304 files changed, 10946 insertions, 5115 deletions

diff --git a/src/audio_core/audio_out.cpp b/src/audio_core/audio_out.cpp
index 50d2a1ed3..8619a3f03 100644
--- a/src/audio_core/audio_out.cpp
+++ b/src/audio_core/audio_out.cpp
@@ -26,14 +26,15 @@ static Stream::Format ChannelsToStreamFormat(u32 num_channels) {
     return {};
 }
 
-StreamPtr AudioOut::OpenStream(u32 sample_rate, u32 num_channels, std::string&& name,
+StreamPtr AudioOut::OpenStream(Core::Timing::CoreTiming& core_timing, u32 sample_rate,
+                               u32 num_channels, std::string&& name,
                                Stream::ReleaseCallback&& release_callback) {
     if (!sink) {
         sink = CreateSinkFromID(Settings::values.sink_id, Settings::values.audio_device_id);
     }
 
     return std::make_shared<Stream>(
-        sample_rate, ChannelsToStreamFormat(num_channels), std::move(release_callback),
+        core_timing, sample_rate, ChannelsToStreamFormat(num_channels), std::move(release_callback),
         sink->AcquireSinkStream(sample_rate, num_channels, name), std::move(name));
 }
 
diff --git a/src/audio_core/audio_out.h b/src/audio_core/audio_out.h
index df9607ac7..b07588287 100644
--- a/src/audio_core/audio_out.h
+++ b/src/audio_core/audio_out.h
@@ -13,6 +13,10 @@
 #include "audio_core/stream.h"
 #include "common/common_types.h"
 
+namespace Core::Timing {
+class CoreTiming;
+}
+
 namespace AudioCore {
 
 /**
@@ -21,8 +25,8 @@ namespace AudioCore {
 class AudioOut {
 public:
     /// Opens a new audio stream
-    StreamPtr OpenStream(u32 sample_rate, u32 num_channels, std::string&& name,
-                         Stream::ReleaseCallback&& release_callback);
+    StreamPtr OpenStream(Core::Timing::CoreTiming& core_timing, u32 sample_rate, u32 num_channels,
+                         std::string&& name, Stream::ReleaseCallback&& release_callback);
 
     /// Returns a vector of recently released buffers specified by tag for the specified stream
     std::vector<Buffer::Tag> GetTagsAndReleaseBuffers(StreamPtr stream, std::size_t max_count);
diff --git a/src/audio_core/audio_renderer.cpp b/src/audio_core/audio_renderer.cpp
index 00c026511..9a0939883 100644
--- a/src/audio_core/audio_renderer.cpp
+++ b/src/audio_core/audio_renderer.cpp
@@ -8,6 +8,7 @@
 #include "audio_core/codec.h"
 #include "common/assert.h"
 #include "common/logging/log.h"
+#include "core/core.h"
 #include "core/hle/kernel/writable_event.h"
 #include "core/memory.h"
 
@@ -71,14 +72,14 @@ private:
     EffectOutStatus out_status{};
     EffectInStatus info{};
 };
-AudioRenderer::AudioRenderer(AudioRendererParameter params,
+AudioRenderer::AudioRenderer(Core::Timing::CoreTiming& core_timing, AudioRendererParameter params,
                              Kernel::SharedPtr<Kernel::WritableEvent> buffer_event)
     : worker_params{params}, buffer_event{buffer_event}, voices(params.voice_count),
       effects(params.effect_count) {
 
     audio_out = std::make_unique<AudioCore::AudioOut>();
-    stream = audio_out->OpenStream(STREAM_SAMPLE_RATE, STREAM_NUM_CHANNELS, "AudioRenderer",
-                                   [=]() { buffer_event->Signal(); });
+    stream = audio_out->OpenStream(core_timing, STREAM_SAMPLE_RATE, STREAM_NUM_CHANNELS,
+                                   "AudioRenderer", [=]() { buffer_event->Signal(); });
     audio_out->StartStream(stream);
 
     QueueMixedBuffer(0);
diff --git a/src/audio_core/audio_renderer.h b/src/audio_core/audio_renderer.h
index 7826881bf..b2e5d336c 100644
--- a/src/audio_core/audio_renderer.h
+++ b/src/audio_core/audio_renderer.h
@@ -14,6 +14,10 @@
 #include "common/swap.h"
 #include "core/hle/kernel/object.h"
 
+namespace Core::Timing {
+class CoreTiming;
+}
+
 namespace Kernel {
 class WritableEvent;
 }
@@ -42,16 +46,18 @@ struct AudioRendererParameter {
     u32_le sample_rate;
     u32_le sample_count;
     u32_le mix_buffer_count;
-    u32_le unknown_c;
+    u32_le submix_count;
     u32_le voice_count;
     u32_le sink_count;
     u32_le effect_count;
-    u32_le unknown_1c;
-    u8 unknown_20;
-    INSERT_PADDING_BYTES(3);
+    u32_le performance_frame_count;
+    u8 is_voice_drop_enabled;
+    u8 unknown_21;
+    u8 unknown_22;
+    u8 execution_mode;
     u32_le splitter_count;
-    u32_le unknown_2c;
-    INSERT_PADDING_WORDS(1);
+    u32_le num_splitter_send_channels;
+    u32_le unknown_30;
     u32_le revision;
 };
 static_assert(sizeof(AudioRendererParameter) == 52, "AudioRendererParameter is an invalid size");
@@ -208,7 +214,7 @@ static_assert(sizeof(UpdateDataHeader) == 0x40, "UpdateDataHeader has wrong size
 
 class AudioRenderer {
 public:
-    AudioRenderer(AudioRendererParameter params,
+    AudioRenderer(Core::Timing::CoreTiming& core_timing, AudioRendererParameter params,
                   Kernel::SharedPtr<Kernel::WritableEvent> buffer_event);
     ~AudioRenderer();
 
diff --git a/src/audio_core/buffer.h b/src/audio_core/buffer.h
index a323b23ec..5ee09e9aa 100644
--- a/src/audio_core/buffer.h
+++ b/src/audio_core/buffer.h
@@ -21,7 +21,7 @@ public:
     Buffer(Tag tag, std::vector<s16>&& samples) : tag{tag}, samples{std::move(samples)} {}
 
     /// Returns the raw audio data for the buffer
-    std::vector<s16>& Samples() {
+    std::vector<s16>& GetSamples() {
         return samples;
     }
 
diff --git a/src/audio_core/codec.cpp b/src/audio_core/codec.cpp
index 454de798b..c5a0d98ce 100644
--- a/src/audio_core/codec.cpp
+++ b/src/audio_core/codec.cpp
@@ -68,8 +68,8 @@ std::vector<s16> DecodeADPCM(const u8* const data, std::size_t size, const ADPCM
         }
     }
 
-    state.yn1 = yn1;
-    state.yn2 = yn2;
+    state.yn1 = static_cast<s16>(yn1);
+    state.yn2 = static_cast<s16>(yn2);
 
     return ret;
 }
diff --git a/src/audio_core/cubeb_sink.cpp b/src/audio_core/cubeb_sink.cpp
index 097328901..7047ed9cf 100644
--- a/src/audio_core/cubeb_sink.cpp
+++ b/src/audio_core/cubeb_sink.cpp
@@ -12,6 +12,10 @@
 #include "common/ring_buffer.h"
 #include "core/settings.h"
 
+#ifdef _WIN32
+#include <objbase.h>
+#endif
+
 namespace AudioCore {
 
 class CubebSinkStream final : public SinkStream {
@@ -46,7 +50,7 @@ public:
         }
     }
 
-    ~CubebSinkStream() {
+    ~CubebSinkStream() override {
         if (!ctx) {
             return;
         }
@@ -75,11 +79,11 @@ public:
         queue.Push(samples);
     }
 
-    std::size_t SamplesInQueue(u32 num_channels) const override {
+    std::size_t SamplesInQueue(u32 channel_count) const override {
         if (!ctx)
             return 0;
 
-        return queue.Size() / num_channels;
+        return queue.Size() / channel_count;
     }
 
     void Flush() override {
@@ -98,7 +102,7 @@ private:
     u32 num_channels{};
 
     Common::RingBuffer<s16, 0x10000> queue;
-    std::array<s16, 2> last_frame;
+    std::array<s16, 2> last_frame{};
     std::atomic<bool> should_flush{};
     TimeStretcher time_stretch;
 
@@ -108,6 +112,11 @@ private:
 };
 
 CubebSink::CubebSink(std::string_view target_device_name) {
+    // Cubeb requires COM to be initialized on the thread calling cubeb_init on Windows
+#ifdef _WIN32
+    com_init_result = CoInitializeEx(nullptr, COINIT_MULTITHREADED);
+#endif
+
     if (cubeb_init(&ctx, "yuzu", nullptr) != CUBEB_OK) {
         LOG_CRITICAL(Audio_Sink, "cubeb_init failed");
         return;
@@ -142,6 +151,12 @@ CubebSink::~CubebSink() {
     }
 
     cubeb_destroy(ctx);
+
+#ifdef _WIN32
+    if (SUCCEEDED(com_init_result)) {
+        CoUninitialize();
+    }
+#endif
 }
 
 SinkStream& CubebSink::AcquireSinkStream(u32 sample_rate, u32 num_channels,
diff --git a/src/audio_core/cubeb_sink.h b/src/audio_core/cubeb_sink.h
index efb9d1634..7ce850f47 100644
--- a/src/audio_core/cubeb_sink.h
+++ b/src/audio_core/cubeb_sink.h
@@ -25,6 +25,10 @@ private:
     cubeb* ctx{};
     cubeb_devid output_device{};
     std::vector<SinkStreamPtr> sink_streams;
+
+#ifdef _WIN32
+    u32 com_init_result = 0;
+#endif
 };
 
 std::vector<std::string> ListCubebSinkDevices();
diff --git a/src/audio_core/stream.cpp b/src/audio_core/stream.cpp
index 4ce2d374e..4b66a6786 100644
--- a/src/audio_core/stream.cpp
+++ b/src/audio_core/stream.cpp
@@ -32,12 +32,12 @@ u32 Stream::GetNumChannels() const {
     return {};
 }
 
-Stream::Stream(u32 sample_rate, Format format, ReleaseCallback&& release_callback,
-               SinkStream& sink_stream, std::string&& name_)
+Stream::Stream(Core::Timing::CoreTiming& core_timing, u32 sample_rate, Format format,
+               ReleaseCallback&& release_callback, SinkStream& sink_stream, std::string&& name_)
     : sample_rate{sample_rate}, format{format}, release_callback{std::move(release_callback)},
-      sink_stream{sink_stream}, name{std::move(name_)} {
+      sink_stream{sink_stream}, core_timing{core_timing}, name{std::move(name_)} {
 
-    release_event = CoreTiming::RegisterEvent(
+    release_event = core_timing.RegisterEvent(
         name, [this](u64 userdata, int cycles_late) { ReleaseActiveBuffer(); });
 }
 
@@ -57,7 +57,7 @@ Stream::State Stream::GetState() const {
 
 s64 Stream::GetBufferReleaseCycles(const Buffer& buffer) const {
     const std::size_t num_samples{buffer.GetSamples().size() / GetNumChannels()};
-    return CoreTiming::usToCycles((static_cast<u64>(num_samples) * 1000000) / sample_rate);
+    return Core::Timing::usToCycles((static_cast<u64>(num_samples) * 1000000) / sample_rate);
 }
 
 static void VolumeAdjustSamples(std::vector<s16>& samples) {
@@ -95,11 +95,11 @@ void Stream::PlayNextBuffer() {
     active_buffer = queued_buffers.front();
     queued_buffers.pop();
 
-    VolumeAdjustSamples(active_buffer->Samples());
+    VolumeAdjustSamples(active_buffer->GetSamples());
 
     sink_stream.EnqueueSamples(GetNumChannels(), active_buffer->GetSamples());
 
-    CoreTiming::ScheduleEventThreadsafe(GetBufferReleaseCycles(*active_buffer), release_event, {});
+    core_timing.ScheduleEventThreadsafe(GetBufferReleaseCycles(*active_buffer), release_event, {});
 }
 
 void Stream::ReleaseActiveBuffer() {
diff --git a/src/audio_core/stream.h b/src/audio_core/stream.h
index aebfeb51d..05071243b 100644
--- a/src/audio_core/stream.h
+++ b/src/audio_core/stream.h
@@ -13,9 +13,10 @@
 #include "audio_core/buffer.h"
 #include "common/common_types.h"
 
-namespace CoreTiming {
+namespace Core::Timing {
+class CoreTiming;
 struct EventType;
-}
+} // namespace Core::Timing
 
 namespace AudioCore {
 
@@ -42,8 +43,8 @@ public:
     /// Callback function type, used to change guest state on a buffer being released
     using ReleaseCallback = std::function<void()>;
 
-    Stream(u32 sample_rate, Format format, ReleaseCallback&& release_callback,
-           SinkStream& sink_stream, std::string&& name_);
+    Stream(Core::Timing::CoreTiming& core_timing, u32 sample_rate, Format format,
+           ReleaseCallback&& release_callback, SinkStream& sink_stream, std::string&& name_);
 
     /// Plays the audio stream
     void Play();
@@ -91,16 +92,17 @@ private:
     /// Gets the number of core cycles when the specified buffer will be released
     s64 GetBufferReleaseCycles(const Buffer& buffer) const;
 
-    u32 sample_rate;                        ///< Sample rate of the stream
-    Format format;                          ///< Format of the stream
-    ReleaseCallback release_callback;       ///< Buffer release callback for the stream
-    State state{State::Stopped};            ///< Playback state of the stream
-    CoreTiming::EventType* release_event{}; ///< Core timing release event for the stream
-    BufferPtr active_buffer;                ///< Actively playing buffer in the stream
-    std::queue<BufferPtr> queued_buffers;   ///< Buffers queued to be played in the stream
-    std::queue<BufferPtr> released_buffers; ///< Buffers recently released from the stream
-    SinkStream& sink_stream;                ///< Output sink for the stream
-    std::string name;                       ///< Name of the stream, must be unique
+    u32 sample_rate;                          ///< Sample rate of the stream
+    Format format;                            ///< Format of the stream
+    ReleaseCallback release_callback;         ///< Buffer release callback for the stream
+    State state{State::Stopped};              ///< Playback state of the stream
+    Core::Timing::EventType* release_event{}; ///< Core timing release event for the stream
+    BufferPtr active_buffer;                  ///< Actively playing buffer in the stream
+    std::queue<BufferPtr> queued_buffers;     ///< Buffers queued to be played in the stream
+    std::queue<BufferPtr> released_buffers;   ///< Buffers recently released from the stream
+    SinkStream& sink_stream;                  ///< Output sink for the stream
+    Core::Timing::CoreTiming& core_timing;    ///< Core timing instance.
+    std::string name;                         ///< Name of the stream, must be unique
 };
 
 using StreamPtr = std::shared_ptr<Stream>;
diff --git a/src/common/CMakeLists.txt b/src/common/CMakeLists.txt
index 845626fc5..43ae8a9e7 100644
--- a/src/common/CMakeLists.txt
+++ b/src/common/CMakeLists.txt
@@ -1,42 +1,70 @@
-# Generate cpp with Git revision from template
-# Also if this is a CI build, add the build name (ie: Nightly, Canary) to the scm_rev file as well
-set(REPO_NAME "")
-set(BUILD_VERSION "0")
-if ($ENV{CI})
-  if ($ENV{TRAVIS})
+# Add a custom command to generate a new shader_cache_version hash when any of the following files change
+# NOTE: This is an approximation of what files affect shader generation, its possible something else
+# could affect the result, but much more unlikely than the following files. Keeping a list of files
+# like this allows for much better caching since it doesn't force the user to recompile binary shaders every update
+set(VIDEO_CORE "${CMAKE_SOURCE_DIR}/src/video_core")
+if (DEFINED ENV{CI})
+  if (DEFINED ENV{TRAVIS})
     set(BUILD_REPOSITORY $ENV{TRAVIS_REPO_SLUG})
     set(BUILD_TAG $ENV{TRAVIS_TAG})
-  elseif($ENV{APPVEYOR})
+  elseif(DEFINED ENV{APPVEYOR})
     set(BUILD_REPOSITORY $ENV{APPVEYOR_REPO_NAME})
     set(BUILD_TAG $ENV{APPVEYOR_REPO_TAG_NAME})
   endif()
-  # regex capture the string nightly or canary into CMAKE_MATCH_1
-  string(REGEX MATCH "yuzu-emu/yuzu-?(.*)" OUTVAR ${BUILD_REPOSITORY})
-  if (${CMAKE_MATCH_COUNT} GREATER 0)
-    # capitalize the first letter of each word in the repo name.
-    string(REPLACE "-" ";" REPO_NAME_LIST ${CMAKE_MATCH_1})
-    foreach(WORD ${REPO_NAME_LIST})
-      string(SUBSTRING ${WORD} 0 1 FIRST_LETTER)
-      string(SUBSTRING ${WORD} 1 -1 REMAINDER)
-      string(TOUPPER ${FIRST_LETTER} FIRST_LETTER)
-      set(REPO_NAME "${REPO_NAME}${FIRST_LETTER}${REMAINDER}")
-    endforeach()
-    if (BUILD_TAG)
-      string(REGEX MATCH "${CMAKE_MATCH_1}-([0-9]+)" OUTVAR ${BUILD_TAG})
-      if (${CMAKE_MATCH_COUNT} GREATER 0)
-        set(BUILD_VERSION ${CMAKE_MATCH_1})
-      endif()
-      if (BUILD_VERSION)
-        # This leaves a trailing space on the last word, but we actually want that
-        # because of how it's styled in the title bar.
-        set(BUILD_FULLNAME "${REPO_NAME} ${BUILD_VERSION} ")
-      else()
-        set(BUILD_FULLNAME "")
-      endif()
-    endif()
-  endif()
 endif()
-configure_file("${CMAKE_CURRENT_SOURCE_DIR}/scm_rev.cpp.in" "${CMAKE_CURRENT_SOURCE_DIR}/scm_rev.cpp" @ONLY)
+add_custom_command(OUTPUT scm_rev.cpp
+    COMMAND ${CMAKE_COMMAND}
+      -DSRC_DIR="${CMAKE_SOURCE_DIR}"
+      -DBUILD_REPOSITORY="${BUILD_REPOSITORY}"
+      -DBUILD_TAG="${BUILD_TAG}"
+      -P "${CMAKE_SOURCE_DIR}/CMakeModules/GenerateSCMRev.cmake"
+    DEPENDS
+      # WARNING! It was too much work to try and make a common location for this list,
+      # so if you need to change it, please update CMakeModules/GenerateSCMRev.cmake as well
+      "${VIDEO_CORE}/renderer_opengl/gl_shader_cache.cpp"
+      "${VIDEO_CORE}/renderer_opengl/gl_shader_cache.h"
+      "${VIDEO_CORE}/renderer_opengl/gl_shader_decompiler.cpp"
+      "${VIDEO_CORE}/renderer_opengl/gl_shader_decompiler.h"
+      "${VIDEO_CORE}/renderer_opengl/gl_shader_disk_cache.cpp"
+      "${VIDEO_CORE}/renderer_opengl/gl_shader_disk_cache.h"
+      "${VIDEO_CORE}/renderer_opengl/gl_shader_gen.cpp"
+      "${VIDEO_CORE}/renderer_opengl/gl_shader_gen.h"
+      "${VIDEO_CORE}/shader/decode/arithmetic.cpp"
+      "${VIDEO_CORE}/shader/decode/arithmetic_half.cpp"
+      "${VIDEO_CORE}/shader/decode/arithmetic_half_immediate.cpp"
+      "${VIDEO_CORE}/shader/decode/arithmetic_immediate.cpp"
+      "${VIDEO_CORE}/shader/decode/arithmetic_integer.cpp"
+      "${VIDEO_CORE}/shader/decode/arithmetic_integer_immediate.cpp"
+      "${VIDEO_CORE}/shader/decode/bfe.cpp"
+      "${VIDEO_CORE}/shader/decode/bfi.cpp"
+      "${VIDEO_CORE}/shader/decode/conversion.cpp"
+      "${VIDEO_CORE}/shader/decode/ffma.cpp"
+      "${VIDEO_CORE}/shader/decode/float_set.cpp"
+      "${VIDEO_CORE}/shader/decode/float_set_predicate.cpp"
+      "${VIDEO_CORE}/shader/decode/half_set.cpp"
+      "${VIDEO_CORE}/shader/decode/half_set_predicate.cpp"
+      "${VIDEO_CORE}/shader/decode/hfma2.cpp"
+      "${VIDEO_CORE}/shader/decode/integer_set.cpp"
+      "${VIDEO_CORE}/shader/decode/integer_set_predicate.cpp"
+      "${VIDEO_CORE}/shader/decode/memory.cpp"
+      "${VIDEO_CORE}/shader/decode/texture.cpp"
+      "${VIDEO_CORE}/shader/decode/other.cpp"
+      "${VIDEO_CORE}/shader/decode/predicate_set_predicate.cpp"
+      "${VIDEO_CORE}/shader/decode/predicate_set_register.cpp"
+      "${VIDEO_CORE}/shader/decode/register_set_predicate.cpp"
+      "${VIDEO_CORE}/shader/decode/shift.cpp"
+      "${VIDEO_CORE}/shader/decode/video.cpp"
+      "${VIDEO_CORE}/shader/decode/xmad.cpp"
+      "${VIDEO_CORE}/shader/decode.cpp"
+      "${VIDEO_CORE}/shader/shader_ir.cpp"
+      "${VIDEO_CORE}/shader/shader_ir.h"
+      "${VIDEO_CORE}/shader/track.cpp"
+      # and also check that the scm_rev files haven't changed
+      "${CMAKE_CURRENT_SOURCE_DIR}/scm_rev.cpp.in"
+      "${CMAKE_CURRENT_SOURCE_DIR}/scm_rev.h"
+      # technically we should regenerate if the git version changed, but its not worth the effort imo
+      "${CMAKE_SOURCE_DIR}/CMakeModules/GenerateSCMRev.cmake"
+)
 
 add_library(common STATIC
     alignment.h
@@ -64,10 +92,14 @@ add_library(common STATIC
     logging/text_formatter.cpp
     logging/text_formatter.h
     math_util.h
+    memory_hook.cpp
+    memory_hook.h
     microprofile.cpp
     microprofile.h
     microprofileui.h
     misc.cpp
+    page_table.cpp
+    page_table.h
     param_package.cpp
     param_package.h
     quaternion.h
@@ -86,6 +118,8 @@ add_library(common STATIC
     threadsafe_queue.h
     timer.cpp
     timer.h
+    uint128.cpp
+    uint128.h
     vector_math.h
     web_result.h
 )
diff --git a/src/common/bit_field.h b/src/common/bit_field.h
index 639efe22d..8e35c463f 100644
--- a/src/common/bit_field.h
+++ b/src/common/bit_field.h
@@ -162,9 +162,13 @@ public:
     BitField(T val) = delete;
     BitField& operator=(T val) = delete;
 
-    // Force default constructor to be created
-    // so that we can use this within unions
-    constexpr BitField() = default;
+    constexpr BitField() noexcept = default;
+
+    constexpr BitField(const BitField&) noexcept = default;
+    constexpr BitField& operator=(const BitField&) noexcept = default;
+
+    constexpr BitField(BitField&&) noexcept = default;
+    constexpr BitField& operator=(BitField&&) noexcept = default;
 
     constexpr FORCE_INLINE operator T() const {
         return Value();
diff --git a/src/common/color.h b/src/common/color.h
index 0379040be..3a2222077 100644
--- a/src/common/color.h
+++ b/src/common/color.h
@@ -55,36 +55,36 @@ constexpr u8 Convert8To6(u8 value) {
 /**
  * Decode a color stored in RGBA8 format
  * @param bytes Pointer to encoded source color
- * @return Result color decoded as Math::Vec4<u8>
+ * @return Result color decoded as Common::Vec4<u8>
  */
-inline Math::Vec4<u8> DecodeRGBA8(const u8* bytes) {
+inline Common::Vec4<u8> DecodeRGBA8(const u8* bytes) {
     return {bytes[3], bytes[2], bytes[1], bytes[0]};
 }
 
 /**
  * Decode a color stored in RGB8 format
  * @param bytes Pointer to encoded source color
- * @return Result color decoded as Math::Vec4<u8>
+ * @return Result color decoded as Common::Vec4<u8>
  */
-inline Math::Vec4<u8> DecodeRGB8(const u8* bytes) {
+inline Common::Vec4<u8> DecodeRGB8(const u8* bytes) {
     return {bytes[2], bytes[1], bytes[0], 255};
 }
 
 /**
  * Decode a color stored in RG8 (aka HILO8) format
  * @param bytes Pointer to encoded source color
- * @return Result color decoded as Math::Vec4<u8>
+ * @return Result color decoded as Common::Vec4<u8>
  */
-inline Math::Vec4<u8> DecodeRG8(const u8* bytes) {
+inline Common::Vec4<u8> DecodeRG8(const u8* bytes) {
     return {bytes[1], bytes[0], 0, 255};
 }
 
 /**
  * Decode a color stored in RGB565 format
  * @param bytes Pointer to encoded source color
- * @return Result color decoded as Math::Vec4<u8>
+ * @return Result color decoded as Common::Vec4<u8>
  */
-inline Math::Vec4<u8> DecodeRGB565(const u8* bytes) {
+inline Common::Vec4<u8> DecodeRGB565(const u8* bytes) {
     u16_le pixel;
     std::memcpy(&pixel, bytes, sizeof(pixel));
     return {Convert5To8((pixel >> 11) & 0x1F), Convert6To8((pixel >> 5) & 0x3F),
@@ -94,9 +94,9 @@ inline Math::Vec4<u8> DecodeRGB565(const u8* bytes) {
 /**
  * Decode a color stored in RGB5A1 format
  * @param bytes Pointer to encoded source color
- * @return Result color decoded as Math::Vec4<u8>
+ * @return Result color decoded as Common::Vec4<u8>
  */
-inline Math::Vec4<u8> DecodeRGB5A1(const u8* bytes) {
+inline Common::Vec4<u8> DecodeRGB5A1(const u8* bytes) {
     u16_le pixel;
     std::memcpy(&pixel, bytes, sizeof(pixel));
     return {Convert5To8((pixel >> 11) & 0x1F), Convert5To8((pixel >> 6) & 0x1F),
@@ -106,9 +106,9 @@ inline Math::Vec4<u8> DecodeRGB5A1(const u8* bytes) {
 /**
  * Decode a color stored in RGBA4 format
  * @param bytes Pointer to encoded source color
- * @return Result color decoded as Math::Vec4<u8>
+ * @return Result color decoded as Common::Vec4<u8>
  */
-inline Math::Vec4<u8> DecodeRGBA4(const u8* bytes) {
+inline Common::Vec4<u8> DecodeRGBA4(const u8* bytes) {
     u16_le pixel;
     std::memcpy(&pixel, bytes, sizeof(pixel));
     return {Convert4To8((pixel >> 12) & 0xF), Convert4To8((pixel >> 8) & 0xF),
@@ -138,9 +138,9 @@ inline u32 DecodeD24(const u8* bytes) {
 /**
  * Decode a depth value and a stencil value stored in D24S8 format
  * @param bytes Pointer to encoded source values
- * @return Resulting values stored as a Math::Vec2
+ * @return Resulting values stored as a Common::Vec2
  */
-inline Math::Vec2<u32> DecodeD24S8(const u8* bytes) {
+inline Common::Vec2<u32> DecodeD24S8(const u8* bytes) {
     return {static_cast<u32>((bytes[2] << 16) | (bytes[1] << 8) | bytes[0]), bytes[3]};
 }
 
@@ -149,7 +149,7 @@ inline Math::Vec2<u32> DecodeD24S8(const u8* bytes) {
  * @param color Source color to encode
  * @param bytes Destination pointer to store encoded color
  */
-inline void EncodeRGBA8(const Math::Vec4<u8>& color, u8* bytes) {
+inline void EncodeRGBA8(const Common::Vec4<u8>& color, u8* bytes) {
     bytes[3] = color.r();
     bytes[2] = color.g();
     bytes[1] = color.b();
@@ -161,7 +161,7 @@ inline void EncodeRGBA8(const Math::Vec4<u8>& color, u8* bytes) {
  * @param color Source color to encode
  * @param bytes Destination pointer to store encoded color
  */
-inline void EncodeRGB8(const Math::Vec4<u8>& color, u8* bytes) {
+inline void EncodeRGB8(const Common::Vec4<u8>& color, u8* bytes) {
     bytes[2] = color.r();
     bytes[1] = color.g();
     bytes[0] = color.b();
@@ -172,7 +172,7 @@ inline void EncodeRGB8(const Math::Vec4<u8>& color, u8* bytes) {
  * @param color Source color to encode
  * @param bytes Destination pointer to store encoded color
  */
-inline void EncodeRG8(const Math::Vec4<u8>& color, u8* bytes) {
+inline void EncodeRG8(const Common::Vec4<u8>& color, u8* bytes) {
     bytes[1] = color.r();
     bytes[0] = color.g();
 }
@@ -181,7 +181,7 @@ inline void EncodeRG8(const Math::Vec4<u8>& color, u8* bytes) {
  * @param color Source color to encode
  * @param bytes Destination pointer to store encoded color
  */
-inline void EncodeRGB565(const Math::Vec4<u8>& color, u8* bytes) {
+inline void EncodeRGB565(const Common::Vec4<u8>& color, u8* bytes) {
     const u16_le data =
         (Convert8To5(color.r()) << 11) | (Convert8To6(color.g()) << 5) | Convert8To5(color.b());
 
@@ -193,7 +193,7 @@ inline void EncodeRGB565(const Math::Vec4<u8>& color, u8* bytes) {
  * @param color Source color to encode
  * @param bytes Destination pointer to store encoded color
  */
-inline void EncodeRGB5A1(const Math::Vec4<u8>& color, u8* bytes) {
+inline void EncodeRGB5A1(const Common::Vec4<u8>& color, u8* bytes) {
     const u16_le data = (Convert8To5(color.r()) << 11) | (Convert8To5(color.g()) << 6) |
                         (Convert8To5(color.b()) << 1) | Convert8To1(color.a());
 
@@ -205,7 +205,7 @@ inline void EncodeRGB5A1(const Math::Vec4<u8>& color, u8* bytes) {
  * @param color Source color to encode
  * @param bytes Destination pointer to store encoded color
  */
-inline void EncodeRGBA4(const Math::Vec4<u8>& color, u8* bytes) {
+inline void EncodeRGBA4(const Common::Vec4<u8>& color, u8* bytes) {
     const u16 data = (Convert8To4(color.r()) << 12) | (Convert8To4(color.g()) << 8) |
                      (Convert8To4(color.b()) << 4) | Convert8To4(color.a());
 
diff --git a/src/common/common_paths.h b/src/common/common_paths.h
index 4f88de768..076752d3b 100644
--- a/src/common/common_paths.h
+++ b/src/common/common_paths.h
@@ -35,6 +35,7 @@
 #define KEYS_DIR "keys"
 #define LOAD_DIR "load"
 #define DUMP_DIR "dump"
+#define SHADER_DIR "shader"
 #define LOG_DIR "log"
 
 // Filenames
diff --git a/src/common/file_util.cpp b/src/common/file_util.cpp
index b52492da6..aecb66c32 100644
--- a/src/common/file_util.cpp
+++ b/src/common/file_util.cpp
@@ -710,6 +710,7 @@ const std::string& GetUserPath(UserPath path, const std::string& new_path) {
         paths.emplace(UserPath::NANDDir, user_path + NAND_DIR DIR_SEP);
         paths.emplace(UserPath::LoadDir, user_path + LOAD_DIR DIR_SEP);
         paths.emplace(UserPath::DumpDir, user_path + DUMP_DIR DIR_SEP);
+        paths.emplace(UserPath::ShaderDir, user_path + SHADER_DIR DIR_SEP);
         paths.emplace(UserPath::SysDataDir, user_path + SYSDATA_DIR DIR_SEP);
         paths.emplace(UserPath::KeysDir, user_path + KEYS_DIR DIR_SEP);
         // TODO: Put the logs in a better location for each OS
diff --git a/src/common/file_util.h b/src/common/file_util.h
index 571503d2a..38cc7f059 100644
--- a/src/common/file_util.h
+++ b/src/common/file_util.h
@@ -31,6 +31,7 @@ enum class UserPath {
     SDMCDir,
     LoadDir,
     DumpDir,
+    ShaderDir,
     SysDataDir,
     UserDir,
 };
diff --git a/src/common/logging/backend.cpp b/src/common/logging/backend.cpp
index 12f6d0114..4462ff3fb 100644
--- a/src/common/logging/backend.cpp
+++ b/src/common/logging/backend.cpp
@@ -39,10 +39,10 @@ public:
     Impl(Impl const&) = delete;
     const Impl& operator=(Impl const&) = delete;
 
-    void PushEntry(Entry e) {
-        std::lock_guard<std::mutex> lock(message_mutex);
-        message_queue.Push(std::move(e));
-        message_cv.notify_one();
+    void PushEntry(Class log_class, Level log_level, const char* filename, unsigned int line_num,
+                   const char* function, std::string message) {
+        message_queue.Push(
+            CreateEntry(log_class, log_level, filename, line_num, function, std::move(message)));
     }
 
     void AddBackend(std::unique_ptr<Backend> backend) {
@@ -86,15 +86,13 @@ private:
                 }
             };
             while (true) {
-                {
-                    std::unique_lock<std::mutex> lock(message_mutex);
-                    message_cv.wait(lock, [&] { return !running || message_queue.Pop(entry); });
-                }
-                if (!running) {
+                entry = message_queue.PopWait();
+                if (entry.final_entry) {
                     break;
                 }
                 write_logs(entry);
             }
+
             // Drain the logging queue. Only writes out up to MAX_LOGS_TO_WRITE to prevent a case
             // where a system is repeatedly spamming logs even on close.
             const int MAX_LOGS_TO_WRITE = filter.IsDebug() ? INT_MAX : 100;
@@ -106,18 +104,36 @@ private:
     }
 
     ~Impl() {
-        running = false;
-        message_cv.notify_one();
+        Entry entry;
+        entry.final_entry = true;
+        message_queue.Push(entry);
         backend_thread.join();
     }
 
-    std::atomic_bool running{true};
-    std::mutex message_mutex, writing_mutex;
-    std::condition_variable message_cv;
+    Entry CreateEntry(Class log_class, Level log_level, const char* filename, unsigned int line_nr,
+                      const char* function, std::string message) const {
+        using std::chrono::duration_cast;
+        using std::chrono::steady_clock;
+
+        Entry entry;
+        entry.timestamp =
+            duration_cast<std::chrono::microseconds>(steady_clock::now() - time_origin);
+        entry.log_class = log_class;
+        entry.log_level = log_level;
+        entry.filename = Common::TrimSourcePath(filename);
+        entry.line_num = line_nr;
+        entry.function = function;
+        entry.message = std::move(message);
+
+        return entry;
+    }
+
+    std::mutex writing_mutex;
     std::thread backend_thread;
     std::vector<std::unique_ptr<Backend>> backends;
     Common::MPSCQueue<Log::Entry> message_queue;
     Filter filter;
+    std::chrono::steady_clock::time_point time_origin{std::chrono::steady_clock::now()};
 };
 
 void ConsoleBackend::Write(const Entry& entry) {
@@ -232,6 +248,7 @@ void DebuggerBackend::Write(const Entry& entry) {
     CLS(Render)                                                                                    \
     SUB(Render, Software)                                                                          \
     SUB(Render, OpenGL)                                                                            \
+    SUB(Render, Vulkan)                                                                            \
     CLS(Audio)                                                                                     \
     SUB(Audio, DSP)                                                                                \
     SUB(Audio, Sink)                                                                               \
@@ -275,25 +292,6 @@ const char* GetLevelName(Level log_level) {
 #undef LVL
 }
 
-Entry CreateEntry(Class log_class, Level log_level, const char* filename, unsigned int line_nr,
-                  const char* function, std::string message) {
-    using std::chrono::duration_cast;
-    using std::chrono::steady_clock;
-
-    static steady_clock::time_point time_origin = steady_clock::now();
-
-    Entry entry;
-    entry.timestamp = duration_cast<std::chrono::microseconds>(steady_clock::now() - time_origin);
-    entry.log_class = log_class;
-    entry.log_level = log_level;
-    entry.filename = Common::TrimSourcePath(filename);
-    entry.line_num = line_nr;
-    entry.function = function;
-    entry.message = std::move(message);
-
-    return entry;
-}
-
 void SetGlobalFilter(const Filter& filter) {
     Impl::Instance().SetGlobalFilter(filter);
 }
@@ -318,9 +316,7 @@ void FmtLogMessageImpl(Class log_class, Level log_level, const char* filename,
     if (!filter.CheckMessage(log_class, log_level))
         return;
 
-    Entry entry =
-        CreateEntry(log_class, log_level, filename, line_num, function, fmt::vformat(format, args));
-
-    instance.PushEntry(std::move(entry));
+    instance.PushEntry(log_class, log_level, filename, line_num, function,
+                       fmt::vformat(format, args));
 }
 } // namespace Log
diff --git a/src/common/logging/backend.h b/src/common/logging/backend.h
index 91bb0c309..fca0267a1 100644
--- a/src/common/logging/backend.h
+++ b/src/common/logging/backend.h
@@ -27,6 +27,7 @@ struct Entry {
     unsigned int line_num;
     std::string function;
     std::string message;
+    bool final_entry = false;
 
     Entry() = default;
     Entry(Entry&& o) = default;
@@ -134,10 +135,6 @@ const char* GetLogClassName(Class log_class);
  */
 const char* GetLevelName(Level log_level);
 
-/// Creates a log entry by formatting the given source location, and message.
-Entry CreateEntry(Class log_class, Level log_level, const char* filename, unsigned int line_nr,
-                  const char* function, std::string message);
-
 /**
  * The global filter will prevent any messages from even being processed if they are filtered. Each
  * backend can have a filter, but if the level is lower than the global filter, the backend will
diff --git a/src/common/logging/log.h b/src/common/logging/log.h
index d4ec31ec3..8ed6d5050 100644
--- a/src/common/logging/log.h
+++ b/src/common/logging/log.h
@@ -112,6 +112,7 @@ enum class Class : ClassType {
     Render,            ///< Emulator video output and hardware acceleration
     Render_Software,   ///< Software renderer backend
     Render_OpenGL,     ///< OpenGL backend
+    Render_Vulkan,     ///< Vulkan backend
     Audio,             ///< Audio emulation
     Audio_DSP,         ///< The HLE implementation of the DSP
     Audio_Sink,        ///< Emulator audio output backend
diff --git a/src/common/math_util.h b/src/common/math_util.h
index 94b4394c5..cff3d48c5 100644
--- a/src/common/math_util.h
+++ b/src/common/math_util.h
@@ -7,7 +7,7 @@
 #include <cstdlib>
 #include <type_traits>
 
-namespace MathUtil {
+namespace Common {
 
 constexpr float PI = 3.14159265f;
 
@@ -41,4 +41,4 @@ struct Rectangle {
     }
 };
 
-} // namespace MathUtil
+} // namespace Common
diff --git a/src/core/memory_hook.cpp b/src/common/memory_hook.cpp
index c61c6c1fb..3986986d6 100644
--- a/src/core/memory_hook.cpp
+++ b/src/common/memory_hook.cpp
@@ -2,10 +2,10 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
-#include "core/memory_hook.h"
+#include "common/memory_hook.h"
 
-namespace Memory {
+namespace Common {
 
 MemoryHook::~MemoryHook() = default;
 
-} // namespace Memory
+} // namespace Common
diff --git a/src/core/memory_hook.h b/src/common/memory_hook.h
index 940777107..adaa4c2c5 100644
--- a/src/core/memory_hook.h
+++ b/src/common/memory_hook.h
@@ -9,7 +9,7 @@
 
 #include "common/common_types.h"
 
-namespace Memory {
+namespace Common {
 
 /**
  * Memory hooks have two purposes:
@@ -44,4 +44,4 @@ public:
 };
 
 using MemoryHookPointer = std::shared_ptr<MemoryHook>;
-} // namespace Memory
+} // namespace Common
diff --git a/src/common/page_table.cpp b/src/common/page_table.cpp
new file mode 100644
index 000000000..8eba1c3f1
--- /dev/null
+++ b/src/common/page_table.cpp
@@ -0,0 +1,29 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/page_table.h"
+
+namespace Common {
+
+PageTable::PageTable(std::size_t page_size_in_bits) : page_size_in_bits{page_size_in_bits} {}
+
+PageTable::~PageTable() = default;
+
+void PageTable::Resize(std::size_t address_space_width_in_bits) {
+    const std::size_t num_page_table_entries = 1ULL
+                                               << (address_space_width_in_bits - page_size_in_bits);
+
+    pointers.resize(num_page_table_entries);
+    attributes.resize(num_page_table_entries);
+
+    // The default is a 39-bit address space, which causes an initial 1GB allocation size. If the
+    // vector size is subsequently decreased (via resize), the vector might not automatically
+    // actually reallocate/resize its underlying allocation, which wastes up to ~800 MB for
+    // 36-bit titles. Call shrink_to_fit to reduce capacity to what's actually in use.
+
+    pointers.shrink_to_fit();
+    attributes.shrink_to_fit();
+}
+
+} // namespace Common
diff --git a/src/common/page_table.h b/src/common/page_table.h
new file mode 100644
index 000000000..8339f2890
--- /dev/null
+++ b/src/common/page_table.h
@@ -0,0 +1,80 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <vector>
+#include <boost/icl/interval_map.hpp>
+#include "common/common_types.h"
+#include "common/memory_hook.h"
+
+namespace Common {
+
+enum class PageType : u8 {
+    /// Page is unmapped and should cause an access error.
+    Unmapped,
+    /// Page is mapped to regular memory. This is the only type you can get pointers to.
+    Memory,
+    /// Page is mapped to regular memory, but also needs to check for rasterizer cache flushing and
+    /// invalidation
+    RasterizerCachedMemory,
+    /// Page is mapped to a I/O region. Writing and reading to this page is handled by functions.
+    Special,
+};
+
+struct SpecialRegion {
+    enum class Type {
+        DebugHook,
+        IODevice,
+    } type;
+
+    MemoryHookPointer handler;
+
+    bool operator<(const SpecialRegion& other) const {
+        return std::tie(type, handler) < std::tie(other.type, other.handler);
+    }
+
+    bool operator==(const SpecialRegion& other) const {
+        return std::tie(type, handler) == std::tie(other.type, other.handler);
+    }
+};
+
+/**
+ * A (reasonably) fast way of allowing switchable and remappable process address spaces. It loosely
+ * mimics the way a real CPU page table works.
+ */
+struct PageTable {
+    explicit PageTable(std::size_t page_size_in_bits);
+    ~PageTable();
+
+    /**
+     * Resizes the page table to be able to accomodate enough pages within
+     * a given address space.
+     *
+     * @param address_space_width_in_bits The address size width in bits.
+     */
+    void Resize(std::size_t address_space_width_in_bits);
+
+    /**
+     * Vector of memory pointers backing each page. An entry can only be non-null if the
+     * corresponding entry in the `attributes` vector is of type `Memory`.
+     */
+    std::vector<u8*> pointers;
+
+    /**
+     * Contains MMIO handlers that back memory regions whose entries in the `attribute` vector is
+     * of type `Special`.
+     */
+    boost::icl::interval_map<VAddr, std::set<SpecialRegion>> special_regions;
+
+    /**
+     * Vector of fine grained page attributes. If it is set to any value other than `Memory`, then
+     * the corresponding entry in `pointers` MUST be set to null.
+     */
+    std::vector<PageType> attributes;
+
+    const std::size_t page_size_in_bits{};
+};
+
+} // namespace Common
diff --git a/src/common/quaternion.h b/src/common/quaternion.h
index c528c0b68..370198ae0 100644
--- a/src/common/quaternion.h
+++ b/src/common/quaternion.h
@@ -6,12 +6,12 @@
 
 #include "common/vector_math.h"
 
-namespace Math {
+namespace Common {
 
 template <typename T>
 class Quaternion {
 public:
-    Math::Vec3<T> xyz;
+    Vec3<T> xyz;
     T w{};
 
     Quaternion<decltype(-T{})> Inverse() const {
@@ -38,12 +38,12 @@ public:
 };
 
 template <typename T>
-auto QuaternionRotate(const Quaternion<T>& q, const Math::Vec3<T>& v) {
+auto QuaternionRotate(const Quaternion<T>& q, const Vec3<T>& v) {
     return v + 2 * Cross(q.xyz, Cross(q.xyz, v) + v * q.w);
 }
 
-inline Quaternion<float> MakeQuaternion(const Math::Vec3<float>& axis, float angle) {
+inline Quaternion<float> MakeQuaternion(const Vec3<float>& axis, float angle) {
     return {axis * std::sin(angle / 2), std::cos(angle / 2)};
 }
 
-} // namespace Math
+} // namespace Common
diff --git a/src/common/scm_rev.cpp.in b/src/common/scm_rev.cpp.in
index 2b1727769..d69038f65 100644
--- a/src/common/scm_rev.cpp.in
+++ b/src/common/scm_rev.cpp.in
@@ -11,6 +11,7 @@
 #define BUILD_DATE   "@BUILD_DATE@"
 #define BUILD_FULLNAME "@BUILD_FULLNAME@"
 #define BUILD_VERSION "@BUILD_VERSION@"
+#define SHADER_CACHE_VERSION "@SHADER_CACHE_VERSION@"
 
 namespace Common {
 
@@ -21,6 +22,7 @@ const char g_build_name[]   = BUILD_NAME;
 const char g_build_date[]   = BUILD_DATE;
 const char g_build_fullname[] = BUILD_FULLNAME;
 const char g_build_version[]  = BUILD_VERSION;
+const char g_shader_cache_version[] = SHADER_CACHE_VERSION;
 
 } // namespace
 
diff --git a/src/common/scm_rev.h b/src/common/scm_rev.h
index af9a9daed..666bf0367 100644
--- a/src/common/scm_rev.h
+++ b/src/common/scm_rev.h
@@ -13,5 +13,6 @@ extern const char g_build_name[];
 extern const char g_build_date[];
 extern const char g_build_fullname[];
 extern const char g_build_version[];
+extern const char g_shader_cache_version[];
 
 } // namespace Common
diff --git a/src/common/swap.h b/src/common/swap.h
index 4b82865fe..b3eab1324 100644
--- a/src/common/swap.h
+++ b/src/common/swap.h
@@ -30,8 +30,8 @@
 #include <cstring>
 #include "common/common_types.h"
 
-// GCC 4.6+
-#if __GNUC__ >= 5 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 6)
+// GCC
+#ifdef __GNUC__
 
 #if __BYTE_ORDER__ && (__BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__) && !defined(COMMON_LITTLE_ENDIAN)
 #define COMMON_LITTLE_ENDIAN 1
@@ -40,7 +40,7 @@
 #endif
 
 // LLVM/clang
-#elif __clang__
+#elif defined(__clang__)
 
 #if __LITTLE_ENDIAN__ && !defined(COMMON_LITTLE_ENDIAN)
 #define COMMON_LITTLE_ENDIAN 1
diff --git a/src/common/thread_queue_list.h b/src/common/thread_queue_list.h
index e7594db68..791f99a8c 100644
--- a/src/common/thread_queue_list.h
+++ b/src/common/thread_queue_list.h
@@ -6,7 +6,6 @@
 
 #include <array>
 #include <deque>
-#include <boost/range/algorithm_ext/erase.hpp>
 
 namespace Common {
 
@@ -111,8 +110,9 @@ struct ThreadQueueList {
     }
 
     void remove(Priority priority, const T& thread_id) {
-        Queue* cur = &queues[priority];
-        boost::remove_erase(cur->data, thread_id);
+        Queue* const cur = &queues[priority];
+        const auto iter = std::remove(cur->data.begin(), cur->data.end(), thread_id);
+        cur->data.erase(iter, cur->data.end());
     }
 
     void rotate(Priority priority) {
diff --git a/src/common/threadsafe_queue.h b/src/common/threadsafe_queue.h
index edf13bc49..821e8536a 100644
--- a/src/common/threadsafe_queue.h
+++ b/src/common/threadsafe_queue.h
@@ -7,17 +7,17 @@
 // a simple lockless thread-safe,
 // single reader, single writer queue
 
-#include <algorithm>
 #include <atomic>
+#include <condition_variable>
 #include <cstddef>
 #include <mutex>
-#include "common/common_types.h"
+#include <utility>
 
 namespace Common {
-template <typename T, bool NeedSize = true>
+template <typename T>
 class SPSCQueue {
 public:
-    SPSCQueue() : size(0) {
+    SPSCQueue() {
         write_ptr = read_ptr = new ElementPtr();
     }
     ~SPSCQueue() {
@@ -25,13 +25,12 @@ public:
         delete read_ptr;
     }
 
-    u32 Size() const {
-        static_assert(NeedSize, "using Size() on FifoQueue without NeedSize");
+    std::size_t Size() const {
         return size.load();
     }
 
     bool Empty() const {
-        return !read_ptr->next.load();
+        return Size() == 0;
     }
 
     T& Front() const {
@@ -47,13 +46,14 @@ public:
         ElementPtr* new_ptr = new ElementPtr();
         write_ptr->next.store(new_ptr, std::memory_order_release);
         write_ptr = new_ptr;
-        if (NeedSize)
-            size++;
+        cv.notify_one();
+
+        ++size;
     }
 
     void Pop() {
-        if (NeedSize)
-            size--;
+        --size;
+
         ElementPtr* tmpptr = read_ptr;
         // advance the read pointer
         read_ptr = tmpptr->next.load();
@@ -66,8 +66,7 @@ public:
         if (Empty())
             return false;
 
-        if (NeedSize)
-            size--;
+        --size;
 
         ElementPtr* tmpptr = read_ptr;
         read_ptr = tmpptr->next.load(std::memory_order_acquire);
@@ -77,6 +76,16 @@ public:
         return true;
     }
 
+    T PopWait() {
+        if (Empty()) {
+            std::unique_lock<std::mutex> lock(cv_mutex);
+            cv.wait(lock, [this]() { return !Empty(); });
+        }
+        T t;
+        Pop(t);
+        return t;
+    }
+
     // not thread-safe
     void Clear() {
         size.store(0);
@@ -89,7 +98,7 @@ private:
     // and a pointer to the next ElementPtr
     class ElementPtr {
     public:
-        ElementPtr() : next(nullptr) {}
+        ElementPtr() {}
         ~ElementPtr() {
             ElementPtr* next_ptr = next.load();
 
@@ -98,21 +107,23 @@ private:
         }
 
         T current;
-        std::atomic<ElementPtr*> next;
+        std::atomic<ElementPtr*> next{nullptr};
     };
 
     ElementPtr* write_ptr;
     ElementPtr* read_ptr;
-    std::atomic<u32> size;
+    std::atomic_size_t size{0};
+    std::mutex cv_mutex;
+    std::condition_variable cv;
 };
 
 // a simple thread-safe,
 // single reader, multiple writer queue
 
-template <typename T, bool NeedSize = true>
+template <typename T>
 class MPSCQueue {
 public:
-    u32 Size() const {
+    std::size_t Size() const {
         return spsc_queue.Size();
     }
 
@@ -138,13 +149,17 @@ public:
         return spsc_queue.Pop(t);
     }
 
+    T PopWait() {
+        return spsc_queue.PopWait();
+    }
+
     // not thread-safe
     void Clear() {
         spsc_queue.Clear();
     }
 
 private:
-    SPSCQueue<T, NeedSize> spsc_queue;
+    SPSCQueue<T> spsc_queue;
     std::mutex write_lock;
 };
 } // namespace Common
diff --git a/src/common/uint128.cpp b/src/common/uint128.cpp
new file mode 100644
index 000000000..32bf56730
--- /dev/null
+++ b/src/common/uint128.cpp
@@ -0,0 +1,45 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#ifdef _MSC_VER
+#include <intrin.h>
+
+#pragma intrinsic(_umul128)
+#endif
+#include <cstring>
+#include "common/uint128.h"
+
+namespace Common {
+
+u128 Multiply64Into128(u64 a, u64 b) {
+    u128 result;
+#ifdef _MSC_VER
+    result[0] = _umul128(a, b, &result[1]);
+#else
+    unsigned __int128 tmp = a;
+    tmp *= b;
+    std::memcpy(&result, &tmp, sizeof(u128));
+#endif
+    return result;
+}
+
+std::pair<u64, u64> Divide128On32(u128 dividend, u32 divisor) {
+    u64 remainder = dividend[0] % divisor;
+    u64 accum = dividend[0] / divisor;
+    if (dividend[1] == 0)
+        return {accum, remainder};
+    // We ignore dividend[1] / divisor as that overflows
+    const u64 first_segment = (dividend[1] % divisor) << 32;
+    accum += (first_segment / divisor) << 32;
+    const u64 second_segment = (first_segment % divisor) << 32;
+    accum += (second_segment / divisor);
+    remainder += second_segment % divisor;
+    if (remainder >= divisor) {
+        accum++;
+        remainder -= divisor;
+    }
+    return {accum, remainder};
+}
+
+} // namespace Common
diff --git a/src/common/uint128.h b/src/common/uint128.h
new file mode 100644
index 000000000..a3be2a2cb
--- /dev/null
+++ b/src/common/uint128.h
@@ -0,0 +1,19 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <utility>
+#include "common/common_types.h"
+
+namespace Common {
+
+// This function multiplies 2 u64 values and produces a u128 value;
+u128 Multiply64Into128(u64 a, u64 b);
+
+// This function divides a u128 by a u32 value and produces two u64 values:
+// the result of division and the remainder
+std::pair<u64, u64> Divide128On32(u128 dividend, u32 divisor);
+
+} // namespace Common
diff --git a/src/common/vector_math.h b/src/common/vector_math.h
index 8feb49941..429485329 100644
--- a/src/common/vector_math.h
+++ b/src/common/vector_math.h
@@ -33,7 +33,7 @@
 #include <cmath>
 #include <type_traits>
 
-namespace Math {
+namespace Common {
 
 template <typename T>
 class Vec2;
@@ -690,4 +690,4 @@ constexpr Vec4<T> MakeVec(const T& x, const Vec3<T>& yzw) {
     return MakeVec(x, yzw[0], yzw[1], yzw[2]);
 }
 
-} // namespace Math
+} // namespace Common
diff --git a/src/core/CMakeLists.txt b/src/core/CMakeLists.txt
index f61bcd40d..16920e2e9 100644
--- a/src/core/CMakeLists.txt
+++ b/src/core/CMakeLists.txt
@@ -107,6 +107,8 @@ add_library(core STATIC
     hle/kernel/client_port.h
     hle/kernel/client_session.cpp
     hle/kernel/client_session.h
+    hle/kernel/code_set.cpp
+    hle/kernel/code_set.h
     hle/kernel/errors.h
     hle/kernel/handle_table.cpp
     hle/kernel/handle_table.h
@@ -217,6 +219,7 @@ add_library(core STATIC
     hle/service/audio/audren_u.h
     hle/service/audio/codecctl.cpp
     hle/service/audio/codecctl.h
+    hle/service/audio/errors.h
     hle/service/audio/hwopus.cpp
     hle/service/audio/hwopus.h
     hle/service/bcat/bcat.cpp
@@ -400,6 +403,10 @@ add_library(core STATIC
     hle/service/time/time.h
     hle/service/usb/usb.cpp
     hle/service/usb/usb.h
+    hle/service/vi/display/vi_display.cpp
+    hle/service/vi/display/vi_display.h
+    hle/service/vi/layer/vi_layer.cpp
+    hle/service/vi/layer/vi_layer.h
     hle/service/vi/vi.cpp
     hle/service/vi/vi.h
     hle/service/vi/vi_m.cpp
@@ -414,8 +421,6 @@ add_library(core STATIC
     loader/deconstructed_rom_directory.h
     loader/elf.cpp
     loader/elf.h
-    loader/linker.cpp
-    loader/linker.h
     loader/loader.cpp
     loader/loader.h
     loader/nax.cpp
@@ -432,8 +437,6 @@ add_library(core STATIC
     loader/xci.h
     memory.cpp
     memory.h
-    memory_hook.cpp
-    memory_hook.h
     memory_setup.h
     perf_stats.cpp
     perf_stats.h
diff --git a/src/core/arm/dynarmic/arm_dynarmic.cpp b/src/core/arm/dynarmic/arm_dynarmic.cpp
index afbda8d8b..4fdc12f11 100644
--- a/src/core/arm/dynarmic/arm_dynarmic.cpp
+++ b/src/core/arm/dynarmic/arm_dynarmic.cpp
@@ -12,6 +12,7 @@
 #include "core/core.h"
 #include "core/core_cpu.h"
 #include "core/core_timing.h"
+#include "core/core_timing_util.h"
 #include "core/gdbstub/gdbstub.h"
 #include "core/hle/kernel/process.h"
 #include "core/hle/kernel/svc.h"
@@ -112,14 +113,14 @@ public:
         // Always execute at least one tick.
         amortized_ticks = std::max<u64>(amortized_ticks, 1);
 
-        CoreTiming::AddTicks(amortized_ticks);
+        parent.core_timing.AddTicks(amortized_ticks);
         num_interpreted_instructions = 0;
     }
     u64 GetTicksRemaining() override {
-        return std::max(CoreTiming::GetDowncount(), 0);
+        return std::max(parent.core_timing.GetDowncount(), 0);
     }
     u64 GetCNTPCT() override {
-        return CoreTiming::GetTicks();
+        return Timing::CpuCyclesToClockCycles(parent.core_timing.GetTicks());
     }
 
     ARM_Dynarmic& parent;
@@ -151,7 +152,7 @@ std::unique_ptr<Dynarmic::A64::Jit> ARM_Dynarmic::MakeJit() const {
     config.tpidr_el0 = &cb->tpidr_el0;
     config.dczid_el0 = 4;
     config.ctr_el0 = 0x8444c004;
-    config.cntfrq_el0 = 19200000; // Value from fusee.
+    config.cntfrq_el0 = Timing::CNTFREQ;
 
     // Unpredictable instructions
     config.define_unpredictable_behaviour = true;
@@ -172,8 +173,10 @@ void ARM_Dynarmic::Step() {
     cb->InterpreterFallback(jit->GetPC(), 1);
 }
 
-ARM_Dynarmic::ARM_Dynarmic(ExclusiveMonitor& exclusive_monitor, std::size_t core_index)
-    : cb(std::make_unique<ARM_Dynarmic_Callbacks>(*this)), core_index{core_index},
+ARM_Dynarmic::ARM_Dynarmic(Timing::CoreTiming& core_timing, ExclusiveMonitor& exclusive_monitor,
+                           std::size_t core_index)
+    : cb(std::make_unique<ARM_Dynarmic_Callbacks>(*this)), inner_unicorn{core_timing},
+      core_index{core_index}, core_timing{core_timing},
       exclusive_monitor{dynamic_cast<DynarmicExclusiveMonitor&>(exclusive_monitor)} {
     ThreadContext ctx{};
     inner_unicorn.SaveContext(ctx);
diff --git a/src/core/arm/dynarmic/arm_dynarmic.h b/src/core/arm/dynarmic/arm_dynarmic.h
index 512bf8ce9..aada1e862 100644
--- a/src/core/arm/dynarmic/arm_dynarmic.h
+++ b/src/core/arm/dynarmic/arm_dynarmic.h
@@ -12,10 +12,14 @@
 #include "core/arm/exclusive_monitor.h"
 #include "core/arm/unicorn/arm_unicorn.h"
 
-namespace Memory {
+namespace Common {
 struct PageTable;
 }
 
+namespace Core::Timing {
+class CoreTiming;
+}
+
 namespace Core {
 
 class ARM_Dynarmic_Callbacks;
@@ -23,7 +27,8 @@ class DynarmicExclusiveMonitor;
 
 class ARM_Dynarmic final : public ARM_Interface {
 public:
-    ARM_Dynarmic(ExclusiveMonitor& exclusive_monitor, std::size_t core_index);
+    ARM_Dynarmic(Timing::CoreTiming& core_timing, ExclusiveMonitor& exclusive_monitor,
+                 std::size_t core_index);
     ~ARM_Dynarmic();
 
     void MapBackingMemory(VAddr address, std::size_t size, u8* memory,
@@ -62,9 +67,10 @@ private:
     ARM_Unicorn inner_unicorn;
 
     std::size_t core_index;
+    Timing::CoreTiming& core_timing;
     DynarmicExclusiveMonitor& exclusive_monitor;
 
