {
@@ -38,7 +38,7 @@ void ReadableEvent::Clear() {

 ResultCode ReadableEvent::Reset() {
    if (!is_signaled) {
-        LOG_ERROR(Kernel, "Handle is not signaled! object_id={}, object_type={}, object_name={}",
+        LOG_TRACE(Kernel, "Handle is not signaled! object_id={}, object_type={}, object_name={}",
                  GetObjectId(), GetTypeName(), GetName());
        return ERR_INVALID_STATE;
    }
@@ -24,13 +24,9 @@ bool ResourceLimit::Reserve(ResourceType resource, s64 amount, u64 timeout) {
    const std::size_t index{ResourceTypeToIndex(resource)};

    s64 new_value = current[index] + amount;
-    while (new_value > limit[index] && available[index] + amount <= limit[index]) {
+    if (new_value > limit[index] && available[index] + amount <= limit[index]) {
        // TODO(bunnei): This is wrong for multicore, we should wait the calling thread for timeout
        new_value = current[index] + amount;
-
-        if (timeout >= 0) {
-            break;
-        }
    }

    if (new_value <= limit[index]) {
@@ -137,8 +137,8 @@ ResultCode ServerSession::HandleDomainSyncRequest(Kernel::HLERequestContext& con
 ResultCode ServerSession::QueueSyncRequest(std::shared_ptr<Thread> thread,
                                           Core::Memory::Memory& memory) {
    u32* cmd_buf{reinterpret_cast<u32*>(memory.GetPointer(thread->GetTLSAddress()))};
-    std::shared_ptr<Kernel::HLERequestContext> context{
-        std::make_shared<Kernel::HLERequestContext>(SharedFrom(this), std::move(thread))};
+    auto context =
+        std::make_shared<HLERequestContext>(kernel, memory, SharedFrom(this), std::move(thread));

    context->PopulateFromIncomingCommandBuffer(kernel.CurrentProcess()->GetHandleTable(), cmd_buf);
    request_queue.Push(std::move(context));
@@ -4,8 +4,8 @@

 #include <fmt/ostream.h>
 #include <httplib.h>
-#include <json.hpp>
 #include <mbedtls/sha256.h>
+#include <nlohmann/json.hpp>
 #include "common/hex_util.h"
 #include "common/logging/backend.h"
 #include "common/logging/log.h"
@@ -767,7 +767,7 @@ FSP_SRV::FSP_SRV(FileSystemController& fsc, const Core::Reporter& reporter)
        {1014, nullptr, "OutputMultiProgramTagAccessLog"},
        {1100, nullptr, "OverrideSaveDataTransferTokenSignVerificationKey"},
        {1110, nullptr, "CorruptSaveDataFileSystemBySaveDataSpaceId2"},
-        {1200, nullptr, "OpenMultiCommitManager"},
+        {1200, &FSP_SRV::OpenMultiCommitManager, "OpenMultiCommitManager"},
        {1300, nullptr, "OpenBisWiper"},
    };
    // clang-format on
@@ -988,4 +988,40 @@ void FSP_SRV::GetAccessLogVersionInfo(Kernel::HLERequestContext& ctx) {
    rb.Push(access_log_program_index);
 }

+class IMultiCommitManager final : public ServiceFramework<IMultiCommitManager> {
+public:
+    explicit IMultiCommitManager() : ServiceFramework("IMultiCommitManager") {
+        static const FunctionInfo functions[] = {
+            {1, &IMultiCommitManager::Add, "Add"},
+            {2, &IMultiCommitManager::Commit, "Commit"},
+        };
+        RegisterHandlers(functions);
+    }
+
+private:
+    FileSys::VirtualFile backend;
+
+    void Add(Kernel::HLERequestContext& ctx) {
+        LOG_WARNING(Service_FS, "(STUBBED) called");
+
+        IPC::ResponseBuilder rb{ctx, 2};
+        rb.Push(RESULT_SUCCESS);
+    }
+
+    void Commit(Kernel::HLERequestContext& ctx) {
+        LOG_WARNING(Service_FS, "(STUBBED) called");
+
+        IPC::ResponseBuilder rb{ctx, 2};
+        rb.Push(RESULT_SUCCESS);
+    }
+};
+
+void FSP_SRV::OpenMultiCommitManager(Kernel::HLERequestContext& ctx) {
+    LOG_DEBUG(Service_FS, "called");
+
+    IPC::ResponseBuilder rb{ctx, 2, 0, 1};
+    rb.Push(RESULT_SUCCESS);
+    rb.PushIpcInterface<IMultiCommitManager>(std::make_shared<IMultiCommitManager>());
+}
+
 } // namespace Service::FileSystem
@@ -50,6 +50,7 @@ private:
    void OpenPatchDataStorageByCurrentProcess(Kernel::HLERequestContext& ctx);
    void OutputAccessLogToSdCard(Kernel::HLERequestContext& ctx);
    void GetAccessLogVersionInfo(Kernel::HLERequestContext& ctx);
+    void OpenMultiCommitManager(Kernel::HLERequestContext& ctx);

    FileSystemController& fsc;

@@ -38,10 +38,11 @@ void Controller_Keyboard::OnUpdate(const Core::Timing::CoreTiming& core_timing,
    cur_entry.sampling_number = last_entry.sampling_number + 1;
    cur_entry.sampling_number2 = cur_entry.sampling_number;

+    cur_entry.key.fill(0);
+    cur_entry.modifier = 0;
+
    for (std::size_t i = 0; i < keyboard_keys.size(); ++i) {
-        for (std::size_t k = 0; k < KEYS_PER_BYTE; ++k) {
-            cur_entry.key[i / KEYS_PER_BYTE] |= (keyboard_keys[i]->GetStatus() << k);
-        }
+        cur_entry.key[i / KEYS_PER_BYTE] |= (keyboard_keys[i]->GetStatus() << (i % KEYS_PER_BYTE));
    }

    for (std::size_t i = 0; i < keyboard_mods.size(); ++i) {
@@ -157,11 +157,11 @@ Hid::Hid(Core::System& system) : ServiceFramework("hid"), system(system) {
        {11, &Hid::ActivateTouchScreen, "ActivateTouchScreen"},
        {21, &Hid::ActivateMouse, "ActivateMouse"},
        {31, &Hid::ActivateKeyboard, "ActivateKeyboard"},
-        {32, nullptr, "SendKeyboardLockKeyEvent"},
+        {32, &Hid::SendKeyboardLockKeyEvent, "SendKeyboardLockKeyEvent"},
        {40, nullptr, "AcquireXpadIdEventHandle"},
        {41, nullptr, "ReleaseXpadIdEventHandle"},
        {51, &Hid::ActivateXpad, "ActivateXpad"},
-        {55, nullptr, "GetXpadIds"},
+        {55, &Hid::GetXpadIDs, "GetXpadIds"},
        {56, nullptr, "ActivateJoyXpad"},
        {58, nullptr, "GetJoyXpadLifoHandle"},
        {59, nullptr, "GetJoyXpadIds"},
@@ -319,6 +319,17 @@ void Hid::ActivateXpad(Kernel::HLERequestContext& ctx) {
    rb.Push(RESULT_SUCCESS);
 }

+void Hid::GetXpadIDs(Kernel::HLERequestContext& ctx) {
+    IPC::RequestParser rp{ctx};
+    const auto applet_resource_user_id{rp.Pop<u64>()};
+
+    LOG_DEBUG(Service_HID, "(STUBBED) called, applet_resource_user_id={}", applet_resource_user_id);
+
+    IPC::ResponseBuilder rb{ctx, 3};
+    rb.Push(RESULT_SUCCESS);
+    rb.Push(0);
+}
+
 void Hid::ActivateDebugPad(Kernel::HLERequestContext& ctx) {
    IPC::RequestParser rp{ctx};
    const auto applet_resource_user_id{rp.Pop<u64>()};
@@ -871,6 +882,15 @@ void Hid::InitializeSevenSixAxisSensor(Kernel::HLERequestContext& ctx) {
    rb.Push(RESULT_SUCCESS);
 }

+void Hid::SendKeyboardLockKeyEvent(Kernel::HLERequestContext& ctx) {
+    IPC::RequestParser rp{ctx};
+    const auto flags{rp.Pop<u32>()};
+    LOG_WARNING(Service_HID, "(STUBBED) called. flags={}", flags);
+
+    IPC::ResponseBuilder rb{ctx, 2};
+    rb.Push(RESULT_SUCCESS);
+}
+
 class HidDbg final : public ServiceFramework<HidDbg> {
 public:
    explicit HidDbg() : ServiceFramework{"hid:dbg"} {
@@ -86,6 +86,7 @@ public:
 private:
    void CreateAppletResource(Kernel::HLERequestContext& ctx);
    void ActivateXpad(Kernel::HLERequestContext& ctx);
+    void GetXpadIDs(Kernel::HLERequestContext& ctx);
    void ActivateDebugPad(Kernel::HLERequestContext& ctx);
    void ActivateTouchScreen(Kernel::HLERequestContext& ctx);
    void ActivateMouse(Kernel::HLERequestContext& ctx);
@@ -130,6 +131,7 @@ private:
    void SetPalmaBoostMode(Kernel::HLERequestContext& ctx);
    void StopSevenSixAxisSensor(Kernel::HLERequestContext& ctx);
    void InitializeSevenSixAxisSensor(Kernel::HLERequestContext& ctx);
+    void SendKeyboardLockKeyEvent(Kernel::HLERequestContext& ctx);

    std::shared_ptr<IAppletResource> applet_resource;
    Core::System& system;
@@ -177,7 +177,8 @@ private:
    void CreateTemporaryNetworkProfile(Kernel::HLERequestContext& ctx) {
        LOG_DEBUG(Service_NIFM, "called");

-        ASSERT_MSG(ctx.GetReadBufferSize() == 0x17c, "NetworkProfileData is not the correct size");
+        ASSERT_MSG(ctx.GetReadBufferSize() == 0x17c,
+                   "SfNetworkProfileData is not the correct size");
        u128 uuid{};
        auto buffer = ctx.ReadBuffer();
        std::memcpy(&uuid, buffer.data() + 8, sizeof(u128));
@@ -25,7 +25,7 @@ u32 nvhost_ctrl_gpu::ioctl(Ioctl command, const std::vector<u8>& input,
    case IoctlCommand::IocGetCharacteristicsCommand:
        return GetCharacteristics(input, output, output2, version);
    case IoctlCommand::IocGetTPCMasksCommand:
-        return GetTPCMasks(input, output);
+        return GetTPCMasks(input, output, output2, version);
    case IoctlCommand::IocGetActiveSlotMaskCommand:
        return GetActiveSlotMask(input, output);
    case IoctlCommand::IocZcullGetCtxSizeCommand:
@@ -98,17 +98,22 @@ u32 nvhost_ctrl_gpu::GetCharacteristics(const std::vector<u8>& input, std::vecto
    return 0;
 }

-u32 nvhost_ctrl_gpu::GetTPCMasks(const std::vector<u8>& input, std::vector<u8>& output) {
+u32 nvhost_ctrl_gpu::GetTPCMasks(const std::vector<u8>& input, std::vector<u8>& output,
+                                 std::vector<u8>& output2, IoctlVersion version) {
    IoctlGpuGetTpcMasksArgs params{};
    std::memcpy(&params, input.data(), input.size());
-    LOG_INFO(Service_NVDRV, "called, mask=0x{:X}, mask_buf_addr=0x{:X}", params.mask_buf_size,
-             params.mask_buf_addr);
-    // TODO(ogniK): Confirm value on hardware
-    if (params.mask_buf_size)
-        params.tpc_mask_size = 4 * 1; // 4 * num_gpc
-    else
-        params.tpc_mask_size = 0;
-    std::memcpy(output.data(), &params, sizeof(params));
+    LOG_DEBUG(Service_NVDRV, "called, mask_buffer_size=0x{:X}", params.mask_buffer_size);
+    if (params.mask_buffer_size != 0) {
+        params.tcp_mask = 3;
+    }
+
+    if (version == IoctlVersion::Version3) {
+        std::memcpy(output.data(), input.data(), output.size());
+        std::memcpy(output2.data(), &params.tcp_mask, output2.size());
+    } else {
+        std::memcpy(output.data(), &params, output.size());
+    }
+
    return 0;
 }

@@ -92,16 +92,11 @@ private:
                  "IoctlCharacteristics is incorrect size");

    struct IoctlGpuGetTpcMasksArgs {
-        /// [in]  TPC mask buffer size reserved by userspace. Should be at least
-        /// sizeof(__u32) * fls(gpc_mask) to receive TPC mask for each GPC.
-        /// [out] full kernel buffer size
-        u32_le mask_buf_size;
-        u32_le reserved;
-
-        /// [in]  pointer to TPC mask buffer. It will receive one 32-bit TPC mask per GPC or 0 if
-        /// GPC is not enabled or not present. This parameter is ignored if mask_buf_size is 0.
-        u64_le mask_buf_addr;
-        u64_le tpc_mask_size; // Nintendo add this?
+        u32_le mask_buffer_size{};
+        INSERT_PADDING_WORDS(1);
+        u64_le mask_buffer_address{};
+        u32_le tcp_mask{};
+        INSERT_PADDING_WORDS(1);
    };
    static_assert(sizeof(IoctlGpuGetTpcMasksArgs) == 24,
                  "IoctlGpuGetTpcMasksArgs is incorrect size");
@@ -166,7 +161,8 @@ private:

    u32 GetCharacteristics(const std::vector<u8>& input, std::vector<u8>& output,
                           std::vector<u8>& output2, IoctlVersion version);
-    u32 GetTPCMasks(const std::vector<u8>& input, std::vector<u8>& output);
+    u32 GetTPCMasks(const std::vector<u8>& input, std::vector<u8>& output, std::vector<u8>& output2,
+                    IoctlVersion version);
    u32 GetActiveSlotMask(const std::vector<u8>& input, std::vector<u8>& output);
    u32 ZCullGetCtxSize(const std::vector<u8>& input, std::vector<u8>& output);
    u32 ZCullGetInfo(const std::vector<u8>& input, std::vector<u8>& output);
@@ -138,9 +138,7 @@ u32 BufferQueue::Query(QueryType type) {

    switch (type) {
    case QueryType::NativeWindowFormat:
-        // TODO(Subv): Use an enum for this
-        static constexpr u32 FormatABGR8 = 1;
-        return FormatABGR8;
+        return static_cast<u32>(PixelFormat::RGBA8888);
    }

    UNIMPLEMENTED();
@@ -66,6 +66,16 @@ public:
        Rotate270 = 0x07,
    };

+    enum class PixelFormat : u32 {
+        RGBA8888 = 1,
+        RGBX8888 = 2,
+        RGB888 = 3,
+        RGB565 = 4,
+        BGRA8888 = 5,
+        RGBA5551 = 6,
+        RRGBA4444 = 7,
+    };
+
    struct Buffer {
        enum class Status { Free = 0, Queued = 1, Dequeued = 2, Acquired = 3 };

@@ -5,6 +5,7 @@
 #include <chrono>
 #include <ctime>

+#include "common/time_zone.h"
 #include "core/hle/service/time/ephemeral_network_system_clock_context_writer.h"
 #include "core/hle/service/time/local_system_clock_context_writer.h"
 #include "core/hle/service/time/network_system_clock_context_writer.h"
@@ -21,8 +22,16 @@ static std::chrono::seconds GetSecondsSinceEpoch() {
           Settings::values.custom_rtc_differential;
 }

+static s64 GetExternalTimeZoneOffset() {
+    // With "auto" timezone setting, we use the external system's timezone offset
+    if (Settings::GetTimeZoneString() == "auto") {
+        return Common::TimeZone::GetCurrentOffsetSeconds().count();
+    }
+    return 0;
+}
+
 static s64 GetExternalRtcValue() {
-    return GetSecondsSinceEpoch().count();
+    return GetSecondsSinceEpoch().count() + GetExternalTimeZoneOffset();
 }

 TimeManager::TimeManager(Core::System& system)
@@ -5,6 +5,7 @@
 #include <sstream>

 #include "common/logging/log.h"
+#include "common/time_zone.h"
 #include "core/core.h"
 #include "core/file_sys/content_archive.h"
 #include "core/file_sys/nca_metadata.h"
@@ -14,6 +15,7 @@
 #include "core/hle/service/filesystem/filesystem.h"
 #include "core/hle/service/time/time_manager.h"
 #include "core/hle/service/time/time_zone_content_manager.h"
+#include "core/settings.h"

 namespace Service::Time::TimeZone {

@@ -68,10 +70,22 @@ static std::vector<std::string> BuildLocationNameCache(Core::System& system) {

