gl_shader_cache: Fix texture view for cubemaps as cubemap arrays

Cubemaps are considered layered and to create a texture view the texture
mustn't be a layered texture, resulting in cubemaps being bound as
cubemap arrays. To fix this issue this commit introduces an extra
surface parameter called "is_array" and uses this to query for texture
view creation.

Now that texture views for cubemaps are actually being created, this
also fixes the number of layers created for the texture view (since they
have to be 6 to create a texture view of cubemaps).

.github/ISSUE_TEMPLATE.md

@@ -1,16 +1,27 @@
 <!--
 Please keep in mind yuzu is EXPERIMENTAL SOFTWARE.
 
-Please read the FAQ: https://yuzu-emu.org/wiki/faq/
+Please read the FAQ:
+https://yuzu-emu.org/wiki/faq/
 
-When submitting an issue, please do the following:
+THIS IS NOT A SUPPORT FORUM, FOR SUPPORT GO TO:
+https://community.citra-emu.org/
 
-- Provide the version (commit hash) of yuzu you are using.
-- Provide sufficient detail for the issue to be reproduced.
-- Provide:
+If the FAQ does not answer your question, please go to:
+https://community.citra-emu.org/
+
+When submitting an issue, please check the following:
+
+- You have read the above.
+- You have provided the version (commit hash) of yuzu you are using.
+- You have provided sufficient detail for the issue to be reproduced.
+- You have provided system specs (if relevant).
+- Please also provide:
+  - For any issues, a log file
   - For crashes, a backtrace.
   - For graphical issues, comparison screenshots with real hardware.
   - For emulation inaccuracies, a test-case (if able).
+
 -->
 
 

.travis/linux/docker.sh

@@ -1,12 +1,12 @@
 #!/bin/bash -ex
 
 apt-get update
-apt-get install --no-install-recommends -y build-essential git libqt5opengl5-dev libsdl2-dev libssl-dev python qtbase5-dev wget cmake ninja-build ccache
+apt-get install --no-install-recommends -y build-essential git libqt5opengl5-dev libsdl2-dev libssl-dev python qtbase5-dev qtwebengine5-dev wget cmake ninja-build ccache
 
 cd /yuzu
 
 mkdir build && cd build
-cmake .. -DYUZU_USE_BUNDLED_UNICORN=ON -DCMAKE_BUILD_TYPE=Release -DCMAKE_C_COMPILER=/usr/lib/ccache/gcc -DCMAKE_CXX_COMPILER=/usr/lib/ccache/g++ -DYUZU_ENABLE_COMPATIBILITY_REPORTING=${ENABLE_COMPATIBILITY_REPORTING:-"OFF"} -DENABLE_COMPATIBILITY_LIST_DOWNLOAD=ON -DUSE_DISCORD_PRESENCE=ON -G Ninja
+cmake .. -DYUZU_USE_BUNDLED_UNICORN=ON -DYUZU_USE_QT_WEB_ENGINE=ON -DCMAKE_BUILD_TYPE=Release -DCMAKE_C_COMPILER=/usr/lib/ccache/gcc -DCMAKE_CXX_COMPILER=/usr/lib/ccache/g++ -DYUZU_ENABLE_COMPATIBILITY_REPORTING=${ENABLE_COMPATIBILITY_REPORTING:-"OFF"} -DENABLE_COMPATIBILITY_LIST_DOWNLOAD=ON -DUSE_DISCORD_PRESENCE=ON -G Ninja
 ninja
 
 ccache -s

.travis/macos/build.sh

@@ -9,7 +9,7 @@ export PATH="/usr/local/opt/ccache/libexec:$PATH"
 
 mkdir build && cd build
 cmake --version
-cmake .. -DYUZU_USE_BUNDLED_UNICORN=ON -DCMAKE_BUILD_TYPE=Release -DENABLE_COMPATIBILITY_LIST_DOWNLOAD=ON -DYUZU_ENABLE_COMPATIBILITY_REPORTING=${ENABLE_COMPATIBILITY_REPORTING:-"OFF"} -DUSE_DISCORD_PRESENCE=ON
+cmake .. -DYUZU_USE_BUNDLED_UNICORN=ON -DYUZU_USE_QT_WEB_ENGINE=ON -DCMAKE_BUILD_TYPE=Release -DENABLE_COMPATIBILITY_LIST_DOWNLOAD=ON -DYUZU_ENABLE_COMPATIBILITY_REPORTING=${ENABLE_COMPATIBILITY_REPORTING:-"OFF"} -DUSE_DISCORD_PRESENCE=ON
 make -j4
 
 ccache -s

CMakeLists.txt

@@ -19,6 +19,8 @@ option(ENABLE_WEB_SERVICE "Enable web services (telemetry, etc.)" ON)
 
 option(YUZU_USE_BUNDLED_UNICORN "Build/Download bundled Unicorn" ON)
 
+option(YUZU_USE_QT_WEB_ENGINE "Use QtWebEngine for web applet implementation" OFF)
+
 option(ENABLE_CUBEB "Enables the cubeb audio backend" ON)
 
 option(USE_DISCORD_PRESENCE "Enables Discord Rich Presence" OFF)
@@ -302,7 +304,7 @@ endif()
 if (ENABLE_QT)
     if (YUZU_USE_BUNDLED_QT)
         if ((MSVC_VERSION GREATER_EQUAL 1910 AND MSVC_VERSION LESS 1920) AND ARCHITECTURE_x86_64)
-            set(QT_VER qt-5.10.0-msvc2015_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()
@@ -319,6 +321,10 @@ if (ENABLE_QT)
     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

CMakeModules/CopyYuzuQt5Deps.cmake

@@ -5,6 +5,7 @@ function(copy_yuzu_Qt5_deps target_dir)
     set(Qt5_PLATFORMS_DIR "${Qt5_DIR}/../../../plugins/platforms/")
     set(Qt5_STYLES_DIR "${Qt5_DIR}/../../../plugins/styles/")
     set(Qt5_IMAGEFORMATS_DIR "${Qt5_DIR}/../../../plugins/imageformats/")
+    set(Qt5_RESOURCES_DIR "${Qt5_DIR}/../../../resources/")
     set(PLATFORMS ${DLL_DEST}platforms/)
     set(STYLES ${DLL_DEST}styles/)
     set(IMAGEFORMATS ${DLL_DEST}imageformats/)
@@ -17,7 +18,35 @@ function(copy_yuzu_Qt5_deps target_dir)
         Qt5OpenGL$<$<CONFIG:Debug>:d>.*
         Qt5Widgets$<$<CONFIG:Debug>:d>.*
     )
+
+    if (YUZU_USE_QT_WEB_ENGINE)
+        windows_copy_files(${target_dir} ${Qt5_DLL_DIR} ${DLL_DEST} 
+            Qt5Network$<$<CONFIG:Debug>:d>.*
+            Qt5Positioning$<$<CONFIG:Debug>:d>.*
+            Qt5PrintSupport$<$<CONFIG:Debug>:d>.*
+            Qt5Qml$<$<CONFIG:Debug>:d>.*
+            Qt5Quick$<$<CONFIG:Debug>:d>.*
+            Qt5QuickWidgets$<$<CONFIG:Debug>:d>.*
+            Qt5WebChannel$<$<CONFIG:Debug>:d>.*
+            Qt5WebEngine$<$<CONFIG:Debug>:d>.*
+            Qt5WebEngineCore$<$<CONFIG:Debug>:d>.*
+            Qt5WebEngineWidgets$<$<CONFIG:Debug>:d>.*
+            QtWebEngineProcess$<$<CONFIG:Debug>:d>.*
+        )
+
+        windows_copy_files(${target_dir} ${Qt5_RESOURCES_DIR} ${DLL_DEST}
+            qtwebengine_resources.pak
+            qtwebengine_devtools_resources.pak
+            qtwebengine_resources_100p.pak
+            qtwebengine_resources_200p.pak
+            icudtl.dat
+        )
+    endif ()
+
     windows_copy_files(yuzu ${Qt5_PLATFORMS_DIR} ${PLATFORMS} qwindows$<$<CONFIG:Debug>:d>.*)
     windows_copy_files(yuzu ${Qt5_STYLES_DIR} ${STYLES} qwindowsvistastyle$<$<CONFIG:Debug>:d>.*)
-    windows_copy_files(yuzu ${Qt5_IMAGEFORMATS_DIR} ${IMAGEFORMATS} qjpeg$<$<CONFIG:Debug>:d>.*)
+    windows_copy_files(yuzu ${Qt5_IMAGEFORMATS_DIR} ${IMAGEFORMATS}
+        qjpeg$<$<CONFIG:Debug>:d>.*
+        qgif$<$<CONFIG:Debug>:d>.*
+        )
 endfunction(copy_yuzu_Qt5_deps)

appveyor.yml

@@ -42,7 +42,7 @@ before_build:
         $COMPAT = if ($env:ENABLE_COMPATIBILITY_REPORTING -eq $null) {0} else {$env:ENABLE_COMPATIBILITY_REPORTING}
         if ($env:BUILD_TYPE -eq 'msvc') {
           # redirect stderr and change the exit code to prevent powershell from cancelling the build if cmake prints a warning
-          cmd /C 'cmake -G "Visual Studio 15 2017 Win64" -DYUZU_USE_BUNDLED_QT=1 -DYUZU_USE_BUNDLED_SDL2=1 -DYUZU_USE_BUNDLED_UNICORN=1 -DENABLE_COMPATIBILITY_LIST_DOWNLOAD=ON -DYUZU_ENABLE_COMPATIBILITY_REPORTING=${COMPAT} -DUSE_DISCORD_PRESENCE=ON .. 2>&1 && exit 0'
+          cmd /C 'cmake -G "Visual Studio 15 2017 Win64" -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=${COMPAT} -DUSE_DISCORD_PRESENCE=ON .. 2>&1 && exit 0'
         } else {
           C:\msys64\usr\bin\bash.exe -lc "cmake -G 'MSYS Makefiles' -DYUZU_BUILD_UNICORN=1 -DCMAKE_BUILD_TYPE=Release -DENABLE_COMPATIBILITY_LIST_DOWNLOAD=ON -DYUZU_ENABLE_COMPATIBILITY_REPORTING=${COMPAT} -DUSE_DISCORD_PRESENCE=ON .. 2>&1"
         }
@@ -94,6 +94,7 @@ after_build:
           Copy-Item "$BUILD_DIR\*" -Destination $RELEASE_DIST -Recurse
           rm "$RELEASE_DIST\*.exe"
           Get-ChildItem "$BUILD_DIR" -Recurse -Filter "yuzu*.exe" | Copy-Item -destination $RELEASE_DIST
+          Get-ChildItem "$BUILD_DIR" -Recurse -Filter "QtWebEngineProcess*.exe" | Copy-Item -destination $RELEASE_DIST
           Copy-Item .\license.txt -Destination $RELEASE_DIST
           Copy-Item .\README.md -Destination $RELEASE_DIST
           7z a -tzip $MSVC_BUILD_ZIP $RELEASE_DIST\*

src/audio_core/stream.cpp

@@ -68,7 +68,7 @@ static void VolumeAdjustSamples(std::vector<s16>& samples) {
     }
 
     // Implementation of a volume slider with a dynamic range of 60 dB
-    const float volume_scale_factor{std::exp(6.90775f * volume) * 0.001f};
+    const float volume_scale_factor = volume == 0 ? 0 : std::exp(6.90775f * volume) * 0.001f;
     for (auto& sample : samples) {
         sample = static_cast<s16>(sample * volume_scale_factor);
     }

src/audio_core/time_stretch.cpp

@@ -53,8 +53,8 @@ std::size_t TimeStretcher::Process(const s16* in, std::size_t num_in, s16* out,
     const double lpf_gain = 1.0 - std::exp(-time_delta / lpf_time_scale);
     m_stretch_ratio += lpf_gain * (current_ratio - m_stretch_ratio);
 
-    // Place a lower limit of 5% speed.  When a game boots up, there will be
-    // many silence samples.  These do not need to be timestretched.
+    // Place a lower limit of 5% speed. When a game boots up, there will be
+    // many silence samples. These do not need to be timestretched.
     m_stretch_ratio = std::max(m_stretch_ratio, 0.05);
     m_sound_touch.setTempo(m_stretch_ratio);
 

src/core/CMakeLists.txt

@@ -88,11 +88,15 @@ add_library(core STATIC
     frontend/applets/profile_select.h
     frontend/applets/software_keyboard.cpp
     frontend/applets/software_keyboard.h
+    frontend/applets/web_browser.cpp
+    frontend/applets/web_browser.h
     frontend/emu_window.cpp
     frontend/emu_window.h
     frontend/framebuffer_layout.cpp
     frontend/framebuffer_layout.h
     frontend/input.h
+    frontend/scope_acquire_window_context.cpp
+    frontend/scope_acquire_window_context.h
     gdbstub/gdbstub.cpp
     gdbstub/gdbstub.h
     hle/ipc.h
@@ -173,6 +177,8 @@ add_library(core STATIC
     hle/service/am/applets/software_keyboard.h
     hle/service/am/applets/stub_applet.cpp
     hle/service/am/applets/stub_applet.h
+    hle/service/am/applets/web_browser.cpp
+    hle/service/am/applets/web_browser.h
     hle/service/am/idle.cpp
     hle/service/am/idle.h
     hle/service/am/omm.cpp

src/core/core.cpp

@@ -30,8 +30,11 @@
 #include "core/hle/service/sm/sm.h"
 #include "core/loader/loader.h"
 #include "core/perf_stats.h"
+#include "core/settings.h"
 #include "core/telemetry_session.h"
+#include "frontend/applets/profile_select.h"
 #include "frontend/applets/software_keyboard.h"
+#include "frontend/applets/web_browser.h"
 #include "video_core/debug_utils/debug_utils.h"
 #include "video_core/gpu.h"
 #include "video_core/renderer_base.h"
@@ -94,6 +97,11 @@ struct System::Impl {
         CoreTiming::Init();
         kernel.Initialize();
 
+        const auto current_time = std::chrono::duration_cast<std::chrono::seconds>(
+            std::chrono::system_clock::now().time_since_epoch());
+        Settings::values.custom_rtc_differential =
+            Settings::values.custom_rtc.value_or(current_time) - current_time;
+
         // Create a default fs if one doesn't already exist.
         if (virtual_filesystem == nullptr)
             virtual_filesystem = std::make_shared<FileSys::RealVfsFilesystem>();
@@ -103,6 +111,8 @@ struct System::Impl {
             profile_selector = std::make_unique<Core::Frontend::DefaultProfileSelectApplet>();
         if (software_keyboard == nullptr)
             software_keyboard = std::make_unique<Core::Frontend::DefaultSoftwareKeyboardApplet>();
+        if (web_browser == nullptr)
+            web_browser = std::make_unique<Core::Frontend::DefaultWebBrowserApplet>();
 
         auto main_process = Kernel::Process::Create(kernel, "main");
         kernel.MakeCurrentProcess(main_process.get());
@@ -199,6 +209,11 @@ struct System::Impl {
         // Close app loader
         app_loader.reset();
 
+        // Clear all applets
+        profile_selector.reset();
+        software_keyboard.reset();
+        web_browser.reset();
+
         LOG_DEBUG(Core, "Shutdown OK");
     }
 
@@ -233,6 +248,7 @@ struct System::Impl {
     /// Frontend applets
     std::unique_ptr<Core::Frontend::ProfileSelectApplet> profile_selector;
     std::unique_ptr<Core::Frontend::SoftwareKeyboardApplet> software_keyboard;
+    std::unique_ptr<Core::Frontend::WebBrowserApplet> web_browser;
 
     /// Service manager
     std::shared_ptr<Service::SM::ServiceManager> service_manager;
@@ -427,22 +443,34 @@ std::shared_ptr<FileSys::VfsFilesystem> System::GetFilesystem() const {
     return impl->virtual_filesystem;
 }
 
-void System::SetProfileSelector(std::unique_ptr<Core::Frontend::ProfileSelectApplet> applet) {
+void System::SetProfileSelector(std::unique_ptr<Frontend::ProfileSelectApplet> applet) {
     impl->profile_selector = std::move(applet);
 }
 
-const Core::Frontend::ProfileSelectApplet& System::GetProfileSelector() const {
+const Frontend::ProfileSelectApplet& System::GetProfileSelector() const {
     return *impl->profile_selector;
 }
 
-void System::SetSoftwareKeyboard(std::unique_ptr<Core::Frontend::SoftwareKeyboardApplet> applet) {
+void System::SetSoftwareKeyboard(std::unique_ptr<Frontend::SoftwareKeyboardApplet> applet) {
     impl->software_keyboard = std::move(applet);
 }
 
-const Core::Frontend::SoftwareKeyboardApplet& System::GetSoftwareKeyboard() const {
+const Frontend::SoftwareKeyboardApplet& System::GetSoftwareKeyboard() const {
     return *impl->software_keyboard;
 }
 
+void System::SetWebBrowser(std::unique_ptr<Frontend::WebBrowserApplet> applet) {
+    impl->web_browser = std::move(applet);
+}
+
+Frontend::WebBrowserApplet& System::GetWebBrowser() {
+    return *impl->web_browser;
+}
+
+const Frontend::WebBrowserApplet& System::GetWebBrowser() const {
+    return *impl->web_browser;
+}
+
 System::ResultStatus System::Init(Frontend::EmuWindow& emu_window) {
     return impl->Init(*this, emu_window);
 }

src/core/core.h

@@ -11,11 +11,12 @@
 #include "common/common_types.h"
 #include "core/file_sys/vfs_types.h"
 #include "core/hle/kernel/object.h"
-#include "frontend/applets/profile_select.h"
 
 namespace Core::Frontend {
 class EmuWindow;
+class ProfileSelectApplet;
 class SoftwareKeyboardApplet;
+class WebBrowserApplet;
 } // namespace Core::Frontend
 
 namespace FileSys {
@@ -242,13 +243,18 @@ public:
 
     std::shared_ptr<FileSys::VfsFilesystem> GetFilesystem() const;
 
-    void SetProfileSelector(std::unique_ptr<Core::Frontend::ProfileSelectApplet> applet);
+    void SetProfileSelector(std::unique_ptr<Frontend::ProfileSelectApplet> applet);
 
-    const Core::Frontend::ProfileSelectApplet& GetProfileSelector() const;
+    const Frontend::ProfileSelectApplet& GetProfileSelector() const;
 
-    void SetSoftwareKeyboard(std::unique_ptr<Core::Frontend::SoftwareKeyboardApplet> applet);
+    void SetSoftwareKeyboard(std::unique_ptr<Frontend::SoftwareKeyboardApplet> applet);
 
-    const Core::Frontend::SoftwareKeyboardApplet& GetSoftwareKeyboard() const;
+    const Frontend::SoftwareKeyboardApplet& GetSoftwareKeyboard() const;
+
+    void SetWebBrowser(std::unique_ptr<Frontend::WebBrowserApplet> applet);
+
+    Frontend::WebBrowserApplet& GetWebBrowser();
+    const Frontend::WebBrowserApplet& GetWebBrowser() const;
 
 private:
     System();

src/core/file_sys/content_archive.cpp

@@ -359,6 +359,8 @@ bool NCA::ReadPFS0Section(const NCASectionHeader& section, const NCASectionTable
             dirs.push_back(std::move(npfs));
             if (IsDirectoryExeFS(dirs.back()))
                 exefs = dirs.back();
+            else if (IsDirectoryLogoPartition(dirs.back()))
+                logo = dirs.back();
         } else {
             if (has_rights_id)
                 status = Loader::ResultStatus::ErrorIncorrectTitlekeyOrTitlekek;
@@ -546,4 +548,8 @@ u64 NCA::GetBaseIVFCOffset() const {
     return ivfc_offset;
 }
 
+VirtualDir NCA::GetLogoPartition() const {
+    return logo;
+}
+
 } // namespace FileSys

src/core/file_sys/content_archive.h

@@ -74,6 +74,13 @@ inline bool IsDirectoryExeFS(const std::shared_ptr<VfsDirectory>& pfs) {
     return pfs->GetFile("main") != nullptr && pfs->GetFile("main.npdm") != nullptr;
 }
 
+inline bool IsDirectoryLogoPartition(const VirtualDir& pfs) {
+    // NintendoLogo is the static image in the top left corner while StartupMovie is the animation
+    // in the bottom right corner.
+    return pfs->GetFile("NintendoLogo.png") != nullptr &&
+           pfs->GetFile("StartupMovie.gif") != nullptr;
+}
+
 // An implementation of VfsDirectory that represents a Nintendo Content Archive (NCA) conatiner.
 // After construction, use GetStatus to determine if the file is valid and ready to be used.
 class NCA : public ReadOnlyVfsDirectory {
@@ -102,6 +109,8 @@ public:
     // Returns the base ivfc offset used in BKTR patching.
     u64 GetBaseIVFCOffset() const;
 
+    VirtualDir GetLogoPartition() const;
+
 private:
     bool CheckSupportedNCA(const NCAHeader& header);
     bool HandlePotentialHeaderDecryption();
@@ -122,6 +131,7 @@ private:
 
     VirtualFile romfs = nullptr;
     VirtualDir exefs = nullptr;
+    VirtualDir logo = nullptr;
     VirtualFile file;
     VirtualFile bktr_base_romfs;
     u64 ivfc_offset = 0;

src/core/file_sys/directory.h

@@ -39,27 +39,4 @@ static_assert(sizeof(Entry) == 0x310, "Directory Entry struct isn't exactly 0x31
 static_assert(offsetof(Entry, type) == 0x304, "Wrong offset for type in Entry.");
 static_assert(offsetof(Entry, file_size) == 0x308, "Wrong offset for file_size in Entry.");
 
-class DirectoryBackend : NonCopyable {
-public:
-    DirectoryBackend() {}
-    virtual ~DirectoryBackend() {}
-
-    /**
-     * List files contained in the directory
-     * @param count Number of entries to return at once in entries
-     * @param entries Buffer to read data into
-     * @return Number of entries listed
-     */
-    virtual u64 Read(const u64 count, Entry* entries) = 0;
-
-    /// Returns the number of entries still left to read.
-    virtual u64 GetEntryCount() const = 0;
-
-    /**
-     * Close the directory
-     * @return true if the directory closed correctly
-     */
-    virtual bool Close() const = 0;
-};
-
 } // namespace FileSys

src/core/file_sys/romfs.cpp

@@ -119,6 +119,9 @@ VirtualDir ExtractRomFS(VirtualFile file, RomFSExtractionType type) {
 
     VirtualDir out = std::move(root);
 
+    if (type == RomFSExtractionType::SingleDiscard)
+        return out->GetSubdirectories().front();
+
     while (out->GetSubdirectories().size() == 1 && out->GetFiles().empty()) {
         if (out->GetSubdirectories().front()->GetName() == "data" &&
             type == RomFSExtractionType::Truncated)

src/core/file_sys/romfs.h

@@ -33,8 +33,9 @@ struct IVFCHeader {
 static_assert(sizeof(IVFCHeader) == 0xE0, "IVFCHeader has incorrect size.");
 
 enum class RomFSExtractionType {
-    Full,      // Includes data directory
-    Truncated, // Traverses into data directory
+    Full,          // Includes data directory
+    Truncated,     // Traverses into data directory
+    SingleDiscard, // Traverses into the first subdirectory of root
 };
 
 // Converts a RomFS binary blob to VFS Filesystem

src/core/frontend/applets/web_browser.cpp

@@ -0,0 +1,24 @@
+// Copyright 2018 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/logging/log.h"
+#include "core/frontend/applets/web_browser.h"
+
+namespace Core::Frontend {
+
+WebBrowserApplet::~WebBrowserApplet() = default;
+
+DefaultWebBrowserApplet::~DefaultWebBrowserApplet() = default;
+
+void DefaultWebBrowserApplet::OpenPage(std::string_view filename,
+                                       std::function<void()> unpack_romfs_callback,
+                                       std::function<void()> finished_callback) {
+    LOG_INFO(Service_AM,
+             "(STUBBED) called - No suitable web browser implementation found to open website page "
+             "at '{}'!",
+             filename);
+    finished_callback();
+}
+
+} // namespace Core::Frontend

src/core/frontend/applets/web_browser.h

@@ -0,0 +1,28 @@
+// Copyright 2018 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <functional>
+#include <string_view>
+
+namespace Core::Frontend {
+
+class WebBrowserApplet {
+public:
+    virtual ~WebBrowserApplet();
+
+    virtual void OpenPage(std::string_view url, std::function<void()> unpack_romfs_callback,
+                          std::function<void()> finished_callback) = 0;
+};
+
+class DefaultWebBrowserApplet final : public WebBrowserApplet {
+public:
+    ~DefaultWebBrowserApplet() override;
+
+    void OpenPage(std::string_view url, std::function<void()> unpack_romfs_callback,
+                  std::function<void()> finished_callback) override;
+};
+
+} // namespace Core::Frontend

src/core/frontend/scope_acquire_window_context.cpp

@@ -0,0 +1,18 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "core/frontend/emu_window.h"
+#include "core/frontend/scope_acquire_window_context.h"
+
+namespace Core::Frontend {
+
+ScopeAcquireWindowContext::ScopeAcquireWindowContext(Core::Frontend::EmuWindow& emu_window_)
+    : emu_window{emu_window_} {
+    emu_window.MakeCurrent();
+}
+ScopeAcquireWindowContext::~ScopeAcquireWindowContext() {
+    emu_window.DoneCurrent();
+}
+
+} // namespace Core::Frontend

src/core/frontend/scope_acquire_window_context.h

@@ -0,0 +1,23 @@
+// Copyright 2019 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 Core::Frontend {
+
+class EmuWindow;
+
+/// Helper class to acquire/release window context within a given scope
+class ScopeAcquireWindowContext : NonCopyable {
+public:
+    explicit ScopeAcquireWindowContext(Core::Frontend::EmuWindow& window);
+    ~ScopeAcquireWindowContext();
+
+private:
+    Core::Frontend::EmuWindow& emu_window;
+};
+
+} // namespace Core::Frontend

src/core/hle/ipc_helpers.h

@@ -216,6 +216,11 @@ private:
 
 /// Push ///
 
+template <>
+inline void ResponseBuilder::Push(s32 value) {
+    cmdbuf[index++] = static_cast<u32>(value);
+}
+
 template <>
 inline void ResponseBuilder::Push(u32 value) {
     cmdbuf[index++] = value;
@@ -234,6 +239,22 @@ inline void ResponseBuilder::Push(ResultCode value) {
     Push<u32>(0);
 }
 
+template <>
+inline void ResponseBuilder::Push(s8 value) {
+    PushRaw(value);
+}
+
+template <>
+inline void ResponseBuilder::Push(s16 value) {
+    PushRaw(value);
+}
+
+template <>
+inline void ResponseBuilder::Push(s64 value) {
+    Push(static_cast<u32>(value));
+    Push(static_cast<u32>(value >> 32));
+}
+
 template <>
 inline void ResponseBuilder::Push(u8 value) {
     PushRaw(value);

src/core/hle/kernel/object.h

@@ -36,7 +36,6 @@ enum class HandleType : u32 {
 enum class ResetType {
     OneShot, ///< Reset automatically on object acquisition
     Sticky,  ///< Never reset automatically
-    Pulse,   ///< Reset automatically on wakeup
 };
 
 class Object : NonCopyable {

src/core/hle/kernel/readable_event.cpp

@@ -46,9 +46,6 @@ ResultCode ReadableEvent::Reset() {
 
 void ReadableEvent::WakeupAllWaitingThreads() {
     WaitObject::WakeupAllWaitingThreads();
-
-    if (reset_type == ResetType::Pulse)
-        signaled = false;
 }
 
 } // namespace Kernel

src/core/hle/kernel/svc.cpp

@@ -597,6 +597,7 @@ enum class BreakType : u32 {
     PostNROLoad = 4,
     PreNROUnload = 5,
     PostNROUnload = 6,
+    CppException = 7,
 };
 
 struct BreakReason {
@@ -669,6 +670,9 @@ static void Break(u32 reason, u64 info1, u64 info2) {
                     "Signalling debugger, Unloaded an NRO at 0x{:016X} with size 0x{:016X}", info1,
                     info2);
         break;
+    case BreakType::CppException:
+        LOG_CRITICAL(Debug_Emulated, "Signalling debugger. Uncaught C++ exception encountered.");
+        break;
     default:
         LOG_WARNING(
             Debug_Emulated,

src/core/hle/kernel/timer.cpp

@@ -68,9 +68,6 @@ void Timer::Clear() {
 
 void Timer::WakeupAllWaitingThreads() {
     WaitObject::WakeupAllWaitingThreads();
-
-    if (reset_type == ResetType::Pulse)
-        signaled = false;
 }
 
 void Timer::Signal(int cycles_late) {

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

@@ -23,6 +23,7 @@
 #include "core/hle/service/am/applets/profile_select.h"
 #include "core/hle/service/am/applets/software_keyboard.h"
 #include "core/hle/service/am/applets/stub_applet.h"
+#include "core/hle/service/am/applets/web_browser.h"
 #include "core/hle/service/am/idle.h"
 #include "core/hle/service/am/omm.h"
 #include "core/hle/service/am/spsm.h"
@@ -38,11 +39,13 @@
 namespace Service::AM {
 
 constexpr ResultCode ERR_NO_DATA_IN_CHANNEL{ErrorModule::AM, 0x2};
+constexpr ResultCode ERR_NO_MESSAGES{ErrorModule::AM, 0x3};
 constexpr ResultCode ERR_SIZE_OUT_OF_BOUNDS{ErrorModule::AM, 0x1F7};
 
 enum class AppletId : u32 {
     ProfileSelect = 0x10,
     SoftwareKeyboard = 0x11,
+    LibAppletOff = 0x17,
 };
 
 constexpr u32 POP_LAUNCH_PARAMETER_MAGIC = 0xC79497CA;
@@ -460,9 +463,17 @@ void ICommonStateGetter::GetEventHandle(Kernel::HLERequestContext& ctx) {
 void ICommonStateGetter::ReceiveMessage(Kernel::HLERequestContext& ctx) {
     LOG_DEBUG(Service_AM, "called");
 
+    const auto message = msg_queue->PopMessage();
     IPC::ResponseBuilder rb{ctx, 3};
+
+    if (message == AppletMessageQueue::AppletMessage::NoMessage) {
+        LOG_ERROR(Service_AM, "Message queue is empty");
+        rb.Push(ERR_NO_MESSAGES);
+        rb.PushEnum<AppletMessageQueue::AppletMessage>(message);
+        return;
+    }
     rb.Push(RESULT_SUCCESS);
-    rb.PushEnum<AppletMessageQueue::AppletMessage>(msg_queue->PopMessage());
+    rb.PushEnum<AppletMessageQueue::AppletMessage>(message);
 }
 
 void ICommonStateGetter::GetCurrentFocusState(Kernel::HLERequestContext& ctx) {
@@ -721,10 +732,10 @@ void IStorageAccessor::Write(Kernel::HLERequestContext& ctx) {
     IPC::RequestParser rp{ctx};
 
     const u64 offset{rp.Pop<u64>()};
-    LOG_DEBUG(Service_AM, "called, offset={}", offset);
-
     const std::vector<u8> data{ctx.ReadBuffer()};
 
+    LOG_DEBUG(Service_AM, "called, offset={}, size={}", offset, data.size());
+
     if (data.size() > backing.buffer.size() - offset) {
         LOG_ERROR(Service_AM,
                   "offset is out of bounds, backing_buffer_sz={}, data_size={}, offset={}",
@@ -744,10 +755,10 @@ void IStorageAccessor::Read(Kernel::HLERequestContext& ctx) {
     IPC::RequestParser rp{ctx};
 
     const u64 offset{rp.Pop<u64>()};
-    LOG_DEBUG(Service_AM, "called, offset={}", offset);
-
     const std::size_t size{ctx.GetWriteBufferSize()};
 
+    LOG_DEBUG(Service_AM, "called, offset={}, size={}", offset, size);
+
     if (size > backing.buffer.size() - offset) {
         LOG_ERROR(Service_AM, "offset is out of bounds, backing_buffer_sz={}, size={}, offset={}",
                   backing.buffer.size(), size, offset);
@@ -782,6 +793,8 @@ static std::shared_ptr<Applets::Applet> GetAppletFromId(AppletId id) {
         return std::make_shared<Applets::ProfileSelect>();
     case AppletId::SoftwareKeyboard:
         return std::make_shared<Applets::SoftwareKeyboard>();
+    case AppletId::LibAppletOff:
+        return std::make_shared<Applets::WebBrowser>();
     default:
         LOG_ERROR(Service_AM, "Unimplemented AppletId [{:08X}]! -- Falling back to stub!",
                   static_cast<u32>(id));