-    Memory::PageTable* current_page_table = nullptr;
+    Common::PageTable* current_page_table = nullptr;
 };
 
 class DynarmicExclusiveMonitor final : public ExclusiveMonitor {
diff --git a/src/core/arm/unicorn/arm_unicorn.cpp b/src/core/arm/unicorn/arm_unicorn.cpp
index c455c81fb..a542a098b 100644
--- a/src/core/arm/unicorn/arm_unicorn.cpp
+++ b/src/core/arm/unicorn/arm_unicorn.cpp
@@ -72,7 +72,7 @@ static bool UnmappedMemoryHook(uc_engine* uc, uc_mem_type type, u64 addr, int si
     return {};
 }
 
-ARM_Unicorn::ARM_Unicorn() {
+ARM_Unicorn::ARM_Unicorn(Timing::CoreTiming& core_timing) : core_timing{core_timing} {
     CHECKED(uc_open(UC_ARCH_ARM64, UC_MODE_ARM, &uc));
 
     auto fpv = 3 << 20;
@@ -177,7 +177,7 @@ void ARM_Unicorn::Run() {
     if (GDBStub::IsServerEnabled()) {
         ExecuteInstructions(std::max(4000000, 0));
     } else {
-        ExecuteInstructions(std::max(CoreTiming::GetDowncount(), 0));
+        ExecuteInstructions(std::max(core_timing.GetDowncount(), 0));
     }
 }
 
@@ -190,7 +190,7 @@ MICROPROFILE_DEFINE(ARM_Jit_Unicorn, "ARM JIT", "Unicorn", MP_RGB(255, 64, 64));
 void ARM_Unicorn::ExecuteInstructions(int num_instructions) {
     MICROPROFILE_SCOPE(ARM_Jit_Unicorn);
     CHECKED(uc_emu_start(uc, GetPC(), 1ULL << 63, 0, num_instructions));
-    CoreTiming::AddTicks(num_instructions);
+    core_timing.AddTicks(num_instructions);
     if (GDBStub::IsServerEnabled()) {
         if (last_bkpt_hit) {
             uc_reg_write(uc, UC_ARM64_REG_PC, &last_bkpt.address);
diff --git a/src/core/arm/unicorn/arm_unicorn.h b/src/core/arm/unicorn/arm_unicorn.h
index 75761950b..dbd6955ea 100644
--- a/src/core/arm/unicorn/arm_unicorn.h
+++ b/src/core/arm/unicorn/arm_unicorn.h
@@ -9,12 +9,17 @@
 #include "core/arm/arm_interface.h"
 #include "core/gdbstub/gdbstub.h"
 
+namespace Core::Timing {
+class CoreTiming;
+}
+
 namespace Core {
 
 class ARM_Unicorn final : public ARM_Interface {
 public:
-    ARM_Unicorn();
+    explicit ARM_Unicorn(Timing::CoreTiming& core_timing);
     ~ARM_Unicorn();
+
     void MapBackingMemory(VAddr address, std::size_t size, u8* memory,
                           Kernel::VMAPermission perms) override;
     void UnmapMemory(VAddr address, std::size_t size) override;
@@ -43,6 +48,7 @@ public:
 
 private:
     uc_engine* uc{};
+    Timing::CoreTiming& core_timing;
     GDBStub::BreakpointAddress last_bkpt{};
     bool last_bkpt_hit;
 };
diff --git a/src/core/core.cpp b/src/core/core.cpp
index 572814e4b..89b3fb418 100644
--- a/src/core/core.cpp
+++ b/src/core/core.cpp
@@ -36,7 +36,8 @@
 #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/gpu_asynch.h"
+#include "video_core/gpu_synch.h"
 #include "video_core/renderer_base.h"
 #include "video_core/video_core.h"
 
@@ -78,6 +79,7 @@ FileSys::VirtualFile GetGameFileFromPath(const FileSys::VirtualFilesystem& vfs,
     return vfs->OpenFile(path, FileSys::Mode::Read);
 }
 struct System::Impl {
+    explicit Impl(System& system) : kernel{system} {}
 
     Cpu& CurrentCpuCore() {
         return cpu_core_manager.GetCurrentCore();
@@ -94,7 +96,7 @@ struct System::Impl {
     ResultStatus Init(System& system, Frontend::EmuWindow& emu_window) {
         LOG_DEBUG(HW_Memory, "initialized OK");
 
-        CoreTiming::Init();
+        core_timing.Initialize();
         kernel.Initialize();
 
         const auto current_time = std::chrono::duration_cast<std::chrono::seconds>(
@@ -114,24 +116,30 @@ struct System::Impl {
         if (web_browser == nullptr)
             web_browser = std::make_unique<Core::Frontend::DefaultWebBrowserApplet>();
 
-        auto main_process = Kernel::Process::Create(kernel, "main");
+        auto main_process = Kernel::Process::Create(system, "main");
         kernel.MakeCurrentProcess(main_process.get());
 
         telemetry_session = std::make_unique<Core::TelemetrySession>();
         service_manager = std::make_shared<Service::SM::ServiceManager>();
 
-        Service::Init(service_manager, *virtual_filesystem);
+        Service::Init(service_manager, system, *virtual_filesystem);
         GDBStub::Init();
 
-        renderer = VideoCore::CreateRenderer(emu_window);
+        renderer = VideoCore::CreateRenderer(emu_window, system);
         if (!renderer->Init()) {
             return ResultStatus::ErrorVideoCore;
         }
 
-        gpu_core = std::make_unique<Tegra::GPU>(renderer->Rasterizer());
+        is_powered_on = true;
+
+        if (Settings::values.use_asynchronous_gpu_emulation) {
+            gpu_core = std::make_unique<VideoCommon::GPUAsynch>(system, *renderer);
+        } else {
+            gpu_core = std::make_unique<VideoCommon::GPUSynch>(system, *renderer);
+        }
 
         cpu_core_manager.Initialize(system);
-        is_powered_on = true;
+
         LOG_DEBUG(Core, "Initialized OK");
 
         // Reset counters and set time origin to current frame
@@ -175,19 +183,20 @@ struct System::Impl {
             return static_cast<ResultStatus>(static_cast<u32>(ResultStatus::ErrorLoader) +
                                              static_cast<u32>(load_result));
         }
+
         status = ResultStatus::Success;
         return status;
     }
 
     void Shutdown() {
         // Log last frame performance stats
-        auto perf_results = GetAndResetPerfStats();
-        Telemetry().AddField(Telemetry::FieldType::Performance, "Shutdown_EmulationSpeed",
-                             perf_results.emulation_speed * 100.0);
-        Telemetry().AddField(Telemetry::FieldType::Performance, "Shutdown_Framerate",
-                             perf_results.game_fps);
-        Telemetry().AddField(Telemetry::FieldType::Performance, "Shutdown_Frametime",
-                             perf_results.frametime * 1000.0);
+        const auto perf_results = GetAndResetPerfStats();
+        telemetry_session->AddField(Telemetry::FieldType::Performance, "Shutdown_EmulationSpeed",
+                                    perf_results.emulation_speed * 100.0);
+        telemetry_session->AddField(Telemetry::FieldType::Performance, "Shutdown_Framerate",
+                                    perf_results.game_fps);
+        telemetry_session->AddField(Telemetry::FieldType::Performance, "Shutdown_Frametime",
+                                    perf_results.frametime * 1000.0);
 
         is_powered_on = false;
 
@@ -204,7 +213,7 @@ struct System::Impl {
 
         // Shutdown kernel and core timing
         kernel.Shutdown();
-        CoreTiming::Shutdown();
+        core_timing.Shutdown();
 
         // Close app loader
         app_loader.reset();
@@ -231,9 +240,10 @@ struct System::Impl {
     }
 
     PerfStatsResults GetAndResetPerfStats() {
-        return perf_stats.GetAndResetStats(CoreTiming::GetGlobalTimeUs());
+        return perf_stats.GetAndResetStats(core_timing.GetGlobalTimeUs());
     }
 
+    Timing::CoreTiming core_timing;
     Kernel::KernelCore kernel;
     /// RealVfsFilesystem instance
     FileSys::VirtualFilesystem virtual_filesystem;
@@ -263,7 +273,7 @@ struct System::Impl {
     Core::FrameLimiter frame_limiter;
 };
 
-System::System() : impl{std::make_unique<Impl>()} {}
+System::System() : impl{std::make_unique<Impl>(*this)} {}
 System::~System() = default;
 
 Cpu& System::CurrentCpuCore() {
@@ -395,6 +405,14 @@ const Kernel::KernelCore& System::Kernel() const {
     return impl->kernel;
 }
 
+Timing::CoreTiming& System::CoreTiming() {
+    return impl->core_timing;
+}
+
+const Timing::CoreTiming& System::CoreTiming() const {
+    return impl->core_timing;
+}
+
 Core::PerfStats& System::GetPerfStats() {
     return impl->perf_stats;
 }
diff --git a/src/core/core.h b/src/core/core.h
index 511a5ad3a..ba76a41d8 100644
--- a/src/core/core.h
+++ b/src/core/core.h
@@ -47,6 +47,10 @@ namespace VideoCore {
 class RendererBase;
 } // namespace VideoCore
 
+namespace Core::Timing {
+class CoreTiming;
+}
+
 namespace Core {
 
 class ARM_Interface;
@@ -205,6 +209,12 @@ public:
     /// Provides a constant pointer to the current process.
     const Kernel::Process* CurrentProcess() const;
 
+    /// Provides a reference to the core timing instance.
+    Timing::CoreTiming& CoreTiming();
+
+    /// Provides a constant reference to the core timing instance.
+    const Timing::CoreTiming& CoreTiming() const;
+
     /// Provides a reference to the kernel instance.
     Kernel::KernelCore& Kernel();
 
@@ -283,10 +293,6 @@ inline ARM_Interface& CurrentArmInterface() {
     return System::GetInstance().CurrentArmInterface();
 }
 
-inline TelemetrySession& Telemetry() {
-    return System::GetInstance().TelemetrySession();
-}
-
 inline Kernel::Process* CurrentProcess() {
     return System::GetInstance().CurrentProcess();
 }
diff --git a/src/core/core_cpu.cpp b/src/core/core_cpu.cpp
index fffda8a99..1eefed6d0 100644
--- a/src/core/core_cpu.cpp
+++ b/src/core/core_cpu.cpp
@@ -11,6 +11,7 @@
 #endif
 #include "core/arm/exclusive_monitor.h"
 #include "core/arm/unicorn/arm_unicorn.h"
+#include "core/core.h"
 #include "core/core_cpu.h"
 #include "core/core_timing.h"
 #include "core/hle/kernel/scheduler.h"
@@ -49,20 +50,21 @@ bool CpuBarrier::Rendezvous() {
     return false;
 }
 
-Cpu::Cpu(ExclusiveMonitor& exclusive_monitor, CpuBarrier& cpu_barrier, std::size_t core_index)
-    : cpu_barrier{cpu_barrier}, core_index{core_index} {
+Cpu::Cpu(System& system, ExclusiveMonitor& exclusive_monitor, CpuBarrier& cpu_barrier,
+         std::size_t core_index)
+    : cpu_barrier{cpu_barrier}, core_timing{system.CoreTiming()}, core_index{core_index} {
     if (Settings::values.use_cpu_jit) {
 #ifdef ARCHITECTURE_x86_64
-        arm_interface = std::make_unique<ARM_Dynarmic>(exclusive_monitor, core_index);
+        arm_interface = std::make_unique<ARM_Dynarmic>(core_timing, exclusive_monitor, core_index);
 #else
         arm_interface = std::make_unique<ARM_Unicorn>();
         LOG_WARNING(Core, "CPU JIT requested, but Dynarmic not available");
 #endif
     } else {
-        arm_interface = std::make_unique<ARM_Unicorn>();
+        arm_interface = std::make_unique<ARM_Unicorn>(core_timing);
     }
 
-    scheduler = std::make_unique<Kernel::Scheduler>(*arm_interface);
+    scheduler = std::make_unique<Kernel::Scheduler>(system, *arm_interface);
 }
 
 Cpu::~Cpu() = default;
@@ -93,14 +95,14 @@ void Cpu::RunLoop(bool tight_loop) {
 
         if (IsMainCore()) {
             // TODO(Subv): Only let CoreTiming idle if all 4 cores are idling.
-            CoreTiming::Idle();
-            CoreTiming::Advance();
+            core_timing.Idle();
+            core_timing.Advance();
         }
 
         PrepareReschedule();
     } else {
         if (IsMainCore()) {
-            CoreTiming::Advance();
+            core_timing.Advance();
         }
 
         if (tight_loop) {
diff --git a/src/core/core_cpu.h b/src/core/core_cpu.h
index 1d2bdc6cd..7589beb8c 100644
--- a/src/core/core_cpu.h
+++ b/src/core/core_cpu.h
@@ -16,6 +16,14 @@ class Scheduler;
 }
 
 namespace Core {
+class System;
+}
+
+namespace Core::Timing {
+class CoreTiming;
+}
+
+namespace Core {
 
 class ARM_Interface;
 class ExclusiveMonitor;
@@ -41,7 +49,8 @@ private:
 
 class Cpu {
 public:
-    Cpu(ExclusiveMonitor& exclusive_monitor, CpuBarrier& cpu_barrier, std::size_t core_index);
+    Cpu(System& system, ExclusiveMonitor& exclusive_monitor, CpuBarrier& cpu_barrier,
+        std::size_t core_index);
     ~Cpu();
 
     void RunLoop(bool tight_loop = true);
@@ -82,6 +91,7 @@ private:
     std::unique_ptr<ARM_Interface> arm_interface;
     CpuBarrier& cpu_barrier;
     std::unique_ptr<Kernel::Scheduler> scheduler;
+    Timing::CoreTiming& core_timing;
 
     std::atomic<bool> reschedule_pending = false;
     std::size_t core_index;
diff --git a/src/core/core_timing.cpp b/src/core/core_timing.cpp
index 7953c8720..a0dd5db24 100644
--- a/src/core/core_timing.cpp
+++ b/src/core/core_timing.cpp
@@ -8,149 +8,98 @@
 #include <mutex>
 #include <string>
 #include <tuple>
-#include <unordered_map>
-#include <vector>
+
 #include "common/assert.h"
 #include "common/thread.h"
-#include "common/threadsafe_queue.h"
 #include "core/core_timing_util.h"
 
-namespace CoreTiming {
-
-static s64 global_timer;
-static int slice_length;
-static int downcount;
+namespace Core::Timing {
 
-struct EventType {
-    TimedCallback callback;
-    const std::string* name;
-};
+constexpr int MAX_SLICE_LENGTH = 20000;
 
-struct Event {
+struct CoreTiming::Event {
     s64 time;
     u64 fifo_order;
     u64 userdata;
     const EventType* type;
-};
-
-// Sort by time, unless the times are the same, in which case sort by the order added to the queue
-static bool operator>(const Event& left, const Event& right) {
-    return std::tie(left.time, left.fifo_order) > std::tie(right.time, right.fifo_order);
-}
-
-static bool operator<(const Event& left, const Event& right) {
-    return std::tie(left.time, left.fifo_order) < std::tie(right.time, right.fifo_order);
-}
-
-// unordered_map stores each element separately as a linked list node so pointers to elements
-// remain stable regardless of rehashes/resizing.
-static std::unordered_map<std::string, EventType> event_types;
 
-// The queue is a min-heap using std::make_heap/push_heap/pop_heap.
-// We don't use std::priority_queue because we need to be able to serialize, unserialize and
-// erase arbitrary events (RemoveEvent()) regardless of the queue order. These aren't accomodated
-// by the standard adaptor class.
-static std::vector<Event> event_queue;
-static u64 event_fifo_id;
-// the queue for storing the events from other threads threadsafe until they will be added
-// to the event_queue by the emu thread
-static Common::MPSCQueue<Event, false> ts_queue;
-
-// the queue for unscheduling the events from other threads threadsafe
-static Common::MPSCQueue<std::pair<const EventType*, u64>, false> unschedule_queue;
-
-constexpr int MAX_SLICE_LENGTH = 20000;
-
-static s64 idled_cycles;
-
-// Are we in a function that has been called from Advance()
-// If events are sheduled from a function that gets called from Advance(),
-// don't change slice_length and downcount.
-static bool is_global_timer_sane;
-
-static EventType* ev_lost = nullptr;
-
-static void EmptyTimedCallback(u64 userdata, s64 cyclesLate) {}
-
-EventType* RegisterEvent(const std::string& name, TimedCallback callback) {
-    // check for existing type with same name.
-    // we want event type names to remain unique so that we can use them for serialization.
-    ASSERT_MSG(event_types.find(name) == event_types.end(),
-               "CoreTiming Event \"{}\" is already registered. Events should only be registered "
-               "during Init to avoid breaking save states.",
-               name.c_str());
+    // Sort by time, unless the times are the same, in which case sort by
+    // the order added to the queue
+    friend bool operator>(const Event& left, const Event& right) {
+        return std::tie(left.time, left.fifo_order) > std::tie(right.time, right.fifo_order);
+    }
 
-    auto info = event_types.emplace(name, EventType{callback, nullptr});
-    EventType* event_type = &info.first->second;
-    event_type->name = &info.first->first;
-    return event_type;
-}
+    friend bool operator<(const Event& left, const Event& right) {
+        return std::tie(left.time, left.fifo_order) < std::tie(right.time, right.fifo_order);
+    }
+};
 
-void UnregisterAllEvents() {
-    ASSERT_MSG(event_queue.empty(), "Cannot unregister events with events pending");
-    event_types.clear();
-}
+CoreTiming::CoreTiming() = default;
+CoreTiming::~CoreTiming() = default;
 
-void Init() {
+void CoreTiming::Initialize() {
     downcount = MAX_SLICE_LENGTH;
     slice_length = MAX_SLICE_LENGTH;
     global_timer = 0;
     idled_cycles = 0;
 
-    // The time between CoreTiming being intialized and the first call to Advance() is considered
+    // The time between CoreTiming being initialized and the first call to Advance() is considered
     // the slice boundary between slice -1 and slice 0. Dispatcher loops must call Advance() before
     // executing the first cycle of each slice to prepare the slice length and downcount for
     // that slice.
     is_global_timer_sane = true;
 
     event_fifo_id = 0;
-    ev_lost = RegisterEvent("_lost_event", &EmptyTimedCallback);
+
+    const auto empty_timed_callback = [](u64, s64) {};
+    ev_lost = RegisterEvent("_lost_event", empty_timed_callback);
 }
 
-void Shutdown() {
+void CoreTiming::Shutdown() {
     MoveEvents();
     ClearPendingEvents();
     UnregisterAllEvents();
 }
 
-// This should only be called from the CPU thread. If you are calling
-// it from any other thread, you are doing something evil
-u64 GetTicks() {
-    u64 ticks = static_cast<u64>(global_timer);
-    if (!is_global_timer_sane) {
-        ticks += slice_length - downcount;
-    }
-    return ticks;
-}
-
-void AddTicks(u64 ticks) {
-    downcount -= static_cast<int>(ticks);
-}
+EventType* CoreTiming::RegisterEvent(const std::string& name, TimedCallback callback) {
+    // check for existing type with same name.
+    // we want event type names to remain unique so that we can use them for serialization.
+    ASSERT_MSG(event_types.find(name) == event_types.end(),
+               "CoreTiming Event \"{}\" is already registered. Events should only be registered "
+               "during Init to avoid breaking save states.",
+               name.c_str());
 
-u64 GetIdleTicks() {
-    return static_cast<u64>(idled_cycles);
+    auto info = event_types.emplace(name, EventType{callback, nullptr});
+    EventType* event_type = &info.first->second;
+    event_type->name = &info.first->first;
+    return event_type;
 }
 
-void ClearPendingEvents() {
-    event_queue.clear();
+void CoreTiming::UnregisterAllEvents() {
+    ASSERT_MSG(event_queue.empty(), "Cannot unregister events with events pending");
+    event_types.clear();
 }
 
-void ScheduleEvent(s64 cycles_into_future, const EventType* event_type, u64 userdata) {
+void CoreTiming::ScheduleEvent(s64 cycles_into_future, const EventType* event_type, u64 userdata) {
     ASSERT(event_type != nullptr);
-    s64 timeout = GetTicks() + cycles_into_future;
+    const s64 timeout = GetTicks() + cycles_into_future;
+
     // If this event needs to be scheduled before the next advance(), force one early
-    if (!is_global_timer_sane)
+    if (!is_global_timer_sane) {
         ForceExceptionCheck(cycles_into_future);
+    }
+
     event_queue.emplace_back(Event{timeout, event_fifo_id++, userdata, event_type});
     std::push_heap(event_queue.begin(), event_queue.end(), std::greater<>());
 }
 
-void ScheduleEventThreadsafe(s64 cycles_into_future, const EventType* event_type, u64 userdata) {
+void CoreTiming::ScheduleEventThreadsafe(s64 cycles_into_future, const EventType* event_type,
+                                         u64 userdata) {
     ts_queue.Push(Event{global_timer + cycles_into_future, 0, userdata, event_type});
 }
 
-void UnscheduleEvent(const EventType* event_type, u64 userdata) {
-    auto itr = std::remove_if(event_queue.begin(), event_queue.end(), [&](const Event& e) {
+void CoreTiming::UnscheduleEvent(const EventType* event_type, u64 userdata) {
+    const auto itr = std::remove_if(event_queue.begin(), event_queue.end(), [&](const Event& e) {
         return e.type == event_type && e.userdata == userdata;
     });
 
@@ -161,13 +110,33 @@ void UnscheduleEvent(const EventType* event_type, u64 userdata) {
     }
 }
 
-void UnscheduleEventThreadsafe(const EventType* event_type, u64 userdata) {
+void CoreTiming::UnscheduleEventThreadsafe(const EventType* event_type, u64 userdata) {
     unschedule_queue.Push(std::make_pair(event_type, userdata));
 }
 
-void RemoveEvent(const EventType* event_type) {
-    auto itr = std::remove_if(event_queue.begin(), event_queue.end(),
-                              [&](const Event& e) { return e.type == event_type; });
+u64 CoreTiming::GetTicks() const {
+    u64 ticks = static_cast<u64>(global_timer);
+    if (!is_global_timer_sane) {
+        ticks += slice_length - downcount;
+    }
+    return ticks;
+}
+
+u64 CoreTiming::GetIdleTicks() const {
+    return static_cast<u64>(idled_cycles);
+}
+
+void CoreTiming::AddTicks(u64 ticks) {
+    downcount -= static_cast<int>(ticks);
+}
+
+void CoreTiming::ClearPendingEvents() {
+    event_queue.clear();
+}
+
+void CoreTiming::RemoveEvent(const EventType* event_type) {
+    const auto itr = std::remove_if(event_queue.begin(), event_queue.end(),
+                                    [&](const Event& e) { return e.type == event_type; });
 
     // Removing random items breaks the invariant so we have to re-establish it.
     if (itr != event_queue.end()) {
@@ -176,22 +145,24 @@ void RemoveEvent(const EventType* event_type) {
     }
 }
 
-void RemoveNormalAndThreadsafeEvent(const EventType* event_type) {
+void CoreTiming::RemoveNormalAndThreadsafeEvent(const EventType* event_type) {
     MoveEvents();
     RemoveEvent(event_type);
 }
 
-void ForceExceptionCheck(s64 cycles) {
+void CoreTiming::ForceExceptionCheck(s64 cycles) {
     cycles = std::max<s64>(0, cycles);
-    if (downcount > cycles) {
-        // downcount is always (much) smaller than MAX_INT so we can safely cast cycles to an int
-        // here. Account for cycles already executed by adjusting the g.slice_length
-        slice_length -= downcount - static_cast<int>(cycles);
-        downcount = static_cast<int>(cycles);
+    if (downcount <= cycles) {
+        return;
     }
+
+    // downcount is always (much) smaller than MAX_INT so we can safely cast cycles to an int
+    // here. Account for cycles already executed by adjusting the g.slice_length
+    slice_length -= downcount - static_cast<int>(cycles);
+    downcount = static_cast<int>(cycles);
 }
 
-void MoveEvents() {
+void CoreTiming::MoveEvents() {
     for (Event ev; ts_queue.Pop(ev);) {
         ev.fifo_order = event_fifo_id++;
         event_queue.emplace_back(std::move(ev));
@@ -199,13 +170,13 @@ void MoveEvents() {
     }
 }
 
-void Advance() {
+void CoreTiming::Advance() {
     MoveEvents();
     for (std::pair<const EventType*, u64> ev; unschedule_queue.Pop(ev);) {
         UnscheduleEvent(ev.first, ev.second);
     }
 
-    int cycles_executed = slice_length - downcount;
+    const int cycles_executed = slice_length - downcount;
     global_timer += cycles_executed;
     slice_length = MAX_SLICE_LENGTH;
 
@@ -229,17 +200,17 @@ void Advance() {
     downcount = slice_length;
 }
 
-void Idle() {
+void CoreTiming::Idle() {
     idled_cycles += downcount;
     downcount = 0;
 }
 
-std::chrono::microseconds GetGlobalTimeUs() {
+std::chrono::microseconds CoreTiming::GetGlobalTimeUs() const {
     return std::chrono::microseconds{GetTicks() * 1000000 / BASE_CLOCK_RATE};
 }
 
-int GetDowncount() {
+int CoreTiming::GetDowncount() const {
     return downcount;
 }
 
-} // namespace CoreTiming
+} // namespace Core::Timing
diff --git a/src/core/core_timing.h b/src/core/core_timing.h
index 9ed757bd7..59163bae1 100644
--- a/src/core/core_timing.h
+++ b/src/core/core_timing.h
@@ -4,6 +4,27 @@
 
 #pragma once
 
+#include <chrono>
+#include <functional>
+#include <string>
+#include <unordered_map>
+#include <vector>
+#include "common/common_types.h"
+#include "common/threadsafe_queue.h"
+
+namespace Core::Timing {
+
+/// A callback that may be scheduled for a particular core timing event.
+using TimedCallback = std::function<void(u64 userdata, int cycles_late)>;
+
+/// Contains the characteristics of a particular event.
+struct EventType {
+    /// The event's callback function.
+    TimedCallback callback;
+    /// A pointer to the name of the event.
+    const std::string* name;
+};
+
 /**
  * This is a system to schedule events into the emulated machine's future. Time is measured
  * in main CPU clock cycles.
@@ -16,80 +37,120 @@
  * inside callback:
  *   ScheduleEvent(periodInCycles - cyclesLate, callback, "whatever")
  */
-
-#include <chrono>
-#include <functional>
-#include <string>
-#include "common/common_types.h"
-
-namespace CoreTiming {
-
-struct EventType;
-
-using TimedCallback = std::function<void(u64 userdata, int cycles_late)>;
-
-/**
- * CoreTiming begins at the boundary of timing slice -1. An initial call to Advance() is
- * required to end slice -1 and start slice 0 before the first cycle of code is executed.
- */
-void Init();
-void Shutdown();
-
-/**
- * This should only be called from the emu thread, if you are calling it any other thread, you are
- * doing something evil
- */
-u64 GetTicks();
-u64 GetIdleTicks();
-void AddTicks(u64 ticks);
-
-/**
- * Returns the event_type identifier. if name is not unique, it will assert.
- */
-EventType* RegisterEvent(const std::string& name, TimedCallback callback);
-void UnregisterAllEvents();
-
-/**
- * After the first Advance, the slice lengths and the downcount will be reduced whenever an event
- * is scheduled earlier than the current values.
- * Scheduling from a callback will not update the downcount until the Advance() completes.
- */
-void ScheduleEvent(s64 cycles_into_future, const EventType* event_type, u64 userdata = 0);
-
-/**
- * This is to be called when outside of hle threads, such as the graphics thread, wants to
- * schedule things to be executed on the main thread.
- * Not that this doesn't change slice_length and thus events scheduled by this might be called
- * with a delay of up to MAX_SLICE_LENGTH
- */
-void ScheduleEventThreadsafe(s64 cycles_into_future, const EventType* event_type, u64 userdata);
-
-void UnscheduleEvent(const EventType* event_type, u64 userdata);
-void UnscheduleEventThreadsafe(const EventType* event_type, u64 userdata);
-
-/// We only permit one event of each type in the queue at a time.
-void RemoveEvent(const EventType* event_type);
-void RemoveNormalAndThreadsafeEvent(const EventType* event_type);
-
-/** Advance must be called at the beginning of dispatcher loops, not the end. Advance() ends
- * the previous timing slice and begins the next one, you must Advance from the previous
- * slice to the current one before executing any cycles. CoreTiming starts in slice -1 so an
- * Advance() is required to initialize the slice length before the first cycle of emulated
- * instructions is executed.
- */
-void Advance();
-void MoveEvents();
-
-/// Pretend that the main CPU has executed enough cycles to reach the next event.
-void Idle();
-
-/// Clear all pending events. This should ONLY be done on exit.
-void ClearPendingEvents();
-
-void ForceExceptionCheck(s64 cycles);
-
-std::chrono::microseconds GetGlobalTimeUs();
-
-int GetDowncount();
-
-} // namespace CoreTiming
+class CoreTiming {
+public:
+    CoreTiming();
+    ~CoreTiming();
+
+    CoreTiming(const CoreTiming&) = delete;
+    CoreTiming(CoreTiming&&) = delete;
+
+    CoreTiming& operator=(const CoreTiming&) = delete;
+    CoreTiming& operator=(CoreTiming&&) = delete;
+
+    /// CoreTiming begins at the boundary of timing slice -1. An initial call to Advance() is
+    /// required to end slice - 1 and start slice 0 before the first cycle of code is executed.
+    void Initialize();
+
+    /// Tears down all timing related functionality.
+    void Shutdown();
+
+    /// Registers a core timing event with the given name and callback.
+    ///
+    /// @param name     The name of the core timing event to register.
+    /// @param callback The callback to execute for the event.
+    ///
+    /// @returns An EventType instance representing the registered event.
+    ///
+    /// @pre The name of the event being registered must be unique among all
+    ///      registered events.
+    ///
+    EventType* RegisterEvent(const std::string& name, TimedCallback callback);
+
+    /// Unregisters all registered events thus far.
+    void UnregisterAllEvents();
+
+    /// After the first Advance, the slice lengths and the downcount will be reduced whenever an
+    /// event is scheduled earlier than the current values.
+    ///
+    /// Scheduling from a callback will not update the downcount until the Advance() completes.
+    void ScheduleEvent(s64 cycles_into_future, const EventType* event_type, u64 userdata = 0);
+
+    /// This is to be called when outside of hle threads, such as the graphics thread, wants to
+    /// schedule things to be executed on the main thread.
+    ///
+    /// @note This doesn't change slice_length and thus events scheduled by this might be
+    /// called with a delay of up to MAX_SLICE_LENGTH
+    void ScheduleEventThreadsafe(s64 cycles_into_future, const EventType* event_type,
+                                 u64 userdata = 0);
+
+    void UnscheduleEvent(const EventType* event_type, u64 userdata);
+    void UnscheduleEventThreadsafe(const EventType* event_type, u64 userdata);
+
+    /// We only permit one event of each type in the queue at a time.
+    void RemoveEvent(const EventType* event_type);
+    void RemoveNormalAndThreadsafeEvent(const EventType* event_type);
+
+    void ForceExceptionCheck(s64 cycles);
+
+    /// This should only be called from the emu thread, if you are calling it any other thread,
+    /// you are doing something evil
+    u64 GetTicks() const;
+
+    u64 GetIdleTicks() const;
+
+    void AddTicks(u64 ticks);
+
+    /// Advance must be called at the beginning of dispatcher loops, not the end. Advance() ends
+    /// the previous timing slice and begins the next one, you must Advance from the previous
+    /// slice to the current one before executing any cycles. CoreTiming starts in slice -1 so an
+    /// Advance() is required to initialize the slice length before the first cycle of emulated
+    /// instructions is executed.
+    void Advance();
+
+    /// Pretend that the main CPU has executed enough cycles to reach the next event.
+    void Idle();
+
+    std::chrono::microseconds GetGlobalTimeUs() const;
+
+    int GetDowncount() const;
+
+private:
+    struct Event;
+
+    /// Clear all pending events. This should ONLY be done on exit.
+    void ClearPendingEvents();
+    void MoveEvents();
+
+    s64 global_timer = 0;
+    s64 idled_cycles = 0;
+    int slice_length = 0;
+    int downcount = 0;
+
+    // Are we in a function that has been called from Advance()
+    // If events are scheduled from a function that gets called from Advance(),
+    // don't change slice_length and downcount.
+    bool is_global_timer_sane = false;
+
+    // The queue is a min-heap using std::make_heap/push_heap/pop_heap.
+    // We don't use std::priority_queue because we need to be able to serialize, unserialize and
+    // erase arbitrary events (RemoveEvent()) regardless of the queue order. These aren't
+    // accomodated by the standard adaptor class.
+    std::vector<Event> event_queue;
+    u64 event_fifo_id = 0;
+
+    // Stores each element separately as a linked list node so pointers to elements
+    // remain stable regardless of rehashes/resizing.
+    std::unordered_map<std::string, EventType> event_types;
+
+    // The queue for storing the events from other threads threadsafe until they will be added
+    // to the event_queue by the emu thread
+    Common::MPSCQueue<Event> ts_queue;
+
+    // The queue for unscheduling the events from other threads threadsafe
+    Common::MPSCQueue<std::pair<const EventType*, u64>> unschedule_queue;
+
+    EventType* ev_lost = nullptr;
+};
+
+} // namespace Core::Timing
diff --git a/src/core/core_timing_util.cpp b/src/core/core_timing_util.cpp
index 73dea4edb..7942f30d6 100644
--- a/src/core/core_timing_util.cpp
+++ b/src/core/core_timing_util.cpp
@@ -7,8 +7,9 @@
 #include <cinttypes>
 #include <limits>
 #include "common/logging/log.h"
+#include "common/uint128.h"
 
-namespace CoreTiming {
+namespace Core::Timing {
 
 constexpr u64 MAX_VALUE_TO_MULTIPLY = std::numeric_limits<s64>::max() / BASE_CLOCK_RATE;
 
@@ -60,4 +61,9 @@ s64 nsToCycles(u64 ns) {
     return (BASE_CLOCK_RATE * static_cast<s64>(ns)) / 1000000000;
 }
 
-} // namespace CoreTiming
+u64 CpuCyclesToClockCycles(u64 ticks) {
+    const u128 temporal = Common::Multiply64Into128(ticks, CNTFREQ);
+    return Common::Divide128On32(temporal, static_cast<u32>(BASE_CLOCK_RATE)).first;
+}
+
+} // namespace Core::Timing
diff --git a/src/core/core_timing_util.h b/src/core/core_timing_util.h
index 5c3718782..679aa3123 100644
--- a/src/core/core_timing_util.h
+++ b/src/core/core_timing_util.h
@@ -6,11 +6,12 @@
 
 #include "common/common_types.h"
 
-namespace CoreTiming {
+namespace Core::Timing {
 
 // The below clock rate is based on Switch's clockspeed being widely known as 1.020GHz
 // The exact value used is of course unverified.
 constexpr u64 BASE_CLOCK_RATE = 1019215872; // Switch clock speed is 1020MHz un/docked
+constexpr u64 CNTFREQ = 19200000;           // Value from fusee.
 
 inline s64 msToCycles(int ms) {
     // since ms is int there is no way to overflow
@@ -61,4 +62,6 @@ inline u64 cyclesToMs(s64 cycles) {
     return cycles * 1000 / BASE_CLOCK_RATE;
 }
 
-} // namespace CoreTiming
+u64 CpuCyclesToClockCycles(u64 ticks);
+
+} // namespace Core::Timing
diff --git a/src/core/cpu_core_manager.cpp b/src/core/cpu_core_manager.cpp
index 769a6fefa..93bc5619c 100644
--- a/src/core/cpu_core_manager.cpp
+++ b/src/core/cpu_core_manager.cpp
@@ -27,7 +27,7 @@ void CpuCoreManager::Initialize(System& system) {
     exclusive_monitor = Cpu::MakeExclusiveMonitor(cores.size());
 
     for (std::size_t index = 0; index < cores.size(); ++index) {
-        cores[index] = std::make_unique<Cpu>(*exclusive_monitor, *barrier, index);
+        cores[index] = std::make_unique<Cpu>(system, *exclusive_monitor, *barrier, index);
     }
 
     // Create threads for CPU cores 1-3, and build thread_to_cpu map
diff --git a/src/core/crypto/key_manager.cpp b/src/core/crypto/key_manager.cpp
index ca12fb4ab..dfac9a4b3 100644
--- a/src/core/crypto/key_manager.cpp
+++ b/src/core/crypto/key_manager.cpp
@@ -398,7 +398,8 @@ static bool ValidCryptoRevisionString(std::string_view base, size_t begin, size_
 }
 
 void KeyManager::LoadFromFile(const std::string& filename, bool is_title_keys) {
-    std::ifstream file(filename);
+    std::ifstream file;
+    OpenFStream(file, filename, std::ios_base::in);
     if (!file.is_open())
         return;
 
diff --git a/src/core/file_sys/content_archive.h b/src/core/file_sys/content_archive.h
index 5d4d05c82..15b9e6624 100644
--- a/src/core/file_sys/content_archive.h
+++ b/src/core/file_sys/content_archive.h
@@ -24,13 +24,26 @@ namespace FileSys {
 
 union NCASectionHeader;
 
+/// Describes the type of content within an NCA archive.
 enum class NCAContentType : u8 {
+    /// Executable-related data
     Program = 0,
+
+    /// Metadata.
     Meta = 1,
+
+    /// Access control data.
     Control = 2,
+
+    /// Information related to the game manual
+    /// e.g. Legal information, etc.
     Manual = 3,
+
+    /// System data.
     Data = 4,
-    Data_Unknown5 = 5, ///< Seems to be used on some system archives
+
+    /// Data that can be accessed by applications.
+    PublicData = 5,
 };
 
 enum class NCASectionCryptoType : u8 {
diff --git a/src/core/file_sys/registered_cache.cpp b/src/core/file_sys/registered_cache.cpp
index 128199063..1c6bacace 100644
--- a/src/core/file_sys/registered_cache.cpp
+++ b/src/core/file_sys/registered_cache.cpp
@@ -94,7 +94,7 @@ static ContentRecordType GetCRTypeFromNCAType(NCAContentType type) {
     case NCAContentType::Control:
         return ContentRecordType::Control;
     case NCAContentType::Data:
-    case NCAContentType::Data_Unknown5:
+    case NCAContentType::PublicData:
         return ContentRecordType::Data;
     case NCAContentType::Manual:
         // TODO(DarkLordZach): Peek at NCA contents to differentiate Manual and Legal.
diff --git a/src/core/file_sys/vfs_vector.cpp b/src/core/file_sys/vfs_vector.cpp
index 515626658..75fc04302 100644
--- a/src/core/file_sys/vfs_vector.cpp
+++ b/src/core/file_sys/vfs_vector.cpp
@@ -47,7 +47,7 @@ std::size_t VectorVfsFile::Write(const u8* data_, std::size_t length, std::size_
     if (offset + length > data.size())
         data.resize(offset + length);
     const auto write = std::min(length, data.size() - offset);
-    std::memcpy(data.data(), data_, write);
+    std::memcpy(data.data() + offset, data_, write);
     return write;
 }
 
diff --git a/src/core/frontend/emu_window.cpp b/src/core/frontend/emu_window.cpp
index 9dd493efb..e29afd630 100644
--- a/src/core/frontend/emu_window.cpp
+++ b/src/core/frontend/emu_window.cpp
@@ -67,7 +67,7 @@ static bool IsWithinTouchscreen(const Layout::FramebufferLayout& layout, unsigne
             framebuffer_x >= layout.screen.left && framebuffer_x < layout.screen.right);
 }
 
-std::tuple<unsigned, unsigned> EmuWindow::ClipToTouchScreen(unsigned new_x, unsigned new_y) {
+std::tuple<unsigned, unsigned> EmuWindow::ClipToTouchScreen(unsigned new_x, unsigned new_y) const {
     new_x = std::max(new_x, framebuffer_layout.screen.left);
     new_x = std::min(new_x, framebuffer_layout.screen.right - 1);
 
diff --git a/src/core/frontend/emu_window.h b/src/core/frontend/emu_window.h
index 7006a37b3..d0bcb4660 100644
--- a/src/core/frontend/emu_window.h
+++ b/src/core/frontend/emu_window.h
@@ -166,7 +166,7 @@ private:
     /**
      * Clip the provided coordinates to be inside the touchscreen area.
      */
-    std::tuple<unsigned, unsigned> ClipToTouchScreen(unsigned new_x, unsigned new_y);
+    std::tuple<unsigned, unsigned> ClipToTouchScreen(unsigned new_x, unsigned new_y) const;
 };
 
 } // namespace Core::Frontend
diff --git a/src/core/frontend/framebuffer_layout.cpp b/src/core/frontend/framebuffer_layout.cpp
index f8662d193..a1357179f 100644
--- a/src/core/frontend/framebuffer_layout.cpp
+++ b/src/core/frontend/framebuffer_layout.cpp
@@ -12,12 +12,12 @@ namespace Layout {
 
 // Finds the largest size subrectangle contained in window area that is confined to the aspect ratio
 template <class T>
-static MathUtil::Rectangle<T> maxRectangle(MathUtil::Rectangle<T> window_area,
-                                           float screen_aspect_ratio) {
+static Common::Rectangle<T> MaxRectangle(Common::Rectangle<T> window_area,
+                                         float screen_aspect_ratio) {
     float scale = std::min(static_cast<float>(window_area.GetWidth()),
                            window_area.GetHeight() / screen_aspect_ratio);
-    return MathUtil::Rectangle<T>{0, 0, static_cast<T>(std::round(scale)),
-                                  static_cast<T>(std::round(scale * screen_aspect_ratio))};
+    return Common::Rectangle<T>{0, 0, static_cast<T>(std::round(scale)),
+                                static_cast<T>(std::round(scale * screen_aspect_ratio))};
 }
 
 FramebufferLayout DefaultFrameLayout(unsigned width, unsigned height) {
@@ -29,8 +29,8 @@ FramebufferLayout DefaultFrameLayout(unsigned width, unsigned height) {
 
     const float emulation_aspect_ratio{static_cast<float>(ScreenUndocked::Height) /
                                        ScreenUndocked::Width};
-    MathUtil::Rectangle<unsigned> screen_window_area{0, 0, width, height};
-    MathUtil::Rectangle<unsigned> screen = maxRectangle(screen_window_area, emulation_aspect_ratio);
+    Common::Rectangle<unsigned> screen_window_area{0, 0, width, height};
+    Common::Rectangle<unsigned> screen = MaxRectangle(screen_window_area, emulation_aspect_ratio);
 
     float window_aspect_ratio = static_cast<float>(height) / width;
 
diff --git a/src/core/frontend/framebuffer_layout.h b/src/core/frontend/framebuffer_layout.h
index e06647794..c2c63d08c 100644
--- a/src/core/frontend/framebuffer_layout.h
+++ b/src/core/frontend/framebuffer_layout.h
@@ -16,7 +16,7 @@ struct FramebufferLayout {
     unsigned width{ScreenUndocked::Width};
     unsigned height{ScreenUndocked::Height};
 
-    MathUtil::Rectangle<unsigned> screen;
+    Common::Rectangle<unsigned> screen;
 
     /**
      * Returns the ration of pixel size of the screen, compared to the native size of the undocked
diff --git a/src/core/frontend/input.h b/src/core/frontend/input.h
index 16fdcd376..7c11d7546 100644
--- a/src/core/frontend/input.h
+++ b/src/core/frontend/input.h
@@ -124,7 +124,7 @@ using AnalogDevice = InputDevice<std::tuple<float, float>>;
  *   Orientation is determined by right-hand rule.
  *   Units: deg/sec
  */
-using MotionDevice = InputDevice<std::tuple<Math::Vec3<float>, Math::Vec3<float>>>;
+using MotionDevice = InputDevice<std::tuple<Common::Vec3<float>, Common::Vec3<float>>>;
 
 /**
  * A touch device is an input device that returns a tuple of two floats and a bool. The floats are
diff --git a/src/core/gdbstub/gdbstub.cpp b/src/core/gdbstub/gdbstub.cpp
index a1cad4fcb..dafb32aae 100644
--- a/src/core/gdbstub/gdbstub.cpp
+++ b/src/core/gdbstub/gdbstub.cpp
@@ -507,8 +507,11 @@ static void RemoveBreakpoint(BreakpointType type, VAddr addr) {
 
     LOG_DEBUG(Debug_GDBStub, "gdb: removed a breakpoint: {:016X} bytes at {:016X} of type {}",
               bp->second.len, bp->second.addr, static_cast<int>(type));
-    Memory::WriteBlock(bp->second.addr, bp->second.inst.data(), bp->second.inst.size());
-    Core::System::GetInstance().InvalidateCpuInstructionCaches();
+
+    if (type == BreakpointType::Execute) {
+        Memory::WriteBlock(bp->second.addr, bp->second.inst.data(), bp->second.inst.size());
+        Core::System::GetInstance().InvalidateCpuInstructionCaches();
+    }
     p.erase(addr);
 }
 
@@ -1057,9 +1060,12 @@ static bool CommitBreakpoint(BreakpointType type, VAddr addr, u64 len) {
     breakpoint.addr = addr;
     breakpoint.len = len;
     Memory::ReadBlock(addr, breakpoint.inst.data(), breakpoint.inst.size());
+
     static constexpr std::array<u8, 4> btrap{0x00, 0x7d, 0x20, 0xd4};
-    Memory::WriteBlock(addr, btrap.data(), btrap.size());
-    Core::System::GetInstance().InvalidateCpuInstructionCaches();
+    if (type == BreakpointType::Execute) {
+        Memory::WriteBlock(addr, btrap.data(), btrap.size());
+        Core::System::GetInstance().InvalidateCpuInstructionCaches();
+    }
     p.insert({addr, breakpoint});
 
     LOG_DEBUG(Debug_GDBStub, "gdb: added {} breakpoint: {:016X} bytes at {:016X}",
diff --git a/src/core/hle/ipc.h b/src/core/hle/ipc.h
index 96c8677d2..fae54bcc7 100644
--- a/src/core/hle/ipc.h
+++ b/src/core/hle/ipc.h
@@ -4,10 +4,10 @@
 
 #pragma once
 
+#include "common/bit_field.h"
+#include "common/common_funcs.h"
 #include "common/common_types.h"
 #include "common/swap.h"
-#include "core/hle/kernel/errors.h"
-#include "core/memory.h"
 
 namespace IPC {
 
diff --git a/src/core/hle/ipc_helpers.h b/src/core/hle/ipc_helpers.h
index 90f276ee8..68406eb63 100644
--- a/src/core/hle/ipc_helpers.h
+++ b/src/core/hle/ipc_helpers.h
@@ -19,9 +19,12 @@
 #include "core/hle/kernel/hle_ipc.h"
 #include "core/hle/kernel/object.h"
 #include "core/hle/kernel/server_session.h"
+#include "core/hle/result.h"
 
 namespace IPC {
 
+constexpr ResultCode ERR_REMOTE_PROCESS_DEAD{ErrorModule::HIPC, 301};
+
 class RequestHelperBase {
 protected:
     Kernel::HLERequestContext* context = nullptr;
@@ -272,6 +275,20 @@ inline void ResponseBuilder::Push(u64 value) {
 }
 
 template <>
+inline void ResponseBuilder::Push(float value) {
+    u32 integral;
+    std::memcpy(&integral, &value, sizeof(u32));
+    Push(integral);
+}
+
+template <>
+inline void ResponseBuilder::Push(double value) {
+    u64 integral;
+    std::memcpy(&integral, &value, sizeof(u64));
+    Push(integral);
+}
+
+template <>
 inline void ResponseBuilder::Push(bool value) {
     Push(static_cast<u8>(value));
 }
@@ -350,7 +367,7 @@ public:
     template <class T>
     std::shared_ptr<T> PopIpcInterface() {
         ASSERT(context->Session()->IsDomain());
-        ASSERT(context->GetDomainMessageHeader()->input_object_count > 0);
+        ASSERT(context->GetDomainMessageHeader().input_object_count > 0);
         return context->GetDomainRequestHandler<T>(Pop<u32>() - 1);
     }
 };
@@ -362,6 +379,11 @@ inline u32 RequestParser::Pop() {
     return cmdbuf[index++];
 }
 
+template <>
+inline s32 RequestParser::Pop() {
+    return static_cast<s32>(Pop<u32>());
+}
+
 template <typename T>
 void RequestParser::PopRaw(T& value) {
     std::memcpy(&value, cmdbuf + index, sizeof(T));
@@ -393,11 +415,37 @@ inline u64 RequestParser::Pop() {
 }
 
 template <>
+inline s8 RequestParser::Pop() {
+    return static_cast<s8>(Pop<u8>());
+}
+
+template <>
+inline s16 RequestParser::Pop() {
+    return static_cast<s16>(Pop<u16>());
+}
+
+template <>
 inline s64 RequestParser::Pop() {
     return static_cast<s64>(Pop<u64>());
 }
 
 template <>
+inline float RequestParser::Pop() {
+    const u32 value = Pop<u32>();
+    float real;
+    std::memcpy(&real, &value, sizeof(real));
+    return real;
+}
+
+template <>
+inline double RequestParser::Pop() {
+    const u64 value = Pop<u64>();
+    float real;
+    std::memcpy(&real, &value, sizeof(real));
+    return real;
+}
+
+template <>
 inline bool RequestParser::Pop() {
     return Pop<u8>() != 0;
 }
diff --git a/src/core/hle/kernel/address_arbiter.cpp b/src/core/hle/kernel/address_arbiter.cpp
index 57157beb4..352190da8 100644
--- a/src/core/hle/kernel/address_arbiter.cpp
+++ b/src/core/hle/kernel/address_arbiter.cpp
@@ -9,6 +9,7 @@
 #include "common/common_types.h"
 #include "core/core.h"
 #include "core/core_cpu.h"
+#include "core/hle/kernel/address_arbiter.h"
 #include "core/hle/kernel/errors.h"
 #include "core/hle/kernel/object.h"
 #include "core/hle/kernel/process.h"
@@ -18,58 +19,15 @@
 #include "core/memory.h"
 
 namespace Kernel {
-namespace AddressArbiter {
-
-// Performs actual address waiting logic.
-static ResultCode WaitForAddress(VAddr address, s64 timeout) {
-    SharedPtr<Thread> current_thread = GetCurrentThread();
-    current_thread->SetArbiterWaitAddress(address);
-    current_thread->SetStatus(ThreadStatus::WaitArb);
-    current_thread->InvalidateWakeupCallback();
-
-    current_thread->WakeAfterDelay(timeout);
-
-    Core::System::GetInstance().CpuCore(current_thread->GetProcessorID()).PrepareReschedule();
-    return RESULT_TIMEOUT;
-}
-
-// Gets the threads waiting on an address.
-static std::vector<SharedPtr<Thread>> GetThreadsWaitingOnAddress(VAddr address) {
-    const auto RetrieveWaitingThreads = [](std::size_t core_index,
-                                           std::vector<SharedPtr<Thread>>& waiting_threads,
-                                           VAddr arb_addr) {
-        const auto& scheduler = Core::System::GetInstance().Scheduler(core_index);
-        const auto& thread_list = scheduler.GetThreadList();
-
-        for (const auto& thread : thread_list) {
-            if (thread->GetArbiterWaitAddress() == arb_addr)
-                waiting_threads.push_back(thread);
-        }
-    };
-
-    // Retrieve all threads that are waiting for this address.
-    std::vector<SharedPtr<Thread>> threads;
-    RetrieveWaitingThreads(0, threads, address);
-    RetrieveWaitingThreads(1, threads, address);
-    RetrieveWaitingThreads(2, threads, address);
-    RetrieveWaitingThreads(3, threads, address);
-
-    // Sort them by priority, such that the highest priority ones come first.
-    std::sort(threads.begin(), threads.end(),
-              [](const SharedPtr<Thread>& lhs, const SharedPtr<Thread>& rhs) {
-                  return lhs->GetPriority() < rhs->GetPriority();
-              });
-
-    return threads;
-}
-
+namespace {
 // Wake up num_to_wake (or all) threads in a vector.
-static void WakeThreads(std::vector<SharedPtr<Thread>>& waiting_threads, s32 num_to_wake) {
+void WakeThreads(const std::vector<SharedPtr<Thread>>& waiting_threads, s32 num_to_wake) {
     // Only process up to 'target' threads, unless 'target' is <= 0, in which case process
     // them all.
     std::size_t last = waiting_threads.size();
-    if (num_to_wake > 0)
+    if (num_to_wake > 0) {
         last = num_to_wake;
+    }
 
     // Signal the waiting threads.
     for (std::size_t i = 0; i < last; i++) {
@@ -79,42 +37,55 @@ static void WakeThreads(std::vector<SharedPtr<Thread>>& waiting_threads, s32 num
         waiting_threads[i]->ResumeFromWait();
     }
 }
+} // Anonymous namespace
+
+AddressArbiter::AddressArbiter(Core::System& system) : system{system} {}
+AddressArbiter::~AddressArbiter() = default;
+
+ResultCode AddressArbiter::SignalToAddress(VAddr address, SignalType type, s32 value,
+                                           s32 num_to_wake) {
+    switch (type) {
+    case SignalType::Signal:
+        return SignalToAddressOnly(address, num_to_wake);
+    case SignalType::IncrementAndSignalIfEqual:
+        return IncrementAndSignalToAddressIfEqual(address, value, num_to_wake);
+    case SignalType::ModifyByWaitingCountAndSignalIfEqual:
+        return ModifyByWaitingCountAndSignalToAddressIfEqual(address, value, num_to_wake);
+    default:
+        return ERR_INVALID_ENUM_VALUE;
+    }
+}
 
-// Signals an address being waited on.
-ResultCode SignalToAddress(VAddr address, s32 num_to_wake) {
-    std::vector<SharedPtr<Thread>> waiting_threads = GetThreadsWaitingOnAddress(address);
-
+ResultCode AddressArbiter::SignalToAddressOnly(VAddr address, s32 num_to_wake) {
+    const std::vector<SharedPtr<Thread>> waiting_threads = GetThreadsWaitingOnAddress(address);
     WakeThreads(waiting_threads, num_to_wake);
     return RESULT_SUCCESS;
 }
 
-// Signals an address being waited on and increments its value if equal to the value argument.
-ResultCode IncrementAndSignalToAddressIfEqual(VAddr address, s32 value, s32 num_to_wake) {
+ResultCode AddressArbiter::IncrementAndSignalToAddressIfEqual(VAddr address, s32 value,
+                                                              s32 num_to_wake) {
     // Ensure that we can write to the address.
     if (!Memory::IsValidVirtualAddress(address)) {
         return ERR_INVALID_ADDRESS_STATE;
     }
 
-    if (static_cast<s32>(Memory::Read32(address)) == value) {
-        Memory::Write32(address, static_cast<u32>(value + 1));
-    } else {
+    if (static_cast<s32>(Memory::Read32(address)) != value) {
         return ERR_INVALID_STATE;
     }
 
-    return SignalToAddress(address, num_to_wake);
+    Memory::Write32(address, static_cast<u32>(value + 1));
+    return SignalToAddressOnly(address, num_to_wake);
 }
 
-// Signals an address being waited on and modifies its value based on waiting thread count if equal
-// to the value argument.
-ResultCode ModifyByWaitingCountAndSignalToAddressIfEqual(VAddr address, s32 value,
-                                                         s32 num_to_wake) {
+ResultCode AddressArbiter::ModifyByWaitingCountAndSignalToAddressIfEqual(VAddr address, s32 value,
+                                                                         s32 num_to_wake) {
     // Ensure that we can write to the address.
     if (!Memory::IsValidVirtualAddress(address)) {
         return ERR_INVALID_ADDRESS_STATE;
     }
 
     // Get threads waiting on the address.
-    std::vector<SharedPtr<Thread>> waiting_threads = GetThreadsWaitingOnAddress(address);
+    const std::vector<SharedPtr<Thread>> waiting_threads = GetThreadsWaitingOnAddress(address);
 
     // Determine the modified value depending on the waiting count.
     s32 updated_value;
@@ -126,41 +97,54 @@ ResultCode ModifyByWaitingCountAndSignalToAddressIfEqual(VAddr address, s32 valu
         updated_value = value;
     }
 
-    if (static_cast<s32>(Memory::Read32(address)) == value) {
-        Memory::Write32(address, static_cast<u32>(updated_value));
-    } else {
+    if (static_cast<s32>(Memory::Read32(address)) != value) {
         return ERR_INVALID_STATE;
     }
 
+    Memory::Write32(address, static_cast<u32>(updated_value));
     WakeThreads(waiting_threads, num_to_wake);
     return RESULT_SUCCESS;
 }
 
-// Waits on an address if the value passed is less than the argument value, optionally decrementing.
-ResultCode WaitForAddressIfLessThan(VAddr address, s32 value, s64 timeout, bool should_decrement) {
+ResultCode AddressArbiter::WaitForAddress(VAddr address, ArbitrationType type, s32 value,
+                                          s64 timeout_ns) {
+    switch (type) {
+    case ArbitrationType::WaitIfLessThan:
+        return WaitForAddressIfLessThan(address, value, timeout_ns, false);
+    case ArbitrationType::DecrementAndWaitIfLessThan:
+        return WaitForAddressIfLessThan(address, value, timeout_ns, true);
+    case ArbitrationType::WaitIfEqual:
+        return WaitForAddressIfEqual(address, value, timeout_ns);
+    default:
+        return ERR_INVALID_ENUM_VALUE;
+    }
+}
+
+ResultCode AddressArbiter::WaitForAddressIfLessThan(VAddr address, s32 value, s64 timeout,
+                                                    bool should_decrement) {
     // Ensure that we can read the address.
     if (!Memory::IsValidVirtualAddress(address)) {
         return ERR_INVALID_ADDRESS_STATE;
     }
 
-    s32 cur_value = static_cast<s32>(Memory::Read32(address));
-    if (cur_value < value) {
-        if (should_decrement) {
-            Memory::Write32(address, static_cast<u32>(cur_value - 1));
-        }
-    } else {
+    const s32 cur_value = static_cast<s32>(Memory::Read32(address));
+    if (cur_value >= value) {
         return ERR_INVALID_STATE;
     }
+
+    if (should_decrement) {
+        Memory::Write32(address, static_cast<u32>(cur_value - 1));
+    }
+
     // Short-circuit without rescheduling, if timeout is zero.
     if (timeout == 0) {
         return RESULT_TIMEOUT;
     }
 
-    return WaitForAddress(address, timeout);
+    return WaitForAddressImpl(address, timeout);
 }
 
-// Waits on an address if the value passed is equal to the argument value.
-ResultCode WaitForAddressIfEqual(VAddr address, s32 value, s64 timeout) {
+ResultCode AddressArbiter::WaitForAddressIfEqual(VAddr address, s32 value, s64 timeout) {
     // Ensure that we can read the address.
     if (!Memory::IsValidVirtualAddress(address)) {
         return ERR_INVALID_ADDRESS_STATE;
@@ -174,7 +158,48 @@ ResultCode WaitForAddressIfEqual(VAddr address, s32 value, s64 timeout) {
         return RESULT_TIMEOUT;
     }
 
-    return WaitForAddress(address, timeout);
+    return WaitForAddressImpl(address, timeout);
+}
+
+ResultCode AddressArbiter::WaitForAddressImpl(VAddr address, s64 timeout) {
+    SharedPtr<Thread> current_thread = system.CurrentScheduler().GetCurrentThread();
+    current_thread->SetArbiterWaitAddress(address);
+    current_thread->SetStatus(ThreadStatus::WaitArb);
+    current_thread->InvalidateWakeupCallback();
+
+    current_thread->WakeAfterDelay(timeout);
+
+    system.CpuCore(current_thread->GetProcessorID()).PrepareReschedule();
+    return RESULT_TIMEOUT;
+}
+
+std::vector<SharedPtr<Thread>> AddressArbiter::GetThreadsWaitingOnAddress(VAddr address) const {
+    const auto RetrieveWaitingThreads = [this](std::size_t core_index,
+                                               std::vector<SharedPtr<Thread>>& waiting_threads,
+                                               VAddr arb_addr) {
+        const auto& scheduler = system.Scheduler(core_index);
+        const auto& thread_list = scheduler.GetThreadList();
+
+        for (const auto& thread : thread_list) {
+            if (thread->GetArbiterWaitAddress() == arb_addr) {
+                waiting_threads.push_back(thread);
+            }
+        }
+    };
+
+    // Retrieve all threads that are waiting for this address.
+    std::vector<SharedPtr<Thread>> threads;
+    RetrieveWaitingThreads(0, threads, address);
+    RetrieveWaitingThreads(1, threads, address);
+    RetrieveWaitingThreads(2, threads, address);
+    RetrieveWaitingThreads(3, threads, address);
+
+    // Sort them by priority, such that the highest priority ones come first.
+    std::sort(threads.begin(), threads.end(),
+              [](const SharedPtr<Thread>& lhs, const SharedPtr<Thread>& rhs) {
+                  return lhs->GetPriority() < rhs->GetPriority();
+              });
+
+    return threads;
 }
-} // namespace AddressArbiter
 } // namespace Kernel
diff --git a/src/core/hle/kernel/address_arbiter.h b/src/core/hle/kernel/address_arbiter.h
index e3657b8e9..ed0d0e69f 100644
--- a/src/core/hle/kernel/address_arbiter.h
+++ b/src/core/hle/kernel/address_arbiter.h
@@ -4,31 +4,77 @@
 
 #pragma once
 
+#include <vector>
+
 #include "common/common_types.h"
+#include "core/hle/kernel/object.h"
 
 union ResultCode;
 
+namespace Core {
+class System;
+}
+
 namespace Kernel {
 
-namespace AddressArbiter {
-enum class ArbitrationType {
-    WaitIfLessThan = 0,
-    DecrementAndWaitIfLessThan = 1,
-    WaitIfEqual = 2,
-};
+class Thread;
 
-enum class SignalType {
-    Signal = 0,
-    IncrementAndSignalIfEqual = 1,
-    ModifyByWaitingCountAndSignalIfEqual = 2,
-};
+class AddressArbiter {
+public:
+    enum class ArbitrationType {
+        WaitIfLessThan = 0,
+        DecrementAndWaitIfLessThan = 1,
+        WaitIfEqual = 2,
+    };
+
+    enum class SignalType {
+        Signal = 0,
+        IncrementAndSignalIfEqual = 1,
+        ModifyByWaitingCountAndSignalIfEqual = 2,
+    };
+
+    explicit AddressArbiter(Core::System& system);
+    ~AddressArbiter();
+
+    AddressArbiter(const AddressArbiter&) = delete;
+    AddressArbiter& operator=(const AddressArbiter&) = delete;
+
+    AddressArbiter(AddressArbiter&&) = default;
+    AddressArbiter& operator=(AddressArbiter&&) = delete;
+
+    /// Signals an address being waited on with a particular signaling type.
+    ResultCode SignalToAddress(VAddr address, SignalType type, s32 value, s32 num_to_wake);
 
-ResultCode SignalToAddress(VAddr address, s32 num_to_wake);
-ResultCode IncrementAndSignalToAddressIfEqual(VAddr address, s32 value, s32 num_to_wake);
-ResultCode ModifyByWaitingCountAndSignalToAddressIfEqual(VAddr address, s32 value, s32 num_to_wake);
+    /// Waits on an address with a particular arbitration type.
+    ResultCode WaitForAddress(VAddr address, ArbitrationType type, s32 value, s64 timeout_ns);
 
-ResultCode WaitForAddressIfLessThan(VAddr address, s32 value, s64 timeout, bool should_decrement);
-ResultCode WaitForAddressIfEqual(VAddr address, s32 value, s64 timeout);
-} // namespace AddressArbiter
+private:
+    /// Signals an address being waited on.
+    ResultCode SignalToAddressOnly(VAddr address, s32 num_to_wake);
+
+    /// Signals an address being waited on and increments its value if equal to the value argument.
+    ResultCode IncrementAndSignalToAddressIfEqual(VAddr address, s32 value, s32 num_to_wake);
+
+    /// Signals an address being waited on and modifies its value based on waiting thread count if
+    /// equal to the value argument.
+    ResultCode ModifyByWaitingCountAndSignalToAddressIfEqual(VAddr address, s32 value,
+                                                             s32 num_to_wake);
+
+    /// Waits on an address if the value passed is less than the argument value,
+    /// optionally decrementing.
+    ResultCode WaitForAddressIfLessThan(VAddr address, s32 value, s64 timeout,
+                                        bool should_decrement);
+
+    /// Waits on an address if the value passed is equal to the argument value.
+    ResultCode WaitForAddressIfEqual(VAddr address, s32 value, s64 timeout);
+
+    // Waits on the given address with a timeout in nanoseconds
+    ResultCode WaitForAddressImpl(VAddr address, s64 timeout);
+
+    // Gets the threads waiting on an address.
+    std::vector<SharedPtr<Thread>> GetThreadsWaitingOnAddress(VAddr address) const;
+
+    Core::System& system;
+};
 
 } // namespace Kernel
diff --git a/src/core/hle/kernel/client_port.cpp b/src/core/hle/kernel/client_port.cpp
index d4c91d529..aa432658e 100644
--- a/src/core/hle/kernel/client_port.cpp
+++ b/src/core/hle/kernel/client_port.cpp
@@ -33,10 +33,11 @@ ResultVal<SharedPtr<ClientSession>> ClientPort::Connect() {
     // Create a new session pair, let the created sessions inherit the parent port's HLE handler.
     auto sessions = ServerSession::CreateSessionPair(kernel, server_port->GetName(), this);
 
-    if (server_port->hle_handler)
-        server_port->hle_handler->ClientConnected(std::get<SharedPtr<ServerSession>>(sessions));
-    else
-        server_port->pending_sessions.push_back(std::get<SharedPtr<ServerSession>>(sessions));
+    if (server_port->HasHLEHandler()) {
+        server_port->GetHLEHandler()->ClientConnected(std::get<SharedPtr<ServerSession>>(sessions));
+    } else {
+        server_port->AppendPendingSession(std::get<SharedPtr<ServerSession>>(sessions));
+    }
 
     // Wake the threads waiting on the ServerPort
     server_port->WakeupAllWaitingThreads();
diff --git a/src/core/hle/kernel/client_session.cpp b/src/core/hle/kernel/client_session.cpp
index 704e82824..c17baa50a 100644
--- a/src/core/hle/kernel/client_session.cpp
+++ b/src/core/hle/kernel/client_session.cpp
@@ -17,21 +17,11 @@ ClientSession::~ClientSession() {
     // This destructor will be called automatically when the last ClientSession handle is closed by
     // the emulated application.
 