 TimeZoneContentManager::TimeZoneContentManager(TimeManager& time_manager, Core::System& system)
    : system{system}, location_name_cache{BuildLocationNameCache(system)} {
-    if (FileSys::VirtualFile vfs_file; GetTimeZoneInfoFile("GMT", vfs_file) == RESULT_SUCCESS) {
+
+    std::string location_name;
+    const auto timezone_setting = Settings::GetTimeZoneString();
+    if (timezone_setting == "auto") {
+        location_name = Common::TimeZone::GetDefaultTimeZone();
+    } else if (timezone_setting == "default") {
+        location_name = location_name;
+    } else {
+        location_name = timezone_setting;
+    }
+
+    if (FileSys::VirtualFile vfs_file;
+        GetTimeZoneInfoFile(location_name, vfs_file) == RESULT_SUCCESS) {
        const auto time_point{
            time_manager.GetStandardSteadyClockCore().GetCurrentTimePoint(system)};
-        time_manager.SetupTimeZoneManager("GMT", time_point, location_name_cache.size(), {},
+        time_manager.SetupTimeZoneManager(location_name, time_point, location_name_cache.size(), {},
                                          vfs_file);
    } else {
        time_zone_manager.MarkAsInitialized();
@@ -113,6 +127,12 @@ ResultCode TimeZoneContentManager::GetTimeZoneInfoFile(const std::string& locati
    }

    vfs_file = zoneinfo_dir->GetFile(location_name);
+    if (!vfs_file) {
+        LOG_ERROR(Service_Time, "{:016X} has no file \"{}\"! Using default timezone.",
+                  time_zone_binary_titleid, location_name);
+        vfs_file = zoneinfo_dir->GetFile(Common::TimeZone::GetDefaultTimeZone());
+    }
+
    if (!vfs_file) {
        LOG_ERROR(Service_Time, "{:016X} has no file \"{}\"!", time_zone_binary_titleid,
                  location_name);
@@ -4,11 +4,12 @@

 #include <ctime>
 #include <fstream>
+#include <iomanip>

 #include <fmt/chrono.h>
 #include <fmt/format.h>
 #include <fmt/ostream.h>
-#include <json.hpp>
+#include <nlohmann/json.hpp>

 #include "common/file_util.h"
 #include "common/hex_util.h"
@@ -63,6 +63,21 @@ const std::array<const char*, NumMouseButtons> mapping = {{

 Values values = {};

+std::string GetTimeZoneString() {
+    static constexpr std::array<const char*, 46> timezones{{
+        "auto",      "default",   "CET", "CST6CDT", "Cuba",    "EET",    "Egypt",     "Eire",
+        "EST",       "EST5EDT",   "GB",  "GB-Eire", "GMT",     "GMT+0",  "GMT-0",     "GMT0",
+        "Greenwich", "Hongkong",  "HST", "Iceland", "Iran",    "Israel", "Jamaica",   "Japan",
+        "Kwajalein", "Libya",     "MET", "MST",     "MST7MDT", "Navajo", "NZ",        "NZ-CHAT",
+        "Poland",    "Portugal",  "PRC", "PST8PDT", "ROC",     "ROK",    "Singapore", "Turkey",
+        "UCT",       "Universal", "UTC", "W-SU",    "WET",     "Zulu",
+    }};
+
+    ASSERT(Settings::values.time_zone_index < timezones.size());
+
+    return timezones[Settings::values.time_zone_index];
+}
+
 void Apply() {
    GDBStub::SetServerPort(values.gdbstub_port);
    GDBStub::ToggleServer(values.use_gdbstub);
@@ -87,6 +102,7 @@ void LogSettings() {
    LogSetting("System_CurrentUser", Settings::values.current_user);
    LogSetting("System_LanguageIndex", Settings::values.language_index);
    LogSetting("System_RegionIndex", Settings::values.region_index);
+    LogSetting("System_TimeZoneIndex", Settings::values.time_zone_index);
    LogSetting("Core_UseMultiCore", Settings::values.use_multi_core);
    LogSetting("Renderer_UseResolutionFactor", Settings::values.resolution_factor);
    LogSetting("Renderer_UseFrameLimit", Settings::values.use_frame_limit);
@@ -96,6 +112,7 @@ void LogSettings() {
    LogSetting("Renderer_UseAsynchronousGpuEmulation",
               Settings::values.use_asynchronous_gpu_emulation);
    LogSetting("Renderer_UseVsync", Settings::values.use_vsync);
+    LogSetting("Renderer_UseAssemblyShaders", Settings::values.use_assembly_shaders);
    LogSetting("Renderer_AnisotropicFilteringLevel", Settings::values.max_anisotropy);
    LogSetting("Audio_OutputEngine", Settings::values.sink_id);
    LogSetting("Audio_EnableAudioStretching", Settings::values.enable_audio_stretching);
@@ -394,6 +394,7 @@ struct Values {
    s32 current_user;
    s32 language_index;
    s32 region_index;
+    s32 time_zone_index;
    s32 sound_index;

    // Controls
@@ -436,7 +437,7 @@ struct Values {
    bool renderer_debug;
    int vulkan_device;

-    float resolution_factor;
+    u16 resolution_factor{1};
    int aspect_ratio;
    int max_anisotropy;
    bool use_frame_limit;
@@ -445,6 +446,7 @@ struct Values {
    GPUAccuracy gpu_accuracy;
    bool use_asynchronous_gpu_emulation;
    bool use_vsync;
+    bool use_assembly_shaders;
    bool force_30fps_mode;
    bool use_fast_gpu_time;

@@ -472,6 +474,7 @@ struct Values {
    bool reporting_services;
    bool quest_flag;
    bool disable_cpu_opt;
+    bool disable_macro_jit;

    // BCAT
    std::string bcat_backend;
@@ -490,6 +493,9 @@ struct Values {
 bool IsGPULevelExtreme();
 bool IsGPULevelHigh();

+std::string GetTimeZoneString();
+
 void Apply();
 void LogSettings();
+
 } // namespace Settings
@@ -201,6 +201,7 @@ void TelemetrySession::AddInitialInfo(Loader::AppLoader& app_loader) {
    AddField(field_type, "Renderer_UseAsynchronousGpuEmulation",
             Settings::values.use_asynchronous_gpu_emulation);
    AddField(field_type, "Renderer_UseVsync", Settings::values.use_vsync);
+    AddField(field_type, "Renderer_UseAssemblyShaders", Settings::values.use_assembly_shaders);
    AddField(field_type, "System_UseDockedMode", Settings::values.use_docked_mode);
 }

@@ -1,6 +1,7 @@
 add_library(video_core STATIC
    buffer_cache/buffer_block.h
    buffer_cache/buffer_cache.h
+    buffer_cache/map_interval.cpp
    buffer_cache/map_interval.h
    dirty_flags.cpp
    dirty_flags.h
@@ -24,6 +25,12 @@ add_library(video_core STATIC
    engines/shader_bytecode.h
    engines/shader_header.h
    engines/shader_type.h
+    macro/macro.cpp
+    macro/macro.h
+    macro/macro_interpreter.cpp
+    macro/macro_interpreter.h
+    macro/macro_jit_x64.cpp
+    macro/macro_jit_x64.h
    fence_manager.h
    gpu.cpp
    gpu.h
@@ -35,8 +42,6 @@ add_library(video_core STATIC
    gpu_thread.h
    guest_driver.cpp
    guest_driver.h
-    macro_interpreter.cpp
-    macro_interpreter.h
    memory_manager.cpp
    memory_manager.h
    morton.cpp
@@ -44,11 +49,11 @@ add_library(video_core STATIC
    query_cache.h
    rasterizer_accelerated.cpp
    rasterizer_accelerated.h
-    rasterizer_cache.cpp
-    rasterizer_cache.h
    rasterizer_interface.h
    renderer_base.cpp
    renderer_base.h
+    renderer_opengl/gl_arb_decompiler.cpp
+    renderer_opengl/gl_arb_decompiler.h
    renderer_opengl/gl_buffer_cache.cpp
    renderer_opengl/gl_buffer_cache.h
    renderer_opengl/gl_device.cpp
@@ -88,6 +93,7 @@ add_library(video_core STATIC
    renderer_opengl/utils.h
    sampler_cache.cpp
    sampler_cache.h
+    shader_cache.h
    shader/decode/arithmetic.cpp
    shader/decode/arithmetic_immediate.cpp
    shader/decode/bfe.cpp
@@ -228,7 +234,7 @@ endif()
 create_target_directory_groups(video_core)

 target_link_libraries(video_core PUBLIC common core)
-target_link_libraries(video_core PRIVATE glad)
+target_link_libraries(video_core PRIVATE glad xbyak)

 if (ENABLE_VULKAN)
    target_include_directories(video_core PRIVATE sirit ../../externals/Vulkan-Headers/include)
@@ -15,48 +15,47 @@ namespace VideoCommon {

 class BufferBlock {
 public:
-    bool Overlaps(const VAddr start, const VAddr end) const {
+    bool Overlaps(VAddr start, VAddr end) const {
        return (cpu_addr < end) && (cpu_addr_end > start);
    }

-    bool IsInside(const VAddr other_start, const VAddr other_end) const {
+    bool IsInside(VAddr other_start, VAddr other_end) const {
        return cpu_addr <= other_start && other_end <= cpu_addr_end;
    }

-    std::size_t GetOffset(const VAddr in_addr) {
+    std::size_t Offset(VAddr in_addr) const {
        return static_cast<std::size_t>(in_addr - cpu_addr);
    }

-    VAddr GetCpuAddr() const {
+    VAddr CpuAddr() const {
        return cpu_addr;
    }

-    VAddr GetCpuAddrEnd() const {
+    VAddr CpuAddrEnd() const {
        return cpu_addr_end;
    }

-    void SetCpuAddr(const VAddr new_addr) {
+    void SetCpuAddr(VAddr new_addr) {
        cpu_addr = new_addr;
        cpu_addr_end = new_addr + size;
    }

-    std::size_t GetSize() const {
+    std::size_t Size() const {
        return size;
    }

+    u64 Epoch() const {
+        return epoch;
+    }
+
    void SetEpoch(u64 new_epoch) {
        epoch = new_epoch;
    }

-    u64 GetEpoch() {
-        return epoch;
-    }
-
 protected:
-    explicit BufferBlock(VAddr cpu_addr, const std::size_t size) : size{size} {
-        SetCpuAddr(cpu_addr);
+    explicit BufferBlock(VAddr cpu_addr_, std::size_t size_) : size{size_} {
+        SetCpuAddr(cpu_addr_);
    }
-    ~BufferBlock() = default;

 private:
    VAddr cpu_addr{};
@@ -12,11 +12,12 @@
 #include <utility>
 #include <vector>

-#include <boost/icl/interval_map.hpp>
+#include <boost/container/small_vector.hpp>
 #include <boost/icl/interval_set.hpp>
-#include <boost/range/iterator_range.hpp>
+#include <boost/intrusive/set.hpp>

 #include "common/alignment.h"
+#include "common/assert.h"
 #include "common/common_types.h"
 #include "common/logging/log.h"
 #include "core/core.h"
@@ -29,10 +30,16 @@

 namespace VideoCommon {

-using MapInterval = std::shared_ptr<MapIntervalBase>;
-
-template <typename OwnerBuffer, typename BufferType, typename StreamBuffer>
+template <typename Buffer, typename BufferType, typename StreamBuffer>
 class BufferCache {
+    using IntervalSet = boost::icl::interval_set<VAddr>;
+    using IntervalType = typename IntervalSet::interval_type;
+    using VectorMapInterval = boost::container::small_vector<MapInterval*, 1>;
+
+    static constexpr u64 WRITE_PAGE_BIT = 11;
+    static constexpr u64 BLOCK_PAGE_BITS = 21;
+    static constexpr u64 BLOCK_PAGE_SIZE = 1ULL << BLOCK_PAGE_BITS;
+
 public:
    using BufferInfo = std::pair<BufferType, u64>;

@@ -40,14 +47,12 @@ public:
                            bool is_written = false, bool use_fast_cbuf = false) {
        std::lock_guard lock{mutex};

-        const std::optional<VAddr> cpu_addr_opt =
-            system.GPU().MemoryManager().GpuToCpuAddress(gpu_addr);
-
+        const auto& memory_manager = system.GPU().MemoryManager();
+        const std::optional<VAddr> cpu_addr_opt = memory_manager.GpuToCpuAddress(gpu_addr);
        if (!cpu_addr_opt) {
            return {GetEmptyBuffer(size), 0};
        }
-
-        VAddr cpu_addr = *cpu_addr_opt;
+        const VAddr cpu_addr = *cpu_addr_opt;

        // Cache management is a big overhead, so only cache entries with a given size.
        // TODO: Figure out which size is the best for given games.
@@ -55,76 +60,91 @@ public:
        if (use_fast_cbuf || size < max_stream_size) {
            if (!is_written && !IsRegionWritten(cpu_addr, cpu_addr + size - 1)) {
                auto& memory_manager = system.GPU().MemoryManager();
+                const bool is_granular = memory_manager.IsGranularRange(gpu_addr, size);
                if (use_fast_cbuf) {
-                    if (memory_manager.IsGranularRange(gpu_addr, size)) {
-                        const auto host_ptr = memory_manager.GetPointer(gpu_addr);
-                        return ConstBufferUpload(host_ptr, size);
+                    u8* dest;
+                    if (is_granular) {
+                        dest = memory_manager.GetPointer(gpu_addr);
                    } else {
                        staging_buffer.resize(size);
-                        memory_manager.ReadBlockUnsafe(gpu_addr, staging_buffer.data(), size);
-                        return ConstBufferUpload(staging_buffer.data(), size);
+                        dest = staging_buffer.data();
+                        memory_manager.ReadBlockUnsafe(gpu_addr, dest, size);
                    }
+                    return ConstBufferUpload(dest, size);
+                }
+                if (is_granular) {
+                    u8* const host_ptr = memory_manager.GetPointer(gpu_addr);
+                    return StreamBufferUpload(size, alignment, [host_ptr, size](u8* dest) {
+                        std::memcpy(dest, host_ptr, size);
+                    });
                } else {
-                    if (memory_manager.IsGranularRange(gpu_addr, size)) {
-                        const auto host_ptr = memory_manager.GetPointer(gpu_addr);
-                        return StreamBufferUpload(host_ptr, size, alignment);
-                    } else {
-                        staging_buffer.resize(size);
-                        memory_manager.ReadBlockUnsafe(gpu_addr, staging_buffer.data(), size);
-                        return StreamBufferUpload(staging_buffer.data(), size, alignment);
-                    }
+                    return StreamBufferUpload(
+                        size, alignment, [&memory_manager, gpu_addr, size](u8* dest) {
+                            memory_manager.ReadBlockUnsafe(gpu_addr, dest, size);
+                        });
                }
            }
        }

-        auto block = GetBlock(cpu_addr, size);
-        auto map = MapAddress(block, gpu_addr, cpu_addr, size);
+        Buffer* const block = GetBlock(cpu_addr, size);
+        MapInterval* const map = MapAddress(block, gpu_addr, cpu_addr, size);
+        if (!map) {
+            return {GetEmptyBuffer(size), 0};
+        }
        if (is_written) {
            map->MarkAsModified(true, GetModifiedTicks());
            if (Settings::IsGPULevelHigh() && Settings::values.use_asynchronous_gpu_emulation) {
                MarkForAsyncFlush(map);
            }
-            if (!map->IsWritten()) {
-                map->MarkAsWritten(true);
-                MarkRegionAsWritten(map->GetStart(), map->GetEnd() - 1);
+            if (!map->is_written) {
+                map->is_written = true;
+                MarkRegionAsWritten(map->start, map->end - 1);
            }
        }

-        return {ToHandle(block), static_cast<u64>(block->GetOffset(cpu_addr))};
+        return {block->Handle(), static_cast<u64>(block->Offset(cpu_addr))};
    }

    /// Uploads from a host memory. Returns the OpenGL buffer where it's located and its offset.
    BufferInfo UploadHostMemory(const void* raw_pointer, std::size_t size,
                                std::size_t alignment = 4) {
        std::lock_guard lock{mutex};
-        return StreamBufferUpload(raw_pointer, size, alignment);
+        return StreamBufferUpload(size, alignment, [raw_pointer, size](u8* dest) {
+            std::memcpy(dest, raw_pointer, size);
+        });
    }

-    void Map(std::size_t max_size) {
+    /// Prepares the buffer cache for data uploading
+    /// @param max_size Maximum number of bytes that will be uploaded
+    /// @return True when a stream buffer invalidation was required, false otherwise
+    bool Map(std::size_t max_size) {
        std::lock_guard lock{mutex};

+        bool invalidated;
        std::tie(buffer_ptr, buffer_offset_base, invalidated) = stream_buffer->Map(max_size, 4);
        buffer_offset = buffer_offset_base;
+
+        return invalidated;
    }

-    /// Finishes the upload stream, returns true on bindings invalidation.
-    bool Unmap() {
+    /// Finishes the upload stream
+    void Unmap() {
        std::lock_guard lock{mutex};
-
        stream_buffer->Unmap(buffer_offset - buffer_offset_base);
-        return std::exchange(invalidated, false);
    }