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

@@ -7,7 +7,7 @@
 #include "common/assert.h"
 #include "common/string_util.h"
 #include "core/core.h"
-#include "core/frontend/applets/software_keyboard.h"
+#include "core/frontend/applets/profile_select.h"
 #include "core/hle/service/am/am.h"
 #include "core/hle/service/am/applets/profile_select.h"
 

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

@@ -0,0 +1,190 @@
+// Copyright 2018 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <array>
+#include <cstring>
+#include <vector>
+
+#include "common/assert.h"
+#include "common/common_funcs.h"
+#include "common/common_paths.h"
+#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/mode.h"
+#include "core/file_sys/nca_metadata.h"
+#include "core/file_sys/registered_cache.h"
+#include "core/file_sys/romfs.h"
+#include "core/file_sys/vfs_types.h"
+#include "core/frontend/applets/web_browser.h"
+#include "core/hle/kernel/process.h"
+#include "core/hle/service/am/applets/web_browser.h"
+#include "core/hle/service/filesystem/filesystem.h"
+#include "core/loader/loader.h"
+
+namespace Service::AM::Applets {
+
+// TODO(DarkLordZach): There are other arguments in the WebBuffer structure that are currently not
+// parsed, for example footer mode and left stick mode. Some of these are not particularly relevant,
+// but some may be worth an implementation.
+constexpr u16 WEB_ARGUMENT_URL_TYPE = 0x6;
+
+struct WebBufferHeader {
+    u16 count;
+    INSERT_PADDING_BYTES(6);
+};
+static_assert(sizeof(WebBufferHeader) == 0x8, "WebBufferHeader has incorrect size.");
+
+struct WebArgumentHeader {
+    u16 type;
+    u16 size;
+    u32 offset;
+};
+static_assert(sizeof(WebArgumentHeader) == 0x8, "WebArgumentHeader has incorrect size.");
+
+struct WebArgumentResult {
+    u32 result_code;
+    std::array<char, 0x1000> last_url;
+    u64 last_url_size;
+};
+static_assert(sizeof(WebArgumentResult) == 0x1010, "WebArgumentResult has incorrect size.");
+
+static std::vector<u8> GetArgumentDataForTagType(const std::vector<u8>& data, u16 type) {
+    WebBufferHeader header;
+    ASSERT(sizeof(WebBufferHeader) <= data.size());
+    std::memcpy(&header, data.data(), sizeof(WebBufferHeader));
+
+    u64 offset = sizeof(WebBufferHeader);
+    for (u16 i = 0; i < header.count; ++i) {
+        WebArgumentHeader arg;
+        ASSERT(offset + sizeof(WebArgumentHeader) <= data.size());
+        std::memcpy(&arg, data.data() + offset, sizeof(WebArgumentHeader));
+        offset += sizeof(WebArgumentHeader);
+
+        if (arg.type == type) {
+            std::vector<u8> out(arg.size);
+            offset += arg.offset;
+            ASSERT(offset + arg.size <= data.size());
+            std::memcpy(out.data(), data.data() + offset, out.size());
+            return out;
+        }
+
+        offset += arg.offset + arg.size;
+    }
+
+    return {};
+}
+
+static FileSys::VirtualFile GetManualRomFS() {
+    auto& loader{Core::System::GetInstance().GetAppLoader()};
+
+    FileSys::VirtualFile out;
+    if (loader.ReadManualRomFS(out) == Loader::ResultStatus::Success)
+        return out;
+
+    const auto& installed{FileSystem::GetUnionContents()};
+    const auto res = installed.GetEntry(Core::System::GetInstance().CurrentProcess()->GetTitleID(),
+                                        FileSys::ContentRecordType::Manual);
+
+    if (res != nullptr)
+        return res->GetRomFS();
+    return nullptr;
+}
+
+WebBrowser::WebBrowser() = default;
+
+WebBrowser::~WebBrowser() = default;
+
+void WebBrowser::Initialize() {
+    Applet::Initialize();
+
+    complete = false;
+    temporary_dir.clear();
+    filename.clear();
+    status = RESULT_SUCCESS;
+
+    const auto web_arg_storage = broker.PopNormalDataToApplet();
+    ASSERT(web_arg_storage != nullptr);
+    const auto& web_arg = web_arg_storage->GetData();
+
+    const auto url_data = GetArgumentDataForTagType(web_arg, WEB_ARGUMENT_URL_TYPE);
+    filename = Common::StringFromFixedZeroTerminatedBuffer(
+        reinterpret_cast<const char*>(url_data.data()), url_data.size());
+
+    temporary_dir = FileUtil::SanitizePath(FileUtil::GetUserPath(FileUtil::UserPath::CacheDir) +
+                                               "web_applet_manual",
+                                           FileUtil::DirectorySeparator::PlatformDefault);
+    FileUtil::DeleteDirRecursively(temporary_dir);
+
+    manual_romfs = GetManualRomFS();
+    if (manual_romfs == nullptr) {
+        status = ResultCode(-1);
+        LOG_ERROR(Service_AM, "Failed to find manual for current process!");
+    }
+
+    filename =
+        FileUtil::SanitizePath(temporary_dir + DIR_SEP + "html-document" + DIR_SEP + filename,
+                               FileUtil::DirectorySeparator::PlatformDefault);
+}
+
+bool WebBrowser::TransactionComplete() const {
+    return complete;
+}
+
+ResultCode WebBrowser::GetStatus() const {
+    return status;
+}
+
+void WebBrowser::ExecuteInteractive() {
+    UNIMPLEMENTED_MSG("Unexpected interactive data recieved!");
+}
+
+void WebBrowser::Execute() {
+    if (complete)
+        return;
+
+    if (status != RESULT_SUCCESS) {
+        complete = true;
+        return;
+    }
+
+    auto& frontend{Core::System::GetInstance().GetWebBrowser()};
+
+    frontend.OpenPage(filename, [this] { UnpackRomFS(); }, [this] { Finalize(); });
+}
+
+void WebBrowser::UnpackRomFS() {
+    if (unpacked)
+        return;
+
+    ASSERT(manual_romfs != nullptr);
+    const auto dir =
+        FileSys::ExtractRomFS(manual_romfs, FileSys::RomFSExtractionType::SingleDiscard);
+    const auto& vfs{Core::System::GetInstance().GetFilesystem()};
+    const auto temp_dir = vfs->CreateDirectory(temporary_dir, FileSys::Mode::ReadWrite);
+    FileSys::VfsRawCopyD(dir, temp_dir);
+
+    unpacked = true;
+}
+
+void WebBrowser::Finalize() {
+    complete = true;
+
+    WebArgumentResult out{};
+    out.result_code = 0;
+    out.last_url_size = 0;
+
+    std::vector<u8> data(sizeof(WebArgumentResult));
+    std::memcpy(data.data(), &out, sizeof(WebArgumentResult));
+
+    broker.PushNormalDataFromApplet(IStorage{data});
+    broker.SignalStateChanged();
+
+    FileUtil::DeleteDirRecursively(temporary_dir);
+}
+
+} // namespace Service::AM::Applets

src/core/hle/service/am/applets/web_browser.h

@@ -0,0 +1,44 @@
+// Copyright 2018 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "core/file_sys/vfs_types.h"
+#include "core/hle/service/am/am.h"
+#include "core/hle/service/am/applets/applets.h"
+
+namespace Service::AM::Applets {
+
+class WebBrowser final : public Applet {
+public:
+    WebBrowser();
+    ~WebBrowser() override;
+
+    void Initialize() override;
+
+    bool TransactionComplete() const override;
+    ResultCode GetStatus() const override;
+    void ExecuteInteractive() override;
+    void Execute() override;
+
+    // Callback to be fired when the frontend needs the manual RomFS unpacked to temporary
+    // directory. This is a blocking call and may take a while as some manuals can be up to 100MB in
+    // size. Attempting to access files at filename before invocation is likely to not work.
+    void UnpackRomFS();
+
+    // Callback to be fired when the frontend is finished browsing. This will delete the temporary
+    // manual RomFS extracted files, so ensure this is only called at actual finalization.
+    void Finalize();
+
+private:
+    bool complete = false;
+    bool unpacked = false;
+    ResultCode status = RESULT_SUCCESS;
+
+    FileSys::VirtualFile manual_romfs;
+    std::string temporary_dir;
+    std::string filename;
+};
+
+} // namespace Service::AM::Applets

src/core/hle/service/hid/controllers/npad.cpp

@@ -410,6 +410,8 @@ void Controller_NPad::OnUpdate(u8* data, std::size_t data_len) {
         libnx_entry.pad.pad_states.raw = pad_state.pad_states.raw;
         libnx_entry.pad.l_stick = pad_state.l_stick;
         libnx_entry.pad.r_stick = pad_state.r_stick;
+
+        press_state |= static_cast<u32>(pad_state.pad_states.raw);
     }
     std::memcpy(data + NPAD_OFFSET, shared_memory_entries.data(),
                 shared_memory_entries.size() * sizeof(NPadEntry));
@@ -636,6 +638,10 @@ void Controller_NPad::ClearAllControllers() {
                   });
 }
 
+u32 Controller_NPad::GetAndResetPressState() {
+    return std::exchange(press_state, 0);
+}
+
 bool Controller_NPad::IsControllerSupported(NPadControllerType controller) const {
     const bool support_handheld =
         std::find(supported_npad_id_types.begin(), supported_npad_id_types.end(), NPAD_HANDHELD) !=

src/core/hle/service/hid/controllers/npad.h

@@ -124,6 +124,10 @@ public:
     void ConnectAllDisconnectedControllers();
     void ClearAllControllers();
 
+    // Logical OR for all buttons presses on all controllers
+    // Specifically for cheat engine and other features.
+    u32 GetAndResetPressState();
+
     static std::size_t NPadIdToIndex(u32 npad_id);
     static u32 IndexToNPad(std::size_t index);
 
@@ -292,6 +296,8 @@ private:
         bool is_connected;
     };
 
+    u32 press_state{};
+
     NPadType style{};
     std::array<NPadEntry, 10> shared_memory_entries{};
     std::array<

src/core/hle/service/hid/hid.h

@@ -4,12 +4,122 @@
 
 #pragma once
 
+#include "controllers/controller_base.h"
+#include "core/hle/service/service.h"
+
+namespace CoreTiming {
+struct EventType;
+}
+
+namespace Kernel {
+class SharedMemory;
+}
+
 namespace SM {
 class ServiceManager;
 }
 
 namespace Service::HID {
 
+enum class HidController : std::size_t {
+    DebugPad,
+    Touchscreen,
+    Mouse,
+    Keyboard,
+    XPad,
+    Unknown1,
+    Unknown2,
+    Unknown3,
+    SixAxisSensor,
+    NPad,
+    Gesture,
+
+    MaxControllers,
+};
+
+class IAppletResource final : public ServiceFramework<IAppletResource> {
+public:
+    IAppletResource();
+    ~IAppletResource() override;
+
+    void ActivateController(HidController controller);
+    void DeactivateController(HidController controller);
+
+    template <typename T>
+    T& GetController(HidController controller) {
+        return static_cast<T&>(*controllers[static_cast<size_t>(controller)]);
+    }
+
+    template <typename T>
+    const T& GetController(HidController controller) const {
+        return static_cast<T&>(*controllers[static_cast<size_t>(controller)]);
+    }
+
+private:
+    template <typename T>
+    void MakeController(HidController controller) {
+        controllers[static_cast<std::size_t>(controller)] = std::make_unique<T>();
+    }
+
+    void GetSharedMemoryHandle(Kernel::HLERequestContext& ctx);
+    void UpdateControllers(u64 userdata, int cycles_late);
+
+    Kernel::SharedPtr<Kernel::SharedMemory> shared_mem;
+
+    CoreTiming::EventType* pad_update_event;
+
+    std::array<std::unique_ptr<ControllerBase>, static_cast<size_t>(HidController::MaxControllers)>
+        controllers{};
+};
+
+class Hid final : public ServiceFramework<Hid> {
+public:
+    Hid();
+    ~Hid() override;
+
+    std::shared_ptr<IAppletResource> GetAppletResource();
+
+private:
+    void CreateAppletResource(Kernel::HLERequestContext& ctx);
+    void ActivateXpad(Kernel::HLERequestContext& ctx);
+    void ActivateDebugPad(Kernel::HLERequestContext& ctx);
+    void ActivateTouchScreen(Kernel::HLERequestContext& ctx);
+    void ActivateMouse(Kernel::HLERequestContext& ctx);
+    void ActivateKeyboard(Kernel::HLERequestContext& ctx);
+    void ActivateGesture(Kernel::HLERequestContext& ctx);
+    void ActivateNpadWithRevision(Kernel::HLERequestContext& ctx);
+    void StartSixAxisSensor(Kernel::HLERequestContext& ctx);
+    void SetGyroscopeZeroDriftMode(Kernel::HLERequestContext& ctx);
+    void IsSixAxisSensorAtRest(Kernel::HLERequestContext& ctx);
+    void SetSupportedNpadStyleSet(Kernel::HLERequestContext& ctx);
+    void GetSupportedNpadStyleSet(Kernel::HLERequestContext& ctx);
+    void SetSupportedNpadIdType(Kernel::HLERequestContext& ctx);
+    void ActivateNpad(Kernel::HLERequestContext& ctx);
+    void AcquireNpadStyleSetUpdateEventHandle(Kernel::HLERequestContext& ctx);
+    void DisconnectNpad(Kernel::HLERequestContext& ctx);
+    void GetPlayerLedPattern(Kernel::HLERequestContext& ctx);
+    void SetNpadJoyHoldType(Kernel::HLERequestContext& ctx);
+    void GetNpadJoyHoldType(Kernel::HLERequestContext& ctx);
+    void SetNpadJoyAssignmentModeSingleByDefault(Kernel::HLERequestContext& ctx);
+    void BeginPermitVibrationSession(Kernel::HLERequestContext& ctx);
+    void EndPermitVibrationSession(Kernel::HLERequestContext& ctx);
+    void SendVibrationValue(Kernel::HLERequestContext& ctx);
+    void SendVibrationValues(Kernel::HLERequestContext& ctx);
+    void GetActualVibrationValue(Kernel::HLERequestContext& ctx);
+    void SetNpadJoyAssignmentModeDual(Kernel::HLERequestContext& ctx);
+    void MergeSingleJoyAsDualJoy(Kernel::HLERequestContext& ctx);
+    void SetNpadHandheldActivationMode(Kernel::HLERequestContext& ctx);
+    void GetVibrationDeviceInfo(Kernel::HLERequestContext& ctx);
+    void CreateActiveVibrationDeviceList(Kernel::HLERequestContext& ctx);
+    void ActivateConsoleSixAxisSensor(Kernel::HLERequestContext& ctx);
+    void StartConsoleSixAxisSensor(Kernel::HLERequestContext& ctx);
+    void StopSixAxisSensor(Kernel::HLERequestContext& ctx);
+    void SetIsPalmaAllConnectable(Kernel::HLERequestContext& ctx);
+    void SetPalmaBoostMode(Kernel::HLERequestContext& ctx);
+
+    std::shared_ptr<IAppletResource> applet_resource;
+};
+
 /// Reload input devices. Used when input configuration changed
 void ReloadInputDevices();
 

src/core/hle/service/nvflinger/nvflinger.cpp

@@ -166,7 +166,7 @@ Layer::~Layer() = default;
 
 Display::Display(u64 id, std::string name) : id(id), name(std::move(name)) {
     auto& kernel = Core::System::GetInstance().Kernel();
-    vsync_event = Kernel::WritableEvent::CreateEventPair(kernel, Kernel::ResetType::Pulse,
+    vsync_event = Kernel::WritableEvent::CreateEventPair(kernel, Kernel::ResetType::Sticky,
                                                          fmt::format("Display VSync Event {}", id));
 }
 

src/core/hle/service/pm/pm.cpp

@@ -13,7 +13,7 @@ public:
     explicit BootMode() : ServiceFramework{"pm:bm"} {
         static const FunctionInfo functions[] = {
             {0, &BootMode::GetBootMode, "GetBootMode"},
-            {1, nullptr, "SetMaintenanceBoot"},
+            {1, &BootMode::SetMaintenanceBoot, "SetMaintenanceBoot"},
         };
         RegisterHandlers(functions);
     }
@@ -24,8 +24,19 @@ private:
 
         IPC::ResponseBuilder rb{ctx, 3};
         rb.Push(RESULT_SUCCESS);
-        rb.Push<u32>(static_cast<u32>(SystemBootMode::Normal)); // Normal boot mode
+        rb.PushEnum(boot_mode);
     }
+
+    void SetMaintenanceBoot(Kernel::HLERequestContext& ctx) {
+        LOG_DEBUG(Service_PM, "called");
+
+        boot_mode = SystemBootMode::Maintenance;
+
+        IPC::ResponseBuilder rb{ctx, 2};
+        rb.Push(RESULT_SUCCESS);
+    }
+
+    SystemBootMode boot_mode = SystemBootMode::Normal;
 };
 
 class DebugMonitor final : public ServiceFramework<DebugMonitor> {

src/core/hle/service/pm/pm.h

@@ -9,7 +9,12 @@ class ServiceManager;
 }
 
 namespace Service::PM {
-enum class SystemBootMode : u32 { Normal = 0, Maintenance = 1 };
+
+enum class SystemBootMode {
+    Normal,
+    Maintenance,
+};
+
 /// Registers all PM services with the specified service manager.
 void InstallInterfaces(SM::ServiceManager& service_manager);
 

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

@@ -12,9 +12,16 @@
 #include "core/hle/kernel/client_session.h"
 #include "core/hle/service/time/interface.h"
 #include "core/hle/service/time/time.h"
+#include "core/settings.h"
 
 namespace Service::Time {
 
+static std::chrono::seconds GetSecondsSinceEpoch() {
+    return std::chrono::duration_cast<std::chrono::seconds>(
+               std::chrono::system_clock::now().time_since_epoch()) +
+           Settings::values.custom_rtc_differential;
+}
+
 static void PosixToCalendar(u64 posix_time, CalendarTime& calendar_time,
                             CalendarAdditionalInfo& additional_info,
                             [[maybe_unused]] const TimeZoneRule& /*rule*/) {
@@ -68,9 +75,7 @@ public:
 
 private:
     void GetCurrentTime(Kernel::HLERequestContext& ctx) {
-        const s64 time_since_epoch{std::chrono::duration_cast<std::chrono::seconds>(
-                                       std::chrono::system_clock::now().time_since_epoch())
-                                       .count()};
+        const s64 time_since_epoch{GetSecondsSinceEpoch().count()};
         LOG_DEBUG(Service_Time, "called");
 
         IPC::ResponseBuilder rb{ctx, 4};
@@ -266,10 +271,7 @@ void Module::Interface::GetClockSnapshot(Kernel::HLERequestContext& ctx) {
     IPC::RequestParser rp{ctx};
     const auto initial_type = rp.PopRaw<u8>();
 
-    const s64 time_since_epoch{std::chrono::duration_cast<std::chrono::seconds>(
-                                   std::chrono::system_clock::now().time_since_epoch())
-                                   .count()};
-
+    const s64 time_since_epoch{GetSecondsSinceEpoch().count()};
     const std::time_t time(time_since_epoch);
     const std::tm* tm = std::localtime(&time);
     if (tm == nullptr) {

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

@@ -32,12 +32,26 @@
 
 namespace Service::VI {
 
+constexpr ResultCode ERR_OPERATION_FAILED{ErrorModule::VI, 1};
+constexpr ResultCode ERR_UNSUPPORTED{ErrorModule::VI, 6};
+
 struct DisplayInfo {
+    /// The name of this particular display.
     char display_name[0x40]{"Default"};
-    u64 unknown_1{1};
-    u64 unknown_2{1};
-    u64 width{1280};
-    u64 height{720};
+
+    /// Whether or not the display has a limited number of layers.
+    u8 has_limited_layers{1};
+    INSERT_PADDING_BYTES(7){};
+
+    /// Indicates the total amount of layers supported by the display.
+    /// @note This is only valid if has_limited_layers is set.
+    u64 max_layers{1};
+
+    /// Maximum width in pixels.
+    u64 width{1920};
+
+    /// Maximum height in pixels.
+    u64 height{1080};
 };
 static_assert(sizeof(DisplayInfo) == 0x60, "DisplayInfo has wrong size");
 
@@ -690,13 +704,14 @@ private:
         rb.Push(RESULT_SUCCESS);
     }
 
+    // This function currently does nothing but return a success error code in
+    // the vi library itself, so do the same thing, but log out the passed in values.
     void SetLayerVisibility(Kernel::HLERequestContext& ctx) {
         IPC::RequestParser rp{ctx};
         const u64 layer_id = rp.Pop<u64>();
         const bool visibility = rp.Pop<bool>();
 
-        LOG_WARNING(Service_VI, "(STUBBED) called, layer_id=0x{:08X}, visibility={}", layer_id,
-                    visibility);
+        LOG_DEBUG(Service_VI, "called, layer_id=0x{:08X}, visibility={}", layer_id, visibility);
 
         IPC::ResponseBuilder rb{ctx, 2};
         rb.Push(RESULT_SUCCESS);
@@ -863,6 +878,22 @@ public:
     ~IApplicationDisplayService() = default;
 
 private:
+    enum class ConvertedScaleMode : u64 {
+        Freeze = 0,
+        ScaleToWindow = 1,
+        ScaleAndCrop = 2,
+        None = 3,
+        PreserveAspectRatio = 4,
+    };
+
+    enum class NintendoScaleMode : u32 {
+        None = 0,
+        Freeze = 1,
+        ScaleToWindow = 2,
+        ScaleAndCrop = 3,
+        PreserveAspectRatio = 4,
+    };
+
     void GetRelayService(Kernel::HLERequestContext& ctx) {
         LOG_WARNING(Service_VI, "(STUBBED) called");
 
@@ -947,33 +978,43 @@ private:
         IPC::RequestParser rp{ctx};
         const u64 display_id = rp.Pop<u64>();
 
-        LOG_WARNING(Service_VI, "(STUBBED) called. display_id=0x{:016X}", display_id);
+        LOG_DEBUG(Service_VI, "called. display_id=0x{:016X}", display_id);
 
         IPC::ResponseBuilder rb{ctx, 6};
         rb.Push(RESULT_SUCCESS);
 
-        if (Settings::values.use_docked_mode) {
-            rb.Push(static_cast<u64>(DisplayResolution::DockedWidth) *
-                    static_cast<u32>(Settings::values.resolution_factor));
-            rb.Push(static_cast<u64>(DisplayResolution::DockedHeight) *
-                    static_cast<u32>(Settings::values.resolution_factor));
-        } else {
-            rb.Push(static_cast<u64>(DisplayResolution::UndockedWidth) *
-                    static_cast<u32>(Settings::values.resolution_factor));
-            rb.Push(static_cast<u64>(DisplayResolution::UndockedHeight) *
-                    static_cast<u32>(Settings::values.resolution_factor));
-        }
+        // This only returns the fixed values of 1280x720 and makes no distinguishing
+        // between docked and undocked dimensions. We take the liberty of applying
+        // the resolution scaling factor here.
+        rb.Push(static_cast<u64>(DisplayResolution::UndockedWidth) *
+                static_cast<u32>(Settings::values.resolution_factor));
+        rb.Push(static_cast<u64>(DisplayResolution::UndockedHeight) *
+                static_cast<u32>(Settings::values.resolution_factor));
     }
 
     void SetLayerScalingMode(Kernel::HLERequestContext& ctx) {
         IPC::RequestParser rp{ctx};
-        const u32 scaling_mode = rp.Pop<u32>();
+        const auto scaling_mode = rp.PopEnum<NintendoScaleMode>();
         const u64 unknown = rp.Pop<u64>();
 
-        LOG_WARNING(Service_VI, "(STUBBED) called. scaling_mode=0x{:08X}, unknown=0x{:016X}",
-                    scaling_mode, unknown);
+        LOG_DEBUG(Service_VI, "called. scaling_mode=0x{:08X}, unknown=0x{:016X}",
+                  static_cast<u32>(scaling_mode), unknown);
 
         IPC::ResponseBuilder rb{ctx, 2};
+
+        if (scaling_mode > NintendoScaleMode::PreserveAspectRatio) {
+            LOG_ERROR(Service_VI, "Invalid scaling mode provided.");
+            rb.Push(ERR_OPERATION_FAILED);
+            return;
+        }
+
+        if (scaling_mode != NintendoScaleMode::ScaleToWindow &&
+            scaling_mode != NintendoScaleMode::PreserveAspectRatio) {
+            LOG_ERROR(Service_VI, "Unsupported scaling mode supplied.");
+            rb.Push(ERR_UNSUPPORTED);
+            return;
+        }
+
         rb.Push(RESULT_SUCCESS);
     }
 
@@ -1054,51 +1095,37 @@ private:
         rb.PushCopyObjects(vsync_event);
     }
 
-    enum class ConvertedScaleMode : u64 {
-        None = 0, // VI seems to name this as "Unknown" but lots of games pass it, assume it's no
-                  // scaling/default
-        Freeze = 1,
-        ScaleToWindow = 2,
-        Crop = 3,
-        NoCrop = 4,
-    };
-
-    // This struct is different, currently it's 1:1 but this might change in the future.
-    enum class NintendoScaleMode : u32 {
-        None = 0,
-        Freeze = 1,
-        ScaleToWindow = 2,
-        Crop = 3,
-        NoCrop = 4,
-    };
-
     void ConvertScalingMode(Kernel::HLERequestContext& ctx) {
         IPC::RequestParser rp{ctx};
-        auto mode = rp.PopEnum<NintendoScaleMode>();
+        const auto mode = rp.PopEnum<NintendoScaleMode>();
         LOG_DEBUG(Service_VI, "called mode={}", static_cast<u32>(mode));
 
-        IPC::ResponseBuilder rb{ctx, 4};
-        rb.Push(RESULT_SUCCESS);
+        const auto converted_mode = ConvertScalingModeImpl(mode);
+
+        if (converted_mode.Succeeded()) {
+            IPC::ResponseBuilder rb{ctx, 4};
+            rb.Push(RESULT_SUCCESS);
+            rb.PushEnum(*converted_mode);
+        } else {
+            IPC::ResponseBuilder rb{ctx, 2};
+            rb.Push(converted_mode.Code());
+        }
+    }
+
+    static ResultVal<ConvertedScaleMode> ConvertScalingModeImpl(NintendoScaleMode mode) {
         switch (mode) {
         case NintendoScaleMode::None:
-            rb.PushEnum(ConvertedScaleMode::None);
-            break;
+            return MakeResult(ConvertedScaleMode::None);
         case NintendoScaleMode::Freeze:
-            rb.PushEnum(ConvertedScaleMode::Freeze);
-            break;
+            return MakeResult(ConvertedScaleMode::Freeze);
         case NintendoScaleMode::ScaleToWindow:
-            rb.PushEnum(ConvertedScaleMode::ScaleToWindow);
-            break;
-        case NintendoScaleMode::Crop:
-            rb.PushEnum(ConvertedScaleMode::Crop);
-            break;
-        case NintendoScaleMode::NoCrop:
-            rb.PushEnum(ConvertedScaleMode::NoCrop);
-            break;
+            return MakeResult(ConvertedScaleMode::ScaleToWindow);
+        case NintendoScaleMode::ScaleAndCrop:
+            return MakeResult(ConvertedScaleMode::ScaleAndCrop);
+        case NintendoScaleMode::PreserveAspectRatio:
+            return MakeResult(ConvertedScaleMode::PreserveAspectRatio);
         default:
-            UNIMPLEMENTED_MSG("Unknown scaling mode {}", static_cast<u32>(mode));
-            rb.PushEnum(ConvertedScaleMode::None);
-            break;
+            return ERR_OPERATION_FAILED;
         }
     }
 

src/core/loader/loader.h

@@ -178,6 +178,8 @@ public:
 
     /**
      * Get the banner (typically banner section) of the application
+     * In the context of NX, this is the animation that displays in the bottom right of the screen
+     * when a game boots. Stored in GIF format.
      * @param buffer Reference to buffer to store data
      * @return ResultStatus result of function
      */
@@ -187,6 +189,8 @@ public:
 
     /**
      * Get the logo (typically logo section) of the application
+     * In the context of NX, this is the static image that displays in the top left of the screen
+     * when a game boots. Stored in JPEG format.
      * @param buffer Reference to buffer to store data
      * @return ResultStatus result of function
      */
@@ -259,6 +263,15 @@ public:
         return ResultStatus::ErrorNotImplemented;
     }
 
+    /**
+     * Get the RomFS of the manual of the application
+     * @param file The raw manual RomFS of the game
+     * @return ResultStatus result of function
+     */
+    virtual ResultStatus ReadManualRomFS(FileSys::VirtualFile& file) {
+        return ResultStatus::ErrorNotImplemented;
+    }
+
 protected:
     FileSys::VirtualFile file;
     bool is_loaded = false;

src/core/loader/nax.cpp

@@ -79,4 +79,13 @@ u64 AppLoader_NAX::ReadRomFSIVFCOffset() const {
 ResultStatus AppLoader_NAX::ReadProgramId(u64& out_program_id) {
     return nca_loader->ReadProgramId(out_program_id);
 }
+
+ResultStatus AppLoader_NAX::ReadBanner(std::vector<u8>& buffer) {
+    return nca_loader->ReadBanner(buffer);
+}
+
+ResultStatus AppLoader_NAX::ReadLogo(std::vector<u8>& buffer) {
+    return nca_loader->ReadLogo(buffer);
+}
+
 } // namespace Loader

src/core/loader/nax.h

@@ -39,6 +39,9 @@ public:
     u64 ReadRomFSIVFCOffset() const override;
     ResultStatus ReadProgramId(u64& out_program_id) override;
 
+    ResultStatus ReadBanner(std::vector<u8>& buffer) override;
+    ResultStatus ReadLogo(std::vector<u8>& buffer) override;
+
 private:
     std::unique_ptr<FileSys::NAX> nax;
     std::unique_ptr<AppLoader_NCA> nca_loader;

src/core/loader/nca.cpp

@@ -84,4 +84,23 @@ ResultStatus AppLoader_NCA::ReadProgramId(u64& out_program_id) {
     return ResultStatus::Success;
 }
 
+ResultStatus AppLoader_NCA::ReadBanner(std::vector<u8>& buffer) {
+    if (nca == nullptr || nca->GetStatus() != ResultStatus::Success)
+        return ResultStatus::ErrorNotInitialized;
+    const auto logo = nca->GetLogoPartition();
+    if (logo == nullptr)
+        return ResultStatus::ErrorNoIcon;
+    buffer = logo->GetFile("StartupMovie.gif")->ReadAllBytes();
+    return ResultStatus::Success;
+}
+
+ResultStatus AppLoader_NCA::ReadLogo(std::vector<u8>& buffer) {
+    if (nca == nullptr || nca->GetStatus() != ResultStatus::Success)
+        return ResultStatus::ErrorNotInitialized;
+    const auto logo = nca->GetLogoPartition();
+    if (logo == nullptr)
+        return ResultStatus::ErrorNoIcon;
+    buffer = logo->GetFile("NintendoLogo.png")->ReadAllBytes();
+    return ResultStatus::Success;
+}
 } // namespace Loader

src/core/loader/nca.h

@@ -39,6 +39,9 @@ public:
     u64 ReadRomFSIVFCOffset() const override;
     ResultStatus ReadProgramId(u64& out_program_id) override;
 