-    // Local references to ServerSession and SessionRequestHandler are necessary to guarantee they
+    // A local reference to the ServerSession is necessary to guarantee it
     // will be kept alive until after ClientDisconnected() returns.
     SharedPtr<ServerSession> server = parent->server;
     if (server) {
-        std::shared_ptr<SessionRequestHandler> hle_handler = server->hle_handler;
-        if (hle_handler)
-            hle_handler->ClientDisconnected(server);
-
-        // TODO(Subv): Force a wake up of all the ServerSession's waiting threads and set
-        // their WaitSynchronization result to 0xC920181A.
-
-        // Clean up the list of client threads with pending requests, they are unneeded now that the
-        // client endpoint is closed.
-        server->pending_requesting_threads.clear();
-        server->currently_handling = nullptr;
+        server->ClientDisconnected();
     }
 
     parent->client = nullptr;
diff --git a/src/core/hle/kernel/client_session.h b/src/core/hle/kernel/client_session.h
index 4c18de69c..b1f39aad7 100644
--- a/src/core/hle/kernel/client_session.h
+++ b/src/core/hle/kernel/client_session.h
@@ -36,14 +36,15 @@ public:
 
     ResultCode SendSyncRequest(SharedPtr<Thread> thread);
 
-    std::string name; ///< Name of client port (optional)
+private:
+    explicit ClientSession(KernelCore& kernel);
+    ~ClientSession() override;
 
     /// The parent session, which links to the server endpoint.
     std::shared_ptr<Session> parent;
 
-private:
-    explicit ClientSession(KernelCore& kernel);
-    ~ClientSession() override;
+    /// Name of the client session (optional)
+    std::string name;
 };
 
 } // namespace Kernel
diff --git a/src/core/hle/kernel/code_set.cpp b/src/core/hle/kernel/code_set.cpp
new file mode 100644
index 000000000..1f434e9af
--- /dev/null
+++ b/src/core/hle/kernel/code_set.cpp
@@ -0,0 +1,12 @@
+// Copyright 2019 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "core/hle/kernel/code_set.h"
+
+namespace Kernel {
+
+CodeSet::CodeSet() = default;
+CodeSet::~CodeSet() = default;
+
+} // namespace Kernel
diff --git a/src/core/hle/kernel/code_set.h b/src/core/hle/kernel/code_set.h
new file mode 100644
index 000000000..834fd23d2
--- /dev/null
+++ b/src/core/hle/kernel/code_set.h
@@ -0,0 +1,90 @@
+// Copyright 2019 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <cstddef>
+#include <memory>
+#include <vector>
+
+#include "common/common_types.h"
+
+namespace Kernel {
+
+/**
+ * Represents executable data that may be loaded into a kernel process.
+ *
+ * A code set consists of three basic segments:
+ *   - A code (AKA text) segment,
+ *   - A read-only data segment (rodata)
+ *   - A data segment
+ *
+ * The code segment is the portion of the object file that contains
+ * executable instructions.
+ *
+ * The read-only data segment in the portion of the object file that
+ * contains (as one would expect) read-only data, such as fixed constant
+ * values and data structures.
+ *
+ * The data segment is similar to the read-only data segment -- it contains
+ * variables and data structures that have predefined values, however,
+ * entities within this segment can be modified.
+ */
+struct CodeSet final {
+    /// A single segment within a code set.
+    struct Segment final {
+        /// The byte offset that this segment is located at.
+        std::size_t offset = 0;
+
+        /// The address to map this segment to.
+        VAddr addr = 0;
+
+        /// The size of this segment in bytes.
+        u32 size = 0;
+    };
+
+    explicit CodeSet();
+    ~CodeSet();
+
+    CodeSet(const CodeSet&) = delete;
+    CodeSet& operator=(const CodeSet&) = delete;
+
+    CodeSet(CodeSet&&) = default;
+    CodeSet& operator=(CodeSet&&) = default;
+
+    Segment& CodeSegment() {
+        return segments[0];
+    }
+
+    const Segment& CodeSegment() const {
+        return segments[0];
+    }
+
+    Segment& RODataSegment() {
+        return segments[1];
+    }
+
+    const Segment& RODataSegment() const {
+        return segments[1];
+    }
+
+    Segment& DataSegment() {
+        return segments[2];
+    }
+
+    const Segment& DataSegment() const {
+        return segments[2];
+    }
+
+    /// The overall data that backs this code set.
+    std::shared_ptr<std::vector<u8>> memory;
+
+    /// The segments that comprise this code set.
+    std::array<Segment, 3> segments;
+
+    /// The entry point address for this code set.
+    VAddr entrypoint = 0;
+};
+
+} // namespace Kernel
diff --git a/src/core/hle/kernel/errors.h b/src/core/hle/kernel/errors.h
index d17eb0cb6..8097b3863 100644
--- a/src/core/hle/kernel/errors.h
+++ b/src/core/hle/kernel/errors.h
@@ -14,6 +14,7 @@ constexpr ResultCode ERR_MAX_CONNECTIONS_REACHED{ErrorModule::Kernel, 7};
 constexpr ResultCode ERR_INVALID_CAPABILITY_DESCRIPTOR{ErrorModule::Kernel, 14};
 constexpr ResultCode ERR_INVALID_SIZE{ErrorModule::Kernel, 101};
 constexpr ResultCode ERR_INVALID_ADDRESS{ErrorModule::Kernel, 102};
+constexpr ResultCode ERR_OUT_OF_MEMORY{ErrorModule::Kernel, 104};
 constexpr ResultCode ERR_HANDLE_TABLE_FULL{ErrorModule::Kernel, 105};
 constexpr ResultCode ERR_INVALID_ADDRESS_STATE{ErrorModule::Kernel, 106};
 constexpr ResultCode ERR_INVALID_MEMORY_PERMISSIONS{ErrorModule::Kernel, 108};
diff --git a/src/core/hle/kernel/handle_table.cpp b/src/core/hle/kernel/handle_table.cpp
index c8acde5b1..bdfaa977f 100644
--- a/src/core/hle/kernel/handle_table.cpp
+++ b/src/core/hle/kernel/handle_table.cpp
@@ -14,32 +14,47 @@
 namespace Kernel {
 namespace {
 constexpr u16 GetSlot(Handle handle) {
-    return handle >> 15;
+    return static_cast<u16>(handle >> 15);
 }
 
 constexpr u16 GetGeneration(Handle handle) {
-    return handle & 0x7FFF;
+    return static_cast<u16>(handle & 0x7FFF);
 }
 } // Anonymous namespace
 
 HandleTable::HandleTable() {
-    next_generation = 1;
     Clear();
 }
 
 HandleTable::~HandleTable() = default;
 
+ResultCode HandleTable::SetSize(s32 handle_table_size) {
+    if (static_cast<u32>(handle_table_size) > MAX_COUNT) {
+        return ERR_OUT_OF_MEMORY;
+    }
+
+    // Values less than or equal to zero indicate to use the maximum allowable
+    // size for the handle table in the actual kernel, so we ignore the given
+    // value in that case, since we assume this by default unless this function
+    // is called.
+    if (handle_table_size > 0) {
+        table_size = static_cast<u16>(handle_table_size);
+    }
+
+    return RESULT_SUCCESS;
+}
+
 ResultVal<Handle> HandleTable::Create(SharedPtr<Object> obj) {
     DEBUG_ASSERT(obj != nullptr);
 
-    u16 slot = next_free_slot;
-    if (slot >= generations.size()) {
+    const u16 slot = next_free_slot;
+    if (slot >= table_size) {
         LOG_ERROR(Kernel, "Unable to allocate Handle, too many slots in use.");
         return ERR_HANDLE_TABLE_FULL;
     }
     next_free_slot = generations[slot];
 
-    u16 generation = next_generation++;
+    const u16 generation = next_generation++;
 
     // Overflow count so it fits in the 15 bits dedicated to the generation in the handle.
     // Horizon OS uses zero to represent an invalid handle, so skip to 1.
@@ -64,10 +79,11 @@ ResultVal<Handle> HandleTable::Duplicate(Handle handle) {
 }
 
 ResultCode HandleTable::Close(Handle handle) {
-    if (!IsValid(handle))
+    if (!IsValid(handle)) {
         return ERR_INVALID_HANDLE;
+    }
 
-    u16 slot = GetSlot(handle);
+    const u16 slot = GetSlot(handle);
 
     objects[slot] = nullptr;
 
@@ -77,10 +93,10 @@ ResultCode HandleTable::Close(Handle handle) {
 }
 
 bool HandleTable::IsValid(Handle handle) const {
-    std::size_t slot = GetSlot(handle);
-    u16 generation = GetGeneration(handle);
+    const std::size_t slot = GetSlot(handle);
+    const u16 generation = GetGeneration(handle);
 
-    return slot < MAX_COUNT && objects[slot] != nullptr && generations[slot] == generation;
+    return slot < table_size && objects[slot] != nullptr && generations[slot] == generation;
 }
 
 SharedPtr<Object> HandleTable::GetGeneric(Handle handle) const {
@@ -97,7 +113,7 @@ SharedPtr<Object> HandleTable::GetGeneric(Handle handle) const {
 }
 
 void HandleTable::Clear() {
-    for (u16 i = 0; i < MAX_COUNT; ++i) {
+    for (u16 i = 0; i < table_size; ++i) {
         generations[i] = i + 1;
         objects[i] = nullptr;
     }
diff --git a/src/core/hle/kernel/handle_table.h b/src/core/hle/kernel/handle_table.h
index 89a3bc740..44901391b 100644
--- a/src/core/hle/kernel/handle_table.h
+++ b/src/core/hle/kernel/handle_table.h
@@ -50,6 +50,20 @@ public:
     ~HandleTable();
 
     /**
+     * Sets the number of handles that may be in use at one time
+     * for this handle table.
+     *
+     * @param handle_table_size The desired size to limit the handle table to.
+     *
+     * @returns an error code indicating if initialization was successful.
+     *          If initialization was not successful, then ERR_OUT_OF_MEMORY
+     *          will be returned.
+     *
+     * @pre handle_table_size must be within the range [0, 1024]
+     */
+    ResultCode SetSize(s32 handle_table_size);
+
+    /**
      * Allocates a handle for the given object.
      * @return The created Handle or one of the following errors:
      *           - `ERR_HANDLE_TABLE_FULL`: the maximum number of handles has been exceeded.
@@ -104,13 +118,20 @@ private:
     std::array<u16, MAX_COUNT> generations;
 
     /**
+     * The limited size of the handle table. This can be specified by process
+     * capabilities in order to restrict the overall number of handles that
+     * can be created in a process instance
+     */
+    u16 table_size = static_cast<u16>(MAX_COUNT);
+
+    /**
      * Global counter of the number of created handles. Stored in `generations` when a handle is
      * created, and wraps around to 1 when it hits 0x8000.
      */
-    u16 next_generation;
+    u16 next_generation = 1;
 
     /// Head of the free slots linked list.
-    u16 next_free_slot;
+    u16 next_free_slot = 0;
 };
 
 } // namespace Kernel
diff --git a/src/core/hle/kernel/hle_ipc.cpp b/src/core/hle/kernel/hle_ipc.cpp
index 5dd855db8..fe710eb6e 100644
--- a/src/core/hle/kernel/hle_ipc.cpp
+++ b/src/core/hle/kernel/hle_ipc.cpp
@@ -86,7 +86,7 @@ HLERequestContext::~HLERequestContext() = default;
 void HLERequestContext::ParseCommandBuffer(const HandleTable& handle_table, u32_le* src_cmdbuf,
                                            bool incoming) {
     IPC::RequestParser rp(src_cmdbuf);
-    command_header = std::make_shared<IPC::CommandHeader>(rp.PopRaw<IPC::CommandHeader>());
+    command_header = rp.PopRaw<IPC::CommandHeader>();
 
     if (command_header->type == IPC::CommandType::Close) {
         // Close does not populate the rest of the IPC header
@@ -95,8 +95,7 @@ void HLERequestContext::ParseCommandBuffer(const HandleTable& handle_table, u32_
 
     // If handle descriptor is present, add size of it
     if (command_header->enable_handle_descriptor) {
-        handle_descriptor_header =
-            std::make_shared<IPC::HandleDescriptorHeader>(rp.PopRaw<IPC::HandleDescriptorHeader>());
+        handle_descriptor_header = rp.PopRaw<IPC::HandleDescriptorHeader>();
         if (handle_descriptor_header->send_current_pid) {
             rp.Skip(2, false);
         }
@@ -140,16 +139,15 @@ void HLERequestContext::ParseCommandBuffer(const HandleTable& handle_table, u32_
         // If this is an incoming message, only CommandType "Request" has a domain header
         // All outgoing domain messages have the domain header, if only incoming has it
         if (incoming || domain_message_header) {
-            domain_message_header =
-                std::make_shared<IPC::DomainMessageHeader>(rp.PopRaw<IPC::DomainMessageHeader>());
+            domain_message_header = rp.PopRaw<IPC::DomainMessageHeader>();
         } else {
-            if (Session()->IsDomain())
+            if (Session()->IsDomain()) {
                 LOG_WARNING(IPC, "Domain request has no DomainMessageHeader!");
+            }
         }
     }
 
-    data_payload_header =
-        std::make_shared<IPC::DataPayloadHeader>(rp.PopRaw<IPC::DataPayloadHeader>());
+    data_payload_header = rp.PopRaw<IPC::DataPayloadHeader>();
 
     data_payload_offset = rp.GetCurrentOffset();
 
@@ -264,11 +262,11 @@ ResultCode HLERequestContext::WriteToOutgoingCommandBuffer(Thread& thread) {
         // Write the domain objects to the command buffer, these go after the raw untranslated data.
         // TODO(Subv): This completely ignores C buffers.
         std::size_t domain_offset = size - domain_message_header->num_objects;
-        auto& request_handlers = server_session->domain_request_handlers;
 
-        for (auto& object : domain_objects) {
-            request_handlers.emplace_back(object);
-            dst_cmdbuf[domain_offset++] = static_cast<u32_le>(request_handlers.size());
+        for (const auto& object : domain_objects) {
+            server_session->AppendDomainRequestHandler(object);
+            dst_cmdbuf[domain_offset++] =
+                static_cast<u32_le>(server_session->NumDomainRequestHandlers());
         }
     }
 
diff --git a/src/core/hle/kernel/hle_ipc.h b/src/core/hle/kernel/hle_ipc.h
index cb1c5aff3..2bdd9f02c 100644
--- a/src/core/hle/kernel/hle_ipc.h
+++ b/src/core/hle/kernel/hle_ipc.h
@@ -6,6 +6,7 @@
 
 #include <array>
 #include <memory>
+#include <optional>
 #include <string>
 #include <type_traits>
 #include <vector>
@@ -15,6 +16,8 @@
 #include "core/hle/ipc.h"
 #include "core/hle/kernel/object.h"
 
+union ResultCode;
+
 namespace Service {
 class ServiceFrameworkBase;
 }
@@ -166,12 +169,12 @@ public:
         return buffer_c_desciptors;
     }
 
-    const IPC::DomainMessageHeader* GetDomainMessageHeader() const {
-        return domain_message_header.get();
+    const IPC::DomainMessageHeader& GetDomainMessageHeader() const {
+        return domain_message_header.value();
     }
 
     bool HasDomainMessageHeader() const {
-        return domain_message_header != nullptr;
+        return domain_message_header.has_value();
     }
 
     /// Helper function to read a buffer using the appropriate buffer descriptor
@@ -208,14 +211,12 @@ public:
 
     template <typename T>
     SharedPtr<T> GetCopyObject(std::size_t index) {
-        ASSERT(index < copy_objects.size());
-        return DynamicObjectCast<T>(copy_objects[index]);
+        return DynamicObjectCast<T>(copy_objects.at(index));
     }
 
     template <typename T>
     SharedPtr<T> GetMoveObject(std::size_t index) {
-        ASSERT(index < move_objects.size());
-        return DynamicObjectCast<T>(move_objects[index]);
+        return DynamicObjectCast<T>(move_objects.at(index));
     }
 
     void AddMoveObject(SharedPtr<Object> object) {
@@ -232,7 +233,7 @@ public:
 
     template <typename T>
     std::shared_ptr<T> GetDomainRequestHandler(std::size_t index) const {
-        return std::static_pointer_cast<T>(domain_request_handlers[index]);
+        return std::static_pointer_cast<T>(domain_request_handlers.at(index));
     }
 
     void SetDomainRequestHandlers(
@@ -272,10 +273,10 @@ private:
     boost::container::small_vector<SharedPtr<Object>, 8> copy_objects;
     boost::container::small_vector<std::shared_ptr<SessionRequestHandler>, 8> domain_objects;
 
-    std::shared_ptr<IPC::CommandHeader> command_header;
-    std::shared_ptr<IPC::HandleDescriptorHeader> handle_descriptor_header;
-    std::shared_ptr<IPC::DataPayloadHeader> data_payload_header;
-    std::shared_ptr<IPC::DomainMessageHeader> domain_message_header;
+    std::optional<IPC::CommandHeader> command_header;
+    std::optional<IPC::HandleDescriptorHeader> handle_descriptor_header;
+    std::optional<IPC::DataPayloadHeader> data_payload_header;
+    std::optional<IPC::DomainMessageHeader> domain_message_header;
     std::vector<IPC::BufferDescriptorX> buffer_x_desciptors;
     std::vector<IPC::BufferDescriptorABW> buffer_a_desciptors;
     std::vector<IPC::BufferDescriptorABW> buffer_b_desciptors;
diff --git a/src/core/hle/kernel/kernel.cpp b/src/core/hle/kernel/kernel.cpp
index 7a524ce5a..4d224d01d 100644
--- a/src/core/hle/kernel/kernel.cpp
+++ b/src/core/hle/kernel/kernel.cpp
@@ -12,6 +12,7 @@
 
 #include "core/core.h"
 #include "core/core_timing.h"
+#include "core/hle/kernel/address_arbiter.h"
 #include "core/hle/kernel/client_port.h"
 #include "core/hle/kernel/handle_table.h"
 #include "core/hle/kernel/kernel.h"
@@ -86,6 +87,8 @@ static void ThreadWakeupCallback(u64 thread_handle, [[maybe_unused]] int cycles_
 }
 
 struct KernelCore::Impl {
+    explicit Impl(Core::System& system) : system{system} {}
+
     void Initialize(KernelCore& kernel) {
         Shutdown();
 
@@ -124,7 +127,7 @@ struct KernelCore::Impl {
 
     void InitializeThreads() {
         thread_wakeup_event_type =
-            CoreTiming::RegisterEvent("ThreadWakeupCallback", ThreadWakeupCallback);
+            system.CoreTiming().RegisterEvent("ThreadWakeupCallback", ThreadWakeupCallback);
     }
 
     std::atomic<u32> next_object_id{0};
@@ -137,7 +140,7 @@ struct KernelCore::Impl {
 
     SharedPtr<ResourceLimit> system_resource_limit;
 
-    CoreTiming::EventType* thread_wakeup_event_type = nullptr;
+    Core::Timing::EventType* thread_wakeup_event_type = nullptr;
     // TODO(yuriks): This can be removed if Thread objects are explicitly pooled in the future,
     // allowing us to simply use a pool index or similar.
     Kernel::HandleTable thread_wakeup_callback_handle_table;
@@ -145,9 +148,12 @@ struct KernelCore::Impl {
     /// Map of named ports managed by the kernel, which can be retrieved using
     /// the ConnectToPort SVC.
     NamedPortTable named_ports;
+
+    // System context
+    Core::System& system;
 };
 
-KernelCore::KernelCore() : impl{std::make_unique<Impl>()} {}
+KernelCore::KernelCore(Core::System& system) : impl{std::make_unique<Impl>(system)} {}
 KernelCore::~KernelCore() {
     Shutdown();
 }
@@ -213,7 +219,7 @@ u64 KernelCore::CreateNewProcessID() {
     return impl->next_process_id++;
 }
 
-CoreTiming::EventType* KernelCore::ThreadWakeupCallbackEventType() const {
+Core::Timing::EventType* KernelCore::ThreadWakeupCallbackEventType() const {
     return impl->thread_wakeup_event_type;
 }
 
diff --git a/src/core/hle/kernel/kernel.h b/src/core/hle/kernel/kernel.h
index c643a6401..ff17ff865 100644
--- a/src/core/hle/kernel/kernel.h
+++ b/src/core/hle/kernel/kernel.h
@@ -11,12 +11,18 @@
 template <typename T>
 class ResultVal;
 
-namespace CoreTiming {
-struct EventType;
+namespace Core {
+class System;
 }
 
+namespace Core::Timing {
+class CoreTiming;
+struct EventType;
+} // namespace Core::Timing
+
 namespace Kernel {
 
+class AddressArbiter;
 class ClientPort;
 class HandleTable;
 class Process;
@@ -29,7 +35,14 @@ private:
     using NamedPortTable = std::unordered_map<std::string, SharedPtr<ClientPort>>;
 
 public:
-    KernelCore();
+    /// Constructs an instance of the kernel using the given System
+    /// instance as a context for any necessary system-related state,
+    /// such as threads, CPU core state, etc.
+    ///
+    /// @post After execution of the constructor, the provided System
+    ///       object *must* outlive the kernel instance itself.
+    ///
+    explicit KernelCore(Core::System& system);
     ~KernelCore();
 
     KernelCore(const KernelCore&) = delete;
@@ -89,7 +102,7 @@ private:
     u64 CreateNewThreadID();
 
     /// Retrieves the event type used for thread wakeup callbacks.
-    CoreTiming::EventType* ThreadWakeupCallbackEventType() const;
+    Core::Timing::EventType* ThreadWakeupCallbackEventType() const;
 
     /// Provides a reference to the thread wakeup callback handle table.
     Kernel::HandleTable& ThreadWakeupCallbackHandleTable();
diff --git a/src/core/hle/kernel/process.cpp b/src/core/hle/kernel/process.cpp
index c5aa19afa..15a16ae14 100644
--- a/src/core/hle/kernel/process.cpp
+++ b/src/core/hle/kernel/process.cpp
@@ -9,6 +9,7 @@
 #include "common/logging/log.h"
 #include "core/core.h"
 #include "core/file_sys/program_metadata.h"
+#include "core/hle/kernel/code_set.h"
 #include "core/hle/kernel/errors.h"
 #include "core/hle/kernel/kernel.h"
 #include "core/hle/kernel/process.h"
@@ -31,7 +32,7 @@ namespace {
  */
 void SetupMainThread(Process& owner_process, KernelCore& kernel, VAddr entry_point, u32 priority) {
     // Setup page table so we can write to memory
-    SetCurrentPageTable(&owner_process.VMManager().page_table);
+    Memory::SetCurrentPageTable(&owner_process.VMManager().page_table);
 
     // Initialize new "main" thread
     const VAddr stack_top = owner_process.VMManager().GetTLSIORegionEndAddress();
@@ -50,12 +51,10 @@ void SetupMainThread(Process& owner_process, KernelCore& kernel, VAddr entry_poi
 }
 } // Anonymous namespace
 
-CodeSet::CodeSet() = default;
-CodeSet::~CodeSet() = default;
-
-SharedPtr<Process> Process::Create(KernelCore& kernel, std::string&& name) {
-    SharedPtr<Process> process(new Process(kernel));
+SharedPtr<Process> Process::Create(Core::System& system, std::string&& name) {
+    auto& kernel = system.Kernel();
 
+    SharedPtr<Process> process(new Process(system));
     process->name = std::move(name);
     process->resource_limit = kernel.GetSystemResourceLimit();
     process->status = ProcessStatus::Created;
@@ -99,7 +98,13 @@ ResultCode Process::LoadFromMetadata(const FileSys::ProgramMetadata& metadata) {
     vm_manager.Reset(metadata.GetAddressSpaceType());
 
     const auto& caps = metadata.GetKernelCapabilities();
-    return capabilities.InitializeForUserProcess(caps.data(), caps.size(), vm_manager);
+    const auto capability_init_result =
+        capabilities.InitializeForUserProcess(caps.data(), caps.size(), vm_manager);
+    if (capability_init_result.IsError()) {
+        return capability_init_result;
+    }
+
+    return handle_table.SetSize(capabilities.GetHandleTableSize());
 }
 
 void Process::Run(VAddr entry_point, s32 main_thread_priority, u32 stack_size) {
@@ -126,7 +131,7 @@ void Process::PrepareForTermination() {
             if (thread->GetOwnerProcess() != this)
                 continue;
 
-            if (thread == GetCurrentThread())
+            if (thread == system.CurrentScheduler().GetCurrentThread())
                 continue;
 
             // TODO(Subv): When are the other running/ready threads terminated?
@@ -138,7 +143,6 @@ void Process::PrepareForTermination() {
         }
     };
 
-    const auto& system = Core::System::GetInstance();
     stop_threads(system.Scheduler(0).GetThreadList());
     stop_threads(system.Scheduler(1).GetThreadList());
     stop_threads(system.Scheduler(2).GetThreadList());
@@ -206,7 +210,7 @@ void Process::FreeTLSSlot(VAddr tls_address) {
 }
 
 void Process::LoadModule(CodeSet module_, VAddr base_addr) {
-    const auto MapSegment = [&](CodeSet::Segment& segment, VMAPermission permissions,
+    const auto MapSegment = [&](const CodeSet::Segment& segment, VMAPermission permissions,
                                 MemoryState memory_state) {
         const auto vma = vm_manager
                              .MapMemoryBlock(segment.addr + base_addr, module_.memory,
@@ -221,14 +225,12 @@ void Process::LoadModule(CodeSet module_, VAddr base_addr) {
     MapSegment(module_.DataSegment(), VMAPermission::ReadWrite, MemoryState::CodeMutable);
 
     // Clear instruction cache in CPU JIT
-    Core::System::GetInstance().ArmInterface(0).ClearInstructionCache();
-    Core::System::GetInstance().ArmInterface(1).ClearInstructionCache();
-    Core::System::GetInstance().ArmInterface(2).ClearInstructionCache();
-    Core::System::GetInstance().ArmInterface(3).ClearInstructionCache();
+    system.InvalidateCpuInstructionCaches();
 }
 
-Kernel::Process::Process(KernelCore& kernel) : WaitObject{kernel} {}
-Kernel::Process::~Process() {}
+Process::Process(Core::System& system)
+    : WaitObject{system.Kernel()}, address_arbiter{system}, system{system} {}
+Process::~Process() = default;
 
 void Process::Acquire(Thread* thread) {
     ASSERT_MSG(!ShouldWait(thread), "Object unavailable!");
diff --git a/src/core/hle/kernel/process.h b/src/core/hle/kernel/process.h
index dcc57ae9f..3ae7c922c 100644
--- a/src/core/hle/kernel/process.h
+++ b/src/core/hle/kernel/process.h
@@ -7,17 +7,21 @@
 #include <array>
 #include <bitset>
 #include <cstddef>
-#include <memory>
 #include <string>
 #include <vector>
 #include <boost/container/static_vector.hpp>
 #include "common/common_types.h"
+#include "core/hle/kernel/address_arbiter.h"
 #include "core/hle/kernel/handle_table.h"
 #include "core/hle/kernel/process_capability.h"
 #include "core/hle/kernel/vm_manager.h"
 #include "core/hle/kernel/wait_object.h"
 #include "core/hle/result.h"
 
+namespace Core {
+class System;
+}
+
 namespace FileSys {
 class ProgramMetadata;
 }
@@ -28,6 +32,8 @@ class KernelCore;
 class ResourceLimit;
 class Thread;
 
+struct CodeSet;
+
 struct AddressMapping {
     // Address and size must be page-aligned
     VAddr address;
@@ -60,46 +66,6 @@ enum class ProcessStatus {
     DebugBreak,
 };
 
-struct CodeSet final {
-    struct Segment {
-        std::size_t offset = 0;
-        VAddr addr = 0;
-        u32 size = 0;
-    };
-
-    explicit CodeSet();
-    ~CodeSet();
-
-    Segment& CodeSegment() {
-        return segments[0];
-    }
-
-    const Segment& CodeSegment() const {
-        return segments[0];
-    }
-
-    Segment& RODataSegment() {
-        return segments[1];
-    }
-
-    const Segment& RODataSegment() const {
-        return segments[1];
-    }
-
-    Segment& DataSegment() {
-        return segments[2];
-    }
-
-    const Segment& DataSegment() const {
-        return segments[2];
-    }
-
-    std::shared_ptr<std::vector<u8>> memory;
-
-    std::array<Segment, 3> segments;
-    VAddr entrypoint = 0;
-};
-
 class Process final : public WaitObject {
 public:
     enum : u64 {
@@ -116,7 +82,7 @@ public:
 
     static constexpr std::size_t RANDOM_ENTROPY_SIZE = 4;
 
-    static SharedPtr<Process> Create(KernelCore& kernel, std::string&& name);
+    static SharedPtr<Process> Create(Core::System& system, std::string&& name);
 
     std::string GetTypeName() const override {
         return "Process";
@@ -150,6 +116,16 @@ public:
         return handle_table;
     }
 
+    /// Gets a reference to the process' address arbiter.
+    AddressArbiter& GetAddressArbiter() {
+        return address_arbiter;
+    }
+
+    /// Gets a const reference to the process' address arbiter.
+    const AddressArbiter& GetAddressArbiter() const {
+        return address_arbiter;
+    }
+
     /// Gets the current status of the process
     ProcessStatus GetStatus() const {
         return status;
@@ -251,7 +227,7 @@ public:
     void FreeTLSSlot(VAddr tls_address);
 
 private:
-    explicit Process(KernelCore& kernel);
+    explicit Process(Core::System& system);
     ~Process() override;
 
     /// Checks if the specified thread should wait until this process is available.
@@ -309,9 +285,16 @@ private:
     /// Per-process handle table for storing created object handles in.
     HandleTable handle_table;
 
+    /// Per-process address arbiter.
+    AddressArbiter address_arbiter;
+
     /// Random values for svcGetInfo RandomEntropy
     std::array<u64, RANDOM_ENTROPY_SIZE> random_entropy;
 
+    /// System context
+    Core::System& system;
+
+    /// Name of this process
     std::string name;
 };
 
diff --git a/src/core/hle/kernel/process_capability.cpp b/src/core/hle/kernel/process_capability.cpp
index 3a2164b25..583e35b79 100644
--- a/src/core/hle/kernel/process_capability.cpp
+++ b/src/core/hle/kernel/process_capability.cpp
@@ -96,7 +96,7 @@ void ProcessCapabilities::InitializeForMetadatalessProcess() {
     interrupt_capabilities.set();
 
     // Allow using the maximum possible amount of handles
-    handle_table_size = static_cast<u32>(HandleTable::MAX_COUNT);
+    handle_table_size = static_cast<s32>(HandleTable::MAX_COUNT);
 
     // Allow all debugging capabilities.
     is_debuggable = true;
@@ -337,7 +337,7 @@ ResultCode ProcessCapabilities::HandleHandleTableFlags(u32 flags) {
         return ERR_RESERVED_VALUE;
     }
 
-    handle_table_size = (flags >> 16) & 0x3FF;
+    handle_table_size = static_cast<s32>((flags >> 16) & 0x3FF);
     return RESULT_SUCCESS;
 }
 
diff --git a/src/core/hle/kernel/process_capability.h b/src/core/hle/kernel/process_capability.h
index fbc8812a3..5cdd80747 100644
--- a/src/core/hle/kernel/process_capability.h
+++ b/src/core/hle/kernel/process_capability.h
@@ -156,7 +156,7 @@ public:
     }
 
     /// Gets the number of total allowable handles for the process' handle table.
-    u32 GetHandleTableSize() const {
+    s32 GetHandleTableSize() const {
         return handle_table_size;
     }
 
@@ -252,7 +252,7 @@ private:
     u64 core_mask = 0;
     u64 priority_mask = 0;
 
-    u32 handle_table_size = 0;
+    s32 handle_table_size = 0;
     u32 kernel_version = 0;
 
     ProgramType program_type = ProgramType::SysModule;
diff --git a/src/core/hle/kernel/scheduler.cpp b/src/core/hle/kernel/scheduler.cpp
index df4d6cf0a..cc189cc64 100644
--- a/src/core/hle/kernel/scheduler.cpp
+++ b/src/core/hle/kernel/scheduler.cpp
@@ -19,7 +19,8 @@ namespace Kernel {
 
 std::mutex Scheduler::scheduler_mutex;
 
-Scheduler::Scheduler(Core::ARM_Interface& cpu_core) : cpu_core(cpu_core) {}
+Scheduler::Scheduler(Core::System& system, Core::ARM_Interface& cpu_core)
+    : cpu_core{cpu_core}, system{system} {}
 
 Scheduler::~Scheduler() {
     for (auto& thread : thread_list) {
@@ -61,7 +62,7 @@ Thread* Scheduler::PopNextReadyThread() {
 
 void Scheduler::SwitchContext(Thread* new_thread) {
     Thread* const previous_thread = GetCurrentThread();
-    Process* const previous_process = Core::CurrentProcess();
+    Process* const previous_process = system.Kernel().CurrentProcess();
 
     UpdateLastContextSwitchTime(previous_thread, previous_process);
 
@@ -94,8 +95,8 @@ void Scheduler::SwitchContext(Thread* new_thread) {
 
         auto* const thread_owner_process = current_thread->GetOwnerProcess();
         if (previous_process != thread_owner_process) {
-            Core::System::GetInstance().Kernel().MakeCurrentProcess(thread_owner_process);
-            SetCurrentPageTable(&Core::CurrentProcess()->VMManager().page_table);
+            system.Kernel().MakeCurrentProcess(thread_owner_process);
+            Memory::SetCurrentPageTable(&thread_owner_process->VMManager().page_table);
         }
 
         cpu_core.LoadContext(new_thread->GetContext());
@@ -111,7 +112,7 @@ void Scheduler::SwitchContext(Thread* new_thread) {
 
 void Scheduler::UpdateLastContextSwitchTime(Thread* thread, Process* process) {
     const u64 prev_switch_ticks = last_context_switch_time;
-    const u64 most_recent_switch_ticks = CoreTiming::GetTicks();
+    const u64 most_recent_switch_ticks = system.CoreTiming().GetTicks();
     const u64 update_ticks = most_recent_switch_ticks - prev_switch_ticks;
 
     if (thread != nullptr) {
@@ -198,8 +199,7 @@ void Scheduler::YieldWithoutLoadBalancing(Thread* thread) {
     ASSERT(thread->GetPriority() < THREADPRIO_COUNT);
 
     // Yield this thread -- sleep for zero time and force reschedule to different thread
-    WaitCurrentThread_Sleep();
-    GetCurrentThread()->WakeAfterDelay(0);
+    GetCurrentThread()->Sleep(0);
 }
 
 void Scheduler::YieldWithLoadBalancing(Thread* thread) {
@@ -214,8 +214,7 @@ void Scheduler::YieldWithLoadBalancing(Thread* thread) {
     ASSERT(priority < THREADPRIO_COUNT);
 
     // Sleep for zero time to be able to force reschedule to different thread
-    WaitCurrentThread_Sleep();
-    GetCurrentThread()->WakeAfterDelay(0);
+    GetCurrentThread()->Sleep(0);
 
     Thread* suggested_thread = nullptr;
 
@@ -223,8 +222,7 @@ void Scheduler::YieldWithLoadBalancing(Thread* thread) {
     // Take the first non-nullptr one
     for (unsigned cur_core = 0; cur_core < Core::NUM_CPU_CORES; ++cur_core) {
         const auto res =
-            Core::System::GetInstance().CpuCore(cur_core).Scheduler().GetNextSuggestedThread(
-                core, priority);
+            system.CpuCore(cur_core).Scheduler().GetNextSuggestedThread(core, priority);
 
         // If scheduler provides a suggested thread
         if (res != nullptr) {
diff --git a/src/core/hle/kernel/scheduler.h b/src/core/hle/kernel/scheduler.h
index 97ced4dfc..1c5bf57d9 100644
--- a/src/core/hle/kernel/scheduler.h
+++ b/src/core/hle/kernel/scheduler.h
@@ -13,7 +13,8 @@
 
 namespace Core {
 class ARM_Interface;
-}
+class System;
+} // namespace Core
 
 namespace Kernel {
 
@@ -21,7 +22,7 @@ class Process;
 
 class Scheduler final {
 public:
-    explicit Scheduler(Core::ARM_Interface& cpu_core);
+    explicit Scheduler(Core::System& system, Core::ARM_Interface& cpu_core);
     ~Scheduler();
 
     /// Returns whether there are any threads that are ready to run.
@@ -162,6 +163,7 @@ private:
     Core::ARM_Interface& cpu_core;
     u64 last_context_switch_time = 0;
 
+    Core::System& system;
     static std::mutex scheduler_mutex;
 };
 
diff --git a/src/core/hle/kernel/server_port.cpp b/src/core/hle/kernel/server_port.cpp
index d6ceeb2da..0e1515c89 100644
--- a/src/core/hle/kernel/server_port.cpp
+++ b/src/core/hle/kernel/server_port.cpp
@@ -26,6 +26,10 @@ ResultVal<SharedPtr<ServerSession>> ServerPort::Accept() {
     return MakeResult(std::move(session));
 }
 
+void ServerPort::AppendPendingSession(SharedPtr<ServerSession> pending_session) {
+    pending_sessions.push_back(std::move(pending_session));
+}
+
 bool ServerPort::ShouldWait(Thread* thread) const {
     // If there are no pending sessions, we wait until a new one is added.
     return pending_sessions.empty();
diff --git a/src/core/hle/kernel/server_port.h b/src/core/hle/kernel/server_port.h
index e52f8245f..9bc667cf2 100644
--- a/src/core/hle/kernel/server_port.h
+++ b/src/core/hle/kernel/server_port.h
@@ -22,6 +22,8 @@ class SessionRequestHandler;
 
 class ServerPort final : public WaitObject {
 public:
+    using HLEHandler = std::shared_ptr<SessionRequestHandler>;
+
     /**
      * Creates a pair of ServerPort and an associated ClientPort.
      *
@@ -51,22 +53,27 @@ public:
      */
     ResultVal<SharedPtr<ServerSession>> Accept();
 
+    /// Whether or not this server port has an HLE handler available.
+    bool HasHLEHandler() const {
+        return hle_handler != nullptr;
+    }
+
+    /// Gets the HLE handler for this port.
+    HLEHandler GetHLEHandler() const {
+        return hle_handler;
+    }
+
     /**
      * Sets the HLE handler template for the port. ServerSessions crated by connecting to this port
      * will inherit a reference to this handler.
      */
-    void SetHleHandler(std::shared_ptr<SessionRequestHandler> hle_handler_) {
+    void SetHleHandler(HLEHandler hle_handler_) {
         hle_handler = std::move(hle_handler_);
     }
 
-    std::string name; ///< Name of port (optional)
-
-    /// ServerSessions waiting to be accepted by the port
-    std::vector<SharedPtr<ServerSession>> pending_sessions;
-
-    /// This session's HLE request handler template (optional)
-    /// ServerSessions created from this port inherit a reference to this handler.
-    std::shared_ptr<SessionRequestHandler> hle_handler;
+    /// Appends a ServerSession to the collection of ServerSessions
+    /// waiting to be accepted by this port.
+    void AppendPendingSession(SharedPtr<ServerSession> pending_session);
 
     bool ShouldWait(Thread* thread) const override;
     void Acquire(Thread* thread) override;
@@ -74,6 +81,16 @@ public:
 private:
     explicit ServerPort(KernelCore& kernel);
     ~ServerPort() override;
+
+    /// ServerSessions waiting to be accepted by the port
+    std::vector<SharedPtr<ServerSession>> pending_sessions;
+
+    /// This session's HLE request handler template (optional)
+    /// ServerSessions created from this port inherit a reference to this handler.
+    HLEHandler hle_handler;
+
+    /// Name of the port (optional)
+    std::string name;
 };
 
 } // namespace Kernel
diff --git a/src/core/hle/kernel/server_session.cpp b/src/core/hle/kernel/server_session.cpp
index 027434f92..4d8a337a7 100644
--- a/src/core/hle/kernel/server_session.cpp
+++ b/src/core/hle/kernel/server_session.cpp
@@ -63,42 +63,71 @@ void ServerSession::Acquire(Thread* thread) {
     pending_requesting_threads.pop_back();
 }
 
+void ServerSession::ClientDisconnected() {
+    // We keep a shared pointer to the hle handler to keep it alive throughout
+    // the call to ClientDisconnected, as ClientDisconnected invalidates the
+    // hle_handler member itself during the course of the function executing.
+    std::shared_ptr<SessionRequestHandler> handler = hle_handler;
+    if (handler) {
+        // Note that after this returns, this server session's hle_handler is
+        // invalidated (set to null).
+        handler->ClientDisconnected(this);
+    }
+
+    // TODO(Subv): Force a wake up of all the ServerSession's waiting threads and set
+    // their WaitSynchronization result to 0xC920181A.
+
+    // Clean up the list of client threads with pending requests, they are unneeded now that the
+    // client endpoint is closed.
+    pending_requesting_threads.clear();
+    currently_handling = nullptr;
+}
+
+void ServerSession::AppendDomainRequestHandler(std::shared_ptr<SessionRequestHandler> handler) {
+    domain_request_handlers.push_back(std::move(handler));
+}
+
+std::size_t ServerSession::NumDomainRequestHandlers() const {
+    return domain_request_handlers.size();
+}
+
 ResultCode ServerSession::HandleDomainSyncRequest(Kernel::HLERequestContext& context) {
-    auto* const domain_message_header = context.GetDomainMessageHeader();
-    if (domain_message_header) {
-        // Set domain handlers in HLE context, used for domain objects (IPC interfaces) as inputs
-        context.SetDomainRequestHandlers(domain_request_handlers);
-
-        // If there is a DomainMessageHeader, then this is CommandType "Request"
-        const u32 object_id{context.GetDomainMessageHeader()->object_id};
-        switch (domain_message_header->command) {
-        case IPC::DomainMessageHeader::CommandType::SendMessage:
-            if (object_id > domain_request_handlers.size()) {
-                LOG_CRITICAL(IPC,
-                             "object_id {} is too big! This probably means a recent service call "
-                             "to {} needed to return a new interface!",
-                             object_id, name);
-                UNREACHABLE();
-                return RESULT_SUCCESS; // Ignore error if asserts are off
-            }
-            return domain_request_handlers[object_id - 1]->HandleSyncRequest(context);
-
-        case IPC::DomainMessageHeader::CommandType::CloseVirtualHandle: {
-            LOG_DEBUG(IPC, "CloseVirtualHandle, object_id=0x{:08X}", object_id);
-
-            domain_request_handlers[object_id - 1] = nullptr;
-
-            IPC::ResponseBuilder rb{context, 2};
-            rb.Push(RESULT_SUCCESS);
-            return RESULT_SUCCESS;
-        }
+    if (!context.HasDomainMessageHeader()) {
+        return RESULT_SUCCESS;
+    }
+
+    // Set domain handlers in HLE context, used for domain objects (IPC interfaces) as inputs
+    context.SetDomainRequestHandlers(domain_request_handlers);
+
+    // If there is a DomainMessageHeader, then this is CommandType "Request"
+    const auto& domain_message_header = context.GetDomainMessageHeader();
+    const u32 object_id{domain_message_header.object_id};
+    switch (domain_message_header.command) {
+    case IPC::DomainMessageHeader::CommandType::SendMessage:
+        if (object_id > domain_request_handlers.size()) {
+            LOG_CRITICAL(IPC,
+                         "object_id {} is too big! This probably means a recent service call "
+                         "to {} needed to return a new interface!",
+                         object_id, name);
+            UNREACHABLE();
+            return RESULT_SUCCESS; // Ignore error if asserts are off
         }
+        return domain_request_handlers[object_id - 1]->HandleSyncRequest(context);
 
-        LOG_CRITICAL(IPC, "Unknown domain command={}",
-                     static_cast<int>(domain_message_header->command.Value()));
-        ASSERT(false);
+    case IPC::DomainMessageHeader::CommandType::CloseVirtualHandle: {
+        LOG_DEBUG(IPC, "CloseVirtualHandle, object_id=0x{:08X}", object_id);
+
+        domain_request_handlers[object_id - 1] = nullptr;
+
+        IPC::ResponseBuilder rb{context, 2};
+        rb.Push(RESULT_SUCCESS);
+        return RESULT_SUCCESS;
+    }
     }
 
+    LOG_CRITICAL(IPC, "Unknown domain command={}",
+                 static_cast<int>(domain_message_header.command.Value()));
+    ASSERT(false);
     return RESULT_SUCCESS;
 }
 
diff --git a/src/core/hle/kernel/server_session.h b/src/core/hle/kernel/server_session.h
index e0e9d64c8..aea4ccfeb 100644
--- a/src/core/hle/kernel/server_session.h
+++ b/src/core/hle/kernel/server_session.h
@@ -46,6 +46,14 @@ public:
         return HANDLE_TYPE;
     }
 
+    Session* GetParent() {
+        return parent.get();
+    }
+
+    const Session* GetParent() const {
+        return parent.get();
+    }
+
     using SessionPair = std::tuple<SharedPtr<ServerSession>, SharedPtr<ClientSession>>;
 
     /**
@@ -78,23 +86,16 @@ public:
 
     void Acquire(Thread* thread) override;
 
-    std::string name;                ///< The name of this session (optional)
-    std::shared_ptr<Session> parent; ///< The parent session, which links to the client endpoint.
-    std::shared_ptr<SessionRequestHandler>
-        hle_handler; ///< This session's HLE request handler (applicable when not a domain)
+    /// Called when a client disconnection occurs.
+    void ClientDisconnected();
 
-    /// This is the list of domain request handlers (after conversion to a domain)
-    std::vector<std::shared_ptr<SessionRequestHandler>> domain_request_handlers;
-
-    /// List of threads that are pending a response after a sync request. This list is processed in
-    /// a LIFO manner, thus, the last request will be dispatched first.
-    /// TODO(Subv): Verify if this is indeed processed in LIFO using a hardware test.
-    std::vector<SharedPtr<Thread>> pending_requesting_threads;
+    /// Adds a new domain request handler to the collection of request handlers within
+    /// this ServerSession instance.
+    void AppendDomainRequestHandler(std::shared_ptr<SessionRequestHandler> handler);
 
-    /// Thread whose request is currently being handled. A request is considered "handled" when a
-    /// response is sent via svcReplyAndReceive.
-    /// TODO(Subv): Find a better name for this.
-    SharedPtr<Thread> currently_handling;
+    /// Retrieves the total number of domain request handlers that have been
+    /// appended to this ServerSession instance.
+    std::size_t NumDomainRequestHandlers() const;
 
     /// Returns true if the session has been converted to a domain, otherwise False
     bool IsDomain() const {
@@ -129,8 +130,30 @@ private:
     /// object handle.
     ResultCode HandleDomainSyncRequest(Kernel::HLERequestContext& context);
 
+    /// The parent session, which links to the client endpoint.
+    std::shared_ptr<Session> parent;
+
+    /// This session's HLE request handler (applicable when not a domain)
+    std::shared_ptr<SessionRequestHandler> hle_handler;
+
+    /// This is the list of domain request handlers (after conversion to a domain)
+    std::vector<std::shared_ptr<SessionRequestHandler>> domain_request_handlers;
+
+    /// List of threads that are pending a response after a sync request. This list is processed in
+    /// a LIFO manner, thus, the last request will be dispatched first.
+    /// TODO(Subv): Verify if this is indeed processed in LIFO using a hardware test.
+    std::vector<SharedPtr<Thread>> pending_requesting_threads;
+
+    /// Thread whose request is currently being handled. A request is considered "handled" when a
+    /// response is sent via svcReplyAndReceive.
+    /// TODO(Subv): Find a better name for this.
+    SharedPtr<Thread> currently_handling;
+
     /// When set to True, converts the session to a domain at the end of the command
     bool convert_to_domain{};
+
+    /// The name of this session (optional)
+    std::string name;
 };
 
 } // namespace Kernel
diff --git a/src/core/hle/kernel/shared_memory.cpp b/src/core/hle/kernel/shared_memory.cpp
index 22d0c1dd5..62861da36 100644
--- a/src/core/hle/kernel/shared_memory.cpp
+++ b/src/core/hle/kernel/shared_memory.cpp
@@ -6,7 +6,6 @@
 
 #include "common/assert.h"
 #include "common/logging/log.h"
-#include "core/core.h"
 #include "core/hle/kernel/errors.h"
 #include "core/hle/kernel/kernel.h"
 #include "core/hle/kernel/shared_memory.h"
@@ -34,8 +33,8 @@ SharedPtr<SharedMemory> SharedMemory::Create(KernelCore& kernel, Process* owner_
         shared_memory->backing_block_offset = 0;
 
         // Refresh the address mappings for the current process.
-        if (Core::CurrentProcess() != nullptr) {
-            Core::CurrentProcess()->VMManager().RefreshMemoryBlockMappings(
+        if (kernel.CurrentProcess() != nullptr) {
+            kernel.CurrentProcess()->VMManager().RefreshMemoryBlockMappings(
                 shared_memory->backing_block.get());
         }
     } else {
diff --git a/src/core/hle/kernel/svc.cpp b/src/core/hle/kernel/svc.cpp
index 7cfecb68c..047fa0c19 100644
--- a/src/core/hle/kernel/svc.cpp
+++ b/src/core/hle/kernel/svc.cpp
@@ -20,6 +20,7 @@
 #include "core/hle/kernel/address_arbiter.h"
 #include "core/hle/kernel/client_port.h"
 #include "core/hle/kernel/client_session.h"
+#include "core/hle/kernel/errors.h"
 #include "core/hle/kernel/handle_table.h"
 #include "core/hle/kernel/kernel.h"
 #include "core/hle/kernel/mutex.h"
@@ -47,23 +48,6 @@ constexpr bool IsValidAddressRange(VAddr address, u64 size) {
     return address + size > address;
 }
 
-// Checks if a given address range lies within a larger address range.
-constexpr bool IsInsideAddressRange(VAddr address, u64 size, VAddr address_range_begin,
-                                    VAddr address_range_end) {
-    const VAddr end_address = address + size - 1;
-    return address_range_begin <= address && end_address <= address_range_end - 1;
-}
-
-bool IsInsideAddressSpace(const VMManager& vm, VAddr address, u64 size) {
-    return IsInsideAddressRange(address, size, vm.GetAddressSpaceBaseAddress(),
-                                vm.GetAddressSpaceEndAddress());
-}
-
-bool IsInsideNewMapRegion(const VMManager& vm, VAddr address, u64 size) {
-    return IsInsideAddressRange(address, size, vm.GetNewMapRegionBaseAddress(),
-                                vm.GetNewMapRegionEndAddress());
-}
-
 // 8 GiB
 constexpr u64 MAIN_MEMORY_SIZE = 0x200000000;
 
@@ -105,14 +89,14 @@ ResultCode MapUnmapMemorySanityChecks(const VMManager& vm_manager, VAddr dst_add
         return ERR_INVALID_ADDRESS_STATE;
     }
 
-    if (!IsInsideAddressSpace(vm_manager, src_addr, size)) {
+    if (!vm_manager.IsWithinAddressSpace(src_addr, size)) {
         LOG_ERROR(Kernel_SVC,
                   "Source is not within the address space, addr=0x{:016X}, size=0x{:016X}",
                   src_addr, size);
         return ERR_INVALID_ADDRESS_STATE;
     }
 
-    if (!IsInsideNewMapRegion(vm_manager, dst_addr, size)) {
+    if (!vm_manager.IsWithinNewMapRegion(dst_addr, size)) {
         LOG_ERROR(Kernel_SVC,
                   "Destination is not within the new map region, addr=0x{:016X}, size=0x{:016X}",
                   dst_addr, size);
@@ -238,7 +222,7 @@ static ResultCode SetMemoryPermission(VAddr addr, u64 size, u32 prot) {
     auto* const current_process = Core::CurrentProcess();
     auto& vm_manager = current_process->VMManager();
 
-    if (!IsInsideAddressSpace(vm_manager, addr, size)) {
+    if (!vm_manager.IsWithinAddressSpace(addr, size)) {
         LOG_ERROR(Kernel_SVC,
                   "Source is not within the address space, addr=0x{:016X}, size=0x{:016X}", addr,
                   size);
@@ -299,7 +283,7 @@ static ResultCode SetMemoryAttribute(VAddr address, u64 size, u32 mask, u32 attr
     }
 
     auto& vm_manager = Core::CurrentProcess()->VMManager();
-    if (!IsInsideAddressSpace(vm_manager, address, size)) {
+    if (!vm_manager.IsWithinAddressSpace(address, size)) {
         LOG_ERROR(Kernel_SVC,
                   "Given address (0x{:016X}) is outside the bounds of the address space.", address);
         return ERR_INVALID_ADDRESS_STATE;
@@ -918,6 +902,7 @@ static ResultCode GetInfo(u64* result, u64 info_id, u64 handle, u64 info_sub_id)
         }
 
         const auto& system = Core::System::GetInstance();
+        const auto& core_timing = system.CoreTiming();
         const auto& scheduler = system.CurrentScheduler();
         const auto* const current_thread = scheduler.GetCurrentThread();
         const bool same_thread = current_thread == thread;
@@ -927,9 +912,9 @@ static ResultCode GetInfo(u64* result, u64 info_id, u64 handle, u64 info_sub_id)
         if (same_thread && info_sub_id == 0xFFFFFFFFFFFFFFFF) {
             const u64 thread_ticks = current_thread->GetTotalCPUTimeTicks();
 
-            out_ticks = thread_ticks + (CoreTiming::GetTicks() - prev_ctx_ticks);
+            out_ticks = thread_ticks + (core_timing.GetTicks() - prev_ctx_ticks);
         } else if (same_thread && info_sub_id == system.CurrentCoreIndex()) {
-            out_ticks = CoreTiming::GetTicks() - prev_ctx_ticks;
+            out_ticks = core_timing.GetTicks() - prev_ctx_ticks;
         }
 
         *result = out_ticks;
@@ -1299,10 +1284,14 @@ static ResultCode StartThread(Handle thread_handle) {
 
 /// Called when a thread exits
 static void ExitThread() {
-    LOG_TRACE(Kernel_SVC, "called, pc=0x{:08X}", Core::CurrentArmInterface().GetPC());
+    auto& system = Core::System::GetInstance();
 
-    ExitCurrentThread();
-    Core::System::GetInstance().PrepareReschedule();
+    LOG_TRACE(Kernel_SVC, "called, pc=0x{:08X}", system.CurrentArmInterface().GetPC());
+
+    auto* const current_thread = system.CurrentScheduler().GetCurrentThread();
+    current_thread->Stop();
+    system.CurrentScheduler().RemoveThread(current_thread);
+    system.PrepareReschedule();
 }
 