+    /// Function called at the end of each frame, inteded for deferred operations
    void TickFrame() {
        ++epoch;
+
        while (!pending_destruction.empty()) {
            // Delay at least 4 frames before destruction.
            // This is due to triple buffering happening on some drivers.
            static constexpr u64 epochs_to_destroy = 5;
-            if (pending_destruction.front()->GetEpoch() + epochs_to_destroy > epoch) {
+            if (pending_destruction.front()->Epoch() + epochs_to_destroy > epoch) {
                break;
            }
-            pending_destruction.pop_front();
+            pending_destruction.pop();
        }
    }

@@ -132,12 +152,11 @@ public:
    void FlushRegion(VAddr addr, std::size_t size) {
        std::lock_guard lock{mutex};

-        std::vector<MapInterval> objects = GetMapsInRange(addr, size);
-        std::sort(objects.begin(), objects.end(), [](const MapInterval& a, const MapInterval& b) {
-            return a->GetModificationTick() < b->GetModificationTick();
-        });
-        for (auto& object : objects) {
-            if (object->IsModified() && object->IsRegistered()) {
+        VectorMapInterval objects = GetMapsInRange(addr, size);
+        std::sort(objects.begin(), objects.end(),
+                  [](MapInterval* lhs, MapInterval* rhs) { return lhs->ticks < rhs->ticks; });
+        for (MapInterval* object : objects) {
+            if (object->is_modified && object->is_registered) {
                mutex.unlock();
                FlushMap(object);
                mutex.lock();
@@ -148,9 +167,9 @@ public:
    bool MustFlushRegion(VAddr addr, std::size_t size) {
        std::lock_guard lock{mutex};

-        const std::vector<MapInterval> objects = GetMapsInRange(addr, size);
-        return std::any_of(objects.cbegin(), objects.cend(), [](const MapInterval& map) {
-            return map->IsModified() && map->IsRegistered();
+        const VectorMapInterval objects = GetMapsInRange(addr, size);
+        return std::any_of(objects.cbegin(), objects.cend(), [](const MapInterval* map) {
+            return map->is_modified && map->is_registered;
        });
    }

@@ -158,9 +177,8 @@ public:
    void InvalidateRegion(VAddr addr, u64 size) {
        std::lock_guard lock{mutex};

-        std::vector<MapInterval> objects = GetMapsInRange(addr, size);
-        for (auto& object : objects) {
-            if (object->IsRegistered()) {
+        for (auto& object : GetMapsInRange(addr, size)) {
+            if (object->is_registered) {
                Unregister(object);
            }
        }
@@ -169,10 +187,10 @@ public:
    void OnCPUWrite(VAddr addr, std::size_t size) {
        std::lock_guard lock{mutex};

-        for (const auto& object : GetMapsInRange(addr, size)) {
-            if (object->IsMemoryMarked() && object->IsRegistered()) {
+        for (MapInterval* object : GetMapsInRange(addr, size)) {
+            if (object->is_memory_marked && object->is_registered) {
                UnmarkMemory(object);
-                object->SetSyncPending(true);
+                object->is_sync_pending = true;
                marked_for_unregister.emplace_back(object);
            }
        }
@@ -181,9 +199,9 @@ public:
    void SyncGuestHost() {
        std::lock_guard lock{mutex};

-        for (const auto& object : marked_for_unregister) {
-            if (object->IsRegistered()) {
-                object->SetSyncPending(false);
+        for (auto& object : marked_for_unregister) {
+            if (object->is_registered) {
+                object->is_sync_pending = false;
                Unregister(object);
            }
        }
@@ -192,9 +210,9 @@ public:

    void CommitAsyncFlushes() {
        if (uncommitted_flushes) {
-            auto commit_list = std::make_shared<std::list<MapInterval>>();
-            for (auto& map : *uncommitted_flushes) {
-                if (map->IsRegistered() && map->IsModified()) {
+            auto commit_list = std::make_shared<std::list<MapInterval*>>();
+            for (MapInterval* map : *uncommitted_flushes) {
+                if (map->is_registered && map->is_modified) {
                    // TODO(Blinkhawk): Implement backend asynchronous flushing
                    // AsyncFlushMap(map)
                    commit_list->push_back(map);
@@ -228,8 +246,8 @@ public:
            committed_flushes.pop_front();
            return;
        }
-        for (MapInterval& map : *flush_list) {
-            if (map->IsRegistered()) {
+        for (MapInterval* map : *flush_list) {
+            if (map->is_registered) {
                // TODO(Blinkhawk): Replace this for reading the asynchronous flush
                FlushMap(map);
            }
@@ -241,23 +259,21 @@ public:

 protected:
    explicit BufferCache(VideoCore::RasterizerInterface& rasterizer, Core::System& system,
-                         std::unique_ptr<StreamBuffer> stream_buffer)
-        : rasterizer{rasterizer}, system{system}, stream_buffer{std::move(stream_buffer)},
-          stream_buffer_handle{this->stream_buffer->GetHandle()} {}
+                         std::unique_ptr<StreamBuffer> stream_buffer_)
+        : rasterizer{rasterizer}, system{system}, stream_buffer{std::move(stream_buffer_)},
+          stream_buffer_handle{stream_buffer->Handle()} {}

    ~BufferCache() = default;

-    virtual BufferType ToHandle(const OwnerBuffer& storage) = 0;
+    virtual std::shared_ptr<Buffer> CreateBlock(VAddr cpu_addr, std::size_t size) = 0;

-    virtual OwnerBuffer CreateBlock(VAddr cpu_addr, std::size_t size) = 0;
-
-    virtual void UploadBlockData(const OwnerBuffer& buffer, std::size_t offset, std::size_t size,
+    virtual void UploadBlockData(const Buffer& buffer, std::size_t offset, std::size_t size,
                                 const u8* data) = 0;

-    virtual void DownloadBlockData(const OwnerBuffer& buffer, std::size_t offset, std::size_t size,
+    virtual void DownloadBlockData(const Buffer& buffer, std::size_t offset, std::size_t size,
                                   u8* data) = 0;

-    virtual void CopyBlock(const OwnerBuffer& src, const OwnerBuffer& dst, std::size_t src_offset,
+    virtual void CopyBlock(const Buffer& src, const Buffer& dst, std::size_t src_offset,
                           std::size_t dst_offset, std::size_t size) = 0;

    virtual BufferInfo ConstBufferUpload(const void* raw_pointer, std::size_t size) {
@@ -265,76 +281,74 @@ protected:
    }

    /// Register an object into the cache
-    void Register(const MapInterval& new_map, bool inherit_written = false) {
-        const VAddr cpu_addr = new_map->GetStart();
+    MapInterval* Register(MapInterval new_map, bool inherit_written = false) {
+        const VAddr cpu_addr = new_map.start;
        if (!cpu_addr) {
            LOG_CRITICAL(HW_GPU, "Failed to register buffer with unmapped gpu_address 0x{:016x}",
-                         new_map->GetGpuAddress());
-            return;
+                         new_map.gpu_addr);
+            return nullptr;
        }
-        const std::size_t size = new_map->GetEnd() - new_map->GetStart();
-        new_map->MarkAsRegistered(true);
-        const IntervalType interval{new_map->GetStart(), new_map->GetEnd()};
-        mapped_addresses.insert({interval, new_map});
+        const std::size_t size = new_map.end - new_map.start;
+        new_map.is_registered = true;
        rasterizer.UpdatePagesCachedCount(cpu_addr, size, 1);
-        new_map->SetMemoryMarked(true);
+        new_map.is_memory_marked = true;
        if (inherit_written) {
-            MarkRegionAsWritten(new_map->GetStart(), new_map->GetEnd() - 1);
-            new_map->MarkAsWritten(true);
+            MarkRegionAsWritten(new_map.start, new_map.end - 1);
+            new_map.is_written = true;
        }
+        MapInterval* const storage = mapped_addresses_allocator.Allocate();
+        *storage = new_map;
+        mapped_addresses.insert(*storage);
+        return storage;
    }

-    void UnmarkMemory(const MapInterval& map) {
-        if (!map->IsMemoryMarked()) {
+    void UnmarkMemory(MapInterval* map) {
+        if (!map->is_memory_marked) {
            return;
        }
-        const std::size_t size = map->GetEnd() - map->GetStart();
-        rasterizer.UpdatePagesCachedCount(map->GetStart(), size, -1);
-        map->SetMemoryMarked(false);
+        const std::size_t size = map->end - map->start;
+        rasterizer.UpdatePagesCachedCount(map->start, size, -1);
+        map->is_memory_marked = false;
    }

    /// Unregisters an object from the cache
-    void Unregister(const MapInterval& map) {
+    void Unregister(MapInterval* map) {
        UnmarkMemory(map);
-        map->MarkAsRegistered(false);
-        if (map->IsSyncPending()) {
+        map->is_registered = false;
+        if (map->is_sync_pending) {
+            map->is_sync_pending = false;
            marked_for_unregister.remove(map);
-            map->SetSyncPending(false);
        }
-        if (map->IsWritten()) {
-            UnmarkRegionAsWritten(map->GetStart(), map->GetEnd() - 1);
+        if (map->is_written) {
+            UnmarkRegionAsWritten(map->start, map->end - 1);
        }
-        const IntervalType delete_interval{map->GetStart(), map->GetEnd()};
-        mapped_addresses.erase(delete_interval);
+        const auto it = mapped_addresses.find(*map);
+        ASSERT(it != mapped_addresses.end());
+        mapped_addresses.erase(it);
+        mapped_addresses_allocator.Release(map);
    }

 private:
-    MapInterval CreateMap(const VAddr start, const VAddr end, const GPUVAddr gpu_addr) {
-        return std::make_shared<MapIntervalBase>(start, end, gpu_addr);
-    }
-
-    MapInterval MapAddress(const OwnerBuffer& block, const GPUVAddr gpu_addr, const VAddr cpu_addr,
-                           const std::size_t size) {
-        std::vector<MapInterval> overlaps = GetMapsInRange(cpu_addr, size);
+    MapInterval* MapAddress(const Buffer* block, GPUVAddr gpu_addr, VAddr cpu_addr,
+                            std::size_t size) {
+        const VectorMapInterval overlaps = GetMapsInRange(cpu_addr, size);
        if (overlaps.empty()) {
            auto& memory_manager = system.GPU().MemoryManager();
            const VAddr cpu_addr_end = cpu_addr + size;
-            MapInterval new_map = CreateMap(cpu_addr, cpu_addr_end, gpu_addr);
            if (memory_manager.IsGranularRange(gpu_addr, size)) {
                u8* host_ptr = memory_manager.GetPointer(gpu_addr);
-                UploadBlockData(block, block->GetOffset(cpu_addr), size, host_ptr);
+                UploadBlockData(*block, block->Offset(cpu_addr), size, host_ptr);
            } else {
                staging_buffer.resize(size);
                memory_manager.ReadBlockUnsafe(gpu_addr, staging_buffer.data(), size);
-                UploadBlockData(block, block->GetOffset(cpu_addr), size, staging_buffer.data());
+                UploadBlockData(*block, block->Offset(cpu_addr), size, staging_buffer.data());
            }
-            Register(new_map);
-            return new_map;
+            return Register(MapInterval(cpu_addr, cpu_addr_end, gpu_addr));
        }

        const VAddr cpu_addr_end = cpu_addr + size;
        if (overlaps.size() == 1) {
-            MapInterval& current_map = overlaps[0];
+            MapInterval* const current_map = overlaps[0];
            if (current_map->IsInside(cpu_addr, cpu_addr_end)) {
                return current_map;
            }
@@ -344,60 +358,70 @@ private:
        bool write_inheritance = false;
        bool modified_inheritance = false;
        // Calculate new buffer parameters
-        for (auto& overlap : overlaps) {
-            new_start = std::min(overlap->GetStart(), new_start);
-            new_end = std::max(overlap->GetEnd(), new_end);
-            write_inheritance |= overlap->IsWritten();
-            modified_inheritance |= overlap->IsModified();
+        for (MapInterval* overlap : overlaps) {
+            new_start = std::min(overlap->start, new_start);
+            new_end = std::max(overlap->end, new_end);
+            write_inheritance |= overlap->is_written;
+            modified_inheritance |= overlap->is_modified;
        }
        GPUVAddr new_gpu_addr = gpu_addr + new_start - cpu_addr;
        for (auto& overlap : overlaps) {
            Unregister(overlap);
        }
        UpdateBlock(block, new_start, new_end, overlaps);
-        MapInterval new_map = CreateMap(new_start, new_end, new_gpu_addr);
+
+        const MapInterval new_map{new_start, new_end, new_gpu_addr};
+        MapInterval* const map = Register(new_map, write_inheritance);
+        if (!map) {
+            return nullptr;
+        }
        if (modified_inheritance) {
-            new_map->MarkAsModified(true, GetModifiedTicks());
+            map->MarkAsModified(true, GetModifiedTicks());
            if (Settings::IsGPULevelHigh() && Settings::values.use_asynchronous_gpu_emulation) {
-                MarkForAsyncFlush(new_map);
+                MarkForAsyncFlush(map);
            }
        }
-        Register(new_map, write_inheritance);
-        return new_map;
+        return map;
    }

-    void UpdateBlock(const OwnerBuffer& block, VAddr start, VAddr end,
-                     std::vector<MapInterval>& overlaps) {
+    void UpdateBlock(const Buffer* block, VAddr start, VAddr end,
+                     const VectorMapInterval& overlaps) {
        const IntervalType base_interval{start, end};
        IntervalSet interval_set{};
        interval_set.add(base_interval);
        for (auto& overlap : overlaps) {
-            const IntervalType subtract{overlap->GetStart(), overlap->GetEnd()};
+            const IntervalType subtract{overlap->start, overlap->end};
            interval_set.subtract(subtract);
        }
        for (auto& interval : interval_set) {
-            std::size_t size = interval.upper() - interval.lower();
-            if (size > 0) {
-                staging_buffer.resize(size);
-                system.Memory().ReadBlockUnsafe(interval.lower(), staging_buffer.data(), size);
-                UploadBlockData(block, block->GetOffset(interval.lower()), size,
-                                staging_buffer.data());
+            const std::size_t size = interval.upper() - interval.lower();
+            if (size == 0) {
+                continue;
            }
+            staging_buffer.resize(size);
+            system.Memory().ReadBlockUnsafe(interval.lower(), staging_buffer.data(), size);
+            UploadBlockData(*block, block->Offset(interval.lower()), size, staging_buffer.data());
        }
    }

-    std::vector<MapInterval> GetMapsInRange(VAddr addr, std::size_t size) {
+    VectorMapInterval GetMapsInRange(VAddr addr, std::size_t size) {
+        VectorMapInterval result;
        if (size == 0) {
-            return {};
+            return result;
        }

-        std::vector<MapInterval> objects{};
-        const IntervalType interval{addr, addr + size};
-        for (auto& pair : boost::make_iterator_range(mapped_addresses.equal_range(interval))) {
-            objects.push_back(pair.second);
+        const VAddr addr_end = addr + size;
+        auto it = mapped_addresses.lower_bound(addr);
+        if (it != mapped_addresses.begin()) {
+            --it;
        }
-
-        return objects;
+        while (it != mapped_addresses.end() && it->start < addr_end) {
+            if (it->Overlaps(addr, addr_end)) {
+                result.push_back(&*it);
+            }
+            ++it;
+        }
+        return result;
    }

    /// Returns a ticks counter used for tracking when cached objects were last modified
@@ -405,20 +429,24 @@ private:
        return ++modified_ticks;
    }

-    void FlushMap(MapInterval map) {
-        std::size_t size = map->GetEnd() - map->GetStart();
-        OwnerBuffer block = blocks[map->GetStart() >> block_page_bits];
+    void FlushMap(MapInterval* map) {
+        const auto it = blocks.find(map->start >> BLOCK_PAGE_BITS);
+        ASSERT_OR_EXECUTE(it != blocks.end(), return;);
+
+        std::shared_ptr<Buffer> block = it->second;
+
+        const std::size_t size = map->end - map->start;
        staging_buffer.resize(size);
-        DownloadBlockData(block, block->GetOffset(map->GetStart()), size, staging_buffer.data());
-        system.Memory().WriteBlockUnsafe(map->GetStart(), staging_buffer.data(), size);
+        DownloadBlockData(*block, block->Offset(map->start), size, staging_buffer.data());
+        system.Memory().WriteBlockUnsafe(map->start, staging_buffer.data(), size);
        map->MarkAsModified(false, 0);
    }

-    BufferInfo StreamBufferUpload(const void* raw_pointer, std::size_t size,
-                                  std::size_t alignment) {
+    template <typename Callable>
+    BufferInfo StreamBufferUpload(std::size_t size, std::size_t alignment, Callable&& callable) {
        AlignBuffer(alignment);
        const std::size_t uploaded_offset = buffer_offset;
-        std::memcpy(buffer_ptr, raw_pointer, size);
+        callable(buffer_ptr);

        buffer_ptr += size;
        buffer_offset += size;
@@ -432,97 +460,89 @@ private:
        buffer_offset = offset_aligned;
    }