+    ResultStatus ReadBanner(std::vector<u8>& buffer) override;
+    ResultStatus ReadLogo(std::vector<u8>& buffer) override;
+
 private:
     std::unique_ptr<FileSys::NCA> nca;
     std::unique_ptr<AppLoader_DeconstructedRomDirectory> directory_loader;

src/core/loader/nsp.cpp

@@ -158,4 +158,21 @@ ResultStatus AppLoader_NSP::ReadControlData(FileSys::NACP& nacp) {
     nacp = *nacp_file;
     return ResultStatus::Success;
 }
+
+ResultStatus AppLoader_NSP::ReadManualRomFS(FileSys::VirtualFile& file) {
+    const auto nca = nsp->GetNCA(nsp->GetProgramTitleID(), FileSys::ContentRecordType::Manual);
+    if (nsp->GetStatus() != ResultStatus::Success || nca == nullptr)
+        return ResultStatus::ErrorNoRomFS;
+    file = nca->GetRomFS();
+    return file == nullptr ? ResultStatus::ErrorNoRomFS : ResultStatus::Success;
+}
+
+ResultStatus AppLoader_NSP::ReadBanner(std::vector<u8>& buffer) {
+    return secondary_loader->ReadBanner(buffer);
+}
+
+ResultStatus AppLoader_NSP::ReadLogo(std::vector<u8>& buffer) {
+    return secondary_loader->ReadLogo(buffer);
+}
+
 } // namespace Loader

src/core/loader/nsp.h

@@ -44,6 +44,10 @@ public:
     ResultStatus ReadIcon(std::vector<u8>& buffer) override;
     ResultStatus ReadTitle(std::string& title) override;
     ResultStatus ReadControlData(FileSys::NACP& nacp) override;
+    ResultStatus ReadManualRomFS(FileSys::VirtualFile& file) override;
+
+    ResultStatus ReadBanner(std::vector<u8>& buffer) override;
+    ResultStatus ReadLogo(std::vector<u8>& buffer) override;
 
 private:
     std::unique_ptr<FileSys::NSP> nsp;

src/core/loader/xci.cpp

@@ -128,4 +128,21 @@ ResultStatus AppLoader_XCI::ReadControlData(FileSys::NACP& control) {
     return ResultStatus::Success;
 }
 
+ResultStatus AppLoader_XCI::ReadManualRomFS(FileSys::VirtualFile& file) {
+    const auto nca = xci->GetSecurePartitionNSP()->GetNCA(xci->GetProgramTitleID(),
+                                                          FileSys::ContentRecordType::Manual);
+    if (xci->GetStatus() != ResultStatus::Success || nca == nullptr)
+        return ResultStatus::ErrorXCIMissingPartition;
+    file = nca->GetRomFS();
+    return file == nullptr ? ResultStatus::ErrorNoRomFS : ResultStatus::Success;
+}
+
+ResultStatus AppLoader_XCI::ReadBanner(std::vector<u8>& buffer) {
+    return nca_loader->ReadBanner(buffer);
+}
+
+ResultStatus AppLoader_XCI::ReadLogo(std::vector<u8>& buffer) {
+    return nca_loader->ReadLogo(buffer);
+}
+
 } // namespace Loader

src/core/loader/xci.h

@@ -44,6 +44,10 @@ public:
     ResultStatus ReadIcon(std::vector<u8>& buffer) override;
     ResultStatus ReadTitle(std::string& title) override;
     ResultStatus ReadControlData(FileSys::NACP& control) override;
+    ResultStatus ReadManualRomFS(FileSys::VirtualFile& file) override;
+
+    ResultStatus ReadBanner(std::vector<u8>& buffer) override;
+    ResultStatus ReadLogo(std::vector<u8>& buffer) override;
 
 private:
     std::unique_ptr<FileSys::XCI> xci;

src/core/settings.cpp

@@ -74,4 +74,33 @@ void Apply() {
     Service::HID::ReloadInputDevices();
 }
 
+template <typename T>
+void LogSetting(const std::string& name, const T& value) {
+    LOG_INFO(Config, "{}: {}", name, value);
+}
+
+void LogSettings() {
+    LOG_INFO(Config, "yuzu Configuration:");
+    LogSetting("System_UseDockedMode", Settings::values.use_docked_mode);
+    LogSetting("System_EnableNfc", Settings::values.enable_nfc);
+    LogSetting("System_RngSeed", Settings::values.rng_seed.value_or(0));
+    LogSetting("System_CurrentUser", Settings::values.current_user);
+    LogSetting("System_LanguageIndex", Settings::values.language_index);
+    LogSetting("Core_UseCpuJit", Settings::values.use_cpu_jit);
+    LogSetting("Core_UseMultiCore", Settings::values.use_multi_core);
+    LogSetting("Renderer_UseResolutionFactor", Settings::values.resolution_factor);
+    LogSetting("Renderer_UseFrameLimit", Settings::values.use_frame_limit);
+    LogSetting("Renderer_FrameLimit", Settings::values.frame_limit);
+    LogSetting("Renderer_UseAccurateGpuEmulation", Settings::values.use_accurate_gpu_emulation);
+    LogSetting("Audio_OutputEngine", Settings::values.sink_id);
+    LogSetting("Audio_EnableAudioStretching", Settings::values.enable_audio_stretching);
+    LogSetting("Audio_OutputDevice", Settings::values.audio_device_id);
+    LogSetting("DataStorage_UseVirtualSd", Settings::values.use_virtual_sd);
+    LogSetting("DataStorage_NandDir", Settings::values.nand_dir);
+    LogSetting("DataStorage_SdmcDir", Settings::values.sdmc_dir);
+    LogSetting("Debugging_UseGdbstub", Settings::values.use_gdbstub);
+    LogSetting("Debugging_GdbstubPort", Settings::values.gdbstub_port);
+    LogSetting("Debugging_ProgramArgs", Settings::values.program_args);
+}
+
 } // namespace Settings

src/core/settings.h

@@ -6,6 +6,7 @@
 
 #include <array>
 #include <atomic>
+#include <chrono>
 #include <map>
 #include <optional>
 #include <string>
@@ -350,6 +351,11 @@ struct Values {
     bool use_docked_mode;
     bool enable_nfc;
     std::optional<u32> rng_seed;
+    // Measured in seconds since epoch
+    std::optional<std::chrono::seconds> custom_rtc;
+    // Set on game boot, reset on stop. Seconds difference between current time and `custom_rtc`
+    std::chrono::seconds custom_rtc_differential;
+
     s32 current_user;
     s32 language_index;
 
@@ -419,4 +425,5 @@ struct Values {
 } extern values;
 
 void Apply();
+void LogSettings();
 } // namespace Settings

src/video_core/CMakeLists.txt

@@ -30,6 +30,8 @@ add_library(video_core STATIC
     renderer_base.h
     renderer_opengl/gl_buffer_cache.cpp
     renderer_opengl/gl_buffer_cache.h
+    renderer_opengl/gl_global_cache.cpp
+    renderer_opengl/gl_global_cache.h
     renderer_opengl/gl_primitive_assembler.cpp
     renderer_opengl/gl_primitive_assembler.h
     renderer_opengl/gl_rasterizer.cpp
@@ -57,6 +59,34 @@ add_library(video_core STATIC
     renderer_opengl/renderer_opengl.h
     renderer_opengl/utils.cpp
     renderer_opengl/utils.h
+    shader/decode/arithmetic.cpp
+    shader/decode/arithmetic_immediate.cpp
+    shader/decode/bfe.cpp
+    shader/decode/bfi.cpp
+    shader/decode/shift.cpp
+    shader/decode/arithmetic_integer.cpp
+    shader/decode/arithmetic_integer_immediate.cpp
+    shader/decode/arithmetic_half.cpp
+    shader/decode/arithmetic_half_immediate.cpp
+    shader/decode/ffma.cpp
+    shader/decode/hfma2.cpp
+    shader/decode/conversion.cpp
+    shader/decode/memory.cpp
+    shader/decode/float_set_predicate.cpp
+    shader/decode/integer_set_predicate.cpp
+    shader/decode/half_set_predicate.cpp
+    shader/decode/predicate_set_register.cpp
+    shader/decode/predicate_set_predicate.cpp
+    shader/decode/register_set_predicate.cpp
+    shader/decode/float_set.cpp
+    shader/decode/integer_set.cpp
+    shader/decode/half_set.cpp
+    shader/decode/video.cpp
+    shader/decode/xmad.cpp
+    shader/decode/other.cpp
+    shader/decode.cpp
+    shader/shader_ir.cpp
+    shader/shader_ir.h
     surface.cpp
     surface.h
     textures/astc.cpp

src/video_core/engines/maxwell_3d.cpp

@@ -37,6 +37,7 @@ void Maxwell3D::InitializeRegisterDefaults() {
         regs.viewports[viewport].depth_range_near = 0.0f;
         regs.viewports[viewport].depth_range_far = 1.0f;
     }
+
     // Doom and Bomberman seems to use the uninitialized registers and just enable blend
     // so initialize blend registers with sane values
     regs.blend.equation_rgb = Regs::Blend::Equation::Add;
@@ -66,6 +67,7 @@ void Maxwell3D::InitializeRegisterDefaults() {
     regs.stencil_back_func_func = Regs::ComparisonOp::Always;
     regs.stencil_back_func_mask = 0xFFFFFFFF;
     regs.stencil_back_mask = 0xFFFFFFFF;
+
     // TODO(Rodrigo): Most games do not set a point size. I think this is a case of a
     // register carrying a default value. Assume it's OpenGL's default (1).
     regs.point_size = 1.0f;
@@ -78,6 +80,9 @@ void Maxwell3D::InitializeRegisterDefaults() {
         regs.color_mask[color_mask].B.Assign(1);
         regs.color_mask[color_mask].A.Assign(1);
     }
+
+    // Commercial games seem to assume this value is enabled and nouveau sets this value manually.
+    regs.rt_separate_frag_data = 1;
 }
 
 void Maxwell3D::CallMacroMethod(u32 method, std::vector<u32> parameters) {
@@ -135,6 +140,33 @@ void Maxwell3D::CallMethod(const GPU::MethodCall& method_call) {
 
     if (regs.reg_array[method_call.method] != method_call.argument) {
         regs.reg_array[method_call.method] = method_call.argument;
+        // Color buffers
+        constexpr u32 first_rt_reg = MAXWELL3D_REG_INDEX(rt);
+        constexpr u32 registers_per_rt = sizeof(regs.rt[0]) / sizeof(u32);
+        if (method_call.method >= first_rt_reg &&
+            method_call.method < first_rt_reg + registers_per_rt * Regs::NumRenderTargets) {
+            const std::size_t rt_index = (method_call.method - first_rt_reg) / registers_per_rt;
+            dirty_flags.color_buffer |= 1u << static_cast<u32>(rt_index);
+        }
+
+        // Zeta buffer
+        constexpr u32 registers_in_zeta = sizeof(regs.zeta) / sizeof(u32);
+        if (method_call.method == MAXWELL3D_REG_INDEX(zeta_enable) ||
+            method_call.method == MAXWELL3D_REG_INDEX(zeta_width) ||
+            method_call.method == MAXWELL3D_REG_INDEX(zeta_height) ||
+            (method_call.method >= MAXWELL3D_REG_INDEX(zeta) &&
+             method_call.method < MAXWELL3D_REG_INDEX(zeta) + registers_in_zeta)) {
+            dirty_flags.zeta_buffer = true;
+        }
+
+        // Shader
+        constexpr u32 shader_registers_count =
+            sizeof(regs.shader_config[0]) * Regs::MaxShaderProgram / sizeof(u32);
+        if (method_call.method >= MAXWELL3D_REG_INDEX(shader_config[0]) &&
+            method_call.method < MAXWELL3D_REG_INDEX(shader_config[0]) + shader_registers_count) {
+            dirty_flags.shaders = true;
+        }
+
         // Vertex format
         if (method_call.method >= MAXWELL3D_REG_INDEX(vertex_attrib_format) &&
             method_call.method <

src/video_core/engines/maxwell_3d.h

@@ -1089,10 +1089,18 @@ public:
     MemoryManager& memory_manager;
 
     struct DirtyFlags {
+        u8 color_buffer = 0xFF;
+        bool zeta_buffer = true;
+
+        bool shaders = true;
+
         bool vertex_attrib_format = true;
         u32 vertex_array = 0xFFFFFFFF;
 
         void OnMemoryWrite() {
+            color_buffer = 0xFF;
+            zeta_buffer = true;
+            shaders = true;
             vertex_array = 0xFFFFFFFF;
         }
     };

src/video_core/engines/shader_bytecode.h

@@ -397,6 +397,10 @@ struct IpaMode {
     bool operator!=(const IpaMode& a) const {
         return !operator==(a);
     }
+    bool operator<(const IpaMode& a) const {
+        return std::tie(interpolation_mode, sampling_mode) <
+               std::tie(a.interpolation_mode, a.sampling_mode);
+    }
 };
 
 enum class SystemVariable : u64 {
@@ -644,6 +648,7 @@ union Instruction {
             BitField<37, 2, HalfPrecision> precision;
             BitField<32, 1, u64> saturate;
 
+            BitField<31, 1, u64> negate_b;
             BitField<30, 1, u64> negate_c;
             BitField<35, 2, HalfType> type_c;
         } rr;
@@ -1431,6 +1436,7 @@ public:
         PredicateSetRegister,
         RegisterSetPredicate,
         Conversion,
+        Video,
         Xmad,
         Unknown,
     };
@@ -1562,8 +1568,8 @@ private:
             INST("11100000--------", Id::IPA, Type::Trivial, "IPA"),
             INST("1111101111100---", Id::OUT_R, Type::Trivial, "OUT_R"),
             INST("1110111111010---", Id::ISBERD, Type::Trivial, "ISBERD"),
-            INST("01011111--------", Id::VMAD, Type::Trivial, "VMAD"),
-            INST("0101000011110---", Id::VSETP, Type::Trivial, "VSETP"),
+            INST("01011111--------", Id::VMAD, Type::Video, "VMAD"),
+            INST("0101000011110---", Id::VSETP, Type::Video, "VSETP"),
             INST("0011001-1-------", Id::FFMA_IMM, Type::Ffma, "FFMA_IMM"),
             INST("010010011-------", Id::FFMA_CR, Type::Ffma, "FFMA_CR"),
             INST("010100011-------", Id::FFMA_RC, Type::Ffma, "FFMA_RC"),

src/video_core/engines/shader_header.h

@@ -106,7 +106,7 @@ struct Header {
         } ps;
     };
 
-    u64 GetLocalMemorySize() {
+    u64 GetLocalMemorySize() const {
         return (common1.shader_local_memory_low_size |
                 (common2.shader_local_memory_high_size << 24));
     }

src/video_core/rasterizer_interface.h

@@ -4,6 +4,7 @@
 
 #pragma once
 
+#include <functional>
 #include "common/common_types.h"
 #include "video_core/engines/fermi_2d.h"
 #include "video_core/gpu.h"
@@ -11,6 +12,14 @@
 
 namespace VideoCore {
 
+enum class LoadCallbackStage {
+    Prepare,
+    Decompile,
+    Build,
+    Complete,
+};
+using DiskResourceLoadCallback = std::function<void(LoadCallbackStage, std::size_t, std::size_t)>;
+
 class RasterizerInterface {
 public:
     virtual ~RasterizerInterface() {}

src/video_core/renderer_opengl/gl_buffer_cache.cpp

@@ -14,7 +14,7 @@
 namespace OpenGL {
 
 OGLBufferCache::OGLBufferCache(RasterizerOpenGL& rasterizer, std::size_t size)
-    : RasterizerCache{rasterizer}, stream_buffer(GL_ARRAY_BUFFER, size) {}
+    : RasterizerCache{rasterizer}, stream_buffer(size, true) {}
 
 GLintptr OGLBufferCache::UploadMemory(Tegra::GPUVAddr gpu_addr, std::size_t size,
                                       std::size_t alignment, bool cache) {

src/video_core/renderer_opengl/gl_global_cache.cpp

@@ -0,0 +1,24 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <glad/glad.h>
+
+#include "video_core/renderer_opengl/gl_global_cache.h"
+#include "video_core/renderer_opengl/gl_rasterizer.h"
+#include "video_core/renderer_opengl/utils.h"
+
+namespace OpenGL {
+
+CachedGlobalRegion::CachedGlobalRegion(VAddr addr, u32 size) : addr{addr}, size{size} {
+    buffer.Create();
+    // Bind and unbind the buffer so it gets allocated by the driver
+    glBindBuffer(GL_SHADER_STORAGE_BUFFER, buffer.handle);
+    glBindBuffer(GL_SHADER_STORAGE_BUFFER, 0);
+    LabelGLObject(GL_BUFFER, buffer.handle, addr, "GlobalMemory");
+}
+
+GlobalRegionCacheOpenGL::GlobalRegionCacheOpenGL(RasterizerOpenGL& rasterizer)
+    : RasterizerCache{rasterizer} {}
+
+} // namespace OpenGL

src/video_core/renderer_opengl/gl_global_cache.h

@@ -0,0 +1,60 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <memory>
+#include <glad/glad.h>
+
+#include "common/common_types.h"
+#include "video_core/rasterizer_cache.h"
+#include "video_core/renderer_opengl/gl_resource_manager.h"
+
+namespace OpenGL {
+
+namespace GLShader {
+class GlobalMemoryEntry;
+} // namespace GLShader
+
+class RasterizerOpenGL;
+class CachedGlobalRegion;
+using GlobalRegion = std::shared_ptr<CachedGlobalRegion>;
+
+class CachedGlobalRegion final : public RasterizerCacheObject {
+public:
+    explicit CachedGlobalRegion(VAddr addr, u32 size);
+
+    /// Gets the address of the shader in guest memory, required for cache management
+    VAddr GetAddr() const {
+        return addr;
+    }
+
+    /// Gets the size of the shader in guest memory, required for cache management
+    std::size_t GetSizeInBytes() const {
+        return size;
+    }
+
+    /// Gets the GL program handle for the buffer
+    GLuint GetBufferHandle() const {
+        return buffer.handle;
+    }
+
+    // TODO(Rodrigo): When global memory is written (STG), implement flushing
+    void Flush() override {
+        UNIMPLEMENTED();
+    }
+
+private:
+    VAddr addr{};
+    u32 size{};
+
+    OGLBuffer buffer;
+};
+
+class GlobalRegionCacheOpenGL final : public RasterizerCache<GlobalRegion> {
+public:
+    explicit GlobalRegionCacheOpenGL(RasterizerOpenGL& rasterizer);
+};
+
+} // namespace OpenGL

src/video_core/renderer_opengl/gl_rasterizer.cpp

@@ -101,7 +101,7 @@ struct FramebufferCacheKey {
 
 RasterizerOpenGL::RasterizerOpenGL(Core::Frontend::EmuWindow& window, ScreenInfo& info)
     : res_cache{*this}, shader_cache{*this}, emu_window{window}, screen_info{info},
-      buffer_cache(*this, STREAM_BUFFER_SIZE) {
+      buffer_cache(*this, STREAM_BUFFER_SIZE), global_cache{*this} {
     // Create sampler objects
     for (std::size_t i = 0; i < texture_samplers.size(); ++i) {
         texture_samplers[i].Create();
@@ -135,27 +135,31 @@ void RasterizerOpenGL::CheckExtensions() {
     }
 }
 
-void RasterizerOpenGL::SetupVertexFormat() {
+GLuint RasterizerOpenGL::SetupVertexFormat() {
     auto& gpu = Core::System::GetInstance().GPU().Maxwell3D();
     const auto& regs = gpu.regs;
 
-    if (!gpu.dirty_flags.vertex_attrib_format)
-        return;
+    if (!gpu.dirty_flags.vertex_attrib_format) {
+        return state.draw.vertex_array;
+    }
     gpu.dirty_flags.vertex_attrib_format = false;
 
     MICROPROFILE_SCOPE(OpenGL_VAO);
 
     auto [iter, is_cache_miss] = vertex_array_cache.try_emplace(regs.vertex_attrib_format);
-    auto& VAO = iter->second;
+    auto& vao_entry = iter->second;
 
     if (is_cache_miss) {
-        VAO.Create();
-        state.draw.vertex_array = VAO.handle;
-        state.ApplyVertexBufferState();
+        vao_entry.Create();
+        const GLuint vao = vao_entry.handle;
 
-        // The index buffer binding is stored within the VAO. Stupid OpenGL, but easy to work
-        // around.
-        glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, buffer_cache.GetHandle());
+        // Eventhough we are using DSA to create this vertex array, there is a bug on Intel's blob
+        // that fails to properly create the vertex array if it's not bound even after creating it
+        // with glCreateVertexArrays
+        state.draw.vertex_array = vao;
+        state.ApplyVertexArrayState();
+
+        glVertexArrayElementBuffer(vao, buffer_cache.GetHandle());
 
         // Use the vertex array as-is, assumes that the data is formatted correctly for OpenGL.
         // Enables the first 16 vertex attributes always, as we don't know which ones are actually
@@ -163,7 +167,7 @@ void RasterizerOpenGL::SetupVertexFormat() {
         // for now to avoid OpenGL errors.
         // TODO(Subv): Analyze the shader to identify which attributes are actually used and don't
         // assume every shader uses them all.
-        for (unsigned index = 0; index < 16; ++index) {
+        for (u32 index = 0; index < 16; ++index) {
             const auto& attrib = regs.vertex_attrib_format[index];
 
             // Ignore invalid attributes.
@@ -178,28 +182,29 @@ void RasterizerOpenGL::SetupVertexFormat() {
 
             ASSERT(buffer.IsEnabled());
 
-            glEnableVertexAttribArray(index);
+            glEnableVertexArrayAttrib(vao, index);
             if (attrib.type == Tegra::Engines::Maxwell3D::Regs::VertexAttribute::Type::SignedInt ||
                 attrib.type ==
                     Tegra::Engines::Maxwell3D::Regs::VertexAttribute::Type::UnsignedInt) {
-                glVertexAttribIFormat(index, attrib.ComponentCount(),
-                                      MaxwellToGL::VertexType(attrib), attrib.offset);
+                glVertexArrayAttribIFormat(vao, index, attrib.ComponentCount(),
+                                           MaxwellToGL::VertexType(attrib), attrib.offset);
             } else {
-                glVertexAttribFormat(index, attrib.ComponentCount(),
-                                     MaxwellToGL::VertexType(attrib),
-                                     attrib.IsNormalized() ? GL_TRUE : GL_FALSE, attrib.offset);
+                glVertexArrayAttribFormat(
+                    vao, index, attrib.ComponentCount(), MaxwellToGL::VertexType(attrib),
+                    attrib.IsNormalized() ? GL_TRUE : GL_FALSE, attrib.offset);
             }
-            glVertexAttribBinding(index, attrib.buffer);
+            glVertexArrayAttribBinding(vao, index, attrib.buffer);
         }
     }
-    state.draw.vertex_array = VAO.handle;
-    state.ApplyVertexBufferState();
 
     // Rebinding the VAO invalidates the vertex buffer bindings.
     gpu.dirty_flags.vertex_array = 0xFFFFFFFF;
+
+    state.draw.vertex_array = vao_entry.handle;
+    return vao_entry.handle;
 }
 
-void RasterizerOpenGL::SetupVertexBuffer() {
+void RasterizerOpenGL::SetupVertexBuffer(GLuint vao) {
     auto& gpu = Core::System::GetInstance().GPU().Maxwell3D();
     const auto& regs = gpu.regs;
 
@@ -217,7 +222,7 @@ void RasterizerOpenGL::SetupVertexBuffer() {
         if (!vertex_array.IsEnabled())
             continue;
 
-        Tegra::GPUVAddr start = vertex_array.StartAddress();
+        const Tegra::GPUVAddr start = vertex_array.StartAddress();
         const Tegra::GPUVAddr end = regs.vertex_array_limit[index].LimitAddress();
 
         ASSERT(end > start);
@@ -225,21 +230,18 @@ void RasterizerOpenGL::SetupVertexBuffer() {
         const GLintptr vertex_buffer_offset = buffer_cache.UploadMemory(start, size);
 
         // Bind the vertex array to the buffer at the current offset.
-        glBindVertexBuffer(index, buffer_cache.GetHandle(), vertex_buffer_offset,
-                           vertex_array.stride);
+        glVertexArrayVertexBuffer(vao, index, buffer_cache.GetHandle(), vertex_buffer_offset,
+                                  vertex_array.stride);
 
         if (regs.instanced_arrays.IsInstancingEnabled(index) && vertex_array.divisor != 0) {
             // Enable vertex buffer instancing with the specified divisor.
-            glVertexBindingDivisor(index, vertex_array.divisor);
+            glVertexArrayBindingDivisor(vao, index, vertex_array.divisor);
         } else {
             // Disable the vertex buffer instancing.
-            glVertexBindingDivisor(index, 0);
+            glVertexArrayBindingDivisor(vao, index, 0);
         }
     }
 
-    // Implicit set by glBindVertexBuffer. Stupid glstate handling...
-    state.draw.vertex_buffer = buffer_cache.GetHandle();
-
     gpu.dirty_flags.vertex_array = 0;
 }
 
@@ -293,7 +295,7 @@ DrawParameters RasterizerOpenGL::SetupDraw() {
 
 void RasterizerOpenGL::SetupShaders(GLenum primitive_mode) {
     MICROPROFILE_SCOPE(OpenGL_Shader);
-    const auto& gpu = Core::System::GetInstance().GPU().Maxwell3D();
+    auto& gpu = Core::System::GetInstance().GPU().Maxwell3D();
 
     // Next available bindpoints to use when uploading the const buffers and textures to the GLSL
     // shaders. The constbuffer bindpoint starts after the shader stage configuration bind points.
@@ -365,7 +367,7 @@ void RasterizerOpenGL::SetupShaders(GLenum primitive_mode) {
         // (sometimes it's half the screen, sometimes three quarters). To avoid this, enable the
         // clip distances only when it's written by a shader stage.
         for (std::size_t i = 0; i < Maxwell::NumClipDistances; ++i) {
-            clip_distances[i] |= shader->GetShaderEntries().clip_distances[i];
+            clip_distances[i] = clip_distances[i] || shader->GetShaderEntries().clip_distances[i];
         }
 
         // When VertexA is enabled, we have dual vertex shaders
@@ -376,6 +378,8 @@ void RasterizerOpenGL::SetupShaders(GLenum primitive_mode) {
     }
 
     SyncClipEnabled(clip_distances);
+
+    gpu.dirty_flags.shaders = false;
 }
 
 void RasterizerOpenGL::SetupCachedFramebuffer(const FramebufferCacheKey& fbkey,
@@ -486,17 +490,26 @@ void RasterizerOpenGL::ConfigureFramebuffers(OpenGLState& current_state, bool us
                                              bool using_depth_fb, bool preserve_contents,
                                              std::optional<std::size_t> single_color_target) {
     MICROPROFILE_SCOPE(OpenGL_Framebuffer);
-    const auto& regs = Core::System::GetInstance().GPU().Maxwell3D().regs;
+    const auto& gpu = Core::System::GetInstance().GPU().Maxwell3D();
+    const auto& regs = gpu.regs;
+
+    const FramebufferConfigState fb_config_state{using_color_fb, using_depth_fb, preserve_contents,
+                                                 single_color_target};
+    if (fb_config_state == current_framebuffer_config_state && gpu.dirty_flags.color_buffer == 0 &&
+        !gpu.dirty_flags.zeta_buffer) {
+        // Only skip if the previous ConfigureFramebuffers call was from the same kind (multiple or
+        // single color targets). This is done because the guest registers may not change but the
+        // host framebuffer may contain different attachments
+        return;
+    }
+    current_framebuffer_config_state = fb_config_state;
 
     Surface depth_surface;
     if (using_depth_fb) {
         depth_surface = res_cache.GetDepthBufferSurface(preserve_contents);
     }
 
-    // TODO(bunnei): Figure out how the below register works. According to envytools, this should be
-    // used to enable multiple render targets. However, it is left unset on all games that I have
-    // tested.
-    UNIMPLEMENTED_IF(regs.rt_separate_frag_data != 0);
+    UNIMPLEMENTED_IF(regs.rt_separate_frag_data == 0);
 
     // Bind the framebuffer surfaces
     current_state.framebuffer_srgb.enabled = regs.framebuffer_srgb != 0;
@@ -630,8 +643,6 @@ void RasterizerOpenGL::Clear() {
         return;
     }
 
-    ScopeAcquireGLContext acquire_context{emu_window};
-
     ConfigureFramebuffers(clear_state, use_color, use_depth || use_stencil, false,
                           regs.clear_buffers.RT.Value());
     if (regs.clear_flags.scissor) {
@@ -665,8 +676,6 @@ void RasterizerOpenGL::DrawArrays() {
     auto& gpu = Core::System::GetInstance().GPU().Maxwell3D();
     const auto& regs = gpu.regs;
 
-    ScopeAcquireGLContext acquire_context{emu_window};
-
     ConfigureFramebuffers(state);
     SyncColorMask();
     SyncFragmentColorClampState();
@@ -689,9 +698,6 @@ void RasterizerOpenGL::DrawArrays() {
     // Draw the vertex batch
     const bool is_indexed = accelerate_draw == AccelDraw::Indexed;
 
-    state.draw.vertex_buffer = buffer_cache.GetHandle();
-    state.ApplyVertexBufferState();
-
     std::size_t buffer_size = CalculateVertexArraysSize();
 
     // Add space for index buffer (keeping in mind non-core primitives)
@@ -721,8 +727,9 @@ void RasterizerOpenGL::DrawArrays() {
         gpu.dirty_flags.vertex_array = 0xFFFFFFFF;
     }
 
-    SetupVertexFormat();
-    SetupVertexBuffer();
+    const GLuint vao = SetupVertexFormat();
+    SetupVertexBuffer(vao);
+
     DrawParameters params = SetupDraw();
     SetupShaders(params.primitive_mode);
 
@@ -761,6 +768,7 @@ void RasterizerOpenGL::InvalidateRegion(VAddr addr, u64 size) {
     MICROPROFILE_SCOPE(OpenGL_CacheManagement);
     res_cache.InvalidateRegion(addr, size);
     shader_cache.InvalidateRegion(addr, size);
+    global_cache.InvalidateRegion(addr, size);
     buffer_cache.InvalidateRegion(addr, size);
 }
 
@@ -922,7 +930,7 @@ u32 RasterizerOpenGL::SetupConstBuffers(Maxwell::ShaderStage stage, Shader& shad
     const auto& gpu = Core::System::GetInstance().GPU();
     const auto& maxwell3d = gpu.Maxwell3D();
     const auto& shader_stage = maxwell3d.state.shader_stages[static_cast<std::size_t>(stage)];
-    const auto& entries = shader->GetShaderEntries().const_buffer_entries;
+    const auto& entries = shader->GetShaderEntries().const_buffers;
 
     constexpr u64 max_binds = Tegra::Engines::Maxwell3D::Regs::MaxConstBuffers;
     std::array<GLuint, max_binds> bind_buffers;
@@ -990,7 +998,7 @@ u32 RasterizerOpenGL::SetupTextures(Maxwell::ShaderStage stage, Shader& shader,
     MICROPROFILE_SCOPE(OpenGL_Texture);
     const auto& gpu = Core::System::GetInstance().GPU();
     const auto& maxwell3d = gpu.Maxwell3D();
-    const auto& entries = shader->GetShaderEntries().texture_samplers;
+    const auto& entries = shader->GetShaderEntries().samplers;
 
     ASSERT_MSG(current_unit + entries.size() <= std::size(state.texture_units),
                "Exceeded the number of active textures.");

src/video_core/renderer_opengl/gl_rasterizer.h

@@ -23,6 +23,7 @@
 #include "video_core/rasterizer_cache.h"
 #include "video_core/rasterizer_interface.h"
 #include "video_core/renderer_opengl/gl_buffer_cache.h"
+#include "video_core/renderer_opengl/gl_global_cache.h"
 #include "video_core/renderer_opengl/gl_primitive_assembler.h"
 #include "video_core/renderer_opengl/gl_rasterizer_cache.h"
 #include "video_core/renderer_opengl/gl_resource_manager.h"
@@ -66,6 +67,10 @@ public:
     static_assert(MaxConstbufferSize % sizeof(GLvec4) == 0,
                   "The maximum size of a constbuffer must be a multiple of the size of GLvec4");
 
+    static constexpr std::size_t MaxGlobalMemorySize = 0x10000;
+    static_assert(MaxGlobalMemorySize % sizeof(float) == 0,
+                  "The maximum size of a global memory must be a multiple of the size of float");
+
 private:
     class SamplerInfo {
     public:
@@ -94,6 +99,23 @@ private:
         float max_anisotropic = 1.0f;
     };
 