 /// Sleep the current thread
@@ -1315,32 +1304,32 @@ static void SleepThread(s64 nanoseconds) {
         YieldAndWaitForLoadBalancing = -2,
     };
 
+    auto& system = Core::System::GetInstance();
+    auto& scheduler = system.CurrentScheduler();
+    auto* const current_thread = scheduler.GetCurrentThread();
+
     if (nanoseconds <= 0) {
-        auto& scheduler{Core::System::GetInstance().CurrentScheduler()};
         switch (static_cast<SleepType>(nanoseconds)) {
         case SleepType::YieldWithoutLoadBalancing:
-            scheduler.YieldWithoutLoadBalancing(GetCurrentThread());
+            scheduler.YieldWithoutLoadBalancing(current_thread);
             break;
         case SleepType::YieldWithLoadBalancing:
-            scheduler.YieldWithLoadBalancing(GetCurrentThread());
+            scheduler.YieldWithLoadBalancing(current_thread);
             break;
         case SleepType::YieldAndWaitForLoadBalancing:
-            scheduler.YieldAndWaitForLoadBalancing(GetCurrentThread());
+            scheduler.YieldAndWaitForLoadBalancing(current_thread);
             break;
         default:
             UNREACHABLE_MSG("Unimplemented sleep yield type '{:016X}'!", nanoseconds);
         }
     } else {
-        // Sleep current thread and check for next thread to schedule
-        WaitCurrentThread_Sleep();
-
-        // Create an event to wake the thread up after the specified nanosecond delay has passed
-        GetCurrentThread()->WakeAfterDelay(nanoseconds);
+        current_thread->Sleep(nanoseconds);
     }
 
     // Reschedule all CPU cores
-    for (std::size_t i = 0; i < Core::NUM_CPU_CORES; ++i)
-        Core::System::GetInstance().CpuCore(i).PrepareReschedule();
+    for (std::size_t i = 0; i < Core::NUM_CPU_CORES; ++i) {
+        system.CpuCore(i).PrepareReschedule();
+    }
 }
 
 /// Wait process wide key atomic
@@ -1494,20 +1483,10 @@ static ResultCode WaitForAddress(VAddr address, u32 type, s32 value, s64 timeout
         return ERR_INVALID_ADDRESS;
     }
 
-    switch (static_cast<AddressArbiter::ArbitrationType>(type)) {
-    case AddressArbiter::ArbitrationType::WaitIfLessThan:
-        return AddressArbiter::WaitForAddressIfLessThan(address, value, timeout, false);
-    case AddressArbiter::ArbitrationType::DecrementAndWaitIfLessThan:
-        return AddressArbiter::WaitForAddressIfLessThan(address, value, timeout, true);
-    case AddressArbiter::ArbitrationType::WaitIfEqual:
-        return AddressArbiter::WaitForAddressIfEqual(address, value, timeout);
-    default:
-        LOG_ERROR(Kernel_SVC,
-                  "Invalid arbitration type, expected WaitIfLessThan, DecrementAndWaitIfLessThan "
-                  "or WaitIfEqual but got {}",
-                  type);
-        return ERR_INVALID_ENUM_VALUE;
-    }
+    const auto arbitration_type = static_cast<AddressArbiter::ArbitrationType>(type);
+    auto& address_arbiter =
+        Core::System::GetInstance().Kernel().CurrentProcess()->GetAddressArbiter();
+    return address_arbiter.WaitForAddress(address, arbitration_type, value, timeout);
 }
 
 // Signals to an address (via Address Arbiter)
@@ -1525,31 +1504,21 @@ static ResultCode SignalToAddress(VAddr address, u32 type, s32 value, s32 num_to
         return ERR_INVALID_ADDRESS;
     }
 
-    switch (static_cast<AddressArbiter::SignalType>(type)) {
-    case AddressArbiter::SignalType::Signal:
-        return AddressArbiter::SignalToAddress(address, num_to_wake);
-    case AddressArbiter::SignalType::IncrementAndSignalIfEqual:
-        return AddressArbiter::IncrementAndSignalToAddressIfEqual(address, value, num_to_wake);
-    case AddressArbiter::SignalType::ModifyByWaitingCountAndSignalIfEqual:
-        return AddressArbiter::ModifyByWaitingCountAndSignalToAddressIfEqual(address, value,
-                                                                             num_to_wake);
-    default:
-        LOG_ERROR(Kernel_SVC,
-                  "Invalid signal type, expected Signal, IncrementAndSignalIfEqual "
-                  "or ModifyByWaitingCountAndSignalIfEqual but got {}",
-                  type);
-        return ERR_INVALID_ENUM_VALUE;
-    }
+    const auto signal_type = static_cast<AddressArbiter::SignalType>(type);
+    auto& address_arbiter =
+        Core::System::GetInstance().Kernel().CurrentProcess()->GetAddressArbiter();
+    return address_arbiter.SignalToAddress(address, signal_type, value, num_to_wake);
 }
 
 /// This returns the total CPU ticks elapsed since the CPU was powered-on
 static u64 GetSystemTick() {
     LOG_TRACE(Kernel_SVC, "called");
 
-    const u64 result{CoreTiming::GetTicks()};
+    auto& core_timing = Core::System::GetInstance().CoreTiming();
+    const u64 result{core_timing.GetTicks()};
 
     // Advance time to defeat dumb games that busy-wait for the frame to end.
-    CoreTiming::AddTicks(400);
+    core_timing.AddTicks(400);
 
     return result;
 }
diff --git a/src/core/hle/kernel/thread.cpp b/src/core/hle/kernel/thread.cpp
index d3984dfc4..3b22e8e0d 100644
--- a/src/core/hle/kernel/thread.cpp
+++ b/src/core/hle/kernel/thread.cpp
@@ -7,8 +7,6 @@
 #include <optional>
 #include <vector>
 
-#include <boost/range/algorithm_ext/erase.hpp>
-
 #include "common/assert.h"
 #include "common/common_types.h"
 #include "common/logging/log.h"
@@ -43,7 +41,8 @@ Thread::~Thread() = default;
 
 void Thread::Stop() {
     // Cancel any outstanding wakeup events for this thread
-    CoreTiming::UnscheduleEvent(kernel.ThreadWakeupCallbackEventType(), callback_handle);
+    Core::System::GetInstance().CoreTiming().UnscheduleEvent(kernel.ThreadWakeupCallbackEventType(),
+                                                             callback_handle);
     kernel.ThreadWakeupCallbackHandleTable().Close(callback_handle);
     callback_handle = 0;
 
@@ -67,17 +66,6 @@ void Thread::Stop() {
     owner_process->FreeTLSSlot(tls_address);
 }
 
-void WaitCurrentThread_Sleep() {
-    Thread* thread = GetCurrentThread();
-    thread->SetStatus(ThreadStatus::WaitSleep);
-}
-
-void ExitCurrentThread() {
-    Thread* thread = GetCurrentThread();
-    thread->Stop();
-    Core::System::GetInstance().CurrentScheduler().RemoveThread(thread);
-}
-
 void Thread::WakeAfterDelay(s64 nanoseconds) {
     // Don't schedule a wakeup if the thread wants to wait forever
     if (nanoseconds == -1)
@@ -85,12 +73,14 @@ void Thread::WakeAfterDelay(s64 nanoseconds) {
 
     // This function might be called from any thread so we have to be cautious and use the
     // thread-safe version of ScheduleEvent.
-    CoreTiming::ScheduleEventThreadsafe(CoreTiming::nsToCycles(nanoseconds),
-                                        kernel.ThreadWakeupCallbackEventType(), callback_handle);
+    Core::System::GetInstance().CoreTiming().ScheduleEventThreadsafe(
+        Core::Timing::nsToCycles(nanoseconds), kernel.ThreadWakeupCallbackEventType(),
+        callback_handle);
 }
 
 void Thread::CancelWakeupTimer() {
-    CoreTiming::UnscheduleEventThreadsafe(kernel.ThreadWakeupCallbackEventType(), callback_handle);
+    Core::System::GetInstance().CoreTiming().UnscheduleEventThreadsafe(
+        kernel.ThreadWakeupCallbackEventType(), callback_handle);
 }
 
 static std::optional<s32> GetNextProcessorId(u64 mask) {
@@ -181,14 +171,13 @@ ResultVal<SharedPtr<Thread>> Thread::Create(KernelCore& kernel, std::string name
         return ERR_INVALID_PROCESSOR_ID;
     }
 
-    // TODO(yuriks): Other checks, returning 0xD9001BEA
-
     if (!Memory::IsValidVirtualAddress(owner_process, entry_point)) {
         LOG_ERROR(Kernel_SVC, "(name={}): invalid entry {:016X}", name, entry_point);
         // TODO (bunnei): Find the correct error code to use here
         return ResultCode(-1);
     }
 
+    auto& system = Core::System::GetInstance();
     SharedPtr<Thread> thread(new Thread(kernel));
 
     thread->thread_id = kernel.CreateNewThreadID();
@@ -197,7 +186,7 @@ ResultVal<SharedPtr<Thread>> Thread::Create(KernelCore& kernel, std::string name
     thread->stack_top = stack_top;
     thread->tpidr_el0 = 0;
     thread->nominal_priority = thread->current_priority = priority;
-    thread->last_running_ticks = CoreTiming::GetTicks();
+    thread->last_running_ticks = system.CoreTiming().GetTicks();
     thread->processor_id = processor_id;
     thread->ideal_core = processor_id;
     thread->affinity_mask = 1ULL << processor_id;
@@ -208,7 +197,7 @@ ResultVal<SharedPtr<Thread>> Thread::Create(KernelCore& kernel, std::string name
     thread->name = std::move(name);
     thread->callback_handle = kernel.ThreadWakeupCallbackHandleTable().Create(thread).Unwrap();
     thread->owner_process = &owner_process;
-    thread->scheduler = &Core::System::GetInstance().Scheduler(processor_id);
+    thread->scheduler = &system.Scheduler(processor_id);
     thread->scheduler->AddThread(thread, priority);
     thread->tls_address = thread->owner_process->MarkNextAvailableTLSSlotAsUsed(*thread);
 
@@ -257,7 +246,7 @@ void Thread::SetStatus(ThreadStatus new_status) {
     }
 
     if (status == ThreadStatus::Running) {
-        last_running_ticks = CoreTiming::GetTicks();
+        last_running_ticks = Core::System::GetInstance().CoreTiming().GetTicks();
     }
 
     status = new_status;
@@ -267,8 +256,8 @@ void Thread::AddMutexWaiter(SharedPtr<Thread> thread) {
     if (thread->lock_owner == this) {
         // If the thread is already waiting for this thread to release the mutex, ensure that the
         // waiters list is consistent and return without doing anything.
-        auto itr = std::find(wait_mutex_threads.begin(), wait_mutex_threads.end(), thread);
-        ASSERT(itr != wait_mutex_threads.end());
+        const auto iter = std::find(wait_mutex_threads.begin(), wait_mutex_threads.end(), thread);
+        ASSERT(iter != wait_mutex_threads.end());
         return;
     }
 
@@ -276,11 +265,16 @@ void Thread::AddMutexWaiter(SharedPtr<Thread> thread) {
     ASSERT(thread->lock_owner == nullptr);
 
     // Ensure that the thread is not already in the list of mutex waiters
-    auto itr = std::find(wait_mutex_threads.begin(), wait_mutex_threads.end(), thread);
-    ASSERT(itr == wait_mutex_threads.end());
-
+    const auto iter = std::find(wait_mutex_threads.begin(), wait_mutex_threads.end(), thread);
+    ASSERT(iter == wait_mutex_threads.end());
+
+    // Keep the list in an ordered fashion
+    const auto insertion_point = std::find_if(
+        wait_mutex_threads.begin(), wait_mutex_threads.end(),
+        [&thread](const auto& entry) { return entry->GetPriority() > thread->GetPriority(); });
+    wait_mutex_threads.insert(insertion_point, thread);
     thread->lock_owner = this;
-    wait_mutex_threads.emplace_back(std::move(thread));
+
     UpdatePriority();
 }
 
@@ -288,32 +282,44 @@ void Thread::RemoveMutexWaiter(SharedPtr<Thread> thread) {
     ASSERT(thread->lock_owner == this);
 
     // Ensure that the thread is in the list of mutex waiters
-    auto itr = std::find(wait_mutex_threads.begin(), wait_mutex_threads.end(), thread);
-    ASSERT(itr != wait_mutex_threads.end());
+    const auto iter = std::find(wait_mutex_threads.begin(), wait_mutex_threads.end(), thread);
+    ASSERT(iter != wait_mutex_threads.end());
+
+    wait_mutex_threads.erase(iter);
 
-    boost::remove_erase(wait_mutex_threads, thread);
     thread->lock_owner = nullptr;
     UpdatePriority();
 }
 
 void Thread::UpdatePriority() {
-    // Find the highest priority among all the threads that are waiting for this thread's lock
+    // If any of the threads waiting on the mutex have a higher priority
+    // (taking into account priority inheritance), then this thread inherits
+    // that thread's priority.
     u32 new_priority = nominal_priority;
-    for (const auto& thread : wait_mutex_threads) {
-        if (thread->nominal_priority < new_priority)
-            new_priority = thread->nominal_priority;
+    if (!wait_mutex_threads.empty()) {
+        if (wait_mutex_threads.front()->current_priority < new_priority) {
+            new_priority = wait_mutex_threads.front()->current_priority;
+        }
     }
 
-    if (new_priority == current_priority)
+    if (new_priority == current_priority) {
         return;
+    }
 
     scheduler->SetThreadPriority(this, new_priority);
-
     current_priority = new_priority;
 
+    if (!lock_owner) {
+        return;
+    }
+
+    // Ensure that the thread is within the correct location in the waiting list.
+    auto old_owner = lock_owner;
+    lock_owner->RemoveMutexWaiter(this);
+    old_owner->AddMutexWaiter(this);
+
     // Recursively update the priority of the thread that depends on the priority of this one.
-    if (lock_owner)
-        lock_owner->UpdatePriority();
+    lock_owner->UpdatePriority();
 }
 
 void Thread::ChangeCore(u32 core, u64 mask) {
@@ -389,6 +395,14 @@ void Thread::SetActivity(ThreadActivity value) {
     }
 }
 
+void Thread::Sleep(s64 nanoseconds) {
+    // Sleep current thread and check for next thread to schedule
+    SetStatus(ThreadStatus::WaitSleep);
+
+    // Create an event to wake the thread up after the specified nanosecond delay has passed
+    WakeAfterDelay(nanoseconds);
+}
+
 ////////////////////////////////////////////////////////////////////////////////////////////////////
 
 /**
diff --git a/src/core/hle/kernel/thread.h b/src/core/hle/kernel/thread.h
index c48b21aba..faad5f391 100644
--- a/src/core/hle/kernel/thread.h
+++ b/src/core/hle/kernel/thread.h
@@ -383,6 +383,9 @@ public:
 
     void SetActivity(ThreadActivity value);
 
+    /// Sleeps this thread for the given amount of nanoseconds.
+    void Sleep(s64 nanoseconds);
+
 private:
     explicit Thread(KernelCore& kernel);
     ~Thread() override;
@@ -398,8 +401,14 @@ private:
     VAddr entry_point = 0;
     VAddr stack_top = 0;
 
-    u32 nominal_priority = 0; ///< Nominal thread priority, as set by the emulated application
-    u32 current_priority = 0; ///< Current thread priority, can be temporarily changed
+    /// Nominal thread priority, as set by the emulated application.
+    /// The nominal priority is the thread priority without priority
+    /// inheritance taken into account.
+    u32 nominal_priority = 0;
+
+    /// Current thread priority. This may change over the course of the
+    /// thread's lifetime in order to facilitate priority inheritance.
+    u32 current_priority = 0;
 
     u64 total_cpu_time_ticks = 0; ///< Total CPU running ticks.
     u64 last_running_ticks = 0;   ///< CPU tick when thread was last running
@@ -460,14 +469,4 @@ private:
  */
 Thread* GetCurrentThread();
 
-/**
- * Waits the current thread on a sleep
- */
-void WaitCurrentThread_Sleep();
-
-/**
- * Stops the current thread and removes it from the thread_list
- */
-void ExitCurrentThread();
-
 } // namespace Kernel
diff --git a/src/core/hle/kernel/vm_manager.cpp b/src/core/hle/kernel/vm_manager.cpp
index 10ad94aa6..3def3e52c 100644
--- a/src/core/hle/kernel/vm_manager.cpp
+++ b/src/core/hle/kernel/vm_manager.cpp
@@ -7,18 +7,18 @@
 #include <utility>
 #include "common/assert.h"
 #include "common/logging/log.h"
+#include "common/memory_hook.h"
 #include "core/arm/arm_interface.h"
 #include "core/core.h"
 #include "core/file_sys/program_metadata.h"
 #include "core/hle/kernel/errors.h"
 #include "core/hle/kernel/vm_manager.h"
 #include "core/memory.h"
-#include "core/memory_hook.h"
 #include "core/memory_setup.h"
 
 namespace Kernel {
-
-static const char* GetMemoryStateName(MemoryState state) {
+namespace {
+const char* GetMemoryStateName(MemoryState state) {
     static constexpr const char* names[] = {
         "Unmapped",         "Io",
         "Normal",           "CodeStatic",
@@ -35,6 +35,14 @@ static const char* GetMemoryStateName(MemoryState state) {
     return names[ToSvcMemoryState(state)];
 }
 
+// Checks if a given address range lies within a larger address range.
+constexpr bool IsInsideAddressRange(VAddr address, u64 size, VAddr address_range_begin,
+                                    VAddr address_range_end) {
+    const VAddr end_address = address + size - 1;
+    return address_range_begin <= address && end_address <= address_range_end - 1;
+}
+} // Anonymous namespace
+
 bool VirtualMemoryArea::CanBeMergedWith(const VirtualMemoryArea& next) const {
     ASSERT(base + size == next.base);
     if (permissions != next.permissions || state != next.state || attribute != next.attribute ||
@@ -169,7 +177,7 @@ ResultVal<VAddr> VMManager::FindFreeRegion(u64 size) const {
 
 ResultVal<VMManager::VMAHandle> VMManager::MapMMIO(VAddr target, PAddr paddr, u64 size,
                                                    MemoryState state,
-                                                   Memory::MemoryHookPointer mmio_handler) {
+                                                   Common::MemoryHookPointer mmio_handler) {
     // This is the appropriately sized VMA that will turn into our allocation.
     CASCADE_RESULT(VMAIter vma_handle, CarveVMA(target, size));
     VirtualMemoryArea& final_vma = vma_handle->second;
@@ -249,8 +257,7 @@ ResultCode VMManager::ReprotectRange(VAddr target, u64 size, VMAPermission new_p
 }
 
 ResultVal<VAddr> VMManager::HeapAllocate(VAddr target, u64 size, VMAPermission perms) {
-    if (target < GetHeapRegionBaseAddress() || target + size > GetHeapRegionEndAddress() ||
-        target + size < target) {
+    if (!IsWithinHeapRegion(target, size)) {
         return ERR_INVALID_ADDRESS;
     }
 
@@ -285,8 +292,7 @@ ResultVal<VAddr> VMManager::HeapAllocate(VAddr target, u64 size, VMAPermission p
 }
 
 ResultCode VMManager::HeapFree(VAddr target, u64 size) {
-    if (target < GetHeapRegionBaseAddress() || target + size > GetHeapRegionEndAddress() ||
-        target + size < target) {
+    if (!IsWithinHeapRegion(target, size)) {
         return ERR_INVALID_ADDRESS;
     }
 
@@ -618,7 +624,7 @@ void VMManager::ClearPageTable() {
     std::fill(page_table.pointers.begin(), page_table.pointers.end(), nullptr);
     page_table.special_regions.clear();
     std::fill(page_table.attributes.begin(), page_table.attributes.end(),
-              Memory::PageType::Unmapped);
+              Common::PageType::Unmapped);
 }
 
 VMManager::CheckResults VMManager::CheckRangeState(VAddr address, u64 size, MemoryState state_mask,
@@ -706,6 +712,11 @@ u64 VMManager::GetAddressSpaceWidth() const {
     return address_space_width;
 }
 
+bool VMManager::IsWithinAddressSpace(VAddr address, u64 size) const {
+    return IsInsideAddressRange(address, size, GetAddressSpaceBaseAddress(),
+                                GetAddressSpaceEndAddress());
+}
+
 VAddr VMManager::GetASLRRegionBaseAddress() const {
     return aslr_region_base;
 }
@@ -750,6 +761,11 @@ u64 VMManager::GetCodeRegionSize() const {
     return code_region_end - code_region_base;
 }
 
+bool VMManager::IsWithinCodeRegion(VAddr address, u64 size) const {
+    return IsInsideAddressRange(address, size, GetCodeRegionBaseAddress(),
+                                GetCodeRegionEndAddress());
+}
+
 VAddr VMManager::GetHeapRegionBaseAddress() const {
     return heap_region_base;
 }
@@ -762,6 +778,11 @@ u64 VMManager::GetHeapRegionSize() const {
     return heap_region_end - heap_region_base;
 }
 
+bool VMManager::IsWithinHeapRegion(VAddr address, u64 size) const {
+    return IsInsideAddressRange(address, size, GetHeapRegionBaseAddress(),
+                                GetHeapRegionEndAddress());
+}
+
 VAddr VMManager::GetMapRegionBaseAddress() const {
     return map_region_base;
 }
@@ -774,6 +795,10 @@ u64 VMManager::GetMapRegionSize() const {
     return map_region_end - map_region_base;
 }
 
+bool VMManager::IsWithinMapRegion(VAddr address, u64 size) const {
+    return IsInsideAddressRange(address, size, GetMapRegionBaseAddress(), GetMapRegionEndAddress());
+}
+
 VAddr VMManager::GetNewMapRegionBaseAddress() const {
     return new_map_region_base;
 }
@@ -786,6 +811,11 @@ u64 VMManager::GetNewMapRegionSize() const {
     return new_map_region_end - new_map_region_base;
 }
 
+bool VMManager::IsWithinNewMapRegion(VAddr address, u64 size) const {
+    return IsInsideAddressRange(address, size, GetNewMapRegionBaseAddress(),
+                                GetNewMapRegionEndAddress());
+}
+
 VAddr VMManager::GetTLSIORegionBaseAddress() const {
     return tls_io_region_base;
 }
@@ -798,4 +828,9 @@ u64 VMManager::GetTLSIORegionSize() const {
     return tls_io_region_end - tls_io_region_base;
 }
 
+bool VMManager::IsWithinTLSIORegion(VAddr address, u64 size) const {
+    return IsInsideAddressRange(address, size, GetTLSIORegionBaseAddress(),
+                                GetTLSIORegionEndAddress());
+}
+
 } // namespace Kernel
diff --git a/src/core/hle/kernel/vm_manager.h b/src/core/hle/kernel/vm_manager.h
index 6091533bc..b96980f8f 100644
--- a/src/core/hle/kernel/vm_manager.h
+++ b/src/core/hle/kernel/vm_manager.h
@@ -9,9 +9,10 @@
 #include <tuple>
 #include <vector>
 #include "common/common_types.h"
+#include "common/memory_hook.h"
+#include "common/page_table.h"
 #include "core/hle/result.h"
 #include "core/memory.h"
-#include "core/memory_hook.h"
 
 namespace FileSys {
 enum class ProgramAddressSpaceType : u8;
@@ -290,7 +291,7 @@ struct VirtualMemoryArea {
     // Settings for type = MMIO
     /// Physical address of the register area this VMA maps to.
     PAddr paddr = 0;
-    Memory::MemoryHookPointer mmio_handler = nullptr;
+    Common::MemoryHookPointer mmio_handler = nullptr;
 
     /// Tests if this area can be merged to the right with `next`.
     bool CanBeMergedWith(const VirtualMemoryArea& next) const;
@@ -368,7 +369,7 @@ public:
      * @param mmio_handler The handler that will implement read and write for this MMIO region.
      */
     ResultVal<VMAHandle> MapMMIO(VAddr target, PAddr paddr, u64 size, MemoryState state,
-                                 Memory::MemoryHookPointer mmio_handler);
+                                 Common::MemoryHookPointer mmio_handler);
 
     /// Unmaps a range of addresses, splitting VMAs as necessary.
     ResultCode UnmapRange(VAddr target, u64 size);
@@ -432,18 +433,21 @@ public:
     /// Gets the address space width in bits.
     u64 GetAddressSpaceWidth() const;
 
+    /// Determines whether or not the given address range lies within the address space.
+    bool IsWithinAddressSpace(VAddr address, u64 size) const;
+
     /// Gets the base address of the ASLR region.
     VAddr GetASLRRegionBaseAddress() const;
 
     /// Gets the end address of the ASLR region.
     VAddr GetASLRRegionEndAddress() const;
 
-    /// Determines whether or not the specified address range is within the ASLR region.
-    bool IsWithinASLRRegion(VAddr address, u64 size) const;
-
     /// Gets the size of the ASLR region
     u64 GetASLRRegionSize() const;
 
+    /// Determines whether or not the specified address range is within the ASLR region.
+    bool IsWithinASLRRegion(VAddr address, u64 size) const;
+
     /// Gets the base address of the code region.
     VAddr GetCodeRegionBaseAddress() const;
 
@@ -453,6 +457,9 @@ public:
     /// Gets the total size of the code region in bytes.
     u64 GetCodeRegionSize() const;
 
+    /// Determines whether or not the specified range is within the code region.
+    bool IsWithinCodeRegion(VAddr address, u64 size) const;
+
     /// Gets the base address of the heap region.
     VAddr GetHeapRegionBaseAddress() const;
 
@@ -462,6 +469,9 @@ public:
     /// Gets the total size of the heap region in bytes.
     u64 GetHeapRegionSize() const;
 
+    /// Determines whether or not the specified range is within the heap region.
+    bool IsWithinHeapRegion(VAddr address, u64 size) const;
+
     /// Gets the base address of the map region.
     VAddr GetMapRegionBaseAddress() const;
 
@@ -471,6 +481,9 @@ public:
     /// Gets the total size of the map region in bytes.
     u64 GetMapRegionSize() const;
 
+    /// Determines whether or not the specified range is within the map region.
+    bool IsWithinMapRegion(VAddr address, u64 size) const;
+
     /// Gets the base address of the new map region.
     VAddr GetNewMapRegionBaseAddress() const;
 
@@ -480,6 +493,9 @@ public:
     /// Gets the total size of the new map region in bytes.
     u64 GetNewMapRegionSize() const;
 
+    /// Determines whether or not the given address range is within the new map region
+    bool IsWithinNewMapRegion(VAddr address, u64 size) const;
+
     /// Gets the base address of the TLS IO region.
     VAddr GetTLSIORegionBaseAddress() const;
 
@@ -489,9 +505,12 @@ public:
     /// Gets the total size of the TLS IO region in bytes.
     u64 GetTLSIORegionSize() const;
 
+    /// Determines if the given address range is within the TLS IO region.
+    bool IsWithinTLSIORegion(VAddr address, u64 size) const;
+
     /// Each VMManager has its own page table, which is set as the main one when the owning process
     /// is scheduled.
-    Memory::PageTable page_table;
+    Common::PageTable page_table{Memory::PAGE_BITS};
 
 private:
     using VMAIter = VMAMap::iterator;
diff --git a/src/core/hle/result.h b/src/core/hle/result.h
index bfb77cc31..ab84f5ddc 100644
--- a/src/core/hle/result.h
+++ b/src/core/hle/result.h
@@ -8,20 +8,11 @@
 #include <utility>
 #include "common/assert.h"
 #include "common/bit_field.h"
-#include "common/common_funcs.h"
 #include "common/common_types.h"
 
 // All the constants in this file come from http://switchbrew.org/index.php?title=Error_codes
 
 /**
- * Detailed description of the error. Code 0 always means success.
- */
-enum class ErrorDescription : u32 {
-    Success = 0,
-    RemoteProcessDead = 301,
-};
-
-/**
  * Identifies the module which caused the error. Error codes can be propagated through a call
  * chain, meaning that this doesn't always correspond to the module where the API call made is
  * contained.
@@ -121,7 +112,7 @@ enum class ErrorModule : u32 {
     ShopN = 811,
 };
 
-/// Encapsulates a CTR-OS error code, allowing it to be separated into its constituent fields.
+/// Encapsulates a Horizon OS error code, allowing it to be separated into its constituent fields.
 union ResultCode {
     u32 raw;
 
@@ -134,17 +125,9 @@ union ResultCode {
 
     constexpr explicit ResultCode(u32 raw) : raw(raw) {}
 
-    constexpr ResultCode(ErrorModule module, ErrorDescription description)
-        : ResultCode(module, static_cast<u32>(description)) {}
-
     constexpr ResultCode(ErrorModule module_, u32 description_)
         : raw(module.FormatValue(module_) | description.FormatValue(description_)) {}
 
-    constexpr ResultCode& operator=(const ResultCode& o) {
-        raw = o.raw;
-        return *this;
-    }
-
     constexpr bool IsSuccess() const {
         return raw == 0;
     }
diff --git a/src/core/hle/service/am/am.cpp b/src/core/hle/service/am/am.cpp
index d1cbe0e44..c750d70ac 100644
--- a/src/core/hle/service/am/am.cpp
+++ b/src/core/hle/service/am/am.cpp
@@ -2,10 +2,10 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#include <algorithm>
 #include <array>
 #include <cinttypes>
 #include <cstring>
-#include <stack>
 #include "audio_core/audio_renderer.h"
 #include "core/core.h"
 #include "core/file_sys/savedata_factory.h"
@@ -93,38 +93,84 @@ void IWindowController::AcquireForegroundRights(Kernel::HLERequestContext& ctx)
 }
 
 IAudioController::IAudioController() : ServiceFramework("IAudioController") {
+    // clang-format off
     static const FunctionInfo functions[] = {
         {0, &IAudioController::SetExpectedMasterVolume, "SetExpectedMasterVolume"},
-        {1, &IAudioController::GetMainAppletExpectedMasterVolume,
-         "GetMainAppletExpectedMasterVolume"},
-        {2, &IAudioController::GetLibraryAppletExpectedMasterVolume,
-         "GetLibraryAppletExpectedMasterVolume"},
-        {3, nullptr, "ChangeMainAppletMasterVolume"},
-        {4, nullptr, "SetTransparentVolumeRate"},
+        {1, &IAudioController::GetMainAppletExpectedMasterVolume, "GetMainAppletExpectedMasterVolume"},
+        {2, &IAudioController::GetLibraryAppletExpectedMasterVolume, "GetLibraryAppletExpectedMasterVolume"},
+        {3, &IAudioController::ChangeMainAppletMasterVolume, "ChangeMainAppletMasterVolume"},
+        {4, &IAudioController::SetTransparentAudioRate, "SetTransparentVolumeRate"},
     };
+    // clang-format on
+
     RegisterHandlers(functions);
 }
 
 IAudioController::~IAudioController() = default;
 
 void IAudioController::SetExpectedMasterVolume(Kernel::HLERequestContext& ctx) {
-    LOG_WARNING(Service_AM, "(STUBBED) called");
+    IPC::RequestParser rp{ctx};
+    const float main_applet_volume_tmp = rp.Pop<float>();
+    const float library_applet_volume_tmp = rp.Pop<float>();
+
+    LOG_DEBUG(Service_AM, "called. main_applet_volume={}, library_applet_volume={}",
+              main_applet_volume_tmp, library_applet_volume_tmp);
+
+    // Ensure the volume values remain within the 0-100% range
+    main_applet_volume = std::clamp(main_applet_volume_tmp, min_allowed_volume, max_allowed_volume);
+    library_applet_volume =
+        std::clamp(library_applet_volume_tmp, min_allowed_volume, max_allowed_volume);
+
     IPC::ResponseBuilder rb{ctx, 2};
     rb.Push(RESULT_SUCCESS);
 }
 
 void IAudioController::GetMainAppletExpectedMasterVolume(Kernel::HLERequestContext& ctx) {
-    LOG_WARNING(Service_AM, "(STUBBED) called");
+    LOG_DEBUG(Service_AM, "called. main_applet_volume={}", main_applet_volume);
     IPC::ResponseBuilder rb{ctx, 3};
     rb.Push(RESULT_SUCCESS);
-    rb.Push(volume);
+    rb.Push(main_applet_volume);
 }
 
 void IAudioController::GetLibraryAppletExpectedMasterVolume(Kernel::HLERequestContext& ctx) {
-    LOG_WARNING(Service_AM, "(STUBBED) called");
+    LOG_DEBUG(Service_AM, "called. library_applet_volume={}", library_applet_volume);
     IPC::ResponseBuilder rb{ctx, 3};
     rb.Push(RESULT_SUCCESS);
-    rb.Push(volume);
+    rb.Push(library_applet_volume);
+}
+
+void IAudioController::ChangeMainAppletMasterVolume(Kernel::HLERequestContext& ctx) {
+    struct Parameters {
+        float volume;
+        s64 fade_time_ns;
+    };
+    static_assert(sizeof(Parameters) == 16);
+
+    IPC::RequestParser rp{ctx};
+    const auto parameters = rp.PopRaw<Parameters>();
+
+    LOG_DEBUG(Service_AM, "called. volume={}, fade_time_ns={}", parameters.volume,
+              parameters.fade_time_ns);
+
+    main_applet_volume = std::clamp(parameters.volume, min_allowed_volume, max_allowed_volume);
+    fade_time_ns = std::chrono::nanoseconds{parameters.fade_time_ns};
+
+    IPC::ResponseBuilder rb{ctx, 2};
+    rb.Push(RESULT_SUCCESS);
+}
+
+void IAudioController::SetTransparentAudioRate(Kernel::HLERequestContext& ctx) {
+    IPC::RequestParser rp{ctx};
+    const float transparent_volume_rate_tmp = rp.Pop<float>();
+
+    LOG_DEBUG(Service_AM, "called. transparent_volume_rate={}", transparent_volume_rate_tmp);
+
+    // Clamp volume range to 0-100%.
+    transparent_volume_rate =
+        std::clamp(transparent_volume_rate_tmp, min_allowed_volume, max_allowed_volume);
+
+    IPC::ResponseBuilder rb{ctx, 2};
+    rb.Push(RESULT_SUCCESS);
 }
 
 IDisplayController::IDisplayController() : ServiceFramework("IDisplayController") {
@@ -322,14 +368,15 @@ void ISelfController::SetScreenShotImageOrientation(Kernel::HLERequestContext& c
 
 void ISelfController::CreateManagedDisplayLayer(Kernel::HLERequestContext& ctx) {
     LOG_WARNING(Service_AM, "(STUBBED) called");
+
     // TODO(Subv): Find out how AM determines the display to use, for now just
     // create the layer in the Default display.
-    u64 display_id = nvflinger->OpenDisplay("Default");
-    u64 layer_id = nvflinger->CreateLayer(display_id);
+    const auto display_id = nvflinger->OpenDisplay("Default");
+    const auto layer_id = nvflinger->CreateLayer(*display_id);
 
     IPC::ResponseBuilder rb{ctx, 4};
     rb.Push(RESULT_SUCCESS);
-    rb.Push(layer_id);
+    rb.Push(*layer_id);
 }
 
 void ISelfController::SetHandlesRequestToDisplay(Kernel::HLERequestContext& ctx) {
diff --git a/src/core/hle/service/am/am.h b/src/core/hle/service/am/am.h
index b6113cfdd..565dd8e9e 100644
--- a/src/core/hle/service/am/am.h
+++ b/src/core/hle/service/am/am.h
@@ -4,6 +4,7 @@
 
 #pragma once
 
+#include <chrono>
 #include <memory>
 #include <queue>
 #include "core/hle/kernel/writable_event.h"
@@ -81,8 +82,21 @@ private:
     void SetExpectedMasterVolume(Kernel::HLERequestContext& ctx);
     void GetMainAppletExpectedMasterVolume(Kernel::HLERequestContext& ctx);
     void GetLibraryAppletExpectedMasterVolume(Kernel::HLERequestContext& ctx);
+    void ChangeMainAppletMasterVolume(Kernel::HLERequestContext& ctx);
+    void SetTransparentAudioRate(Kernel::HLERequestContext& ctx);
 
-    u32 volume{100};
+    static constexpr float min_allowed_volume = 0.0f;
+    static constexpr float max_allowed_volume = 1.0f;
+
+    float main_applet_volume{0.25f};
+    float library_applet_volume{max_allowed_volume};
+    float transparent_volume_rate{min_allowed_volume};
+
+    // Volume transition fade time in nanoseconds.
+    // e.g. If the main applet volume was 0% and was changed to 50%
+    //      with a fade of 50ns, then over the course of 50ns,
+    //      the volume will gradually fade up to 50%
+    std::chrono::nanoseconds fade_time_ns{0};
 };
 
 class IDisplayController final : public ServiceFramework<IDisplayController> {
diff --git a/src/core/hle/service/am/applets/software_keyboard.cpp b/src/core/hle/service/am/applets/software_keyboard.cpp
index f255f74b5..8c5bd6059 100644
--- a/src/core/hle/service/am/applets/software_keyboard.cpp
+++ b/src/core/hle/service/am/applets/software_keyboard.cpp
@@ -7,6 +7,7 @@
 #include "common/string_util.h"
 #include "core/core.h"
 #include "core/frontend/applets/software_keyboard.h"
+#include "core/hle/result.h"
 #include "core/hle/service/am/am.h"
 #include "core/hle/service/am/applets/software_keyboard.h"
 
diff --git a/src/core/hle/service/am/applets/software_keyboard.h b/src/core/hle/service/am/applets/software_keyboard.h
index efd5753a1..b93a30d28 100644
--- a/src/core/hle/service/am/applets/software_keyboard.h
+++ b/src/core/hle/service/am/applets/software_keyboard.h
@@ -9,10 +9,13 @@
 #include <vector>
 
 #include "common/common_funcs.h"
+#include "common/common_types.h"
 #include "common/swap.h"
 #include "core/hle/service/am/am.h"
 #include "core/hle/service/am/applets/applets.h"
 
+union ResultCode;
+
 namespace Service::AM::Applets {
 
 enum class KeysetDisable : u32 {
diff --git a/src/core/hle/service/audio/audout_u.cpp b/src/core/hle/service/audio/audout_u.cpp
index dc6a6b188..21f5e64c7 100644
--- a/src/core/hle/service/audio/audout_u.cpp
+++ b/src/core/hle/service/audio/audout_u.cpp
@@ -18,17 +18,11 @@
 #include "core/hle/kernel/readable_event.h"
 #include "core/hle/kernel/writable_event.h"
 #include "core/hle/service/audio/audout_u.h"
+#include "core/hle/service/audio/errors.h"
 #include "core/memory.h"
 
 namespace Service::Audio {
 
-namespace ErrCodes {
-enum {
-    ErrorUnknown = 2,
-    BufferCountExceeded = 8,
-};
-}
-
 constexpr std::array<char, 10> DefaultDevice{{"DeviceOut"}};
 constexpr int DefaultSampleRate{48000};
 
@@ -68,12 +62,12 @@ public:
         RegisterHandlers(functions);
 
         // This is the event handle used to check if the audio buffer was released
-        auto& kernel = Core::System::GetInstance().Kernel();
-        buffer_event = Kernel::WritableEvent::CreateEventPair(kernel, Kernel::ResetType::Sticky,
-                                                              "IAudioOutBufferReleased");
+        auto& system = Core::System::GetInstance();
+        buffer_event = Kernel::WritableEvent::CreateEventPair(
+            system.Kernel(), Kernel::ResetType::Sticky, "IAudioOutBufferReleased");
 
-        stream = audio_core.OpenStream(audio_params.sample_rate, audio_params.channel_count,
-                                       std::move(unique_name),
+        stream = audio_core.OpenStream(system.CoreTiming(), audio_params.sample_rate,
+                                       audio_params.channel_count, std::move(unique_name),
                                        [=]() { buffer_event.writable->Signal(); });
     }
 
@@ -100,7 +94,7 @@ private:
 
         if (stream->IsPlaying()) {
             IPC::ResponseBuilder rb{ctx, 2};
-            rb.Push(ResultCode(ErrorModule::Audio, ErrCodes::ErrorUnknown));
+            rb.Push(ERR_OPERATION_FAILED);
             return;
         }
 
@@ -113,7 +107,9 @@ private:
     void StopAudioOut(Kernel::HLERequestContext& ctx) {
         LOG_DEBUG(Service_Audio, "called");
 
-        audio_core.StopStream(stream);
+        if (stream->IsPlaying()) {
+            audio_core.StopStream(stream);
+        }
 
         IPC::ResponseBuilder rb{ctx, 2};
         rb.Push(RESULT_SUCCESS);
@@ -143,7 +139,8 @@ private:
 
         if (!audio_core.QueueBuffer(stream, tag, std::move(samples))) {
             IPC::ResponseBuilder rb{ctx, 2};
-            rb.Push(ResultCode(ErrorModule::Audio, ErrCodes::BufferCountExceeded));
+            rb.Push(ERR_BUFFER_COUNT_EXCEEDED);
+            return;
         }
 
         IPC::ResponseBuilder rb{ctx, 2};
diff --git a/src/core/hle/service/audio/audren_u.cpp b/src/core/hle/service/audio/audren_u.cpp
index 76cc48254..c9de10a24 100644
--- a/src/core/hle/service/audio/audren_u.cpp
+++ b/src/core/hle/service/audio/audren_u.cpp
@@ -17,6 +17,7 @@
 #include "core/hle/kernel/readable_event.h"
 #include "core/hle/kernel/writable_event.h"
 #include "core/hle/service/audio/audren_u.h"
+#include "core/hle/service/audio/errors.h"
 
 namespace Service::Audio {
 
@@ -37,15 +38,16 @@ public:
             {8, &IAudioRenderer::SetRenderingTimeLimit, "SetRenderingTimeLimit"},
             {9, &IAudioRenderer::GetRenderingTimeLimit, "GetRenderingTimeLimit"},
             {10, &IAudioRenderer::RequestUpdateImpl, "RequestUpdateAuto"},
-            {11, nullptr, "ExecuteAudioRendererRendering"},
+            {11, &IAudioRenderer::ExecuteAudioRendererRendering, "ExecuteAudioRendererRendering"},
         };
         // clang-format on
         RegisterHandlers(functions);
 
-        auto& kernel = Core::System::GetInstance().Kernel();
-        system_event = Kernel::WritableEvent::CreateEventPair(kernel, Kernel::ResetType::Sticky,
-                                                              "IAudioRenderer:SystemEvent");
-        renderer = std::make_unique<AudioCore::AudioRenderer>(audren_params, system_event.writable);
+        auto& system = Core::System::GetInstance();
+        system_event = Kernel::WritableEvent::CreateEventPair(
+            system.Kernel(), Kernel::ResetType::Sticky, "IAudioRenderer:SystemEvent");
+        renderer = std::make_unique<AudioCore::AudioRenderer>(system.CoreTiming(), audren_params,
+                                                              system_event.writable);
     }
 
 private:
@@ -137,6 +139,17 @@ private:
         rb.Push(rendering_time_limit_percent);
     }
 
+    void ExecuteAudioRendererRendering(Kernel::HLERequestContext& ctx) {
+        LOG_DEBUG(Service_Audio, "called");
+
+        // This service command currently only reports an unsupported operation
+        // error code, or aborts. Given that, we just always return an error
+        // code in this case.
+
+        IPC::ResponseBuilder rb{ctx, 2};
+        rb.Push(ERR_NOT_SUPPORTED);
+    }
+
     Kernel::EventPair system_event;
     std::unique_ptr<AudioCore::AudioRenderer> renderer;
     u32 rendering_time_limit_percent = 100;
@@ -234,7 +247,7 @@ AudRenU::AudRenU() : ServiceFramework("audren:u") {
         {0, &AudRenU::OpenAudioRenderer, "OpenAudioRenderer"},
         {1, &AudRenU::GetAudioRendererWorkBufferSize, "GetAudioRendererWorkBufferSize"},
         {2, &AudRenU::GetAudioDeviceService, "GetAudioDeviceService"},
-        {3, nullptr, "OpenAudioRendererAuto"},
+        {3, &AudRenU::OpenAudioRendererAuto, "OpenAudioRendererAuto"},
         {4, &AudRenU::GetAudioDeviceServiceWithRevisionInfo, "GetAudioDeviceServiceWithRevisionInfo"},
     };
     // clang-format on
@@ -247,12 +260,7 @@ AudRenU::~AudRenU() = default;
 void AudRenU::OpenAudioRenderer(Kernel::HLERequestContext& ctx) {
     LOG_DEBUG(Service_Audio, "called");
 
-    IPC::RequestParser rp{ctx};
-    auto params = rp.PopRaw<AudioCore::AudioRendererParameter>();
-    IPC::ResponseBuilder rb{ctx, 2, 0, 1};
-
-    rb.Push(RESULT_SUCCESS);
-    rb.PushIpcInterface<Audio::IAudioRenderer>(std::move(params));
+    OpenAudioRendererImpl(ctx);
 }
 
 void AudRenU::GetAudioRendererWorkBufferSize(Kernel::HLERequestContext& ctx) {
@@ -261,20 +269,20 @@ void AudRenU::GetAudioRendererWorkBufferSize(Kernel::HLERequestContext& ctx) {
     LOG_DEBUG(Service_Audio, "called");
 
     u64 buffer_sz = Common::AlignUp(4 * params.mix_buffer_count, 0x40);
-    buffer_sz += params.unknown_c * 1024;
-    buffer_sz += 0x940 * (params.unknown_c + 1);
+    buffer_sz += params.submix_count * 1024;
+    buffer_sz += 0x940 * (params.submix_count + 1);
     buffer_sz += 0x3F0 * params.voice_count;
-    buffer_sz += Common::AlignUp(8 * (params.unknown_c + 1), 0x10);
+    buffer_sz += Common::AlignUp(8 * (params.submix_count + 1), 0x10);
     buffer_sz += Common::AlignUp(8 * params.voice_count, 0x10);
-    buffer_sz +=
-        Common::AlignUp((0x3C0 * (params.sink_count + params.unknown_c) + 4 * params.sample_count) *
-                            (params.mix_buffer_count + 6),
-                        0x40);
+    buffer_sz += Common::AlignUp(
+        (0x3C0 * (params.sink_count + params.submix_count) + 4 * params.sample_count) *
+            (params.mix_buffer_count + 6),
+        0x40);
 
     if (IsFeatureSupported(AudioFeatures::Splitter, params.revision)) {
-        u32 count = params.unknown_c + 1;
+        const u32 count = params.submix_count + 1;
         u64 node_count = Common::AlignUp(count, 0x40);
-        u64 node_state_buffer_sz =
+        const u64 node_state_buffer_sz =
             4 * (node_count * node_count) + 0xC * node_count + 2 * (node_count / 8);
         u64 edge_matrix_buffer_sz = 0;
         node_count = Common::AlignUp(count * count, 0x40);
@@ -288,19 +296,19 @@ void AudRenU::GetAudioRendererWorkBufferSize(Kernel::HLERequestContext& ctx) {
 
     buffer_sz += 0x20 * (params.effect_count + 4 * params.voice_count) + 0x50;
     if (IsFeatureSupported(AudioFeatures::Splitter, params.revision)) {
-        buffer_sz += 0xE0 * params.unknown_2c;
+        buffer_sz += 0xE0 * params.num_splitter_send_channels;
         buffer_sz += 0x20 * params.splitter_count;
-        buffer_sz += Common::AlignUp(4 * params.unknown_2c, 0x10);
+        buffer_sz += Common::AlignUp(4 * params.num_splitter_send_channels, 0x10);
     }
     buffer_sz = Common::AlignUp(buffer_sz, 0x40) + 0x170 * params.sink_count;
     u64 output_sz = buffer_sz + 0x280 * params.sink_count + 0x4B0 * params.effect_count +
                     ((params.voice_count * 256) | 0x40);
 
-    if (params.unknown_1c >= 1) {
+    if (params.performance_frame_count >= 1) {
         output_sz = Common::AlignUp(((16 * params.sink_count + 16 * params.effect_count +
                                       16 * params.voice_count + 16) +
                                      0x658) *
-                                            (params.unknown_1c + 1) +
+                                            (params.performance_frame_count + 1) +
                                         0xc0,
                                     0x40) +
                     output_sz;
@@ -324,6 +332,12 @@ void AudRenU::GetAudioDeviceService(Kernel::HLERequestContext& ctx) {
     rb.PushIpcInterface<Audio::IAudioDevice>();
 }
 
+void AudRenU::OpenAudioRendererAuto(Kernel::HLERequestContext& ctx) {
+    LOG_DEBUG(Service_Audio, "called");
+
+    OpenAudioRendererImpl(ctx);
+}
+
 void AudRenU::GetAudioDeviceServiceWithRevisionInfo(Kernel::HLERequestContext& ctx) {
     LOG_WARNING(Service_Audio, "(STUBBED) called");
 
@@ -334,6 +348,15 @@ void AudRenU::GetAudioDeviceServiceWithRevisionInfo(Kernel::HLERequestContext& c
                                                 // based on the current revision
 }
 
+void AudRenU::OpenAudioRendererImpl(Kernel::HLERequestContext& ctx) {
+    IPC::RequestParser rp{ctx};
+    const auto params = rp.PopRaw<AudioCore::AudioRendererParameter>();
+    IPC::ResponseBuilder rb{ctx, 2, 0, 1};
+
+    rb.Push(RESULT_SUCCESS);
+    rb.PushIpcInterface<IAudioRenderer>(params);
+}
+
 bool AudRenU::IsFeatureSupported(AudioFeatures feature, u32_le revision) const {
     u32_be version_num = (revision - Common::MakeMagic('R', 'E', 'V', '0')); // Byte swap
     switch (feature) {
diff --git a/src/core/hle/service/audio/audren_u.h b/src/core/hle/service/audio/audren_u.h
index 3d63388fb..e55d25973 100644
--- a/src/core/hle/service/audio/audren_u.h
+++ b/src/core/hle/service/audio/audren_u.h
@@ -21,8 +21,11 @@ private:
     void OpenAudioRenderer(Kernel::HLERequestContext& ctx);
     void GetAudioRendererWorkBufferSize(Kernel::HLERequestContext& ctx);
     void GetAudioDeviceService(Kernel::HLERequestContext& ctx);
+    void OpenAudioRendererAuto(Kernel::HLERequestContext& ctx);
     void GetAudioDeviceServiceWithRevisionInfo(Kernel::HLERequestContext& ctx);
 
+    void OpenAudioRendererImpl(Kernel::HLERequestContext& ctx);
+
     enum class AudioFeatures : u32 {
         Splitter,
     };
diff --git a/src/core/hle/service/audio/errors.h b/src/core/hle/service/audio/errors.h
new file mode 100644
index 000000000..6f8c09bcf
--- /dev/null
+++ b/src/core/hle/service/audio/errors.h
@@ -0,0 +1,15 @@
+// Copyright 2019 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "core/hle/result.h"
+
+namespace Service::Audio {
+
+constexpr ResultCode ERR_OPERATION_FAILED{ErrorModule::Audio, 2};
+constexpr ResultCode ERR_BUFFER_COUNT_EXCEEDED{ErrorModule::Audio, 8};
+constexpr ResultCode ERR_NOT_SUPPORTED{ErrorModule::Audio, 513};
+
+} // namespace Service::Audio
diff --git a/src/core/hle/service/audio/hwopus.cpp b/src/core/hle/service/audio/hwopus.cpp
index 11eba4a12..cb4a1160d 100644
--- a/src/core/hle/service/audio/hwopus.cpp
+++ b/src/core/hle/service/audio/hwopus.cpp
@@ -8,44 +8,34 @@
 #include <vector>
 
 #include <opus.h>
+#include <opus_multistream.h>
 
-#include "common/common_funcs.h"
+#include "common/assert.h"
 #include "common/logging/log.h"
 #include "core/hle/ipc_helpers.h"
 #include "core/hle/kernel/hle_ipc.h"
 #include "core/hle/service/audio/hwopus.h"
 
 namespace Service::Audio {
-
+namespace {
 struct OpusDeleter {
-    void operator()(void* ptr) const {
-        operator delete(ptr);
+    void operator()(OpusMSDecoder* ptr) const {
+        opus_multistream_decoder_destroy(ptr);
     }
 };
 
-class IHardwareOpusDecoderManager final : public ServiceFramework<IHardwareOpusDecoderManager> {
-public:
-    IHardwareOpusDecoderManager(std::unique_ptr<OpusDecoder, OpusDeleter> decoder, u32 sample_rate,
-                                u32 channel_count)
-        : ServiceFramework("IHardwareOpusDecoderManager"), decoder(std::move(decoder)),
-          sample_rate(sample_rate), channel_count(channel_count) {
-        // clang-format off
-        static const FunctionInfo functions[] = {
-            {0, &IHardwareOpusDecoderManager::DecodeInterleavedOld, "DecodeInterleavedOld"},
-            {1, nullptr, "SetContext"},
-            {2, nullptr, "DecodeInterleavedForMultiStreamOld"},
-            {3, nullptr, "SetContextForMultiStream"},
-            {4, &IHardwareOpusDecoderManager::DecodeInterleavedWithPerfOld, "DecodeInterleavedWithPerfOld"},
-            {5, nullptr, "DecodeInterleavedForMultiStreamWithPerfOld"},
-            {6, &IHardwareOpusDecoderManager::DecodeInterleaved, "DecodeInterleaved"},
-            {7, nullptr, "DecodeInterleavedForMultiStream"},
-        };
-        // clang-format on
+using OpusDecoderPtr = std::unique_ptr<OpusMSDecoder, OpusDeleter>;
 
-        RegisterHandlers(functions);
-    }
+struct OpusPacketHeader {
+    // Packet size in bytes.
+    u32_be size;
+    // Indicates the final range of the codec's entropy coder.
+    u32_be final_range;
+};
+static_assert(sizeof(OpusPacketHeader) == 0x8, "OpusHeader is an invalid size");
 
-private:
+class OpusDecoderState {
+public:
     /// Describes extra behavior that may be asked of the decoding context.
     enum class ExtraBehavior {
         /// No extra behavior.
@@ -55,30 +45,27 @@ private:
         ResetContext,
     };
 
-    void DecodeInterleavedOld(Kernel::HLERequestContext& ctx) {
-        LOG_DEBUG(Audio, "called");
-
-        DecodeInterleavedHelper(ctx, nullptr, ExtraBehavior::None);
-    }
-
-    void DecodeInterleavedWithPerfOld(Kernel::HLERequestContext& ctx) {
-        LOG_DEBUG(Audio, "called");
-
-        u64 performance = 0;
-        DecodeInterleavedHelper(ctx, &performance, ExtraBehavior::None);
-    }
-
-    void DecodeInterleaved(Kernel::HLERequestContext& ctx) {
-        LOG_DEBUG(Audio, "called");
-
-        IPC::RequestParser rp{ctx};
-        const auto extra_behavior =
-            rp.Pop<bool>() ? ExtraBehavior::ResetContext : ExtraBehavior::None;
+    enum class PerfTime {
+        Disabled,
+        Enabled,
+    };
 
-        u64 performance = 0;
-        DecodeInterleavedHelper(ctx, &performance, extra_behavior);
+    explicit OpusDecoderState(OpusDecoderPtr decoder, u32 sample_rate, u32 channel_count)
+        : decoder{std::move(decoder)}, sample_rate{sample_rate}, channel_count{channel_count} {}
+
+    // Decodes interleaved Opus packets. Optionally allows reporting time taken to
+    // perform the decoding, as well as any relevant extra behavior.
+    void DecodeInterleaved(Kernel::HLERequestContext& ctx, PerfTime perf_time,
+                           ExtraBehavior extra_behavior) {
+        if (perf_time == PerfTime::Disabled) {
+            DecodeInterleavedHelper(ctx, nullptr, extra_behavior);
+        } else {
+            u64 performance = 0;
+            DecodeInterleavedHelper(ctx, &performance, extra_behavior);
+        }
     }
 
+private:
     void DecodeInterleavedHelper(Kernel::HLERequestContext& ctx, u64* performance,
                                  ExtraBehavior extra_behavior) {
         u32 consumed = 0;
@@ -89,8 +76,7 @@ private:
             ResetDecoderContext();
         }
 
-        if (!Decoder_DecodeInterleaved(consumed, sample_count, ctx.ReadBuffer(), samples,
-                                       performance)) {
+        if (!DecodeOpusData(consumed, sample_count, ctx.ReadBuffer(), samples, performance)) {
             LOG_ERROR(Audio, "Failed to decode opus data");
             IPC::ResponseBuilder rb{ctx, 2};
             // TODO(ogniK): Use correct error code
@@ -109,27 +95,27 @@ private:
         ctx.WriteBuffer(samples.data(), samples.size() * sizeof(s16));
     }
 
-    bool Decoder_DecodeInterleaved(u32& consumed, u32& sample_count, const std::vector<u8>& input,
-                                   std::vector<opus_int16>& output, u64* out_performance_time) {
+    bool DecodeOpusData(u32& consumed, u32& sample_count, const std::vector<u8>& input,
+                        std::vector<opus_int16>& output, u64* out_performance_time) const {
         const auto start_time = std::chrono::high_resolution_clock::now();
         const std::size_t raw_output_sz = output.size() * sizeof(opus_int16);
-        if (sizeof(OpusHeader) > input.size()) {
+        if (sizeof(OpusPacketHeader) > input.size()) {
             LOG_ERROR(Audio, "Input is smaller than the header size, header_sz={}, input_sz={}",
-                      sizeof(OpusHeader), input.size());
+                      sizeof(OpusPacketHeader), input.size());
             return false;
         }
 
-        OpusHeader hdr{};
-        std::memcpy(&hdr, input.data(), sizeof(OpusHeader));
-        if (sizeof(OpusHeader) + static_cast<u32>(hdr.sz) > input.size()) {
+        OpusPacketHeader hdr{};
+        std::memcpy(&hdr, input.data(), sizeof(OpusPacketHeader));
+        if (sizeof(OpusPacketHeader) + static_cast<u32>(hdr.size) > input.size()) {
             LOG_ERROR(Audio, "Input does not fit in the opus header size. data_sz={}, input_sz={}",
-                      sizeof(OpusHeader) + static_cast<u32>(hdr.sz), input.size());
+                      sizeof(OpusPacketHeader) + static_cast<u32>(hdr.size), input.size());
             return false;
         }
 
-        const auto frame = input.data() + sizeof(OpusHeader);
+        const auto frame = input.data() + sizeof(OpusPacketHeader);
         const auto decoded_sample_count = opus_packet_get_nb_samples(
-            frame, static_cast<opus_int32>(input.size() - sizeof(OpusHeader)),
+            frame, static_cast<opus_int32>(input.size() - sizeof(OpusPacketHeader)),
             static_cast<opus_int32>(sample_rate));
         if (decoded_sample_count * channel_count * sizeof(u16) > raw_output_sz) {
             LOG_ERROR(
@@ -141,18 +127,18 @@ private:
 
         const int frame_size = (static_cast<int>(raw_output_sz / sizeof(s16) / channel_count));
         const auto out_sample_count =
-            opus_decode(decoder.get(), frame, hdr.sz, output.data(), frame_size, 0);
+            opus_multistream_decode(decoder.get(), frame, hdr.size, output.data(), frame_size, 0);
         if (out_sample_count < 0) {
             LOG_ERROR(Audio,
                       "Incorrect sample count received from opus_decode, "
                       "output_sample_count={}, frame_size={}, data_sz_from_hdr={}",
-                      out_sample_count, frame_size, static_cast<u32>(hdr.sz));
+                      out_sample_count, frame_size, static_cast<u32>(hdr.size));
             return false;
         }
 
         const auto end_time = std::chrono::high_resolution_clock::now() - start_time;
         sample_count = out_sample_count;
-        consumed = static_cast<u32>(sizeof(OpusHeader) + hdr.sz);
+        consumed = static_cast<u32>(sizeof(OpusPacketHeader) + hdr.size);
         if (out_performance_time != nullptr) {
             *out_performance_time =
                 std::chrono::duration_cast<std::chrono::milliseconds>(end_time).count();
@@ -164,25 +150,86 @@ private:
     void ResetDecoderContext() {
         ASSERT(decoder != nullptr);
 
-        opus_decoder_ctl(decoder.get(), OPUS_RESET_STATE);
+        opus_multistream_decoder_ctl(decoder.get(), OPUS_RESET_STATE);
     }
 
-    struct OpusHeader {
-        u32_be sz; // Needs to be BE for some odd reason
-        INSERT_PADDING_WORDS(1);
-    };
-    static_assert(sizeof(OpusHeader) == 0x8, "OpusHeader is an invalid size");
-
-    std::unique_ptr<OpusDecoder, OpusDeleter> decoder;
+    OpusDecoderPtr decoder;
     u32 sample_rate;
     u32 channel_count;
 };
 