-    OwnerBuffer EnlargeBlock(OwnerBuffer buffer) {
-        const std::size_t old_size = buffer->GetSize();
-        const std::size_t new_size = old_size + block_page_size;
-        const VAddr cpu_addr = buffer->GetCpuAddr();
-        OwnerBuffer new_buffer = CreateBlock(cpu_addr, new_size);
-        CopyBlock(buffer, new_buffer, 0, 0, old_size);
-        buffer->SetEpoch(epoch);
-        pending_destruction.push_back(buffer);
+    std::shared_ptr<Buffer> EnlargeBlock(std::shared_ptr<Buffer> buffer) {
+        const std::size_t old_size = buffer->Size();
+        const std::size_t new_size = old_size + BLOCK_PAGE_SIZE;
+        const VAddr cpu_addr = buffer->CpuAddr();
+        std::shared_ptr<Buffer> new_buffer = CreateBlock(cpu_addr, new_size);
+        CopyBlock(*buffer, *new_buffer, 0, 0, old_size);
+        QueueDestruction(std::move(buffer));
+
        const VAddr cpu_addr_end = cpu_addr + new_size - 1;
-        u64 page_start = cpu_addr >> block_page_bits;
-        const u64 page_end = cpu_addr_end >> block_page_bits;
-        while (page_start <= page_end) {
-            blocks[page_start] = new_buffer;
-            ++page_start;
+        const u64 page_end = cpu_addr_end >> BLOCK_PAGE_BITS;
+        for (u64 page_start = cpu_addr >> BLOCK_PAGE_BITS; page_start <= page_end; ++page_start) {
+            blocks.insert_or_assign(page_start, new_buffer);
        }
+
        return new_buffer;
    }

-    OwnerBuffer MergeBlocks(OwnerBuffer first, OwnerBuffer second) {
-        const std::size_t size_1 = first->GetSize();
-        const std::size_t size_2 = second->GetSize();
-        const VAddr first_addr = first->GetCpuAddr();
-        const VAddr second_addr = second->GetCpuAddr();
+    std::shared_ptr<Buffer> MergeBlocks(std::shared_ptr<Buffer> first,
+                                        std::shared_ptr<Buffer> second) {
+        const std::size_t size_1 = first->Size();
+        const std::size_t size_2 = second->Size();
+        const VAddr first_addr = first->CpuAddr();
+        const VAddr second_addr = second->CpuAddr();
        const VAddr new_addr = std::min(first_addr, second_addr);
        const std::size_t new_size = size_1 + size_2;
-        OwnerBuffer new_buffer = CreateBlock(new_addr, new_size);
-        CopyBlock(first, new_buffer, 0, new_buffer->GetOffset(first_addr), size_1);
-        CopyBlock(second, new_buffer, 0, new_buffer->GetOffset(second_addr), size_2);
-        first->SetEpoch(epoch);
-        second->SetEpoch(epoch);
-        pending_destruction.push_back(first);
-        pending_destruction.push_back(second);
+
+        std::shared_ptr<Buffer> new_buffer = CreateBlock(new_addr, new_size);
+        CopyBlock(*first, *new_buffer, 0, new_buffer->Offset(first_addr), size_1);
+        CopyBlock(*second, *new_buffer, 0, new_buffer->Offset(second_addr), size_2);
+        QueueDestruction(std::move(first));
+        QueueDestruction(std::move(second));
+
        const VAddr cpu_addr_end = new_addr + new_size - 1;
-        u64 page_start = new_addr >> block_page_bits;
-        const u64 page_end = cpu_addr_end >> block_page_bits;
-        while (page_start <= page_end) {
-            blocks[page_start] = new_buffer;
-            ++page_start;
+        const u64 page_end = cpu_addr_end >> BLOCK_PAGE_BITS;
+        for (u64 page_start = new_addr >> BLOCK_PAGE_BITS; page_start <= page_end; ++page_start) {
+            blocks.insert_or_assign(page_start, new_buffer);
        }
        return new_buffer;
    }

-    OwnerBuffer GetBlock(const VAddr cpu_addr, const std::size_t size) {
-        OwnerBuffer found;
+    Buffer* GetBlock(VAddr cpu_addr, std::size_t size) {
+        std::shared_ptr<Buffer> found;
+
        const VAddr cpu_addr_end = cpu_addr + size - 1;
-        u64 page_start = cpu_addr >> block_page_bits;
-        const u64 page_end = cpu_addr_end >> block_page_bits;
-        while (page_start <= page_end) {
+        const u64 page_end = cpu_addr_end >> BLOCK_PAGE_BITS;
+        for (u64 page_start = cpu_addr >> BLOCK_PAGE_BITS; page_start <= page_end; ++page_start) {
            auto it = blocks.find(page_start);
            if (it == blocks.end()) {
                if (found) {
                    found = EnlargeBlock(found);
-                } else {
-                    const VAddr start_addr = (page_start << block_page_bits);
-                    found = CreateBlock(start_addr, block_page_size);
-                    blocks[page_start] = found;
-                }
-            } else {
-                if (found) {
-                    if (found == it->second) {
-                        ++page_start;
-                        continue;
-                    }
-                    found = MergeBlocks(found, it->second);
-                } else {
-                    found = it->second;
+                    continue;
                }
+                const VAddr start_addr = page_start << BLOCK_PAGE_BITS;
+                found = CreateBlock(start_addr, BLOCK_PAGE_SIZE);
+                blocks.insert_or_assign(page_start, found);
+                continue;
+            }
+            if (!found) {
+                found = it->second;
+                continue;
+            }
+            if (found != it->second) {
+                found = MergeBlocks(std::move(found), it->second);
            }
-            ++page_start;
        }
-        return found;
+        return found.get();
    }

-    void MarkRegionAsWritten(const VAddr start, const VAddr end) {
-        u64 page_start = start >> write_page_bit;
-        const u64 page_end = end >> write_page_bit;
-        while (page_start <= page_end) {
+    void MarkRegionAsWritten(VAddr start, VAddr end) {
+        const u64 page_end = end >> WRITE_PAGE_BIT;
+        for (u64 page_start = start >> WRITE_PAGE_BIT; page_start <= page_end; ++page_start) {
            auto it = written_pages.find(page_start);
            if (it != written_pages.end()) {
                it->second = it->second + 1;
            } else {
-                written_pages[page_start] = 1;
+                written_pages.insert_or_assign(page_start, 1);
            }
-            page_start++;
        }
    }

-    void UnmarkRegionAsWritten(const VAddr start, const VAddr end) {
-        u64 page_start = start >> write_page_bit;
-        const u64 page_end = end >> write_page_bit;
-        while (page_start <= page_end) {
+    void UnmarkRegionAsWritten(VAddr start, VAddr end) {
+        const u64 page_end = end >> WRITE_PAGE_BIT;
+        for (u64 page_start = start >> WRITE_PAGE_BIT; page_start <= page_end; ++page_start) {
            auto it = written_pages.find(page_start);
            if (it != written_pages.end()) {
                if (it->second > 1) {
@@ -531,25 +551,27 @@ private:
                    written_pages.erase(it);
                }
            }
-            page_start++;
        }
    }

-    bool IsRegionWritten(const VAddr start, const VAddr end) const {
-        u64 page_start = start >> write_page_bit;
-        const u64 page_end = end >> write_page_bit;
-        while (page_start <= page_end) {
+    bool IsRegionWritten(VAddr start, VAddr end) const {
+        const u64 page_end = end >> WRITE_PAGE_BIT;
+        for (u64 page_start = start >> WRITE_PAGE_BIT; page_start <= page_end; ++page_start) {
            if (written_pages.count(page_start) > 0) {
                return true;
            }
-            page_start++;
        }
        return false;
    }

-    void MarkForAsyncFlush(MapInterval& map) {
+    void QueueDestruction(std::shared_ptr<Buffer> buffer) {
+        buffer->SetEpoch(epoch);
+        pending_destruction.push(std::move(buffer));
+    }
+
+    void MarkForAsyncFlush(MapInterval* map) {
        if (!uncommitted_flushes) {
-            uncommitted_flushes = std::make_shared<std::unordered_set<MapInterval>>();
+            uncommitted_flushes = std::make_shared<std::unordered_set<MapInterval*>>();
        }
        uncommitted_flushes->insert(map);
    }
@@ -558,35 +580,29 @@ private:
    Core::System& system;

    std::unique_ptr<StreamBuffer> stream_buffer;
-    BufferType stream_buffer_handle{};
-
-    bool invalidated = false;
+    BufferType stream_buffer_handle;

    u8* buffer_ptr = nullptr;
    u64 buffer_offset = 0;
    u64 buffer_offset_base = 0;

-    using IntervalSet = boost::icl::interval_set<VAddr>;
-    using IntervalCache = boost::icl::interval_map<VAddr, MapInterval>;
-    using IntervalType = typename IntervalCache::interval_type;
-    IntervalCache mapped_addresses;
+    MapIntervalAllocator mapped_addresses_allocator;
+    boost::intrusive::set<MapInterval, boost::intrusive::compare<MapIntervalCompare>>
+        mapped_addresses;

-    static constexpr u64 write_page_bit = 11;
    std::unordered_map<u64, u32> written_pages;
+    std::unordered_map<u64, std::shared_ptr<Buffer>> blocks;

-    static constexpr u64 block_page_bits = 21;
-    static constexpr u64 block_page_size = 1ULL << block_page_bits;
-    std::unordered_map<u64, OwnerBuffer> blocks;
-
-    std::list<OwnerBuffer> pending_destruction;
+    std::queue<std::shared_ptr<Buffer>> pending_destruction;
    u64 epoch = 0;
    u64 modified_ticks = 0;

    std::vector<u8> staging_buffer;
-    std::list<MapInterval> marked_for_unregister;

-    std::shared_ptr<std::unordered_set<MapInterval>> uncommitted_flushes{};
-    std::list<std::shared_ptr<std::list<MapInterval>>> committed_flushes;
+    std::list<MapInterval*> marked_for_unregister;
+
+    std::shared_ptr<std::unordered_set<MapInterval*>> uncommitted_flushes;
+    std::list<std::shared_ptr<std::list<MapInterval*>>> committed_flushes;

    std::recursive_mutex mutex;
 };
@@ -0,0 +1,33 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <algorithm>
+#include <array>
+#include <cstddef>
+#include <memory>
+
+#include "video_core/buffer_cache/map_interval.h"
+
+namespace VideoCommon {
+
+MapIntervalAllocator::MapIntervalAllocator() {
+    FillFreeList(first_chunk);
+}
+
+MapIntervalAllocator::~MapIntervalAllocator() = default;
+
+void MapIntervalAllocator::AllocateNewChunk() {
+    *new_chunk = std::make_unique<Chunk>();
+    FillFreeList(**new_chunk);
+    new_chunk = &(*new_chunk)->next;
+}
+
+void MapIntervalAllocator::FillFreeList(Chunk& chunk) {
+    const std::size_t old_size = free_list.size();
+    free_list.resize(old_size + chunk.data.size());
+    std::transform(chunk.data.rbegin(), chunk.data.rend(), free_list.begin() + old_size,
+                   [](MapInterval& interval) { return &interval; });
+}
+
+} // namespace VideoCommon
@@ -4,104 +4,89 @@

 #pragma once

+#include <array>
+#include <cstddef>
+#include <memory>
+#include <vector>
+
+#include <boost/intrusive/set_hook.hpp>
+
 #include "common/common_types.h"
 #include "video_core/gpu.h"

 namespace VideoCommon {

-class MapIntervalBase {
-public:
-    MapIntervalBase(const VAddr start, const VAddr end, const GPUVAddr gpu_addr)
-        : start{start}, end{end}, gpu_addr{gpu_addr} {}
+struct MapInterval : public boost::intrusive::set_base_hook<boost::intrusive::optimize_size<true>> {
+    MapInterval() = default;

-    void SetCpuAddress(VAddr new_cpu_addr) {
-        cpu_addr = new_cpu_addr;
+    /*implicit*/ MapInterval(VAddr start_) noexcept : start{start_} {}
+
+    explicit MapInterval(VAddr start_, VAddr end_, GPUVAddr gpu_addr_) noexcept
+        : start{start_}, end{end_}, gpu_addr{gpu_addr_} {}
+
+    bool IsInside(VAddr other_start, VAddr other_end) const noexcept {
+        return start <= other_start && other_end <= end;
    }

-    VAddr GetCpuAddress() const {
-        return cpu_addr;
+    bool Overlaps(VAddr other_start, VAddr other_end) const noexcept {
+        return start < other_end && other_start < end;
    }

-    GPUVAddr GetGpuAddress() const {
-        return gpu_addr;
-    }
-
-    bool IsInside(const VAddr other_start, const VAddr other_end) const {
-        return (start <= other_start && other_end <= end);
-    }
-
-    bool operator==(const MapIntervalBase& rhs) const {
-        return std::tie(start, end) == std::tie(rhs.start, rhs.end);
-    }
-
-    bool operator!=(const MapIntervalBase& rhs) const {
-        return !operator==(rhs);
-    }
-
-    void MarkAsRegistered(const bool registered) {
-        is_registered = registered;
-    }
-
-    bool IsRegistered() const {
-        return is_registered;
-    }
-
-    void SetMemoryMarked(bool is_memory_marked_) {
-        is_memory_marked = is_memory_marked_;
-    }
-
-    bool IsMemoryMarked() const {
-        return is_memory_marked;
-    }
-
-    void SetSyncPending(bool is_sync_pending_) {
-        is_sync_pending = is_sync_pending_;
-    }
-
-    bool IsSyncPending() const {
-        return is_sync_pending;
-    }
-
-    VAddr GetStart() const {
-        return start;
-    }
-
-    VAddr GetEnd() const {
-        return end;
-    }
-
-    void MarkAsModified(const bool is_modified_, const u64 tick) {
+    void MarkAsModified(bool is_modified_, u64 ticks_) noexcept {
        is_modified = is_modified_;
-        ticks = tick;
+        ticks = ticks_;
    }

-    bool IsModified() const {
-        return is_modified;
+    boost::intrusive::set_member_hook<> member_hook_;
+    VAddr start = 0;
+    VAddr end = 0;
+    GPUVAddr gpu_addr = 0;
+    u64 ticks = 0;
+    bool is_written = false;
+    bool is_modified = false;
+    bool is_registered = false;
+    bool is_memory_marked = false;
+    bool is_sync_pending = false;
+};
+
+struct MapIntervalCompare {
+    constexpr bool operator()(const MapInterval& lhs, const MapInterval& rhs) const noexcept {
+        return lhs.start < rhs.start;
+    }
+};
+
+class MapIntervalAllocator {
+public:
+    MapIntervalAllocator();
+    ~MapIntervalAllocator();
+
+    MapInterval* Allocate() {
+        if (free_list.empty()) {
+            AllocateNewChunk();
+        }
+        MapInterval* const interval = free_list.back();
+        free_list.pop_back();
+        return interval;
    }

-    u64 GetModificationTick() const {
-        return ticks;
-    }
-
-    void MarkAsWritten(const bool is_written_) {
-        is_written = is_written_;
-    }
-
-    bool IsWritten() const {
-        return is_written;
+    void Release(MapInterval* interval) {
+        free_list.push_back(interval);
    }

 private:
-    VAddr start;
-    VAddr end;
-    GPUVAddr gpu_addr;
-    VAddr cpu_addr{};
-    bool is_written{};
-    bool is_modified{};
-    bool is_registered{};
-    bool is_memory_marked{};
-    bool is_sync_pending{};
-    u64 ticks{};
+    struct Chunk {
+        std::unique_ptr<Chunk> next;
+        std::array<MapInterval, 0x8000> data;
+    };
+
+    void AllocateNewChunk();
+
+    void FillFreeList(Chunk& chunk);
+
+    std::vector<MapInterval*> free_list;
+    std::unique_ptr<Chunk>* new_chunk = &first_chunk.next;
+
+    Chunk first_chunk;
 };

 } // namespace VideoCommon
@@ -54,9 +54,7 @@ bool DmaPusher::Step() {
        return true;
    });
    const CommandListHeader command_list_header{command_list[dma_pushbuffer_subindex++]};
-    GPUVAddr dma_get = command_list_header.addr;
-    GPUVAddr dma_put = dma_get + command_list_header.size * sizeof(u32);
-    bool non_main = command_list_header.is_non_main;
+    const GPUVAddr dma_get = command_list_header.addr;

    if (dma_pushbuffer_subindex >= command_list.size()) {
        // We've gone through the current list, remove it from the queue
@@ -133,11 +131,6 @@ bool DmaPusher::Step() {
        index++;
    }

-    if (!non_main) {
-        // TODO (degasus): This is dead code, as dma_mget is never read.
-        dma_mget = dma_put;
-    }
-
    return true;
 }