+    struct FramebufferConfigState {
+        bool using_color_fb{};
+        bool using_depth_fb{};
+        bool preserve_contents{};
+        std::optional<std::size_t> single_color_target;
+
+        bool operator==(const FramebufferConfigState& rhs) const {
+            return std::tie(using_color_fb, using_depth_fb, preserve_contents,
+                            single_color_target) == std::tie(rhs.using_color_fb, rhs.using_depth_fb,
+                                                             rhs.preserve_contents,
+                                                             rhs.single_color_target);
+        }
+        bool operator!=(const FramebufferConfigState& rhs) const {
+            return !operator==(rhs);
+        }
+    };
+
     /**
      * Configures the color and depth framebuffer states.
      * @param use_color_fb If true, configure color framebuffers.
@@ -105,7 +127,7 @@ private:
                                bool using_depth_fb = true, bool preserve_contents = true,
                                std::optional<std::size_t> single_color_target = {});
 
-    /*
+    /**
      * Configures the current constbuffers to use for the draw command.
      * @param stage The shader stage to configure buffers for.
      * @param shader The shader object that contains the specified stage.
@@ -115,7 +137,7 @@ private:
     u32 SetupConstBuffers(Tegra::Engines::Maxwell3D::Regs::ShaderStage stage, Shader& shader,
                           GLenum primitive_mode, u32 current_bindpoint);
 
-    /*
+    /**
      * Configures the current textures to use for the draw command.
      * @param stage The shader stage to configure textures for.
      * @param shader The shader object that contains the specified stage.
@@ -185,6 +207,7 @@ private:
 
     RasterizerCacheOpenGL res_cache;
     ShaderCacheOpenGL shader_cache;
+    GlobalRegionCacheOpenGL global_cache;
 
     Core::Frontend::EmuWindow& emu_window;
 
@@ -197,6 +220,7 @@ private:
         vertex_array_cache;
 
     std::map<FramebufferCacheKey, OGLFramebuffer> framebuffer_cache;
+    FramebufferConfigState current_framebuffer_config_state;
 
     std::array<SamplerInfo, Tegra::Engines::Maxwell3D::Regs::NumTextureSamplers> texture_samplers;
 
@@ -209,8 +233,10 @@ private:
 
     std::size_t CalculateIndexBufferSize() const;
 
-    void SetupVertexFormat();
-    void SetupVertexBuffer();
+    /// Updates and returns a vertex array object representing current vertex format
+    GLuint SetupVertexFormat();
+
+    void SetupVertexBuffer(GLuint vao);
 
     DrawParameters SetupDraw();
 

src/video_core/renderer_opengl/gl_rasterizer_cache.cpp

@@ -167,6 +167,7 @@ std::size_t SurfaceParams::InnerMemorySize(bool force_gl, bool layer_only,
     }
 
     params.is_layered = SurfaceTargetIsLayered(params.target);
+    params.is_array = SurfaceTargetIsArray(params.target);
     params.max_mip_level = config.tic.max_mip_level + 1;
     params.rt = {};
 
@@ -452,7 +453,7 @@ static void CopySurface(const Surface& src_surface, const Surface& dst_surface,
     const std::size_t buffer_size = std::max(src_params.size_in_bytes, dst_params.size_in_bytes);
 
     glBindBuffer(GL_PIXEL_PACK_BUFFER, copy_pbo_handle);
-    glBufferData(GL_PIXEL_PACK_BUFFER, buffer_size, nullptr, GL_STREAM_DRAW);
+    glBufferData(GL_PIXEL_PACK_BUFFER, buffer_size, nullptr, GL_STREAM_COPY);
     if (source_format.compressed) {
         glGetCompressedTextureImage(src_surface->Texture().handle, src_attachment,
                                     static_cast<GLsizei>(src_params.size_in_bytes), nullptr);
@@ -877,10 +878,13 @@ void CachedSurface::EnsureTextureView() {
     UNIMPLEMENTED_IF(gl_is_compressed);
 
     const GLenum target{TargetLayer()};
+    const GLuint num_layers{target == GL_TEXTURE_CUBE_MAP_ARRAY ? 6u : 1u};
+    constexpr GLuint min_layer = 0;
+    constexpr GLuint min_level = 0;
 
     texture_view.Create();
-    glTextureView(texture_view.handle, target, texture.handle, gl_internal_format, 0,
-                  params.max_mip_level, 0, 1);
+    glTextureView(texture_view.handle, target, texture.handle, gl_internal_format, min_level,
+                  params.max_mip_level, min_layer, num_layers);
 
     OpenGLState cur_state = OpenGLState::GetCurState();
     const auto& old_tex = cur_state.texture_units[0];
@@ -919,9 +923,16 @@ Surface RasterizerCacheOpenGL::GetTextureSurface(const Tegra::Texture::FullTextu
 }
 
 Surface RasterizerCacheOpenGL::GetDepthBufferSurface(bool preserve_contents) {
-    const auto& regs{Core::System::GetInstance().GPU().Maxwell3D().regs};
+    auto& gpu{Core::System::GetInstance().GPU().Maxwell3D()};
+    const auto& regs{gpu.regs};
+
+    if (!gpu.dirty_flags.zeta_buffer) {
+        return last_depth_buffer;
+    }
+    gpu.dirty_flags.zeta_buffer = false;
+
     if (!regs.zeta.Address() || !regs.zeta_enable) {
-        return {};
+        return last_depth_buffer = {};
     }
 
     SurfaceParams depth_params{SurfaceParams::CreateForDepthBuffer(
@@ -929,25 +940,31 @@ Surface RasterizerCacheOpenGL::GetDepthBufferSurface(bool preserve_contents) {
         regs.zeta.memory_layout.block_width, regs.zeta.memory_layout.block_height,
         regs.zeta.memory_layout.block_depth, regs.zeta.memory_layout.type)};
 
-    return GetSurface(depth_params, preserve_contents);
+    return last_depth_buffer = GetSurface(depth_params, preserve_contents);
 }
 
 Surface RasterizerCacheOpenGL::GetColorBufferSurface(std::size_t index, bool preserve_contents) {
-    const auto& regs{Core::System::GetInstance().GPU().Maxwell3D().regs};
+    auto& gpu{Core::System::GetInstance().GPU().Maxwell3D()};
+    const auto& regs{gpu.regs};
+
+    if ((gpu.dirty_flags.color_buffer & (1u << static_cast<u32>(index))) == 0) {
+        return last_color_buffers[index];
+    }
+    gpu.dirty_flags.color_buffer &= ~(1u << static_cast<u32>(index));
 
     ASSERT(index < Tegra::Engines::Maxwell3D::Regs::NumRenderTargets);
 
     if (index >= regs.rt_control.count) {
-        return {};
+        return last_color_buffers[index] = {};
     }
 
     if (regs.rt[index].Address() == 0 || regs.rt[index].format == Tegra::RenderTargetFormat::NONE) {
-        return {};
+        return last_color_buffers[index] = {};
     }
 
     const SurfaceParams color_params{SurfaceParams::CreateForFramebuffer(index)};
 
-    return GetSurface(color_params, preserve_contents);
+    return last_color_buffers[index] = GetSurface(color_params, preserve_contents);
 }
 
 void RasterizerCacheOpenGL::LoadSurface(const Surface& surface) {

src/video_core/renderer_opengl/gl_rasterizer_cache.h

@@ -225,6 +225,7 @@ struct SurfaceParams {
     SurfaceTarget target;
     u32 max_mip_level;
     bool is_layered;
+    bool is_array;
     bool srgb_conversion;
     // Parameters used for caching
     VAddr addr;
@@ -294,7 +295,7 @@ public:
     }
 
     const OGLTexture& TextureLayer() {
-        if (params.is_layered) {
+        if (params.is_array) {
             return Texture();
         }
         EnsureTextureView();
@@ -396,6 +397,9 @@ private:
     /// Use a Pixel Buffer Object to download the previous texture and then upload it to the new one
     /// using the new format.
     OGLBuffer copy_pbo;
+
+    std::array<Surface, Tegra::Engines::Maxwell3D::Regs::NumRenderTargets> last_color_buffers;
+    Surface last_depth_buffer;
 };
 
 } // namespace OpenGL

src/video_core/renderer_opengl/gl_resource_manager.cpp

@@ -117,7 +117,7 @@ void OGLBuffer::Create() {
         return;
 
     MICROPROFILE_SCOPE(OpenGL_ResourceCreation);
-    glGenBuffers(1, &handle);
+    glCreateBuffers(1, &handle);
 }
 
 void OGLBuffer::Release() {
@@ -126,7 +126,6 @@ void OGLBuffer::Release() {
 
     MICROPROFILE_SCOPE(OpenGL_ResourceDeletion);
     glDeleteBuffers(1, &handle);
-    OpenGLState::GetCurState().ResetBuffer(handle).Apply();
     handle = 0;
 }
 
@@ -152,7 +151,7 @@ void OGLVertexArray::Create() {
         return;
 
     MICROPROFILE_SCOPE(OpenGL_ResourceCreation);
-    glGenVertexArrays(1, &handle);
+    glCreateVertexArrays(1, &handle);
 }
 
 void OGLVertexArray::Release() {

src/video_core/renderer_opengl/gl_shader_cache.cpp

@@ -10,11 +10,15 @@
 #include "video_core/engines/maxwell_3d.h"
 #include "video_core/renderer_opengl/gl_rasterizer.h"
 #include "video_core/renderer_opengl/gl_shader_cache.h"
+#include "video_core/renderer_opengl/gl_shader_decompiler.h"
 #include "video_core/renderer_opengl/gl_shader_manager.h"
 #include "video_core/renderer_opengl/utils.h"
+#include "video_core/shader/shader_ir.h"
 
 namespace OpenGL {
 
+using VideoCommon::Shader::ProgramCode;
+
 /// Gets the address for the specified shader stage program
 static VAddr GetShaderAddress(Maxwell::ShaderProgram program) {
     const auto& gpu = Core::System::GetInstance().GPU().Maxwell3D();
@@ -24,8 +28,8 @@ static VAddr GetShaderAddress(Maxwell::ShaderProgram program) {
 }
 
 /// Gets the shader program code from memory for the specified address
-static GLShader::ProgramCode GetShaderCode(VAddr addr) {
-    GLShader::ProgramCode program_code(GLShader::MAX_PROGRAM_CODE_LENGTH);
+static ProgramCode GetShaderCode(VAddr addr) {
+    ProgramCode program_code(VideoCommon::Shader::MAX_PROGRAM_LENGTH);
     Memory::ReadBlock(addr, program_code.data(), program_code.size() * sizeof(u64));
     return program_code;
 }
@@ -188,6 +192,10 @@ void CachedShader::CalculateProperties() {
 ShaderCacheOpenGL::ShaderCacheOpenGL(RasterizerOpenGL& rasterizer) : RasterizerCache{rasterizer} {}
 
 Shader ShaderCacheOpenGL::GetStageProgram(Maxwell::ShaderProgram program) {
+    if (!Core::System::GetInstance().GPU().Maxwell3D().dirty_flags.shaders) {
+        return last_shaders[static_cast<u32>(program)];
+    }
+
     const VAddr program_addr{GetShaderAddress(program)};
 
     // Look up shader in the cache based on address
@@ -199,7 +207,7 @@ Shader ShaderCacheOpenGL::GetStageProgram(Maxwell::ShaderProgram program) {
         Register(shader);
     }
 
-    return shader;
+    return last_shaders[static_cast<u32>(program)] = shader;
 }
 
 } // namespace OpenGL

src/video_core/renderer_opengl/gl_shader_cache.h

@@ -4,6 +4,7 @@
 
 #pragma once
 
+#include <array>
 #include <map>
 #include <memory>
 
@@ -11,6 +12,7 @@
 #include "common/common_types.h"
 #include "video_core/rasterizer_cache.h"
 #include "video_core/renderer_opengl/gl_resource_manager.h"
+#include "video_core/renderer_opengl/gl_shader_decompiler.h"
 #include "video_core/renderer_opengl/gl_shader_gen.h"
 
 namespace OpenGL {
@@ -115,6 +117,9 @@ public:
 
     /// Gets the current specified shader stage program
     Shader GetStageProgram(Maxwell::ShaderProgram program);
+
+private:
+    std::array<Shader, Maxwell::MaxShaderProgram> last_shaders;
 };
 
 } // namespace OpenGL

src/video_core/renderer_opengl/gl_shader_decompiler.h

@@ -5,21 +5,84 @@
 #pragma once
 
 #include <array>
-#include <functional>
-#include <optional>
 #include <string>
+#include <utility>
+#include <vector>
 #include "common/common_types.h"
 #include "video_core/engines/maxwell_3d.h"
-#include "video_core/renderer_opengl/gl_shader_gen.h"
+#include "video_core/shader/shader_ir.h"
 
-namespace OpenGL::GLShader::Decompiler {
+namespace VideoCommon::Shader {
+class ShaderIR;
+}
 
-using Tegra::Engines::Maxwell3D;
+namespace OpenGL::GLShader {
+
+using Maxwell = Tegra::Engines::Maxwell3D::Regs;
+
+class ConstBufferEntry : public VideoCommon::Shader::ConstBuffer {
+public:
+    explicit ConstBufferEntry(const VideoCommon::Shader::ConstBuffer& entry,
+                              Maxwell::ShaderStage stage, const std::string& name, u32 index)
+        : VideoCommon::Shader::ConstBuffer{entry}, stage{stage}, name{name}, index{index} {}
+
+    const std::string& GetName() const {
+        return name;
+    }
+
+    Maxwell::ShaderStage GetStage() const {
+        return stage;
+    }
+
+    u32 GetIndex() const {
+        return index;
+    }
+
+    u32 GetHash() const {
+        return (static_cast<u32>(stage) << 16) | index;
+    }
+
+private:
+    std::string name;
+    Maxwell::ShaderStage stage{};
+    u32 index{};
+};
+
+class SamplerEntry : public VideoCommon::Shader::Sampler {
+public:
+    explicit SamplerEntry(const VideoCommon::Shader::Sampler& entry, Maxwell::ShaderStage stage,
+                          const std::string& name)
+        : VideoCommon::Shader::Sampler{entry}, stage{stage}, name{name} {}
+
+    const std::string& GetName() const {
+        return name;
+    }
+
+    Maxwell::ShaderStage GetStage() const {
+        return stage;
+    }
+
+    u32 GetHash() const {
+        return (static_cast<u32>(stage) << 16) | static_cast<u32>(GetIndex());
+    }
+
+private:
+    std::string name;
+    Maxwell::ShaderStage stage{};
+};
+
+struct ShaderEntries {
+    std::vector<ConstBufferEntry> const_buffers;
+    std::vector<SamplerEntry> samplers;
+    std::array<bool, Maxwell::NumClipDistances> clip_distances{};
+    std::size_t shader_length{};
+};
+
+using ProgramResult = std::pair<std::string, ShaderEntries>;
 
 std::string GetCommonDeclarations();
 
-std::optional<ProgramResult> DecompileProgram(const ProgramCode& program_code, u32 main_offset,
-                                              Maxwell3D::Regs::ShaderStage stage,
-                                              const std::string& suffix);
+ProgramResult Decompile(const VideoCommon::Shader::ShaderIR& ir, Maxwell::ShaderStage stage,
+                        const std::string& suffix);
 
-} // namespace OpenGL::GLShader::Decompiler
+} // namespace OpenGL::GLShader

src/video_core/renderer_opengl/gl_shader_gen.cpp

@@ -7,22 +7,25 @@
 #include "video_core/engines/maxwell_3d.h"
 #include "video_core/renderer_opengl/gl_shader_decompiler.h"
 #include "video_core/renderer_opengl/gl_shader_gen.h"
+#include "video_core/shader/shader_ir.h"
 
 namespace OpenGL::GLShader {
 
 using Tegra::Engines::Maxwell3D;
+using VideoCommon::Shader::ProgramCode;
+using VideoCommon::Shader::ShaderIR;
 
 static constexpr u32 PROGRAM_OFFSET{10};
 
 ProgramResult GenerateVertexShader(const ShaderSetup& setup) {
+    const std::string id = fmt::format("{:016x}", setup.program.unique_identifier);
+
     std::string out = "#version 430 core\n";
     out += "#extension GL_ARB_separate_shader_objects : enable\n\n";
-    const std::string id = fmt::format("{:016x}", setup.program.unique_identifier);
     out += "// Shader Unique Id: VS" + id + "\n\n";
-    out += Decompiler::GetCommonDeclarations();
+    out += GetCommonDeclarations();
 
     out += R"(
-
 layout (location = 0) out vec4 position;
 
 layout(std140) uniform vs_config {
@@ -30,40 +33,32 @@ layout(std140) uniform vs_config {
     uvec4 config_pack; // instance_id, flip_stage, y_direction, padding
     uvec4 alpha_test;
 };
+
 )";
-
-    if (setup.IsDualProgram()) {
-        out += "bool exec_vertex_b();\n";
-    }
-
-    ProgramResult program =
-        Decompiler::DecompileProgram(setup.program.code, PROGRAM_OFFSET,
-                                     Maxwell3D::Regs::ShaderStage::Vertex, "vertex")
-            .value_or(ProgramResult());
+    ShaderIR program_ir(setup.program.code, PROGRAM_OFFSET);
+    ProgramResult program = Decompile(program_ir, Maxwell3D::Regs::ShaderStage::Vertex, "vertex");
 
     out += program.first;
 
     if (setup.IsDualProgram()) {
+        ShaderIR program_ir_b(setup.program.code_b, PROGRAM_OFFSET);
         ProgramResult program_b =
-            Decompiler::DecompileProgram(setup.program.code_b, PROGRAM_OFFSET,
-                                         Maxwell3D::Regs::ShaderStage::Vertex, "vertex_b")
-                .value_or(ProgramResult());
+            Decompile(program_ir_b, Maxwell3D::Regs::ShaderStage::Vertex, "vertex_b");
+
         out += program_b.first;
     }
 
     out += R"(
-
 void main() {
     position = vec4(0.0, 0.0, 0.0, 0.0);
-    exec_vertex();
+    execute_vertex();
 )";
 
     if (setup.IsDualProgram()) {
-        out += "    exec_vertex_b();";
+        out += "    execute_vertex_b();";
     }
 
     out += R"(
-
     // Check if the flip stage is VertexB
     // Config pack's second value is flip_stage
     if (config_pack[1] == 1) {
@@ -77,30 +72,20 @@ void main() {
     if (config_pack[1] == 1) {
         position.w = 1.0;
     }
-}
-
-)";
+})";
 
     return {out, program.second};
 }
 
 ProgramResult GenerateGeometryShader(const ShaderSetup& setup) {
     // Version is intentionally skipped in shader generation, it's added by the lazy compilation.
-    std::string out = "#extension GL_ARB_separate_shader_objects : enable\n\n";
     const std::string id = fmt::format("{:016x}", setup.program.unique_identifier);
+
+    std::string out = "#extension GL_ARB_separate_shader_objects : enable\n\n";
     out += "// Shader Unique Id: GS" + id + "\n\n";
-    out += Decompiler::GetCommonDeclarations();
-    out += "bool exec_geometry();\n";
+    out += GetCommonDeclarations();
 
-    ProgramResult program =
-        Decompiler::DecompileProgram(setup.program.code, PROGRAM_OFFSET,
-                                     Maxwell3D::Regs::ShaderStage::Geometry, "geometry")
-            .value_or(ProgramResult());
     out += R"(
-out gl_PerVertex {
-    vec4 gl_Position;
-};
-
 layout (location = 0) in vec4 gs_position[];
 layout (location = 0) out vec4 position;
 
@@ -110,36 +95,37 @@ layout (std140) uniform gs_config {
     uvec4 alpha_test;
 };
 
-void main() {
-    exec_geometry();
-}
-
 )";
+    ShaderIR program_ir(setup.program.code, PROGRAM_OFFSET);
+    ProgramResult program =
+        Decompile(program_ir, Maxwell3D::Regs::ShaderStage::Geometry, "geometry");
     out += program.first;
+
+    out += R"(
+void main() {
+    execute_geometry();
+};)";
+
     return {out, program.second};
 }
 
 ProgramResult GenerateFragmentShader(const ShaderSetup& setup) {
+    const std::string id = fmt::format("{:016x}", setup.program.unique_identifier);
+
     std::string out = "#version 430 core\n";
     out += "#extension GL_ARB_separate_shader_objects : enable\n\n";
-    const std::string id = fmt::format("{:016x}", setup.program.unique_identifier);
     out += "// Shader Unique Id: FS" + id + "\n\n";
-    out += Decompiler::GetCommonDeclarations();
-    out += "bool exec_fragment();\n";
+    out += GetCommonDeclarations();
 
-    ProgramResult program =
-        Decompiler::DecompileProgram(setup.program.code, PROGRAM_OFFSET,
-                                     Maxwell3D::Regs::ShaderStage::Fragment, "fragment")
-            .value_or(ProgramResult());
     out += R"(
-layout(location = 0) out vec4 FragColor0;
-layout(location = 1) out vec4 FragColor1;
-layout(location = 2) out vec4 FragColor2;
-layout(location = 3) out vec4 FragColor3;
-layout(location = 4) out vec4 FragColor4;
-layout(location = 5) out vec4 FragColor5;
-layout(location = 6) out vec4 FragColor6;
-layout(location = 7) out vec4 FragColor7;
+layout (location = 0) out vec4 FragColor0;
+layout (location = 1) out vec4 FragColor1;
+layout (location = 2) out vec4 FragColor2;
+layout (location = 3) out vec4 FragColor3;
+layout (location = 4) out vec4 FragColor4;
+layout (location = 5) out vec4 FragColor5;
+layout (location = 6) out vec4 FragColor6;
+layout (location = 7) out vec4 FragColor7;
 
 layout (location = 0) in vec4 position;
 
@@ -173,12 +159,20 @@ bool AlphaFunc(in float value) {
     }
 }
 
+)";
+    ShaderIR program_ir(setup.program.code, PROGRAM_OFFSET);
+    ProgramResult program =
+        Decompile(program_ir, Maxwell3D::Regs::ShaderStage::Fragment, "fragment");
+
+    out += program.first;
+
+    out += R"(
 void main() {
-    exec_fragment();
+    execute_fragment();
 }
 
 )";
-    out += program.first;
     return {out, program.second};
 }
-} // namespace OpenGL::GLShader
+
+} // namespace OpenGL::GLShader

src/video_core/renderer_opengl/gl_shader_gen.h

@@ -10,164 +10,12 @@
 
 #include "common/common_types.h"
 #include "video_core/engines/shader_bytecode.h"
+#include "video_core/renderer_opengl/gl_shader_decompiler.h"
+#include "video_core/shader/shader_ir.h"
 
 namespace OpenGL::GLShader {
 
-constexpr std::size_t MAX_PROGRAM_CODE_LENGTH{0x1000};
-using ProgramCode = std::vector<u64>;
-
-enum : u32 { POSITION_VARYING_LOCATION = 0, GENERIC_VARYING_START_LOCATION = 1 };
-
-class ConstBufferEntry {
-    using Maxwell = Tegra::Engines::Maxwell3D::Regs;
-
-public:
-    void MarkAsUsed(u64 index, u64 offset, Maxwell::ShaderStage stage) {
-        is_used = true;
-        this->index = static_cast<unsigned>(index);
-        this->stage = stage;
-        max_offset = std::max(max_offset, static_cast<unsigned>(offset));
-    }
-
-    void MarkAsUsedIndirect(u64 index, Maxwell::ShaderStage stage) {
-        is_used = true;
-        is_indirect = true;
-        this->index = static_cast<unsigned>(index);
-        this->stage = stage;
-    }
-
-    bool IsUsed() const {
-        return is_used;
-    }
-
-    bool IsIndirect() const {
-        return is_indirect;
-    }
-
-    unsigned GetIndex() const {
-        return index;
-    }
-
-    unsigned GetSize() const {
-        return max_offset + 1;
-    }
-
-    std::string GetName() const {
-        return BufferBaseNames[static_cast<std::size_t>(stage)] + std::to_string(index);
-    }
-
-    u32 GetHash() const {
-        return (static_cast<u32>(stage) << 16) | index;
-    }
-
-private:
-    static constexpr std::array<const char*, Maxwell::MaxShaderStage> BufferBaseNames = {
-        "buffer_vs_c", "buffer_tessc_c", "buffer_tesse_c", "buffer_gs_c", "buffer_fs_c",
-    };
-
-    bool is_used{};
-    bool is_indirect{};
-    unsigned index{};
-    unsigned max_offset{};
-    Maxwell::ShaderStage stage;
-};
-
-class SamplerEntry {
-    using Maxwell = Tegra::Engines::Maxwell3D::Regs;
-
-public:
-    SamplerEntry(Maxwell::ShaderStage stage, std::size_t offset, std::size_t index,
-                 Tegra::Shader::TextureType type, bool is_array, bool is_shadow)
-        : offset(offset), stage(stage), sampler_index(index), type(type), is_array(is_array),
-          is_shadow(is_shadow) {}
-
-    std::size_t GetOffset() const {
-        return offset;
-    }
-
-    std::size_t GetIndex() const {
-        return sampler_index;
-    }
-
-    Maxwell::ShaderStage GetStage() const {
-        return stage;
-    }
-
-    std::string GetName() const {
-        return std::string(TextureSamplerNames[static_cast<std::size_t>(stage)]) + '_' +
-               std::to_string(sampler_index);
-    }
-
-    std::string GetTypeString() const {
-        using Tegra::Shader::TextureType;
-        std::string glsl_type;
-
-        switch (type) {
-        case TextureType::Texture1D:
-            glsl_type = "sampler1D";
-            break;
-        case TextureType::Texture2D:
-            glsl_type = "sampler2D";
-            break;
-        case TextureType::Texture3D:
-            glsl_type = "sampler3D";
-            break;
-        case TextureType::TextureCube:
-            glsl_type = "samplerCube";
-            break;
-        default:
-            UNIMPLEMENTED();
-        }
-        if (is_array)
-            glsl_type += "Array";
-        if (is_shadow)
-            glsl_type += "Shadow";
-        return glsl_type;
-    }
-
-    Tegra::Shader::TextureType GetType() const {
-        return type;
-    }
-
-    bool IsArray() const {
-        return is_array;
-    }
-
-    bool IsShadow() const {
-        return is_shadow;
-    }
-
-    u32 GetHash() const {
-        return (static_cast<u32>(stage) << 16) | static_cast<u32>(sampler_index);
-    }
-
-    static std::string GetArrayName(Maxwell::ShaderStage stage) {
-        return TextureSamplerNames[static_cast<std::size_t>(stage)];
-    }
-
-private:
-    static constexpr std::array<const char*, Maxwell::MaxShaderStage> TextureSamplerNames = {
-        "tex_vs", "tex_tessc", "tex_tesse", "tex_gs", "tex_fs",
-    };
-
-    /// Offset in TSC memory from which to read the sampler object, as specified by the sampling
-    /// instruction.
-    std::size_t offset;
-    Maxwell::ShaderStage stage;      ///< Shader stage where this sampler was used.
-    std::size_t sampler_index;       ///< Value used to index into the generated GLSL sampler array.
-    Tegra::Shader::TextureType type; ///< The type used to sample this texture (Texture2D, etc)
-    bool is_array;  ///< Whether the texture is being sampled as an array texture or not.
-    bool is_shadow; ///< Whether the texture is being sampled as a depth texture or not.
-};
-
-struct ShaderEntries {
-    std::vector<ConstBufferEntry> const_buffer_entries;
-    std::vector<SamplerEntry> texture_samplers;
-    std::array<bool, Tegra::Engines::Maxwell3D::Regs::NumClipDistances> clip_distances;
-    std::size_t shader_length;
-};
-
-using ProgramResult = std::pair<std::string, ShaderEntries>;
+using VideoCommon::Shader::ProgramCode;
 
 struct ShaderSetup {
     explicit ShaderSetup(ProgramCode program_code) {

src/video_core/renderer_opengl/gl_state.cpp

@@ -83,8 +83,6 @@ OpenGLState::OpenGLState() {
     draw.read_framebuffer = 0;
     draw.draw_framebuffer = 0;
     draw.vertex_array = 0;
-    draw.vertex_buffer = 0;
-    draw.uniform_buffer = 0;
     draw.shader_program = 0;
     draw.program_pipeline = 0;
 
@@ -505,7 +503,6 @@ void OpenGLState::ApplySamplers() const {
 }
 
 void OpenGLState::ApplyFramebufferState() const {
-    // Framebuffer
     if (draw.read_framebuffer != cur_state.draw.read_framebuffer) {
         glBindFramebuffer(GL_READ_FRAMEBUFFER, draw.read_framebuffer);
     }
@@ -514,16 +511,10 @@ void OpenGLState::ApplyFramebufferState() const {
     }
 }
 
-void OpenGLState::ApplyVertexBufferState() const {
-    // Vertex array
+void OpenGLState::ApplyVertexArrayState() const {
     if (draw.vertex_array != cur_state.draw.vertex_array) {
         glBindVertexArray(draw.vertex_array);
     }
-
-    // Vertex buffer
-    if (draw.vertex_buffer != cur_state.draw.vertex_buffer) {
-        glBindBuffer(GL_ARRAY_BUFFER, draw.vertex_buffer);
-    }
 }
 
 void OpenGLState::ApplyDepthClamp() const {
@@ -543,11 +534,7 @@ void OpenGLState::ApplyDepthClamp() const {
 
 void OpenGLState::Apply() const {
     ApplyFramebufferState();
-    ApplyVertexBufferState();
-    // Uniform buffer
-    if (draw.uniform_buffer != cur_state.draw.uniform_buffer) {
-        glBindBuffer(GL_UNIFORM_BUFFER, draw.uniform_buffer);
-    }
+    ApplyVertexArrayState();
 
     // Shader program
     if (draw.shader_program != cur_state.draw.shader_program) {
@@ -638,16 +625,6 @@ OpenGLState& OpenGLState::ResetPipeline(GLuint handle) {
     return *this;
 }
 
-OpenGLState& OpenGLState::ResetBuffer(GLuint handle) {
-    if (draw.vertex_buffer == handle) {
-        draw.vertex_buffer = 0;
-    }
-    if (draw.uniform_buffer == handle) {
-        draw.uniform_buffer = 0;
-    }
-    return *this;
-}
-
 OpenGLState& OpenGLState::ResetVertexArray(GLuint handle) {
     if (draw.vertex_array == handle) {
         draw.vertex_array = 0;

src/video_core/renderer_opengl/gl_state.h

@@ -154,8 +154,6 @@ public:
         GLuint read_framebuffer; // GL_READ_FRAMEBUFFER_BINDING
         GLuint draw_framebuffer; // GL_DRAW_FRAMEBUFFER_BINDING
         GLuint vertex_array;     // GL_VERTEX_ARRAY_BINDING
-        GLuint vertex_buffer;    // GL_ARRAY_BUFFER_BINDING
-        GLuint uniform_buffer;   // GL_UNIFORM_BUFFER_BINDING
         GLuint shader_program;   // GL_CURRENT_PROGRAM
         GLuint program_pipeline; // GL_PROGRAM_PIPELINE_BINDING
     } draw;
@@ -206,10 +204,10 @@ public:
     }
     /// Apply this state as the current OpenGL state
     void Apply() const;
-    /// Apply only the state afecting the framebuffer
+    /// Apply only the state affecting the framebuffer
     void ApplyFramebufferState() const;
-    /// Apply only the state afecting the vertex buffer
-    void ApplyVertexBufferState() const;
+    /// Apply only the state affecting the vertex array
+    void ApplyVertexArrayState() const;
     /// Set the initial OpenGL state
     static void ApplyDefaultState();
     /// Resets any references to the given resource
@@ -217,7 +215,6 @@ public:
     OpenGLState& ResetSampler(GLuint handle);
     OpenGLState& ResetProgram(GLuint handle);
     OpenGLState& ResetPipeline(GLuint handle);
-    OpenGLState& ResetBuffer(GLuint handle);
     OpenGLState& ResetVertexArray(GLuint handle);
     OpenGLState& ResetFramebuffer(GLuint handle);
     void EmulateViewportWithScissor();

src/video_core/renderer_opengl/gl_stream_buffer.cpp

@@ -15,13 +15,12 @@ MICROPROFILE_DEFINE(OpenGL_StreamBuffer, "OpenGL", "Stream Buffer Orphaning",
 
 namespace OpenGL {
 
-OGLStreamBuffer::OGLStreamBuffer(GLenum target, GLsizeiptr size, bool prefer_coherent)
-    : gl_target(target), buffer_size(size) {
+OGLStreamBuffer::OGLStreamBuffer(GLsizeiptr size, bool vertex_data_usage, bool prefer_coherent)
+    : buffer_size(size) {
     gl_buffer.Create();
-    glBindBuffer(gl_target, gl_buffer.handle);
 