-static std::size_t WorkerBufferSize(u32 channel_count) {
+class IHardwareOpusDecoderManager final : public ServiceFramework<IHardwareOpusDecoderManager> {
+public:
+    explicit IHardwareOpusDecoderManager(OpusDecoderState decoder_state)
+        : ServiceFramework("IHardwareOpusDecoderManager"), decoder_state{std::move(decoder_state)} {
+        // clang-format off
+        static const FunctionInfo functions[] = {
+            {0, &IHardwareOpusDecoderManager::DecodeInterleavedOld, "DecodeInterleavedOld"},
+            {1, nullptr, "SetContext"},
+            {2, nullptr, "DecodeInterleavedForMultiStreamOld"},
+            {3, nullptr, "SetContextForMultiStream"},
+            {4, &IHardwareOpusDecoderManager::DecodeInterleavedWithPerfOld, "DecodeInterleavedWithPerfOld"},
+            {5, nullptr, "DecodeInterleavedForMultiStreamWithPerfOld"},
+            {6, &IHardwareOpusDecoderManager::DecodeInterleaved, "DecodeInterleaved"},
+            {7, nullptr, "DecodeInterleavedForMultiStream"},
+        };
+        // clang-format on
+
+        RegisterHandlers(functions);
+    }
+
+private:
+    void DecodeInterleavedOld(Kernel::HLERequestContext& ctx) {
+        LOG_DEBUG(Audio, "called");
+
+        decoder_state.DecodeInterleaved(ctx, OpusDecoderState::PerfTime::Disabled,
+                                        OpusDecoderState::ExtraBehavior::None);
+    }
+
+    void DecodeInterleavedWithPerfOld(Kernel::HLERequestContext& ctx) {
+        LOG_DEBUG(Audio, "called");
+
+        decoder_state.DecodeInterleaved(ctx, OpusDecoderState::PerfTime::Enabled,
+                                        OpusDecoderState::ExtraBehavior::None);
+    }
+
+    void DecodeInterleaved(Kernel::HLERequestContext& ctx) {
+        LOG_DEBUG(Audio, "called");
+
+        IPC::RequestParser rp{ctx};
+        const auto extra_behavior = rp.Pop<bool>() ? OpusDecoderState::ExtraBehavior::ResetContext
+                                                   : OpusDecoderState::ExtraBehavior::None;
+
+        decoder_state.DecodeInterleaved(ctx, OpusDecoderState::PerfTime::Enabled, extra_behavior);
+    }
+
+    OpusDecoderState decoder_state;
+};
+
+std::size_t WorkerBufferSize(u32 channel_count) {
     ASSERT_MSG(channel_count == 1 || channel_count == 2, "Invalid channel count");
-    return opus_decoder_get_size(static_cast<int>(channel_count));
+    constexpr int num_streams = 1;
+    const int num_stereo_streams = channel_count == 2 ? 1 : 0;
+    return opus_multistream_decoder_get_size(num_streams, num_stereo_streams);
 }
 
+// Creates the mapping table that maps the input channels to the particular
+// output channels. In the stereo case, we map the left and right input channels
+// to the left and right output channels respectively.
+//
+// However, in the monophonic case, we only map the one available channel
+// to the sole output channel. We specify 255 for the would-be right channel
+// as this is a special value defined by Opus to indicate to the decoder to
+// ignore that channel.
+std::array<u8, 2> CreateMappingTable(u32 channel_count) {
+    if (channel_count == 2) {
+        return {{0, 1}};
+    }
+
+    return {{0, 255}};
+}
+} // Anonymous namespace
+
 void HwOpus::GetWorkBufferSize(Kernel::HLERequestContext& ctx) {
     IPC::RequestParser rp{ctx};
     const auto sample_rate = rp.Pop<u32>();
@@ -220,10 +267,15 @@ void HwOpus::OpenOpusDecoder(Kernel::HLERequestContext& ctx) {
     const std::size_t worker_sz = WorkerBufferSize(channel_count);
     ASSERT_MSG(buffer_sz >= worker_sz, "Worker buffer too large");
 
-    std::unique_ptr<OpusDecoder, OpusDeleter> decoder{
-        static_cast<OpusDecoder*>(operator new(worker_sz))};
-    if (const int err = opus_decoder_init(decoder.get(), sample_rate, channel_count)) {
-        LOG_ERROR(Audio, "Failed to init opus decoder with error={}", err);
+    const int num_stereo_streams = channel_count == 2 ? 1 : 0;
+    const auto mapping_table = CreateMappingTable(channel_count);
+
+    int error = 0;
+    OpusDecoderPtr decoder{
+        opus_multistream_decoder_create(sample_rate, static_cast<int>(channel_count), 1,
+                                        num_stereo_streams, mapping_table.data(), &error)};
+    if (error != OPUS_OK || decoder == nullptr) {
+        LOG_ERROR(Audio, "Failed to create Opus decoder (error={}).", error);
         IPC::ResponseBuilder rb{ctx, 2};
         // TODO(ogniK): Use correct error code
         rb.Push(ResultCode(-1));
@@ -232,8 +284,8 @@ void HwOpus::OpenOpusDecoder(Kernel::HLERequestContext& ctx) {
 
     IPC::ResponseBuilder rb{ctx, 2, 0, 1};
     rb.Push(RESULT_SUCCESS);
-    rb.PushIpcInterface<IHardwareOpusDecoderManager>(std::move(decoder), sample_rate,
-                                                     channel_count);
+    rb.PushIpcInterface<IHardwareOpusDecoderManager>(
+        OpusDecoderState{std::move(decoder), sample_rate, channel_count});
 }
 
 HwOpus::HwOpus() : ServiceFramework("hwopus") {
diff --git a/src/core/hle/service/filesystem/fsp_srv.cpp b/src/core/hle/service/filesystem/fsp_srv.cpp
index 54959edd8..f03fb629c 100644
--- a/src/core/hle/service/filesystem/fsp_srv.cpp
+++ b/src/core/hle/service/filesystem/fsp_srv.cpp
@@ -733,7 +733,10 @@ FSP_SRV::FSP_SRV() : ServiceFramework("fsp-srv") {
 FSP_SRV::~FSP_SRV() = default;
 
 void FSP_SRV::SetCurrentProcess(Kernel::HLERequestContext& ctx) {
-    LOG_WARNING(Service_FS, "(STUBBED) called");
+    IPC::RequestParser rp{ctx};
+    current_process_id = rp.Pop<u64>();
+
+    LOG_DEBUG(Service_FS, "called. current_process_id=0x{:016X}", current_process_id);
 
     IPC::ResponseBuilder rb{ctx, 2};
     rb.Push(RESULT_SUCCESS);
diff --git a/src/core/hle/service/filesystem/fsp_srv.h b/src/core/hle/service/filesystem/fsp_srv.h
index 3a5f4e200..d7572ba7a 100644
--- a/src/core/hle/service/filesystem/fsp_srv.h
+++ b/src/core/hle/service/filesystem/fsp_srv.h
@@ -32,6 +32,7 @@ private:
     void OpenPatchDataStorageByCurrentProcess(Kernel::HLERequestContext& ctx);
 
     FileSys::VirtualFile romfs;
+    u64 current_process_id = 0;
 };
 
 } // namespace Service::FileSystem
diff --git a/src/core/hle/service/hid/controllers/controller_base.h b/src/core/hle/service/hid/controllers/controller_base.h
index f0e092b1b..5e5097a03 100644
--- a/src/core/hle/service/hid/controllers/controller_base.h
+++ b/src/core/hle/service/hid/controllers/controller_base.h
@@ -7,6 +7,10 @@
 #include "common/common_types.h"
 #include "common/swap.h"
 
+namespace Core::Timing {
+class CoreTiming;
+}
+
 namespace Service::HID {
 class ControllerBase {
 public:
@@ -20,7 +24,8 @@ public:
     virtual void OnRelease() = 0;
 
     // When the controller is requesting an update for the shared memory
-    virtual void OnUpdate(u8* data, std::size_t size) = 0;
+    virtual void OnUpdate(const Core::Timing::CoreTiming& core_timing, u8* data,
+                          std::size_t size) = 0;
 
     // Called when input devices should be loaded
     virtual void OnLoadInputDevices() = 0;
diff --git a/src/core/hle/service/hid/controllers/debug_pad.cpp b/src/core/hle/service/hid/controllers/debug_pad.cpp
index c22357d8c..c5c2e032a 100644
--- a/src/core/hle/service/hid/controllers/debug_pad.cpp
+++ b/src/core/hle/service/hid/controllers/debug_pad.cpp
@@ -21,8 +21,9 @@ void Controller_DebugPad::OnInit() {}
 
 void Controller_DebugPad::OnRelease() {}
 
-void Controller_DebugPad::OnUpdate(u8* data, std::size_t size) {
-    shared_memory.header.timestamp = CoreTiming::GetTicks();
+void Controller_DebugPad::OnUpdate(const Core::Timing::CoreTiming& core_timing, u8* data,
+                                   std::size_t size) {
+    shared_memory.header.timestamp = core_timing.GetTicks();
     shared_memory.header.total_entry_count = 17;
 
     if (!IsControllerActivated()) {
diff --git a/src/core/hle/service/hid/controllers/debug_pad.h b/src/core/hle/service/hid/controllers/debug_pad.h
index 2b60ead12..e584b92ec 100644
--- a/src/core/hle/service/hid/controllers/debug_pad.h
+++ b/src/core/hle/service/hid/controllers/debug_pad.h
@@ -26,7 +26,7 @@ public:
     void OnRelease() override;
 
     // When the controller is requesting an update for the shared memory
-    void OnUpdate(u8* data, std::size_t size) override;
+    void OnUpdate(const Core::Timing::CoreTiming& core_timing, u8* data, std::size_t size) override;
 
     // Called when input devices should be loaded
     void OnLoadInputDevices() override;
diff --git a/src/core/hle/service/hid/controllers/gesture.cpp b/src/core/hle/service/hid/controllers/gesture.cpp
index 898572277..a179252e3 100644
--- a/src/core/hle/service/hid/controllers/gesture.cpp
+++ b/src/core/hle/service/hid/controllers/gesture.cpp
@@ -17,8 +17,9 @@ void Controller_Gesture::OnInit() {}
 
 void Controller_Gesture::OnRelease() {}
 
-void Controller_Gesture::OnUpdate(u8* data, std::size_t size) {
-    shared_memory.header.timestamp = CoreTiming::GetTicks();
+void Controller_Gesture::OnUpdate(const Core::Timing::CoreTiming& core_timing, u8* data,
+                                  std::size_t size) {
+    shared_memory.header.timestamp = core_timing.GetTicks();
     shared_memory.header.total_entry_count = 17;
 
     if (!IsControllerActivated()) {
diff --git a/src/core/hle/service/hid/controllers/gesture.h b/src/core/hle/service/hid/controllers/gesture.h
index 1056ffbcd..f305fe90f 100644
--- a/src/core/hle/service/hid/controllers/gesture.h
+++ b/src/core/hle/service/hid/controllers/gesture.h
@@ -22,7 +22,7 @@ public:
     void OnRelease() override;
 
     // When the controller is requesting an update for the shared memory
-    void OnUpdate(u8* data, size_t size) override;
+    void OnUpdate(const Core::Timing::CoreTiming& core_timing, u8* data, size_t size) override;
 
     // Called when input devices should be loaded
     void OnLoadInputDevices() override;
diff --git a/src/core/hle/service/hid/controllers/keyboard.cpp b/src/core/hle/service/hid/controllers/keyboard.cpp
index ca75adc2b..92d7bfb52 100644
--- a/src/core/hle/service/hid/controllers/keyboard.cpp
+++ b/src/core/hle/service/hid/controllers/keyboard.cpp
@@ -19,8 +19,9 @@ void Controller_Keyboard::OnInit() {}
 
 void Controller_Keyboard::OnRelease() {}
 
-void Controller_Keyboard::OnUpdate(u8* data, std::size_t size) {
-    shared_memory.header.timestamp = CoreTiming::GetTicks();
+void Controller_Keyboard::OnUpdate(const Core::Timing::CoreTiming& core_timing, u8* data,
+                                   std::size_t size) {
+    shared_memory.header.timestamp = core_timing.GetTicks();
     shared_memory.header.total_entry_count = 17;
 
     if (!IsControllerActivated()) {
diff --git a/src/core/hle/service/hid/controllers/keyboard.h b/src/core/hle/service/hid/controllers/keyboard.h
index f52775456..73cd2c7bb 100644
--- a/src/core/hle/service/hid/controllers/keyboard.h
+++ b/src/core/hle/service/hid/controllers/keyboard.h
@@ -25,7 +25,7 @@ public:
     void OnRelease() override;
 
     // When the controller is requesting an update for the shared memory
-    void OnUpdate(u8* data, std::size_t size) override;
+    void OnUpdate(const Core::Timing::CoreTiming& core_timing, u8* data, std::size_t size) override;
 
     // Called when input devices should be loaded
     void OnLoadInputDevices() override;
diff --git a/src/core/hle/service/hid/controllers/mouse.cpp b/src/core/hle/service/hid/controllers/mouse.cpp
index 63391dbe9..11ab096d9 100644
--- a/src/core/hle/service/hid/controllers/mouse.cpp
+++ b/src/core/hle/service/hid/controllers/mouse.cpp
@@ -17,8 +17,9 @@ Controller_Mouse::~Controller_Mouse() = default;
 void Controller_Mouse::OnInit() {}
 void Controller_Mouse::OnRelease() {}
 
-void Controller_Mouse::OnUpdate(u8* data, std::size_t size) {
-    shared_memory.header.timestamp = CoreTiming::GetTicks();
+void Controller_Mouse::OnUpdate(const Core::Timing::CoreTiming& core_timing, u8* data,
+                                std::size_t size) {
+    shared_memory.header.timestamp = core_timing.GetTicks();
     shared_memory.header.total_entry_count = 17;
 
     if (!IsControllerActivated()) {
diff --git a/src/core/hle/service/hid/controllers/mouse.h b/src/core/hle/service/hid/controllers/mouse.h
index 70b654d07..9d46eecbe 100644
--- a/src/core/hle/service/hid/controllers/mouse.h
+++ b/src/core/hle/service/hid/controllers/mouse.h
@@ -24,7 +24,7 @@ public:
     void OnRelease() override;
 
     // When the controller is requesting an update for the shared memory
-    void OnUpdate(u8* data, std::size_t size) override;
+    void OnUpdate(const Core::Timing::CoreTiming& core_timing, u8* data, std::size_t size) override;
 
     // Called when input devices should be loaded
     void OnLoadInputDevices() override;
diff --git a/src/core/hle/service/hid/controllers/npad.cpp b/src/core/hle/service/hid/controllers/npad.cpp
index 04c8c35a8..e7fc7a619 100644
--- a/src/core/hle/service/hid/controllers/npad.cpp
+++ b/src/core/hle/service/hid/controllers/npad.cpp
@@ -288,7 +288,8 @@ void Controller_NPad::RequestPadStateUpdate(u32 npad_id) {
     rstick_entry.y = static_cast<s32>(stick_r_y_f * HID_JOYSTICK_MAX);
 }
 
-void Controller_NPad::OnUpdate(u8* data, std::size_t data_len) {
+void Controller_NPad::OnUpdate(const Core::Timing::CoreTiming& core_timing, u8* data,
+                               std::size_t data_len) {
     if (!IsControllerActivated())
         return;
     for (std::size_t i = 0; i < shared_memory_entries.size(); i++) {
@@ -308,7 +309,7 @@ void Controller_NPad::OnUpdate(u8* data, std::size_t data_len) {
             const auto& last_entry =
                 main_controller->npad[main_controller->common.last_entry_index];
 
-            main_controller->common.timestamp = CoreTiming::GetTicks();
+            main_controller->common.timestamp = core_timing.GetTicks();
             main_controller->common.last_entry_index =
                 (main_controller->common.last_entry_index + 1) % 17;
 
diff --git a/src/core/hle/service/hid/controllers/npad.h b/src/core/hle/service/hid/controllers/npad.h
index ce057da82..4ff50b3cd 100644
--- a/src/core/hle/service/hid/controllers/npad.h
+++ b/src/core/hle/service/hid/controllers/npad.h
@@ -30,7 +30,7 @@ public:
     void OnRelease() override;
 
     // When the controller is requesting an update for the shared memory
-    void OnUpdate(u8* data, std::size_t size) override;
+    void OnUpdate(const Core::Timing::CoreTiming& core_timing, u8* data, std::size_t size) override;
 
     // Called when input devices should be loaded
     void OnLoadInputDevices() override;
diff --git a/src/core/hle/service/hid/controllers/stubbed.cpp b/src/core/hle/service/hid/controllers/stubbed.cpp
index 02fcfadd9..946948f5e 100644
--- a/src/core/hle/service/hid/controllers/stubbed.cpp
+++ b/src/core/hle/service/hid/controllers/stubbed.cpp
@@ -16,13 +16,14 @@ void Controller_Stubbed::OnInit() {}
 
 void Controller_Stubbed::OnRelease() {}
 
-void Controller_Stubbed::OnUpdate(u8* data, std::size_t size) {
+void Controller_Stubbed::OnUpdate(const Core::Timing::CoreTiming& core_timing, u8* data,
+                                  std::size_t size) {
     if (!smart_update) {
         return;
     }
 
     CommonHeader header{};
-    header.timestamp = CoreTiming::GetTicks();
+    header.timestamp = core_timing.GetTicks();
     header.total_entry_count = 17;
     header.entry_count = 0;
     header.last_entry_index = 0;
diff --git a/src/core/hle/service/hid/controllers/stubbed.h b/src/core/hle/service/hid/controllers/stubbed.h
index 4a21c643e..24469f03e 100644
--- a/src/core/hle/service/hid/controllers/stubbed.h
+++ b/src/core/hle/service/hid/controllers/stubbed.h
@@ -20,7 +20,7 @@ public:
     void OnRelease() override;
 
     // When the controller is requesting an update for the shared memory
-    void OnUpdate(u8* data, std::size_t size) override;
+    void OnUpdate(const Core::Timing::CoreTiming& core_timing, u8* data, std::size_t size) override;
 
     // Called when input devices should be loaded
     void OnLoadInputDevices() override;
diff --git a/src/core/hle/service/hid/controllers/touchscreen.cpp b/src/core/hle/service/hid/controllers/touchscreen.cpp
index f666b1bd8..1a8445a43 100644
--- a/src/core/hle/service/hid/controllers/touchscreen.cpp
+++ b/src/core/hle/service/hid/controllers/touchscreen.cpp
@@ -20,8 +20,9 @@ void Controller_Touchscreen::OnInit() {}
 
 void Controller_Touchscreen::OnRelease() {}
 
-void Controller_Touchscreen::OnUpdate(u8* data, std::size_t size) {
-    shared_memory.header.timestamp = CoreTiming::GetTicks();
+void Controller_Touchscreen::OnUpdate(const Core::Timing::CoreTiming& core_timing, u8* data,
+                                      std::size_t size) {
+    shared_memory.header.timestamp = core_timing.GetTicks();
     shared_memory.header.total_entry_count = 17;
 
     if (!IsControllerActivated()) {
@@ -48,7 +49,7 @@ void Controller_Touchscreen::OnUpdate(u8* data, std::size_t size) {
         touch_entry.diameter_x = Settings::values.touchscreen.diameter_x;
         touch_entry.diameter_y = Settings::values.touchscreen.diameter_y;
         touch_entry.rotation_angle = Settings::values.touchscreen.rotation_angle;
-        const u64 tick = CoreTiming::GetTicks();
+        const u64 tick = core_timing.GetTicks();
         touch_entry.delta_time = tick - last_touch;
         last_touch = tick;
         touch_entry.finger = Settings::values.touchscreen.finger;
diff --git a/src/core/hle/service/hid/controllers/touchscreen.h b/src/core/hle/service/hid/controllers/touchscreen.h
index be2583864..76fc340e9 100644
--- a/src/core/hle/service/hid/controllers/touchscreen.h
+++ b/src/core/hle/service/hid/controllers/touchscreen.h
@@ -24,7 +24,7 @@ public:
     void OnRelease() override;
 
     // When the controller is requesting an update for the shared memory
-    void OnUpdate(u8* data, std::size_t size) override;
+    void OnUpdate(const Core::Timing::CoreTiming& core_timing, u8* data, std::size_t size) override;
 
     // Called when input devices should be loaded
     void OnLoadInputDevices() override;
diff --git a/src/core/hle/service/hid/controllers/xpad.cpp b/src/core/hle/service/hid/controllers/xpad.cpp
index cd397c70b..1a9da9576 100644
--- a/src/core/hle/service/hid/controllers/xpad.cpp
+++ b/src/core/hle/service/hid/controllers/xpad.cpp
@@ -17,9 +17,10 @@ void Controller_XPad::OnInit() {}
 
 void Controller_XPad::OnRelease() {}
 
-void Controller_XPad::OnUpdate(u8* data, std::size_t size) {
+void Controller_XPad::OnUpdate(const Core::Timing::CoreTiming& core_timing, u8* data,
+                               std::size_t size) {
     for (auto& xpad_entry : shared_memory.shared_memory_entries) {
-        xpad_entry.header.timestamp = CoreTiming::GetTicks();
+        xpad_entry.header.timestamp = core_timing.GetTicks();
         xpad_entry.header.total_entry_count = 17;
 
         if (!IsControllerActivated()) {
diff --git a/src/core/hle/service/hid/controllers/xpad.h b/src/core/hle/service/hid/controllers/xpad.h
index ff836989f..2864e6617 100644
--- a/src/core/hle/service/hid/controllers/xpad.h
+++ b/src/core/hle/service/hid/controllers/xpad.h
@@ -22,7 +22,7 @@ public:
     void OnRelease() override;
 
     // When the controller is requesting an update for the shared memory
-    void OnUpdate(u8* data, std::size_t size) override;
+    void OnUpdate(const Core::Timing::CoreTiming& core_timing, u8* data, std::size_t size) override;
 
     // Called when input devices should be loaded
     void OnLoadInputDevices() override;
diff --git a/src/core/hle/service/hid/hid.cpp b/src/core/hle/service/hid/hid.cpp
index 008bf3f02..8a6de83a2 100644
--- a/src/core/hle/service/hid/hid.cpp
+++ b/src/core/hle/service/hid/hid.cpp
@@ -36,9 +36,9 @@ namespace Service::HID {
 
 // Updating period for each HID device.
 // TODO(ogniK): Find actual polling rate of hid
-constexpr u64 pad_update_ticks = CoreTiming::BASE_CLOCK_RATE / 66;
-constexpr u64 accelerometer_update_ticks = CoreTiming::BASE_CLOCK_RATE / 100;
-constexpr u64 gyroscope_update_ticks = CoreTiming::BASE_CLOCK_RATE / 100;
+constexpr u64 pad_update_ticks = Core::Timing::BASE_CLOCK_RATE / 66;
+constexpr u64 accelerometer_update_ticks = Core::Timing::BASE_CLOCK_RATE / 100;
+constexpr u64 gyroscope_update_ticks = Core::Timing::BASE_CLOCK_RATE / 100;
 constexpr std::size_t SHARED_MEMORY_SIZE = 0x40000;
 
 IAppletResource::IAppletResource() : ServiceFramework("IAppletResource") {
@@ -73,14 +73,15 @@ IAppletResource::IAppletResource() : ServiceFramework("IAppletResource") {
     GetController<Controller_Stubbed>(HidController::Unknown3).SetCommonHeaderOffset(0x5000);
 
     // Register update callbacks
+    auto& core_timing = Core::System::GetInstance().CoreTiming();
     pad_update_event =
-        CoreTiming::RegisterEvent("HID::UpdatePadCallback", [this](u64 userdata, int cycles_late) {
+        core_timing.RegisterEvent("HID::UpdatePadCallback", [this](u64 userdata, int cycles_late) {
             UpdateControllers(userdata, cycles_late);
         });
 
     // TODO(shinyquagsire23): Other update callbacks? (accel, gyro?)
 
-    CoreTiming::ScheduleEvent(pad_update_ticks, pad_update_event);
+    core_timing.ScheduleEvent(pad_update_ticks, pad_update_event);
 
     ReloadInputDevices();
 }
@@ -94,7 +95,7 @@ void IAppletResource::DeactivateController(HidController controller) {
 }
 
 IAppletResource ::~IAppletResource() {
-    CoreTiming::UnscheduleEvent(pad_update_event, 0);
+    Core::System::GetInstance().CoreTiming().UnscheduleEvent(pad_update_event, 0);
 }
 
 void IAppletResource::GetSharedMemoryHandle(Kernel::HLERequestContext& ctx) {
@@ -106,15 +107,17 @@ void IAppletResource::GetSharedMemoryHandle(Kernel::HLERequestContext& ctx) {
 }
 
 void IAppletResource::UpdateControllers(u64 userdata, int cycles_late) {
+    auto& core_timing = Core::System::GetInstance().CoreTiming();
+
     const bool should_reload = Settings::values.is_device_reload_pending.exchange(false);
     for (const auto& controller : controllers) {
         if (should_reload) {
             controller->OnLoadInputDevices();
         }
-        controller->OnUpdate(shared_mem->GetPointer(), SHARED_MEMORY_SIZE);
+        controller->OnUpdate(core_timing, shared_mem->GetPointer(), SHARED_MEMORY_SIZE);
     }
 
-    CoreTiming::ScheduleEvent(pad_update_ticks - cycles_late, pad_update_event);
+    core_timing.ScheduleEvent(pad_update_ticks - cycles_late, pad_update_event);
 }
 
 class IActiveVibrationDeviceList final : public ServiceFramework<IActiveVibrationDeviceList> {
diff --git a/src/core/hle/service/hid/hid.h b/src/core/hle/service/hid/hid.h
index eca27c056..7cc58db4c 100644
--- a/src/core/hle/service/hid/hid.h
+++ b/src/core/hle/service/hid/hid.h
@@ -7,7 +7,7 @@
 #include "controllers/controller_base.h"
 #include "core/hle/service/service.h"
 
-namespace CoreTiming {
+namespace Core::Timing {
 struct EventType;
 }
 
@@ -15,7 +15,7 @@ namespace Kernel {
 class SharedMemory;
 }
 
-namespace SM {
+namespace Service::SM {
 class ServiceManager;
 }
 
@@ -66,7 +66,7 @@ private:
 
     Kernel::SharedPtr<Kernel::SharedMemory> shared_mem;
 
-    CoreTiming::EventType* pad_update_event;
+    Core::Timing::EventType* pad_update_event;
 
     std::array<std::unique_ptr<ControllerBase>, static_cast<size_t>(HidController::MaxControllers)>
         controllers{};
diff --git a/src/core/hle/service/hid/irs.cpp b/src/core/hle/service/hid/irs.cpp
index 3c7f8b1ee..2c4625c99 100644
--- a/src/core/hle/service/hid/irs.cpp
+++ b/src/core/hle/service/hid/irs.cpp
@@ -98,7 +98,7 @@ void IRS::GetImageTransferProcessorState(Kernel::HLERequestContext& ctx) {
 
     IPC::ResponseBuilder rb{ctx, 5};
     rb.Push(RESULT_SUCCESS);
-    rb.PushRaw<u64>(CoreTiming::GetTicks());
+    rb.PushRaw<u64>(Core::System::GetInstance().CoreTiming().GetTicks());
     rb.PushRaw<u32>(0);
 }
 
diff --git a/src/core/hle/service/nvdrv/devices/nvdisp_disp0.cpp b/src/core/hle/service/nvdrv/devices/nvdisp_disp0.cpp
index 92acc57b1..20c7c39aa 100644
--- a/src/core/hle/service/nvdrv/devices/nvdisp_disp0.cpp
+++ b/src/core/hle/service/nvdrv/devices/nvdisp_disp0.cpp
@@ -23,11 +23,11 @@ u32 nvdisp_disp0::ioctl(Ioctl command, const std::vector<u8>& input, std::vector
 
 void nvdisp_disp0::flip(u32 buffer_handle, u32 offset, u32 format, u32 width, u32 height,
                         u32 stride, NVFlinger::BufferQueue::BufferTransformFlags transform,
-                        const MathUtil::Rectangle<int>& crop_rect) {
+                        const Common::Rectangle<int>& crop_rect) {
     VAddr addr = nvmap_dev->GetObjectAddress(buffer_handle);
-    LOG_WARNING(Service,
-                "Drawing from address {:X} offset {:08X} Width {} Height {} Stride {} Format {}",
-                addr, offset, width, height, stride, format);
+    LOG_TRACE(Service,
+              "Drawing from address {:X} offset {:08X} Width {} Height {} Stride {} Format {}",
+              addr, offset, width, height, stride, format);
 
     using PixelFormat = Tegra::FramebufferConfig::PixelFormat;
     const Tegra::FramebufferConfig framebuffer{
@@ -36,7 +36,7 @@ void nvdisp_disp0::flip(u32 buffer_handle, u32 offset, u32 format, u32 width, u3
 
     auto& instance = Core::System::GetInstance();
     instance.GetPerfStats().EndGameFrame();
-    instance.Renderer().SwapBuffers(framebuffer);
+    instance.GPU().SwapBuffers(framebuffer);
 }
 
 } // namespace Service::Nvidia::Devices
diff --git a/src/core/hle/service/nvdrv/devices/nvdisp_disp0.h b/src/core/hle/service/nvdrv/devices/nvdisp_disp0.h
index a45086e45..ace71169f 100644
--- a/src/core/hle/service/nvdrv/devices/nvdisp_disp0.h
+++ b/src/core/hle/service/nvdrv/devices/nvdisp_disp0.h
@@ -25,7 +25,7 @@ public:
     /// Performs a screen flip, drawing the buffer pointed to by the handle.
     void flip(u32 buffer_handle, u32 offset, u32 format, u32 width, u32 height, u32 stride,
               NVFlinger::BufferQueue::BufferTransformFlags transform,
-              const MathUtil::Rectangle<int>& crop_rect);
+              const Common::Rectangle<int>& crop_rect);
 
 private:
     std::shared_ptr<nvmap> nvmap_dev;
diff --git a/src/core/hle/service/nvdrv/devices/nvhost_as_gpu.cpp b/src/core/hle/service/nvdrv/devices/nvhost_as_gpu.cpp
index 466db7ccd..b031ebc66 100644
--- a/src/core/hle/service/nvdrv/devices/nvhost_as_gpu.cpp
+++ b/src/core/hle/service/nvdrv/devices/nvhost_as_gpu.cpp
@@ -10,6 +10,7 @@
 #include "core/core.h"
 #include "core/hle/service/nvdrv/devices/nvhost_as_gpu.h"
 #include "core/hle/service/nvdrv/devices/nvmap.h"
+#include "core/memory.h"
 #include "video_core/memory_manager.h"
 #include "video_core/rasterizer_interface.h"
 #include "video_core/renderer_base.h"
@@ -178,7 +179,7 @@ u32 nvhost_as_gpu::UnmapBuffer(const std::vector<u8>& input, std::vector<u8>& ou
     auto& gpu = system_instance.GPU();
     auto cpu_addr = gpu.MemoryManager().GpuToCpuAddress(params.offset);
     ASSERT(cpu_addr);
-    system_instance.Renderer().Rasterizer().FlushAndInvalidateRegion(*cpu_addr, itr->second.size);
+    gpu.FlushAndInvalidateRegion(ToCacheAddr(Memory::GetPointer(*cpu_addr)), itr->second.size);
 
     params.offset = gpu.MemoryManager().UnmapBuffer(params.offset, itr->second.size);
 
diff --git a/src/core/hle/service/nvdrv/devices/nvhost_ctrl_gpu.cpp b/src/core/hle/service/nvdrv/devices/nvhost_ctrl_gpu.cpp
index d57a54ee8..45812d238 100644
--- a/src/core/hle/service/nvdrv/devices/nvhost_ctrl_gpu.cpp
+++ b/src/core/hle/service/nvdrv/devices/nvhost_ctrl_gpu.cpp
@@ -5,6 +5,7 @@
 #include <cstring>
 #include "common/assert.h"
 #include "common/logging/log.h"
+#include "core/core.h"
 #include "core/core_timing.h"
 #include "core/core_timing_util.h"
 #include "core/hle/service/nvdrv/devices/nvhost_ctrl_gpu.h"
@@ -184,7 +185,7 @@ u32 nvhost_ctrl_gpu::GetGpuTime(const std::vector<u8>& input, std::vector<u8>& o
 
     IoctlGetGpuTime params{};
     std::memcpy(&params, input.data(), input.size());
-    params.gpu_time = CoreTiming::cyclesToNs(CoreTiming::GetTicks());
+    params.gpu_time = Core::Timing::cyclesToNs(Core::System::GetInstance().CoreTiming().GetTicks());
     std::memcpy(output.data(), &params, output.size());
     return 0;
 }
diff --git a/src/core/hle/service/nvdrv/devices/nvhost_gpu.cpp b/src/core/hle/service/nvdrv/devices/nvhost_gpu.cpp
index 0a650f36c..8ce7bc7a5 100644
--- a/src/core/hle/service/nvdrv/devices/nvhost_gpu.cpp
+++ b/src/core/hle/service/nvdrv/devices/nvhost_gpu.cpp
@@ -136,16 +136,6 @@ u32 nvhost_gpu::AllocateObjectContext(const std::vector<u8>& input, std::vector<
     return 0;
 }
 
-static void PushGPUEntries(Tegra::CommandList&& entries) {
-    if (entries.empty()) {
-        return;
-    }
-
-    auto& dma_pusher{Core::System::GetInstance().GPU().DmaPusher()};
-    dma_pusher.Push(std::move(entries));
-    dma_pusher.DispatchCalls();
-}
-
 u32 nvhost_gpu::SubmitGPFIFO(const std::vector<u8>& input, std::vector<u8>& output) {
     if (input.size() < sizeof(IoctlSubmitGpfifo)) {
         UNIMPLEMENTED();
@@ -163,7 +153,7 @@ u32 nvhost_gpu::SubmitGPFIFO(const std::vector<u8>& input, std::vector<u8>& outp
     std::memcpy(entries.data(), &input[sizeof(IoctlSubmitGpfifo)],
                 params.num_entries * sizeof(Tegra::CommandListHeader));
 
-    PushGPUEntries(std::move(entries));
+    Core::System::GetInstance().GPU().PushGPUEntries(std::move(entries));
 
     params.fence_out.id = 0;
     params.fence_out.value = 0;
@@ -184,7 +174,7 @@ u32 nvhost_gpu::KickoffPB(const std::vector<u8>& input, std::vector<u8>& output)
     Memory::ReadBlock(params.address, entries.data(),
                       params.num_entries * sizeof(Tegra::CommandListHeader));
 
-    PushGPUEntries(std::move(entries));
+    Core::System::GetInstance().GPU().PushGPUEntries(std::move(entries));
 
     params.fence_out.id = 0;
     params.fence_out.value = 0;
diff --git a/src/core/hle/service/nvflinger/buffer_queue.cpp b/src/core/hle/service/nvflinger/buffer_queue.cpp
index fc07d9bb8..4d150fc71 100644
--- a/src/core/hle/service/nvflinger/buffer_queue.cpp
+++ b/src/core/hle/service/nvflinger/buffer_queue.cpp
@@ -63,7 +63,7 @@ const IGBPBuffer& BufferQueue::RequestBuffer(u32 slot) const {
 }
 
 void BufferQueue::QueueBuffer(u32 slot, BufferTransformFlags transform,
-                              const MathUtil::Rectangle<int>& crop_rect) {
+                              const Common::Rectangle<int>& crop_rect) {
     auto itr = std::find_if(queue.begin(), queue.end(),
                             [&](const Buffer& buffer) { return buffer.slot == slot; });
     ASSERT(itr != queue.end());
diff --git a/src/core/hle/service/nvflinger/buffer_queue.h b/src/core/hle/service/nvflinger/buffer_queue.h
index b171f256c..e1ccb6171 100644
--- a/src/core/hle/service/nvflinger/buffer_queue.h
+++ b/src/core/hle/service/nvflinger/buffer_queue.h
@@ -13,10 +13,6 @@
 #include "core/hle/kernel/object.h"
 #include "core/hle/kernel/writable_event.h"
 
-namespace CoreTiming {
-struct EventType;
-}
-
 namespace Service::NVFlinger {
 
 struct IGBPBuffer {
@@ -71,14 +67,14 @@ public:
         Status status = Status::Free;
         IGBPBuffer igbp_buffer;
         BufferTransformFlags transform;
-        MathUtil::Rectangle<int> crop_rect;
+        Common::Rectangle<int> crop_rect;
     };
 
     void SetPreallocatedBuffer(u32 slot, const IGBPBuffer& igbp_buffer);
     std::optional<u32> DequeueBuffer(u32 width, u32 height);
     const IGBPBuffer& RequestBuffer(u32 slot) const;
     void QueueBuffer(u32 slot, BufferTransformFlags transform,
-                     const MathUtil::Rectangle<int>& crop_rect);
+                     const Common::Rectangle<int>& crop_rect);
     std::optional<std::reference_wrapper<const Buffer>> AcquireBuffer();
     void ReleaseBuffer(u32 slot);
     u32 Query(QueryType type);
diff --git a/src/core/hle/service/nvflinger/nvflinger.cpp b/src/core/hle/service/nvflinger/nvflinger.cpp
index 8dfc0df03..fc496b654 100644
--- a/src/core/hle/service/nvflinger/nvflinger.cpp
+++ b/src/core/hle/service/nvflinger/nvflinger.cpp
@@ -14,135 +14,170 @@
 #include "core/core_timing_util.h"
 #include "core/hle/kernel/kernel.h"
 #include "core/hle/kernel/readable_event.h"
-#include "core/hle/kernel/writable_event.h"
 #include "core/hle/service/nvdrv/devices/nvdisp_disp0.h"
 #include "core/hle/service/nvdrv/nvdrv.h"
 #include "core/hle/service/nvflinger/buffer_queue.h"
 #include "core/hle/service/nvflinger/nvflinger.h"
+#include "core/hle/service/vi/display/vi_display.h"
+#include "core/hle/service/vi/layer/vi_layer.h"
 #include "core/perf_stats.h"
 #include "video_core/renderer_base.h"
 
 namespace Service::NVFlinger {
 
 constexpr std::size_t SCREEN_REFRESH_RATE = 60;
-constexpr u64 frame_ticks = static_cast<u64>(CoreTiming::BASE_CLOCK_RATE / SCREEN_REFRESH_RATE);
+constexpr u64 frame_ticks = static_cast<u64>(Core::Timing::BASE_CLOCK_RATE / SCREEN_REFRESH_RATE);
+
+NVFlinger::NVFlinger(Core::Timing::CoreTiming& core_timing) : core_timing{core_timing} {
+    displays.emplace_back(0, "Default");
+    displays.emplace_back(1, "External");
+    displays.emplace_back(2, "Edid");
+    displays.emplace_back(3, "Internal");
+    displays.emplace_back(4, "Null");
 
-NVFlinger::NVFlinger() {
     // Schedule the screen composition events
     composition_event =
-        CoreTiming::RegisterEvent("ScreenComposition", [this](u64 userdata, int cycles_late) {
+        core_timing.RegisterEvent("ScreenComposition", [this](u64 userdata, int cycles_late) {
             Compose();
-            CoreTiming::ScheduleEvent(frame_ticks - cycles_late, composition_event);
+            this->core_timing.ScheduleEvent(frame_ticks - cycles_late, composition_event);
         });
 
-    CoreTiming::ScheduleEvent(frame_ticks, composition_event);
+    core_timing.ScheduleEvent(frame_ticks, composition_event);
 }
 
 NVFlinger::~NVFlinger() {
-    CoreTiming::UnscheduleEvent(composition_event, 0);
+    core_timing.UnscheduleEvent(composition_event, 0);
 }
 
 void NVFlinger::SetNVDrvInstance(std::shared_ptr<Nvidia::Module> instance) {
     nvdrv = std::move(instance);
 }
 
-u64 NVFlinger::OpenDisplay(std::string_view name) {
+std::optional<u64> NVFlinger::OpenDisplay(std::string_view name) {
     LOG_DEBUG(Service, "Opening \"{}\" display", name);
 
     // TODO(Subv): Currently we only support the Default display.
     ASSERT(name == "Default");
 
-    const auto itr = std::find_if(displays.begin(), displays.end(),
-                                  [&](const Display& display) { return display.name == name; });
-
-    ASSERT(itr != displays.end());
+    const auto itr =
+        std::find_if(displays.begin(), displays.end(),
+                     [&](const VI::Display& display) { return display.GetName() == name; });
+    if (itr == displays.end()) {
+        return {};
+    }
 
-    return itr->id;
+    return itr->GetID();
 }
 
-u64 NVFlinger::CreateLayer(u64 display_id) {
-    auto& display = FindDisplay(display_id);
+std::optional<u64> NVFlinger::CreateLayer(u64 display_id) {
+    auto* const display = FindDisplay(display_id);
 
-    ASSERT_MSG(display.layers.empty(), "Only one layer is supported per display at the moment");
+    if (display == nullptr) {
+        return {};
+    }
 
     const u64 layer_id = next_layer_id++;
     const u32 buffer_queue_id = next_buffer_queue_id++;
-    auto buffer_queue = std::make_shared<BufferQueue>(buffer_queue_id, layer_id);
-    display.layers.emplace_back(layer_id, buffer_queue);
-    buffer_queues.emplace_back(std::move(buffer_queue));
+    buffer_queues.emplace_back(buffer_queue_id, layer_id);
+    display->CreateLayer(layer_id, buffer_queues.back());
     return layer_id;
 }
 
-u32 NVFlinger::FindBufferQueueId(u64 display_id, u64 layer_id) const {
-    const auto& layer = FindLayer(display_id, layer_id);
-    return layer.buffer_queue->GetId();
+std::optional<u32> NVFlinger::FindBufferQueueId(u64 display_id, u64 layer_id) const {
+    const auto* const layer = FindLayer(display_id, layer_id);
+
+    if (layer == nullptr) {
+        return {};
+    }
+
+    return layer->GetBufferQueue().GetId();
 }
 
-Kernel::SharedPtr<Kernel::ReadableEvent> NVFlinger::GetVsyncEvent(u64 display_id) {
-    return FindDisplay(display_id).vsync_event.readable;
+Kernel::SharedPtr<Kernel::ReadableEvent> NVFlinger::FindVsyncEvent(u64 display_id) const {
+    auto* const display = FindDisplay(display_id);
+
+    if (display == nullptr) {
+        return nullptr;
+    }
+
+    return display->GetVSyncEvent();
 }
 
-std::shared_ptr<BufferQueue> NVFlinger::FindBufferQueue(u32 id) const {
+BufferQueue& NVFlinger::FindBufferQueue(u32 id) {
     const auto itr = std::find_if(buffer_queues.begin(), buffer_queues.end(),
-                                  [&](const auto& queue) { return queue->GetId() == id; });
+                                  [id](const auto& queue) { return queue.GetId() == id; });
 
     ASSERT(itr != buffer_queues.end());
     return *itr;
 }
 
-Display& NVFlinger::FindDisplay(u64 display_id) {
-    const auto itr = std::find_if(displays.begin(), displays.end(),
-                                  [&](const Display& display) { return display.id == display_id; });
+const BufferQueue& NVFlinger::FindBufferQueue(u32 id) const {
+    const auto itr = std::find_if(buffer_queues.begin(), buffer_queues.end(),
+                                  [id](const auto& queue) { return queue.GetId() == id; });
 
-    ASSERT(itr != displays.end());
+    ASSERT(itr != buffer_queues.end());
     return *itr;
 }
 
-const Display& NVFlinger::FindDisplay(u64 display_id) const {
-    const auto itr = std::find_if(displays.begin(), displays.end(),
-                                  [&](const Display& display) { return display.id == display_id; });
+VI::Display* NVFlinger::FindDisplay(u64 display_id) {
+    const auto itr =
+        std::find_if(displays.begin(), displays.end(),
+                     [&](const VI::Display& display) { return display.GetID() == display_id; });
 
-    ASSERT(itr != displays.end());
-    return *itr;
+    if (itr == displays.end()) {
+        return nullptr;
+    }
+
+    return &*itr;
 }
 
-Layer& NVFlinger::FindLayer(u64 display_id, u64 layer_id) {
-    auto& display = FindDisplay(display_id);
+const VI::Display* NVFlinger::FindDisplay(u64 display_id) const {
+    const auto itr =
+        std::find_if(displays.begin(), displays.end(),
+                     [&](const VI::Display& display) { return display.GetID() == display_id; });
 
-    const auto itr = std::find_if(display.layers.begin(), display.layers.end(),
-                                  [&](const Layer& layer) { return layer.id == layer_id; });
+    if (itr == displays.end()) {
+        return nullptr;
+    }
 
-    ASSERT(itr != display.layers.end());
-    return *itr;
+    return &*itr;
 }
 
-const Layer& NVFlinger::FindLayer(u64 display_id, u64 layer_id) const {
-    const auto& display = FindDisplay(display_id);
+VI::Layer* NVFlinger::FindLayer(u64 display_id, u64 layer_id) {
+    auto* const display = FindDisplay(display_id);
 
-    const auto itr = std::find_if(display.layers.begin(), display.layers.end(),
-                                  [&](const Layer& layer) { return layer.id == layer_id; });
+    if (display == nullptr) {
+        return nullptr;
+    }
 
-    ASSERT(itr != display.layers.end());
-    return *itr;
+    return display->FindLayer(layer_id);
+}
+
+const VI::Layer* NVFlinger::FindLayer(u64 display_id, u64 layer_id) const {
+    const auto* const display = FindDisplay(display_id);
+
+    if (display == nullptr) {
+        return nullptr;
+    }
+
+    return display->FindLayer(layer_id);
 }
 
 void NVFlinger::Compose() {
     for (auto& display : displays) {
         // Trigger vsync for this display at the end of drawing
-        SCOPE_EXIT({ display.vsync_event.writable->Signal(); });
+        SCOPE_EXIT({ display.SignalVSyncEvent(); });
 
         // Don't do anything for displays without layers.
-        if (display.layers.empty())
+        if (!display.HasLayers())
             continue;
 
         // TODO(Subv): Support more than 1 layer.
-        ASSERT_MSG(display.layers.size() == 1, "Max 1 layer per display is supported");
-
-        Layer& layer = display.layers[0];
-        auto& buffer_queue = layer.buffer_queue;
+        VI::Layer& layer = display.GetLayer(0);
+        auto& buffer_queue = layer.GetBufferQueue();
 
         // Search for a queued buffer and acquire it
-        auto buffer = buffer_queue->AcquireBuffer();
+        auto buffer = buffer_queue.AcquireBuffer();
 
         MicroProfileFlip();
 
@@ -151,7 +186,7 @@ void NVFlinger::Compose() {
 
             // There was no queued buffer to draw, render previous frame
             system_instance.GetPerfStats().EndGameFrame();
-            system_instance.Renderer().SwapBuffers({});
+            system_instance.GPU().SwapBuffers({});
             continue;
         }
 
@@ -167,19 +202,8 @@ void NVFlinger::Compose() {
                      igbp_buffer.width, igbp_buffer.height, igbp_buffer.stride,
                      buffer->get().transform, buffer->get().crop_rect);
 
-        buffer_queue->ReleaseBuffer(buffer->get().slot);
+        buffer_queue.ReleaseBuffer(buffer->get().slot);
     }
 }
 
-Layer::Layer(u64 id, std::shared_ptr<BufferQueue> queue) : id(id), buffer_queue(std::move(queue)) {}
-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::Sticky,
-                                                         fmt::format("Display VSync Event {}", id));
-}
-
-Display::~Display() = default;
-
 } // namespace Service::NVFlinger
diff --git a/src/core/hle/service/nvflinger/nvflinger.h b/src/core/hle/service/nvflinger/nvflinger.h
index 83e974ed3..c0a83fffb 100644
--- a/src/core/hle/service/nvflinger/nvflinger.h
+++ b/src/core/hle/service/nvflinger/nvflinger.h
@@ -4,8 +4,8 @@
 
 #pragma once
 
-#include <array>
 #include <memory>
+#include <optional>
 #include <string>
 #include <string_view>
 #include <vector>
@@ -13,9 +13,10 @@
 #include "common/common_types.h"
 #include "core/hle/kernel/object.h"
 
-namespace CoreTiming {
+namespace Core::Timing {
+class CoreTiming;
 struct EventType;
-}
+} // namespace Core::Timing
 
 namespace Kernel {
 class ReadableEvent;
@@ -24,53 +25,50 @@ class WritableEvent;
 
 namespace Service::Nvidia {
 class Module;
-}
+} // namespace Service::Nvidia
+
+namespace Service::VI {
+class Display;
+class Layer;
+} // namespace Service::VI
 
 namespace Service::NVFlinger {
 
 class BufferQueue;
 
-struct Layer {
-    Layer(u64 id, std::shared_ptr<BufferQueue> queue);
-    ~Layer();
-
-    u64 id;
-    std::shared_ptr<BufferQueue> buffer_queue;
-};
-
-struct Display {
-    Display(u64 id, std::string name);
-    ~Display();
-
-    u64 id;
-    std::string name;
-
-    std::vector<Layer> layers;
-    Kernel::EventPair vsync_event;
-};
-
 class NVFlinger final {
 public:
-    NVFlinger();
+    explicit NVFlinger(Core::Timing::CoreTiming& core_timing);
     ~NVFlinger();
 
     /// Sets the NVDrv module instance to use to send buffers to the GPU.
     void SetNVDrvInstance(std::shared_ptr<Nvidia::Module> instance);
 
     /// Opens the specified display and returns the ID.
-    u64 OpenDisplay(std::string_view name);
+    ///
+    /// If an invalid display name is provided, then an empty optional is returned.
+    std::optional<u64> OpenDisplay(std::string_view name);
 
     /// Creates a layer on the specified display and returns the layer ID.
-    u64 CreateLayer(u64 display_id);
+    ///
+    /// If an invalid display ID is specified, then an empty optional is returned.
+    std::optional<u64> CreateLayer(u64 display_id);
 
     /// Finds the buffer queue ID of the specified layer in the specified display.
-    u32 FindBufferQueueId(u64 display_id, u64 layer_id) const;
+    ///
+    /// If an invalid display ID or layer ID is provided, then an empty optional is returned.
+    std::optional<u32> FindBufferQueueId(u64 display_id, u64 layer_id) const;
 
     /// Gets the vsync event for the specified display.
-    Kernel::SharedPtr<Kernel::ReadableEvent> GetVsyncEvent(u64 display_id);
+    ///
+    /// If an invalid display ID is provided, then nullptr is returned.
+    Kernel::SharedPtr<Kernel::ReadableEvent> FindVsyncEvent(u64 display_id) const;
 
     /// Obtains a buffer queue identified by the ID.
-    std::shared_ptr<BufferQueue> FindBufferQueue(u32 id) const;
+    BufferQueue& FindBufferQueue(u32 id);
+
+    /// Obtains a buffer queue identified by the ID.
+    const BufferQueue& FindBufferQueue(u32 id) const;
 
     /// Performs a composition request to the emulated nvidia GPU and triggers the vsync events when
     /// finished.
@@ -78,27 +76,21 @@ public:
 
 private:
     /// Finds the display identified by the specified ID.
-    Display& FindDisplay(u64 display_id);
+    VI::Display* FindDisplay(u64 display_id);
 
     /// Finds the display identified by the specified ID.
-    const Display& FindDisplay(u64 display_id) const;
+    const VI::Display* FindDisplay(u64 display_id) const;
 
     /// Finds the layer identified by the specified ID in the desired display.
-    Layer& FindLayer(u64 display_id, u64 layer_id);
+    VI::Layer* FindLayer(u64 display_id, u64 layer_id);
 
     /// Finds the layer identified by the specified ID in the desired display.
-    const Layer& FindLayer(u64 display_id, u64 layer_id) const;
+    const VI::Layer* FindLayer(u64 display_id, u64 layer_id) const;
 
     std::shared_ptr<Nvidia::Module> nvdrv;
 
-    std::array<Display, 5> displays{{
-        {0, "Default"},
-        {1, "External"},
-        {2, "Edid"},
-        {3, "Internal"},
-        {4, "Null"},
-    }};
-    std::vector<std::shared_ptr<BufferQueue>> buffer_queues;
+    std::vector<VI::Display> displays;
+    std::vector<BufferQueue> buffer_queues;
 
     /// Id to use for the next layer that is created, this counter is shared among all displays.
     u64 next_layer_id = 1;
@@ -106,8 +98,11 @@ private:
     /// layers.
     u32 next_buffer_queue_id = 1;
 
-    /// CoreTiming event that handles screen composition.
-    CoreTiming::EventType* composition_event;
+    /// Event that handles screen composition.
+    Core::Timing::EventType* composition_event;
+
+    /// Core timing instance for registering/unregistering the composition event.
+    Core::Timing::CoreTiming& core_timing;
 };
 
 } // namespace Service::NVFlinger
diff --git a/src/core/hle/service/service.cpp b/src/core/hle/service/service.cpp
index d25b80ab0..00806b0ed 100644
--- a/src/core/hle/service/service.cpp
+++ b/src/core/hle/service/service.cpp
@@ -11,7 +11,6 @@
 #include "core/hle/ipc.h"
 #include "core/hle/ipc_helpers.h"
 #include "core/hle/kernel/client_port.h"
-#include "core/hle/kernel/handle_table.h"
 #include "core/hle/kernel/kernel.h"
 #include "core/hle/kernel/process.h"
 #include "core/hle/kernel/server_port.h"
@@ -76,7 +75,8 @@ namespace Service {
  * Creates a function string for logging, complete with the name (or header code, depending
  * on what's passed in) the port name, and all the cmd_buff arguments.
  */
-[[maybe_unused]] static std::string MakeFunctionString(const char* name, const char* port_name,
+[[maybe_unused]] static std::string MakeFunctionString(std::string_view name,
+                                                       std::string_view port_name,
                                                        const u32* cmd_buff) {
     // Number of params == bits 0-5 + bits 6-11
     int num_params = (cmd_buff[0] & 0x3F) + ((cmd_buff[0] >> 6) & 0x3F);
@@ -158,9 +158,7 @@ void ServiceFrameworkBase::InvokeRequest(Kernel::HLERequestContext& ctx) {
         return ReportUnimplementedFunction(ctx, info);
     }
 
-    LOG_TRACE(
-        Service, "{}",
-        MakeFunctionString(info->name, GetServiceName().c_str(), ctx.CommandBuffer()).c_str());
+    LOG_TRACE(Service, "{}", MakeFunctionString(info->name, GetServiceName(), ctx.CommandBuffer()));
     handler_invoker(this, info->handler_callback, ctx);
 }
 
@@ -169,7 +167,7 @@ ResultCode ServiceFrameworkBase::HandleSyncRequest(Kernel::HLERequestContext& co
     case IPC::CommandType::Close: {
         IPC::ResponseBuilder rb{context, 2};
         rb.Push(RESULT_SUCCESS);
-        return ResultCode(ErrorModule::HIPC, ErrorDescription::RemoteProcessDead);
+        return IPC::ERR_REMOTE_PROCESS_DEAD;
     }
     case IPC::CommandType::ControlWithContext:
     case IPC::CommandType::Control: {
@@ -194,10 +192,11 @@ ResultCode ServiceFrameworkBase::HandleSyncRequest(Kernel::HLERequestContext& co
 // Module interface
 
 /// Initialize ServiceManager
-void Init(std::shared_ptr<SM::ServiceManager>& sm, FileSys::VfsFilesystem& vfs) {
+void Init(std::shared_ptr<SM::ServiceManager>& sm, Core::System& system,
+          FileSys::VfsFilesystem& vfs) {
     // NVFlinger needs to be accessed by several services like Vi and AppletOE so we instantiate it
     // here and pass it into the respective InstallInterfaces functions.
-    auto nv_flinger = std::make_shared<NVFlinger::NVFlinger>();
+    auto nv_flinger = std::make_shared<NVFlinger::NVFlinger>(system.CoreTiming());
 
     SM::ServiceManager::InstallInterfaces(sm);
 
diff --git a/src/core/hle/service/service.h b/src/core/hle/service/service.h
index 029533628..830790269 100644
--- a/src/core/hle/service/service.h
+++ b/src/core/hle/service/service.h
@@ -14,6 +14,14 @@
 ////////////////////////////////////////////////////////////////////////////////////////////////////
 // Namespace Service
 
+namespace Core {
+class System;
+}
+
+namespace FileSys {
+class VfsFilesystem;
+}
+
 namespace Kernel {
 class ClientPort;
 class ServerPort;
@@ -21,10 +29,6 @@ class ServerSession;
 class HLERequestContext;
 } // namespace Kernel
 
-namespace FileSys {
-class VfsFilesystem;
-}
-
 namespace Service {
 
 namespace SM {
@@ -178,7 +182,8 @@ private:
 };
 
 /// Initialize ServiceManager
-void Init(std::shared_ptr<SM::ServiceManager>& sm, FileSys::VfsFilesystem& vfs);
+void Init(std::shared_ptr<SM::ServiceManager>& sm, Core::System& system,
+          FileSys::VfsFilesystem& vfs);
 
 /// Shutdown ServiceManager
 void Shutdown();
diff --git a/src/core/hle/service/sm/controller.cpp b/src/core/hle/service/sm/controller.cpp
index 74da4d5e6..e9ee73710 100644
--- a/src/core/hle/service/sm/controller.cpp
+++ b/src/core/hle/service/sm/controller.cpp
@@ -30,7 +30,7 @@ void Controller::DuplicateSession(Kernel::HLERequestContext& ctx) {
 
     IPC::ResponseBuilder rb{ctx, 2, 0, 1, IPC::ResponseBuilder::Flags::AlwaysMoveHandles};
     rb.Push(RESULT_SUCCESS);
-    Kernel::SharedPtr<Kernel::ClientSession> session{ctx.Session()->parent->client};
+    Kernel::SharedPtr<Kernel::ClientSession> session{ctx.Session()->GetParent()->client};
     rb.PushMoveObjects(session);
 
     LOG_DEBUG(Service, "session={}", session->GetObjectId());
diff --git a/src/core/hle/service/sm/sm.h b/src/core/hle/service/sm/sm.h
index bef25433e..b9d6381b4 100644
--- a/src/core/hle/service/sm/sm.h
+++ b/src/core/hle/service/sm/sm.h
@@ -67,7 +67,7 @@ public:
         if (port == nullptr) {
             return nullptr;
         }
-        return std::static_pointer_cast<T>(port->hle_handler);
+        return std::static_pointer_cast<T>(port->GetHLEHandler());
     }
 
     void InvokeControlRequest(Kernel::HLERequestContext& context);
diff --git a/src/core/hle/service/time/time.cpp b/src/core/hle/service/time/time.cpp
index c13640ad8..aa115935d 100644
--- a/src/core/hle/service/time/time.cpp
+++ b/src/core/hle/service/time/time.cpp
@@ -5,6 +5,7 @@
 #include <chrono>
 #include <ctime>
 #include "common/logging/log.h"
+#include "core/core.h"
 #include "core/core_timing.h"
 #include "core/core_timing_util.h"
 #include "core/hle/ipc_helpers.h"
@@ -106,8 +107,9 @@ private:
     void GetCurrentTimePoint(Kernel::HLERequestContext& ctx) {
         LOG_DEBUG(Service_Time, "called");
 
-        SteadyClockTimePoint steady_clock_time_point{
-            CoreTiming::cyclesToMs(CoreTiming::GetTicks()) / 1000};
+        const auto& core_timing = Core::System::GetInstance().CoreTiming();
+        const SteadyClockTimePoint steady_clock_time_point{
+            Core::Timing::cyclesToMs(core_timing.GetTicks()) / 1000};
         IPC::ResponseBuilder rb{ctx, (sizeof(SteadyClockTimePoint) / 4) + 2};
         rb.Push(RESULT_SUCCESS);
         rb.PushRaw(steady_clock_time_point);
@@ -281,8 +283,9 @@ void Module::Interface::GetClockSnapshot(Kernel::HLERequestContext& ctx) {
         return;
     }
 
+    const auto& core_timing = Core::System::GetInstance().CoreTiming();
     const SteadyClockTimePoint steady_clock_time_point{
-        CoreTiming::cyclesToMs(CoreTiming::GetTicks()) / 1000, {}};
+        Core::Timing::cyclesToMs(core_timing.GetTicks()) / 1000, {}};
 