@@ -102,7 +102,6 @@ private:
    DmaState dma_state{};
    bool dma_increment_once{};

-    GPUVAddr dma_mget{};  ///< main pushbuffer last read address
    bool ib_enable{true}; ///< IB mode enabled

    std::array<Tegra::Engines::EngineInterface*, max_subchannels> subchannels{};
@@ -93,6 +93,7 @@ public:
    virtual SamplerDescriptor AccessBoundSampler(ShaderType stage, u64 offset) const = 0;
    virtual SamplerDescriptor AccessBindlessSampler(ShaderType stage, u64 const_buffer,
                                                    u64 offset) const = 0;
+    virtual SamplerDescriptor AccessSampler(u32 handle) const = 0;
    virtual u32 GetBoundBuffer() const = 0;

    virtual VideoCore::GuestDriverProfile& AccessGuestDriverProfile() = 0;
@@ -92,8 +92,11 @@ SamplerDescriptor KeplerCompute::AccessBindlessSampler(ShaderType stage, u64 con
    ASSERT(stage == ShaderType::Compute);
    const auto& tex_info_buffer = launch_description.const_buffer_config[const_buffer];
    const GPUVAddr tex_info_address = tex_info_buffer.Address() + offset;
+    return AccessSampler(memory_manager.Read<u32>(tex_info_address));
+}

-    const Texture::TextureHandle tex_handle{memory_manager.Read<u32>(tex_info_address)};
+SamplerDescriptor KeplerCompute::AccessSampler(u32 handle) const {
+    const Texture::TextureHandle tex_handle{handle};
    const Texture::FullTextureInfo tex_info = GetTextureInfo(tex_handle);
    SamplerDescriptor result = SamplerDescriptor::FromTIC(tex_info.tic);
    result.is_shadow.Assign(tex_info.tsc.depth_compare_enabled.Value());
@@ -219,6 +219,8 @@ public:
    SamplerDescriptor AccessBindlessSampler(ShaderType stage, u64 const_buffer,
                                            u64 offset) const override;

+    SamplerDescriptor AccessSampler(u32 handle) const override;
+
    u32 GetBoundBuffer() const override {
        return regs.tex_cb_index;
    }
@@ -25,9 +25,8 @@ constexpr u32 MacroRegistersStart = 0xE00;
 Maxwell3D::Maxwell3D(Core::System& system, VideoCore::RasterizerInterface& rasterizer,
                     MemoryManager& memory_manager)
    : system{system}, rasterizer{rasterizer}, memory_manager{memory_manager},
-      macro_interpreter{*this}, upload_state{memory_manager, regs.upload} {
+      macro_engine{GetMacroEngine(*this)}, upload_state{memory_manager, regs.upload} {
    dirty.flags.flip();
-
    InitializeRegisterDefaults();
 }

@@ -44,6 +43,12 @@ void Maxwell3D::InitializeRegisterDefaults() {
        viewport.depth_range_near = 0.0f;
        viewport.depth_range_far = 1.0f;
    }
+    for (auto& viewport : regs.viewport_transform) {
+        viewport.swizzle.x.Assign(Regs::ViewportSwizzle::PositiveX);
+        viewport.swizzle.y.Assign(Regs::ViewportSwizzle::PositiveY);
+        viewport.swizzle.z.Assign(Regs::ViewportSwizzle::PositiveZ);
+        viewport.swizzle.w.Assign(Regs::ViewportSwizzle::PositiveW);
+    }

    // Doom and Bomberman seems to use the uninitialized registers and just enable blend
    // so initialize blend registers with sane values
@@ -100,7 +105,11 @@ void Maxwell3D::InitializeRegisterDefaults() {
    regs.rasterize_enable = 1;
    regs.rt_separate_frag_data = 1;
    regs.framebuffer_srgb = 1;
+    regs.line_width_aliased = 1.0f;
+    regs.line_width_smooth = 1.0f;
    regs.front_face = Maxwell3D::Regs::FrontFace::ClockWise;
+    regs.polygon_mode_back = Maxwell3D::Regs::PolygonMode::Fill;
+    regs.polygon_mode_front = Maxwell3D::Regs::PolygonMode::Fill;

    shadow_state = regs;

@@ -110,7 +119,7 @@ void Maxwell3D::InitializeRegisterDefaults() {
    mme_inline[MAXWELL3D_REG_INDEX(index_array.count)] = true;
 }

-void Maxwell3D::CallMacroMethod(u32 method, std::size_t num_parameters, const u32* parameters) {
+void Maxwell3D::CallMacroMethod(u32 method, const std::vector<u32>& parameters) {
    // Reset the current macro.
    executing_macro = 0;

@@ -119,7 +128,7 @@ void Maxwell3D::CallMacroMethod(u32 method, std::size_t num_parameters, const u3
        ((method - MacroRegistersStart) >> 1) % static_cast<u32>(macro_positions.size());

    // Execute the current macro.
-    macro_interpreter.Execute(macro_positions[entry], num_parameters, parameters);
+    macro_engine->Execute(macro_positions[entry], parameters);
    if (mme_draw.current_mode != MMEDrawMode::Undefined) {
        FlushMMEInlineDraw();
    }
@@ -155,7 +164,7 @@ void Maxwell3D::CallMethod(u32 method, u32 method_argument, bool is_last_call) {

        // Call the macro when there are no more parameters in the command buffer
        if (is_last_call) {
-            CallMacroMethod(executing_macro, macro_params.size(), macro_params.data());
+            CallMacroMethod(executing_macro, macro_params);
            macro_params.clear();
        }
        return;
@@ -191,7 +200,7 @@ void Maxwell3D::CallMethod(u32 method, u32 method_argument, bool is_last_call) {
        break;
    }
    case MAXWELL3D_REG_INDEX(macros.data): {
-        ProcessMacroUpload(arg);
+        macro_engine->AddCode(regs.macros.upload_address, arg);
        break;
    }
    case MAXWELL3D_REG_INDEX(macros.bind): {
@@ -300,7 +309,7 @@ void Maxwell3D::CallMultiMethod(u32 method, const u32* base_start, u32 amount,

        // Call the macro when there are no more parameters in the command buffer
        if (amount == methods_pending) {
-            CallMacroMethod(executing_macro, macro_params.size(), macro_params.data());
+            CallMacroMethod(executing_macro, macro_params);
            macro_params.clear();
        }
        return;
@@ -414,9 +423,7 @@ void Maxwell3D::FlushMMEInlineDraw() {
 }

 void Maxwell3D::ProcessMacroUpload(u32 data) {
-    ASSERT_MSG(regs.macros.upload_address < macro_memory.size(),
-               "upload_address exceeded macro_memory size!");
-    macro_memory[regs.macros.upload_address++] = data;
+    macro_engine->AddCode(regs.macros.upload_address++, data);
 }

 void Maxwell3D::ProcessMacroBind(u32 data) {
@@ -451,8 +458,9 @@ void Maxwell3D::StampQueryResult(u64 payload, bool long_query) {

 void Maxwell3D::ProcessQueryGet() {
    // TODO(Subv): Support the other query units.
-    ASSERT_MSG(regs.query.query_get.unit == Regs::QueryUnit::Crop,
-               "Units other than CROP are unimplemented");
+    if (regs.query.query_get.unit != Regs::QueryUnit::Crop) {
+        LOG_DEBUG(HW_GPU, "Units other than CROP are unimplemented");
+    }

    switch (regs.query.query_get.operation) {
    case Regs::QueryOperation::Release:
@@ -528,8 +536,8 @@ void Maxwell3D::ProcessCounterReset() {
        rasterizer.ResetCounter(QueryType::SamplesPassed);
        break;
    default:
-        LOG_WARNING(Render_OpenGL, "Unimplemented counter reset={}",
-                    static_cast<int>(regs.counter_reset));
+        LOG_DEBUG(Render_OpenGL, "Unimplemented counter reset={}",
+                  static_cast<int>(regs.counter_reset));
        break;
    }
 }
@@ -586,8 +594,8 @@ std::optional<u64> Maxwell3D::GetQueryResult() {
                         system.GPU().GetTicks());
        return {};
    default:
-        UNIMPLEMENTED_MSG("Unimplemented query select type {}",
-                          static_cast<u32>(regs.query.query_get.select.Value()));
+        LOG_DEBUG(HW_GPU, "Unimplemented query select type {}",
+                  static_cast<u32>(regs.query.query_get.select.Value()));
        return 1;
    }
 }
@@ -732,8 +740,11 @@ SamplerDescriptor Maxwell3D::AccessBindlessSampler(ShaderType stage, u64 const_b
    const auto& shader = state.shader_stages[static_cast<std::size_t>(stage)];
    const auto& tex_info_buffer = shader.const_buffers[const_buffer];
    const GPUVAddr tex_info_address = tex_info_buffer.address + offset;
+    return AccessSampler(memory_manager.Read<u32>(tex_info_address));
+}

-    const Texture::TextureHandle tex_handle{memory_manager.Read<u32>(tex_info_address)};
+SamplerDescriptor Maxwell3D::AccessSampler(u32 handle) const {
+    const Texture::TextureHandle tex_handle{handle};
    const Texture::FullTextureInfo tex_info = GetTextureInfo(tex_handle);
    SamplerDescriptor result = SamplerDescriptor::FromTIC(tex_info.tic);
    result.is_shadow.Assign(tex_info.tsc.depth_compare_enabled.Value());
@@ -23,7 +23,7 @@
 #include "video_core/engines/engine_upload.h"
 #include "video_core/engines/shader_type.h"
 #include "video_core/gpu.h"
-#include "video_core/macro_interpreter.h"
+#include "video_core/macro/macro.h"
 #include "video_core/textures/texture.h"

 namespace Core {
@@ -576,6 +576,17 @@ public:
            Replay = 3,
        };

+        enum class ViewportSwizzle : u32 {
+            PositiveX = 0,
+            NegativeX = 1,
+            PositiveY = 2,
+            NegativeY = 3,
+            PositiveZ = 4,
+            NegativeZ = 5,
+            PositiveW = 6,
+            NegativeW = 7,
+        };
+
        struct RenderTargetConfig {
            u32 address_high;
            u32 address_low;
@@ -587,6 +598,7 @@ public:
                BitField<4, 3, u32> block_height;
                BitField<8, 3, u32> block_depth;
                BitField<12, 1, InvMemoryLayout> type;
+                BitField<16, 1, u32> is_3d;
            } memory_layout;
            union {
                BitField<0, 16, u32> layers;
@@ -619,7 +631,14 @@ public:
            f32 translate_x;
            f32 translate_y;
            f32 translate_z;
-            INSERT_UNION_PADDING_WORDS(2);
+            union {
+                u32 raw;
+                BitField<0, 3, ViewportSwizzle> x;
+                BitField<4, 3, ViewportSwizzle> y;
+                BitField<8, 3, ViewportSwizzle> z;
+                BitField<12, 3, ViewportSwizzle> w;
+            } swizzle;
+            INSERT_UNION_PADDING_WORDS(1);

            Common::Rectangle<f32> GetRect() const {
                return {
@@ -1385,6 +1404,8 @@ public:
    SamplerDescriptor AccessBindlessSampler(ShaderType stage, u64 const_buffer,
                                            u64 offset) const override;

+    SamplerDescriptor AccessSampler(u32 handle) const override;
+
    u32 GetBoundBuffer() const override {
        return regs.tex_cb_index;
    }
@@ -1393,15 +1414,6 @@ public:

    const VideoCore::GuestDriverProfile& AccessGuestDriverProfile() const override;

-    /// Memory for macro code - it's undetermined how big this is, however 1MB is much larger than
-    /// we've seen used.
-    using MacroMemory = std::array<u32, 0x40000>;
-
-    /// Gets a reference to macro memory.
-    const MacroMemory& GetMacroMemory() const {
-        return macro_memory;
-    }
-
    bool ShouldExecute() const {
        return execute_on;
    }
@@ -1450,16 +1462,13 @@ private:

    std::array<bool, Regs::NUM_REGS> mme_inline{};

-    /// Memory for macro code
-    MacroMemory macro_memory;
-
    /// Macro method that is currently being executed / being fed parameters.
    u32 executing_macro = 0;
    /// Parameters that have been submitted to the macro call so far.
    std::vector<u32> macro_params;

    /// Interpreter for the macro codes uploaded to the GPU.
-    MacroInterpreter macro_interpreter;
+    std::unique_ptr<MacroEngine> macro_engine;

    static constexpr u32 null_cb_data = 0xFFFFFFFF;
    struct {
@@ -1488,7 +1497,7 @@ private:
     * @param num_parameters Number of arguments
     * @param parameters Arguments to the method call
     */
-    void CallMacroMethod(u32 method, std::size_t num_parameters, const u32* parameters);
+    void CallMacroMethod(u32 method, const std::vector<u32>& parameters);

    /// Handles writes to the macro uploading register.
    void ProcessMacroUpload(u32 data);
@@ -168,18 +168,22 @@ enum class Pred : u64 {
 };

 enum class PredCondition : u64 {
-    LessThan = 1,
-    Equal = 2,
-    LessEqual = 3,
-    GreaterThan = 4,
-    NotEqual = 5,
-    GreaterEqual = 6,
-    LessThanWithNan = 9,
-    LessEqualWithNan = 11,
-    GreaterThanWithNan = 12,
-    NotEqualWithNan = 13,
-    GreaterEqualWithNan = 14,
-    // TODO(Subv): Other condition types
+    F = 0,    // Always false
+    LT = 1,   // Ordered less than
+    EQ = 2,   // Ordered equal
+    LE = 3,   // Ordered less than or equal
+    GT = 4,   // Ordered greater than
+    NE = 5,   // Ordered not equal
+    GE = 6,   // Ordered greater than or equal
+    NUM = 7,  // Ordered
+    NAN_ = 8, // Unordered
+    LTU = 9,  // Unordered less than
+    EQU = 10, // Unordered equal
+    LEU = 11, // Unordered less than or equal
+    GTU = 12, // Unordered greater than
+    NEU = 13, // Unordered not equal
+    GEU = 14, // Unordered greater than or equal
+    T = 15,   // Always true
 };