     GLsizeiptr allocate_size = size;
-    if (target == GL_ARRAY_BUFFER) {
+    if (vertex_data_usage) {
         // On AMD GPU there is a strange crash in indexed drawing. The crash happens when the buffer
         // read position is near the end and is an out-of-bound access to the vertex buffer. This is
         // probably a bug in the driver and is related to the usage of vec3<byte> attributes in the
@@ -35,18 +34,17 @@ OGLStreamBuffer::OGLStreamBuffer(GLenum target, GLsizeiptr size, bool prefer_coh
         coherent = prefer_coherent;
         const GLbitfield flags =
             GL_MAP_WRITE_BIT | GL_MAP_PERSISTENT_BIT | (coherent ? GL_MAP_COHERENT_BIT : 0);
-        glBufferStorage(gl_target, allocate_size, nullptr, flags);
-        mapped_ptr = static_cast<u8*>(glMapBufferRange(
-            gl_target, 0, buffer_size, flags | (coherent ? 0 : GL_MAP_FLUSH_EXPLICIT_BIT)));
+        glNamedBufferStorage(gl_buffer.handle, allocate_size, nullptr, flags);
+        mapped_ptr = static_cast<u8*>(glMapNamedBufferRange(
+            gl_buffer.handle, 0, buffer_size, flags | (coherent ? 0 : GL_MAP_FLUSH_EXPLICIT_BIT)));
     } else {
-        glBufferData(gl_target, allocate_size, nullptr, GL_STREAM_DRAW);
+        glNamedBufferData(gl_buffer.handle, allocate_size, nullptr, GL_STREAM_DRAW);
     }
 }
 
 OGLStreamBuffer::~OGLStreamBuffer() {
     if (persistent) {
-        glBindBuffer(gl_target, gl_buffer.handle);
-        glUnmapBuffer(gl_target);
+        glUnmapNamedBuffer(gl_buffer.handle);
     }
     gl_buffer.Release();
 }
@@ -74,7 +72,7 @@ std::tuple<u8*, GLintptr, bool> OGLStreamBuffer::Map(GLsizeiptr size, GLintptr a
         invalidate = true;
 
         if (persistent) {
-            glUnmapBuffer(gl_target);
+            glUnmapNamedBuffer(gl_buffer.handle);
         }
     }
 
@@ -84,7 +82,7 @@ std::tuple<u8*, GLintptr, bool> OGLStreamBuffer::Map(GLsizeiptr size, GLintptr a
                            (coherent ? GL_MAP_COHERENT_BIT : GL_MAP_FLUSH_EXPLICIT_BIT) |
                            (invalidate ? GL_MAP_INVALIDATE_BUFFER_BIT : GL_MAP_UNSYNCHRONIZED_BIT);
         mapped_ptr = static_cast<u8*>(
-            glMapBufferRange(gl_target, buffer_pos, buffer_size - buffer_pos, flags));
+            glMapNamedBufferRange(gl_buffer.handle, buffer_pos, buffer_size - buffer_pos, flags));
         mapped_offset = buffer_pos;
     }
 
@@ -95,11 +93,11 @@ void OGLStreamBuffer::Unmap(GLsizeiptr size) {
     ASSERT(size <= mapped_size);
 
     if (!coherent && size > 0) {
-        glFlushMappedBufferRange(gl_target, buffer_pos - mapped_offset, size);
+        glFlushMappedNamedBufferRange(gl_buffer.handle, buffer_pos - mapped_offset, size);
     }
 
     if (!persistent) {
-        glUnmapBuffer(gl_target);
+        glUnmapNamedBuffer(gl_buffer.handle);
     }
 
     buffer_pos += size;

src/video_core/renderer_opengl/gl_stream_buffer.h

@@ -13,7 +13,7 @@ namespace OpenGL {
 
 class OGLStreamBuffer : private NonCopyable {
 public:
-    explicit OGLStreamBuffer(GLenum target, GLsizeiptr size, bool prefer_coherent = false);
+    explicit OGLStreamBuffer(GLsizeiptr size, bool vertex_data_usage, bool prefer_coherent = false);
     ~OGLStreamBuffer();
 
     GLuint GetHandle() const;
@@ -33,7 +33,6 @@ public:
 
 private:
     OGLBuffer gl_buffer;
-    GLenum gl_target;
 
     bool coherent = false;
     bool persistent = false;

src/video_core/renderer_opengl/renderer_opengl.cpp

@@ -14,6 +14,7 @@
 #include "core/core.h"
 #include "core/core_timing.h"
 #include "core/frontend/emu_window.h"
+#include "core/frontend/scope_acquire_window_context.h"
 #include "core/memory.h"
 #include "core/perf_stats.h"
 #include "core/settings.h"
@@ -97,18 +98,6 @@ static std::array<GLfloat, 3 * 2> MakeOrthographicMatrix(const float width, cons
     return matrix;
 }
 
-ScopeAcquireGLContext::ScopeAcquireGLContext(Core::Frontend::EmuWindow& emu_window_)
-    : emu_window{emu_window_} {
-    if (Settings::values.use_multi_core) {
-        emu_window.MakeCurrent();
-    }
-}
-ScopeAcquireGLContext::~ScopeAcquireGLContext() {
-    if (Settings::values.use_multi_core) {
-        emu_window.DoneCurrent();
-    }
-}
-
 RendererOpenGL::RendererOpenGL(Core::Frontend::EmuWindow& window)
     : VideoCore::RendererBase{window} {}
 
@@ -117,7 +106,6 @@ RendererOpenGL::~RendererOpenGL() = default;
 /// Swap buffers (render frame)
 void RendererOpenGL::SwapBuffers(
     std::optional<std::reference_wrapper<const Tegra::FramebufferConfig>> framebuffer) {
-    ScopeAcquireGLContext acquire_context{render_window};
 
     Core::System::GetInstance().GetPerfStats().EndSystemFrame();
 
@@ -245,20 +233,20 @@ void RendererOpenGL::InitOpenGLObjects() {
 
     // Generate VAO
     vertex_array.Create();
-
     state.draw.vertex_array = vertex_array.handle;
-    state.draw.vertex_buffer = vertex_buffer.handle;
-    state.draw.uniform_buffer = 0;
-    state.Apply();
 
     // Attach vertex data to VAO
-    glBufferData(GL_ARRAY_BUFFER, sizeof(ScreenRectVertex) * 4, nullptr, GL_STREAM_DRAW);
-    glVertexAttribPointer(attrib_position, 2, GL_FLOAT, GL_FALSE, sizeof(ScreenRectVertex),
-                          (GLvoid*)offsetof(ScreenRectVertex, position));
-    glVertexAttribPointer(attrib_tex_coord, 2, GL_FLOAT, GL_FALSE, sizeof(ScreenRectVertex),
-                          (GLvoid*)offsetof(ScreenRectVertex, tex_coord));
-    glEnableVertexAttribArray(attrib_position);
-    glEnableVertexAttribArray(attrib_tex_coord);
+    glNamedBufferData(vertex_buffer.handle, sizeof(ScreenRectVertex) * 4, nullptr, GL_STREAM_DRAW);
+    glVertexArrayAttribFormat(vertex_array.handle, attrib_position, 2, GL_FLOAT, GL_FALSE,
+                              offsetof(ScreenRectVertex, position));
+    glVertexArrayAttribFormat(vertex_array.handle, attrib_tex_coord, 2, GL_FLOAT, GL_FALSE,
+                              offsetof(ScreenRectVertex, tex_coord));
+    glVertexArrayAttribBinding(vertex_array.handle, attrib_position, 0);
+    glVertexArrayAttribBinding(vertex_array.handle, attrib_tex_coord, 0);
+    glEnableVertexArrayAttrib(vertex_array.handle, attrib_position);
+    glEnableVertexArrayAttrib(vertex_array.handle, attrib_tex_coord);
+    glVertexArrayVertexBuffer(vertex_array.handle, 0, vertex_buffer.handle, 0,
+                              sizeof(ScreenRectVertex));
 
     // Allocate textures for the screen
     screen_info.texture.resource.Create();
@@ -370,14 +358,12 @@ void RendererOpenGL::DrawScreenTriangles(const ScreenInfo& screen_info, float x,
     state.texture_units[0].texture = screen_info.display_texture;
     state.texture_units[0].swizzle = {GL_RED, GL_GREEN, GL_BLUE, GL_ALPHA};
     // Workaround brigthness problems in SMO by enabling sRGB in the final output
-    // if it has been used in the frame
-    // Needed because of this bug in QT
-    // QTBUG-50987
+    // if it has been used in the frame. Needed because of this bug in QT: QTBUG-50987
     state.framebuffer_srgb.enabled = OpenGLState::GetsRGBUsed();
     state.Apply();
-    glBufferSubData(GL_ARRAY_BUFFER, 0, sizeof(vertices), vertices.data());
+    glNamedBufferSubData(vertex_buffer.handle, 0, sizeof(vertices), vertices.data());
     glDrawArrays(GL_TRIANGLE_STRIP, 0, 4);
-    // restore default state
+    // Restore default state
     state.framebuffer_srgb.enabled = false;
     state.texture_units[0].texture = 0;
     state.Apply();
@@ -508,7 +494,7 @@ static void APIENTRY DebugHandler(GLenum source, GLenum type, GLuint id, GLenum
 
 /// Initialize the renderer
 bool RendererOpenGL::Init() {
-    ScopeAcquireGLContext acquire_context{render_window};
+    Core::Frontend::ScopeAcquireWindowContext acquire_context{render_window};
 
     if (GLAD_GL_KHR_debug) {
         glEnable(GL_DEBUG_OUTPUT);

src/video_core/renderer_opengl/renderer_opengl.h

@@ -39,16 +39,6 @@ struct ScreenInfo {
     TextureInfo texture;
 };
 
-/// Helper class to acquire/release OpenGL context within a given scope
-class ScopeAcquireGLContext : NonCopyable {
-public:
-    explicit ScopeAcquireGLContext(Core::Frontend::EmuWindow& window);
-    ~ScopeAcquireGLContext();
-
-private:
-    Core::Frontend::EmuWindow& emu_window;
-};
-
 class RendererOpenGL : public VideoCore::RendererBase {
 public:
     explicit RendererOpenGL(Core::Frontend::EmuWindow& window);

src/video_core/shader/decode.cpp

@@ -0,0 +1,206 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <cstring>
+#include <set>
+
+#include <fmt/format.h>
+
+#include "common/assert.h"
+#include "common/common_types.h"
+#include "video_core/engines/shader_bytecode.h"
+#include "video_core/engines/shader_header.h"
+#include "video_core/shader/shader_ir.h"
+
+namespace VideoCommon::Shader {
+
+using Tegra::Shader::Instruction;
+using Tegra::Shader::OpCode;
+
+namespace {
+
+/// Merges exit method of two parallel branches.
+constexpr ExitMethod ParallelExit(ExitMethod a, ExitMethod b) {
+    if (a == ExitMethod::Undetermined) {
+        return b;
+    }
+    if (b == ExitMethod::Undetermined) {
+        return a;
+    }
+    if (a == b) {
+        return a;
+    }
+    return ExitMethod::Conditional;
+}
+
+/**
+ * Returns whether the instruction at the specified offset is a 'sched' instruction.
+ * Sched instructions always appear before a sequence of 3 instructions.
+ */
+constexpr bool IsSchedInstruction(u32 offset, u32 main_offset) {
+    constexpr u32 SchedPeriod = 4;
+    u32 absolute_offset = offset - main_offset;
+
+    return (absolute_offset % SchedPeriod) == 0;
+}
+
+} // namespace
+
+void ShaderIR::Decode() {
+    std::memcpy(&header, program_code.data(), sizeof(Tegra::Shader::Header));
+
+    std::set<u32> labels;
+    const ExitMethod exit_method = Scan(main_offset, MAX_PROGRAM_LENGTH, labels);
+    if (exit_method != ExitMethod::AlwaysEnd) {
+        UNREACHABLE_MSG("Program does not always end");
+    }
+
+    if (labels.empty()) {
+        basic_blocks.insert({main_offset, DecodeRange(main_offset, MAX_PROGRAM_LENGTH)});
+        return;
+    }
+
+    labels.insert(main_offset);
+
+    for (const u32 label : labels) {
+        const auto next_it = labels.lower_bound(label + 1);
+        const u32 next_label = next_it == labels.end() ? MAX_PROGRAM_LENGTH : *next_it;
+
+        basic_blocks.insert({label, DecodeRange(label, next_label)});
+    }
+}
+
+ExitMethod ShaderIR::Scan(u32 begin, u32 end, std::set<u32>& labels) {
+    const auto [iter, inserted] =
+        exit_method_map.emplace(std::make_pair(begin, end), ExitMethod::Undetermined);
+    ExitMethod& exit_method = iter->second;
+    if (!inserted)
+        return exit_method;
+
+    for (u32 offset = begin; offset != end && offset != MAX_PROGRAM_LENGTH; ++offset) {
+        coverage_begin = std::min(coverage_begin, offset);
+        coverage_end = std::max(coverage_end, offset + 1);
+
+        const Instruction instr = {program_code[offset]};
+        const auto opcode = OpCode::Decode(instr);
+        if (!opcode)
+            continue;
+        switch (opcode->get().GetId()) {
+        case OpCode::Id::EXIT: {
+            // The EXIT instruction can be predicated, which means that the shader can conditionally
+            // end on this instruction. We have to consider the case where the condition is not met
+            // and check the exit method of that other basic block.
+            using Tegra::Shader::Pred;
+            if (instr.pred.pred_index == static_cast<u64>(Pred::UnusedIndex)) {
+                return exit_method = ExitMethod::AlwaysEnd;
+            } else {
+                const ExitMethod not_met = Scan(offset + 1, end, labels);
+                return exit_method = ParallelExit(ExitMethod::AlwaysEnd, not_met);
+            }
+        }
+        case OpCode::Id::BRA: {
+            const u32 target = offset + instr.bra.GetBranchTarget();
+            labels.insert(target);
+            const ExitMethod no_jmp = Scan(offset + 1, end, labels);
+            const ExitMethod jmp = Scan(target, end, labels);
+            return exit_method = ParallelExit(no_jmp, jmp);
+        }
+        case OpCode::Id::SSY:
+        case OpCode::Id::PBK: {
+            // The SSY and PBK use a similar encoding as the BRA instruction.
+            UNIMPLEMENTED_IF_MSG(instr.bra.constant_buffer != 0,
+                                 "Constant buffer branching is not supported");
+            const u32 target = offset + instr.bra.GetBranchTarget();
+            labels.insert(target);
+            // Continue scanning for an exit method.
+            break;
+        }
+        }
+    }
+    return exit_method = ExitMethod::AlwaysReturn;
+}
+
+BasicBlock ShaderIR::DecodeRange(u32 begin, u32 end) {
+    BasicBlock basic_block;
+    for (u32 pc = begin; pc < (begin > end ? MAX_PROGRAM_LENGTH : end);) {
+        pc = DecodeInstr(basic_block, pc);
+    }
+    return basic_block;
+}
+
+u32 ShaderIR::DecodeInstr(BasicBlock& bb, u32 pc) {
+    // Ignore sched instructions when generating code.
+    if (IsSchedInstruction(pc, main_offset)) {
+        return pc + 1;
+    }
+
+    const Instruction instr = {program_code[pc]};
+    const auto opcode = OpCode::Decode(instr);
+
+    // Decoding failure
+    if (!opcode) {
+        UNIMPLEMENTED_MSG("Unhandled instruction: {0:x}", instr.value);
+        return pc + 1;
+    }
+
+    bb.push_back(
+        Comment(fmt::format("{}: {} (0x{:016x})", pc, opcode->get().GetName(), instr.value)));
+
+    using Tegra::Shader::Pred;
+    UNIMPLEMENTED_IF_MSG(instr.pred.full_pred == Pred::NeverExecute,
+                         "NeverExecute predicate not implemented");
+
+    static const std::map<OpCode::Type, u32 (ShaderIR::*)(BasicBlock&, const BasicBlock&, u32)>
+        decoders = {
+            {OpCode::Type::Arithmetic, &ShaderIR::DecodeArithmetic},
+            {OpCode::Type::ArithmeticImmediate, &ShaderIR::DecodeArithmeticImmediate},
+            {OpCode::Type::Bfe, &ShaderIR::DecodeBfe},
+            {OpCode::Type::Bfi, &ShaderIR::DecodeBfi},
+            {OpCode::Type::Shift, &ShaderIR::DecodeShift},
+            {OpCode::Type::ArithmeticInteger, &ShaderIR::DecodeArithmeticInteger},
+            {OpCode::Type::ArithmeticIntegerImmediate, &ShaderIR::DecodeArithmeticIntegerImmediate},
+            {OpCode::Type::ArithmeticHalf, &ShaderIR::DecodeArithmeticHalf},
+            {OpCode::Type::ArithmeticHalfImmediate, &ShaderIR::DecodeArithmeticHalfImmediate},
+            {OpCode::Type::Ffma, &ShaderIR::DecodeFfma},
+            {OpCode::Type::Hfma2, &ShaderIR::DecodeHfma2},
+            {OpCode::Type::Conversion, &ShaderIR::DecodeConversion},
+            {OpCode::Type::Memory, &ShaderIR::DecodeMemory},
+            {OpCode::Type::FloatSetPredicate, &ShaderIR::DecodeFloatSetPredicate},
+            {OpCode::Type::IntegerSetPredicate, &ShaderIR::DecodeIntegerSetPredicate},
+            {OpCode::Type::HalfSetPredicate, &ShaderIR::DecodeHalfSetPredicate},
+            {OpCode::Type::PredicateSetRegister, &ShaderIR::DecodePredicateSetRegister},
+            {OpCode::Type::PredicateSetPredicate, &ShaderIR::DecodePredicateSetPredicate},
+            {OpCode::Type::RegisterSetPredicate, &ShaderIR::DecodeRegisterSetPredicate},
+            {OpCode::Type::FloatSet, &ShaderIR::DecodeFloatSet},
+            {OpCode::Type::IntegerSet, &ShaderIR::DecodeIntegerSet},
+            {OpCode::Type::HalfSet, &ShaderIR::DecodeHalfSet},
+            {OpCode::Type::Video, &ShaderIR::DecodeVideo},
+            {OpCode::Type::Xmad, &ShaderIR::DecodeXmad},
+        };
+
+    std::vector<Node> tmp_block;
+    if (const auto decoder = decoders.find(opcode->get().GetType()); decoder != decoders.end()) {
+        pc = (this->*decoder->second)(tmp_block, bb, pc);
+    } else {
+        pc = DecodeOther(tmp_block, bb, pc);
+    }
+
+    // Some instructions (like SSY) don't have a predicate field, they are always unconditionally
+    // executed.
+    const bool can_be_predicated = OpCode::IsPredicatedInstruction(opcode->get().GetId());
+    const auto pred_index = static_cast<u32>(instr.pred.pred_index);
+
+    if (can_be_predicated && pred_index != static_cast<u32>(Pred::UnusedIndex)) {
+        bb.push_back(
+            Conditional(GetPredicate(pred_index, instr.negate_pred != 0), std::move(tmp_block)));
+    } else {
+        for (auto& node : tmp_block) {
+            bb.push_back(std::move(node));
+        }
+    }
+
+    return pc + 1;
+}
+
+} // namespace VideoCommon::Shader

src/video_core/shader/decode/arithmetic.cpp

@@ -0,0 +1,155 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/assert.h"
+#include "common/common_types.h"
+#include "video_core/engines/shader_bytecode.h"
+#include "video_core/shader/shader_ir.h"
+
+namespace VideoCommon::Shader {
+
+using Tegra::Shader::Instruction;
+using Tegra::Shader::OpCode;
+using Tegra::Shader::SubOp;
+
+u32 ShaderIR::DecodeArithmetic(BasicBlock& bb, const BasicBlock& code, u32 pc) {
+    const Instruction instr = {program_code[pc]};
+    const auto opcode = OpCode::Decode(instr);
+
+    Node op_a = GetRegister(instr.gpr8);
+
+    Node op_b = [&]() -> Node {
+        if (instr.is_b_imm) {
+            return GetImmediate19(instr);
+        } else if (instr.is_b_gpr) {
+            return GetRegister(instr.gpr20);
+        } else {
+            return GetConstBuffer(instr.cbuf34.index, instr.cbuf34.offset);
+        }
+    }();
+
+    switch (opcode->get().GetId()) {
+    case OpCode::Id::MOV_C:
+    case OpCode::Id::MOV_R: {
+        // MOV does not have neither 'abs' nor 'neg' bits.
+        SetRegister(bb, instr.gpr0, op_b);
+        break;
+    }
+    case OpCode::Id::FMUL_C:
+    case OpCode::Id::FMUL_R:
+    case OpCode::Id::FMUL_IMM: {
+        // FMUL does not have 'abs' bits and only the second operand has a 'neg' bit.
+        UNIMPLEMENTED_IF_MSG(instr.fmul.tab5cb8_2 != 0, "FMUL tab5cb8_2({}) is not implemented",
+                             instr.fmul.tab5cb8_2.Value());
+        UNIMPLEMENTED_IF_MSG(
+            instr.fmul.tab5c68_0 != 1, "FMUL tab5cb8_0({}) is not implemented",
+            instr.fmul.tab5c68_0.Value()); // SMO typical sends 1 here which seems to be the default
+
+        op_b = GetOperandAbsNegFloat(op_b, false, instr.fmul.negate_b);
+
+        // TODO(Rodrigo): Should precise be used when there's a postfactor?
+        Node value = Operation(OperationCode::FMul, PRECISE, op_a, op_b);
+
+        if (instr.fmul.postfactor != 0) {
+            auto postfactor = static_cast<s32>(instr.fmul.postfactor);
+
+            // Postfactor encoded as 3-bit 1's complement in instruction, interpreted with below
+            // logic.
+            if (postfactor >= 4) {
+                postfactor = 7 - postfactor;
+            } else {
+                postfactor = 0 - postfactor;
+            }
+
+            if (postfactor > 0) {
+                value = Operation(OperationCode::FMul, NO_PRECISE, value,
+                                  Immediate(static_cast<f32>(1 << postfactor)));
+            } else {
+                value = Operation(OperationCode::FDiv, NO_PRECISE, value,
+                                  Immediate(static_cast<f32>(1 << -postfactor)));
+            }
+        }
+
+        value = GetSaturatedFloat(value, instr.alu.saturate_d);
+
+        SetInternalFlagsFromFloat(bb, value, instr.generates_cc);
+        SetRegister(bb, instr.gpr0, value);
+        break;
+    }
+    case OpCode::Id::FADD_C:
+    case OpCode::Id::FADD_R:
+    case OpCode::Id::FADD_IMM: {
+        op_a = GetOperandAbsNegFloat(op_a, instr.alu.abs_a, instr.alu.negate_a);
+        op_b = GetOperandAbsNegFloat(op_b, instr.alu.abs_b, instr.alu.negate_b);
+
+        Node value = Operation(OperationCode::FAdd, PRECISE, op_a, op_b);
+        value = GetSaturatedFloat(value, instr.alu.saturate_d);
+
+        SetInternalFlagsFromFloat(bb, value, instr.generates_cc);
+        SetRegister(bb, instr.gpr0, value);
+        break;
+    }
+    case OpCode::Id::MUFU: {
+        op_a = GetOperandAbsNegFloat(op_a, instr.alu.abs_a, instr.alu.negate_a);
+
+        Node value = [&]() {
+            switch (instr.sub_op) {
+            case SubOp::Cos:
+                return Operation(OperationCode::FCos, PRECISE, op_a);
+            case SubOp::Sin:
+                return Operation(OperationCode::FSin, PRECISE, op_a);
+            case SubOp::Ex2:
+                return Operation(OperationCode::FExp2, PRECISE, op_a);
+            case SubOp::Lg2:
+                return Operation(OperationCode::FLog2, PRECISE, op_a);
+            case SubOp::Rcp:
+                return Operation(OperationCode::FDiv, PRECISE, Immediate(1.0f), op_a);
+            case SubOp::Rsq:
+                return Operation(OperationCode::FInverseSqrt, PRECISE, op_a);
+            case SubOp::Sqrt:
+                return Operation(OperationCode::FSqrt, PRECISE, op_a);
+            default:
+                UNIMPLEMENTED_MSG("Unhandled MUFU sub op={0:x}",
+                                  static_cast<unsigned>(instr.sub_op.Value()));
+                return Immediate(0);
+            }
+        }();
+        value = GetSaturatedFloat(value, instr.alu.saturate_d);
+
+        SetRegister(bb, instr.gpr0, value);
+        break;
+    }
+    case OpCode::Id::FMNMX_C:
+    case OpCode::Id::FMNMX_R:
+    case OpCode::Id::FMNMX_IMM: {
+        op_a = GetOperandAbsNegFloat(op_a, instr.alu.abs_a, instr.alu.negate_a);
+        op_b = GetOperandAbsNegFloat(op_b, instr.alu.abs_b, instr.alu.negate_b);
+
+        const Node condition = GetPredicate(instr.alu.fmnmx.pred, instr.alu.fmnmx.negate_pred != 0);
+
+        const Node min = Operation(OperationCode::FMin, NO_PRECISE, op_a, op_b);
+        const Node max = Operation(OperationCode::FMax, NO_PRECISE, op_a, op_b);
+        const Node value = Operation(OperationCode::Select, NO_PRECISE, condition, min, max);
+
+        SetInternalFlagsFromFloat(bb, value, instr.generates_cc);
+        SetRegister(bb, instr.gpr0, value);
+        break;
+    }
+    case OpCode::Id::RRO_C:
+    case OpCode::Id::RRO_R:
+    case OpCode::Id::RRO_IMM: {
+        // Currently RRO is only implemented as a register move.
+        op_b = GetOperandAbsNegFloat(op_b, instr.alu.abs_b, instr.alu.negate_b);
+        SetRegister(bb, instr.gpr0, op_b);
+        LOG_WARNING(HW_GPU, "RRO instruction is incomplete");
+        break;
+    }
+    default:
+        UNIMPLEMENTED_MSG("Unhandled arithmetic instruction: {}", opcode->get().GetName());
+    }
+
+    return pc;
+}
+
+} // namespace VideoCommon::Shader

src/video_core/shader/decode/arithmetic_half.cpp

@@ -0,0 +1,70 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/assert.h"
+#include "common/common_types.h"
+#include "video_core/engines/shader_bytecode.h"
+#include "video_core/shader/shader_ir.h"
+
+namespace VideoCommon::Shader {
+
+using Tegra::Shader::Instruction;
+using Tegra::Shader::OpCode;
+
+u32 ShaderIR::DecodeArithmeticHalf(BasicBlock& bb, const BasicBlock& code, u32 pc) {
+    const Instruction instr = {program_code[pc]};
+    const auto opcode = OpCode::Decode(instr);
+
+    if (opcode->get().GetId() == OpCode::Id::HADD2_C ||
+        opcode->get().GetId() == OpCode::Id::HADD2_R) {
+        UNIMPLEMENTED_IF(instr.alu_half.ftz != 0);
+    }
+    UNIMPLEMENTED_IF_MSG(instr.alu_half.saturate != 0, "Half float saturation not implemented");
+
+    const bool negate_a =
+        opcode->get().GetId() != OpCode::Id::HMUL2_R && instr.alu_half.negate_a != 0;
+    const bool negate_b =
+        opcode->get().GetId() != OpCode::Id::HMUL2_C && instr.alu_half.negate_b != 0;
+
+    const Node op_a = GetOperandAbsNegHalf(GetRegister(instr.gpr8), instr.alu_half.abs_a, negate_a);
+
+    // instr.alu_half.type_a
+
+    Node op_b = [&]() {
+        switch (opcode->get().GetId()) {
+        case OpCode::Id::HADD2_C:
+        case OpCode::Id::HMUL2_C:
+            return GetConstBuffer(instr.cbuf34.index, instr.cbuf34.offset);
+        case OpCode::Id::HADD2_R:
+        case OpCode::Id::HMUL2_R:
+            return GetRegister(instr.gpr20);
+        default:
+            UNREACHABLE();
+            return Immediate(0);
+        }
+    }();
+    op_b = GetOperandAbsNegHalf(op_b, instr.alu_half.abs_b, negate_b);
+
+    Node value = [&]() {
+        MetaHalfArithmetic meta{true, {instr.alu_half_imm.type_a, instr.alu_half.type_b}};
+        switch (opcode->get().GetId()) {
+        case OpCode::Id::HADD2_C:
+        case OpCode::Id::HADD2_R:
+            return Operation(OperationCode::HAdd, meta, op_a, op_b);
+        case OpCode::Id::HMUL2_C:
+        case OpCode::Id::HMUL2_R:
+            return Operation(OperationCode::HMul, meta, op_a, op_b);
+        default:
+            UNIMPLEMENTED_MSG("Unhandled half float instruction: {}", opcode->get().GetName());
+            return Immediate(0);
+        }
+    }();
+    value = HalfMerge(GetRegister(instr.gpr0), value, instr.alu_half.merge);
+
+    SetRegister(bb, instr.gpr0, value);
+
+    return pc;
+}
+
+} // namespace VideoCommon::Shader

src/video_core/shader/decode/arithmetic_half_immediate.cpp

@@ -0,0 +1,51 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/assert.h"
+#include "common/common_types.h"
+#include "video_core/engines/shader_bytecode.h"
+#include "video_core/shader/shader_ir.h"
+
+namespace VideoCommon::Shader {
+
+using Tegra::Shader::Instruction;
+using Tegra::Shader::OpCode;
+
+u32 ShaderIR::DecodeArithmeticHalfImmediate(BasicBlock& bb, const BasicBlock& code, u32 pc) {
+    const Instruction instr = {program_code[pc]};
+    const auto opcode = OpCode::Decode(instr);
+
+    if (opcode->get().GetId() == OpCode::Id::HADD2_IMM) {
+        UNIMPLEMENTED_IF(instr.alu_half_imm.ftz != 0);
+    } else {
+        UNIMPLEMENTED_IF(instr.alu_half_imm.precision != Tegra::Shader::HalfPrecision::None);
+    }
+    UNIMPLEMENTED_IF_MSG(instr.alu_half_imm.saturate != 0,
+                         "Half float immediate saturation not implemented");
+
+    Node op_a = GetRegister(instr.gpr8);
+    op_a = GetOperandAbsNegHalf(op_a, instr.alu_half_imm.abs_a, instr.alu_half_imm.negate_a);
+
+    const Node op_b = UnpackHalfImmediate(instr, true);
+
+    Node value = [&]() {
+        MetaHalfArithmetic meta{true, {instr.alu_half_imm.type_a}};
+        switch (opcode->get().GetId()) {
+        case OpCode::Id::HADD2_IMM:
+            return Operation(OperationCode::HAdd, meta, op_a, op_b);
+        case OpCode::Id::HMUL2_IMM:
+            return Operation(OperationCode::HMul, meta, op_a, op_b);
+        default:
+            UNREACHABLE();
+            return Immediate(0);
+        }
+    }();
+    value = HalfMerge(GetRegister(instr.gpr0), value, instr.alu_half_imm.merge);
+
+    SetRegister(bb, instr.gpr0, value);
+
+    return pc;
+}
+
+} // namespace VideoCommon::Shader

src/video_core/shader/decode/arithmetic_immediate.cpp

@@ -0,0 +1,52 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/assert.h"
+#include "common/common_types.h"
+#include "video_core/engines/shader_bytecode.h"
+#include "video_core/shader/shader_ir.h"
+
+namespace VideoCommon::Shader {
+
+using Tegra::Shader::Instruction;
+using Tegra::Shader::OpCode;
+
+u32 ShaderIR::DecodeArithmeticImmediate(BasicBlock& bb, const BasicBlock& code, u32 pc) {
+    const Instruction instr = {program_code[pc]};
+    const auto opcode = OpCode::Decode(instr);
+
+    switch (opcode->get().GetId()) {
+    case OpCode::Id::MOV32_IMM: {
+        SetRegister(bb, instr.gpr0, GetImmediate32(instr));
+        break;
+    }
+    case OpCode::Id::FMUL32_IMM: {
+        Node value =
+            Operation(OperationCode::FMul, PRECISE, GetRegister(instr.gpr8), GetImmediate32(instr));
+        value = GetSaturatedFloat(value, instr.fmul32.saturate);
+
+        SetInternalFlagsFromFloat(bb, value, instr.op_32.generates_cc);
+        SetRegister(bb, instr.gpr0, value);
+        break;
+    }
+    case OpCode::Id::FADD32I: {
+        const Node op_a = GetOperandAbsNegFloat(GetRegister(instr.gpr8), instr.fadd32i.abs_a,
+                                                instr.fadd32i.negate_a);
+        const Node op_b = GetOperandAbsNegFloat(GetImmediate32(instr), instr.fadd32i.abs_b,
+                                                instr.fadd32i.negate_b);
+
+        const Node value = Operation(OperationCode::FAdd, PRECISE, op_a, op_b);
+        SetInternalFlagsFromFloat(bb, value, instr.op_32.generates_cc);
+        SetRegister(bb, instr.gpr0, value);
+        break;
+    }
+    default:
+        UNIMPLEMENTED_MSG("Unhandled arithmetic immediate instruction: {}",
+                          opcode->get().GetName());
+    }
+
+    return pc;
+}
+
+} // namespace VideoCommon::Shader