     CalendarTime calendar_time{};
     calendar_time.year = tm->tm_year + 1900;
diff --git a/src/core/hle/service/vi/display/vi_display.cpp b/src/core/hle/service/vi/display/vi_display.cpp
new file mode 100644
index 000000000..01d80311b
--- /dev/null
+++ b/src/core/hle/service/vi/display/vi_display.cpp
@@ -0,0 +1,71 @@
+// Copyright 2019 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <algorithm>
+#include <utility>
+
+#include <fmt/format.h>
+
+#include "common/assert.h"
+#include "core/core.h"
+#include "core/hle/kernel/readable_event.h"
+#include "core/hle/service/vi/display/vi_display.h"
+#include "core/hle/service/vi/layer/vi_layer.h"
+
+namespace Service::VI {
+
+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::Sticky,
+                                                         fmt::format("Display VSync Event {}", id));
+}
+
+Display::~Display() = default;
+
+Layer& Display::GetLayer(std::size_t index) {
+    return layers.at(index);
+}
+
+const Layer& Display::GetLayer(std::size_t index) const {
+    return layers.at(index);
+}
+
+Kernel::SharedPtr<Kernel::ReadableEvent> Display::GetVSyncEvent() const {
+    return vsync_event.readable;
+}
+
+void Display::SignalVSyncEvent() {
+    vsync_event.writable->Signal();
+}
+
+void Display::CreateLayer(u64 id, NVFlinger::BufferQueue& buffer_queue) {
+    // TODO(Subv): Support more than 1 layer.
+    ASSERT_MSG(layers.empty(), "Only one layer is supported per display at the moment");
+
+    layers.emplace_back(id, buffer_queue);
+}
+
+Layer* Display::FindLayer(u64 id) {
+    const auto itr = std::find_if(layers.begin(), layers.end(),
+                                  [id](const VI::Layer& layer) { return layer.GetID() == id; });
+
+    if (itr == layers.end()) {
+        return nullptr;
+    }
+
+    return &*itr;
+}
+
+const Layer* Display::FindLayer(u64 id) const {
+    const auto itr = std::find_if(layers.begin(), layers.end(),
+                                  [id](const VI::Layer& layer) { return layer.GetID() == id; });
+
+    if (itr == layers.end()) {
+        return nullptr;
+    }
+
+    return &*itr;
+}
+
+} // namespace Service::VI
diff --git a/src/core/hle/service/vi/display/vi_display.h b/src/core/hle/service/vi/display/vi_display.h
new file mode 100644
index 000000000..2acd46ff8
--- /dev/null
+++ b/src/core/hle/service/vi/display/vi_display.h
@@ -0,0 +1,98 @@
+// Copyright 2019 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <string>
+#include <vector>
+
+#include "common/common_types.h"
+#include "core/hle/kernel/writable_event.h"
+
+namespace Service::NVFlinger {
+class BufferQueue;
+}
+
+namespace Service::VI {
+
+class Layer;
+
+/// Represents a single display type
+class Display {
+public:
+    /// Constructs a display with a given unique ID and name.
+    ///
+    /// @param id   The unique ID for this display.
+    /// @param name The name for this display.
+    ///
+    Display(u64 id, std::string name);
+    ~Display();
+
+    Display(const Display&) = delete;
+    Display& operator=(const Display&) = delete;
+
+    Display(Display&&) = default;
+    Display& operator=(Display&&) = default;
+
+    /// Gets the unique ID assigned to this display.
+    u64 GetID() const {
+        return id;
+    }
+
+    /// Gets the name of this display
+    const std::string& GetName() const {
+        return name;
+    }
+
+    /// Whether or not this display has any layers added to it.
+    bool HasLayers() const {
+        return !layers.empty();
+    }
+
+    /// Gets a layer for this display based off an index.
+    Layer& GetLayer(std::size_t index);
+
+    /// Gets a layer for this display based off an index.
+    const Layer& GetLayer(std::size_t index) const;
+
+    /// Gets the readable vsync event.
+    Kernel::SharedPtr<Kernel::ReadableEvent> GetVSyncEvent() const;
+
+    /// Signals the internal vsync event.
+    void SignalVSyncEvent();
+
+    /// Creates and adds a layer to this display with the given ID.
+    ///
+    /// @param id           The ID to assign to the created layer.
+    /// @param buffer_queue The buffer queue for the layer instance to use.
+    ///
+    void CreateLayer(u64 id, NVFlinger::BufferQueue& buffer_queue);
+
+    /// Attempts to find a layer with the given ID.
+    ///
+    /// @param id The layer ID.
+    ///
+    /// @returns If found, the Layer instance with the given ID.
+    ///          If not found, then nullptr is returned.
+    ///
+    Layer* FindLayer(u64 id);
+
+    /// Attempts to find a layer with the given ID.
+    ///
+    /// @param id The layer ID.
+    ///
+    /// @returns If found, the Layer instance with the given ID.
+    ///          If not found, then nullptr is returned.
+    ///
+    const Layer* FindLayer(u64 id) const;
+
+private:
+    u64 id;
+    std::string name;
+
+    std::vector<Layer> layers;
+    Kernel::EventPair vsync_event;
+};
+
+} // namespace Service::VI
diff --git a/src/core/hle/service/vi/layer/vi_layer.cpp b/src/core/hle/service/vi/layer/vi_layer.cpp
new file mode 100644
index 000000000..954225c26
--- /dev/null
+++ b/src/core/hle/service/vi/layer/vi_layer.cpp
@@ -0,0 +1,13 @@
+// Copyright 2019 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "core/hle/service/vi/layer/vi_layer.h"
+
+namespace Service::VI {
+
+Layer::Layer(u64 id, NVFlinger::BufferQueue& queue) : id{id}, buffer_queue{queue} {}
+
+Layer::~Layer() = default;
+
+} // namespace Service::VI
diff --git a/src/core/hle/service/vi/layer/vi_layer.h b/src/core/hle/service/vi/layer/vi_layer.h
new file mode 100644
index 000000000..c6bfd01f6
--- /dev/null
+++ b/src/core/hle/service/vi/layer/vi_layer.h
@@ -0,0 +1,52 @@
+// Copyright 2019 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "common/common_types.h"
+
+namespace Service::NVFlinger {
+class BufferQueue;
+}
+
+namespace Service::VI {
+
+/// Represents a single display layer.
+class Layer {
+public:
+    /// Constructs a layer with a given ID and buffer queue.
+    ///
+    /// @param id    The ID to assign to this layer.
+    /// @param queue The buffer queue for this layer to use.
+    ///
+    Layer(u64 id, NVFlinger::BufferQueue& queue);
+    ~Layer();
+
+    Layer(const Layer&) = delete;
+    Layer& operator=(const Layer&) = delete;
+
+    Layer(Layer&&) = default;
+    Layer& operator=(Layer&&) = delete;
+
+    /// Gets the ID for this layer.
+    u64 GetID() const {
+        return id;
+    }
+
+    /// Gets a reference to the buffer queue this layer is using.
+    NVFlinger::BufferQueue& GetBufferQueue() {
+        return buffer_queue;
+    }
+
+    /// Gets a const reference to the buffer queue this layer is using.
+    const NVFlinger::BufferQueue& GetBufferQueue() const {
+        return buffer_queue;
+    }
+
+private:
+    u64 id;
+    NVFlinger::BufferQueue& buffer_queue;
+};
+
+} // namespace Service::VI
diff --git a/src/core/hle/service/vi/vi.cpp b/src/core/hle/service/vi/vi.cpp
index fe08c38f2..566cd6006 100644
--- a/src/core/hle/service/vi/vi.cpp
+++ b/src/core/hle/service/vi/vi.cpp
@@ -24,6 +24,7 @@
 #include "core/hle/service/nvdrv/nvdrv.h"
 #include "core/hle/service/nvflinger/buffer_queue.h"
 #include "core/hle/service/nvflinger/nvflinger.h"
+#include "core/hle/service/service.h"
 #include "core/hle/service/vi/vi.h"
 #include "core/hle/service/vi/vi_m.h"
 #include "core/hle/service/vi/vi_s.h"
@@ -33,7 +34,9 @@
 namespace Service::VI {
 
 constexpr ResultCode ERR_OPERATION_FAILED{ErrorModule::VI, 1};
+constexpr ResultCode ERR_PERMISSION_DENIED{ErrorModule::VI, 5};
 constexpr ResultCode ERR_UNSUPPORTED{ErrorModule::VI, 6};
+constexpr ResultCode ERR_NOT_FOUND{ErrorModule::VI, 7};
 
 struct DisplayInfo {
     /// The name of this particular display.
@@ -419,7 +422,7 @@ public:
         u32_le fence_is_valid;
         std::array<Fence, 2> fences;
 
-        MathUtil::Rectangle<int> GetCropRect() const {
+        Common::Rectangle<int> GetCropRect() const {
             return {crop_left, crop_top, crop_right, crop_bottom};
         }
     };
@@ -524,7 +527,7 @@ private:
         LOG_DEBUG(Service_VI, "called. id=0x{:08X} transaction={:X}, flags=0x{:08X}", id,
                   static_cast<u32>(transaction), flags);
 
-        auto buffer_queue = nv_flinger->FindBufferQueue(id);
+        auto& buffer_queue = nv_flinger->FindBufferQueue(id);
 
         if (transaction == TransactionId::Connect) {
             IGBPConnectRequestParcel request{ctx.ReadBuffer()};
@@ -537,7 +540,7 @@ private:
         } else if (transaction == TransactionId::SetPreallocatedBuffer) {
             IGBPSetPreallocatedBufferRequestParcel request{ctx.ReadBuffer()};
 
-            buffer_queue->SetPreallocatedBuffer(request.data.slot, request.buffer);
+            buffer_queue.SetPreallocatedBuffer(request.data.slot, request.buffer);
 
             IGBPSetPreallocatedBufferResponseParcel response{};
             ctx.WriteBuffer(response.Serialize());
@@ -545,7 +548,7 @@ private:
             IGBPDequeueBufferRequestParcel request{ctx.ReadBuffer()};
             const u32 width{request.data.width};
             const u32 height{request.data.height};
-            std::optional<u32> slot = buffer_queue->DequeueBuffer(width, height);
+            std::optional<u32> slot = buffer_queue.DequeueBuffer(width, height);
 
             if (slot) {
                 // Buffer is available
@@ -558,8 +561,8 @@ private:
                     [=](Kernel::SharedPtr<Kernel::Thread> thread, Kernel::HLERequestContext& ctx,
                         Kernel::ThreadWakeupReason reason) {
                         // Repeat TransactParcel DequeueBuffer when a buffer is available
-                        auto buffer_queue = nv_flinger->FindBufferQueue(id);
-                        std::optional<u32> slot = buffer_queue->DequeueBuffer(width, height);
+                        auto& buffer_queue = nv_flinger->FindBufferQueue(id);
+                        std::optional<u32> slot = buffer_queue.DequeueBuffer(width, height);
                         ASSERT_MSG(slot != std::nullopt, "Could not dequeue buffer.");
 
                         IGBPDequeueBufferResponseParcel response{*slot};
@@ -567,28 +570,28 @@ private:
                         IPC::ResponseBuilder rb{ctx, 2};
                         rb.Push(RESULT_SUCCESS);
                     },
-                    buffer_queue->GetWritableBufferWaitEvent());
+                    buffer_queue.GetWritableBufferWaitEvent());
             }
         } else if (transaction == TransactionId::RequestBuffer) {
             IGBPRequestBufferRequestParcel request{ctx.ReadBuffer()};
 
-            auto& buffer = buffer_queue->RequestBuffer(request.slot);
+            auto& buffer = buffer_queue.RequestBuffer(request.slot);
 
             IGBPRequestBufferResponseParcel response{buffer};
             ctx.WriteBuffer(response.Serialize());
         } else if (transaction == TransactionId::QueueBuffer) {
             IGBPQueueBufferRequestParcel request{ctx.ReadBuffer()};
 
-            buffer_queue->QueueBuffer(request.data.slot, request.data.transform,
-                                      request.data.GetCropRect());
+            buffer_queue.QueueBuffer(request.data.slot, request.data.transform,
+                                     request.data.GetCropRect());
 
             IGBPQueueBufferResponseParcel response{1280, 720};
             ctx.WriteBuffer(response.Serialize());
         } else if (transaction == TransactionId::Query) {
             IGBPQueryRequestParcel request{ctx.ReadBuffer()};
 
-            u32 value =
-                buffer_queue->Query(static_cast<NVFlinger::BufferQueue::QueryType>(request.type));
+            const u32 value =
+                buffer_queue.Query(static_cast<NVFlinger::BufferQueue::QueryType>(request.type));
 
             IGBPQueryResponseParcel response{value};
             ctx.WriteBuffer(response.Serialize());
@@ -628,12 +631,12 @@ private:
 
         LOG_WARNING(Service_VI, "(STUBBED) called id={}, unknown={:08X}", id, unknown);
 
-        const auto buffer_queue = nv_flinger->FindBufferQueue(id);
+        const auto& buffer_queue = nv_flinger->FindBufferQueue(id);
 
         // TODO(Subv): Find out what this actually is.
         IPC::ResponseBuilder rb{ctx, 2, 1};
         rb.Push(RESULT_SUCCESS);
-        rb.PushCopyObjects(buffer_queue->GetBufferWaitEvent());
+        rb.PushCopyObjects(buffer_queue.GetBufferWaitEvent());
     }
 
     std::shared_ptr<NVFlinger::NVFlinger> nv_flinger;
@@ -751,6 +754,7 @@ public:
             {1102, nullptr, "GetDisplayResolution"},
             {2010, &IManagerDisplayService::CreateManagedLayer, "CreateManagedLayer"},
             {2011, nullptr, "DestroyManagedLayer"},
+            {2012, nullptr, "CreateStrayLayer"},
             {2050, nullptr, "CreateIndirectLayer"},
             {2051, nullptr, "DestroyIndirectLayer"},
             {2052, nullptr, "CreateIndirectProducerEndPoint"},
@@ -838,11 +842,16 @@ private:
                     "(STUBBED) called. unknown=0x{:08X}, display=0x{:016X}, aruid=0x{:016X}",
                     unknown, display, aruid);
 
-        const u64 layer_id = nv_flinger->CreateLayer(display);
+        const auto layer_id = nv_flinger->CreateLayer(display);
+        if (!layer_id) {
+            IPC::ResponseBuilder rb{ctx, 2};
+            rb.Push(ERR_NOT_FOUND);
+            return;
+        }
 
         IPC::ResponseBuilder rb{ctx, 4};
         rb.Push(RESULT_SUCCESS);
-        rb.Push(layer_id);
+        rb.Push(*layer_id);
     }
 
     void AddToLayerStack(Kernel::HLERequestContext& ctx) {
@@ -950,9 +959,16 @@ private:
 
         ASSERT_MSG(name == "Default", "Non-default displays aren't supported yet");
 
+        const auto display_id = nv_flinger->OpenDisplay(name);
+        if (!display_id) {
+            IPC::ResponseBuilder rb{ctx, 2};
+            rb.Push(ERR_NOT_FOUND);
+            return;
+        }
+
         IPC::ResponseBuilder rb{ctx, 4};
         rb.Push(RESULT_SUCCESS);
-        rb.Push<u64>(nv_flinger->OpenDisplay(name));
+        rb.Push<u64>(*display_id);
     }
 
     void CloseDisplay(Kernel::HLERequestContext& ctx) {
@@ -1043,10 +1059,21 @@ private:
 
         LOG_DEBUG(Service_VI, "called. layer_id=0x{:016X}, aruid=0x{:016X}", layer_id, aruid);
 
-        const u64 display_id = nv_flinger->OpenDisplay(display_name);
-        const u32 buffer_queue_id = nv_flinger->FindBufferQueueId(display_id, layer_id);
+        const auto display_id = nv_flinger->OpenDisplay(display_name);
+        if (!display_id) {
+            IPC::ResponseBuilder rb{ctx, 2};
+            rb.Push(ERR_NOT_FOUND);
+            return;
+        }
+
+        const auto buffer_queue_id = nv_flinger->FindBufferQueueId(*display_id, layer_id);
+        if (!buffer_queue_id) {
+            IPC::ResponseBuilder rb{ctx, 2};
+            rb.Push(ERR_NOT_FOUND);
+            return;
+        }
 
-        NativeWindow native_window{buffer_queue_id};
+        NativeWindow native_window{*buffer_queue_id};
         IPC::ResponseBuilder rb{ctx, 4};
         rb.Push(RESULT_SUCCESS);
         rb.Push<u64>(ctx.WriteBuffer(native_window.Serialize()));
@@ -1062,13 +1089,24 @@ private:
 
         // TODO(Subv): What's the difference between a Stray and a Managed layer?
 
-        const u64 layer_id = nv_flinger->CreateLayer(display_id);
-        const u32 buffer_queue_id = nv_flinger->FindBufferQueueId(display_id, layer_id);
+        const auto layer_id = nv_flinger->CreateLayer(display_id);
+        if (!layer_id) {
+            IPC::ResponseBuilder rb{ctx, 2};
+            rb.Push(ERR_NOT_FOUND);
+            return;
+        }
+
+        const auto buffer_queue_id = nv_flinger->FindBufferQueueId(display_id, *layer_id);
+        if (!buffer_queue_id) {
+            IPC::ResponseBuilder rb{ctx, 2};
+            rb.Push(ERR_NOT_FOUND);
+            return;
+        }
 
-        NativeWindow native_window{buffer_queue_id};
+        NativeWindow native_window{*buffer_queue_id};
         IPC::ResponseBuilder rb{ctx, 6};
         rb.Push(RESULT_SUCCESS);
-        rb.Push(layer_id);
+        rb.Push(*layer_id);
         rb.Push<u64>(ctx.WriteBuffer(native_window.Serialize()));
     }
 
@@ -1088,7 +1126,12 @@ private:
 
         LOG_WARNING(Service_VI, "(STUBBED) called. display_id=0x{:016X}", display_id);
 
-        const auto vsync_event = nv_flinger->GetVsyncEvent(display_id);
+        const auto vsync_event = nv_flinger->FindVsyncEvent(display_id);
+        if (!vsync_event) {
+            IPC::ResponseBuilder rb{ctx, 2};
+            rb.Push(ERR_NOT_FOUND);
+            return;
+        }
 
         IPC::ResponseBuilder rb{ctx, 2, 1};
         rb.Push(RESULT_SUCCESS);
@@ -1162,26 +1205,40 @@ IApplicationDisplayService::IApplicationDisplayService(
     RegisterHandlers(functions);
 }
 
-Module::Interface::Interface(std::shared_ptr<Module> module, const char* name,
-                             std::shared_ptr<NVFlinger::NVFlinger> nv_flinger)
-    : ServiceFramework(name), module(std::move(module)), nv_flinger(std::move(nv_flinger)) {}
+static bool IsValidServiceAccess(Permission permission, Policy policy) {
+    if (permission == Permission::User) {
+        return policy == Policy::User;
+    }
+
+    if (permission == Permission::System || permission == Permission::Manager) {
+        return policy == Policy::User || policy == Policy::Compositor;
+    }
 
-Module::Interface::~Interface() = default;
+    return false;
+}
 
-void Module::Interface::GetDisplayService(Kernel::HLERequestContext& ctx) {
-    LOG_WARNING(Service_VI, "(STUBBED) called");
+void detail::GetDisplayServiceImpl(Kernel::HLERequestContext& ctx,
+                                   std::shared_ptr<NVFlinger::NVFlinger> nv_flinger,
+                                   Permission permission) {
+    IPC::RequestParser rp{ctx};
+    const auto policy = rp.PopEnum<Policy>();
+
+    if (!IsValidServiceAccess(permission, policy)) {
+        IPC::ResponseBuilder rb{ctx, 2};
+        rb.Push(ERR_PERMISSION_DENIED);
+        return;
+    }
 
     IPC::ResponseBuilder rb{ctx, 2, 0, 1};
     rb.Push(RESULT_SUCCESS);
-    rb.PushIpcInterface<IApplicationDisplayService>(nv_flinger);
+    rb.PushIpcInterface<IApplicationDisplayService>(std::move(nv_flinger));
 }
 
 void InstallInterfaces(SM::ServiceManager& service_manager,
                        std::shared_ptr<NVFlinger::NVFlinger> nv_flinger) {
-    auto module = std::make_shared<Module>();
-    std::make_shared<VI_M>(module, nv_flinger)->InstallAsService(service_manager);
-    std::make_shared<VI_S>(module, nv_flinger)->InstallAsService(service_manager);
-    std::make_shared<VI_U>(module, nv_flinger)->InstallAsService(service_manager);
+    std::make_shared<VI_M>(nv_flinger)->InstallAsService(service_manager);
+    std::make_shared<VI_S>(nv_flinger)->InstallAsService(service_manager);
+    std::make_shared<VI_U>(nv_flinger)->InstallAsService(service_manager);
 }
 
 } // namespace Service::VI
diff --git a/src/core/hle/service/vi/vi.h b/src/core/hle/service/vi/vi.h
index e3963502a..6b66f8b81 100644
--- a/src/core/hle/service/vi/vi.h
+++ b/src/core/hle/service/vi/vi.h
@@ -4,12 +4,21 @@
 
 #pragma once
 
-#include "core/hle/service/service.h"
+#include <memory>
+#include "common/common_types.h"
+
+namespace Kernel {
+class HLERequestContext;
+}
 
 namespace Service::NVFlinger {
 class NVFlinger;
 }
 
+namespace Service::SM {
+class ServiceManager;
+}
+
 namespace Service::VI {
 
 enum class DisplayResolution : u32 {
@@ -19,22 +28,25 @@ enum class DisplayResolution : u32 {
     UndockedHeight = 720,
 };
 
-class Module final {
-public:
-    class Interface : public ServiceFramework<Interface> {
-    public:
-        explicit Interface(std::shared_ptr<Module> module, const char* name,
-                           std::shared_ptr<NVFlinger::NVFlinger> nv_flinger);
-        ~Interface() override;
-
-        void GetDisplayService(Kernel::HLERequestContext& ctx);
+/// Permission level for a particular VI service instance
+enum class Permission {
+    User,
+    System,
+    Manager,
+};
 
-    protected:
-        std::shared_ptr<Module> module;
-        std::shared_ptr<NVFlinger::NVFlinger> nv_flinger;
-    };
+/// A policy type that may be requested via GetDisplayService and
+/// GetDisplayServiceWithProxyNameExchange
+enum class Policy {
+    User,
+    Compositor,
 };
 
+namespace detail {
+void GetDisplayServiceImpl(Kernel::HLERequestContext& ctx,
+                           std::shared_ptr<NVFlinger::NVFlinger> nv_flinger, Permission permission);
+} // namespace detail
+
 /// Registers all VI services with the specified service manager.
 void InstallInterfaces(SM::ServiceManager& service_manager,
                        std::shared_ptr<NVFlinger::NVFlinger> nv_flinger);
diff --git a/src/core/hle/service/vi/vi_m.cpp b/src/core/hle/service/vi/vi_m.cpp
index 207c06b16..06070087f 100644
--- a/src/core/hle/service/vi/vi_m.cpp
+++ b/src/core/hle/service/vi/vi_m.cpp
@@ -2,12 +2,14 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#include "common/logging/log.h"
+#include "core/hle/service/vi/vi.h"
 #include "core/hle/service/vi/vi_m.h"
 
 namespace Service::VI {
 
-VI_M::VI_M(std::shared_ptr<Module> module, std::shared_ptr<NVFlinger::NVFlinger> nv_flinger)
-    : Module::Interface(std::move(module), "vi:m", std::move(nv_flinger)) {
+VI_M::VI_M(std::shared_ptr<NVFlinger::NVFlinger> nv_flinger)
+    : ServiceFramework{"vi:m"}, nv_flinger{std::move(nv_flinger)} {
     static const FunctionInfo functions[] = {
         {2, &VI_M::GetDisplayService, "GetDisplayService"},
         {3, nullptr, "GetDisplayServiceWithProxyNameExchange"},
@@ -17,4 +19,10 @@ VI_M::VI_M(std::shared_ptr<Module> module, std::shared_ptr<NVFlinger::NVFlinger>
 
 VI_M::~VI_M() = default;
 
+void VI_M::GetDisplayService(Kernel::HLERequestContext& ctx) {
+    LOG_DEBUG(Service_VI, "called");
+
+    detail::GetDisplayServiceImpl(ctx, nv_flinger, Permission::Manager);
+}
+
 } // namespace Service::VI
diff --git a/src/core/hle/service/vi/vi_m.h b/src/core/hle/service/vi/vi_m.h
index 487d58d50..290e06689 100644
--- a/src/core/hle/service/vi/vi_m.h
+++ b/src/core/hle/service/vi/vi_m.h
@@ -4,14 +4,27 @@
 
 #pragma once
 
-#include "core/hle/service/vi/vi.h"
+#include "core/hle/service/service.h"
+
+namespace Kernel {
+class HLERequestContext;
+}
+
+namespace Service::NVFlinger {
+class NVFlinger;
+}
 
 namespace Service::VI {
 
-class VI_M final : public Module::Interface {
+class VI_M final : public ServiceFramework<VI_M> {
 public:
-    explicit VI_M(std::shared_ptr<Module> module, std::shared_ptr<NVFlinger::NVFlinger> nv_flinger);
+    explicit VI_M(std::shared_ptr<NVFlinger::NVFlinger> nv_flinger);
     ~VI_M() override;
+
+private:
+    void GetDisplayService(Kernel::HLERequestContext& ctx);
+
+    std::shared_ptr<NVFlinger::NVFlinger> nv_flinger;
 };
 
 } // namespace Service::VI
diff --git a/src/core/hle/service/vi/vi_s.cpp b/src/core/hle/service/vi/vi_s.cpp
index 920e6a1f6..57c596cc4 100644
--- a/src/core/hle/service/vi/vi_s.cpp
+++ b/src/core/hle/service/vi/vi_s.cpp
@@ -2,12 +2,14 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#include "common/logging/log.h"
+#include "core/hle/service/vi/vi.h"
 #include "core/hle/service/vi/vi_s.h"
 
 namespace Service::VI {
 
-VI_S::VI_S(std::shared_ptr<Module> module, std::shared_ptr<NVFlinger::NVFlinger> nv_flinger)
-    : Module::Interface(std::move(module), "vi:s", std::move(nv_flinger)) {
+VI_S::VI_S(std::shared_ptr<NVFlinger::NVFlinger> nv_flinger)
+    : ServiceFramework{"vi:s"}, nv_flinger{std::move(nv_flinger)} {
     static const FunctionInfo functions[] = {
         {1, &VI_S::GetDisplayService, "GetDisplayService"},
         {3, nullptr, "GetDisplayServiceWithProxyNameExchange"},
@@ -17,4 +19,10 @@ VI_S::VI_S(std::shared_ptr<Module> module, std::shared_ptr<NVFlinger::NVFlinger>
 
 VI_S::~VI_S() = default;
 
+void VI_S::GetDisplayService(Kernel::HLERequestContext& ctx) {
+    LOG_DEBUG(Service_VI, "called");
+
+    detail::GetDisplayServiceImpl(ctx, nv_flinger, Permission::System);
+}
+
 } // namespace Service::VI
diff --git a/src/core/hle/service/vi/vi_s.h b/src/core/hle/service/vi/vi_s.h
index bbc31148f..47804dc0b 100644
--- a/src/core/hle/service/vi/vi_s.h
+++ b/src/core/hle/service/vi/vi_s.h
@@ -4,14 +4,27 @@
 
 #pragma once
 
-#include "core/hle/service/vi/vi.h"
+#include "core/hle/service/service.h"
+
+namespace Kernel {
+class HLERequestContext;
+}
+
+namespace Service::NVFlinger {
+class NVFlinger;
+}
 
 namespace Service::VI {
 
-class VI_S final : public Module::Interface {
+class VI_S final : public ServiceFramework<VI_S> {
 public:
-    explicit VI_S(std::shared_ptr<Module> module, std::shared_ptr<NVFlinger::NVFlinger> nv_flinger);
+    explicit VI_S(std::shared_ptr<NVFlinger::NVFlinger> nv_flinger);
     ~VI_S() override;
+
+private:
+    void GetDisplayService(Kernel::HLERequestContext& ctx);
+
+    std::shared_ptr<NVFlinger::NVFlinger> nv_flinger;
 };
 
 } // namespace Service::VI
diff --git a/src/core/hle/service/vi/vi_u.cpp b/src/core/hle/service/vi/vi_u.cpp
index d81e410d6..9d5ceb608 100644
--- a/src/core/hle/service/vi/vi_u.cpp
+++ b/src/core/hle/service/vi/vi_u.cpp
@@ -2,12 +2,14 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#include "common/logging/log.h"
+#include "core/hle/service/vi/vi.h"
 #include "core/hle/service/vi/vi_u.h"
 
 namespace Service::VI {
 
-VI_U::VI_U(std::shared_ptr<Module> module, std::shared_ptr<NVFlinger::NVFlinger> nv_flinger)
-    : Module::Interface(std::move(module), "vi:u", std::move(nv_flinger)) {
+VI_U::VI_U(std::shared_ptr<NVFlinger::NVFlinger> nv_flinger)
+    : ServiceFramework{"vi:u"}, nv_flinger{std::move(nv_flinger)} {
     static const FunctionInfo functions[] = {
         {0, &VI_U::GetDisplayService, "GetDisplayService"},
     };
@@ -16,4 +18,10 @@ VI_U::VI_U(std::shared_ptr<Module> module, std::shared_ptr<NVFlinger::NVFlinger>
 
 VI_U::~VI_U() = default;
 
+void VI_U::GetDisplayService(Kernel::HLERequestContext& ctx) {
+    LOG_DEBUG(Service_VI, "called");
+
+    detail::GetDisplayServiceImpl(ctx, nv_flinger, Permission::User);
+}
+
 } // namespace Service::VI
diff --git a/src/core/hle/service/vi/vi_u.h b/src/core/hle/service/vi/vi_u.h
index b92f28c92..19bdb73b0 100644
--- a/src/core/hle/service/vi/vi_u.h
+++ b/src/core/hle/service/vi/vi_u.h
@@ -4,14 +4,27 @@
 
 #pragma once
 
-#include "core/hle/service/vi/vi.h"
+#include "core/hle/service/service.h"
+
+namespace Kernel {
+class HLERequestContext;
+}
+
+namespace Service::NVFlinger {
+class NVFlinger;
+}
 
 namespace Service::VI {
 
-class VI_U final : public Module::Interface {
+class VI_U final : public ServiceFramework<VI_U> {
 public:
-    explicit VI_U(std::shared_ptr<Module> module, std::shared_ptr<NVFlinger::NVFlinger> nv_flinger);
+    explicit VI_U(std::shared_ptr<NVFlinger::NVFlinger> nv_flinger);
     ~VI_U() override;
+
+private:
+    void GetDisplayService(Kernel::HLERequestContext& ctx);
+
+    std::shared_ptr<NVFlinger::NVFlinger> nv_flinger;
 };
 
 } // namespace Service::VI
diff --git a/src/core/loader/elf.cpp b/src/core/loader/elf.cpp
index 6057c7f26..8b1920f22 100644
--- a/src/core/loader/elf.cpp
+++ b/src/core/loader/elf.cpp
@@ -9,6 +9,7 @@
 #include "common/common_types.h"
 #include "common/file_util.h"
 #include "common/logging/log.h"
+#include "core/hle/kernel/code_set.h"
 #include "core/hle/kernel/process.h"
 #include "core/hle/kernel/vm_manager.h"
 #include "core/loader/elf.h"
diff --git a/src/core/loader/linker.cpp b/src/core/loader/linker.cpp
deleted file mode 100644
index 57ca8c3ee..000000000
--- a/src/core/loader/linker.cpp
+++ /dev/null
@@ -1,147 +0,0 @@
-// Copyright 2018 yuzu emulator team
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#include <vector>
-
-#include "common/common_funcs.h"
-#include "common/logging/log.h"
-#include "common/swap.h"
-#include "core/loader/linker.h"
-#include "core/memory.h"
-
-namespace Loader {
-
-enum class RelocationType : u32 { ABS64 = 257, GLOB_DAT = 1025, JUMP_SLOT = 1026, RELATIVE = 1027 };
-
-enum DynamicType : u32 {
-    DT_NULL = 0,
-    DT_PLTRELSZ = 2,
-    DT_STRTAB = 5,
-    DT_SYMTAB = 6,
-    DT_RELA = 7,
-    DT_RELASZ = 8,
-    DT_STRSZ = 10,
-    DT_JMPREL = 23,
-};
-
-struct Elf64_Rela {
-    u64_le offset;
-    RelocationType type;
-    u32_le symbol;
-    s64_le addend;
-};
-static_assert(sizeof(Elf64_Rela) == 0x18, "Elf64_Rela has incorrect size.");
-
-struct Elf64_Dyn {
-    u64_le tag;
-    u64_le value;
-};
-static_assert(sizeof(Elf64_Dyn) == 0x10, "Elf64_Dyn has incorrect size.");
-
-struct Elf64_Sym {
-    u32_le name;
-    INSERT_PADDING_BYTES(0x2);
-    u16_le shndx;
-    u64_le value;
-    u64_le size;
-};
-static_assert(sizeof(Elf64_Sym) == 0x18, "Elf64_Sym has incorrect size.");
-
-void Linker::WriteRelocations(std::vector<u8>& program_image, const std::vector<Symbol>& symbols,
-                              u64 relocation_offset, u64 size, VAddr load_base) {
-    for (u64 i = 0; i < size; i += sizeof(Elf64_Rela)) {
-        Elf64_Rela rela;
-        std::memcpy(&rela, &program_image[relocation_offset + i], sizeof(Elf64_Rela));
-
-        const Symbol& symbol = symbols[rela.symbol];
-        switch (rela.type) {
-        case RelocationType::RELATIVE: {
-            const u64 value = load_base + rela.addend;
-            if (!symbol.name.empty()) {
-                exports[symbol.name] = value;
-            }
-            std::memcpy(&program_image[rela.offset], &value, sizeof(u64));
-            break;
-        }
-        case RelocationType::JUMP_SLOT:
-        case RelocationType::GLOB_DAT:
-            if (!symbol.value) {
-                imports[symbol.name] = {rela.offset + load_base, 0};
-            } else {
-                exports[symbol.name] = symbol.value;
-                std::memcpy(&program_image[rela.offset], &symbol.value, sizeof(u64));
-            }
-            break;
-        case RelocationType::ABS64:
-            if (!symbol.value) {
-                imports[symbol.name] = {rela.offset + load_base, rela.addend};
-            } else {
-                const u64 value = symbol.value + rela.addend;
-                exports[symbol.name] = value;
-                std::memcpy(&program_image[rela.offset], &value, sizeof(u64));
-            }
-            break;
-        default:
-            LOG_CRITICAL(Loader, "Unknown relocation type: {}", static_cast<int>(rela.type));
-            break;
-        }
-    }
-}
-
-void Linker::Relocate(std::vector<u8>& program_image, u32 dynamic_section_offset, VAddr load_base) {
-    std::map<u64, u64> dynamic;
-    while (dynamic_section_offset < program_image.size()) {
-        Elf64_Dyn dyn;
-        std::memcpy(&dyn, &program_image[dynamic_section_offset], sizeof(Elf64_Dyn));
-        dynamic_section_offset += sizeof(Elf64_Dyn);
-
-        if (dyn.tag == DT_NULL) {
-            break;
-        }
-        dynamic[dyn.tag] = dyn.value;
-    }
-
-    u64 offset = dynamic[DT_SYMTAB];
-    std::vector<Symbol> symbols;
-    while (offset < program_image.size()) {
-        Elf64_Sym sym;
-        std::memcpy(&sym, &program_image[offset], sizeof(Elf64_Sym));
-        offset += sizeof(Elf64_Sym);
-
-        if (sym.name >= dynamic[DT_STRSZ]) {
-            break;
-        }
-
-        std::string name = reinterpret_cast<char*>(&program_image[dynamic[DT_STRTAB] + sym.name]);
-        if (sym.value) {
-            exports[name] = load_base + sym.value;
-            symbols.emplace_back(std::move(name), load_base + sym.value);
-        } else {
-            symbols.emplace_back(std::move(name), 0);
-        }
-    }
-
-    if (dynamic.find(DT_RELA) != dynamic.end()) {
-        WriteRelocations(program_image, symbols, dynamic[DT_RELA], dynamic[DT_RELASZ], load_base);
-    }
-
-    if (dynamic.find(DT_JMPREL) != dynamic.end()) {
-        WriteRelocations(program_image, symbols, dynamic[DT_JMPREL], dynamic[DT_PLTRELSZ],
-                         load_base);
-    }
-}
-
-void Linker::ResolveImports() {
-    // Resolve imports
-    for (const auto& import : imports) {
-        const auto& search = exports.find(import.first);
-        if (search != exports.end()) {
-            Memory::Write64(import.second.ea, search->second + import.second.addend);
-        } else {
-            LOG_ERROR(Loader, "Unresolved import: {}", import.first);
-        }
-    }
-}
-
-} // namespace Loader
diff --git a/src/core/loader/linker.h b/src/core/loader/linker.h
deleted file mode 100644
index 107625837..000000000
--- a/src/core/loader/linker.h
+++ /dev/null
@@ -1,36 +0,0 @@
-// Copyright 2018 yuzu emulator team
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#pragma once
-
-#include <map>
-#include <string>
-#include "common/common_types.h"
-
-namespace Loader {
-
-class Linker {
-protected:
-    struct Symbol {
-        Symbol(std::string&& name, u64 value) : name(std::move(name)), value(value) {}
-        std::string name;
-        u64 value;
-    };
-
-    struct Import {
-        VAddr ea;
-        s64 addend;
-    };
-
-    void WriteRelocations(std::vector<u8>& program_image, const std::vector<Symbol>& symbols,
-                          u64 relocation_offset, u64 size, VAddr load_base);
-    void Relocate(std::vector<u8>& program_image, u32 dynamic_section_offset, VAddr load_base);
-
-    void ResolveImports();
-
-    std::map<std::string, Import> imports;
-    std::map<std::string, VAddr> exports;
-};
-
-} // namespace Loader
diff --git a/src/core/loader/nro.cpp b/src/core/loader/nro.cpp
index 4fad0c0dd..5de02a94b 100644
--- a/src/core/loader/nro.cpp
+++ b/src/core/loader/nro.cpp
@@ -14,6 +14,7 @@
 #include "core/file_sys/romfs_factory.h"
 #include "core/file_sys/vfs_offset.h"
 #include "core/gdbstub/gdbstub.h"
+#include "core/hle/kernel/code_set.h"
 #include "core/hle/kernel/process.h"
 #include "core/hle/kernel/vm_manager.h"
 #include "core/hle/service/filesystem/filesystem.h"
diff --git a/src/core/loader/nro.h b/src/core/loader/nro.h
index 013d629c0..85b0ed644 100644
--- a/src/core/loader/nro.h
+++ b/src/core/loader/nro.h
@@ -4,10 +4,10 @@
 
 #pragma once
 
+#include <memory>
 #include <string>
 #include <vector>
 #include "common/common_types.h"
-#include "core/loader/linker.h"
 #include "core/loader/loader.h"
 
 namespace FileSys {
@@ -21,7 +21,7 @@ class Process;
 namespace Loader {
 
 /// Loads an NRO file
-class AppLoader_NRO final : public AppLoader, Linker {
+class AppLoader_NRO final : public AppLoader {
 public:
     explicit AppLoader_NRO(FileSys::VirtualFile file);
     ~AppLoader_NRO() override;
diff --git a/src/core/loader/nso.cpp b/src/core/loader/nso.cpp
index 6ded0b707..e1c8908a1 100644
--- a/src/core/loader/nso.cpp
+++ b/src/core/loader/nso.cpp
@@ -11,6 +11,7 @@
 #include "common/swap.h"
 #include "core/file_sys/patch_manager.h"
 #include "core/gdbstub/gdbstub.h"
+#include "core/hle/kernel/code_set.h"
 #include "core/hle/kernel/process.h"
 #include "core/hle/kernel/vm_manager.h"
 #include "core/loader/nso.h"
diff --git a/src/core/loader/nso.h b/src/core/loader/nso.h
index 135b6ea5a..167c8a694 100644
--- a/src/core/loader/nso.h
+++ b/src/core/loader/nso.h
@@ -6,8 +6,8 @@
 
 #include <optional>
 #include "common/common_types.h"
+#include "common/swap.h"
 #include "core/file_sys/patch_manager.h"
-#include "core/loader/linker.h"
 #include "core/loader/loader.h"
 
 namespace Kernel {
@@ -26,7 +26,7 @@ struct NSOArgumentHeader {
 static_assert(sizeof(NSOArgumentHeader) == 0x20, "NSOArgumentHeader has incorrect size.");
 
 /// Loads an NSO file
-class AppLoader_NSO final : public AppLoader, Linker {
+class AppLoader_NSO final : public AppLoader {
 public:
     explicit AppLoader_NSO(FileSys::VirtualFile file);
 
diff --git a/src/core/memory.cpp b/src/core/memory.cpp
index e9166dbd9..365ac82b4 100644
--- a/src/core/memory.cpp
+++ b/src/core/memory.cpp
@@ -10,6 +10,7 @@
 #include "common/assert.h"
 #include "common/common_types.h"
 #include "common/logging/log.h"
+#include "common/page_table.h"
 #include "common/swap.h"
 #include "core/arm/arm_interface.h"
 #include "core/core.h"
@@ -18,13 +19,14 @@
 #include "core/hle/lock.h"
 #include "core/memory.h"
 #include "core/memory_setup.h"
+#include "video_core/gpu.h"
 #include "video_core/renderer_base.h"
 
 namespace Memory {
 
-static PageTable* current_page_table = nullptr;
+static Common::PageTable* current_page_table = nullptr;
 
-void SetCurrentPageTable(PageTable* page_table) {
+void SetCurrentPageTable(Common::PageTable* page_table) {
     current_page_table = page_table;
 
     auto& system = Core::System::GetInstance();
@@ -36,88 +38,80 @@ void SetCurrentPageTable(PageTable* page_table) {
     }
 }
 
-PageTable* GetCurrentPageTable() {
+Common::PageTable* GetCurrentPageTable() {
     return current_page_table;
 }
 
-PageTable::PageTable() = default;
-
-PageTable::PageTable(std::size_t address_space_width_in_bits) {
-    Resize(address_space_width_in_bits);
-}
-
-PageTable::~PageTable() = default;
-
-void PageTable::Resize(std::size_t address_space_width_in_bits) {
-    const std::size_t num_page_table_entries = 1ULL << (address_space_width_in_bits - PAGE_BITS);
-
-    pointers.resize(num_page_table_entries);
-    attributes.resize(num_page_table_entries);
-
-    // The default is a 39-bit address space, which causes an initial 1GB allocation size. If the
-    // vector size is subsequently decreased (via resize), the vector might not automatically
-    // actually reallocate/resize its underlying allocation, which wastes up to ~800 MB for
-    // 36-bit titles. Call shrink_to_fit to reduce capacity to what's actually in use.
-
-    pointers.shrink_to_fit();
-    attributes.shrink_to_fit();
-}
-
-static void MapPages(PageTable& page_table, VAddr base, u64 size, u8* memory, PageType type) {
+static void MapPages(Common::PageTable& page_table, VAddr base, u64 size, u8* memory,
+                     Common::PageType type) {
     LOG_DEBUG(HW_Memory, "Mapping {} onto {:016X}-{:016X}", fmt::ptr(memory), base * PAGE_SIZE,
               (base + size) * PAGE_SIZE);
 
-    RasterizerFlushVirtualRegion(base << PAGE_BITS, size * PAGE_SIZE,
-                                 FlushMode::FlushAndInvalidate);
+    // During boot, current_page_table might not be set yet, in which case we need not flush
+    if (current_page_table) {
+        Core::System::GetInstance().GPU().FlushAndInvalidateRegion(base << PAGE_BITS,
+                                                                   size * PAGE_SIZE);
+    }
 
     VAddr end = base + size;
-    while (base != end) {
-        ASSERT_MSG(base < page_table.pointers.size(), "out of range mapping at {:016X}", base);
+    ASSERT_MSG(end <= page_table.pointers.size(), "out of range mapping at {:016X}",
+               base + page_table.pointers.size());
 
-        page_table.attributes[base] = type;
-        page_table.pointers[base] = memory;
+    std::fill(page_table.attributes.begin() + base, page_table.attributes.begin() + end, type);
 
-        base += 1;
-        if (memory != nullptr)
+    if (memory == nullptr) {
+        std::fill(page_table.pointers.begin() + base, page_table.pointers.begin() + end, memory);
+    } else {
+        while (base != end) {
+            page_table.pointers[base] = memory;
+
+            base += 1;
             memory += PAGE_SIZE;
+        }
     }
 }
 
-void MapMemoryRegion(PageTable& page_table, VAddr base, u64 size, u8* target) {
+void MapMemoryRegion(Common::PageTable& page_table, VAddr base, u64 size, u8* target) {
     ASSERT_MSG((size & PAGE_MASK) == 0, "non-page aligned size: {:016X}", size);
     ASSERT_MSG((base & PAGE_MASK) == 0, "non-page aligned base: {:016X}", base);
-    MapPages(page_table, base / PAGE_SIZE, size / PAGE_SIZE, target, PageType::Memory);
+    MapPages(page_table, base / PAGE_SIZE, size / PAGE_SIZE, target, Common::PageType::Memory);
 }
 
-void MapIoRegion(PageTable& page_table, VAddr base, u64 size, MemoryHookPointer mmio_handler) {
+void MapIoRegion(Common::PageTable& page_table, VAddr base, u64 size,
+                 Common::MemoryHookPointer mmio_handler) {
     ASSERT_MSG((size & PAGE_MASK) == 0, "non-page aligned size: {:016X}", size);
     ASSERT_MSG((base & PAGE_MASK) == 0, "non-page aligned base: {:016X}", base);
-    MapPages(page_table, base / PAGE_SIZE, size / PAGE_SIZE, nullptr, PageType::Special);
+    MapPages(page_table, base / PAGE_SIZE, size / PAGE_SIZE, nullptr, Common::PageType::Special);
 
     auto interval = boost::icl::discrete_interval<VAddr>::closed(base, base + size - 1);
-    SpecialRegion region{SpecialRegion::Type::IODevice, std::move(mmio_handler)};
-    page_table.special_regions.add(std::make_pair(interval, std::set<SpecialRegion>{region}));
+    Common::SpecialRegion region{Common::SpecialRegion::Type::IODevice, std::move(mmio_handler)};
+    page_table.special_regions.add(
+        std::make_pair(interval, std::set<Common::SpecialRegion>{region}));
 }
 
-void UnmapRegion(PageTable& page_table, VAddr base, u64 size) {
+void UnmapRegion(Common::PageTable& page_table, VAddr base, u64 size) {
     ASSERT_MSG((size & PAGE_MASK) == 0, "non-page aligned size: {:016X}", size);
     ASSERT_MSG((base & PAGE_MASK) == 0, "non-page aligned base: {:016X}", base);
-    MapPages(page_table, base / PAGE_SIZE, size / PAGE_SIZE, nullptr, PageType::Unmapped);
+    MapPages(page_table, base / PAGE_SIZE, size / PAGE_SIZE, nullptr, Common::PageType::Unmapped);
 
     auto interval = boost::icl::discrete_interval<VAddr>::closed(base, base + size - 1);
     page_table.special_regions.erase(interval);
 }
 
-void AddDebugHook(PageTable& page_table, VAddr base, u64 size, MemoryHookPointer hook) {
+void AddDebugHook(Common::PageTable& page_table, VAddr base, u64 size,
+                  Common::MemoryHookPointer hook) {
     auto interval = boost::icl::discrete_interval<VAddr>::closed(base, base + size - 1);
-    SpecialRegion region{SpecialRegion::Type::DebugHook, std::move(hook)};
-    page_table.special_regions.add(std::make_pair(interval, std::set<SpecialRegion>{region}));
+    Common::SpecialRegion region{Common::SpecialRegion::Type::DebugHook, std::move(hook)};
+    page_table.special_regions.add(
+        std::make_pair(interval, std::set<Common::SpecialRegion>{region}));
 }
 
-void RemoveDebugHook(PageTable& page_table, VAddr base, u64 size, MemoryHookPointer hook) {
+void RemoveDebugHook(Common::PageTable& page_table, VAddr base, u64 size,
+                     Common::MemoryHookPointer hook) {
     auto interval = boost::icl::discrete_interval<VAddr>::closed(base, base + size - 1);
-    SpecialRegion region{SpecialRegion::Type::DebugHook, std::move(hook)};
-    page_table.special_regions.subtract(std::make_pair(interval, std::set<SpecialRegion>{region}));
+    Common::SpecialRegion region{Common::SpecialRegion::Type::DebugHook, std::move(hook)};
+    page_table.special_regions.subtract(
+        std::make_pair(interval, std::set<Common::SpecialRegion>{region}));
 }
 
 /**
@@ -166,22 +160,19 @@ T Read(const VAddr vaddr) {
         return value;
     }
 
-    // The memory access might do an MMIO or cached access, so we have to lock the HLE kernel state
-    std::lock_guard<std::recursive_mutex> lock(HLE::g_hle_lock);
-
-    PageType type = current_page_table->attributes[vaddr >> PAGE_BITS];
+    Common::PageType type = current_page_table->attributes[vaddr >> PAGE_BITS];
     switch (type) {
-    case PageType::Unmapped:
+    case Common::PageType::Unmapped:
         LOG_ERROR(HW_Memory, "Unmapped Read{} @ 0x{:08X}", sizeof(T) * 8, vaddr);
         return 0;
-    case PageType::Memory:
+    case Common::PageType::Memory:
         ASSERT_MSG(false, "Mapped memory page without a pointer @ {:016X}", vaddr);
         break;
-    case PageType::RasterizerCachedMemory: {
-        RasterizerFlushVirtualRegion(vaddr, sizeof(T), FlushMode::Flush);
-
+    case Common::PageType::RasterizerCachedMemory: {
+        auto host_ptr{GetPointerFromVMA(vaddr)};
+        Core::System::GetInstance().GPU().FlushRegion(ToCacheAddr(host_ptr), sizeof(T));
         T value;
-        std::memcpy(&value, GetPointerFromVMA(vaddr), sizeof(T));
+        std::memcpy(&value, host_ptr, sizeof(T));
         return value;
     }
     default:
@@ -199,21 +190,19 @@ void Write(const VAddr vaddr, const T data) {
         return;
     }
 
-    // The memory access might do an MMIO or cached access, so we have to lock the HLE kernel state
-    std::lock_guard<std::recursive_mutex> lock(HLE::g_hle_lock);
-
-    PageType type = current_page_table->attributes[vaddr >> PAGE_BITS];
+    Common::PageType type = current_page_table->attributes[vaddr >> PAGE_BITS];
     switch (type) {
-    case PageType::Unmapped:
+    case Common::PageType::Unmapped:
         LOG_ERROR(HW_Memory, "Unmapped Write{} 0x{:08X} @ 0x{:016X}", sizeof(data) * 8,
                   static_cast<u32>(data), vaddr);
         return;
-    case PageType::Memory:
+    case Common::PageType::Memory:
         ASSERT_MSG(false, "Mapped memory page without a pointer @ {:016X}", vaddr);
         break;
-    case PageType::RasterizerCachedMemory: {
-        RasterizerFlushVirtualRegion(vaddr, sizeof(T), FlushMode::Invalidate);
-        std::memcpy(GetPointerFromVMA(vaddr), &data, sizeof(T));
+    case Common::PageType::RasterizerCachedMemory: {
+        auto host_ptr{GetPointerFromVMA(vaddr)};
+        Core::System::GetInstance().GPU().InvalidateRegion(ToCacheAddr(host_ptr), sizeof(T));
+        std::memcpy(host_ptr, &data, sizeof(T));
         break;
     }
     default:
@@ -228,10 +217,10 @@ bool IsValidVirtualAddress(const Kernel::Process& process, const VAddr vaddr) {
     if (page_pointer)
         return true;
 
-    if (page_table.attributes[vaddr >> PAGE_BITS] == PageType::RasterizerCachedMemory)
+    if (page_table.attributes[vaddr >> PAGE_BITS] == Common::PageType::RasterizerCachedMemory)
         return true;
 
-    if (page_table.attributes[vaddr >> PAGE_BITS] != PageType::Special)
+    if (page_table.attributes[vaddr >> PAGE_BITS] != Common::PageType::Special)
         return false;
 
     return false;
@@ -251,7 +240,8 @@ u8* GetPointer(const VAddr vaddr) {
         return page_pointer + (vaddr & PAGE_MASK);
     }
 
-    if (current_page_table->attributes[vaddr >> PAGE_BITS] == PageType::RasterizerCachedMemory) {
+    if (current_page_table->attributes[vaddr >> PAGE_BITS] ==
+        Common::PageType::RasterizerCachedMemory) {
         return GetPointerFromVMA(vaddr);
     }
 
@@ -285,20 +275,20 @@ void RasterizerMarkRegionCached(VAddr vaddr, u64 size, bool cached) {
 
     u64 num_pages = ((vaddr + size - 1) >> PAGE_BITS) - (vaddr >> PAGE_BITS) + 1;
     for (unsigned i = 0; i < num_pages; ++i, vaddr += PAGE_SIZE) {
-        PageType& page_type = current_page_table->attributes[vaddr >> PAGE_BITS];
+        Common::PageType& page_type = current_page_table->attributes[vaddr >> PAGE_BITS];
 
         if (cached) {
             // Switch page type to cached if now cached
             switch (page_type) {
-            case PageType::Unmapped:
+            case Common::PageType::Unmapped:
                 // It is not necessary for a process to have this region mapped into its address
                 // space, for example, a system module need not have a VRAM mapping.
                 break;
-            case PageType::Memory:
-                page_type = PageType::RasterizerCachedMemory;
+            case Common::PageType::Memory:
+                page_type = Common::PageType::RasterizerCachedMemory;
                 current_page_table->pointers[vaddr >> PAGE_BITS] = nullptr;
                 break;
-            case PageType::RasterizerCachedMemory:
+            case Common::PageType::RasterizerCachedMemory:
                 // There can be more than one GPU region mapped per CPU region, so it's common that
                 // this area is already marked as cached.
                 break;
@@ -308,23 +298,23 @@ void RasterizerMarkRegionCached(VAddr vaddr, u64 size, bool cached) {
         } else {
             // Switch page type to uncached if now uncached
             switch (page_type) {
-            case PageType::Unmapped:
+            case Common::PageType::Unmapped:
                 // It is not necessary for a process to have this region mapped into its address
                 // space, for example, a system module need not have a VRAM mapping.
                 break;
-            case PageType::Memory:
+            case Common::PageType::Memory:
                 // There can be more than one GPU region mapped per CPU region, so it's common that
                 // this area is already unmarked as cached.
                 break;
-            case PageType::RasterizerCachedMemory: {
+            case Common::PageType::RasterizerCachedMemory: {
                 u8* pointer = GetPointerFromVMA(vaddr & ~PAGE_MASK);
                 if (pointer == nullptr) {
                     // It's possible that this function has been called while updating the pagetable
                     // after unmapping a VMA. In that case the underlying VMA will no longer exist,
                     // and we should just leave the pagetable entry blank.
-                    page_type = PageType::Unmapped;
+                    page_type = Common::PageType::Unmapped;
                 } else {
-                    page_type = PageType::Memory;
+                    page_type = Common::PageType::Memory;
                     current_page_table->pointers[vaddr >> PAGE_BITS] = pointer;
                 }
                 break;
@@ -336,47 +326,6 @@ void RasterizerMarkRegionCached(VAddr vaddr, u64 size, bool cached) {
     }
 }
 
-void RasterizerFlushVirtualRegion(VAddr start, u64 size, FlushMode mode) {
-    auto& system_instance = Core::System::GetInstance();
-
-    // Since pages are unmapped on shutdown after video core is shutdown, the renderer may be
-    // null here
-    if (!system_instance.IsPoweredOn()) {
-        return;
-    }
-
-    const VAddr end = start + size;
-
-    const auto CheckRegion = [&](VAddr region_start, VAddr region_end) {
-        if (start >= region_end || end <= region_start) {
-            // No overlap with region
-            return;
-        }
-
-        const VAddr overlap_start = std::max(start, region_start);
-        const VAddr overlap_end = std::min(end, region_end);
-        const VAddr overlap_size = overlap_end - overlap_start;
-
-        auto& rasterizer = system_instance.Renderer().Rasterizer();
-        switch (mode) {
-        case FlushMode::Flush:
-            rasterizer.FlushRegion(overlap_start, overlap_size);
-            break;
-        case FlushMode::Invalidate:
-            rasterizer.InvalidateRegion(overlap_start, overlap_size);
-            break;
-        case FlushMode::FlushAndInvalidate:
-            rasterizer.FlushAndInvalidateRegion(overlap_start, overlap_size);
-            break;
-        }
-    };
-
-    const auto& vm_manager = Core::CurrentProcess()->VMManager();
-
-    CheckRegion(vm_manager.GetCodeRegionBaseAddress(), vm_manager.GetCodeRegionEndAddress());
-    CheckRegion(vm_manager.GetHeapRegionBaseAddress(), vm_manager.GetHeapRegionEndAddress());
-}
-
 u8 Read8(const VAddr addr) {
     return Read<u8>(addr);
 }
@@ -407,24 +356,24 @@ void ReadBlock(const Kernel::Process& process, const VAddr src_addr, void* dest_
         const VAddr current_vaddr = static_cast<VAddr>((page_index << PAGE_BITS) + page_offset);
 
         switch (page_table.attributes[page_index]) {
-        case PageType::Unmapped: {
+        case Common::PageType::Unmapped: {
             LOG_ERROR(HW_Memory,
                       "Unmapped ReadBlock @ 0x{:016X} (start address = 0x{:016X}, size = {})",
                       current_vaddr, src_addr, size);
             std::memset(dest_buffer, 0, copy_amount);
             break;
         }
-        case PageType::Memory: {
+        case Common::PageType::Memory: {
             DEBUG_ASSERT(page_table.pointers[page_index]);
 
             const u8* src_ptr = page_table.pointers[page_index] + page_offset;
             std::memcpy(dest_buffer, src_ptr, copy_amount);
             break;
         }
-        case PageType::RasterizerCachedMemory: {
-            RasterizerFlushVirtualRegion(current_vaddr, static_cast<u32>(copy_amount),
-                                         FlushMode::Flush);
-            std::memcpy(dest_buffer, GetPointerFromVMA(process, current_vaddr), copy_amount);
+        case Common::PageType::RasterizerCachedMemory: {
+            const auto& host_ptr{GetPointerFromVMA(process, current_vaddr)};
+            Core::System::GetInstance().GPU().FlushRegion(ToCacheAddr(host_ptr), copy_amount);
+            std::memcpy(dest_buffer, host_ptr, copy_amount);
             break;
         }
         default:
@@ -471,23 +420,23 @@ void WriteBlock(const Kernel::Process& process, const VAddr dest_addr, const voi
         const VAddr current_vaddr = static_cast<VAddr>((page_index << PAGE_BITS) + page_offset);
 
         switch (page_table.attributes[page_index]) {
-        case PageType::Unmapped: {
+        case Common::PageType::Unmapped: {
             LOG_ERROR(HW_Memory,
                       "Unmapped WriteBlock @ 0x{:016X} (start address = 0x{:016X}, size = {})",
                       current_vaddr, dest_addr, size);
             break;
         }
-        case PageType::Memory: {
+        case Common::PageType::Memory: {
             DEBUG_ASSERT(page_table.pointers[page_index]);
 
             u8* dest_ptr = page_table.pointers[page_index] + page_offset;
             std::memcpy(dest_ptr, src_buffer, copy_amount);
             break;
         }
-        case PageType::RasterizerCachedMemory: {
-            RasterizerFlushVirtualRegion(current_vaddr, static_cast<u32>(copy_amount),
-                                         FlushMode::Invalidate);
-            std::memcpy(GetPointerFromVMA(process, current_vaddr), src_buffer, copy_amount);
+        case Common::PageType::RasterizerCachedMemory: {
+            const auto& host_ptr{GetPointerFromVMA(process, current_vaddr)};
+            Core::System::GetInstance().GPU().InvalidateRegion(ToCacheAddr(host_ptr), copy_amount);
+            std::memcpy(host_ptr, src_buffer, copy_amount);
             break;
         }
         default:
@@ -517,23 +466,23 @@ void ZeroBlock(const Kernel::Process& process, const VAddr dest_addr, const std:
         const VAddr current_vaddr = static_cast<VAddr>((page_index << PAGE_BITS) + page_offset);
 
         switch (page_table.attributes[page_index]) {
-        case PageType::Unmapped: {
+        case Common::PageType::Unmapped: {
             LOG_ERROR(HW_Memory,
                       "Unmapped ZeroBlock @ 0x{:016X} (start address = 0x{:016X}, size = {})",
                       current_vaddr, dest_addr, size);
             break;
         }
-        case PageType::Memory: {
+        case Common::PageType::Memory: {
             DEBUG_ASSERT(page_table.pointers[page_index]);
 
             u8* dest_ptr = page_table.pointers[page_index] + page_offset;
             std::memset(dest_ptr, 0, copy_amount);
             break;
         }
-        case PageType::RasterizerCachedMemory: {
-            RasterizerFlushVirtualRegion(current_vaddr, static_cast<u32>(copy_amount),
-                                         FlushMode::Invalidate);
-            std::memset(GetPointerFromVMA(process, current_vaddr), 0, copy_amount);
+        case Common::PageType::RasterizerCachedMemory: {
+            const auto& host_ptr{GetPointerFromVMA(process, current_vaddr)};
+            Core::System::GetInstance().GPU().InvalidateRegion(ToCacheAddr(host_ptr), copy_amount);
+            std::memset(host_ptr, 0, copy_amount);
             break;
         }
         default:
@@ -559,23 +508,23 @@ void CopyBlock(const Kernel::Process& process, VAddr dest_addr, VAddr src_addr,
         const VAddr current_vaddr = static_cast<VAddr>((page_index << PAGE_BITS) + page_offset);
 
         switch (page_table.attributes[page_index]) {
-        case PageType::Unmapped: {
+        case Common::PageType::Unmapped: {
             LOG_ERROR(HW_Memory,
                       "Unmapped CopyBlock @ 0x{:016X} (start address = 0x{:016X}, size = {})",
                       current_vaddr, src_addr, size);
             ZeroBlock(process, dest_addr, copy_amount);
             break;
         }
-        case PageType::Memory: {
+        case Common::PageType::Memory: {
             DEBUG_ASSERT(page_table.pointers[page_index]);
             const u8* src_ptr = page_table.pointers[page_index] + page_offset;
             WriteBlock(process, dest_addr, src_ptr, copy_amount);
             break;
         }
-        case PageType::RasterizerCachedMemory: {
-            RasterizerFlushVirtualRegion(current_vaddr, static_cast<u32>(copy_amount),
-                                         FlushMode::Flush);
-            WriteBlock(process, dest_addr, GetPointerFromVMA(process, current_vaddr), copy_amount);
+        case Common::PageType::RasterizerCachedMemory: {
+            const auto& host_ptr{GetPointerFromVMA(process, current_vaddr)};
+            Core::System::GetInstance().GPU().FlushRegion(ToCacheAddr(host_ptr), copy_amount);
+            WriteBlock(process, dest_addr, host_ptr, copy_amount);
             break;
         }
         default:
diff --git a/src/core/memory.h b/src/core/memory.h
index 1acf5ce8c..3f60d868c 100644
--- a/src/core/memory.h
+++ b/src/core/memory.h
@@ -10,7 +10,10 @@
 #include <vector>
 #include <boost/icl/interval_map.hpp>
 #include "common/common_types.h"
-#include "core/memory_hook.h"
+
+namespace Common {
+struct PageTable;
+}
 
 namespace Kernel {
 class Process;
@@ -26,71 +29,6 @@ constexpr std::size_t PAGE_BITS = 12;
 constexpr u64 PAGE_SIZE = 1ULL << PAGE_BITS;
 constexpr u64 PAGE_MASK = PAGE_SIZE - 1;
 