 enum class PredOperation : u64 {
@@ -68,6 +68,7 @@ enum class RenderTargetFormat : u32 {
    BGR5A1_UNORM = 0xE9,
    RG8_UNORM = 0xEA,
    RG8_SNORM = 0xEB,
+    RG8_UINT = 0xED,
    R16_UNORM = 0xEE,
    R16_SNORM = 0xEF,
    R16_SINT = 0xF0,
@@ -0,0 +1,45 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/assert.h"
+#include "common/logging/log.h"
+#include "core/settings.h"
+#include "video_core/macro/macro.h"
+#include "video_core/macro/macro_interpreter.h"
+#include "video_core/macro/macro_jit_x64.h"
+
+namespace Tegra {
+
+void MacroEngine::AddCode(u32 method, u32 data) {
+    uploaded_macro_code[method].push_back(data);
+}
+
+void MacroEngine::Execute(u32 method, const std::vector<u32>& parameters) {
+    auto compiled_macro = macro_cache.find(method);
+    if (compiled_macro != macro_cache.end()) {
+        compiled_macro->second->Execute(parameters, method);
+    } else {
+        // Macro not compiled, check if it's uploaded and if so, compile it
+        auto macro_code = uploaded_macro_code.find(method);
+        if (macro_code == uploaded_macro_code.end()) {
+            UNREACHABLE_MSG("Macro 0x{0:x} was not uploaded", method);
+            return;
+        }
+        macro_cache[method] = Compile(macro_code->second);
+        macro_cache[method]->Execute(parameters, method);
+    }
+}
+
+std::unique_ptr<MacroEngine> GetMacroEngine(Engines::Maxwell3D& maxwell3d) {
+    if (Settings::values.disable_macro_jit) {
+        return std::make_unique<MacroInterpreter>(maxwell3d);
+    }
+#ifdef ARCHITECTURE_x86_64
+    return std::make_unique<MacroJITx64>(maxwell3d);
+#else
+    return std::make_unique<MacroInterpreter>(maxwell3d);
+#endif
+}
+
+} // namespace Tegra
@@ -0,0 +1,128 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <memory>
+#include <unordered_map>
+#include <vector>
+#include "common/bit_field.h"
+#include "common/common_types.h"
+
+namespace Tegra {
+namespace Engines {
+class Maxwell3D;
+}
+namespace Macro {
+constexpr std::size_t NUM_MACRO_REGISTERS = 8;
+enum class Operation : u32 {
+    ALU = 0,
+    AddImmediate = 1,
+    ExtractInsert = 2,
+    ExtractShiftLeftImmediate = 3,
+    ExtractShiftLeftRegister = 4,
+    Read = 5,
+    Unused = 6, // This operation doesn't seem to be a valid encoding.
+    Branch = 7,
+};
+
+enum class ALUOperation : u32 {
+    Add = 0,
+    AddWithCarry = 1,
+    Subtract = 2,
+    SubtractWithBorrow = 3,
+    // Operations 4-7 don't seem to be valid encodings.
+    Xor = 8,
+    Or = 9,
+    And = 10,
+    AndNot = 11,
+    Nand = 12
+};
+
+enum class ResultOperation : u32 {
+    IgnoreAndFetch = 0,
+    Move = 1,
+    MoveAndSetMethod = 2,
+    FetchAndSend = 3,
+    MoveAndSend = 4,
+    FetchAndSetMethod = 5,
+    MoveAndSetMethodFetchAndSend = 6,
+    MoveAndSetMethodSend = 7
+};
+
+enum class BranchCondition : u32 {
+    Zero = 0,
+    NotZero = 1,
+};
+
+union Opcode {
+    u32 raw;
+    BitField<0, 3, Operation> operation;
+    BitField<4, 3, ResultOperation> result_operation;
+    BitField<4, 1, BranchCondition> branch_condition;
+    // If set on a branch, then the branch doesn't have a delay slot.
+    BitField<5, 1, u32> branch_annul;
+    BitField<7, 1, u32> is_exit;
+    BitField<8, 3, u32> dst;
+    BitField<11, 3, u32> src_a;
+    BitField<14, 3, u32> src_b;
+    // The signed immediate overlaps the second source operand and the alu operation.
+    BitField<14, 18, s32> immediate;
+
+    BitField<17, 5, ALUOperation> alu_operation;
+
+    // Bitfield instructions data
+    BitField<17, 5, u32> bf_src_bit;
+    BitField<22, 5, u32> bf_size;
+    BitField<27, 5, u32> bf_dst_bit;
+
+    u32 GetBitfieldMask() const {
+        return (1 << bf_size) - 1;
+    }
+
+    s32 GetBranchTarget() const {
+        return static_cast<s32>(immediate * sizeof(u32));
+    }
+};
+
+union MethodAddress {
+    u32 raw;
+    BitField<0, 12, u32> address;
+    BitField<12, 6, u32> increment;
+};
+
+} // namespace Macro
+
+class CachedMacro {
+public:
+    virtual ~CachedMacro() = default;
+    /**
+     * Executes the macro code with the specified input parameters.
+     * @param code The macro byte code to execute
+     * @param parameters The parameters of the macro
+     */
+    virtual void Execute(const std::vector<u32>& parameters, u32 method) = 0;
+};
+
+class MacroEngine {
+public:
+    virtual ~MacroEngine() = default;
+
+    // Store the uploaded macro code to compile them when they're called.
+    void AddCode(u32 method, u32 data);
+
+    // Compiles the macro if its not in the cache, and executes the compiled macro
+    void Execute(u32 method, const std::vector<u32>& parameters);
+
+protected:
+    virtual std::unique_ptr<CachedMacro> Compile(const std::vector<u32>& code) = 0;
+
+private:
+    std::unordered_map<u32, std::unique_ptr<CachedMacro>> macro_cache;
+    std::unordered_map<u32, std::vector<u32>> uploaded_macro_code;
+};
+
+std::unique_ptr<MacroEngine> GetMacroEngine(Engines::Maxwell3D& maxwell3d);
+
+} // namespace Tegra
@@ -1,4 +1,4 @@
-// Copyright 2018 yuzu Emulator Project
+// Copyright 2020 yuzu Emulator Project
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.

@@ -6,109 +6,46 @@
 #include "common/logging/log.h"
 #include "common/microprofile.h"
 #include "video_core/engines/maxwell_3d.h"
-#include "video_core/macro_interpreter.h"
+#include "video_core/macro/macro_interpreter.h"

 MICROPROFILE_DEFINE(MacroInterp, "GPU", "Execute macro interpreter", MP_RGB(128, 128, 192));

 namespace Tegra {
-namespace {
-enum class Operation : u32 {
-    ALU = 0,
-    AddImmediate = 1,
-    ExtractInsert = 2,
-    ExtractShiftLeftImmediate = 3,
-    ExtractShiftLeftRegister = 4,
-    Read = 5,
-    Unused = 6, // This operation doesn't seem to be a valid encoding.
-    Branch = 7,
-};
-} // Anonymous namespace
-
-enum class MacroInterpreter::ALUOperation : u32 {
-    Add = 0,
-    AddWithCarry = 1,
-    Subtract = 2,
-    SubtractWithBorrow = 3,
-    // Operations 4-7 don't seem to be valid encodings.
-    Xor = 8,
-    Or = 9,
-    And = 10,
-    AndNot = 11,
-    Nand = 12
-};
-
-enum class MacroInterpreter::ResultOperation : u32 {
-    IgnoreAndFetch = 0,
-    Move = 1,
-    MoveAndSetMethod = 2,
-    FetchAndSend = 3,
-    MoveAndSend = 4,
-    FetchAndSetMethod = 5,
-    MoveAndSetMethodFetchAndSend = 6,
-    MoveAndSetMethodSend = 7
-};
-
-enum class MacroInterpreter::BranchCondition : u32 {
-    Zero = 0,
-    NotZero = 1,
-};
-
-union MacroInterpreter::Opcode {
-    u32 raw;
-    BitField<0, 3, Operation> operation;
-    BitField<4, 3, ResultOperation> result_operation;
-    BitField<4, 1, BranchCondition> branch_condition;
-    // If set on a branch, then the branch doesn't have a delay slot.
-    BitField<5, 1, u32> branch_annul;
-    BitField<7, 1, u32> is_exit;
-    BitField<8, 3, u32> dst;
-    BitField<11, 3, u32> src_a;
-    BitField<14, 3, u32> src_b;
-    // The signed immediate overlaps the second source operand and the alu operation.
-    BitField<14, 18, s32> immediate;
-
-    BitField<17, 5, ALUOperation> alu_operation;
-
-    // Bitfield instructions data
-    BitField<17, 5, u32> bf_src_bit;
-    BitField<22, 5, u32> bf_size;
-    BitField<27, 5, u32> bf_dst_bit;
-
-    u32 GetBitfieldMask() const {
-        return (1 << bf_size) - 1;
-    }
-
-    s32 GetBranchTarget() const {
-        return static_cast<s32>(immediate * sizeof(u32));
-    }
-};
-
 MacroInterpreter::MacroInterpreter(Engines::Maxwell3D& maxwell3d) : maxwell3d(maxwell3d) {}

-void MacroInterpreter::Execute(u32 offset, std::size_t num_parameters, const u32* parameters) {
+std::unique_ptr<CachedMacro> MacroInterpreter::Compile(const std::vector<u32>& code) {
+    return std::make_unique<MacroInterpreterImpl>(maxwell3d, code);
+}
+
+MacroInterpreterImpl::MacroInterpreterImpl(Engines::Maxwell3D& maxwell3d,
+                                           const std::vector<u32>& code)
+    : maxwell3d(maxwell3d), code(code) {}
+
+void MacroInterpreterImpl::Execute(const std::vector<u32>& parameters, u32 method) {
    MICROPROFILE_SCOPE(MacroInterp);
    Reset();

    registers[1] = parameters[0];
+    num_parameters = parameters.size();

    if (num_parameters > parameters_capacity) {
        parameters_capacity = num_parameters;
        this->parameters = std::make_unique<u32[]>(num_parameters);
    }
-    std::memcpy(this->parameters.get(), parameters, num_parameters * sizeof(u32));
+    std::memcpy(this->parameters.get(), parameters.data(), num_parameters * sizeof(u32));
    this->num_parameters = num_parameters;

    // Execute the code until we hit an exit condition.
    bool keep_executing = true;
    while (keep_executing) {
-        keep_executing = Step(offset, false);
+        keep_executing = Step(false);
    }

    // Assert the the macro used all the input parameters
    ASSERT(next_parameter_index == num_parameters);
 }

-void MacroInterpreter::Reset() {
+void MacroInterpreterImpl::Reset() {
    registers = {};
    pc = 0;
    delayed_pc = {};
@@ -120,10 +57,10 @@ void MacroInterpreter::Reset() {
    carry_flag = false;
 }

-bool MacroInterpreter::Step(u32 offset, bool is_delay_slot) {
+bool MacroInterpreterImpl::Step(bool is_delay_slot) {
    u32 base_address = pc;

-    Opcode opcode = GetOpcode(offset);
+    Macro::Opcode opcode = GetOpcode();
    pc += 4;

    // Update the program counter if we were delayed
@@ -134,18 +71,18 @@ bool MacroInterpreter::Step(u32 offset, bool is_delay_slot) {
    }

    switch (opcode.operation) {
-    case Operation::ALU: {
+    case Macro::Operation::ALU: {
        u32 result = GetALUResult(opcode.alu_operation, GetRegister(opcode.src_a),
                                  GetRegister(opcode.src_b));
        ProcessResult(opcode.result_operation, opcode.dst, result);
        break;
    }
-    case Operation::AddImmediate: {
+    case Macro::Operation::AddImmediate: {
        ProcessResult(opcode.result_operation, opcode.dst,
                      GetRegister(opcode.src_a) + opcode.immediate);
        break;
    }
-    case Operation::ExtractInsert: {
+    case Macro::Operation::ExtractInsert: {
        u32 dst = GetRegister(opcode.src_a);
        u32 src = GetRegister(opcode.src_b);

@@ -155,7 +92,7 @@ bool MacroInterpreter::Step(u32 offset, bool is_delay_slot) {
        ProcessResult(opcode.result_operation, opcode.dst, dst);
        break;
    }
-    case Operation::ExtractShiftLeftImmediate: {
+    case Macro::Operation::ExtractShiftLeftImmediate: {
        u32 dst = GetRegister(opcode.src_a);
        u32 src = GetRegister(opcode.src_b);

@@ -164,7 +101,7 @@ bool MacroInterpreter::Step(u32 offset, bool is_delay_slot) {
        ProcessResult(opcode.result_operation, opcode.dst, result);
        break;
    }
-    case Operation::ExtractShiftLeftRegister: {
+    case Macro::Operation::ExtractShiftLeftRegister: {
        u32 dst = GetRegister(opcode.src_a);
        u32 src = GetRegister(opcode.src_b);

@@ -173,12 +110,12 @@ bool MacroInterpreter::Step(u32 offset, bool is_delay_slot) {
        ProcessResult(opcode.result_operation, opcode.dst, result);
        break;
    }
-    case Operation::Read: {
+    case Macro::Operation::Read: {
        u32 result = Read(GetRegister(opcode.src_a) + opcode.immediate);
        ProcessResult(opcode.result_operation, opcode.dst, result);
        break;
    }
-    case Operation::Branch: {
+    case Macro::Operation::Branch: {
        ASSERT_MSG(!is_delay_slot, "Executing a branch in a delay slot is not valid");
        u32 value = GetRegister(opcode.src_a);
        bool taken = EvaluateBranchCondition(opcode.branch_condition, value);
@@ -191,7 +128,7 @@ bool MacroInterpreter::Step(u32 offset, bool is_delay_slot) {

            delayed_pc = base_address + opcode.GetBranchTarget();
            // Execute one more instruction due to the delay slot.
-            return Step(offset, true);
+            return Step(true);
        }
        break;
    }
@@ -204,51 +141,44 @@ bool MacroInterpreter::Step(u32 offset, bool is_delay_slot) {
    // cause an exit if it's executed inside a delay slot.
    if (opcode.is_exit && !is_delay_slot) {
        // Exit has a delay slot, execute the next instruction
-        Step(offset, true);
+        Step(true);
        return false;
    }

    return true;
 }

-MacroInterpreter::Opcode MacroInterpreter::GetOpcode(u32 offset) const {
-    const auto& macro_memory{maxwell3d.GetMacroMemory()};
-    ASSERT((pc % sizeof(u32)) == 0);
-    ASSERT((pc + offset) < macro_memory.size() * sizeof(u32));
-    return {macro_memory[offset + pc / sizeof(u32)]};
-}
-
-u32 MacroInterpreter::GetALUResult(ALUOperation operation, u32 src_a, u32 src_b) {
+u32 MacroInterpreterImpl::GetALUResult(Macro::ALUOperation operation, u32 src_a, u32 src_b) {
    switch (operation) {
-    case ALUOperation::Add: {
+    case Macro::ALUOperation::Add: {
        const u64 result{static_cast<u64>(src_a) + src_b};
        carry_flag = result > 0xffffffff;
        return static_cast<u32>(result);
    }
-    case ALUOperation::AddWithCarry: {
+    case Macro::ALUOperation::AddWithCarry: {
        const u64 result{static_cast<u64>(src_a) + src_b + (carry_flag ? 1ULL : 0ULL)};
        carry_flag = result > 0xffffffff;
        return static_cast<u32>(result);
    }
-    case ALUOperation::Subtract: {
+    case Macro::ALUOperation::Subtract: {
        const u64 result{static_cast<u64>(src_a) - src_b};
        carry_flag = result < 0x100000000;
        return static_cast<u32>(result);
    }
-    case ALUOperation::SubtractWithBorrow: {
+    case Macro::ALUOperation::SubtractWithBorrow: {
        const u64 result{static_cast<u64>(src_a) - src_b - (carry_flag ? 0ULL : 1ULL)};
        carry_flag = result < 0x100000000;
        return static_cast<u32>(result);
    }
-    case ALUOperation::Xor:
+    case Macro::ALUOperation::Xor:
        return src_a ^ src_b;
-    case ALUOperation::Or:
+    case Macro::ALUOperation::Or:
        return src_a | src_b;
-    case ALUOperation::And:
+    case Macro::ALUOperation::And:
        return src_a & src_b;
-    case ALUOperation::AndNot:
+    case Macro::ALUOperation::AndNot:
        return src_a & ~src_b;
-    case ALUOperation::Nand:
+    case Macro::ALUOperation::Nand:
        return ~(src_a & src_b);

    default:
@@ -257,43 +187,43 @@ u32 MacroInterpreter::GetALUResult(ALUOperation operation, u32 src_a, u32 src_b)
    }
 }

-void MacroInterpreter::ProcessResult(ResultOperation operation, u32 reg, u32 result) {
+void MacroInterpreterImpl::ProcessResult(Macro::ResultOperation operation, u32 reg, u32 result) {
    switch (operation) {
-    case ResultOperation::IgnoreAndFetch:
+    case Macro::ResultOperation::IgnoreAndFetch:
        // Fetch parameter and ignore result.
        SetRegister(reg, FetchParameter());
        break;
-    case ResultOperation::Move:
+    case Macro::ResultOperation::Move:
        // Move result.
        SetRegister(reg, result);
        break;
-    case ResultOperation::MoveAndSetMethod:
+    case Macro::ResultOperation::MoveAndSetMethod:
        // Move result and use as Method Address.
        SetRegister(reg, result);
        SetMethodAddress(result);
        break;
-    case ResultOperation::FetchAndSend:
+    case Macro::ResultOperation::FetchAndSend:
        // Fetch parameter and send result.
        SetRegister(reg, FetchParameter());
        Send(result);
        break;
-    case ResultOperation::MoveAndSend:
+    case Macro::ResultOperation::MoveAndSend:
        // Move and send result.
        SetRegister(reg, result);
        Send(result);
        break;
-    case ResultOperation::FetchAndSetMethod:
+    case Macro::ResultOperation::FetchAndSetMethod:
        // Fetch parameter and use result as Method Address.
        SetRegister(reg, FetchParameter());
        SetMethodAddress(result);
        break;
-    case ResultOperation::MoveAndSetMethodFetchAndSend:
+    case Macro::ResultOperation::MoveAndSetMethodFetchAndSend:
        // Move result and use as Method Address, then fetch and send parameter.
        SetRegister(reg, result);
        SetMethodAddress(result);
        Send(FetchParameter());
        break;
-    case ResultOperation::MoveAndSetMethodSend:
+    case Macro::ResultOperation::MoveAndSetMethodSend:
        // Move result and use as Method Address, then send bits 12:17 of result.
        SetRegister(reg, result);
        SetMethodAddress(result);
@@ -304,16 +234,28 @@ void MacroInterpreter::ProcessResult(ResultOperation operation, u32 reg, u32 res
    }
 }

-u32 MacroInterpreter::FetchParameter() {
-    ASSERT(next_parameter_index < num_parameters);
-    return parameters[next_parameter_index++];
+bool MacroInterpreterImpl::EvaluateBranchCondition(Macro::BranchCondition cond, u32 value) const {
+    switch (cond) {
+    case Macro::BranchCondition::Zero:
+        return value == 0;
+    case Macro::BranchCondition::NotZero:
+        return value != 0;
+    }
+    UNREACHABLE();
+    return true;
 }

-u32 MacroInterpreter::GetRegister(u32 register_id) const {
+Macro::Opcode MacroInterpreterImpl::GetOpcode() const {
+    ASSERT((pc % sizeof(u32)) == 0);
+    ASSERT(pc < code.size() * sizeof(u32));
+    return {code[pc / sizeof(u32)]};
+}
+
+u32 MacroInterpreterImpl::GetRegister(u32 register_id) const {
    return registers.at(register_id);
 }

-void MacroInterpreter::SetRegister(u32 register_id, u32 value) {
+void MacroInterpreterImpl::SetRegister(u32 register_id, u32 value) {
    // Register 0 is hardwired as the zero register.
    // Ensure no writes to it actually occur.
    if (register_id == 0) {
@@ -323,30 +265,24 @@ void MacroInterpreter::SetRegister(u32 register_id, u32 value) {
    registers.at(register_id) = value;
 }

-void MacroInterpreter::SetMethodAddress(u32 address) {
+void MacroInterpreterImpl::SetMethodAddress(u32 address) {
    method_address.raw = address;
 }

-void MacroInterpreter::Send(u32 value) {
+void MacroInterpreterImpl::Send(u32 value) {
    maxwell3d.CallMethodFromMME(method_address.address, value);
    // Increment the method address by the method increment.
    method_address.address.Assign(method_address.address.Value() +
                                  method_address.increment.Value());
 }

-u32 MacroInterpreter::Read(u32 method) const {
+u32 MacroInterpreterImpl::Read(u32 method) const {
    return maxwell3d.GetRegisterValue(method);
 }

-bool MacroInterpreter::EvaluateBranchCondition(BranchCondition cond, u32 value) const {
-    switch (cond) {
-    case BranchCondition::Zero:
-        return value == 0;
-    case BranchCondition::NotZero:
-        return value != 0;
-    }
-    UNREACHABLE();
-    return true;
+u32 MacroInterpreterImpl::FetchParameter() {
+    ASSERT(next_parameter_index < num_parameters);
+    return parameters[next_parameter_index++];
 }

 } // namespace Tegra
@@ -1,44 +1,37 @@
-// Copyright 2018 yuzu Emulator Project
+// Copyright 2020 yuzu Emulator Project
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.