src/video_core/shader/decode/arithmetic_integer.cpp

@@ -0,0 +1,287 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/assert.h"
+#include "common/common_types.h"
+#include "video_core/engines/shader_bytecode.h"
+#include "video_core/shader/shader_ir.h"
+
+namespace VideoCommon::Shader {
+
+using Tegra::Shader::IAdd3Height;
+using Tegra::Shader::Instruction;
+using Tegra::Shader::OpCode;
+using Tegra::Shader::Pred;
+using Tegra::Shader::Register;
+
+u32 ShaderIR::DecodeArithmeticInteger(BasicBlock& bb, const BasicBlock& code, u32 pc) {
+    const Instruction instr = {program_code[pc]};
+    const auto opcode = OpCode::Decode(instr);
+
+    Node op_a = GetRegister(instr.gpr8);
+    Node op_b = [&]() {
+        if (instr.is_b_imm) {
+            return Immediate(instr.alu.GetSignedImm20_20());
+        } else if (instr.is_b_gpr) {
+            return GetRegister(instr.gpr20);
+        } else {
+            return GetConstBuffer(instr.cbuf34.index, instr.cbuf34.offset);
+        }
+    }();
+
+    switch (opcode->get().GetId()) {
+    case OpCode::Id::IADD_C:
+    case OpCode::Id::IADD_R:
+    case OpCode::Id::IADD_IMM: {
+        UNIMPLEMENTED_IF_MSG(instr.alu.saturate_d, "IADD saturation not implemented");
+
+        op_a = GetOperandAbsNegInteger(op_a, false, instr.alu_integer.negate_a, true);
+        op_b = GetOperandAbsNegInteger(op_b, false, instr.alu_integer.negate_b, true);
+
+        const Node value = Operation(OperationCode::IAdd, PRECISE, op_a, op_b);
+
+        SetInternalFlagsFromInteger(bb, value, instr.op_32.generates_cc);
+        SetRegister(bb, instr.gpr0, value);
+        break;
+    }
+    case OpCode::Id::IADD3_C:
+    case OpCode::Id::IADD3_R:
+    case OpCode::Id::IADD3_IMM: {
+        Node op_c = GetRegister(instr.gpr39);
+
+        const auto ApplyHeight = [&](IAdd3Height height, Node value) {
+            switch (height) {
+            case IAdd3Height::None:
+                return value;
+            case IAdd3Height::LowerHalfWord:
+                return BitfieldExtract(value, 0, 16);
+            case IAdd3Height::UpperHalfWord:
+                return BitfieldExtract(value, 16, 16);
+            default:
+                UNIMPLEMENTED_MSG("Unhandled IADD3 height: {}", static_cast<u32>(height));
+                return Immediate(0);
+            }
+        };
+
+        if (opcode->get().GetId() == OpCode::Id::IADD3_R) {
+            op_a = ApplyHeight(instr.iadd3.height_a, op_a);
+            op_b = ApplyHeight(instr.iadd3.height_b, op_b);
+            op_c = ApplyHeight(instr.iadd3.height_c, op_c);
+        }
+
+        op_a = GetOperandAbsNegInteger(op_a, false, instr.iadd3.neg_a, true);
+        op_b = GetOperandAbsNegInteger(op_b, false, instr.iadd3.neg_b, true);
+        op_c = GetOperandAbsNegInteger(op_c, false, instr.iadd3.neg_c, true);
+
+        const Node value = [&]() {
+            const Node add_ab = Operation(OperationCode::IAdd, NO_PRECISE, op_a, op_b);
+            if (opcode->get().GetId() != OpCode::Id::IADD3_R) {
+                return Operation(OperationCode::IAdd, NO_PRECISE, add_ab, op_c);
+            }
+            const Node shifted = [&]() {
+                switch (instr.iadd3.mode) {
+                case Tegra::Shader::IAdd3Mode::RightShift:
+                    // TODO(tech4me): According to
+                    // https://envytools.readthedocs.io/en/latest/hw/graph/maxwell/cuda/int.html?highlight=iadd3
+                    // The addition between op_a and op_b should be done in uint33, more
+                    // investigation required
+                    return Operation(OperationCode::ILogicalShiftRight, NO_PRECISE, add_ab,
+                                     Immediate(16));
+                case Tegra::Shader::IAdd3Mode::LeftShift:
+                    return Operation(OperationCode::ILogicalShiftLeft, NO_PRECISE, add_ab,
+                                     Immediate(16));
+                default:
+                    return add_ab;
+                }
+            }();
+            return Operation(OperationCode::IAdd, NO_PRECISE, shifted, op_c);
+        }();
+
+        SetInternalFlagsFromInteger(bb, value, instr.generates_cc);
+        SetRegister(bb, instr.gpr0, value);
+        break;
+    }
+    case OpCode::Id::ISCADD_C:
+    case OpCode::Id::ISCADD_R:
+    case OpCode::Id::ISCADD_IMM: {
+        UNIMPLEMENTED_IF_MSG(instr.generates_cc,
+                             "Condition codes generation in ISCADD is not implemented");
+
+        op_a = GetOperandAbsNegInteger(op_a, false, instr.alu_integer.negate_a, true);
+        op_b = GetOperandAbsNegInteger(op_b, false, instr.alu_integer.negate_b, true);
+
+        const Node shift = Immediate(static_cast<u32>(instr.alu_integer.shift_amount));
+        const Node shifted_a = Operation(OperationCode::ILogicalShiftLeft, NO_PRECISE, op_a, shift);
+        const Node value = Operation(OperationCode::IAdd, NO_PRECISE, shifted_a, op_b);
+
+        SetInternalFlagsFromInteger(bb, value, instr.generates_cc);
+        SetRegister(bb, instr.gpr0, value);
+        break;
+    }
+    case OpCode::Id::POPC_C:
+    case OpCode::Id::POPC_R:
+    case OpCode::Id::POPC_IMM: {
+        if (instr.popc.invert) {
+            op_b = Operation(OperationCode::IBitwiseNot, NO_PRECISE, op_b);
+        }
+        const Node value = Operation(OperationCode::IBitCount, PRECISE, op_b);
+        SetRegister(bb, instr.gpr0, value);
+        break;
+    }
+    case OpCode::Id::SEL_C:
+    case OpCode::Id::SEL_R:
+    case OpCode::Id::SEL_IMM: {
+        const Node condition = GetPredicate(instr.sel.pred, instr.sel.neg_pred != 0);
+        const Node value = Operation(OperationCode::Select, PRECISE, condition, op_a, op_b);
+        SetRegister(bb, instr.gpr0, value);
+        break;
+    }
+    case OpCode::Id::LOP_C:
+    case OpCode::Id::LOP_R:
+    case OpCode::Id::LOP_IMM: {
+        if (instr.alu.lop.invert_a)
+            op_a = Operation(OperationCode::IBitwiseNot, NO_PRECISE, op_a);
+        if (instr.alu.lop.invert_b)
+            op_b = Operation(OperationCode::IBitwiseNot, NO_PRECISE, op_b);
+
+        WriteLogicOperation(bb, instr.gpr0, instr.alu.lop.operation, op_a, op_b,
+                            instr.alu.lop.pred_result_mode, instr.alu.lop.pred48,
+                            instr.generates_cc);
+        break;
+    }
+    case OpCode::Id::LOP3_C:
+    case OpCode::Id::LOP3_R:
+    case OpCode::Id::LOP3_IMM: {
+        const Node op_c = GetRegister(instr.gpr39);
+        const Node lut = [&]() {
+            if (opcode->get().GetId() == OpCode::Id::LOP3_R) {
+                return Immediate(instr.alu.lop3.GetImmLut28());
+            } else {
+                return Immediate(instr.alu.lop3.GetImmLut48());
+            }
+        }();
+
+        WriteLop3Instruction(bb, instr.gpr0, op_a, op_b, op_c, lut, instr.generates_cc);
+        break;
+    }
+    case OpCode::Id::IMNMX_C:
+    case OpCode::Id::IMNMX_R:
+    case OpCode::Id::IMNMX_IMM: {
+        UNIMPLEMENTED_IF(instr.imnmx.exchange != Tegra::Shader::IMinMaxExchange::None);
+
+        const bool is_signed = instr.imnmx.is_signed;
+
+        const Node condition = GetPredicate(instr.imnmx.pred, instr.imnmx.negate_pred != 0);
+        const Node min = SignedOperation(OperationCode::IMin, is_signed, NO_PRECISE, op_a, op_b);
+        const Node max = SignedOperation(OperationCode::IMax, is_signed, NO_PRECISE, op_a, op_b);
+        const Node value = Operation(OperationCode::Select, NO_PRECISE, condition, min, max);
+
+        SetInternalFlagsFromInteger(bb, value, instr.generates_cc);
+        SetRegister(bb, instr.gpr0, value);
+        break;
+    }
+    case OpCode::Id::LEA_R2:
+    case OpCode::Id::LEA_R1:
+    case OpCode::Id::LEA_IMM:
+    case OpCode::Id::LEA_RZ:
+    case OpCode::Id::LEA_HI: {
+        const auto [op_a, op_b, op_c] = [&]() -> std::tuple<Node, Node, Node> {
+            switch (opcode->get().GetId()) {
+            case OpCode::Id::LEA_R2: {
+                return {GetRegister(instr.gpr20), GetRegister(instr.gpr39),
+                        Immediate(static_cast<u32>(instr.lea.r2.entry_a))};
+            }
+
+            case OpCode::Id::LEA_R1: {
+                const bool neg = instr.lea.r1.neg != 0;
+                return {GetOperandAbsNegInteger(GetRegister(instr.gpr8), false, neg, true),
+                        GetRegister(instr.gpr20),
+                        Immediate(static_cast<u32>(instr.lea.r1.entry_a))};
+            }
+
+            case OpCode::Id::LEA_IMM: {
+                const bool neg = instr.lea.imm.neg != 0;
+                return {Immediate(static_cast<u32>(instr.lea.imm.entry_a)),
+                        GetOperandAbsNegInteger(GetRegister(instr.gpr8), false, neg, true),
+                        Immediate(static_cast<u32>(instr.lea.imm.entry_b))};
+            }
+
+            case OpCode::Id::LEA_RZ: {
+                const bool neg = instr.lea.rz.neg != 0;
+                return {GetConstBuffer(instr.lea.rz.cb_index, instr.lea.rz.cb_offset),
+                        GetOperandAbsNegInteger(GetRegister(instr.gpr8), false, neg, true),
+                        Immediate(static_cast<u32>(instr.lea.rz.entry_a))};
+            }
+
+            case OpCode::Id::LEA_HI:
+            default:
+                UNIMPLEMENTED_MSG("Unhandled LEA subinstruction: {}", opcode->get().GetName());
+
+                return {Immediate(static_cast<u32>(instr.lea.imm.entry_a)), GetRegister(instr.gpr8),
+                        Immediate(static_cast<u32>(instr.lea.imm.entry_b))};
+            }
+        }();
+
+        UNIMPLEMENTED_IF_MSG(instr.lea.pred48 != static_cast<u64>(Pred::UnusedIndex),
+                             "Unhandled LEA Predicate");
+
+        const Node shifted_c =
+            Operation(OperationCode::ILogicalShiftLeft, NO_PRECISE, Immediate(1), op_c);
+        const Node mul_bc = Operation(OperationCode::IMul, NO_PRECISE, op_b, shifted_c);
+        const Node value = Operation(OperationCode::IAdd, NO_PRECISE, op_a, mul_bc);
+
+        SetRegister(bb, instr.gpr0, value);
+
+        break;
+    }
+    default:
+        UNIMPLEMENTED_MSG("Unhandled ArithmeticInteger instruction: {}", opcode->get().GetName());
+    }
+
+    return pc;
+}
+
+void ShaderIR::WriteLop3Instruction(BasicBlock& bb, Register dest, Node op_a, Node op_b, Node op_c,
+                                    Node imm_lut, bool sets_cc) {
+    constexpr u32 lop_iterations = 32;
+    const Node one = Immediate(1);
+    const Node two = Immediate(2);
+
+    Node value{};
+    for (u32 i = 0; i < lop_iterations; ++i) {
+        const Node shift_amount = Immediate(i);
+
+        const Node a = Operation(OperationCode::ILogicalShiftRight, NO_PRECISE, op_c, shift_amount);
+        const Node pack_0 = Operation(OperationCode::IBitwiseAnd, NO_PRECISE, a, one);
+
+        const Node b = Operation(OperationCode::ILogicalShiftRight, NO_PRECISE, op_b, shift_amount);
+        const Node c = Operation(OperationCode::IBitwiseAnd, NO_PRECISE, b, one);
+        const Node pack_1 = Operation(OperationCode::ILogicalShiftLeft, NO_PRECISE, c, one);
+
+        const Node d = Operation(OperationCode::ILogicalShiftRight, NO_PRECISE, op_a, shift_amount);
+        const Node e = Operation(OperationCode::IBitwiseAnd, NO_PRECISE, d, one);
+        const Node pack_2 = Operation(OperationCode::ILogicalShiftLeft, NO_PRECISE, e, two);
+
+        const Node pack_01 = Operation(OperationCode::IBitwiseAnd, NO_PRECISE, pack_0, pack_1);
+        const Node pack_012 = Operation(OperationCode::IBitwiseAnd, NO_PRECISE, pack_01, pack_2);
+
+        const Node shifted_bit =
+            Operation(OperationCode::ILogicalShiftRight, NO_PRECISE, imm_lut, pack_012);
+        const Node bit = Operation(OperationCode::IBitwiseAnd, NO_PRECISE, shifted_bit, one);
+
+        const Node right =
+            Operation(OperationCode::ILogicalShiftLeft, NO_PRECISE, bit, shift_amount);
+
+        if (i > 0) {
+            value = Operation(OperationCode::IBitwiseOr, NO_PRECISE, value, right);
+        } else {
+            value = right;
+        }
+    }
+
+    SetInternalFlagsFromInteger(bb, value, sets_cc);
+    SetRegister(bb, dest, value);
+}
+
+} // namespace VideoCommon::Shader

src/video_core/shader/decode/arithmetic_integer_immediate.cpp

@@ -0,0 +1,96 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/assert.h"
+#include "common/common_types.h"
+#include "video_core/engines/shader_bytecode.h"
+#include "video_core/shader/shader_ir.h"
+
+namespace VideoCommon::Shader {
+
+using Tegra::Shader::Instruction;
+using Tegra::Shader::LogicOperation;
+using Tegra::Shader::OpCode;
+using Tegra::Shader::Pred;
+using Tegra::Shader::PredicateResultMode;
+using Tegra::Shader::Register;
+
+u32 ShaderIR::DecodeArithmeticIntegerImmediate(BasicBlock& bb, const BasicBlock& code, u32 pc) {
+    const Instruction instr = {program_code[pc]};
+    const auto opcode = OpCode::Decode(instr);
+
+    Node op_a = GetRegister(instr.gpr8);
+    Node op_b = Immediate(static_cast<s32>(instr.alu.imm20_32));
+
+    switch (opcode->get().GetId()) {
+    case OpCode::Id::IADD32I: {
+        UNIMPLEMENTED_IF_MSG(instr.iadd32i.saturate, "IADD32I saturation is not implemented");
+
+        op_a = GetOperandAbsNegInteger(op_a, false, instr.iadd32i.negate_a, true);
+
+        const Node value = Operation(OperationCode::IAdd, PRECISE, op_a, op_b);
+
+        SetInternalFlagsFromInteger(bb, value, instr.op_32.generates_cc);
+        SetRegister(bb, instr.gpr0, value);
+        break;
+    }
+    case OpCode::Id::LOP32I: {
+        if (instr.alu.lop32i.invert_a)
+            op_a = Operation(OperationCode::IBitwiseNot, NO_PRECISE, op_a);
+
+        if (instr.alu.lop32i.invert_b)
+            op_b = Operation(OperationCode::IBitwiseNot, NO_PRECISE, op_b);
+
+        WriteLogicOperation(bb, instr.gpr0, instr.alu.lop32i.operation, op_a, op_b,
+                            PredicateResultMode::None, Pred::UnusedIndex, instr.op_32.generates_cc);
+        break;
+    }
+    default:
+        UNIMPLEMENTED_MSG("Unhandled ArithmeticIntegerImmediate instruction: {}",
+                          opcode->get().GetName());
+    }
+
+    return pc;
+}
+
+void ShaderIR::WriteLogicOperation(BasicBlock& bb, Register dest, LogicOperation logic_op,
+                                   Node op_a, Node op_b, PredicateResultMode predicate_mode,
+                                   Pred predicate, bool sets_cc) {
+    const Node result = [&]() {
+        switch (logic_op) {
+        case LogicOperation::And:
+            return Operation(OperationCode::IBitwiseAnd, PRECISE, op_a, op_b);
+        case LogicOperation::Or:
+            return Operation(OperationCode::IBitwiseOr, PRECISE, op_a, op_b);
+        case LogicOperation::Xor:
+            return Operation(OperationCode::IBitwiseXor, PRECISE, op_a, op_b);
+        case LogicOperation::PassB:
+            return op_b;
+        default:
+            UNIMPLEMENTED_MSG("Unimplemented logic operation={}", static_cast<u32>(logic_op));
+            return Immediate(0);
+        }
+    }();
+
+    SetInternalFlagsFromInteger(bb, result, sets_cc);
+    SetRegister(bb, dest, result);
+
+    // Write the predicate value depending on the predicate mode.
+    switch (predicate_mode) {
+    case PredicateResultMode::None:
+        // Do nothing.
+        return;
+    case PredicateResultMode::NotZero: {
+        // Set the predicate to true if the result is not zero.
+        const Node compare = Operation(OperationCode::LogicalINotEqual, result, Immediate(0));
+        SetPredicate(bb, static_cast<u64>(predicate), compare);
+        break;
+    }
+    default:
+        UNIMPLEMENTED_MSG("Unimplemented predicate result mode: {}",
+                          static_cast<u32>(predicate_mode));
+    }
+}
+
+} // namespace VideoCommon::Shader

src/video_core/shader/decode/bfe.cpp

@@ -0,0 +1,49 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/assert.h"
+#include "common/common_types.h"
+#include "video_core/engines/shader_bytecode.h"
+#include "video_core/shader/shader_ir.h"
+
+namespace VideoCommon::Shader {
+
+using Tegra::Shader::Instruction;
+using Tegra::Shader::OpCode;
+
+u32 ShaderIR::DecodeBfe(BasicBlock& bb, const BasicBlock& code, u32 pc) {
+    const Instruction instr = {program_code[pc]};
+    const auto opcode = OpCode::Decode(instr);
+
+    UNIMPLEMENTED_IF(instr.bfe.negate_b);
+
+    Node op_a = GetRegister(instr.gpr8);
+    op_a = GetOperandAbsNegInteger(op_a, false, instr.bfe.negate_a, false);
+
+    switch (opcode->get().GetId()) {
+    case OpCode::Id::BFE_IMM: {
+        UNIMPLEMENTED_IF_MSG(instr.generates_cc,
+                             "Condition codes generation in BFE is not implemented");
+
+        const Node inner_shift_imm = Immediate(static_cast<u32>(instr.bfe.GetLeftShiftValue()));
+        const Node outer_shift_imm =
+            Immediate(static_cast<u32>(instr.bfe.GetLeftShiftValue() + instr.bfe.shift_position));
+
+        const Node inner_shift =
+            Operation(OperationCode::ILogicalShiftLeft, NO_PRECISE, op_a, inner_shift_imm);
+        const Node outer_shift =
+            Operation(OperationCode::ILogicalShiftRight, NO_PRECISE, inner_shift, outer_shift_imm);
+
+        SetInternalFlagsFromInteger(bb, outer_shift, instr.generates_cc);
+        SetRegister(bb, instr.gpr0, outer_shift);
+        break;
+    }
+    default:
+        UNIMPLEMENTED_MSG("Unhandled BFE instruction: {}", opcode->get().GetName());
+    }
+
+    return pc;
+}
+
+} // namespace VideoCommon::Shader

src/video_core/shader/decode/bfi.cpp

@@ -0,0 +1,41 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/assert.h"
+#include "common/common_types.h"
+#include "video_core/engines/shader_bytecode.h"
+#include "video_core/shader/shader_ir.h"
+
+namespace VideoCommon::Shader {
+
+using Tegra::Shader::Instruction;
+using Tegra::Shader::OpCode;
+
+u32 ShaderIR::DecodeBfi(BasicBlock& bb, const BasicBlock& code, u32 pc) {
+    const Instruction instr = {program_code[pc]};
+    const auto opcode = OpCode::Decode(instr);
+
+    const auto [base, packed_shift] = [&]() -> std::tuple<Node, Node> {
+        switch (opcode->get().GetId()) {
+        case OpCode::Id::BFI_IMM_R:
+            return {GetRegister(instr.gpr39), Immediate(instr.alu.GetSignedImm20_20())};
+        default:
+            UNREACHABLE();
+            return {Immediate(0), Immediate(0)};
+        }
+    }();
+    const Node insert = GetRegister(instr.gpr8);
+    const Node offset = BitfieldExtract(packed_shift, 0, 8);
+    const Node bits = BitfieldExtract(packed_shift, 8, 8);
+
+    const Node value =
+        Operation(OperationCode::UBitfieldInsert, PRECISE, base, insert, offset, bits);
+
+    SetInternalFlagsFromInteger(bb, value, instr.generates_cc);
+    SetRegister(bb, instr.gpr0, value);
+
+    return pc;
+}
+
+} // namespace VideoCommon::Shader

src/video_core/shader/decode/conversion.cpp

@@ -0,0 +1,149 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/assert.h"
+#include "common/common_types.h"
+#include "video_core/engines/shader_bytecode.h"
+#include "video_core/shader/shader_ir.h"
+
+namespace VideoCommon::Shader {
+
+using Tegra::Shader::Instruction;
+using Tegra::Shader::OpCode;
+using Tegra::Shader::Register;
+
+u32 ShaderIR::DecodeConversion(BasicBlock& bb, const BasicBlock& code, u32 pc) {
+    const Instruction instr = {program_code[pc]};
+    const auto opcode = OpCode::Decode(instr);
+
+    switch (opcode->get().GetId()) {
+    case OpCode::Id::I2I_R: {
+        UNIMPLEMENTED_IF(instr.conversion.selector);
+
+        const bool input_signed = instr.conversion.is_input_signed;
+        const bool output_signed = instr.conversion.is_output_signed;
+
+        Node value = GetRegister(instr.gpr20);
+        value = ConvertIntegerSize(value, instr.conversion.src_size, input_signed);
+
+        value = GetOperandAbsNegInteger(value, instr.conversion.abs_a, instr.conversion.negate_a,
+                                        input_signed);
+        if (input_signed != output_signed) {
+            value = SignedOperation(OperationCode::ICastUnsigned, output_signed, NO_PRECISE, value);
+        }
+
+        SetInternalFlagsFromInteger(bb, value, instr.generates_cc);
+        SetRegister(bb, instr.gpr0, value);
+        break;
+    }
+    case OpCode::Id::I2F_R:
+    case OpCode::Id::I2F_C: {
+        UNIMPLEMENTED_IF(instr.conversion.dest_size != Register::Size::Word);
+        UNIMPLEMENTED_IF(instr.conversion.selector);
+        UNIMPLEMENTED_IF_MSG(instr.generates_cc,
+                             "Condition codes generation in I2F is not implemented");
+
+        Node value = [&]() {
+            if (instr.is_b_gpr) {
+                return GetRegister(instr.gpr20);
+            } else {
+                return GetConstBuffer(instr.cbuf34.index, instr.cbuf34.offset);
+            }
+        }();
+        const bool input_signed = instr.conversion.is_input_signed;
+        value = ConvertIntegerSize(value, instr.conversion.src_size, input_signed);
+        value = GetOperandAbsNegInteger(value, instr.conversion.abs_a, false, input_signed);
+        value = SignedOperation(OperationCode::FCastInteger, input_signed, PRECISE, value);
+        value = GetOperandAbsNegFloat(value, false, instr.conversion.negate_a);
+
+        SetInternalFlagsFromFloat(bb, value, instr.generates_cc);
+        SetRegister(bb, instr.gpr0, value);
+        break;
+    }
+    case OpCode::Id::F2F_R:
+    case OpCode::Id::F2F_C: {
+        UNIMPLEMENTED_IF(instr.conversion.dest_size != Register::Size::Word);
+        UNIMPLEMENTED_IF(instr.conversion.src_size != Register::Size::Word);
+        UNIMPLEMENTED_IF_MSG(instr.generates_cc,
+                             "Condition codes generation in F2F is not implemented");
+
+        Node value = [&]() {
+            if (instr.is_b_gpr) {
+                return GetRegister(instr.gpr20);
+            } else {
+                return GetConstBuffer(instr.cbuf34.index, instr.cbuf34.offset);
+            }
+        }();
+
+        value = GetOperandAbsNegFloat(value, instr.conversion.abs_a, instr.conversion.negate_a);
+
+        value = [&]() {
+            switch (instr.conversion.f2f.rounding) {
+            case Tegra::Shader::F2fRoundingOp::None:
+                return value;
+            case Tegra::Shader::F2fRoundingOp::Round:
+                return Operation(OperationCode::FRoundEven, PRECISE, value);
+            case Tegra::Shader::F2fRoundingOp::Floor:
+                return Operation(OperationCode::FFloor, PRECISE, value);
+            case Tegra::Shader::F2fRoundingOp::Ceil:
+                return Operation(OperationCode::FCeil, PRECISE, value);
+            case Tegra::Shader::F2fRoundingOp::Trunc:
+                return Operation(OperationCode::FTrunc, PRECISE, value);
+            }
+            UNIMPLEMENTED_MSG("Unimplemented F2F rounding mode {}",
+                              static_cast<u32>(instr.conversion.f2f.rounding.Value()));
+            return Immediate(0);
+        }();
+        value = GetSaturatedFloat(value, instr.alu.saturate_d);
+
+        SetInternalFlagsFromFloat(bb, value, instr.generates_cc);
+        SetRegister(bb, instr.gpr0, value);
+        break;
+    }
+    case OpCode::Id::F2I_R:
+    case OpCode::Id::F2I_C: {
+        UNIMPLEMENTED_IF(instr.conversion.src_size != Register::Size::Word);
+        UNIMPLEMENTED_IF_MSG(instr.generates_cc,
+                             "Condition codes generation in F2I is not implemented");
+        Node value = [&]() {
+            if (instr.is_b_gpr) {
+                return GetRegister(instr.gpr20);
+            } else {
+                return GetConstBuffer(instr.cbuf34.index, instr.cbuf34.offset);
+            }
+        }();
+
+        value = GetOperandAbsNegFloat(value, instr.conversion.abs_a, instr.conversion.negate_a);
+
+        value = [&]() {
+            switch (instr.conversion.f2i.rounding) {
+            case Tegra::Shader::F2iRoundingOp::None:
+                return value;
+            case Tegra::Shader::F2iRoundingOp::Floor:
+                return Operation(OperationCode::FFloor, PRECISE, value);
+            case Tegra::Shader::F2iRoundingOp::Ceil:
+                return Operation(OperationCode::FCeil, PRECISE, value);
+            case Tegra::Shader::F2iRoundingOp::Trunc:
+                return Operation(OperationCode::FTrunc, PRECISE, value);
+            default:
+                UNIMPLEMENTED_MSG("Unimplemented F2I rounding mode {}",
+                                  static_cast<u32>(instr.conversion.f2i.rounding.Value()));
+                return Immediate(0);
+            }
+        }();
+        const bool is_signed = instr.conversion.is_output_signed;
+        value = SignedOperation(OperationCode::ICastFloat, is_signed, PRECISE, value);
+        value = ConvertIntegerSize(value, instr.conversion.dest_size, is_signed);
+
+        SetRegister(bb, instr.gpr0, value);
+        break;
+    }
+    default:
+        UNIMPLEMENTED_MSG("Unhandled conversion instruction: {}", opcode->get().GetName());
+    }
+
+    return pc;
+}
+
+} // namespace VideoCommon::Shader

src/video_core/shader/decode/ffma.cpp

@@ -0,0 +1,59 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/assert.h"
+#include "common/common_types.h"
+#include "video_core/engines/shader_bytecode.h"
+#include "video_core/shader/shader_ir.h"
+
+namespace VideoCommon::Shader {
+
+using Tegra::Shader::Instruction;
+using Tegra::Shader::OpCode;
+
+u32 ShaderIR::DecodeFfma(BasicBlock& bb, const BasicBlock& code, u32 pc) {
+    const Instruction instr = {program_code[pc]};
+    const auto opcode = OpCode::Decode(instr);
+
+    UNIMPLEMENTED_IF_MSG(instr.ffma.cc != 0, "FFMA cc not implemented");
+    UNIMPLEMENTED_IF_MSG(instr.ffma.tab5980_0 != 1, "FFMA tab5980_0({}) not implemented",
+                         instr.ffma.tab5980_0.Value()); // Seems to be 1 by default based on SMO
+    UNIMPLEMENTED_IF_MSG(instr.ffma.tab5980_1 != 0, "FFMA tab5980_1({}) not implemented",
+                         instr.ffma.tab5980_1.Value());
+
+    const Node op_a = GetRegister(instr.gpr8);
+
+    auto [op_b, op_c] = [&]() -> std::tuple<Node, Node> {
+        switch (opcode->get().GetId()) {
+        case OpCode::Id::FFMA_CR: {
+            return {GetConstBuffer(instr.cbuf34.index, instr.cbuf34.offset),
+                    GetRegister(instr.gpr39)};
+        }
+        case OpCode::Id::FFMA_RR:
+            return {GetRegister(instr.gpr20), GetRegister(instr.gpr39)};
+        case OpCode::Id::FFMA_RC: {
+            return {GetRegister(instr.gpr39),
+                    GetConstBuffer(instr.cbuf34.index, instr.cbuf34.offset)};
+        }
+        case OpCode::Id::FFMA_IMM:
+            return {GetImmediate19(instr), GetRegister(instr.gpr39)};
+        default:
+            UNIMPLEMENTED_MSG("Unhandled FFMA instruction: {}", opcode->get().GetName());
+            return {Immediate(0), Immediate(0)};
+        }
+    }();
+
+    op_b = GetOperandAbsNegFloat(op_b, false, instr.ffma.negate_b);
+    op_c = GetOperandAbsNegFloat(op_c, false, instr.ffma.negate_c);
+
+    Node value = Operation(OperationCode::FFma, PRECISE, op_a, op_b, op_c);
+    value = GetSaturatedFloat(value, instr.alu.saturate_d);
+
+    SetInternalFlagsFromFloat(bb, value, instr.generates_cc);
+    SetRegister(bb, instr.gpr0, value);
+
+    return pc;
+}
+
+} // namespace VideoCommon::Shader