-enum class PageType : u8 {
-    /// Page is unmapped and should cause an access error.
-    Unmapped,
-    /// Page is mapped to regular memory. This is the only type you can get pointers to.
-    Memory,
-    /// Page is mapped to regular memory, but also needs to check for rasterizer cache flushing and
-    /// invalidation
-    RasterizerCachedMemory,
-    /// Page is mapped to a I/O region. Writing and reading to this page is handled by functions.
-    Special,
-};
-
-struct SpecialRegion {
-    enum class Type {
-        DebugHook,
-        IODevice,
-    } type;
-
-    MemoryHookPointer handler;
-
-    bool operator<(const SpecialRegion& other) const {
-        return std::tie(type, handler) < std::tie(other.type, other.handler);
-    }
-
-    bool operator==(const SpecialRegion& other) const {
-        return std::tie(type, handler) == std::tie(other.type, other.handler);
-    }
-};
-
-/**
- * A (reasonably) fast way of allowing switchable and remappable process address spaces. It loosely
- * mimics the way a real CPU page table works.
- */
-struct PageTable {
-    explicit PageTable();
-    explicit PageTable(std::size_t address_space_width_in_bits);
-    ~PageTable();
-
-    /**
-     * Resizes the page table to be able to accomodate enough pages within
-     * a given address space.
-     *
-     * @param address_space_width_in_bits The address size width in bits.
-     */
-    void Resize(std::size_t address_space_width_in_bits);
-
-    /**
-     * Vector of memory pointers backing each page. An entry can only be non-null if the
-     * corresponding entry in the `attributes` vector is of type `Memory`.
-     */
-    std::vector<u8*> pointers;
-
-    /**
-     * Contains MMIO handlers that back memory regions whose entries in the `attribute` vector is
-     * of type `Special`.
-     */
-    boost::icl::interval_map<VAddr, std::set<SpecialRegion>> special_regions;
-
-    /**
-     * Vector of fine grained page attributes. If it is set to any value other than `Memory`, then
-     * the corresponding entry in `pointers` MUST be set to null.
-     */
-    std::vector<PageType> attributes;
-};
-
 /// Virtual user-space memory regions
 enum : VAddr {
     /// Read-only page containing kernel and system configuration values.
@@ -116,8 +54,8 @@ enum : VAddr {
 };
 
 /// Currently active page table
-void SetCurrentPageTable(PageTable* page_table);
-PageTable* GetCurrentPageTable();
+void SetCurrentPageTable(Common::PageTable* page_table);
+Common::PageTable* GetCurrentPageTable();
 
 /// Determines if the given VAddr is valid for the specified process.
 bool IsValidVirtualAddress(const Kernel::Process& process, VAddr vaddr);
@@ -161,10 +99,4 @@ enum class FlushMode {
  */
 void RasterizerMarkRegionCached(VAddr vaddr, u64 size, bool cached);
 
-/**
- * Flushes and invalidates any externally cached rasterizer resources touching the given virtual
- * address region.
- */
-void RasterizerFlushVirtualRegion(VAddr start, u64 size, FlushMode mode);
-
 } // namespace Memory
diff --git a/src/core/memory_setup.h b/src/core/memory_setup.h
index 9a1a4f4be..5225ee8e2 100644
--- a/src/core/memory_setup.h
+++ b/src/core/memory_setup.h
@@ -5,7 +5,11 @@
 #pragma once
 
 #include "common/common_types.h"
-#include "core/memory_hook.h"
+#include "common/memory_hook.h"
+
+namespace Common {
+struct PageTable;
+}
 
 namespace Memory {
 
@@ -17,7 +21,7 @@ namespace Memory {
  * @param size The amount of bytes to map. Must be page-aligned.
  * @param target Buffer with the memory backing the mapping. Must be of length at least `size`.
  */
-void MapMemoryRegion(PageTable& page_table, VAddr base, u64 size, u8* target);
+void MapMemoryRegion(Common::PageTable& page_table, VAddr base, u64 size, u8* target);
 
 /**
  * Maps a region of the emulated process address space as a IO region.
@@ -26,11 +30,14 @@ void MapMemoryRegion(PageTable& page_table, VAddr base, u64 size, u8* target);
  * @param size The amount of bytes to map. Must be page-aligned.
  * @param mmio_handler The handler that backs the mapping.
  */
-void MapIoRegion(PageTable& page_table, VAddr base, u64 size, MemoryHookPointer mmio_handler);
+void MapIoRegion(Common::PageTable& page_table, VAddr base, u64 size,
+                 Common::MemoryHookPointer mmio_handler);
 
-void UnmapRegion(PageTable& page_table, VAddr base, u64 size);
+void UnmapRegion(Common::PageTable& page_table, VAddr base, u64 size);
 
-void AddDebugHook(PageTable& page_table, VAddr base, u64 size, MemoryHookPointer hook);
-void RemoveDebugHook(PageTable& page_table, VAddr base, u64 size, MemoryHookPointer hook);
+void AddDebugHook(Common::PageTable& page_table, VAddr base, u64 size,
+                  Common::MemoryHookPointer hook);
+void RemoveDebugHook(Common::PageTable& page_table, VAddr base, u64 size,
+                     Common::MemoryHookPointer hook);
 
 } // namespace Memory
diff --git a/src/core/settings.cpp b/src/core/settings.cpp
index 2e232e1e7..6dd3139cc 100644
--- a/src/core/settings.cpp
+++ b/src/core/settings.cpp
@@ -91,7 +91,10 @@ void LogSettings() {
     LogSetting("Renderer_UseResolutionFactor", Settings::values.resolution_factor);
     LogSetting("Renderer_UseFrameLimit", Settings::values.use_frame_limit);
     LogSetting("Renderer_FrameLimit", Settings::values.frame_limit);
+    LogSetting("Renderer_UseDiskShaderCache", Settings::values.use_disk_shader_cache);
     LogSetting("Renderer_UseAccurateGpuEmulation", Settings::values.use_accurate_gpu_emulation);
+    LogSetting("Renderer_UseAsynchronousGpuEmulation",
+               Settings::values.use_asynchronous_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);
diff --git a/src/core/settings.h b/src/core/settings.h
index c97387fc7..cdfb2f742 100644
--- a/src/core/settings.h
+++ b/src/core/settings.h
@@ -391,7 +391,9 @@ struct Values {
     float resolution_factor;
     bool use_frame_limit;
     u16 frame_limit;
+    bool use_disk_shader_cache;
     bool use_accurate_gpu_emulation;
+    bool use_asynchronous_gpu_emulation;
 
     float bg_red;
     float bg_green;
diff --git a/src/core/telemetry_session.cpp b/src/core/telemetry_session.cpp
index 09ed74d78..e1db06811 100644
--- a/src/core/telemetry_session.cpp
+++ b/src/core/telemetry_session.cpp
@@ -158,8 +158,12 @@ TelemetrySession::TelemetrySession() {
     AddField(Telemetry::FieldType::UserConfig, "Renderer_UseFrameLimit",
              Settings::values.use_frame_limit);
     AddField(Telemetry::FieldType::UserConfig, "Renderer_FrameLimit", Settings::values.frame_limit);
+    AddField(Telemetry::FieldType::UserConfig, "Renderer_UseDiskShaderCache",
+             Settings::values.use_disk_shader_cache);
     AddField(Telemetry::FieldType::UserConfig, "Renderer_UseAccurateGpuEmulation",
              Settings::values.use_accurate_gpu_emulation);
+    AddField(Telemetry::FieldType::UserConfig, "Renderer_UseAsynchronousGpuEmulation",
+             Settings::values.use_asynchronous_gpu_emulation);
     AddField(Telemetry::FieldType::UserConfig, "System_UseDockedMode",
              Settings::values.use_docked_mode);
 }
diff --git a/src/input_common/CMakeLists.txt b/src/input_common/CMakeLists.txt
index 1c7db28c0..5b4e032bd 100644
--- a/src/input_common/CMakeLists.txt
+++ b/src/input_common/CMakeLists.txt
@@ -7,15 +7,18 @@ add_library(input_common STATIC
     main.h
     motion_emu.cpp
     motion_emu.h
-
-    $<$<BOOL:${SDL2_FOUND}>:sdl/sdl.cpp sdl/sdl.h>
+    sdl/sdl.cpp
+    sdl/sdl.h
 )
 
-create_target_directory_groups(input_common)
-
-target_link_libraries(input_common PUBLIC core PRIVATE common)
-
 if(SDL2_FOUND)
+    target_sources(input_common PRIVATE
+        sdl/sdl_impl.cpp
+        sdl/sdl_impl.h
+    )
     target_link_libraries(input_common PRIVATE SDL2)
     target_compile_definitions(input_common PRIVATE HAVE_SDL2)
 endif()
+
+create_target_directory_groups(input_common)
+target_link_libraries(input_common PUBLIC core PRIVATE common)
diff --git a/src/input_common/main.cpp b/src/input_common/main.cpp
index 37f572853..8e66c1b15 100644
--- a/src/input_common/main.cpp
+++ b/src/input_common/main.cpp
@@ -17,10 +17,7 @@ namespace InputCommon {
 
 static std::shared_ptr<Keyboard> keyboard;
 static std::shared_ptr<MotionEmu> motion_emu;
-
-#ifdef HAVE_SDL2
-static std::thread poll_thread;
-#endif
+static std::unique_ptr<SDL::State> sdl;
 
 void Init() {
     keyboard = std::make_shared<Keyboard>();
@@ -30,15 +27,7 @@ void Init() {
     motion_emu = std::make_shared<MotionEmu>();
     Input::RegisterFactory<Input::MotionDevice>("motion_emu", motion_emu);
 
-#ifdef HAVE_SDL2
-    SDL::Init();
-#endif
-}
-
-void StartJoystickEventHandler() {
-#ifdef HAVE_SDL2
-    poll_thread = std::thread(SDL::PollLoop);
-#endif
+    sdl = SDL::Init();
 }
 
 void Shutdown() {
@@ -47,11 +36,7 @@ void Shutdown() {
     Input::UnregisterFactory<Input::AnalogDevice>("analog_from_button");
     Input::UnregisterFactory<Input::MotionDevice>("motion_emu");
     motion_emu.reset();
-
-#ifdef HAVE_SDL2
-    SDL::Shutdown();
-    poll_thread.join();
-#endif
+    sdl.reset();
 }
 
 Keyboard* GetKeyboard() {
@@ -88,7 +73,7 @@ namespace Polling {
 
 std::vector<std::unique_ptr<DevicePoller>> GetPollers(DeviceType type) {
 #ifdef HAVE_SDL2
-    return SDL::Polling::GetPollers(type);
+    return sdl->GetPollers(type);
 #else
     return {};
 #endif
diff --git a/src/input_common/main.h b/src/input_common/main.h
index 9eb13106e..77a0ce90b 100644
--- a/src/input_common/main.h
+++ b/src/input_common/main.h
@@ -20,8 +20,6 @@ void Init();
 /// Deregisters all built-in input device factories and shuts them down.
 void Shutdown();
 
-void StartJoystickEventHandler();
-
 class Keyboard;
 
 /// Gets the keyboard button device factory.
diff --git a/src/input_common/motion_emu.cpp b/src/input_common/motion_emu.cpp
index 9570c060e..6d96d4019 100644
--- a/src/input_common/motion_emu.cpp
+++ b/src/input_common/motion_emu.cpp
@@ -32,12 +32,12 @@ public:
     }
 
     void BeginTilt(int x, int y) {
-        mouse_origin = Math::MakeVec(x, y);
+        mouse_origin = Common::MakeVec(x, y);
         is_tilting = true;
     }
 
     void Tilt(int x, int y) {
-        auto mouse_move = Math::MakeVec(x, y) - mouse_origin;
+        auto mouse_move = Common::MakeVec(x, y) - mouse_origin;
         if (is_tilting) {
             std::lock_guard<std::mutex> guard(tilt_mutex);
             if (mouse_move.x == 0 && mouse_move.y == 0) {
@@ -45,7 +45,7 @@ public:
             } else {
                 tilt_direction = mouse_move.Cast<float>();
                 tilt_angle =
-                    std::clamp(tilt_direction.Normalize() * sensitivity, 0.0f, MathUtil::PI * 0.5f);
+                    std::clamp(tilt_direction.Normalize() * sensitivity, 0.0f, Common::PI * 0.5f);
             }
         }
     }
@@ -56,7 +56,7 @@ public:
         is_tilting = false;
     }
 
-    std::tuple<Math::Vec3<float>, Math::Vec3<float>> GetStatus() {
+    std::tuple<Common::Vec3<float>, Common::Vec3<float>> GetStatus() {
         std::lock_guard<std::mutex> guard(status_mutex);
         return status;
     }
@@ -66,17 +66,17 @@ private:
     const std::chrono::steady_clock::duration update_duration;
     const float sensitivity;
 
-    Math::Vec2<int> mouse_origin;
+    Common::Vec2<int> mouse_origin;
 
     std::mutex tilt_mutex;
-    Math::Vec2<float> tilt_direction;
+    Common::Vec2<float> tilt_direction;
     float tilt_angle = 0;
 
     bool is_tilting = false;
 
     Common::Event shutdown_event;
 
-    std::tuple<Math::Vec3<float>, Math::Vec3<float>> status;
+    std::tuple<Common::Vec3<float>, Common::Vec3<float>> status;
     std::mutex status_mutex;
 
     // Note: always keep the thread declaration at the end so that other objects are initialized
@@ -85,8 +85,8 @@ private:
 
     void MotionEmuThread() {
         auto update_time = std::chrono::steady_clock::now();
-        Math::Quaternion<float> q = MakeQuaternion(Math::Vec3<float>(), 0);
-        Math::Quaternion<float> old_q;
+        Common::Quaternion<float> q = Common::MakeQuaternion(Common::Vec3<float>(), 0);
+        Common::Quaternion<float> old_q;
 
         while (!shutdown_event.WaitUntil(update_time)) {
             update_time += update_duration;
@@ -96,18 +96,18 @@ private:
                 std::lock_guard<std::mutex> guard(tilt_mutex);
 
                 // Find the quaternion describing current 3DS tilting
-                q = MakeQuaternion(Math::MakeVec(-tilt_direction.y, 0.0f, tilt_direction.x),
-                                   tilt_angle);
+                q = Common::MakeQuaternion(
+                    Common::MakeVec(-tilt_direction.y, 0.0f, tilt_direction.x), tilt_angle);
             }
 
             auto inv_q = q.Inverse();
 
             // Set the gravity vector in world space
-            auto gravity = Math::MakeVec(0.0f, -1.0f, 0.0f);
+            auto gravity = Common::MakeVec(0.0f, -1.0f, 0.0f);
 
             // Find the angular rate vector in world space
             auto angular_rate = ((q - old_q) * inv_q).xyz * 2;
-            angular_rate *= 1000 / update_millisecond / MathUtil::PI * 180;
+            angular_rate *= 1000 / update_millisecond / Common::PI * 180;
 
             // Transform the two vectors from world space to 3DS space
             gravity = QuaternionRotate(inv_q, gravity);
@@ -131,7 +131,7 @@ public:
         device = std::make_shared<MotionEmuDevice>(update_millisecond, sensitivity);
     }
 
-    std::tuple<Math::Vec3<float>, Math::Vec3<float>> GetStatus() const override {
+    std::tuple<Common::Vec3<float>, Common::Vec3<float>> GetStatus() const override {
         return device->GetStatus();
     }
 
diff --git a/src/input_common/sdl/sdl.cpp b/src/input_common/sdl/sdl.cpp
index faf3c1fa3..644db3448 100644
--- a/src/input_common/sdl/sdl.cpp
+++ b/src/input_common/sdl/sdl.cpp
@@ -1,631 +1,19 @@
-// Copyright 2017 Citra Emulator Project
+// Copyright 2018 Citra Emulator Project
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
-#include <algorithm>
-#include <atomic>
-#include <cmath>
-#include <functional>
-#include <iterator>
-#include <mutex>
-#include <string>
-#include <thread>
-#include <tuple>
-#include <unordered_map>
-#include <utility>
-#include <vector>
-#include <SDL.h>
-#include "common/assert.h"
-#include "common/logging/log.h"
-#include "common/math_util.h"
-#include "common/param_package.h"
-#include "common/threadsafe_queue.h"
-#include "input_common/main.h"
 #include "input_common/sdl/sdl.h"
+#ifdef HAVE_SDL2
+#include "input_common/sdl/sdl_impl.h"
+#endif
 
-namespace InputCommon {
+namespace InputCommon::SDL {
 
-namespace SDL {
-
-class SDLJoystick;
-class SDLButtonFactory;
-class SDLAnalogFactory;
-
-/// Map of GUID of a list of corresponding virtual Joysticks
-static std::unordered_map<std::string, std::vector<std::shared_ptr<SDLJoystick>>> joystick_map;
-static std::mutex joystick_map_mutex;
-
-static std::shared_ptr<SDLButtonFactory> button_factory;
-static std::shared_ptr<SDLAnalogFactory> analog_factory;
-
-/// Used by the Pollers during config
-static std::atomic<bool> polling;
-static Common::SPSCQueue<SDL_Event> event_queue;
-
-static std::atomic<bool> initialized = false;
-
-static std::string GetGUID(SDL_Joystick* joystick) {
-    SDL_JoystickGUID guid = SDL_JoystickGetGUID(joystick);
-    char guid_str[33];
-    SDL_JoystickGetGUIDString(guid, guid_str, sizeof(guid_str));
-    return guid_str;
-}
-
-class SDLJoystick {
-public:
-    SDLJoystick(std::string guid_, int port_, SDL_Joystick* joystick,
-                decltype(&SDL_JoystickClose) deleter = &SDL_JoystickClose)
-        : guid{std::move(guid_)}, port{port_}, sdl_joystick{joystick, deleter} {}
-
-    void SetButton(int button, bool value) {
-        std::lock_guard<std::mutex> lock(mutex);
-        state.buttons[button] = value;
-    }
-
-    bool GetButton(int button) const {
-        std::lock_guard<std::mutex> lock(mutex);
-        return state.buttons.at(button);
-    }
-
-    void SetAxis(int axis, Sint16 value) {
-        std::lock_guard<std::mutex> lock(mutex);
-        state.axes[axis] = value;
-    }
-
-    float GetAxis(int axis) const {
-        std::lock_guard<std::mutex> lock(mutex);
-        return state.axes.at(axis) / 32767.0f;
-    }
-
-    std::tuple<float, float> GetAnalog(int axis_x, int axis_y) const {
-        float x = GetAxis(axis_x);
-        float y = GetAxis(axis_y);
-        y = -y; // 3DS uses an y-axis inverse from SDL
-
-        // Make sure the coordinates are in the unit circle,
-        // otherwise normalize it.
-        float r = x * x + y * y;
-        if (r > 1.0f) {
-            r = std::sqrt(r);
-            x /= r;
-            y /= r;
-        }
-
-        return std::make_tuple(x, y);
-    }
-
-    void SetHat(int hat, Uint8 direction) {
-        std::lock_guard<std::mutex> lock(mutex);
-        state.hats[hat] = direction;
-    }
-
-    bool GetHatDirection(int hat, Uint8 direction) const {
-        std::lock_guard<std::mutex> lock(mutex);
-        return (state.hats.at(hat) & direction) != 0;
-    }
-    /**
-     * The guid of the joystick
-     */
-    const std::string& GetGUID() const {
-        return guid;
-    }
-
-    /**
-     * The number of joystick from the same type that were connected before this joystick
-     */
-    int GetPort() const {
-        return port;
-    }
-
-    SDL_Joystick* GetSDLJoystick() const {
-        return sdl_joystick.get();
-    }
-
-    void SetSDLJoystick(SDL_Joystick* joystick,
-                        decltype(&SDL_JoystickClose) deleter = &SDL_JoystickClose) {
-        sdl_joystick =
-            std::unique_ptr<SDL_Joystick, decltype(&SDL_JoystickClose)>(joystick, deleter);
-    }
-
-private:
-    struct State {
-        std::unordered_map<int, bool> buttons;
-        std::unordered_map<int, Sint16> axes;
-        std::unordered_map<int, Uint8> hats;
-    } state;
-    std::string guid;
-    int port;
-    std::unique_ptr<SDL_Joystick, decltype(&SDL_JoystickClose)> sdl_joystick;
-    mutable std::mutex mutex;
-};
-
-/**
- * Get the nth joystick with the corresponding GUID
- */
-static std::shared_ptr<SDLJoystick> GetSDLJoystickByGUID(const std::string& guid, int port) {
-    std::lock_guard<std::mutex> lock(joystick_map_mutex);
-    const auto it = joystick_map.find(guid);
-    if (it != joystick_map.end()) {
-        while (it->second.size() <= port) {
-            auto joystick = std::make_shared<SDLJoystick>(guid, it->second.size(), nullptr,
-                                                          [](SDL_Joystick*) {});
-            it->second.emplace_back(std::move(joystick));
-        }
-        return it->second[port];
-    }
-    auto joystick = std::make_shared<SDLJoystick>(guid, 0, nullptr, [](SDL_Joystick*) {});
-    return joystick_map[guid].emplace_back(std::move(joystick));
-}
-
-/**
- * Check how many identical joysticks (by guid) were connected before the one with sdl_id and so tie
- * it to a SDLJoystick with the same guid and that port
- */
-static std::shared_ptr<SDLJoystick> GetSDLJoystickBySDLID(SDL_JoystickID sdl_id) {
-    std::lock_guard<std::mutex> lock(joystick_map_mutex);
-    auto sdl_joystick = SDL_JoystickFromInstanceID(sdl_id);
-    const std::string guid = GetGUID(sdl_joystick);
-    auto map_it = joystick_map.find(guid);
-    if (map_it != joystick_map.end()) {
-        auto vec_it = std::find_if(map_it->second.begin(), map_it->second.end(),
-                                   [&sdl_joystick](const std::shared_ptr<SDLJoystick>& joystick) {
-                                       return sdl_joystick == joystick->GetSDLJoystick();
-                                   });
-        if (vec_it != map_it->second.end()) {
-            // This is the common case: There is already an existing SDL_Joystick maped to a
-            // SDLJoystick. return the SDLJoystick
-            return *vec_it;
-        }
-        // Search for a SDLJoystick without a mapped SDL_Joystick...
-        auto nullptr_it = std::find_if(map_it->second.begin(), map_it->second.end(),
-                                       [](const std::shared_ptr<SDLJoystick>& joystick) {
-                                           return !joystick->GetSDLJoystick();
-                                       });
-        if (nullptr_it != map_it->second.end()) {
-            // ... and map it
-            (*nullptr_it)->SetSDLJoystick(sdl_joystick);
-            return *nullptr_it;
-        }
-        // There is no SDLJoystick without a mapped SDL_Joystick
-        // Create a new SDLJoystick
-        auto joystick = std::make_shared<SDLJoystick>(guid, map_it->second.size(), sdl_joystick);
-        return map_it->second.emplace_back(std::move(joystick));
-    }
-    auto joystick = std::make_shared<SDLJoystick>(guid, 0, sdl_joystick);
-    return joystick_map[guid].emplace_back(std::move(joystick));
-}
-
-void InitJoystick(int joystick_index) {
-    std::lock_guard<std::mutex> lock(joystick_map_mutex);
-    SDL_Joystick* sdl_joystick = SDL_JoystickOpen(joystick_index);
-    if (!sdl_joystick) {
-        LOG_ERROR(Input, "failed to open joystick {}", joystick_index);
-        return;
-    }
-    std::string guid = GetGUID(sdl_joystick);
-    if (joystick_map.find(guid) == joystick_map.end()) {
-        auto joystick = std::make_shared<SDLJoystick>(guid, 0, sdl_joystick);
-        joystick_map[guid].emplace_back(std::move(joystick));
-        return;
-    }
-    auto& joystick_guid_list = joystick_map[guid];
-    const auto it = std::find_if(
-        joystick_guid_list.begin(), joystick_guid_list.end(),
-        [](const std::shared_ptr<SDLJoystick>& joystick) { return !joystick->GetSDLJoystick(); });
-    if (it != joystick_guid_list.end()) {
-        (*it)->SetSDLJoystick(sdl_joystick);
-        return;
-    }
-    auto joystick = std::make_shared<SDLJoystick>(guid, joystick_guid_list.size(), sdl_joystick);
-    joystick_guid_list.emplace_back(std::move(joystick));
-}
-
-void CloseJoystick(SDL_Joystick* sdl_joystick) {
-    std::lock_guard<std::mutex> lock(joystick_map_mutex);
-    std::string guid = GetGUID(sdl_joystick);
-    // This call to guid is save since the joystick is guranteed to be in that map
-    auto& joystick_guid_list = joystick_map[guid];
-    const auto joystick_it =
-        std::find_if(joystick_guid_list.begin(), joystick_guid_list.end(),
-                     [&sdl_joystick](const std::shared_ptr<SDLJoystick>& joystick) {
-                         return joystick->GetSDLJoystick() == sdl_joystick;
-                     });
-    (*joystick_it)->SetSDLJoystick(nullptr, [](SDL_Joystick*) {});
-}
-
-void HandleGameControllerEvent(const SDL_Event& event) {
-    switch (event.type) {
-    case SDL_JOYBUTTONUP: {
-        auto joystick = GetSDLJoystickBySDLID(event.jbutton.which);
-        if (joystick) {
-            joystick->SetButton(event.jbutton.button, false);
-        }
-        break;
-    }
-    case SDL_JOYBUTTONDOWN: {
-        auto joystick = GetSDLJoystickBySDLID(event.jbutton.which);
-        if (joystick) {
-            joystick->SetButton(event.jbutton.button, true);
-        }
-        break;
-    }
-    case SDL_JOYHATMOTION: {
-        auto joystick = GetSDLJoystickBySDLID(event.jhat.which);
-        if (joystick) {
-            joystick->SetHat(event.jhat.hat, event.jhat.value);
-        }
-        break;
-    }
-    case SDL_JOYAXISMOTION: {
-        auto joystick = GetSDLJoystickBySDLID(event.jaxis.which);
-        if (joystick) {
-            joystick->SetAxis(event.jaxis.axis, event.jaxis.value);
-        }
-        break;
-    }
-    case SDL_JOYDEVICEREMOVED:
-        LOG_DEBUG(Input, "Controller removed with Instance_ID {}", event.jdevice.which);
-        CloseJoystick(SDL_JoystickFromInstanceID(event.jdevice.which));
-        break;
-    case SDL_JOYDEVICEADDED:
-        LOG_DEBUG(Input, "Controller connected with device index {}", event.jdevice.which);
-        InitJoystick(event.jdevice.which);
-        break;
-    }
-}
-
-void CloseSDLJoysticks() {
-    std::lock_guard<std::mutex> lock(joystick_map_mutex);
-    joystick_map.clear();
-}
-
-void PollLoop() {
-    if (SDL_Init(SDL_INIT_JOYSTICK) < 0) {
-        LOG_CRITICAL(Input, "SDL_Init(SDL_INIT_JOYSTICK) failed with: {}", SDL_GetError());
-        return;
-    }
-
-    SDL_Event event;
-    while (initialized) {
-        // Wait for 10 ms or until an event happens
-        if (SDL_WaitEventTimeout(&event, 10)) {
-            // Don't handle the event if we are configuring
-            if (polling) {
-                event_queue.Push(event);
-            } else {
-                HandleGameControllerEvent(event);
-            }
-        }
-    }
-    CloseSDLJoysticks();
-    SDL_QuitSubSystem(SDL_INIT_JOYSTICK);
-}
-
-class SDLButton final : public Input::ButtonDevice {
-public:
-    explicit SDLButton(std::shared_ptr<SDLJoystick> joystick_, int button_)
-        : joystick(std::move(joystick_)), button(button_) {}
-
-    bool GetStatus() const override {
-        return joystick->GetButton(button);
-    }
-
-private:
-    std::shared_ptr<SDLJoystick> joystick;
-    int button;
-};
-
-class SDLDirectionButton final : public Input::ButtonDevice {
-public:
-    explicit SDLDirectionButton(std::shared_ptr<SDLJoystick> joystick_, int hat_, Uint8 direction_)
-        : joystick(std::move(joystick_)), hat(hat_), direction(direction_) {}
-
-    bool GetStatus() const override {
-        return joystick->GetHatDirection(hat, direction);
-    }
-
-private:
-    std::shared_ptr<SDLJoystick> joystick;
-    int hat;
-    Uint8 direction;
-};
-
-class SDLAxisButton final : public Input::ButtonDevice {
-public:
-    explicit SDLAxisButton(std::shared_ptr<SDLJoystick> joystick_, int axis_, float threshold_,
-                           bool trigger_if_greater_)
-        : joystick(std::move(joystick_)), axis(axis_), threshold(threshold_),
-          trigger_if_greater(trigger_if_greater_) {}
-
-    bool GetStatus() const override {
-        float axis_value = joystick->GetAxis(axis);
-        if (trigger_if_greater)
-            return axis_value > threshold;
-        return axis_value < threshold;
-    }
-
-private:
-    std::shared_ptr<SDLJoystick> joystick;
-    int axis;
-    float threshold;
-    bool trigger_if_greater;
-};
-
-class SDLAnalog final : public Input::AnalogDevice {
-public:
-    SDLAnalog(std::shared_ptr<SDLJoystick> joystick_, int axis_x_, int axis_y_)
-        : joystick(std::move(joystick_)), axis_x(axis_x_), axis_y(axis_y_) {}
-
-    std::tuple<float, float> GetStatus() const override {
-        return joystick->GetAnalog(axis_x, axis_y);
-    }
-
-private:
-    std::shared_ptr<SDLJoystick> joystick;
-    int axis_x;
-    int axis_y;
-};
-
-/// A button device factory that creates button devices from SDL joystick
-class SDLButtonFactory final : public Input::Factory<Input::ButtonDevice> {
-public:
-    /**
-     * Creates a button device from a joystick button
-     * @param params contains parameters for creating the device:
-     *     - "guid": the guid of the joystick to bind
-     *     - "port": the nth joystick of the same type to bind
-     *     - "button"(optional): the index of the button to bind
-     *     - "hat"(optional): the index of the hat to bind as direction buttons
-     *     - "axis"(optional): the index of the axis to bind
-     *     - "direction"(only used for hat): the direction name of the hat to bind. Can be "up",
-     *         "down", "left" or "right"
-     *     - "threshold"(only used for axis): a float value in (-1.0, 1.0) which the button is
-     *         triggered if the axis value crosses
-     *     - "direction"(only used for axis): "+" means the button is triggered when the axis
-     * value is greater than the threshold; "-" means the button is triggered when the axis
-     * value is smaller than the threshold
-     */
-    std::unique_ptr<Input::ButtonDevice> Create(const Common::ParamPackage& params) override {
-        const std::string guid = params.Get("guid", "0");
-        const int port = params.Get("port", 0);
-
-        auto joystick = GetSDLJoystickByGUID(guid, port);
-
-        if (params.Has("hat")) {
-            const int hat = params.Get("hat", 0);
-            const std::string direction_name = params.Get("direction", "");
-            Uint8 direction;
-            if (direction_name == "up") {
-                direction = SDL_HAT_UP;
-            } else if (direction_name == "down") {
-                direction = SDL_HAT_DOWN;
-            } else if (direction_name == "left") {
-                direction = SDL_HAT_LEFT;
-            } else if (direction_name == "right") {
-                direction = SDL_HAT_RIGHT;
-            } else {
-                direction = 0;
-            }
-            // This is necessary so accessing GetHat with hat won't crash
-            joystick->SetHat(hat, SDL_HAT_CENTERED);
-            return std::make_unique<SDLDirectionButton>(joystick, hat, direction);
-        }
-
-        if (params.Has("axis")) {
-            const int axis = params.Get("axis", 0);
-            const float threshold = params.Get("threshold", 0.5f);
-            const std::string direction_name = params.Get("direction", "");
-            bool trigger_if_greater;
-            if (direction_name == "+") {
-                trigger_if_greater = true;
-            } else if (direction_name == "-") {
-                trigger_if_greater = false;
-            } else {
-                trigger_if_greater = true;
-                LOG_ERROR(Input, "Unknown direction '{}'", direction_name);
-            }
-            // This is necessary so accessing GetAxis with axis won't crash
-            joystick->SetAxis(axis, 0);
-            return std::make_unique<SDLAxisButton>(joystick, axis, threshold, trigger_if_greater);
-        }
-
-        const int button = params.Get("button", 0);
-        // This is necessary so accessing GetButton with button won't crash
-        joystick->SetButton(button, false);
-        return std::make_unique<SDLButton>(joystick, button);
-    }
-};
-
-/// An analog device factory that creates analog devices from SDL joystick
-class SDLAnalogFactory final : public Input::Factory<Input::AnalogDevice> {
-public:
-    /**
-     * Creates analog device from joystick axes
-     * @param params contains parameters for creating the device:
-     *     - "guid": the guid of the joystick to bind
-     *     - "port": the nth joystick of the same type
-     *     - "axis_x": the index of the axis to be bind as x-axis
-     *     - "axis_y": the index of the axis to be bind as y-axis
-     */
-    std::unique_ptr<Input::AnalogDevice> Create(const Common::ParamPackage& params) override {
-        const std::string guid = params.Get("guid", "0");
-        const int port = params.Get("port", 0);
-        const int axis_x = params.Get("axis_x", 0);
-        const int axis_y = params.Get("axis_y", 1);
-
-        auto joystick = GetSDLJoystickByGUID(guid, port);
-
-        // This is necessary so accessing GetAxis with axis_x and axis_y won't crash
-        joystick->SetAxis(axis_x, 0);
-        joystick->SetAxis(axis_y, 0);
-        return std::make_unique<SDLAnalog>(joystick, axis_x, axis_y);
-    }
-};
-
-void Init() {
-    using namespace Input;
-    RegisterFactory<ButtonDevice>("sdl", std::make_shared<SDLButtonFactory>());
-    RegisterFactory<AnalogDevice>("sdl", std::make_shared<SDLAnalogFactory>());
-    polling = false;
-    initialized = true;
-}
-
-void Shutdown() {
-    if (initialized) {
-        using namespace Input;
-        UnregisterFactory<ButtonDevice>("sdl");
-        UnregisterFactory<AnalogDevice>("sdl");
-        initialized = false;
-    }
-}
-
-Common::ParamPackage SDLEventToButtonParamPackage(const SDL_Event& event) {
-    Common::ParamPackage params({{"engine", "sdl"}});
-    switch (event.type) {
-    case SDL_JOYAXISMOTION: {
-        auto joystick = GetSDLJoystickBySDLID(event.jaxis.which);
-        params.Set("port", joystick->GetPort());
-        params.Set("guid", joystick->GetGUID());
-        params.Set("axis", event.jaxis.axis);
-        if (event.jaxis.value > 0) {
-            params.Set("direction", "+");
-            params.Set("threshold", "0.5");
-        } else {
-            params.Set("direction", "-");
-            params.Set("threshold", "-0.5");
-        }
-        break;
-    }
-    case SDL_JOYBUTTONUP: {
-        auto joystick = GetSDLJoystickBySDLID(event.jbutton.which);
-        params.Set("port", joystick->GetPort());
-        params.Set("guid", joystick->GetGUID());
-        params.Set("button", event.jbutton.button);
-        break;
-    }
-    case SDL_JOYHATMOTION: {
-        auto joystick = GetSDLJoystickBySDLID(event.jhat.which);
-        params.Set("port", joystick->GetPort());
-        params.Set("guid", joystick->GetGUID());
-        params.Set("hat", event.jhat.hat);
-        switch (event.jhat.value) {
-        case SDL_HAT_UP:
-            params.Set("direction", "up");
-            break;
-        case SDL_HAT_DOWN:
-            params.Set("direction", "down");
-            break;
-        case SDL_HAT_LEFT:
-            params.Set("direction", "left");
-            break;
-        case SDL_HAT_RIGHT:
-            params.Set("direction", "right");
-            break;
-        default:
-            return {};
-        }
-        break;
-    }
-    }
-    return params;
-}
-
-namespace Polling {
-
-class SDLPoller : public InputCommon::Polling::DevicePoller {
-public:
-    void Start() override {
-        event_queue.Clear();
-        polling = true;
-    }
-
-    void Stop() override {
-        polling = false;
-    }
-};
-
-class SDLButtonPoller final : public SDLPoller {
-public:
-    Common::ParamPackage GetNextInput() override {
-        SDL_Event event;
-        while (event_queue.Pop(event)) {
-            switch (event.type) {
-            case SDL_JOYAXISMOTION:
-                if (std::abs(event.jaxis.value / 32767.0) < 0.5) {
-                    break;
-                }
-            case SDL_JOYBUTTONUP:
-            case SDL_JOYHATMOTION:
-                return SDLEventToButtonParamPackage(event);
-            }
-        }
-        return {};
-    }
-};
-
-class SDLAnalogPoller final : public SDLPoller {
-public:
-    void Start() override {
-        SDLPoller::Start();
-
-        // Reset stored axes
-        analog_xaxis = -1;
-        analog_yaxis = -1;
-        analog_axes_joystick = -1;
-    }
-
-    Common::ParamPackage GetNextInput() override {
-        SDL_Event event;
-        while (event_queue.Pop(event)) {
-            if (event.type != SDL_JOYAXISMOTION || std::abs(event.jaxis.value / 32767.0) < 0.5) {
-                continue;
-            }
-            // An analog device needs two axes, so we need to store the axis for later and wait for
-            // a second SDL event. The axes also must be from the same joystick.
-            int axis = event.jaxis.axis;
-            if (analog_xaxis == -1) {
-                analog_xaxis = axis;
-                analog_axes_joystick = event.jaxis.which;
-            } else if (analog_yaxis == -1 && analog_xaxis != axis &&
-                       analog_axes_joystick == event.jaxis.which) {
-                analog_yaxis = axis;
-            }
-        }
-        Common::ParamPackage params;
-        if (analog_xaxis != -1 && analog_yaxis != -1) {
-            auto joystick = GetSDLJoystickBySDLID(event.jaxis.which);
-            params.Set("engine", "sdl");
-            params.Set("port", joystick->GetPort());
-            params.Set("guid", joystick->GetGUID());
-            params.Set("axis_x", analog_xaxis);
-            params.Set("axis_y", analog_yaxis);
-            analog_xaxis = -1;
-            analog_yaxis = -1;
-            analog_axes_joystick = -1;
-            return params;
-        }
-        return params;
-    }
-
-private:
-    int analog_xaxis = -1;
-    int analog_yaxis = -1;
-    SDL_JoystickID analog_axes_joystick = -1;
-};
-
-std::vector<std::unique_ptr<InputCommon::Polling::DevicePoller>> GetPollers(
-    InputCommon::Polling::DeviceType type) {
-    std::vector<std::unique_ptr<InputCommon::Polling::DevicePoller>> pollers;
-    switch (type) {
-    case InputCommon::Polling::DeviceType::Analog:
-        pollers.push_back(std::make_unique<SDLAnalogPoller>());
-        break;
-    case InputCommon::Polling::DeviceType::Button:
-        pollers.push_back(std::make_unique<SDLButtonPoller>());
-        break;
-    }
-    return pollers;
+std::unique_ptr<State> Init() {
+#ifdef HAVE_SDL2
+    return std::make_unique<SDLState>();
+#else
+    return std::make_unique<NullState>();
+#endif
 }
-} // namespace Polling
-} // namespace SDL
-} // namespace InputCommon
+} // namespace InputCommon::SDL
diff --git a/src/input_common/sdl/sdl.h b/src/input_common/sdl/sdl.h
index 0206860d3..02a8d2e2c 100644
--- a/src/input_common/sdl/sdl.h
+++ b/src/input_common/sdl/sdl.h
@@ -1,4 +1,4 @@
-// Copyright 2017 Citra Emulator Project
+// Copyright 2018 Citra Emulator Project
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
@@ -7,45 +7,36 @@
 #include <memory>
 #include <vector>
 #include "core/frontend/input.h"
+#include "input_common/main.h"
 
 union SDL_Event;
+
 namespace Common {
 class ParamPackage;
-}
-namespace InputCommon {
-namespace Polling {
+} // namespace Common
+
+namespace InputCommon::Polling {
 class DevicePoller;
 enum class DeviceType;
-} // namespace Polling
-} // namespace InputCommon
-
-namespace InputCommon {
-namespace SDL {
-
-/// Initializes and registers SDL device factories
-void Init();
-
-/// Unresisters SDL device factories and shut them down.
-void Shutdown();
-
-/// Needs to be called before SDL_QuitSubSystem.
-void CloseSDLJoysticks();
+} // namespace InputCommon::Polling
 
-/// Handle SDL_Events for joysticks from SDL_PollEvent
-void HandleGameControllerEvent(const SDL_Event& event);
+namespace InputCommon::SDL {
 
-/// A Loop that calls HandleGameControllerEvent until Shutdown is called
-void PollLoop();
+class State {
+public:
+    /// Unresisters SDL device factories and shut them down.
+    virtual ~State() = default;
 
-/// Creates a ParamPackage from an SDL_Event that can directly be used to create a ButtonDevice
-Common::ParamPackage SDLEventToButtonParamPackage(const SDL_Event& event);
+    virtual std::vector<std::unique_ptr<InputCommon::Polling::DevicePoller>> GetPollers(
+        InputCommon::Polling::DeviceType type) = 0;
+};
 
-namespace Polling {
+class NullState : public State {
+public:
+    std::vector<std::unique_ptr<InputCommon::Polling::DevicePoller>> GetPollers(
+        InputCommon::Polling::DeviceType type) override {}
+};
 
-/// Get all DevicePoller that use the SDL backend for a specific device type
-std::vector<std::unique_ptr<InputCommon::Polling::DevicePoller>> GetPollers(
-    InputCommon::Polling::DeviceType type);
+std::unique_ptr<State> Init();
 
-} // namespace Polling
-} // namespace SDL
-} // namespace InputCommon
+} // namespace InputCommon::SDL
diff --git a/src/input_common/sdl/sdl_impl.cpp b/src/input_common/sdl/sdl_impl.cpp
new file mode 100644
index 000000000..6e8376549
--- /dev/null
+++ b/src/input_common/sdl/sdl_impl.cpp
@@ -0,0 +1,668 @@
+// Copyright 2018 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <algorithm>
+#include <atomic>
+#include <cmath>
+#include <functional>
+#include <iterator>
+#include <mutex>
+#include <string>
+#include <thread>
+#include <tuple>
+#include <unordered_map>
+#include <utility>
+#include <vector>
+#include <SDL.h>
+#include "common/assert.h"
+#include "common/logging/log.h"
+#include "common/math_util.h"
+#include "common/param_package.h"
+#include "common/threadsafe_queue.h"
+#include "core/frontend/input.h"
+#include "input_common/sdl/sdl_impl.h"
+
+namespace InputCommon {
+
+namespace SDL {
+
+static std::string GetGUID(SDL_Joystick* joystick) {
+    SDL_JoystickGUID guid = SDL_JoystickGetGUID(joystick);
+    char guid_str[33];
+    SDL_JoystickGetGUIDString(guid, guid_str, sizeof(guid_str));
+    return guid_str;
+}
+
+/// Creates a ParamPackage from an SDL_Event that can directly be used to create a ButtonDevice
+static Common::ParamPackage SDLEventToButtonParamPackage(SDLState& state, const SDL_Event& event);
+
+static int SDLEventWatcher(void* userdata, SDL_Event* event) {
+    SDLState* sdl_state = reinterpret_cast<SDLState*>(userdata);
+    // Don't handle the event if we are configuring
+    if (sdl_state->polling) {
+        sdl_state->event_queue.Push(*event);
+    } else {
+        sdl_state->HandleGameControllerEvent(*event);
+    }
+    return 0;
+}
+
+class SDLJoystick {
+public:
+    SDLJoystick(std::string guid_, int port_, SDL_Joystick* joystick,
+                decltype(&SDL_JoystickClose) deleter = &SDL_JoystickClose)
+        : guid{std::move(guid_)}, port{port_}, sdl_joystick{joystick, deleter} {}
+
+    void SetButton(int button, bool value) {
+        std::lock_guard<std::mutex> lock(mutex);
+        state.buttons[button] = value;
+    }
+
+    bool GetButton(int button) const {
+        std::lock_guard<std::mutex> lock(mutex);
+        return state.buttons.at(button);
+    }
+
+    void SetAxis(int axis, Sint16 value) {
+        std::lock_guard<std::mutex> lock(mutex);
+        state.axes[axis] = value;
+    }
+
+    float GetAxis(int axis) const {
+        std::lock_guard<std::mutex> lock(mutex);
+        return state.axes.at(axis) / 32767.0f;
+    }
+
+    std::tuple<float, float> GetAnalog(int axis_x, int axis_y) const {
+        float x = GetAxis(axis_x);
+        float y = GetAxis(axis_y);
+        y = -y; // 3DS uses an y-axis inverse from SDL
+
+        // Make sure the coordinates are in the unit circle,
+        // otherwise normalize it.
+        float r = x * x + y * y;
+        if (r > 1.0f) {
+            r = std::sqrt(r);
+            x /= r;
+            y /= r;
+        }
+
+        return std::make_tuple(x, y);
+    }
+
+    void SetHat(int hat, Uint8 direction) {
+        std::lock_guard<std::mutex> lock(mutex);
+        state.hats[hat] = direction;
+    }
+
+    bool GetHatDirection(int hat, Uint8 direction) const {
+        std::lock_guard<std::mutex> lock(mutex);
+        return (state.hats.at(hat) & direction) != 0;
+    }
+    /**
+     * The guid of the joystick
+     */
+    const std::string& GetGUID() const {
+        return guid;
+    }
+
+    /**
+     * The number of joystick from the same type that were connected before this joystick
+     */
+    int GetPort() const {
+        return port;
+    }
+
+    SDL_Joystick* GetSDLJoystick() const {
+        return sdl_joystick.get();
+    }
+
+    void SetSDLJoystick(SDL_Joystick* joystick,
+                        decltype(&SDL_JoystickClose) deleter = &SDL_JoystickClose) {
+        sdl_joystick =
+            std::unique_ptr<SDL_Joystick, decltype(&SDL_JoystickClose)>(joystick, deleter);
+    }
+
+private:
+    struct State {
+        std::unordered_map<int, bool> buttons;
+        std::unordered_map<int, Sint16> axes;
+        std::unordered_map<int, Uint8> hats;
+    } state;
+    std::string guid;
+    int port;
+    std::unique_ptr<SDL_Joystick, decltype(&SDL_JoystickClose)> sdl_joystick;
+    mutable std::mutex mutex;
+};
+
+/**
+ * Get the nth joystick with the corresponding GUID
+ */
+std::shared_ptr<SDLJoystick> SDLState::GetSDLJoystickByGUID(const std::string& guid, int port) {
+    std::lock_guard<std::mutex> lock(joystick_map_mutex);
+    const auto it = joystick_map.find(guid);
+    if (it != joystick_map.end()) {
+        while (it->second.size() <= port) {
+            auto joystick = std::make_shared<SDLJoystick>(guid, it->second.size(), nullptr,
+                                                          [](SDL_Joystick*) {});
+            it->second.emplace_back(std::move(joystick));
+        }
+        return it->second[port];
+    }
+    auto joystick = std::make_shared<SDLJoystick>(guid, 0, nullptr, [](SDL_Joystick*) {});
+    return joystick_map[guid].emplace_back(std::move(joystick));
+}
+
+/**
+ * Check how many identical joysticks (by guid) were connected before the one with sdl_id and so tie
+ * it to a SDLJoystick with the same guid and that port
+ */
+std::shared_ptr<SDLJoystick> SDLState::GetSDLJoystickBySDLID(SDL_JoystickID sdl_id) {
+    auto sdl_joystick = SDL_JoystickFromInstanceID(sdl_id);
+    const std::string guid = GetGUID(sdl_joystick);
+    std::lock_guard<std::mutex> lock(joystick_map_mutex);
+    auto map_it = joystick_map.find(guid);
+    if (map_it != joystick_map.end()) {
+        auto vec_it = std::find_if(map_it->second.begin(), map_it->second.end(),
+                                   [&sdl_joystick](const std::shared_ptr<SDLJoystick>& joystick) {
+                                       return sdl_joystick == joystick->GetSDLJoystick();
+                                   });
+        if (vec_it != map_it->second.end()) {
+            // This is the common case: There is already an existing SDL_Joystick maped to a
+            // SDLJoystick. return the SDLJoystick
+            return *vec_it;
+        }
+        // Search for a SDLJoystick without a mapped SDL_Joystick...
+        auto nullptr_it = std::find_if(map_it->second.begin(), map_it->second.end(),
+                                       [](const std::shared_ptr<SDLJoystick>& joystick) {
+                                           return !joystick->GetSDLJoystick();
+                                       });
+        if (nullptr_it != map_it->second.end()) {
+            // ... and map it
+            (*nullptr_it)->SetSDLJoystick(sdl_joystick);
+            return *nullptr_it;
+        }
+        // There is no SDLJoystick without a mapped SDL_Joystick
+        // Create a new SDLJoystick
+        auto joystick = std::make_shared<SDLJoystick>(guid, map_it->second.size(), sdl_joystick);
+        return map_it->second.emplace_back(std::move(joystick));
+    }
+    auto joystick = std::make_shared<SDLJoystick>(guid, 0, sdl_joystick);
+    return joystick_map[guid].emplace_back(std::move(joystick));
+}
+
+void SDLState::InitJoystick(int joystick_index) {
+    SDL_Joystick* sdl_joystick = SDL_JoystickOpen(joystick_index);
+    if (!sdl_joystick) {
+        LOG_ERROR(Input, "failed to open joystick {}", joystick_index);
+        return;
+    }
+    std::string guid = GetGUID(sdl_joystick);
+    std::lock_guard<std::mutex> lock(joystick_map_mutex);
+    if (joystick_map.find(guid) == joystick_map.end()) {
+        auto joystick = std::make_shared<SDLJoystick>(guid, 0, sdl_joystick);
+        joystick_map[guid].emplace_back(std::move(joystick));
+        return;
+    }
+    auto& joystick_guid_list = joystick_map[guid];
+    const auto it = std::find_if(
+        joystick_guid_list.begin(), joystick_guid_list.end(),
+        [](const std::shared_ptr<SDLJoystick>& joystick) { return !joystick->GetSDLJoystick(); });
+    if (it != joystick_guid_list.end()) {
+        (*it)->SetSDLJoystick(sdl_joystick);
+        return;
+    }
+    auto joystick = std::make_shared<SDLJoystick>(guid, joystick_guid_list.size(), sdl_joystick);
+    joystick_guid_list.emplace_back(std::move(joystick));
+}
+
+void SDLState::CloseJoystick(SDL_Joystick* sdl_joystick) {
+    std::string guid = GetGUID(sdl_joystick);
+    std::shared_ptr<SDLJoystick> joystick;
+    {
+        std::lock_guard<std::mutex> lock(joystick_map_mutex);
+        // This call to guid is safe since the joystick is guaranteed to be in the map
+        auto& joystick_guid_list = joystick_map[guid];
+        const auto joystick_it =
+            std::find_if(joystick_guid_list.begin(), joystick_guid_list.end(),
+                         [&sdl_joystick](const std::shared_ptr<SDLJoystick>& joystick) {
+                             return joystick->GetSDLJoystick() == sdl_joystick;
+                         });
+        joystick = *joystick_it;
+    }
+    // Destruct SDL_Joystick outside the lock guard because SDL can internally call event calback
+    // which locks the mutex again
+    joystick->SetSDLJoystick(nullptr, [](SDL_Joystick*) {});
+}
+
+void SDLState::HandleGameControllerEvent(const SDL_Event& event) {
+    switch (event.type) {
+    case SDL_JOYBUTTONUP: {
+        if (auto joystick = GetSDLJoystickBySDLID(event.jbutton.which)) {
+            joystick->SetButton(event.jbutton.button, false);
+        }
+        break;
+    }
+    case SDL_JOYBUTTONDOWN: {
+        if (auto joystick = GetSDLJoystickBySDLID(event.jbutton.which)) {
+            joystick->SetButton(event.jbutton.button, true);
+        }
+        break;
+    }
+    case SDL_JOYHATMOTION: {
+        if (auto joystick = GetSDLJoystickBySDLID(event.jhat.which)) {
+            joystick->SetHat(event.jhat.hat, event.jhat.value);
+        }
+        break;
+    }
+    case SDL_JOYAXISMOTION: {
+        if (auto joystick = GetSDLJoystickBySDLID(event.jaxis.which)) {
+            joystick->SetAxis(event.jaxis.axis, event.jaxis.value);
+        }
+        break;
+    }
+    case SDL_JOYDEVICEREMOVED:
+        LOG_DEBUG(Input, "Controller removed with Instance_ID {}", event.jdevice.which);
+        CloseJoystick(SDL_JoystickFromInstanceID(event.jdevice.which));
+        break;
+    case SDL_JOYDEVICEADDED:
+        LOG_DEBUG(Input, "Controller connected with device index {}", event.jdevice.which);
+        InitJoystick(event.jdevice.which);
+        break;
+    }
+}
+
+void SDLState::CloseJoysticks() {
+    std::lock_guard<std::mutex> lock(joystick_map_mutex);
+    joystick_map.clear();
+}
+
+class SDLButton final : public Input::ButtonDevice {
+public:
+    explicit SDLButton(std::shared_ptr<SDLJoystick> joystick_, int button_)
+        : joystick(std::move(joystick_)), button(button_) {}
+
+    bool GetStatus() const override {
+        return joystick->GetButton(button);
+    }
+
+private:
+    std::shared_ptr<SDLJoystick> joystick;
+    int button;
+};
+
+class SDLDirectionButton final : public Input::ButtonDevice {
+public:
+    explicit SDLDirectionButton(std::shared_ptr<SDLJoystick> joystick_, int hat_, Uint8 direction_)
+        : joystick(std::move(joystick_)), hat(hat_), direction(direction_) {}
+
+    bool GetStatus() const override {
+        return joystick->GetHatDirection(hat, direction);
+    }
+
+private:
+    std::shared_ptr<SDLJoystick> joystick;
+    int hat;
+    Uint8 direction;
+};
+
+class SDLAxisButton final : public Input::ButtonDevice {
+public:
+    explicit SDLAxisButton(std::shared_ptr<SDLJoystick> joystick_, int axis_, float threshold_,
+                           bool trigger_if_greater_)
+        : joystick(std::move(joystick_)), axis(axis_), threshold(threshold_),
+          trigger_if_greater(trigger_if_greater_) {}
+
+    bool GetStatus() const override {
+        float axis_value = joystick->GetAxis(axis);
+        if (trigger_if_greater)
+            return axis_value > threshold;
+        return axis_value < threshold;
+    }
+
+private:
+    std::shared_ptr<SDLJoystick> joystick;
+    int axis;
+    float threshold;
+    bool trigger_if_greater;
+};
+
+class SDLAnalog final : public Input::AnalogDevice {
+public:
+    SDLAnalog(std::shared_ptr<SDLJoystick> joystick_, int axis_x_, int axis_y_, float deadzone_)
+        : joystick(std::move(joystick_)), axis_x(axis_x_), axis_y(axis_y_), deadzone(deadzone_) {}
+
+    std::tuple<float, float> GetStatus() const override {
+        const auto [x, y] = joystick->GetAnalog(axis_x, axis_y);
+        const float r = std::sqrt((x * x) + (y * y));
+        if (r > deadzone) {
+            return std::make_tuple(x / r * (r - deadzone) / (1 - deadzone),
+                                   y / r * (r - deadzone) / (1 - deadzone));
+        }
+        return std::make_tuple<float, float>(0.0f, 0.0f);
+    }
+
+private:
+    std::shared_ptr<SDLJoystick> joystick;
+    const int axis_x;
+    const int axis_y;
+    const float deadzone;
+};
+
+/// A button device factory that creates button devices from SDL joystick
+class SDLButtonFactory final : public Input::Factory<Input::ButtonDevice> {
+public:
+    explicit SDLButtonFactory(SDLState& state_) : state(state_) {}
+
+    /**
+     * Creates a button device from a joystick button
+     * @param params contains parameters for creating the device:
+     *     - "guid": the guid of the joystick to bind
+     *     - "port": the nth joystick of the same type to bind
+     *     - "button"(optional): the index of the button to bind
+     *     - "hat"(optional): the index of the hat to bind as direction buttons
+     *     - "axis"(optional): the index of the axis to bind
+     *     - "direction"(only used for hat): the direction name of the hat to bind. Can be "up",
+     *         "down", "left" or "right"
+     *     - "threshold"(only used for axis): a float value in (-1.0, 1.0) which the button is
+     *         triggered if the axis value crosses
+     *     - "direction"(only used for axis): "+" means the button is triggered when the axis
+     * value is greater than the threshold; "-" means the button is triggered when the axis
+     * value is smaller than the threshold
+     */
+    std::unique_ptr<Input::ButtonDevice> Create(const Common::ParamPackage& params) override {
+        const std::string guid = params.Get("guid", "0");
+        const int port = params.Get("port", 0);
+
+        auto joystick = state.GetSDLJoystickByGUID(guid, port);
+
+        if (params.Has("hat")) {
+            const int hat = params.Get("hat", 0);
+            const std::string direction_name = params.Get("direction", "");
+            Uint8 direction;
+            if (direction_name == "up") {
+                direction = SDL_HAT_UP;
+            } else if (direction_name == "down") {
+                direction = SDL_HAT_DOWN;
+            } else if (direction_name == "left") {
+                direction = SDL_HAT_LEFT;
+            } else if (direction_name == "right") {
+                direction = SDL_HAT_RIGHT;
+            } else {
+                direction = 0;
+            }
+            // This is necessary so accessing GetHat with hat won't crash
+            joystick->SetHat(hat, SDL_HAT_CENTERED);
+            return std::make_unique<SDLDirectionButton>(joystick, hat, direction);
+        }
+
+        if (params.Has("axis")) {
+            const int axis = params.Get("axis", 0);
+            const float threshold = params.Get("threshold", 0.5f);
+            const std::string direction_name = params.Get("direction", "");
+            bool trigger_if_greater;
+            if (direction_name == "+") {
+                trigger_if_greater = true;
+            } else if (direction_name == "-") {
+                trigger_if_greater = false;
+            } else {
+                trigger_if_greater = true;
+                LOG_ERROR(Input, "Unknown direction {}", direction_name);
+            }
+            // This is necessary so accessing GetAxis with axis won't crash
+            joystick->SetAxis(axis, 0);
+            return std::make_unique<SDLAxisButton>(joystick, axis, threshold, trigger_if_greater);
+        }
+
+        const int button = params.Get("button", 0);
+        // This is necessary so accessing GetButton with button won't crash
+        joystick->SetButton(button, false);
+        return std::make_unique<SDLButton>(joystick, button);
+    }
+
+private:
+    SDLState& state;
+};
+
+/// An analog device factory that creates analog devices from SDL joystick
+class SDLAnalogFactory final : public Input::Factory<Input::AnalogDevice> {
+public:
+    explicit SDLAnalogFactory(SDLState& state_) : state(state_) {}
+    /**
+     * Creates analog device from joystick axes
+     * @param params contains parameters for creating the device:
+     *     - "guid": the guid of the joystick to bind
+     *     - "port": the nth joystick of the same type
+     *     - "axis_x": the index of the axis to be bind as x-axis
+     *     - "axis_y": the index of the axis to be bind as y-axis
+     */
+    std::unique_ptr<Input::AnalogDevice> Create(const Common::ParamPackage& params) override {
+        const std::string guid = params.Get("guid", "0");
+        const int port = params.Get("port", 0);
+        const int axis_x = params.Get("axis_x", 0);
+        const int axis_y = params.Get("axis_y", 1);
+        float deadzone = std::clamp(params.Get("deadzone", 0.0f), 0.0f, .99f);
+
+        auto joystick = state.GetSDLJoystickByGUID(guid, port);
+
+        // This is necessary so accessing GetAxis with axis_x and axis_y won't crash
+        joystick->SetAxis(axis_x, 0);
+        joystick->SetAxis(axis_y, 0);
+        return std::make_unique<SDLAnalog>(joystick, axis_x, axis_y, deadzone);
+    }
+
+private:
+    SDLState& state;
+};
+
+SDLState::SDLState() {
+    using namespace Input;
+    RegisterFactory<ButtonDevice>("sdl", std::make_shared<SDLButtonFactory>(*this));
+    RegisterFactory<AnalogDevice>("sdl", std::make_shared<SDLAnalogFactory>(*this));
+
+    // If the frontend is going to manage the event loop, then we dont start one here
+    start_thread = !SDL_WasInit(SDL_INIT_JOYSTICK);
+    if (start_thread && SDL_Init(SDL_INIT_JOYSTICK) < 0) {
+        LOG_CRITICAL(Input, "SDL_Init(SDL_INIT_JOYSTICK) failed with: {}", SDL_GetError());
+        return;
+    }
+    if (SDL_SetHint(SDL_HINT_JOYSTICK_ALLOW_BACKGROUND_EVENTS, "1") == SDL_FALSE) {
+        LOG_ERROR(Input, "Failed to set Hint for background events", SDL_GetError());
+    }
+
+    SDL_AddEventWatch(&SDLEventWatcher, this);
+
+    initialized = true;
+    if (start_thread) {
+        poll_thread = std::thread([this] {
+            using namespace std::chrono_literals;
+            while (initialized) {
+                SDL_PumpEvents();
+                std::this_thread::sleep_for(10ms);
+            }
+        });
+    }
+    // Because the events for joystick connection happens before we have our event watcher added, we
+    // can just open all the joysticks right here
+    for (int i = 0; i < SDL_NumJoysticks(); ++i) {
+        InitJoystick(i);
+    }
+}
+
+SDLState::~SDLState() {
+    using namespace Input;
+    UnregisterFactory<ButtonDevice>("sdl");
+    UnregisterFactory<AnalogDevice>("sdl");
+
+    CloseJoysticks();
+    SDL_DelEventWatch(&SDLEventWatcher, this);
+
+    initialized = false;
+    if (start_thread) {
+        poll_thread.join();
+        SDL_QuitSubSystem(SDL_INIT_JOYSTICK);
+    }
+}
+
+Common::ParamPackage SDLEventToButtonParamPackage(SDLState& state, const SDL_Event& event) {
+    Common::ParamPackage params({{"engine", "sdl"}});
+
+    switch (event.type) {
+    case SDL_JOYAXISMOTION: {
+        auto joystick = state.GetSDLJoystickBySDLID(event.jaxis.which);
+        params.Set("port", joystick->GetPort());
+        params.Set("guid", joystick->GetGUID());
+        params.Set("axis", event.jaxis.axis);
+        if (event.jaxis.value > 0) {
+            params.Set("direction", "+");
+            params.Set("threshold", "0.5");
+        } else {
+            params.Set("direction", "-");
+            params.Set("threshold", "-0.5");
+        }
+        break;
+    }
+    case SDL_JOYBUTTONUP: {
+        auto joystick = state.GetSDLJoystickBySDLID(event.jbutton.which);
+        params.Set("port", joystick->GetPort());
+        params.Set("guid", joystick->GetGUID());
+        params.Set("button", event.jbutton.button);
+        break;
+    }
+    case SDL_JOYHATMOTION: {
+        auto joystick = state.GetSDLJoystickBySDLID(event.jhat.which);
+        params.Set("port", joystick->GetPort());
+        params.Set("guid", joystick->GetGUID());
+        params.Set("hat", event.jhat.hat);
+        switch (event.jhat.value) {
+        case SDL_HAT_UP:
+            params.Set("direction", "up");
+            break;
+        case SDL_HAT_DOWN:
+            params.Set("direction", "down");
+            break;
+        case SDL_HAT_LEFT:
+            params.Set("direction", "left");
+            break;
+        case SDL_HAT_RIGHT:
+            params.Set("direction", "right");
+            break;
+        default:
+            return {};
+        }
+        break;
+    }
+    }
+    return params;
+}
+
+namespace Polling {
+
+class SDLPoller : public InputCommon::Polling::DevicePoller {
+public:
+    explicit SDLPoller(SDLState& state_) : state(state_) {}
+
+    void Start() override {
+        state.event_queue.Clear();
+        state.polling = true;
+    }
+
+    void Stop() override {
+        state.polling = false;
+    }
+
+protected:
+    SDLState& state;
+};
+
+class SDLButtonPoller final : public SDLPoller {
+public:
+    explicit SDLButtonPoller(SDLState& state_) : SDLPoller(state_) {}
+
+    Common::ParamPackage GetNextInput() override {
+        SDL_Event event;
+        while (state.event_queue.Pop(event)) {
+            switch (event.type) {
+            case SDL_JOYAXISMOTION:
+                if (std::abs(event.jaxis.value / 32767.0) < 0.5) {
+                    break;
+                }
+            case SDL_JOYBUTTONUP:
+            case SDL_JOYHATMOTION:
+                return SDLEventToButtonParamPackage(state, event);
+            }
+        }
+        return {};
+    }
+};
+
+class SDLAnalogPoller final : public SDLPoller {
+public:
+    explicit SDLAnalogPoller(SDLState& state_) : SDLPoller(state_) {}
+
+    void Start() override {
+        SDLPoller::Start();
+
+        // Reset stored axes
+        analog_xaxis = -1;
+        analog_yaxis = -1;
+        analog_axes_joystick = -1;
+    }
+
+    Common::ParamPackage GetNextInput() override {
+        SDL_Event event;
+        while (state.event_queue.Pop(event)) {
+            if (event.type != SDL_JOYAXISMOTION || std::abs(event.jaxis.value / 32767.0) < 0.5) {
+                continue;
+            }
+            // An analog device needs two axes, so we need to store the axis for later and wait for
+            // a second SDL event. The axes also must be from the same joystick.
+            int axis = event.jaxis.axis;
+            if (analog_xaxis == -1) {
+                analog_xaxis = axis;
+                analog_axes_joystick = event.jaxis.which;
+            } else if (analog_yaxis == -1 && analog_xaxis != axis &&
+                       analog_axes_joystick == event.jaxis.which) {
+                analog_yaxis = axis;
+            }
+        }
+        Common::ParamPackage params;
+        if (analog_xaxis != -1 && analog_yaxis != -1) {
+            auto joystick = state.GetSDLJoystickBySDLID(event.jaxis.which);
+            params.Set("engine", "sdl");
+            params.Set("port", joystick->GetPort());
+            params.Set("guid", joystick->GetGUID());
+            params.Set("axis_x", analog_xaxis);
+            params.Set("axis_y", analog_yaxis);
+            analog_xaxis = -1;
+            analog_yaxis = -1;
+            analog_axes_joystick = -1;
+            return params;
+        }
+        return params;
+    }
+
+private:
+    int analog_xaxis = -1;
+    int analog_yaxis = -1;
+    SDL_JoystickID analog_axes_joystick = -1;
+};
+} // namespace Polling
+
+std::vector<std::unique_ptr<InputCommon::Polling::DevicePoller>> SDLState::GetPollers(
+    InputCommon::Polling::DeviceType type) {
+    std::vector<std::unique_ptr<InputCommon::Polling::DevicePoller>> pollers;
+    switch (type) {
+    case InputCommon::Polling::DeviceType::Analog:
+        pollers.emplace_back(std::make_unique<Polling::SDLAnalogPoller>(*this));
+        break;
+    case InputCommon::Polling::DeviceType::Button:
+        pollers.emplace_back(std::make_unique<Polling::SDLButtonPoller>(*this));
+        break;
+        return pollers;
+    }
+}
+
+} // namespace SDL
+} // namespace InputCommon
diff --git a/src/input_common/sdl/sdl_impl.h b/src/input_common/sdl/sdl_impl.h
new file mode 100644
index 000000000..fec82fbe6
--- /dev/null
+++ b/src/input_common/sdl/sdl_impl.h
@@ -0,0 +1,64 @@
+// Copyright 2018 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <atomic>
+#include <memory>
+#include <thread>
+#include "common/threadsafe_queue.h"
+#include "input_common/sdl/sdl.h"
+
+union SDL_Event;
+using SDL_Joystick = struct _SDL_Joystick;
+using SDL_JoystickID = s32;
+
+namespace InputCommon::SDL {
+
+class SDLJoystick;
+class SDLButtonFactory;
+class SDLAnalogFactory;
+
+class SDLState : public State {
+public:
+    /// Initializes and registers SDL device factories
+    SDLState();
+
+    /// Unresisters SDL device factories and shut them down.
+    ~SDLState() override;
+
+    /// Handle SDL_Events for joysticks from SDL_PollEvent
+    void HandleGameControllerEvent(const SDL_Event& event);
+
+    std::shared_ptr<SDLJoystick> GetSDLJoystickBySDLID(SDL_JoystickID sdl_id);
+    std::shared_ptr<SDLJoystick> GetSDLJoystickByGUID(const std::string& guid, int port);
+
+    /// Get all DevicePoller that use the SDL backend for a specific device type
+    std::vector<std::unique_ptr<InputCommon::Polling::DevicePoller>> GetPollers(
+        InputCommon::Polling::DeviceType type) override;
+
+    /// Used by the Pollers during config
+    std::atomic<bool> polling = false;
+    Common::SPSCQueue<SDL_Event> event_queue;
+
+private:
+    void InitJoystick(int joystick_index);
+    void CloseJoystick(SDL_Joystick* sdl_joystick);
+
+    /// Needs to be called before SDL_QuitSubSystem.
+    void CloseJoysticks();
+
+    /// Map of GUID of a list of corresponding virtual Joysticks
+    std::unordered_map<std::string, std::vector<std::shared_ptr<SDLJoystick>>> joystick_map;
+    std::mutex joystick_map_mutex;
+
+    std::shared_ptr<SDLButtonFactory> button_factory;
+    std::shared_ptr<SDLAnalogFactory> analog_factory;
+
+    bool start_thread = false;
+    std::atomic<bool> initialized = false;
+
+    std::thread poll_thread;
+};
+} // namespace InputCommon::SDL
diff --git a/src/tests/core/arm/arm_test_common.cpp b/src/tests/core/arm/arm_test_common.cpp
index 9b8a44fa1..3e1a735c3 100644
--- a/src/tests/core/arm/arm_test_common.cpp
+++ b/src/tests/core/arm/arm_test_common.cpp
@@ -4,6 +4,7 @@
 