 #pragma once
-
 #include <array>
 #include <optional>
-
+#include <vector>
 #include "common/bit_field.h"
 #include "common/common_types.h"
+#include "video_core/macro/macro.h"

 namespace Tegra {
 namespace Engines {
 class Maxwell3D;
 }

-class MacroInterpreter final {
+class MacroInterpreter final : public MacroEngine {
 public:
    explicit MacroInterpreter(Engines::Maxwell3D& maxwell3d);

-    /**
-     * Executes the macro code with the specified input parameters.
-     * @param offset Offset to start execution at.
-     * @param parameters The parameters of the macro.
-     */
-    void Execute(u32 offset, std::size_t num_parameters, const u32* parameters);
+protected:
+    std::unique_ptr<CachedMacro> Compile(const std::vector<u32>& code) override;

 private:
-    enum class ALUOperation : u32;
-    enum class BranchCondition : u32;
-    enum class ResultOperation : u32;
+    Engines::Maxwell3D& maxwell3d;
+};

-    union Opcode;
-
-    union MethodAddress {
-        u32 raw;
-        BitField<0, 12, u32> address;
-        BitField<12, 6, u32> increment;
-    };
+class MacroInterpreterImpl : public CachedMacro {
+public:
+    MacroInterpreterImpl(Engines::Maxwell3D& maxwell3d, const std::vector<u32>& code);
+    void Execute(const std::vector<u32>& parameters, u32 method) override;

+private:
    /// Resets the execution engine state, zeroing registers, etc.
    void Reset();

@@ -49,20 +42,20 @@ private:
     * @param is_delay_slot Whether the current step is being executed due to a delay slot in a
     * previous instruction.
     */
-    bool Step(u32 offset, bool is_delay_slot);
+    bool Step(bool is_delay_slot);

    /// Calculates the result of an ALU operation. src_a OP src_b;
-    u32 GetALUResult(ALUOperation operation, u32 src_a, u32 src_b);
+    u32 GetALUResult(Macro::ALUOperation operation, u32 src_a, u32 src_b);

    /// Performs the result operation on the input result and stores it in the specified register
    /// (if necessary).
-    void ProcessResult(ResultOperation operation, u32 reg, u32 result);
+    void ProcessResult(Macro::ResultOperation operation, u32 reg, u32 result);

    /// Evaluates the branch condition and returns whether the branch should be taken or not.
-    bool EvaluateBranchCondition(BranchCondition cond, u32 value) const;
+    bool EvaluateBranchCondition(Macro::BranchCondition cond, u32 value) const;

    /// Reads an opcode at the current program counter location.
-    Opcode GetOpcode(u32 offset) const;
+    Macro::Opcode GetOpcode() const;

    /// Returns the specified register's value. Register 0 is hardcoded to always return 0.
    u32 GetRegister(u32 register_id) const;
@@ -89,13 +82,11 @@ private:
    /// Program counter to execute at after the delay slot is executed.
    std::optional<u32> delayed_pc;

-    static constexpr std::size_t NumMacroRegisters = 8;
-
    /// General purpose macro registers.
-    std::array<u32, NumMacroRegisters> registers = {};
+    std::array<u32, Macro::NUM_MACRO_REGISTERS> registers = {};

    /// Method address to use for the next Send instruction.
-    MethodAddress method_address = {};
+    Macro::MethodAddress method_address = {};

    /// Input parameters of the current macro.
    std::unique_ptr<u32[]> parameters;
@@ -105,5 +96,7 @@ private:
    u32 next_parameter_index = 0;