src/video_core/shader/decode/float_set.cpp

@@ -0,0 +1,58 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/assert.h"
+#include "common/common_types.h"
+#include "video_core/engines/shader_bytecode.h"
+#include "video_core/shader/shader_ir.h"
+
+namespace VideoCommon::Shader {
+
+using Tegra::Shader::Instruction;
+using Tegra::Shader::OpCode;
+
+u32 ShaderIR::DecodeFloatSet(BasicBlock& bb, const BasicBlock& code, u32 pc) {
+    const Instruction instr = {program_code[pc]};
+    const auto opcode = OpCode::Decode(instr);
+
+    const Node op_a = GetOperandAbsNegFloat(GetRegister(instr.gpr8), instr.fset.abs_a != 0,
+                                            instr.fset.neg_a != 0);
+
+    Node op_b = [&]() {
+        if (instr.is_b_imm) {
+            return GetImmediate19(instr);
+        } else if (instr.is_b_gpr) {
+            return GetRegister(instr.gpr20);
+        } else {
+            return GetConstBuffer(instr.cbuf34.index, instr.cbuf34.offset);
+        }
+    }();
+
+    op_b = GetOperandAbsNegFloat(op_b, instr.fset.abs_b != 0, instr.fset.neg_b != 0);
+
+    // The fset instruction sets a register to 1.0 or -1 (depending on the bf bit) if the
+    // condition is true, and to 0 otherwise.
+    const Node second_pred = GetPredicate(instr.fset.pred39, instr.fset.neg_pred != 0);
+
+    const OperationCode combiner = GetPredicateCombiner(instr.fset.op);
+    const Node first_pred = GetPredicateComparisonFloat(instr.fset.cond, op_a, op_b);
+
+    const Node predicate = Operation(combiner, first_pred, second_pred);
+
+    const Node true_value = instr.fset.bf ? Immediate(1.0f) : Immediate(-1);
+    const Node false_value = instr.fset.bf ? Immediate(0.0f) : Immediate(0);
+    const Node value =
+        Operation(OperationCode::Select, PRECISE, predicate, true_value, false_value);
+
+    if (instr.fset.bf) {
+        SetInternalFlagsFromFloat(bb, value, instr.generates_cc);
+    } else {
+        SetInternalFlagsFromInteger(bb, value, instr.generates_cc);
+    }
+    SetRegister(bb, instr.gpr0, value);
+
+    return pc;
+}
+
+} // namespace VideoCommon::Shader

src/video_core/shader/decode/float_set_predicate.cpp

@@ -0,0 +1,56 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/assert.h"
+#include "common/common_types.h"
+#include "video_core/engines/shader_bytecode.h"
+#include "video_core/shader/shader_ir.h"
+
+namespace VideoCommon::Shader {
+
+using Tegra::Shader::Instruction;
+using Tegra::Shader::OpCode;
+using Tegra::Shader::Pred;
+
+u32 ShaderIR::DecodeFloatSetPredicate(BasicBlock& bb, const BasicBlock& code, u32 pc) {
+    const Instruction instr = {program_code[pc]};
+    const auto opcode = OpCode::Decode(instr);
+
+    const Node op_a = GetOperandAbsNegFloat(GetRegister(instr.gpr8), instr.fsetp.abs_a != 0,
+                                            instr.fsetp.neg_a != 0);
+    Node op_b = [&]() {
+        if (instr.is_b_imm) {
+            return GetImmediate19(instr);
+        } else if (instr.is_b_gpr) {
+            return GetRegister(instr.gpr20);
+        } else {
+            return GetConstBuffer(instr.cbuf34.index, instr.cbuf34.offset);
+        }
+    }();
+    op_b = GetOperandAbsNegFloat(op_b, instr.fsetp.abs_b, false);
+
+    // We can't use the constant predicate as destination.
+    ASSERT(instr.fsetp.pred3 != static_cast<u64>(Pred::UnusedIndex));
+
+    const Node predicate = GetPredicateComparisonFloat(instr.fsetp.cond, op_a, op_b);
+    const Node second_pred = GetPredicate(instr.fsetp.pred39, instr.fsetp.neg_pred != 0);
+
+    const OperationCode combiner = GetPredicateCombiner(instr.fsetp.op);
+    const Node value = Operation(combiner, predicate, second_pred);
+
+    // Set the primary predicate to the result of Predicate OP SecondPredicate
+    SetPredicate(bb, instr.fsetp.pred3, value);
+
+    if (instr.fsetp.pred0 != static_cast<u64>(Pred::UnusedIndex)) {
+        // Set the secondary predicate to the result of !Predicate OP SecondPredicate,
+        // if enabled
+        const Node negated_pred = Operation(OperationCode::LogicalNegate, predicate);
+        const Node second_value = Operation(combiner, negated_pred, second_pred);
+        SetPredicate(bb, instr.fsetp.pred0, second_value);
+    }
+
+    return pc;
+}
+
+} // namespace VideoCommon::Shader

src/video_core/shader/decode/half_set.cpp

@@ -0,0 +1,67 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <array>
+
+#include "common/assert.h"
+#include "common/common_types.h"
+#include "video_core/engines/shader_bytecode.h"
+#include "video_core/shader/shader_ir.h"
+
+namespace VideoCommon::Shader {
+
+using Tegra::Shader::Instruction;
+using Tegra::Shader::OpCode;
+
+u32 ShaderIR::DecodeHalfSet(BasicBlock& bb, const BasicBlock& code, u32 pc) {
+    const Instruction instr = {program_code[pc]};
+    const auto opcode = OpCode::Decode(instr);
+
+    UNIMPLEMENTED_IF(instr.hset2.ftz != 0);
+
+    // instr.hset2.type_a
+    // instr.hset2.type_b
+    Node op_a = GetRegister(instr.gpr8);
+    Node op_b = [&]() {
+        switch (opcode->get().GetId()) {
+        case OpCode::Id::HSET2_R:
+            return GetRegister(instr.gpr20);
+        default:
+            UNREACHABLE();
+            return Immediate(0);
+        }
+    }();
+
+    op_a = GetOperandAbsNegHalf(op_a, instr.hset2.abs_a, instr.hset2.negate_a);
+    op_b = GetOperandAbsNegHalf(op_b, instr.hset2.abs_b, instr.hset2.negate_b);
+
+    const Node second_pred = GetPredicate(instr.hset2.pred39, instr.hset2.neg_pred);
+
+    MetaHalfArithmetic meta{false, {instr.hset2.type_a, instr.hset2.type_b}};
+    const Node comparison_pair = GetPredicateComparisonHalf(instr.hset2.cond, meta, op_a, op_b);
+
+    const OperationCode combiner = GetPredicateCombiner(instr.hset2.op);
+
+    // HSET2 operates on each half float in the pack.
+    std::array<Node, 2> values;
+    for (u32 i = 0; i < 2; ++i) {
+        const u32 raw_value = instr.hset2.bf ? 0x3c00 : 0xffff;
+        const Node true_value = Immediate(raw_value << (i * 16));
+        const Node false_value = Immediate(0);
+
+        const Node comparison =
+            Operation(OperationCode::LogicalPick2, comparison_pair, Immediate(i));
+        const Node predicate = Operation(combiner, comparison, second_pred);
+
+        values[i] =
+            Operation(OperationCode::Select, NO_PRECISE, predicate, true_value, false_value);
+    }
+
+    const Node value = Operation(OperationCode::UBitwiseOr, NO_PRECISE, values[0], values[1]);
+    SetRegister(bb, instr.gpr0, value);
+
+    return pc;
+}
+
+} // namespace VideoCommon::Shader

src/video_core/shader/decode/half_set_predicate.cpp

@@ -0,0 +1,62 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/assert.h"
+#include "common/common_types.h"
+#include "video_core/engines/shader_bytecode.h"
+#include "video_core/shader/shader_ir.h"
+
+namespace VideoCommon::Shader {
+
+using Tegra::Shader::Instruction;
+using Tegra::Shader::OpCode;
+using Tegra::Shader::Pred;
+
+u32 ShaderIR::DecodeHalfSetPredicate(BasicBlock& bb, const BasicBlock& code, u32 pc) {
+    const Instruction instr = {program_code[pc]};
+    const auto opcode = OpCode::Decode(instr);
+
+    UNIMPLEMENTED_IF(instr.hsetp2.ftz != 0);
+
+    Node op_a = GetRegister(instr.gpr8);
+    op_a = GetOperandAbsNegHalf(op_a, instr.hsetp2.abs_a, instr.hsetp2.negate_a);
+
+    const Node op_b = [&]() {
+        switch (opcode->get().GetId()) {
+        case OpCode::Id::HSETP2_R:
+            return GetOperandAbsNegHalf(GetRegister(instr.gpr20), instr.hsetp2.abs_a,
+                                        instr.hsetp2.negate_b);
+        default:
+            UNREACHABLE();
+            return Immediate(0);
+        }
+    }();
+
+    // We can't use the constant predicate as destination.
+    ASSERT(instr.hsetp2.pred3 != static_cast<u64>(Pred::UnusedIndex));
+
+    const Node second_pred = GetPredicate(instr.hsetp2.pred39, instr.hsetp2.neg_pred != 0);
+
+    const OperationCode combiner = GetPredicateCombiner(instr.hsetp2.op);
+    const OperationCode pair_combiner =
+        instr.hsetp2.h_and ? OperationCode::LogicalAll2 : OperationCode::LogicalAny2;
+
+    MetaHalfArithmetic meta = {false, {instr.hsetp2.type_a, instr.hsetp2.type_b}};
+    const Node comparison = GetPredicateComparisonHalf(instr.hsetp2.cond, meta, op_a, op_b);
+    const Node first_pred = Operation(pair_combiner, comparison);
+
+    // Set the primary predicate to the result of Predicate OP SecondPredicate
+    const Node value = Operation(combiner, first_pred, second_pred);
+    SetPredicate(bb, instr.hsetp2.pred3, value);
+
+    if (instr.hsetp2.pred0 != static_cast<u64>(Pred::UnusedIndex)) {
+        // Set the secondary predicate to the result of !Predicate OP SecondPredicate, if enabled
+        const Node negated_pred = Operation(OperationCode::LogicalNegate, first_pred);
+        SetPredicate(bb, instr.hsetp2.pred0, Operation(combiner, negated_pred, second_pred));
+    }
+
+    return pc;
+}
+
+} // namespace VideoCommon::Shader

src/video_core/shader/decode/hfma2.cpp

@@ -0,0 +1,76 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <tuple>
+
+#include "common/assert.h"
+#include "common/common_types.h"
+#include "video_core/engines/shader_bytecode.h"
+#include "video_core/shader/shader_ir.h"
+
+namespace VideoCommon::Shader {
+
+using Tegra::Shader::HalfPrecision;
+using Tegra::Shader::HalfType;
+using Tegra::Shader::Instruction;
+using Tegra::Shader::OpCode;
+
+u32 ShaderIR::DecodeHfma2(BasicBlock& bb, const BasicBlock& code, u32 pc) {
+    const Instruction instr = {program_code[pc]};
+    const auto opcode = OpCode::Decode(instr);
+
+    if (opcode->get().GetId() == OpCode::Id::HFMA2_RR) {
+        UNIMPLEMENTED_IF(instr.hfma2.rr.precision != HalfPrecision::None);
+    } else {
+        UNIMPLEMENTED_IF(instr.hfma2.precision != HalfPrecision::None);
+    }
+
+    constexpr auto identity = HalfType::H0_H1;
+
+    const HalfType type_a = instr.hfma2.type_a;
+    const Node op_a = GetRegister(instr.gpr8);
+
+    bool neg_b{}, neg_c{};
+    auto [saturate, type_b, op_b, type_c,
+          op_c] = [&]() -> std::tuple<bool, HalfType, Node, HalfType, Node> {
+        switch (opcode->get().GetId()) {
+        case OpCode::Id::HFMA2_CR:
+            neg_b = instr.hfma2.negate_b;
+            neg_c = instr.hfma2.negate_c;
+            return {instr.hfma2.saturate, instr.hfma2.type_b,
+                    GetConstBuffer(instr.cbuf34.index, instr.cbuf34.offset), instr.hfma2.type_reg39,
+                    GetRegister(instr.gpr39)};
+        case OpCode::Id::HFMA2_RC:
+            neg_b = instr.hfma2.negate_b;
+            neg_c = instr.hfma2.negate_c;
+            return {instr.hfma2.saturate, instr.hfma2.type_reg39, GetRegister(instr.gpr39),
+                    instr.hfma2.type_b, GetConstBuffer(instr.cbuf34.index, instr.cbuf34.offset)};
+        case OpCode::Id::HFMA2_RR:
+            neg_b = instr.hfma2.rr.negate_b;
+            neg_c = instr.hfma2.rr.negate_c;
+            return {instr.hfma2.rr.saturate, instr.hfma2.type_b, GetRegister(instr.gpr20),
+                    instr.hfma2.rr.type_c, GetRegister(instr.gpr39)};
+        case OpCode::Id::HFMA2_IMM_R:
+            neg_c = instr.hfma2.negate_c;
+            return {instr.hfma2.saturate, identity, UnpackHalfImmediate(instr, true),
+                    instr.hfma2.type_reg39, GetRegister(instr.gpr39)};
+        default:
+            return {false, identity, Immediate(0), identity, Immediate(0)};
+        }
+    }();
+    UNIMPLEMENTED_IF_MSG(saturate, "HFMA2 saturation is not implemented");
+
+    op_b = GetOperandAbsNegHalf(op_b, false, neg_b);
+    op_c = GetOperandAbsNegHalf(op_c, false, neg_c);
+
+    MetaHalfArithmetic meta{true, {type_a, type_b, type_c}};
+    Node value = Operation(OperationCode::HFma, meta, op_a, op_b, op_c);
+    value = HalfMerge(GetRegister(instr.gpr0), value, instr.hfma2.merge);
+
+    SetRegister(bb, instr.gpr0, value);
+
+    return pc;
+}
+
+} // namespace VideoCommon::Shader

src/video_core/shader/decode/integer_set.cpp

@@ -0,0 +1,50 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/assert.h"
+#include "common/common_types.h"
+#include "video_core/engines/shader_bytecode.h"
+#include "video_core/shader/shader_ir.h"
+
+namespace VideoCommon::Shader {
+
+using Tegra::Shader::Instruction;
+using Tegra::Shader::OpCode;
+
+u32 ShaderIR::DecodeIntegerSet(BasicBlock& bb, const BasicBlock& code, u32 pc) {
+    const Instruction instr = {program_code[pc]};
+    const auto opcode = OpCode::Decode(instr);
+
+    const Node op_a = GetRegister(instr.gpr8);
+    const Node op_b = [&]() {
+        if (instr.is_b_imm) {
+            return Immediate(instr.alu.GetSignedImm20_20());
+        } else if (instr.is_b_gpr) {
+            return GetRegister(instr.gpr20);
+        } else {
+            return GetConstBuffer(instr.cbuf34.index, instr.cbuf34.offset);
+        }
+    }();
+
+    // The iset instruction sets a register to 1.0 or -1 (depending on the bf bit) if the condition
+    // is true, and to 0 otherwise.
+    const Node second_pred = GetPredicate(instr.iset.pred39, instr.iset.neg_pred != 0);
+    const Node first_pred =
+        GetPredicateComparisonInteger(instr.iset.cond, instr.iset.is_signed, op_a, op_b);
+
+    const OperationCode combiner = GetPredicateCombiner(instr.iset.op);
+
+    const Node predicate = Operation(combiner, first_pred, second_pred);
+
+    const Node true_value = instr.iset.bf ? Immediate(1.0f) : Immediate(-1);
+    const Node false_value = instr.iset.bf ? Immediate(0.0f) : Immediate(0);
+    const Node value =
+        Operation(OperationCode::Select, PRECISE, predicate, true_value, false_value);
+
+    SetRegister(bb, instr.gpr0, value);
+
+    return pc;
+}
+
+} // namespace VideoCommon::Shader

src/video_core/shader/decode/integer_set_predicate.cpp

@@ -0,0 +1,53 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/assert.h"
+#include "common/common_types.h"
+#include "video_core/engines/shader_bytecode.h"
+#include "video_core/shader/shader_ir.h"
+
+namespace VideoCommon::Shader {
+
+using Tegra::Shader::Instruction;
+using Tegra::Shader::OpCode;
+using Tegra::Shader::Pred;
+
+u32 ShaderIR::DecodeIntegerSetPredicate(BasicBlock& bb, const BasicBlock& code, u32 pc) {
+    const Instruction instr = {program_code[pc]};
+    const auto opcode = OpCode::Decode(instr);
+
+    const Node op_a = GetRegister(instr.gpr8);
+
+    const Node op_b = [&]() {
+        if (instr.is_b_imm) {
+            return Immediate(instr.alu.GetSignedImm20_20());
+        } else if (instr.is_b_gpr) {
+            return GetRegister(instr.gpr20);
+        } else {
+            return GetConstBuffer(instr.cbuf34.index, instr.cbuf34.offset);
+        }
+    }();
+
+    // We can't use the constant predicate as destination.
+    ASSERT(instr.isetp.pred3 != static_cast<u64>(Pred::UnusedIndex));
+
+    const Node second_pred = GetPredicate(instr.isetp.pred39, instr.isetp.neg_pred != 0);
+    const Node predicate =
+        GetPredicateComparisonInteger(instr.isetp.cond, instr.isetp.is_signed, op_a, op_b);
+
+    // Set the primary predicate to the result of Predicate OP SecondPredicate
+    const OperationCode combiner = GetPredicateCombiner(instr.isetp.op);
+    const Node value = Operation(combiner, predicate, second_pred);
+    SetPredicate(bb, instr.isetp.pred3, value);
+
+    if (instr.isetp.pred0 != static_cast<u64>(Pred::UnusedIndex)) {
+        // Set the secondary predicate to the result of !Predicate OP SecondPredicate, if enabled
+        const Node negated_pred = Operation(OperationCode::LogicalNegate, predicate);
+        SetPredicate(bb, instr.isetp.pred0, Operation(combiner, negated_pred, second_pred));
+    }
+
+    return pc;
+}
+
+} // namespace VideoCommon::Shader