    bool carry_flag = false;
+    const std::vector<u32>& code;
 };
+
 } // namespace Tegra
@@ -0,0 +1,610 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/assert.h"
+#include "common/logging/log.h"
+#include "common/microprofile.h"
+#include "common/x64/xbyak_util.h"
+#include "video_core/engines/maxwell_3d.h"
+#include "video_core/macro/macro_interpreter.h"
+#include "video_core/macro/macro_jit_x64.h"
+
+MICROPROFILE_DEFINE(MacroJitCompile, "GPU", "Compile macro JIT", MP_RGB(173, 255, 47));
+MICROPROFILE_DEFINE(MacroJitExecute, "GPU", "Execute macro JIT", MP_RGB(255, 255, 0));
+
+namespace Tegra {
+static const Xbyak::Reg64 STATE = Xbyak::util::rbx;
+static const Xbyak::Reg32 RESULT = Xbyak::util::ebp;
+static const Xbyak::Reg64 PARAMETERS = Xbyak::util::r12;
+static const Xbyak::Reg32 METHOD_ADDRESS = Xbyak::util::r14d;
+static const Xbyak::Reg64 BRANCH_HOLDER = Xbyak::util::r15;
+
+static const std::bitset<32> PERSISTENT_REGISTERS = Common::X64::BuildRegSet({
+    STATE,
+    RESULT,
+    PARAMETERS,
+    METHOD_ADDRESS,
+    BRANCH_HOLDER,
+});
+
+MacroJITx64::MacroJITx64(Engines::Maxwell3D& maxwell3d) : maxwell3d(maxwell3d) {}
+
+std::unique_ptr<CachedMacro> MacroJITx64::Compile(const std::vector<u32>& code) {
+    return std::make_unique<MacroJITx64Impl>(maxwell3d, code);
+}
+
+MacroJITx64Impl::MacroJITx64Impl(Engines::Maxwell3D& maxwell3d, const std::vector<u32>& code)
+    : Xbyak::CodeGenerator(MAX_CODE_SIZE), code(code), maxwell3d(maxwell3d) {
+    Compile();
+}
+
+MacroJITx64Impl::~MacroJITx64Impl() = default;
+
+void MacroJITx64Impl::Execute(const std::vector<u32>& parameters, u32 method) {
+    MICROPROFILE_SCOPE(MacroJitExecute);
+    ASSERT_OR_EXECUTE(program != nullptr, { return; });
+    JITState state{};
+    state.maxwell3d = &maxwell3d;
+    state.registers = {};
+    program(&state, parameters.data());
+}
+
+void MacroJITx64Impl::Compile_ALU(Macro::Opcode opcode) {
+    const bool is_a_zero = opcode.src_a == 0;
+    const bool is_b_zero = opcode.src_b == 0;
+    const bool valid_operation = !is_a_zero && !is_b_zero;
+    const bool is_move_operation = !is_a_zero && is_b_zero;
+    const bool has_zero_register = is_a_zero || is_b_zero;
+
+    Xbyak::Reg32 src_a;
+    Xbyak::Reg32 src_b;
+
+    if (!optimizer.zero_reg_skip) {
+        src_a = Compile_GetRegister(opcode.src_a, RESULT);
+        src_b = Compile_GetRegister(opcode.src_b, eax);
+    } else {
+        if (!is_a_zero) {
+            src_a = Compile_GetRegister(opcode.src_a, RESULT);
+        }
+        if (!is_b_zero) {
+            src_b = Compile_GetRegister(opcode.src_b, eax);
+        }
+    }
+    Xbyak::Label skip_carry{};
+
+    bool has_emitted = false;
+
+    switch (opcode.alu_operation) {
+    case Macro::ALUOperation::Add:
+        if (optimizer.zero_reg_skip) {
+            if (valid_operation) {
+                add(src_a, src_b);
+            }
+        } else {
+            add(src_a, src_b);
+        }
+
+        if (!optimizer.can_skip_carry) {
+            setc(byte[STATE + offsetof(JITState, carry_flag)]);
+        }
+        break;
+    case Macro::ALUOperation::AddWithCarry:
+        bt(dword[STATE + offsetof(JITState, carry_flag)], 0);
+        adc(src_a, src_b);
+        setc(byte[STATE + offsetof(JITState, carry_flag)]);
+        break;
+    case Macro::ALUOperation::Subtract:
+        if (optimizer.zero_reg_skip) {
+            if (valid_operation) {
+                sub(src_a, src_b);
+                has_emitted = true;
+            }
+        } else {
+            sub(src_a, src_b);
+            has_emitted = true;
+        }
+        if (!optimizer.can_skip_carry && has_emitted) {
+            setc(byte[STATE + offsetof(JITState, carry_flag)]);
+        }
+        break;
+    case Macro::ALUOperation::SubtractWithBorrow:
+        bt(dword[STATE + offsetof(JITState, carry_flag)], 0);
+        sbb(src_a, src_b);
+        setc(byte[STATE + offsetof(JITState, carry_flag)]);
+        break;
+    case Macro::ALUOperation::Xor:
+        if (optimizer.zero_reg_skip) {
+            if (valid_operation) {
+                xor_(src_a, src_b);
+            }
+        } else {
+            xor_(src_a, src_b);
+        }
+        break;
+    case Macro::ALUOperation::Or:
+        if (optimizer.zero_reg_skip) {
+            if (valid_operation) {
+                or_(src_a, src_b);
+            }
+        } else {
+            or_(src_a, src_b);
+        }
+        break;
+    case Macro::ALUOperation::And:
+        if (optimizer.zero_reg_skip) {
+            if (!has_zero_register) {
+                and_(src_a, src_b);
+            }
+        } else {
+            and_(src_a, src_b);
+        }
+        break;
+    case Macro::ALUOperation::AndNot:
+        if (optimizer.zero_reg_skip) {
+            if (!is_a_zero) {
+                not_(src_b);
+                and_(src_a, src_b);
+            }
+        } else {
+            not_(src_b);
+            and_(src_a, src_b);
+        }
+        break;
+    case Macro::ALUOperation::Nand:
+        if (optimizer.zero_reg_skip) {
+            if (!is_a_zero) {
+                and_(src_a, src_b);
+                not_(src_a);
+            }
+        } else {
+            and_(src_a, src_b);
+            not_(src_a);
+        }
+        break;
+    default:
+        UNIMPLEMENTED_MSG("Unimplemented ALU operation {}",
+                          static_cast<std::size_t>(opcode.alu_operation.Value()));
+        break;
+    }
+    Compile_ProcessResult(opcode.result_operation, opcode.dst);
+}
+
+void MacroJITx64Impl::Compile_AddImmediate(Macro::Opcode opcode) {
+    if (optimizer.skip_dummy_addimmediate) {
+        // Games tend to use this as an exit instruction placeholder. It's to encode an instruction
+        // without doing anything. In our case we can just not emit anything.
+        if (opcode.result_operation == Macro::ResultOperation::Move && opcode.dst == 0) {
+            return;
+        }
+    }
+    // Check for redundant moves
+    if (optimizer.optimize_for_method_move &&
+        opcode.result_operation == Macro::ResultOperation::MoveAndSetMethod) {
+        if (next_opcode.has_value()) {
+            const auto next = *next_opcode;
+            if (next.result_operation == Macro::ResultOperation::MoveAndSetMethod) {
+                return;
+            }
+        }
+    }
+    if (optimizer.zero_reg_skip && opcode.src_a == 0) {
+        if (opcode.immediate == 0) {
+            xor_(RESULT, RESULT);
+        } else {
+            mov(RESULT, opcode.immediate);
+        }
+    } else {
+        auto result = Compile_GetRegister(opcode.src_a, RESULT);
+        if (opcode.immediate > 2) {
+            add(result, opcode.immediate);
+        } else if (opcode.immediate == 1) {
+            inc(result);
+        } else if (opcode.immediate < 0) {
+            sub(result, opcode.immediate * -1);
+        }
+    }
+    Compile_ProcessResult(opcode.result_operation, opcode.dst);
+}
+
+void MacroJITx64Impl::Compile_ExtractInsert(Macro::Opcode opcode) {
+    auto dst = Compile_GetRegister(opcode.src_a, RESULT);
+    auto src = Compile_GetRegister(opcode.src_b, eax);
+
+    if (opcode.bf_src_bit != 0 && opcode.bf_src_bit != 31) {
+        shr(src, opcode.bf_src_bit);
+    } else if (opcode.bf_src_bit == 31) {
+        xor_(src, src);
+    }
+    // Don't bother masking the whole register since we're using a 32 bit register
+    if (opcode.bf_size != 31 && opcode.bf_size != 0) {
+        and_(src, opcode.GetBitfieldMask());
+    } else if (opcode.bf_size == 0) {
+        xor_(src, src);
+    }
+    if (opcode.bf_dst_bit != 31 && opcode.bf_dst_bit != 0) {
+        shl(src, opcode.bf_dst_bit);
+    } else if (opcode.bf_dst_bit == 31) {
+        xor_(src, src);
+    }
+
+    const u32 mask = ~(opcode.GetBitfieldMask() << opcode.bf_dst_bit);
+    if (mask != 0xffffffff) {
+        and_(dst, mask);
+    }
+    or_(dst, src);
+    Compile_ProcessResult(opcode.result_operation, opcode.dst);
+}
+
+void MacroJITx64Impl::Compile_ExtractShiftLeftImmediate(Macro::Opcode opcode) {
+    auto dst = Compile_GetRegister(opcode.src_a, eax);
+    auto src = Compile_GetRegister(opcode.src_b, RESULT);
+
+    shr(src, al);
+    if (opcode.bf_size != 0 && opcode.bf_size != 31) {
+        and_(src, opcode.GetBitfieldMask());
+    } else if (opcode.bf_size == 0) {
+        xor_(src, src);
+    }
+
+    if (opcode.bf_dst_bit != 0 && opcode.bf_dst_bit != 31) {
+        shl(src, opcode.bf_dst_bit);
+    } else if (opcode.bf_dst_bit == 31) {
+        xor_(src, src);
+    }
+    Compile_ProcessResult(opcode.result_operation, opcode.dst);
+}
+
+void MacroJITx64Impl::Compile_ExtractShiftLeftRegister(Macro::Opcode opcode) {
+    auto dst = Compile_GetRegister(opcode.src_a, eax);
+    auto src = Compile_GetRegister(opcode.src_b, RESULT);
+
+    if (opcode.bf_src_bit != 0) {
+        shr(src, opcode.bf_src_bit);
+    }
+
+    if (opcode.bf_size != 31) {
+        and_(src, opcode.GetBitfieldMask());
+    }
+    shl(src, al);
+    Compile_ProcessResult(opcode.result_operation, opcode.dst);
+}
+
+static u32 Read(Engines::Maxwell3D* maxwell3d, u32 method) {
+    return maxwell3d->GetRegisterValue(method);
+}
+
+static void Send(Engines::Maxwell3D* maxwell3d, Macro::MethodAddress method_address, u32 value) {
+    maxwell3d->CallMethodFromMME(method_address.address, value);
+}
+
+void MacroJITx64Impl::Compile_Read(Macro::Opcode opcode) {
+    if (optimizer.zero_reg_skip && opcode.src_a == 0) {
+        if (opcode.immediate == 0) {
+            xor_(RESULT, RESULT);
+        } else {
+            mov(RESULT, opcode.immediate);
+        }
+    } else {
+        auto result = Compile_GetRegister(opcode.src_a, RESULT);
+        if (opcode.immediate > 2) {
+            add(result, opcode.immediate);
+        } else if (opcode.immediate == 1) {
+            inc(result);
+        } else if (opcode.immediate < 0) {
+            sub(result, opcode.immediate * -1);
+        }
+    }
+    Common::X64::ABI_PushRegistersAndAdjustStack(*this, PersistentCallerSavedRegs(), 0);
+    mov(Common::X64::ABI_PARAM1, qword[STATE]);
+    mov(Common::X64::ABI_PARAM2, RESULT);
+    Common::X64::CallFarFunction(*this, &Read);
+    Common::X64::ABI_PopRegistersAndAdjustStack(*this, PersistentCallerSavedRegs(), 0);
+    mov(RESULT, Common::X64::ABI_RETURN.cvt32());
+    Compile_ProcessResult(opcode.result_operation, opcode.dst);
+}
+
+void Tegra::MacroJITx64Impl::Compile_Send(Xbyak::Reg32 value) {
+    Common::X64::ABI_PushRegistersAndAdjustStack(*this, PersistentCallerSavedRegs(), 0);
+    mov(Common::X64::ABI_PARAM1, qword[STATE]);
+    mov(Common::X64::ABI_PARAM2, METHOD_ADDRESS);
+    mov(Common::X64::ABI_PARAM3, value);
+    Common::X64::CallFarFunction(*this, &Send);
+    Common::X64::ABI_PopRegistersAndAdjustStack(*this, PersistentCallerSavedRegs(), 0);
+
+    Xbyak::Label dont_process{};
+    // Get increment
+    test(METHOD_ADDRESS, 0x3f000);
+    // If zero, method address doesn't update
+    je(dont_process);
+
+    mov(ecx, METHOD_ADDRESS);
+    and_(METHOD_ADDRESS, 0xfff);
+    shr(ecx, 12);
+    and_(ecx, 0x3f);
+    lea(eax, ptr[rcx + METHOD_ADDRESS.cvt64()]);
+    sal(ecx, 12);
+    or_(eax, ecx);
+
+    mov(METHOD_ADDRESS, eax);
+
+    L(dont_process);
+}
+
+void Tegra::MacroJITx64Impl::Compile_Branch(Macro::Opcode opcode) {
+    ASSERT_MSG(!is_delay_slot, "Executing a branch in a delay slot is not valid");
+    const s32 jump_address =
+        static_cast<s32>(pc) + static_cast<s32>(opcode.GetBranchTarget() / sizeof(s32));
+
+    Xbyak::Label end;
+    auto value = Compile_GetRegister(opcode.src_a, eax);
+    test(value, value);
+    if (optimizer.has_delayed_pc) {
+        switch (opcode.branch_condition) {
+        case Macro::BranchCondition::Zero:
+            jne(end, T_NEAR);
+            break;
+        case Macro::BranchCondition::NotZero:
+            je(end, T_NEAR);
+            break;
+        }
+
+        if (opcode.branch_annul) {
+            xor_(BRANCH_HOLDER, BRANCH_HOLDER);
+            jmp(labels[jump_address], T_NEAR);
+        } else {
+            Xbyak::Label handle_post_exit{};
+            Xbyak::Label skip{};
+            jmp(skip, T_NEAR);
+            if (opcode.is_exit) {
+                L(handle_post_exit);
+                // Execute 1 instruction
+                mov(BRANCH_HOLDER, end_of_code);
+                // Jump to next instruction to skip delay slot check
+                jmp(labels[jump_address], T_NEAR);
+            } else {
+                L(handle_post_exit);
+                xor_(BRANCH_HOLDER, BRANCH_HOLDER);
+                jmp(labels[jump_address], T_NEAR);
+            }
+            L(skip);
+            mov(BRANCH_HOLDER, handle_post_exit);
+            jmp(delay_skip[pc], T_NEAR);
+        }
+    } else {
+        switch (opcode.branch_condition) {
+        case Macro::BranchCondition::Zero:
+            je(labels[jump_address], T_NEAR);
+            break;
+        case Macro::BranchCondition::NotZero:
+            jne(labels[jump_address], T_NEAR);
+            break;
+        }
+    }
+
+    L(end);
+}
+
+void Tegra::MacroJITx64Impl::Optimizer_ScanFlags() {
+    optimizer.can_skip_carry = true;
+    optimizer.has_delayed_pc = false;
+    for (auto raw_op : code) {
+        Macro::Opcode op{};
+        op.raw = raw_op;
+
+        if (op.operation == Macro::Operation::ALU) {
+            // Scan for any ALU operations which actually use the carry flag, if they don't exist in
+            // our current code we can skip emitting the carry flag handling operations
+            if (op.alu_operation == Macro::ALUOperation::AddWithCarry ||
+                op.alu_operation == Macro::ALUOperation::SubtractWithBorrow) {
+                optimizer.can_skip_carry = false;
+            }
+        }
+
+        if (op.operation == Macro::Operation::Branch) {
+            if (!op.branch_annul) {
+                optimizer.has_delayed_pc = true;
+            }
+        }
+    }
+}
+
+void MacroJITx64Impl::Compile() {
+    MICROPROFILE_SCOPE(MacroJitCompile);
+    bool keep_executing = true;
+    labels.fill(Xbyak::Label());
+
+    Common::X64::ABI_PushRegistersAndAdjustStack(*this, Common::X64::ABI_ALL_CALLEE_SAVED, 8);
+    // JIT state
+    mov(STATE, Common::X64::ABI_PARAM1);
+    mov(PARAMETERS, Common::X64::ABI_PARAM2);
+    xor_(RESULT, RESULT);
+    xor_(METHOD_ADDRESS, METHOD_ADDRESS);
+    xor_(BRANCH_HOLDER, BRANCH_HOLDER);
+
+    mov(dword[STATE + offsetof(JITState, registers) + 4], Compile_FetchParameter());
+
+    // Track get register for zero registers and mark it as no-op
+    optimizer.zero_reg_skip = true;
+
+    // AddImmediate tends to be used as a NOP instruction, if we detect this we can
+    // completely skip the entire code path and no emit anything
+    optimizer.skip_dummy_addimmediate = true;
+
+    // SMO tends to emit a lot of unnecessary method moves, we can mitigate this by only emitting
+    // one if our register isn't "dirty"
+    optimizer.optimize_for_method_move = true;
+
+    // Check to see if we can skip emitting certain instructions
+    Optimizer_ScanFlags();
+
+    const u32 op_count = static_cast<u32>(code.size());
+    for (u32 i = 0; i < op_count; i++) {
+        if (i < op_count - 1) {
+            pc = i + 1;
+            next_opcode = GetOpCode();
+        } else {
+            next_opcode = {};
+        }
+        pc = i;
+        Compile_NextInstruction();
+    }
+
+    L(end_of_code);
+
+    Common::X64::ABI_PopRegistersAndAdjustStack(*this, Common::X64::ABI_ALL_CALLEE_SAVED, 8);
+    ret();
+    ready();
+    program = getCode<ProgramType>();
+}
+
+bool MacroJITx64Impl::Compile_NextInstruction() {
+    const auto opcode = GetOpCode();
+    if (labels[pc].getAddress()) {
+        return false;
+    }
+
+    L(labels[pc]);
+
+    switch (opcode.operation) {
+    case Macro::Operation::ALU:
+        Compile_ALU(opcode);
+        break;
+    case Macro::Operation::AddImmediate:
+        Compile_AddImmediate(opcode);
+        break;
+    case Macro::Operation::ExtractInsert:
+        Compile_ExtractInsert(opcode);
+        break;
+    case Macro::Operation::ExtractShiftLeftImmediate:
+        Compile_ExtractShiftLeftImmediate(opcode);
+        break;
+    case Macro::Operation::ExtractShiftLeftRegister:
+        Compile_ExtractShiftLeftRegister(opcode);
+        break;
+    case Macro::Operation::Read:
+        Compile_Read(opcode);
+        break;
+    case Macro::Operation::Branch:
+        Compile_Branch(opcode);
+        break;
+    default:
+        UNIMPLEMENTED_MSG("Unimplemented opcode {}", opcode.operation.Value());
+        break;
+    }
+
+    if (optimizer.has_delayed_pc) {
+        if (opcode.is_exit) {
+            mov(rax, end_of_code);
+            test(BRANCH_HOLDER, BRANCH_HOLDER);
+            cmove(BRANCH_HOLDER, rax);
+            // Jump to next instruction to skip delay slot check
+            je(labels[pc + 1], T_NEAR);
+        } else {
+            // TODO(ogniK): Optimize delay slot branching
+            Xbyak::Label no_delay_slot{};
+            test(BRANCH_HOLDER, BRANCH_HOLDER);
+            je(no_delay_slot, T_NEAR);
+            mov(rax, BRANCH_HOLDER);
+            xor_(BRANCH_HOLDER, BRANCH_HOLDER);
+            jmp(rax);
+            L(no_delay_slot);
+        }
+        L(delay_skip[pc]);
+        if (opcode.is_exit) {
+            return false;
+        }
+    } else {
+        test(BRANCH_HOLDER, BRANCH_HOLDER);
+        jne(end_of_code, T_NEAR);
+        if (opcode.is_exit) {
+            inc(BRANCH_HOLDER);
+            return false;
+        }
+    }
+    return true;
+}
+
+Xbyak::Reg32 Tegra::MacroJITx64Impl::Compile_FetchParameter() {
+    mov(eax, dword[PARAMETERS]);
+    add(PARAMETERS, sizeof(u32));
+    return eax;
+}
+
+Xbyak::Reg32 MacroJITx64Impl::Compile_GetRegister(u32 index, Xbyak::Reg32 dst) {
+    if (index == 0) {
+        // Register 0 is always zero
+        xor_(dst, dst);
+    } else {
+        mov(dst, dword[STATE + offsetof(JITState, registers) + index * sizeof(u32)]);
+    }
+
+    return dst;
+}
+
+void MacroJITx64Impl::Compile_ProcessResult(Macro::ResultOperation operation, u32 reg) {
+    auto SetRegister = [=](u32 reg, Xbyak::Reg32 result) {
+        // Register 0 is supposed to always return 0. NOP is implemented as a store to the zero
+        // register.
+        if (reg == 0) {
+            return;
+        }
+        mov(dword[STATE + offsetof(JITState, registers) + reg * sizeof(u32)], result);
+    };
+    auto SetMethodAddress = [=](Xbyak::Reg32 reg) { mov(METHOD_ADDRESS, reg); };
+
+    switch (operation) {
+    case Macro::ResultOperation::IgnoreAndFetch:
+        SetRegister(reg, Compile_FetchParameter());
+        break;
+    case Macro::ResultOperation::Move:
+        SetRegister(reg, RESULT);
+        break;
+    case Macro::ResultOperation::MoveAndSetMethod:
+        SetRegister(reg, RESULT);
+        SetMethodAddress(RESULT);
+        break;
+    case Macro::ResultOperation::FetchAndSend:
+        // Fetch parameter and send result.
+        SetRegister(reg, Compile_FetchParameter());
+        Compile_Send(RESULT);
+        break;
+    case Macro::ResultOperation::MoveAndSend:
+        // Move and send result.
+        SetRegister(reg, RESULT);
+        Compile_Send(RESULT);
+        break;
+    case Macro::ResultOperation::FetchAndSetMethod:
+        // Fetch parameter and use result as Method Address.
+        SetRegister(reg, Compile_FetchParameter());
+        SetMethodAddress(RESULT);
+        break;
+    case Macro::ResultOperation::MoveAndSetMethodFetchAndSend:
+        // Move result and use as Method Address, then fetch and send parameter.
+        SetRegister(reg, RESULT);
+        SetMethodAddress(RESULT);
+        Compile_Send(Compile_FetchParameter());
+        break;
+    case Macro::ResultOperation::MoveAndSetMethodSend:
+        // Move result and use as Method Address, then send bits 12:17 of result.
+        SetRegister(reg, RESULT);
+        SetMethodAddress(RESULT);
+        shr(RESULT, 12);
+        and_(RESULT, 0b111111);
+        Compile_Send(RESULT);
+        break;
+    default:
+        UNIMPLEMENTED_MSG("Unimplemented macro operation {}", static_cast<std::size_t>(operation));
+    }
+}
+
+Macro::Opcode MacroJITx64Impl::GetOpCode() const {
+    ASSERT(pc < code.size());
+    return {code[pc]};
+}
+
+std::bitset<32> MacroJITx64Impl::PersistentCallerSavedRegs() const {
+    return PERSISTENT_REGISTERS & Common::X64::ABI_ALL_CALLER_SAVED;
+}
+
+} // namespace Tegra
@@ -0,0 +1,97 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <array>
+#include <bitset>
+#include <xbyak.h>
+#include "common/bit_field.h"
+#include "common/common_types.h"
+#include "common/x64/xbyak_abi.h"
+#include "video_core/macro/macro.h"
+
+namespace Tegra {
+
+namespace Engines {
+class Maxwell3D;
+}
+
+/// MAX_CODE_SIZE is arbitrarily chosen based on current booting games
+constexpr size_t MAX_CODE_SIZE = 0x10000;
+
+class MacroJITx64 final : public MacroEngine {
+public:
+    explicit MacroJITx64(Engines::Maxwell3D& maxwell3d);
+
+protected:
+    std::unique_ptr<CachedMacro> Compile(const std::vector<u32>& code) override;
+
+private:
+    Engines::Maxwell3D& maxwell3d;
+};
+
+class MacroJITx64Impl : public Xbyak::CodeGenerator, public CachedMacro {
+public:
+    MacroJITx64Impl(Engines::Maxwell3D& maxwell3d, const std::vector<u32>& code);
+    ~MacroJITx64Impl();
+
+    void Execute(const std::vector<u32>& parameters, u32 method) override;
+
+    void Compile_ALU(Macro::Opcode opcode);
+    void Compile_AddImmediate(Macro::Opcode opcode);
+    void Compile_ExtractInsert(Macro::Opcode opcode);
+    void Compile_ExtractShiftLeftImmediate(Macro::Opcode opcode);
+    void Compile_ExtractShiftLeftRegister(Macro::Opcode opcode);
+    void Compile_Read(Macro::Opcode opcode);
+    void Compile_Branch(Macro::Opcode opcode);
+
+private:
+    void Optimizer_ScanFlags();
+
+    void Compile();
+    bool Compile_NextInstruction();
+
+    Xbyak::Reg32 Compile_FetchParameter();
+    Xbyak::Reg32 Compile_GetRegister(u32 index, Xbyak::Reg32 dst);
+
+    void Compile_ProcessResult(Macro::ResultOperation operation, u32 reg);
+    void Compile_Send(Xbyak::Reg32 value);
+
+    Macro::Opcode GetOpCode() const;
+    std::bitset<32> PersistentCallerSavedRegs() const;
+
+    struct JITState {
+        Engines::Maxwell3D* maxwell3d{};
+        std::array<u32, Macro::NUM_MACRO_REGISTERS> registers{};
+        u32 carry_flag{};
+    };
+    static_assert(offsetof(JITState, maxwell3d) == 0, "Maxwell3D is not at 0x0");
+    using ProgramType = void (*)(JITState*, const u32*);
+
+    struct OptimizerState {
+        bool can_skip_carry{};
+        bool has_delayed_pc{};
+        bool zero_reg_skip{};
+        bool skip_dummy_addimmediate{};
+        bool optimize_for_method_move{};
+    };
+    OptimizerState optimizer{};
+
+    std::optional<Macro::Opcode> next_opcode{};
+    ProgramType program{nullptr};
+
+    std::array<Xbyak::Label, MAX_CODE_SIZE> labels;
+    std::array<Xbyak::Label, MAX_CODE_SIZE> delay_skip;
+    Xbyak::Label end_of_code{};
+
+    bool is_delay_slot{};
+    u32 pc{};
+    std::optional<u32> delayed_pc;
+
+    const std::vector<u32>& code;
+    Engines::Maxwell3D& maxwell3d;
+};
+
+} // namespace Tegra
@@ -83,6 +83,7 @@ static constexpr ConversionArray morton_to_linear_fns = {
    MortonCopy<true, PixelFormat::RGBA8_SRGB>,
    MortonCopy<true, PixelFormat::RG8U>,
    MortonCopy<true, PixelFormat::RG8S>,
+    MortonCopy<true, PixelFormat::RG8UI>,
    MortonCopy<true, PixelFormat::RG32UI>,
    MortonCopy<true, PixelFormat::RGBX16F>,
    MortonCopy<true, PixelFormat::R32UI>,
@@ -166,6 +167,7 @@ static constexpr ConversionArray linear_to_morton_fns = {
    MortonCopy<false, PixelFormat::RGBA8_SRGB>,
    MortonCopy<false, PixelFormat::RG8U>,
    MortonCopy<false, PixelFormat::RG8S>,
+    MortonCopy<false, PixelFormat::RG8UI>,
    MortonCopy<false, PixelFormat::RG32UI>,
    MortonCopy<false, PixelFormat::RGBX16F>,
    MortonCopy<false, PixelFormat::R32UI>,
--- a/Show More
+++ b/Show More