src/video_core/shader/decode/memory.cpp

@@ -0,0 +1,688 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <algorithm>
+#include <vector>
+
+#include "common/assert.h"
+#include "common/common_types.h"
+#include "video_core/engines/shader_bytecode.h"
+#include "video_core/shader/shader_ir.h"
+
+namespace VideoCommon::Shader {
+
+using Tegra::Shader::Attribute;
+using Tegra::Shader::Instruction;
+using Tegra::Shader::OpCode;
+using Tegra::Shader::Register;
+using Tegra::Shader::TextureMiscMode;
+using Tegra::Shader::TextureProcessMode;
+using Tegra::Shader::TextureType;
+
+static std::size_t GetCoordCount(TextureType texture_type) {
+    switch (texture_type) {
+    case TextureType::Texture1D:
+        return 1;
+    case TextureType::Texture2D:
+        return 2;
+    case TextureType::Texture3D:
+    case TextureType::TextureCube:
+        return 3;
+    default:
+        UNIMPLEMENTED_MSG("Unhandled texture type: {}", static_cast<u32>(texture_type));
+        return 0;
+    }
+}
+
+u32 ShaderIR::DecodeMemory(BasicBlock& bb, const BasicBlock& code, u32 pc) {
+    const Instruction instr = {program_code[pc]};
+    const auto opcode = OpCode::Decode(instr);
+
+    switch (opcode->get().GetId()) {
+    case OpCode::Id::LD_A: {
+        // Note: Shouldn't this be interp mode flat? As in no interpolation made.
+        UNIMPLEMENTED_IF_MSG(instr.gpr8.Value() != Register::ZeroIndex,
+                             "Indirect attribute loads are not supported");
+        UNIMPLEMENTED_IF_MSG((instr.attribute.fmt20.immediate.Value() % sizeof(u32)) != 0,
+                             "Unaligned attribute loads are not supported");
+
+        Tegra::Shader::IpaMode input_mode{Tegra::Shader::IpaInterpMode::Perspective,
+                                          Tegra::Shader::IpaSampleMode::Default};
+
+        u64 next_element = instr.attribute.fmt20.element;
+        auto next_index = static_cast<u64>(instr.attribute.fmt20.index.Value());
+
+        const auto LoadNextElement = [&](u32 reg_offset) {
+            const Node buffer = GetRegister(instr.gpr39);
+            const Node attribute = GetInputAttribute(static_cast<Attribute::Index>(next_index),
+                                                     next_element, input_mode, buffer);
+
+            SetRegister(bb, instr.gpr0.Value() + reg_offset, attribute);
+
+            // Load the next attribute element into the following register. If the element
+            // to load goes beyond the vec4 size, load the first element of the next
+            // attribute.
+            next_element = (next_element + 1) % 4;
+            next_index = next_index + (next_element == 0 ? 1 : 0);
+        };
+
+        const u32 num_words = static_cast<u32>(instr.attribute.fmt20.size.Value()) + 1;
+        for (u32 reg_offset = 0; reg_offset < num_words; ++reg_offset) {
+            LoadNextElement(reg_offset);
+        }
+        break;
+    }
+    case OpCode::Id::LD_C: {
+        UNIMPLEMENTED_IF(instr.ld_c.unknown != 0);
+
+        Node index = GetRegister(instr.gpr8);
+
+        const Node op_a =
+            GetConstBufferIndirect(instr.cbuf36.index, instr.cbuf36.offset + 0, index);
+
+        switch (instr.ld_c.type.Value()) {
+        case Tegra::Shader::UniformType::Single:
+            SetRegister(bb, instr.gpr0, op_a);
+            break;
+
+        case Tegra::Shader::UniformType::Double: {
+            const Node op_b =
+                GetConstBufferIndirect(instr.cbuf36.index, instr.cbuf36.offset + 4, index);
+
+            SetTemporal(bb, 0, op_a);
+            SetTemporal(bb, 1, op_b);
+            SetRegister(bb, instr.gpr0, GetTemporal(0));
+            SetRegister(bb, instr.gpr0.Value() + 1, GetTemporal(1));
+            break;
+        }
+        default:
+            UNIMPLEMENTED_MSG("Unhandled type: {}", static_cast<unsigned>(instr.ld_c.type.Value()));
+        }
+        break;
+    }
+    case OpCode::Id::LD_L: {
+        UNIMPLEMENTED_IF_MSG(instr.ld_l.unknown == 1, "LD_L Unhandled mode: {}",
+                             static_cast<unsigned>(instr.ld_l.unknown.Value()));
+
+        const Node index = Operation(OperationCode::IAdd, GetRegister(instr.gpr8),
+                                     Immediate(static_cast<s32>(instr.smem_imm)));
+        const Node lmem = GetLocalMemory(index);
+
+        switch (instr.ldst_sl.type.Value()) {
+        case Tegra::Shader::StoreType::Bytes32:
+            SetRegister(bb, instr.gpr0, lmem);
+            break;
+        default:
+            UNIMPLEMENTED_MSG("LD_L Unhandled type: {}",
+                              static_cast<unsigned>(instr.ldst_sl.type.Value()));
+        }
+        break;
+    }
+    case OpCode::Id::ST_A: {
+        UNIMPLEMENTED_IF_MSG(instr.gpr8.Value() != Register::ZeroIndex,
+                             "Indirect attribute loads are not supported");
+        UNIMPLEMENTED_IF_MSG((instr.attribute.fmt20.immediate.Value() % sizeof(u32)) != 0,
+                             "Unaligned attribute loads are not supported");
+
+        u64 next_element = instr.attribute.fmt20.element;
+        auto next_index = static_cast<u64>(instr.attribute.fmt20.index.Value());
+
+        const auto StoreNextElement = [&](u32 reg_offset) {
+            const auto dest = GetOutputAttribute(static_cast<Attribute::Index>(next_index),
+                                                 next_element, GetRegister(instr.gpr39));
+            const auto src = GetRegister(instr.gpr0.Value() + reg_offset);
+
+            bb.push_back(Operation(OperationCode::Assign, dest, src));
+
+            // Load the next attribute element into the following register. If the element
+            // to load goes beyond the vec4 size, load the first element of the next
+            // attribute.
+            next_element = (next_element + 1) % 4;
+            next_index = next_index + (next_element == 0 ? 1 : 0);
+        };
+
+        const u32 num_words = static_cast<u32>(instr.attribute.fmt20.size.Value()) + 1;
+        for (u32 reg_offset = 0; reg_offset < num_words; ++reg_offset) {
+            StoreNextElement(reg_offset);
+        }
+
+        break;
+    }
+    case OpCode::Id::ST_L: {
+        UNIMPLEMENTED_IF_MSG(instr.st_l.unknown == 0, "ST_L Unhandled mode: {}",
+                             static_cast<u32>(instr.st_l.unknown.Value()));
+
+        const Node index = Operation(OperationCode::IAdd, NO_PRECISE, GetRegister(instr.gpr8),
+                                     Immediate(static_cast<s32>(instr.smem_imm)));
+
+        switch (instr.ldst_sl.type.Value()) {
+        case Tegra::Shader::StoreType::Bytes32:
+            SetLocalMemory(bb, index, GetRegister(instr.gpr0));
+            break;
+        default:
+            UNIMPLEMENTED_MSG("ST_L Unhandled type: {}",
+                              static_cast<u32>(instr.ldst_sl.type.Value()));
+        }
+        break;
+    }
+    case OpCode::Id::TEX: {
+        UNIMPLEMENTED_IF_MSG(instr.tex.UsesMiscMode(TextureMiscMode::AOFFI),
+                             "AOFFI is not implemented");
+
+        if (instr.tex.UsesMiscMode(TextureMiscMode::NODEP)) {
+            LOG_WARNING(HW_GPU, "TEX.NODEP implementation is incomplete");
+        }
+
+        const TextureType texture_type{instr.tex.texture_type};
+        const bool is_array = instr.tex.array != 0;
+        const bool depth_compare = instr.tex.UsesMiscMode(TextureMiscMode::DC);
+        const auto process_mode = instr.tex.GetTextureProcessMode();
+        WriteTexInstructionFloat(
+            bb, instr, GetTexCode(instr, texture_type, process_mode, depth_compare, is_array));
+        break;
+    }
+    case OpCode::Id::TEXS: {
+        const TextureType texture_type{instr.texs.GetTextureType()};
+        const bool is_array{instr.texs.IsArrayTexture()};
+        const bool depth_compare = instr.texs.UsesMiscMode(TextureMiscMode::DC);
+        const auto process_mode = instr.texs.GetTextureProcessMode();
+
+        if (instr.texs.UsesMiscMode(TextureMiscMode::NODEP)) {
+            LOG_WARNING(HW_GPU, "TEXS.NODEP implementation is incomplete");
+        }
+
+        const Node4 components =
+            GetTexsCode(instr, texture_type, process_mode, depth_compare, is_array);
+
+        if (instr.texs.fp32_flag) {
+            WriteTexsInstructionFloat(bb, instr, components);
+        } else {
+            WriteTexsInstructionHalfFloat(bb, instr, components);
+        }
+        break;
+    }
+    case OpCode::Id::TLD4: {
+        ASSERT(instr.tld4.array == 0);
+        UNIMPLEMENTED_IF_MSG(instr.tld4.UsesMiscMode(TextureMiscMode::AOFFI),
+                             "AOFFI is not implemented");
+        UNIMPLEMENTED_IF_MSG(instr.tld4.UsesMiscMode(TextureMiscMode::NDV),
+                             "NDV is not implemented");
+        UNIMPLEMENTED_IF_MSG(instr.tld4.UsesMiscMode(TextureMiscMode::PTP),
+                             "PTP is not implemented");
+
+        if (instr.tld4.UsesMiscMode(TextureMiscMode::NODEP)) {
+            LOG_WARNING(HW_GPU, "TLD4.NODEP implementation is incomplete");
+        }
+
+        const auto texture_type = instr.tld4.texture_type.Value();
+        const bool depth_compare = instr.tld4.UsesMiscMode(TextureMiscMode::DC);
+        const bool is_array = instr.tld4.array != 0;
+        WriteTexInstructionFloat(bb, instr,
+                                 GetTld4Code(instr, texture_type, depth_compare, is_array));
+        break;
+    }
+    case OpCode::Id::TLD4S: {
+        UNIMPLEMENTED_IF_MSG(instr.tld4s.UsesMiscMode(TextureMiscMode::AOFFI),
+                             "AOFFI is not implemented");
+
+        if (instr.tld4s.UsesMiscMode(TextureMiscMode::NODEP)) {
+            LOG_WARNING(HW_GPU, "TLD4S.NODEP implementation is incomplete");
+        }
+
+        const bool depth_compare = instr.tld4s.UsesMiscMode(TextureMiscMode::DC);
+        const Node op_a = GetRegister(instr.gpr8);
+        const Node op_b = GetRegister(instr.gpr20);
+
+        std::vector<Node> coords;
+
+        // TODO(Subv): Figure out how the sampler type is encoded in the TLD4S instruction.
+        if (depth_compare) {
+            // Note: TLD4S coordinate encoding works just like TEXS's
+            const Node op_y = GetRegister(instr.gpr8.Value() + 1);
+            coords.push_back(op_a);
+            coords.push_back(op_y);
+            coords.push_back(op_b);
+        } else {
+            coords.push_back(op_a);
+            coords.push_back(op_b);
+        }
+        const auto num_coords = static_cast<u32>(coords.size());
+        coords.push_back(Immediate(static_cast<u32>(instr.tld4s.component)));
+
+        const auto& sampler =
+            GetSampler(instr.sampler, TextureType::Texture2D, false, depth_compare);
+
+        Node4 values;
+        for (u32 element = 0; element < values.size(); ++element) {
+            auto params = coords;
+            MetaTexture meta{sampler, element, num_coords};
+            values[element] =
+                Operation(OperationCode::F4TextureGather, std::move(meta), std::move(params));
+        }
+
+        WriteTexsInstructionFloat(bb, instr, values);
+        break;
+    }
+    case OpCode::Id::TXQ: {
+        if (instr.txq.UsesMiscMode(TextureMiscMode::NODEP)) {
+            LOG_WARNING(HW_GPU, "TXQ.NODEP implementation is incomplete");
+        }
+
+        // TODO: The new commits on the texture refactor, change the way samplers work.
+        // Sadly, not all texture instructions specify the type of texture their sampler
+        // uses. This must be fixed at a later instance.
+        const auto& sampler =
+            GetSampler(instr.sampler, Tegra::Shader::TextureType::Texture2D, false, false);
+
+        switch (instr.txq.query_type) {
+        case Tegra::Shader::TextureQueryType::Dimension: {
+            for (u32 element = 0; element < 4; ++element) {
+                MetaTexture meta{sampler, element};
+                const Node value = Operation(OperationCode::F4TextureQueryDimensions,
+                                             std::move(meta), GetRegister(instr.gpr8));
+                SetTemporal(bb, element, value);
+            }
+            for (u32 i = 0; i < 4; ++i) {
+                SetRegister(bb, instr.gpr0.Value() + i, GetTemporal(i));
+            }
+            break;
+        }
+        default:
+            UNIMPLEMENTED_MSG("Unhandled texture query type: {}",
+                              static_cast<u32>(instr.txq.query_type.Value()));
+        }
+        break;
+    }
+    case OpCode::Id::TMML: {
+        UNIMPLEMENTED_IF_MSG(instr.tmml.UsesMiscMode(Tegra::Shader::TextureMiscMode::NDV),
+                             "NDV is not implemented");
+
+        if (instr.tmml.UsesMiscMode(TextureMiscMode::NODEP)) {
+            LOG_WARNING(HW_GPU, "TMML.NODEP implementation is incomplete");
+        }
+
+        auto texture_type = instr.tmml.texture_type.Value();
+        const bool is_array = instr.tmml.array != 0;
+        const auto& sampler = GetSampler(instr.sampler, texture_type, is_array, false);
+
+        std::vector<Node> coords;
+
+        // TODO: Add coordinates for different samplers once other texture types are implemented.
+        switch (texture_type) {
+        case TextureType::Texture1D:
+            coords.push_back(GetRegister(instr.gpr8));
+            break;
+        case TextureType::Texture2D:
+            coords.push_back(GetRegister(instr.gpr8.Value() + 0));
+            coords.push_back(GetRegister(instr.gpr8.Value() + 1));
+            break;
+        default:
+            UNIMPLEMENTED_MSG("Unhandled texture type {}", static_cast<u32>(texture_type));
+
+            // Fallback to interpreting as a 2D texture for now
+            coords.push_back(GetRegister(instr.gpr8.Value() + 0));
+            coords.push_back(GetRegister(instr.gpr8.Value() + 1));
+            texture_type = TextureType::Texture2D;
+        }
+
+        for (u32 element = 0; element < 2; ++element) {
+            auto params = coords;
+            MetaTexture meta_texture{sampler, element, static_cast<u32>(coords.size())};
+            const Node value =
+                Operation(OperationCode::F4TextureQueryLod, meta_texture, std::move(params));
+            SetTemporal(bb, element, value);
+        }
+        for (u32 element = 0; element < 2; ++element) {
+            SetRegister(bb, instr.gpr0.Value() + element, GetTemporal(element));
+        }
+
+        break;
+    }
+    case OpCode::Id::TLDS: {
+        const Tegra::Shader::TextureType texture_type{instr.tlds.GetTextureType()};
+        const bool is_array{instr.tlds.IsArrayTexture()};
+
+        UNIMPLEMENTED_IF_MSG(instr.tlds.UsesMiscMode(TextureMiscMode::AOFFI),
+                             "AOFFI is not implemented");
+        UNIMPLEMENTED_IF_MSG(instr.tlds.UsesMiscMode(TextureMiscMode::MZ), "MZ is not implemented");
+
+        if (instr.tlds.UsesMiscMode(TextureMiscMode::NODEP)) {
+            LOG_WARNING(HW_GPU, "TMML.NODEP implementation is incomplete");
+        }
+
+        WriteTexsInstructionFloat(bb, instr, GetTldsCode(instr, texture_type, is_array));
+        break;
+    }
+    default:
+        UNIMPLEMENTED_MSG("Unhandled memory instruction: {}", opcode->get().GetName());
+    }
+
+    return pc;
+}
+
+const Sampler& ShaderIR::GetSampler(const Tegra::Shader::Sampler& sampler, TextureType type,
+                                    bool is_array, bool is_shadow) {
+    const auto offset = static_cast<std::size_t>(sampler.index.Value());
+
+    // If this sampler has already been used, return the existing mapping.
+    const auto itr =
+        std::find_if(used_samplers.begin(), used_samplers.end(),
+                     [&](const Sampler& entry) { return entry.GetOffset() == offset; });
+    if (itr != used_samplers.end()) {
+        ASSERT(itr->GetType() == type && itr->IsArray() == is_array &&
+               itr->IsShadow() == is_shadow);
+        return *itr;
+    }
+
+    // Otherwise create a new mapping for this sampler
+    const std::size_t next_index = used_samplers.size();
+    const Sampler entry{offset, next_index, type, is_array, is_shadow};
+    return *used_samplers.emplace(entry).first;
+}
+
+void ShaderIR::WriteTexInstructionFloat(BasicBlock& bb, Instruction instr,
+                                        const Node4& components) {
+    u32 dest_elem = 0;
+    for (u32 elem = 0; elem < 4; ++elem) {
+        if (!instr.tex.IsComponentEnabled(elem)) {
+            // Skip disabled components
+            continue;
+        }
+        SetTemporal(bb, dest_elem++, components[elem]);
+    }
+    // After writing values in temporals, move them to the real registers
+    for (u32 i = 0; i < dest_elem; ++i) {
+        SetRegister(bb, instr.gpr0.Value() + i, GetTemporal(i));
+    }
+}
+
+void ShaderIR::WriteTexsInstructionFloat(BasicBlock& bb, Instruction instr,
+                                         const Node4& components) {
+    // TEXS has two destination registers and a swizzle. The first two elements in the swizzle
+    // go into gpr0+0 and gpr0+1, and the rest goes into gpr28+0 and gpr28+1
+
+    u32 dest_elem = 0;
+    for (u32 component = 0; component < 4; ++component) {
+        if (!instr.texs.IsComponentEnabled(component))
+            continue;
+        SetTemporal(bb, dest_elem++, components[component]);
+    }
+
+    for (u32 i = 0; i < dest_elem; ++i) {
+        if (i < 2) {
+            // Write the first two swizzle components to gpr0 and gpr0+1
+            SetRegister(bb, instr.gpr0.Value() + i % 2, GetTemporal(i));
+        } else {
+            ASSERT(instr.texs.HasTwoDestinations());
+            // Write the rest of the swizzle components to gpr28 and gpr28+1
+            SetRegister(bb, instr.gpr28.Value() + i % 2, GetTemporal(i));
+        }
+    }
+}
+
+void ShaderIR::WriteTexsInstructionHalfFloat(BasicBlock& bb, Instruction instr,
+                                             const Node4& components) {
+    // TEXS.F16 destionation registers are packed in two registers in pairs (just like any half
+    // float instruction).
+
+    Node4 values;
+    u32 dest_elem = 0;
+    for (u32 component = 0; component < 4; ++component) {
+        if (!instr.texs.IsComponentEnabled(component))
+            continue;
+        values[dest_elem++] = components[component];
+    }
+    if (dest_elem == 0)
+        return;
+
+    std::generate(values.begin() + dest_elem, values.end(), [&]() { return Immediate(0); });
+
+    const Node first_value = Operation(OperationCode::HPack2, values[0], values[1]);
+    if (dest_elem <= 2) {
+        SetRegister(bb, instr.gpr0, first_value);
+        return;
+    }
+
+    SetTemporal(bb, 0, first_value);
+    SetTemporal(bb, 1, Operation(OperationCode::HPack2, values[2], values[3]));
+
+    SetRegister(bb, instr.gpr0, GetTemporal(0));
+    SetRegister(bb, instr.gpr28, GetTemporal(1));
+}
+
+Node4 ShaderIR::GetTextureCode(Instruction instr, TextureType texture_type,
+                               TextureProcessMode process_mode, bool depth_compare, bool is_array,
+                               std::size_t array_offset, std::size_t bias_offset,
+                               std::vector<Node>&& coords) {
+    UNIMPLEMENTED_IF_MSG(
+        (texture_type == TextureType::Texture3D && (is_array || depth_compare)) ||
+            (texture_type == TextureType::TextureCube && is_array && depth_compare),
+        "This method is not supported.");
+
+    const auto& sampler = GetSampler(instr.sampler, texture_type, is_array, depth_compare);
+
+    const bool lod_needed = process_mode == TextureProcessMode::LZ ||
+                            process_mode == TextureProcessMode::LL ||
+                            process_mode == TextureProcessMode::LLA;
+
+    // LOD selection (either via bias or explicit textureLod) not supported in GL for
+    // sampler2DArrayShadow and samplerCubeArrayShadow.
+    const bool gl_lod_supported =
+        !((texture_type == Tegra::Shader::TextureType::Texture2D && is_array && depth_compare) ||
+          (texture_type == Tegra::Shader::TextureType::TextureCube && is_array && depth_compare));
+
+    const OperationCode read_method =
+        lod_needed && gl_lod_supported ? OperationCode::F4TextureLod : OperationCode::F4Texture;
+
+    UNIMPLEMENTED_IF(process_mode != TextureProcessMode::None && !gl_lod_supported);
+
+    std::optional<u32> array_offset_value;
+    if (is_array)
+        array_offset_value = static_cast<u32>(array_offset);
+
+    const auto coords_count = static_cast<u32>(coords.size());
+
+    if (process_mode != TextureProcessMode::None && gl_lod_supported) {
+        if (process_mode == TextureProcessMode::LZ) {
+            coords.push_back(Immediate(0.0f));
+        } else {
+            // If present, lod or bias are always stored in the register indexed by the gpr20
+            // field with an offset depending on the usage of the other registers
+            coords.push_back(GetRegister(instr.gpr20.Value() + bias_offset));
+        }
+    }
+
+    Node4 values;
+    for (u32 element = 0; element < values.size(); ++element) {
+        auto params = coords;
+        MetaTexture meta{sampler, element, coords_count, array_offset_value};
+        values[element] = Operation(read_method, std::move(meta), std::move(params));
+    }
+
+    return values;
+}
+
+Node4 ShaderIR::GetTexCode(Instruction instr, TextureType texture_type,
+                           TextureProcessMode process_mode, bool depth_compare, bool is_array) {
+    const bool lod_bias_enabled =
+        (process_mode != TextureProcessMode::None && process_mode != TextureProcessMode::LZ);
+
+    const auto [coord_count, total_coord_count] = ValidateAndGetCoordinateElement(
+        texture_type, depth_compare, is_array, lod_bias_enabled, 4, 5);
+    // If enabled arrays index is always stored in the gpr8 field
+    const u64 array_register = instr.gpr8.Value();
+    // First coordinate index is the gpr8 or gpr8 + 1 when arrays are used
+    const u64 coord_register = array_register + (is_array ? 1 : 0);
+
+    std::vector<Node> coords;
+    for (std::size_t i = 0; i < coord_count; ++i) {
+        coords.push_back(GetRegister(coord_register + i));
+    }
+    // 1D.DC in opengl the 2nd component is ignored.
+    if (depth_compare && !is_array && texture_type == TextureType::Texture1D) {
+        coords.push_back(Immediate(0.0f));
+    }
+    std::size_t array_offset{};
+    if (is_array) {
+        array_offset = coords.size();
+        coords.push_back(GetRegister(array_register));
+    }
+    if (depth_compare) {
+        // Depth is always stored in the register signaled by gpr20
+        // or in the next register if lod or bias are used
+        const u64 depth_register = instr.gpr20.Value() + (lod_bias_enabled ? 1 : 0);
+        coords.push_back(GetRegister(depth_register));
+    }
+    // Fill ignored coordinates
+    while (coords.size() < total_coord_count) {
+        coords.push_back(Immediate(0));
+    }
+
+    return GetTextureCode(instr, texture_type, process_mode, depth_compare, is_array, array_offset,
+                          0, std::move(coords));
+}
+
+Node4 ShaderIR::GetTexsCode(Instruction instr, TextureType texture_type,
+                            TextureProcessMode process_mode, bool depth_compare, bool is_array) {
+    const bool lod_bias_enabled =
+        (process_mode != TextureProcessMode::None && process_mode != TextureProcessMode::LZ);
+
+    const auto [coord_count, total_coord_count] = ValidateAndGetCoordinateElement(
+        texture_type, depth_compare, is_array, lod_bias_enabled, 4, 4);
+    // If enabled arrays index is always stored in the gpr8 field
+    const u64 array_register = instr.gpr8.Value();
+    // First coordinate index is stored in gpr8 field or (gpr8 + 1) when arrays are used
+    const u64 coord_register = array_register + (is_array ? 1 : 0);
+    const u64 last_coord_register =
+        (is_array || !(lod_bias_enabled || depth_compare) || (coord_count > 2))
+            ? static_cast<u64>(instr.gpr20.Value())
+            : coord_register + 1;
+
+    std::vector<Node> coords;
+    for (std::size_t i = 0; i < coord_count; ++i) {
+        const bool last = (i == (coord_count - 1)) && (coord_count > 1);
+        coords.push_back(GetRegister(last ? last_coord_register : coord_register + i));
+    }
+
+    std::size_t array_offset{};
+    if (is_array) {
+        array_offset = coords.size();
+        coords.push_back(GetRegister(array_register));
+    }
+    if (depth_compare) {
+        // Depth is always stored in the register signaled by gpr20
+        // or in the next register if lod or bias are used
+        const u64 depth_register = instr.gpr20.Value() + (lod_bias_enabled ? 1 : 0);
+        coords.push_back(GetRegister(depth_register));
+    }
+    // Fill ignored coordinates
+    while (coords.size() < total_coord_count) {
+        coords.push_back(Immediate(0));
+    }
+
+    return GetTextureCode(instr, texture_type, process_mode, depth_compare, is_array, array_offset,
+                          (coord_count > 2 ? 1 : 0), std::move(coords));
+}
+
+Node4 ShaderIR::GetTld4Code(Instruction instr, TextureType texture_type, bool depth_compare,
+                            bool is_array) {
+    const std::size_t coord_count = GetCoordCount(texture_type);
+    const std::size_t total_coord_count = coord_count + (is_array ? 1 : 0);
+    const std::size_t total_reg_count = total_coord_count + (depth_compare ? 1 : 0);
+
+    // If enabled arrays index is always stored in the gpr8 field
+    const u64 array_register = instr.gpr8.Value();
+    // First coordinate index is the gpr8 or gpr8 + 1 when arrays are used
+    const u64 coord_register = array_register + (is_array ? 1 : 0);
+
+    std::vector<Node> coords;
+
+    for (size_t i = 0; i < coord_count; ++i) {
+        coords.push_back(GetRegister(coord_register + i));
+    }
+    std::optional<u32> array_offset;
+    if (is_array) {
+        array_offset = static_cast<u32>(coords.size());
+        coords.push_back(GetRegister(array_register));
+    }
+
+    const auto& sampler = GetSampler(instr.sampler, texture_type, is_array, depth_compare);
+
+    Node4 values;
+    for (u32 element = 0; element < values.size(); ++element) {
+        auto params = coords;
+        MetaTexture meta{sampler, element, static_cast<u32>(coords.size()), array_offset};
+        values[element] =
+            Operation(OperationCode::F4TextureGather, std::move(meta), std::move(params));
+    }
+
+    return values;
+}
+
+Node4 ShaderIR::GetTldsCode(Instruction instr, TextureType texture_type, bool is_array) {
+    const std::size_t type_coord_count = GetCoordCount(texture_type);
+    const std::size_t total_coord_count = type_coord_count + (is_array ? 1 : 0);
+    const bool lod_enabled = instr.tlds.GetTextureProcessMode() == TextureProcessMode::LL;
+
+    // If enabled arrays index is always stored in the gpr8 field
+    const u64 array_register = instr.gpr8.Value();
+    // if is array gpr20 is used
+    const u64 coord_register = is_array ? instr.gpr20.Value() : instr.gpr8.Value();
+
+    const u64 last_coord_register =
+        ((type_coord_count > 2) || (type_coord_count == 2 && !lod_enabled)) && !is_array
+            ? static_cast<u64>(instr.gpr20.Value())
+            : coord_register + 1;
+
+    std::vector<Node> coords;
+
+    for (std::size_t i = 0; i < type_coord_count; ++i) {
+        const bool last = (i == (type_coord_count - 1)) && (type_coord_count > 1);
+        coords.push_back(GetRegister(last ? last_coord_register : coord_register + i));
+    }
+    std::optional<u32> array_offset;
+    if (is_array) {
+        array_offset = static_cast<u32>(coords.size());
+        coords.push_back(GetRegister(array_register));
+    }
+    const auto coords_count = static_cast<u32>(coords.size());
+
+    if (lod_enabled) {
+        // When lod is used always is in grp20
+        coords.push_back(GetRegister(instr.gpr20));
+    } else {
+        coords.push_back(Immediate(0));
+    }
+
+    const auto& sampler = GetSampler(instr.sampler, texture_type, is_array, false);
+
+    Node4 values;
+    for (u32 element = 0; element < values.size(); ++element) {
+        auto params = coords;
+        MetaTexture meta{sampler, element, coords_count, array_offset};
+        values[element] =
+            Operation(OperationCode::F4TexelFetch, std::move(meta), std::move(params));
+    }
+    return values;
+}
+
+std::tuple<std::size_t, std::size_t> ShaderIR::ValidateAndGetCoordinateElement(
+    TextureType texture_type, bool depth_compare, bool is_array, bool lod_bias_enabled,
+    std::size_t max_coords, std::size_t max_inputs) {
+    const std::size_t coord_count = GetCoordCount(texture_type);
+
+    std::size_t total_coord_count = coord_count + (is_array ? 1 : 0) + (depth_compare ? 1 : 0);
+    const std::size_t total_reg_count = total_coord_count + (lod_bias_enabled ? 1 : 0);
+    if (total_coord_count > max_coords || total_reg_count > max_inputs) {
+        UNIMPLEMENTED_MSG("Unsupported Texture operation");
+        total_coord_count = std::min(total_coord_count, max_coords);
+    }
+    // 1D.DC OpenGL is using a vec3 but 2nd component is ignored later.
+    total_coord_count +=
+        (depth_compare && !is_array && texture_type == TextureType::Texture1D) ? 1 : 0;
+
+    return {coord_count, total_coord_count};
+}
+
+} // namespace VideoCommon::Shader

src/video_core/shader/decode/other.cpp

@@ -0,0 +1,178 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/assert.h"
+#include "common/common_types.h"
+#include "video_core/engines/shader_bytecode.h"
+#include "video_core/shader/shader_ir.h"
+
+namespace VideoCommon::Shader {
+
+using Tegra::Shader::ConditionCode;
+using Tegra::Shader::Instruction;
+using Tegra::Shader::OpCode;
+using Tegra::Shader::Register;
+
+u32 ShaderIR::DecodeOther(BasicBlock& bb, const BasicBlock& code, u32 pc) {
+    const Instruction instr = {program_code[pc]};
+    const auto opcode = OpCode::Decode(instr);
+
+    switch (opcode->get().GetId()) {
+    case OpCode::Id::EXIT: {
+        const Tegra::Shader::ConditionCode cc = instr.flow_condition_code;
+        UNIMPLEMENTED_IF_MSG(cc != Tegra::Shader::ConditionCode::T, "EXIT condition code used: {}",
+                             static_cast<u32>(cc));
+
+        switch (instr.flow.cond) {
+        case Tegra::Shader::FlowCondition::Always:
+            bb.push_back(Operation(OperationCode::Exit));
+            if (instr.pred.pred_index == static_cast<u64>(Tegra::Shader::Pred::UnusedIndex)) {
+                // If this is an unconditional exit then just end processing here,
+                // otherwise we have to account for the possibility of the condition
+                // not being met, so continue processing the next instruction.
+                pc = MAX_PROGRAM_LENGTH - 1;
+            }
+            break;
+
+        case Tegra::Shader::FlowCondition::Fcsm_Tr:
+            // TODO(bunnei): What is this used for? If we assume this conditon is not
+            // satisifed, dual vertex shaders in Farming Simulator make more sense
+            UNIMPLEMENTED_MSG("Skipping unknown FlowCondition::Fcsm_Tr");
+            break;
+
+        default:
+            UNIMPLEMENTED_MSG("Unhandled flow condition: {}",
+                              static_cast<u32>(instr.flow.cond.Value()));
+        }
+        break;
+    }
+    case OpCode::Id::KIL: {
+        UNIMPLEMENTED_IF(instr.flow.cond != Tegra::Shader::FlowCondition::Always);
+
+        const Tegra::Shader::ConditionCode cc = instr.flow_condition_code;
+        UNIMPLEMENTED_IF_MSG(cc != Tegra::Shader::ConditionCode::T, "KIL condition code used: {}",
+                             static_cast<u32>(cc));
+
+        bb.push_back(Operation(OperationCode::Discard));
+        break;
+    }
+    case OpCode::Id::MOV_SYS: {
+        switch (instr.sys20) {
+        case Tegra::Shader::SystemVariable::InvocationInfo: {
+            LOG_WARNING(HW_GPU, "MOV_SYS instruction with InvocationInfo is incomplete");
+            SetRegister(bb, instr.gpr0, Immediate(0u));
+            break;
+        }
+        case Tegra::Shader::SystemVariable::Ydirection: {
+            // Config pack's third value is Y_NEGATE's state.
+            SetRegister(bb, instr.gpr0, Operation(OperationCode::YNegate));
+            break;
+        }
+        default:
+            UNIMPLEMENTED_MSG("Unhandled system move: {}", static_cast<u32>(instr.sys20.Value()));
+        }
+        break;
+    }
+    case OpCode::Id::BRA: {
+        UNIMPLEMENTED_IF_MSG(instr.bra.constant_buffer != 0,
+                             "BRA with constant buffers are not implemented");
+
+        const u32 target = pc + instr.bra.GetBranchTarget();
+        const Node branch = Operation(OperationCode::Branch, Immediate(target));
+
+        const Tegra::Shader::ConditionCode cc = instr.flow_condition_code;
+        if (cc != Tegra::Shader::ConditionCode::T) {
+            bb.push_back(Conditional(GetConditionCode(cc), {branch}));
+        } else {
+            bb.push_back(branch);
+        }
+        break;
+    }
+    case OpCode::Id::SSY: {
+        UNIMPLEMENTED_IF_MSG(instr.bra.constant_buffer != 0,
+                             "Constant buffer flow is not supported");
+
+        // The SSY opcode tells the GPU where to re-converge divergent execution paths, it sets the
+        // target of the jump that the SYNC instruction will make. The SSY opcode has a similar
+        // structure to the BRA opcode.
+        const u32 target = pc + instr.bra.GetBranchTarget();
+        bb.push_back(Operation(OperationCode::PushFlowStack, Immediate(target)));
+        break;
+    }
+    case OpCode::Id::PBK: {
+        UNIMPLEMENTED_IF_MSG(instr.bra.constant_buffer != 0,
+                             "Constant buffer PBK is not supported");
+
+        // PBK pushes to a stack the address where BRK will jump to. This shares stack with SSY but
+        // using SYNC on a PBK address will kill the shader execution. We don't emulate this because
+        // it's very unlikely a driver will emit such invalid shader.
+        const u32 target = pc + instr.bra.GetBranchTarget();
+        bb.push_back(Operation(OperationCode::PushFlowStack, Immediate(target)));
+        break;
+    }
+    case OpCode::Id::SYNC: {
+        const Tegra::Shader::ConditionCode cc = instr.flow_condition_code;
+        UNIMPLEMENTED_IF_MSG(cc != Tegra::Shader::ConditionCode::T, "SYNC condition code used: {}",
+                             static_cast<u32>(cc));
+
+        // The SYNC opcode jumps to the address previously set by the SSY opcode
+        bb.push_back(Operation(OperationCode::PopFlowStack));
+        break;
+    }
+    case OpCode::Id::BRK: {
+        const Tegra::Shader::ConditionCode cc = instr.flow_condition_code;
+        UNIMPLEMENTED_IF_MSG(cc != Tegra::Shader::ConditionCode::T, "BRK condition code used: {}",
+                             static_cast<u32>(cc));
+
+        // The BRK opcode jumps to the address previously set by the PBK opcode
+        bb.push_back(Operation(OperationCode::PopFlowStack));
+        break;
+    }
+    case OpCode::Id::IPA: {
+        const auto& attribute = instr.attribute.fmt28;
+        const Tegra::Shader::IpaMode input_mode{instr.ipa.interp_mode.Value(),
+                                                instr.ipa.sample_mode.Value()};
+
+        const Node attr = GetInputAttribute(attribute.index, attribute.element, input_mode);
+        const Node value = GetSaturatedFloat(attr, instr.ipa.saturate);
+
+        SetRegister(bb, instr.gpr0, value);
+        break;
+    }
+    case OpCode::Id::OUT_R: {
+        UNIMPLEMENTED_IF_MSG(instr.gpr20.Value() != Register::ZeroIndex,
+                             "Stream buffer is not supported");
+
+        if (instr.out.emit) {
+            // gpr0 is used to store the next address and gpr8 contains the address to emit.
+            // Hardware uses pointers here but we just ignore it
+            bb.push_back(Operation(OperationCode::EmitVertex));
+            SetRegister(bb, instr.gpr0, Immediate(0));
+        }
+        if (instr.out.cut) {
+            bb.push_back(Operation(OperationCode::EndPrimitive));
+        }
+        break;
+    }
+    case OpCode::Id::ISBERD: {
+        UNIMPLEMENTED_IF(instr.isberd.o != 0);
+        UNIMPLEMENTED_IF(instr.isberd.skew != 0);
+        UNIMPLEMENTED_IF(instr.isberd.shift != Tegra::Shader::IsberdShift::None);
+        UNIMPLEMENTED_IF(instr.isberd.mode != Tegra::Shader::IsberdMode::None);
+        LOG_WARNING(HW_GPU, "ISBERD instruction is incomplete");
+        SetRegister(bb, instr.gpr0, GetRegister(instr.gpr8));
+        break;
+    }
+    case OpCode::Id::DEPBAR: {
+        LOG_WARNING(HW_GPU, "DEPBAR instruction is stubbed");
+        break;
+    }
+    default:
+        UNIMPLEMENTED_MSG("Unhandled instruction: {}", opcode->get().GetName());
+    }
+
+    return pc;
+}
+
+} // namespace VideoCommon::Shader

src/video_core/shader/decode/predicate_set_predicate.cpp

@@ -0,0 +1,67 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/assert.h"
+#include "common/common_types.h"
+#include "video_core/engines/shader_bytecode.h"
+#include "video_core/shader/shader_ir.h"
+
+namespace VideoCommon::Shader {
+
+using Tegra::Shader::Instruction;
+using Tegra::Shader::OpCode;
+using Tegra::Shader::Pred;
+
+u32 ShaderIR::DecodePredicateSetPredicate(BasicBlock& bb, const BasicBlock& code, u32 pc) {
+    const Instruction instr = {program_code[pc]};
+    const auto opcode = OpCode::Decode(instr);
+
+    switch (opcode->get().GetId()) {
+    case OpCode::Id::PSETP: {
+        const Node op_a = GetPredicate(instr.psetp.pred12, instr.psetp.neg_pred12 != 0);
+        const Node op_b = GetPredicate(instr.psetp.pred29, instr.psetp.neg_pred29 != 0);
+
+        // We can't use the constant predicate as destination.
+        ASSERT(instr.psetp.pred3 != static_cast<u64>(Pred::UnusedIndex));
+
+        const Node second_pred = GetPredicate(instr.psetp.pred39, instr.psetp.neg_pred39 != 0);
+
+        const OperationCode combiner = GetPredicateCombiner(instr.psetp.op);
+        const Node predicate = Operation(combiner, op_a, op_b);
+
+        // Set the primary predicate to the result of Predicate OP SecondPredicate
+        SetPredicate(bb, instr.psetp.pred3, Operation(combiner, predicate, second_pred));
+
+        if (instr.psetp.pred0 != static_cast<u64>(Pred::UnusedIndex)) {
+            // Set the secondary predicate to the result of !Predicate OP SecondPredicate, if
+            // enabled
+            SetPredicate(bb, instr.psetp.pred0,
+                         Operation(combiner, Operation(OperationCode::LogicalNegate, predicate),
+                                   second_pred));
+        }
+        break;
+    }
+    case OpCode::Id::CSETP: {
+        const Node pred = GetPredicate(instr.csetp.pred39, instr.csetp.neg_pred39 != 0);
+        const Node condition_code = GetConditionCode(instr.csetp.cc);
+
+        const OperationCode combiner = GetPredicateCombiner(instr.csetp.op);
+
+        if (instr.csetp.pred3 != static_cast<u64>(Pred::UnusedIndex)) {
+            SetPredicate(bb, instr.csetp.pred3, Operation(combiner, condition_code, pred));
+        }
+        if (instr.csetp.pred0 != static_cast<u64>(Pred::UnusedIndex)) {
+            const Node neg_cc = Operation(OperationCode::LogicalNegate, condition_code);
+            SetPredicate(bb, instr.csetp.pred0, Operation(combiner, neg_cc, pred));
+        }
+        break;
+    }
+    default:
+        UNIMPLEMENTED_MSG("Unhandled predicate instruction: {}", opcode->get().GetName());
+    }
+
+    return pc;
+}
+
+} // namespace VideoCommon::Shader

src/video_core/shader/decode/predicate_set_register.cpp

@@ -0,0 +1,46 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/assert.h"
+#include "common/common_types.h"
+#include "video_core/engines/shader_bytecode.h"
+#include "video_core/shader/shader_ir.h"
+
+namespace VideoCommon::Shader {
+
+using Tegra::Shader::Instruction;
+using Tegra::Shader::OpCode;
+
+u32 ShaderIR::DecodePredicateSetRegister(BasicBlock& bb, const BasicBlock& code, u32 pc) {
+    const Instruction instr = {program_code[pc]};
+    const auto opcode = OpCode::Decode(instr);
+
+    UNIMPLEMENTED_IF_MSG(instr.generates_cc,
+                         "Condition codes generation in PSET is not implemented");
+
+    const Node op_a = GetPredicate(instr.pset.pred12, instr.pset.neg_pred12 != 0);
+    const Node op_b = GetPredicate(instr.pset.pred29, instr.pset.neg_pred29 != 0);
+    const Node first_pred = Operation(GetPredicateCombiner(instr.pset.cond), op_a, op_b);
+
+    const Node second_pred = GetPredicate(instr.pset.pred39, instr.pset.neg_pred39 != 0);
+
+    const OperationCode combiner = GetPredicateCombiner(instr.pset.op);
+    const Node predicate = Operation(combiner, first_pred, second_pred);
+
+    const Node true_value = instr.pset.bf ? Immediate(1.0f) : Immediate(0xffffffff);
+    const Node false_value = instr.pset.bf ? Immediate(0.0f) : Immediate(0);
+    const Node value =
+        Operation(OperationCode::Select, PRECISE, predicate, true_value, false_value);
+
+    if (instr.pset.bf) {
+        SetInternalFlagsFromFloat(bb, value, instr.generates_cc);
+    } else {
+        SetInternalFlagsFromInteger(bb, value, instr.generates_cc);
+    }
+    SetRegister(bb, instr.gpr0, value);
+
+    return pc;
+}
+
+} // namespace VideoCommon::Shader

src/video_core/shader/decode/register_set_predicate.cpp

@@ -0,0 +1,51 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/assert.h"
+#include "common/common_types.h"
+#include "video_core/engines/shader_bytecode.h"
+#include "video_core/shader/shader_ir.h"
+
+namespace VideoCommon::Shader {
+
+using Tegra::Shader::Instruction;
+using Tegra::Shader::OpCode;
+
+u32 ShaderIR::DecodeRegisterSetPredicate(BasicBlock& bb, const BasicBlock& code, u32 pc) {
+    const Instruction instr = {program_code[pc]};
+    const auto opcode = OpCode::Decode(instr);
+
+    UNIMPLEMENTED_IF(instr.r2p.mode != Tegra::Shader::R2pMode::Pr);
+
+    const Node apply_mask = [&]() {
+        switch (opcode->get().GetId()) {
+        case OpCode::Id::R2P_IMM:
+            return Immediate(static_cast<u32>(instr.r2p.immediate_mask));
+        default:
+            UNREACHABLE();
+            return Immediate(static_cast<u32>(instr.r2p.immediate_mask));
+        }
+    }();
+    const Node mask = GetRegister(instr.gpr8);
+    const auto offset = static_cast<u32>(instr.r2p.byte) * 8;
+
+    constexpr u32 programmable_preds = 7;
+    for (u64 pred = 0; pred < programmable_preds; ++pred) {
+        const auto shift = static_cast<u32>(pred);
+
+        const Node apply_compare = BitfieldExtract(apply_mask, shift, 1);
+        const Node condition =
+            Operation(OperationCode::LogicalUNotEqual, apply_compare, Immediate(0));
+
+        const Node value_compare = BitfieldExtract(mask, offset + shift, 1);
+        const Node value = Operation(OperationCode::LogicalUNotEqual, value_compare, Immediate(0));
+
+        const Node code = Operation(OperationCode::LogicalAssign, GetPredicate(pred), value);
+        bb.push_back(Conditional(condition, {code}));
+    }
+
+    return pc;
+}
+
+} // namespace VideoCommon::Shader

src/video_core/shader/decode/shift.cpp

@@ -0,0 +1,55 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/assert.h"
+#include "common/common_types.h"
+#include "video_core/engines/shader_bytecode.h"
+#include "video_core/shader/shader_ir.h"
+
+namespace VideoCommon::Shader {
+
+using Tegra::Shader::Instruction;
+using Tegra::Shader::OpCode;
+
+u32 ShaderIR::DecodeShift(BasicBlock& bb, const BasicBlock& code, u32 pc) {
+    const Instruction instr = {program_code[pc]};
+    const auto opcode = OpCode::Decode(instr);
+
+    const Node op_a = GetRegister(instr.gpr8);
+    const Node op_b = [&]() {
+        if (instr.is_b_imm) {
+            return Immediate(instr.alu.GetSignedImm20_20());
+        } else if (instr.is_b_gpr) {
+            return GetRegister(instr.gpr20);
+        } else {
+            return GetConstBuffer(instr.cbuf34.index, instr.cbuf34.offset);
+        }
+    }();
+
+    switch (opcode->get().GetId()) {
+    case OpCode::Id::SHR_C:
+    case OpCode::Id::SHR_R:
+    case OpCode::Id::SHR_IMM: {
+        const Node value = SignedOperation(OperationCode::IArithmeticShiftRight,
+                                           instr.shift.is_signed, PRECISE, op_a, op_b);
+        SetInternalFlagsFromInteger(bb, value, instr.generates_cc);
+        SetRegister(bb, instr.gpr0, value);
+        break;
+    }
+    case OpCode::Id::SHL_C:
+    case OpCode::Id::SHL_R:
+    case OpCode::Id::SHL_IMM: {
+        const Node value = Operation(OperationCode::ILogicalShiftLeft, PRECISE, op_a, op_b);
+        SetInternalFlagsFromInteger(bb, value, instr.generates_cc);
+        SetRegister(bb, instr.gpr0, value);
+        break;
+    }
+    default:
+        UNIMPLEMENTED_MSG("Unhandled shift instruction: {}", opcode->get().GetName());
+    }
+
+    return pc;
+}
+
+} // namespace VideoCommon::Shader