 #include <algorithm>
 
+#include "common/page_table.h"
 #include "core/core.h"
 #include "core/hle/kernel/process.h"
 #include "core/memory.h"
@@ -13,16 +14,16 @@
 namespace ArmTests {
 
 TestEnvironment::TestEnvironment(bool mutable_memory_)
-    : mutable_memory(mutable_memory_), test_memory(std::make_shared<TestMemory>(this)) {
-
-    auto process = Kernel::Process::Create(kernel, "");
+    : mutable_memory(mutable_memory_),
+      test_memory(std::make_shared<TestMemory>(this)), kernel{Core::System::GetInstance()} {
+    auto process = Kernel::Process::Create(Core::System::GetInstance(), "");
     kernel.MakeCurrentProcess(process.get());
-    page_table = &Core::CurrentProcess()->VMManager().page_table;
+    page_table = &process->VMManager().page_table;
 
     std::fill(page_table->pointers.begin(), page_table->pointers.end(), nullptr);
     page_table->special_regions.clear();
     std::fill(page_table->attributes.begin(), page_table->attributes.end(),
-              Memory::PageType::Unmapped);
+              Common::PageType::Unmapped);
 
     Memory::MapIoRegion(*page_table, 0x00000000, 0x80000000, test_memory);
     Memory::MapIoRegion(*page_table, 0x80000000, 0x80000000, test_memory);
diff --git a/src/tests/core/arm/arm_test_common.h b/src/tests/core/arm/arm_test_common.h
index 0b7539601..d145dbfcc 100644
--- a/src/tests/core/arm/arm_test_common.h
+++ b/src/tests/core/arm/arm_test_common.h
@@ -9,10 +9,10 @@
 #include <vector>
 
 #include "common/common_types.h"
+#include "common/memory_hook.h"
 #include "core/hle/kernel/kernel.h"
-#include "core/memory_hook.h"
 
-namespace Memory {
+namespace Common {
 struct PageTable;
 }
 
@@ -58,7 +58,7 @@ public:
 
 private:
     friend struct TestMemory;
-    struct TestMemory final : Memory::MemoryHook {
+    struct TestMemory final : Common::MemoryHook {
         explicit TestMemory(TestEnvironment* env_) : env(env_) {}
         TestEnvironment* env;
 
@@ -86,7 +86,7 @@ private:
     bool mutable_memory;
     std::shared_ptr<TestMemory> test_memory;
     std::vector<WriteRecord> write_records;
-    Memory::PageTable* page_table = nullptr;
+    Common::PageTable* page_table = nullptr;
     Kernel::KernelCore kernel;
 };
 
diff --git a/src/tests/core/core_timing.cpp b/src/tests/core/core_timing.cpp
index 2242c14cf..340d6a272 100644
--- a/src/tests/core/core_timing.cpp
+++ b/src/tests/core/core_timing.cpp
@@ -28,100 +28,103 @@ void CallbackTemplate(u64 userdata, s64 cycles_late) {
     REQUIRE(lateness == cycles_late);
 }
 
-class ScopeInit final {
-public:
+struct ScopeInit final {
     ScopeInit() {
-        CoreTiming::Init();
+        core_timing.Initialize();
     }
     ~ScopeInit() {
-        CoreTiming::Shutdown();
+        core_timing.Shutdown();
     }
+
+    Core::Timing::CoreTiming core_timing;
 };
 
-static void AdvanceAndCheck(u32 idx, int downcount, int expected_lateness = 0,
-                            int cpu_downcount = 0) {
+static void AdvanceAndCheck(Core::Timing::CoreTiming& core_timing, u32 idx, int downcount,
+                            int expected_lateness = 0, int cpu_downcount = 0) {
     callbacks_ran_flags = 0;
     expected_callback = CB_IDS[idx];
     lateness = expected_lateness;
 
-    CoreTiming::AddTicks(CoreTiming::GetDowncount() -
-                         cpu_downcount); // Pretend we executed X cycles of instructions.
-    CoreTiming::Advance();
+    // Pretend we executed X cycles of instructions.
+    core_timing.AddTicks(core_timing.GetDowncount() - cpu_downcount);
+    core_timing.Advance();
 
     REQUIRE(decltype(callbacks_ran_flags)().set(idx) == callbacks_ran_flags);
-    REQUIRE(downcount == CoreTiming::GetDowncount());
+    REQUIRE(downcount == core_timing.GetDowncount());
 }
 
 TEST_CASE("CoreTiming[BasicOrder]", "[core]") {
     ScopeInit guard;
+    auto& core_timing = guard.core_timing;
 
-    CoreTiming::EventType* cb_a = CoreTiming::RegisterEvent("callbackA", CallbackTemplate<0>);
-    CoreTiming::EventType* cb_b = CoreTiming::RegisterEvent("callbackB", CallbackTemplate<1>);
-    CoreTiming::EventType* cb_c = CoreTiming::RegisterEvent("callbackC", CallbackTemplate<2>);
-    CoreTiming::EventType* cb_d = CoreTiming::RegisterEvent("callbackD", CallbackTemplate<3>);
-    CoreTiming::EventType* cb_e = CoreTiming::RegisterEvent("callbackE", CallbackTemplate<4>);
+    Core::Timing::EventType* cb_a = core_timing.RegisterEvent("callbackA", CallbackTemplate<0>);
+    Core::Timing::EventType* cb_b = core_timing.RegisterEvent("callbackB", CallbackTemplate<1>);
+    Core::Timing::EventType* cb_c = core_timing.RegisterEvent("callbackC", CallbackTemplate<2>);
+    Core::Timing::EventType* cb_d = core_timing.RegisterEvent("callbackD", CallbackTemplate<3>);
+    Core::Timing::EventType* cb_e = core_timing.RegisterEvent("callbackE", CallbackTemplate<4>);
 
     // Enter slice 0
-    CoreTiming::Advance();
+    core_timing.Advance();
 
     // D -> B -> C -> A -> E
-    CoreTiming::ScheduleEvent(1000, cb_a, CB_IDS[0]);
-    REQUIRE(1000 == CoreTiming::GetDowncount());
-    CoreTiming::ScheduleEvent(500, cb_b, CB_IDS[1]);
-    REQUIRE(500 == CoreTiming::GetDowncount());
-    CoreTiming::ScheduleEvent(800, cb_c, CB_IDS[2]);
-    REQUIRE(500 == CoreTiming::GetDowncount());
-    CoreTiming::ScheduleEvent(100, cb_d, CB_IDS[3]);
-    REQUIRE(100 == CoreTiming::GetDowncount());
-    CoreTiming::ScheduleEvent(1200, cb_e, CB_IDS[4]);
-    REQUIRE(100 == CoreTiming::GetDowncount());
-
-    AdvanceAndCheck(3, 400);
-    AdvanceAndCheck(1, 300);
-    AdvanceAndCheck(2, 200);
-    AdvanceAndCheck(0, 200);
-    AdvanceAndCheck(4, MAX_SLICE_LENGTH);
+    core_timing.ScheduleEvent(1000, cb_a, CB_IDS[0]);
+    REQUIRE(1000 == core_timing.GetDowncount());
+    core_timing.ScheduleEvent(500, cb_b, CB_IDS[1]);
+    REQUIRE(500 == core_timing.GetDowncount());
+    core_timing.ScheduleEvent(800, cb_c, CB_IDS[2]);
+    REQUIRE(500 == core_timing.GetDowncount());
+    core_timing.ScheduleEvent(100, cb_d, CB_IDS[3]);
+    REQUIRE(100 == core_timing.GetDowncount());
+    core_timing.ScheduleEvent(1200, cb_e, CB_IDS[4]);
+    REQUIRE(100 == core_timing.GetDowncount());
+
+    AdvanceAndCheck(core_timing, 3, 400);
+    AdvanceAndCheck(core_timing, 1, 300);
+    AdvanceAndCheck(core_timing, 2, 200);
+    AdvanceAndCheck(core_timing, 0, 200);
+    AdvanceAndCheck(core_timing, 4, MAX_SLICE_LENGTH);
 }
 
 TEST_CASE("CoreTiming[Threadsave]", "[core]") {
     ScopeInit guard;
+    auto& core_timing = guard.core_timing;
 
-    CoreTiming::EventType* cb_a = CoreTiming::RegisterEvent("callbackA", CallbackTemplate<0>);
-    CoreTiming::EventType* cb_b = CoreTiming::RegisterEvent("callbackB", CallbackTemplate<1>);
-    CoreTiming::EventType* cb_c = CoreTiming::RegisterEvent("callbackC", CallbackTemplate<2>);
-    CoreTiming::EventType* cb_d = CoreTiming::RegisterEvent("callbackD", CallbackTemplate<3>);
-    CoreTiming::EventType* cb_e = CoreTiming::RegisterEvent("callbackE", CallbackTemplate<4>);
+    Core::Timing::EventType* cb_a = core_timing.RegisterEvent("callbackA", CallbackTemplate<0>);
+    Core::Timing::EventType* cb_b = core_timing.RegisterEvent("callbackB", CallbackTemplate<1>);
+    Core::Timing::EventType* cb_c = core_timing.RegisterEvent("callbackC", CallbackTemplate<2>);
+    Core::Timing::EventType* cb_d = core_timing.RegisterEvent("callbackD", CallbackTemplate<3>);
+    Core::Timing::EventType* cb_e = core_timing.RegisterEvent("callbackE", CallbackTemplate<4>);
 
     // Enter slice 0
-    CoreTiming::Advance();
+    core_timing.Advance();
 
     // D -> B -> C -> A -> E
-    CoreTiming::ScheduleEventThreadsafe(1000, cb_a, CB_IDS[0]);
+    core_timing.ScheduleEventThreadsafe(1000, cb_a, CB_IDS[0]);
     // Manually force since ScheduleEventThreadsafe doesn't call it
-    CoreTiming::ForceExceptionCheck(1000);
-    REQUIRE(1000 == CoreTiming::GetDowncount());
-    CoreTiming::ScheduleEventThreadsafe(500, cb_b, CB_IDS[1]);
+    core_timing.ForceExceptionCheck(1000);
+    REQUIRE(1000 == core_timing.GetDowncount());
+    core_timing.ScheduleEventThreadsafe(500, cb_b, CB_IDS[1]);
     // Manually force since ScheduleEventThreadsafe doesn't call it
-    CoreTiming::ForceExceptionCheck(500);
-    REQUIRE(500 == CoreTiming::GetDowncount());
-    CoreTiming::ScheduleEventThreadsafe(800, cb_c, CB_IDS[2]);
+    core_timing.ForceExceptionCheck(500);
+    REQUIRE(500 == core_timing.GetDowncount());
+    core_timing.ScheduleEventThreadsafe(800, cb_c, CB_IDS[2]);
     // Manually force since ScheduleEventThreadsafe doesn't call it
-    CoreTiming::ForceExceptionCheck(800);
-    REQUIRE(500 == CoreTiming::GetDowncount());
-    CoreTiming::ScheduleEventThreadsafe(100, cb_d, CB_IDS[3]);
+    core_timing.ForceExceptionCheck(800);
+    REQUIRE(500 == core_timing.GetDowncount());
+    core_timing.ScheduleEventThreadsafe(100, cb_d, CB_IDS[3]);
     // Manually force since ScheduleEventThreadsafe doesn't call it
-    CoreTiming::ForceExceptionCheck(100);
-    REQUIRE(100 == CoreTiming::GetDowncount());
-    CoreTiming::ScheduleEventThreadsafe(1200, cb_e, CB_IDS[4]);
+    core_timing.ForceExceptionCheck(100);
+    REQUIRE(100 == core_timing.GetDowncount());
+    core_timing.ScheduleEventThreadsafe(1200, cb_e, CB_IDS[4]);
     // Manually force since ScheduleEventThreadsafe doesn't call it
-    CoreTiming::ForceExceptionCheck(1200);
-    REQUIRE(100 == CoreTiming::GetDowncount());
-
-    AdvanceAndCheck(3, 400);
-    AdvanceAndCheck(1, 300);
-    AdvanceAndCheck(2, 200);
-    AdvanceAndCheck(0, 200);
-    AdvanceAndCheck(4, MAX_SLICE_LENGTH);
+    core_timing.ForceExceptionCheck(1200);
+    REQUIRE(100 == core_timing.GetDowncount());
+
+    AdvanceAndCheck(core_timing, 3, 400);
+    AdvanceAndCheck(core_timing, 1, 300);
+    AdvanceAndCheck(core_timing, 2, 200);
+    AdvanceAndCheck(core_timing, 0, 200);
+    AdvanceAndCheck(core_timing, 4, MAX_SLICE_LENGTH);
 }
 
 namespace SharedSlotTest {
@@ -142,59 +145,63 @@ TEST_CASE("CoreTiming[SharedSlot]", "[core]") {
     using namespace SharedSlotTest;
 
     ScopeInit guard;
+    auto& core_timing = guard.core_timing;
 
-    CoreTiming::EventType* cb_a = CoreTiming::RegisterEvent("callbackA", FifoCallback<0>);
-    CoreTiming::EventType* cb_b = CoreTiming::RegisterEvent("callbackB", FifoCallback<1>);
-    CoreTiming::EventType* cb_c = CoreTiming::RegisterEvent("callbackC", FifoCallback<2>);
-    CoreTiming::EventType* cb_d = CoreTiming::RegisterEvent("callbackD", FifoCallback<3>);
-    CoreTiming::EventType* cb_e = CoreTiming::RegisterEvent("callbackE", FifoCallback<4>);
+    Core::Timing::EventType* cb_a = core_timing.RegisterEvent("callbackA", FifoCallback<0>);
+    Core::Timing::EventType* cb_b = core_timing.RegisterEvent("callbackB", FifoCallback<1>);
+    Core::Timing::EventType* cb_c = core_timing.RegisterEvent("callbackC", FifoCallback<2>);
+    Core::Timing::EventType* cb_d = core_timing.RegisterEvent("callbackD", FifoCallback<3>);
+    Core::Timing::EventType* cb_e = core_timing.RegisterEvent("callbackE", FifoCallback<4>);
 
-    CoreTiming::ScheduleEvent(1000, cb_a, CB_IDS[0]);
-    CoreTiming::ScheduleEvent(1000, cb_b, CB_IDS[1]);
-    CoreTiming::ScheduleEvent(1000, cb_c, CB_IDS[2]);
-    CoreTiming::ScheduleEvent(1000, cb_d, CB_IDS[3]);
-    CoreTiming::ScheduleEvent(1000, cb_e, CB_IDS[4]);
+    core_timing.ScheduleEvent(1000, cb_a, CB_IDS[0]);
+    core_timing.ScheduleEvent(1000, cb_b, CB_IDS[1]);
+    core_timing.ScheduleEvent(1000, cb_c, CB_IDS[2]);
+    core_timing.ScheduleEvent(1000, cb_d, CB_IDS[3]);
+    core_timing.ScheduleEvent(1000, cb_e, CB_IDS[4]);
 
     // Enter slice 0
-    CoreTiming::Advance();
-    REQUIRE(1000 == CoreTiming::GetDowncount());
+    core_timing.Advance();
+    REQUIRE(1000 == core_timing.GetDowncount());
 
     callbacks_ran_flags = 0;
     counter = 0;
     lateness = 0;
-    CoreTiming::AddTicks(CoreTiming::GetDowncount());
-    CoreTiming::Advance();
-    REQUIRE(MAX_SLICE_LENGTH == CoreTiming::GetDowncount());
+    core_timing.AddTicks(core_timing.GetDowncount());
+    core_timing.Advance();
+    REQUIRE(MAX_SLICE_LENGTH == core_timing.GetDowncount());
     REQUIRE(0x1FULL == callbacks_ran_flags.to_ullong());
 }
 
-TEST_CASE("CoreTiming[PredictableLateness]", "[core]") {
+TEST_CASE("Core::Timing[PredictableLateness]", "[core]") {
     ScopeInit guard;
+    auto& core_timing = guard.core_timing;
 
-    CoreTiming::EventType* cb_a = CoreTiming::RegisterEvent("callbackA", CallbackTemplate<0>);
-    CoreTiming::EventType* cb_b = CoreTiming::RegisterEvent("callbackB", CallbackTemplate<1>);
+    Core::Timing::EventType* cb_a = core_timing.RegisterEvent("callbackA", CallbackTemplate<0>);
+    Core::Timing::EventType* cb_b = core_timing.RegisterEvent("callbackB", CallbackTemplate<1>);
 
     // Enter slice 0
-    CoreTiming::Advance();
+    core_timing.Advance();
 
-    CoreTiming::ScheduleEvent(100, cb_a, CB_IDS[0]);
-    CoreTiming::ScheduleEvent(200, cb_b, CB_IDS[1]);
+    core_timing.ScheduleEvent(100, cb_a, CB_IDS[0]);
+    core_timing.ScheduleEvent(200, cb_b, CB_IDS[1]);
 
-    AdvanceAndCheck(0, 90, 10, -10); // (100 - 10)
-    AdvanceAndCheck(1, MAX_SLICE_LENGTH, 50, -50);
+    AdvanceAndCheck(core_timing, 0, 90, 10, -10); // (100 - 10)
+    AdvanceAndCheck(core_timing, 1, MAX_SLICE_LENGTH, 50, -50);
 }
 
 namespace ChainSchedulingTest {
 static int reschedules = 0;
 
-static void RescheduleCallback(u64 userdata, s64 cycles_late) {
+static void RescheduleCallback(Core::Timing::CoreTiming& core_timing, u64 userdata,
+                               s64 cycles_late) {
     --reschedules;
     REQUIRE(reschedules >= 0);
     REQUIRE(lateness == cycles_late);
 
-    if (reschedules > 0)
-        CoreTiming::ScheduleEvent(1000, reinterpret_cast<CoreTiming::EventType*>(userdata),
+    if (reschedules > 0) {
+        core_timing.ScheduleEvent(1000, reinterpret_cast<Core::Timing::EventType*>(userdata),
                                   userdata);
+    }
 }
 } // namespace ChainSchedulingTest
 
@@ -202,36 +209,39 @@ TEST_CASE("CoreTiming[ChainScheduling]", "[core]") {
     using namespace ChainSchedulingTest;
 
     ScopeInit guard;
+    auto& core_timing = guard.core_timing;
 
-    CoreTiming::EventType* cb_a = CoreTiming::RegisterEvent("callbackA", CallbackTemplate<0>);
-    CoreTiming::EventType* cb_b = CoreTiming::RegisterEvent("callbackB", CallbackTemplate<1>);
-    CoreTiming::EventType* cb_c = CoreTiming::RegisterEvent("callbackC", CallbackTemplate<2>);
-    CoreTiming::EventType* cb_rs =
-        CoreTiming::RegisterEvent("callbackReschedule", RescheduleCallback);
+    Core::Timing::EventType* cb_a = core_timing.RegisterEvent("callbackA", CallbackTemplate<0>);
+    Core::Timing::EventType* cb_b = core_timing.RegisterEvent("callbackB", CallbackTemplate<1>);
+    Core::Timing::EventType* cb_c = core_timing.RegisterEvent("callbackC", CallbackTemplate<2>);
+    Core::Timing::EventType* cb_rs = core_timing.RegisterEvent(
+        "callbackReschedule", [&core_timing](u64 userdata, s64 cycles_late) {
+            RescheduleCallback(core_timing, userdata, cycles_late);
+        });
 
     // Enter slice 0
-    CoreTiming::Advance();
+    core_timing.Advance();
 
-    CoreTiming::ScheduleEvent(800, cb_a, CB_IDS[0]);
-    CoreTiming::ScheduleEvent(1000, cb_b, CB_IDS[1]);
-    CoreTiming::ScheduleEvent(2200, cb_c, CB_IDS[2]);
-    CoreTiming::ScheduleEvent(1000, cb_rs, reinterpret_cast<u64>(cb_rs));
-    REQUIRE(800 == CoreTiming::GetDowncount());
+    core_timing.ScheduleEvent(800, cb_a, CB_IDS[0]);
+    core_timing.ScheduleEvent(1000, cb_b, CB_IDS[1]);
+    core_timing.ScheduleEvent(2200, cb_c, CB_IDS[2]);
+    core_timing.ScheduleEvent(1000, cb_rs, reinterpret_cast<u64>(cb_rs));
+    REQUIRE(800 == core_timing.GetDowncount());
 
     reschedules = 3;
-    AdvanceAndCheck(0, 200);  // cb_a
-    AdvanceAndCheck(1, 1000); // cb_b, cb_rs
+    AdvanceAndCheck(core_timing, 0, 200);  // cb_a
+    AdvanceAndCheck(core_timing, 1, 1000); // cb_b, cb_rs
     REQUIRE(2 == reschedules);
 
-    CoreTiming::AddTicks(CoreTiming::GetDowncount());
-    CoreTiming::Advance(); // cb_rs
+    core_timing.AddTicks(core_timing.GetDowncount());
+    core_timing.Advance(); // cb_rs
     REQUIRE(1 == reschedules);
-    REQUIRE(200 == CoreTiming::GetDowncount());
+    REQUIRE(200 == core_timing.GetDowncount());
 
-    AdvanceAndCheck(2, 800); // cb_c
+    AdvanceAndCheck(core_timing, 2, 800); // cb_c
 
-    CoreTiming::AddTicks(CoreTiming::GetDowncount());
-    CoreTiming::Advance(); // cb_rs
+    core_timing.AddTicks(core_timing.GetDowncount());
+    core_timing.Advance(); // cb_rs
     REQUIRE(0 == reschedules);
-    REQUIRE(MAX_SLICE_LENGTH == CoreTiming::GetDowncount());
+    REQUIRE(MAX_SLICE_LENGTH == core_timing.GetDowncount());
 }
diff --git a/src/video_core/CMakeLists.txt b/src/video_core/CMakeLists.txt
index 6113e17ff..14b76680f 100644
--- a/src/video_core/CMakeLists.txt
+++ b/src/video_core/CMakeLists.txt
@@ -5,18 +5,24 @@ add_library(video_core STATIC
     debug_utils/debug_utils.h
     engines/fermi_2d.cpp
     engines/fermi_2d.h
+    engines/kepler_compute.cpp
+    engines/kepler_compute.h
     engines/kepler_memory.cpp
     engines/kepler_memory.h
     engines/maxwell_3d.cpp
     engines/maxwell_3d.h
-    engines/maxwell_compute.cpp
-    engines/maxwell_compute.h
     engines/maxwell_dma.cpp
     engines/maxwell_dma.h
     engines/shader_bytecode.h
     engines/shader_header.h
     gpu.cpp
     gpu.h
+    gpu_asynch.cpp
+    gpu_asynch.h
+    gpu_synch.cpp
+    gpu_synch.h
+    gpu_thread.cpp
+    gpu_thread.h
     macro_interpreter.cpp
     macro_interpreter.h
     memory_manager.cpp
@@ -44,6 +50,8 @@ add_library(video_core STATIC
     renderer_opengl/gl_shader_cache.h
     renderer_opengl/gl_shader_decompiler.cpp
     renderer_opengl/gl_shader_decompiler.h
+    renderer_opengl/gl_shader_disk_cache.cpp
+    renderer_opengl/gl_shader_disk_cache.h
     renderer_opengl/gl_shader_gen.cpp
     renderer_opengl/gl_shader_gen.h
     renderer_opengl/gl_shader_manager.cpp
@@ -72,6 +80,7 @@ add_library(video_core STATIC
     shader/decode/hfma2.cpp
     shader/decode/conversion.cpp
     shader/decode/memory.cpp
+    shader/decode/texture.cpp
     shader/decode/float_set_predicate.cpp
     shader/decode/integer_set_predicate.cpp
     shader/decode/half_set_predicate.cpp
@@ -92,6 +101,8 @@ add_library(video_core STATIC
     surface.h
     textures/astc.cpp
     textures/astc.h
+    textures/convert.cpp
+    textures/convert.h
     textures/decoders.cpp
     textures/decoders.h
     textures/texture.h
@@ -99,7 +110,31 @@ add_library(video_core STATIC
     video_core.h
 )
 
+if (ENABLE_VULKAN)
+    target_sources(video_core PRIVATE
+        renderer_vulkan/declarations.h
+        renderer_vulkan/maxwell_to_vk.cpp
+        renderer_vulkan/maxwell_to_vk.h
+        renderer_vulkan/vk_buffer_cache.cpp
+        renderer_vulkan/vk_buffer_cache.h
+        renderer_vulkan/vk_device.cpp
+        renderer_vulkan/vk_device.h
+        renderer_vulkan/vk_memory_manager.cpp
+        renderer_vulkan/vk_memory_manager.h
+        renderer_vulkan/vk_resource_manager.cpp
+        renderer_vulkan/vk_resource_manager.h
+        renderer_vulkan/vk_sampler_cache.cpp
+        renderer_vulkan/vk_sampler_cache.h
+        renderer_vulkan/vk_scheduler.cpp
+        renderer_vulkan/vk_scheduler.h
+        renderer_vulkan/vk_stream_buffer.cpp
+        renderer_vulkan/vk_stream_buffer.h)
+
+    target_include_directories(video_core PRIVATE ../../externals/Vulkan-Headers/include)
+    target_compile_definitions(video_core PRIVATE HAS_VULKAN)
+endif()
+
 create_target_directory_groups(video_core)
 
 target_link_libraries(video_core PUBLIC common core)
-target_link_libraries(video_core PRIVATE glad)
+target_link_libraries(video_core PRIVATE glad lz4_static)
diff --git a/src/video_core/dma_pusher.cpp b/src/video_core/dma_pusher.cpp
index 63a958f11..8b1bea1ae 100644
--- a/src/video_core/dma_pusher.cpp
+++ b/src/video_core/dma_pusher.cpp
@@ -33,16 +33,33 @@ void DmaPusher::DispatchCalls() {
 }
 
 bool DmaPusher::Step() {
-    if (dma_get != dma_put) {
-        // Push buffer non-empty, read a word
-        const CommandHeader command_header{
-            Memory::Read32(*gpu.MemoryManager().GpuToCpuAddress(dma_get))};
+    if (!ib_enable || dma_pushbuffer.empty()) {
+        // pushbuffer empty and IB empty or nonexistent - nothing to do
+        return false;
+    }
 
-        dma_get += sizeof(u32);
+    const CommandList& command_list{dma_pushbuffer.front()};
+    const CommandListHeader command_list_header{command_list[dma_pushbuffer_subindex++]};
+    GPUVAddr dma_get = command_list_header.addr;
+    GPUVAddr dma_put = dma_get + command_list_header.size * sizeof(u32);
+    bool non_main = command_list_header.is_non_main;
 
-        if (!non_main) {
-            dma_mget = dma_get;
-        }
+    if (dma_pushbuffer_subindex >= command_list.size()) {
+        // We've gone through the current list, remove it from the queue
+        dma_pushbuffer.pop();
+        dma_pushbuffer_subindex = 0;
+    }
+
+    if (command_list_header.size == 0) {
+        return true;
+    }
+
+    // Push buffer non-empty, read a word
+    command_headers.resize(command_list_header.size);
+    gpu.MemoryManager().ReadBlock(dma_get, command_headers.data(),
+                                  command_list_header.size * sizeof(u32));
+
+    for (const CommandHeader& command_header : command_headers) {
 
         // now, see if we're in the middle of a command
         if (dma_state.length_pending) {
@@ -89,22 +106,11 @@ bool DmaPusher::Step() {
                 break;
             }
         }
-    } else if (ib_enable && !dma_pushbuffer.empty()) {
-        // Current pushbuffer empty, but we have more IB entries to read
-        const CommandList& command_list{dma_pushbuffer.front()};
-        const CommandListHeader& command_list_header{command_list[dma_pushbuffer_subindex++]};
-        dma_get = command_list_header.addr;
-        dma_put = dma_get + command_list_header.size * sizeof(u32);
-        non_main = command_list_header.is_non_main;
-
-        if (dma_pushbuffer_subindex >= command_list.size()) {
-            // We've gone through the current list, remove it from the queue
-            dma_pushbuffer.pop();
-            dma_pushbuffer_subindex = 0;
-        }
-    } else {
-        // Otherwise, pushbuffer empty and IB empty or nonexistent - nothing to do
-        return {};
+    }
+
+    if (!non_main) {
+        // TODO (degasus): This is dead code, as dma_mget is never read.
+        dma_mget = dma_put;
     }
 
     return true;
diff --git a/src/video_core/dma_pusher.h b/src/video_core/dma_pusher.h
index 1097e5c49..27a36348c 100644
--- a/src/video_core/dma_pusher.h
+++ b/src/video_core/dma_pusher.h
@@ -75,6 +75,8 @@ private:
 
     GPU& gpu;
 
+    std::vector<CommandHeader> command_headers; ///< Buffer for list of commands fetched at once
+
     std::queue<CommandList> dma_pushbuffer; ///< Queue of command lists to be processed
     std::size_t dma_pushbuffer_subindex{};  ///< Index within a command list within the pushbuffer
 
@@ -89,11 +91,8 @@ private:
     DmaState dma_state{};
     bool dma_increment_once{};
 
-    GPUVAddr dma_put{};   ///< pushbuffer current end address
-    GPUVAddr dma_get{};   ///< pushbuffer current read address
     GPUVAddr dma_mget{};  ///< main pushbuffer last read address
     bool ib_enable{true}; ///< IB mode enabled
-    bool non_main{};      ///< non-main pushbuffer active
 };
 
 } // namespace Tegra
diff --git a/src/video_core/engines/fermi_2d.cpp b/src/video_core/engines/fermi_2d.cpp
index 80f70e332..03b7ee5d8 100644
--- a/src/video_core/engines/fermi_2d.cpp
+++ b/src/video_core/engines/fermi_2d.cpp
@@ -2,12 +2,11 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
-#include "core/core.h"
-#include "core/memory.h"
+#include "common/assert.h"
+#include "common/logging/log.h"
+#include "common/math_util.h"
 #include "video_core/engines/fermi_2d.h"
-#include "video_core/engines/maxwell_3d.h"
 #include "video_core/rasterizer_interface.h"
-#include "video_core/textures/decoders.h"
 
 namespace Tegra::Engines {
 
@@ -21,7 +20,9 @@ void Fermi2D::CallMethod(const GPU::MethodCall& method_call) {
     regs.reg_array[method_call.method] = method_call.argument;
 
     switch (method_call.method) {
-    case FERMI2D_REG_INDEX(trigger): {
+    // Trigger the surface copy on the last register write. This is blit_src_y, but this is 64-bit,
+    // so trigger on the second 32-bit write.
+    case FERMI2D_REG_INDEX(blit_src_y) + 1: {
         HandleSurfaceCopy();
         break;
     }
@@ -32,55 +33,23 @@ void Fermi2D::HandleSurfaceCopy() {
     LOG_WARNING(HW_GPU, "Requested a surface copy with operation {}",
                 static_cast<u32>(regs.operation));
 
-    const GPUVAddr source = regs.src.Address();
-    const GPUVAddr dest = regs.dst.Address();
-
-    // TODO(Subv): Only same-format and same-size copies are allowed for now.
-    ASSERT(regs.src.format == regs.dst.format);
-    ASSERT(regs.src.width * regs.src.height == regs.dst.width * regs.dst.height);
-
     // TODO(Subv): Only raw copies are implemented.
     ASSERT(regs.operation == Regs::Operation::SrcCopy);
 
-    const VAddr source_cpu = *memory_manager.GpuToCpuAddress(source);
-    const VAddr dest_cpu = *memory_manager.GpuToCpuAddress(dest);
-
-    u32 src_bytes_per_pixel = RenderTargetBytesPerPixel(regs.src.format);
-    u32 dst_bytes_per_pixel = RenderTargetBytesPerPixel(regs.dst.format);
-
-    if (!rasterizer.AccelerateSurfaceCopy(regs.src, regs.dst)) {
-        // All copies here update the main memory, so mark all rasterizer states as invalid.
-        Core::System::GetInstance().GPU().Maxwell3D().dirty_flags.OnMemoryWrite();
+    const u32 src_blit_x1{static_cast<u32>(regs.blit_src_x >> 32)};
+    const u32 src_blit_y1{static_cast<u32>(regs.blit_src_y >> 32)};
+    const u32 src_blit_x2{
+        static_cast<u32>((regs.blit_src_x + (regs.blit_dst_width * regs.blit_du_dx)) >> 32)};
+    const u32 src_blit_y2{
+        static_cast<u32>((regs.blit_src_y + (regs.blit_dst_height * regs.blit_dv_dy)) >> 32)};
 
-        rasterizer.FlushRegion(source_cpu, src_bytes_per_pixel * regs.src.width * regs.src.height);
-        // We have to invalidate the destination region to evict any outdated surfaces from the
-        // cache. We do this before actually writing the new data because the destination address
-        // might contain a dirty surface that will have to be written back to memory.
-        rasterizer.InvalidateRegion(dest_cpu,
-                                    dst_bytes_per_pixel * regs.dst.width * regs.dst.height);
+    const Common::Rectangle<u32> src_rect{src_blit_x1, src_blit_y1, src_blit_x2, src_blit_y2};
+    const Common::Rectangle<u32> dst_rect{regs.blit_dst_x, regs.blit_dst_y,
+                                          regs.blit_dst_x + regs.blit_dst_width,
+                                          regs.blit_dst_y + regs.blit_dst_height};
 
-        if (regs.src.linear == regs.dst.linear) {
-            // If the input layout and the output layout are the same, just perform a raw copy.
-            ASSERT(regs.src.BlockHeight() == regs.dst.BlockHeight());
-            Memory::CopyBlock(dest_cpu, source_cpu,
-                              src_bytes_per_pixel * regs.dst.width * regs.dst.height);
-            return;
-        }
-        u8* src_buffer = Memory::GetPointer(source_cpu);
-        u8* dst_buffer = Memory::GetPointer(dest_cpu);
-        if (!regs.src.linear && regs.dst.linear) {
-            // If the input is tiled and the output is linear, deswizzle the input and copy it over.
-            Texture::CopySwizzledData(regs.src.width, regs.src.height, regs.src.depth,
-                                      src_bytes_per_pixel, dst_bytes_per_pixel, src_buffer,
-                                      dst_buffer, true, regs.src.BlockHeight(),
-                                      regs.src.BlockDepth(), 0);
-        } else {
-            // If the input is linear and the output is tiled, swizzle the input and copy it over.
-            Texture::CopySwizzledData(regs.src.width, regs.src.height, regs.src.depth,
-                                      src_bytes_per_pixel, dst_bytes_per_pixel, dst_buffer,
-                                      src_buffer, false, regs.dst.BlockHeight(),
-                                      regs.dst.BlockDepth(), 0);
-        }
+    if (!rasterizer.AccelerateSurfaceCopy(regs.src, regs.dst, src_rect, dst_rect)) {
+        UNIMPLEMENTED();
     }
 }
 
diff --git a/src/video_core/engines/fermi_2d.h b/src/video_core/engines/fermi_2d.h
index 50009bf75..80523e320 100644
--- a/src/video_core/engines/fermi_2d.h
+++ b/src/video_core/engines/fermi_2d.h
@@ -5,7 +5,7 @@
 #pragma once
 
 #include <array>
-#include "common/assert.h"
+#include <cstddef>
 #include "common/bit_field.h"
 #include "common/common_funcs.h"
 #include "common/common_types.h"
@@ -94,12 +94,22 @@ public:
 
                 Operation operation;
 
-                INSERT_PADDING_WORDS(0x9);
+                INSERT_PADDING_WORDS(0x177);
 
-                // TODO(Subv): This is only a guess.
-                u32 trigger;
+                u32 blit_control;
 
-                INSERT_PADDING_WORDS(0x1A3);
+                INSERT_PADDING_WORDS(0x8);
+
+                u32 blit_dst_x;
+                u32 blit_dst_y;
+                u32 blit_dst_width;
+                u32 blit_dst_height;
+                u64 blit_du_dx;
+                u64 blit_dv_dy;
+                u64 blit_src_x;
+                u64 blit_src_y;
+
+                INSERT_PADDING_WORDS(0x21);
             };
             std::array<u32, NUM_REGS> reg_array;
         };
@@ -122,7 +132,16 @@ private:
 ASSERT_REG_POSITION(dst, 0x80);
 ASSERT_REG_POSITION(src, 0x8C);
 ASSERT_REG_POSITION(operation, 0xAB);
-ASSERT_REG_POSITION(trigger, 0xB5);
+ASSERT_REG_POSITION(blit_control, 0x223);
+ASSERT_REG_POSITION(blit_dst_x, 0x22c);
+ASSERT_REG_POSITION(blit_dst_y, 0x22d);
+ASSERT_REG_POSITION(blit_dst_width, 0x22e);
+ASSERT_REG_POSITION(blit_dst_height, 0x22f);
+ASSERT_REG_POSITION(blit_du_dx, 0x230);
+ASSERT_REG_POSITION(blit_dv_dy, 0x232);
+ASSERT_REG_POSITION(blit_src_x, 0x234);
+ASSERT_REG_POSITION(blit_src_y, 0x236);
+
 #undef ASSERT_REG_POSITION
 
 } // namespace Tegra::Engines
diff --git a/src/video_core/engines/kepler_compute.cpp b/src/video_core/engines/kepler_compute.cpp
new file mode 100644
index 000000000..b1d950460
--- /dev/null
+++ b/src/video_core/engines/kepler_compute.cpp
@@ -0,0 +1,33 @@
+// 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/logging/log.h"
+#include "video_core/engines/kepler_compute.h"
+#include "video_core/memory_manager.h"
+
+namespace Tegra::Engines {
+
+KeplerCompute::KeplerCompute(MemoryManager& memory_manager) : memory_manager{memory_manager} {}
+
+KeplerCompute::~KeplerCompute() = default;
+
+void KeplerCompute::CallMethod(const GPU::MethodCall& method_call) {
+    ASSERT_MSG(method_call.method < Regs::NUM_REGS,
+               "Invalid KeplerCompute register, increase the size of the Regs structure");
+
+    regs.reg_array[method_call.method] = method_call.argument;
+
+    switch (method_call.method) {
+    case KEPLER_COMPUTE_REG_INDEX(launch):
+        // Abort execution since compute shaders can be used to alter game memory (e.g. CUDA
+        // kernels)
+        UNREACHABLE_MSG("Compute shaders are not implemented");
+        break;
+    default:
+        break;
+    }
+}
+
+} // namespace Tegra::Engines
diff --git a/src/video_core/engines/maxwell_compute.h b/src/video_core/engines/kepler_compute.h
index 1d71f11bd..6575afd0f 100644
--- a/src/video_core/engines/maxwell_compute.h
+++ b/src/video_core/engines/kepler_compute.h
@@ -5,52 +5,52 @@
 #pragma once
 
 #include <array>
-#include "common/assert.h"
-#include "common/bit_field.h"
+#include <cstddef>
 #include "common/common_funcs.h"
 #include "common/common_types.h"
 #include "video_core/gpu.h"
+#include "video_core/memory_manager.h"
 
 namespace Tegra::Engines {
 
-#define MAXWELL_COMPUTE_REG_INDEX(field_name)                                                      \
-    (offsetof(Tegra::Engines::MaxwellCompute::Regs, field_name) / sizeof(u32))
+#define KEPLER_COMPUTE_REG_INDEX(field_name)                                                       \
+    (offsetof(Tegra::Engines::KeplerCompute::Regs, field_name) / sizeof(u32))
 
-class MaxwellCompute final {
+class KeplerCompute final {
 public:
-    MaxwellCompute() = default;
-    ~MaxwellCompute() = default;
+    explicit KeplerCompute(MemoryManager& memory_manager);
+    ~KeplerCompute();
+
+    static constexpr std::size_t NumConstBuffers = 8;
 
     struct Regs {
         static constexpr std::size_t NUM_REGS = 0xCF8;
 
         union {
             struct {
-                INSERT_PADDING_WORDS(0x281);
+                INSERT_PADDING_WORDS(0xAF);
 
-                union {
-                    u32 compute_end;
-                    BitField<0, 1, u32> unknown;
-                } compute;
+                u32 launch;
 
-                INSERT_PADDING_WORDS(0xA76);
+                INSERT_PADDING_WORDS(0xC48);
             };
             std::array<u32, NUM_REGS> reg_array;
         };
     } regs{};
-
     static_assert(sizeof(Regs) == Regs::NUM_REGS * sizeof(u32),
-                  "MaxwellCompute Regs has wrong size");
+                  "KeplerCompute Regs has wrong size");
+
+    MemoryManager& memory_manager;
 
     /// Write the value to the register identified by method.
     void CallMethod(const GPU::MethodCall& method_call);
 };
 
 #define ASSERT_REG_POSITION(field_name, position)                                                  \
-    static_assert(offsetof(MaxwellCompute::Regs, field_name) == position * 4,                      \
+    static_assert(offsetof(KeplerCompute::Regs, field_name) == position * 4,                       \
                   "Field " #field_name " has invalid position")
 
-ASSERT_REG_POSITION(compute, 0x281);
+ASSERT_REG_POSITION(launch, 0xAF);
 
 #undef ASSERT_REG_POSITION
 
diff --git a/src/video_core/engines/kepler_memory.cpp b/src/video_core/engines/kepler_memory.cpp
index 4880191fc..0931b9626 100644
--- a/src/video_core/engines/kepler_memory.cpp
+++ b/src/video_core/engines/kepler_memory.cpp
@@ -2,18 +2,20 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#include "common/assert.h"
 #include "common/logging/log.h"
 #include "core/core.h"
 #include "core/memory.h"
 #include "video_core/engines/kepler_memory.h"
 #include "video_core/engines/maxwell_3d.h"
 #include "video_core/rasterizer_interface.h"
+#include "video_core/renderer_base.h"
 
 namespace Tegra::Engines {
 
-KeplerMemory::KeplerMemory(VideoCore::RasterizerInterface& rasterizer,
+KeplerMemory::KeplerMemory(Core::System& system, VideoCore::RasterizerInterface& rasterizer,
                            MemoryManager& memory_manager)
-    : memory_manager(memory_manager), rasterizer{rasterizer} {}
+    : system{system}, memory_manager(memory_manager), rasterizer{rasterizer} {}
 
 KeplerMemory::~KeplerMemory() = default;
 
@@ -39,17 +41,14 @@ void KeplerMemory::ProcessData(u32 data) {
     ASSERT_MSG(regs.exec.linear, "Non-linear uploads are not supported");
     ASSERT(regs.dest.x == 0 && regs.dest.y == 0 && regs.dest.z == 0);
 
-    GPUVAddr address = regs.dest.Address();
-    VAddr dest_address =
-        *memory_manager.GpuToCpuAddress(address + state.write_offset * sizeof(u32));
-
     // We have to invalidate the destination region to evict any outdated surfaces from the cache.
-    // We do this before actually writing the new data because the destination address might contain
-    // a dirty surface that will have to be written back to memory.
-    rasterizer.InvalidateRegion(dest_address, sizeof(u32));
+    // We do this before actually writing the new data because the destination address might
+    // contain a dirty surface that will have to be written back to memory.
+    const GPUVAddr address{regs.dest.Address() + state.write_offset * sizeof(u32)};
+    rasterizer.InvalidateRegion(ToCacheAddr(memory_manager.GetPointer(address)), sizeof(u32));
+    memory_manager.Write32(address, data);
 
-    Memory::Write32(dest_address, data);
-    Core::System::GetInstance().GPU().Maxwell3D().dirty_flags.OnMemoryWrite();
+    system.GPU().Maxwell3D().dirty_flags.OnMemoryWrite();
 
     state.write_offset++;
 }
diff --git a/src/video_core/engines/kepler_memory.h b/src/video_core/engines/kepler_memory.h
index fe9ebc5b9..9181e9d80 100644
--- a/src/video_core/engines/kepler_memory.h
+++ b/src/video_core/engines/kepler_memory.h
@@ -5,13 +5,17 @@
 #pragma once
 
 #include <array>
-#include "common/assert.h"
+#include <cstddef>
 #include "common/bit_field.h"
 #include "common/common_funcs.h"
 #include "common/common_types.h"
 #include "video_core/gpu.h"
 #include "video_core/memory_manager.h"
 
+namespace Core {
+class System;
+}
+
 namespace VideoCore {
 class RasterizerInterface;
 }
@@ -23,7 +27,8 @@ namespace Tegra::Engines {
 
 class KeplerMemory final {
 public:
-    KeplerMemory(VideoCore::RasterizerInterface& rasterizer, MemoryManager& memory_manager);
+    KeplerMemory(Core::System& system, VideoCore::RasterizerInterface& rasterizer,
+                 MemoryManager& memory_manager);
     ~KeplerMemory();
 
     /// Write the value to the register identified by method.
@@ -76,6 +81,7 @@ public:
     } state{};
 
 private:
+    Core::System& system;
     MemoryManager& memory_manager;
     VideoCore::RasterizerInterface& rasterizer;
 
diff --git a/src/video_core/engines/maxwell_3d.cpp b/src/video_core/engines/maxwell_3d.cpp
index a388b3944..c5d5be4ef 100644
--- a/src/video_core/engines/maxwell_3d.cpp
+++ b/src/video_core/engines/maxwell_3d.cpp
@@ -19,8 +19,10 @@ namespace Tegra::Engines {
 /// First register id that is actually a Macro call.
 constexpr u32 MacroRegistersStart = 0xE00;
 
-Maxwell3D::Maxwell3D(VideoCore::RasterizerInterface& rasterizer, MemoryManager& memory_manager)
-    : memory_manager(memory_manager), rasterizer{rasterizer}, macro_interpreter(*this) {
+Maxwell3D::Maxwell3D(Core::System& system, VideoCore::RasterizerInterface& rasterizer,
+                     MemoryManager& memory_manager)
+    : memory_manager(memory_manager), system{system}, rasterizer{rasterizer},
+      macro_interpreter(*this) {
     InitializeRegisterDefaults();
 }
 
@@ -103,23 +105,25 @@ void Maxwell3D::CallMacroMethod(u32 method, std::vector<u32> parameters) {
 }
 
 void Maxwell3D::CallMethod(const GPU::MethodCall& method_call) {
-    auto debug_context = Core::System::GetInstance().GetGPUDebugContext();
+    auto debug_context = system.GetGPUDebugContext();
+
+    const u32 method = method_call.method;
 
     // It is an error to write to a register other than the current macro's ARG register before it
     // has finished execution.
     if (executing_macro != 0) {
-        ASSERT(method_call.method == executing_macro + 1);
+        ASSERT(method == executing_macro + 1);
     }
 
     // Methods after 0xE00 are special, they're actually triggers for some microcode that was
     // uploaded to the GPU during initialization.
-    if (method_call.method >= MacroRegistersStart) {
+    if (method >= MacroRegistersStart) {
         // We're trying to execute a macro
         if (executing_macro == 0) {
             // A macro call must begin by writing the macro method's register, not its argument.
-            ASSERT_MSG((method_call.method % 2) == 0,
+            ASSERT_MSG((method % 2) == 0,
                        "Can't start macro execution by writing to the ARGS register");
-            executing_macro = method_call.method;
+            executing_macro = method;
         }
 
         macro_params.push_back(method_call.argument);
@@ -131,66 +135,62 @@ void Maxwell3D::CallMethod(const GPU::MethodCall& method_call) {
         return;
     }
 
-    ASSERT_MSG(method_call.method < Regs::NUM_REGS,
+    ASSERT_MSG(method < Regs::NUM_REGS,
                "Invalid Maxwell3D register, increase the size of the Regs structure");
 
     if (debug_context) {
         debug_context->OnEvent(Tegra::DebugContext::Event::MaxwellCommandLoaded, nullptr);
     }
 
-    if (regs.reg_array[method_call.method] != method_call.argument) {
-        regs.reg_array[method_call.method] = method_call.argument;
+    if (regs.reg_array[method] != method_call.argument) {
+        regs.reg_array[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);
+        if (method >= first_rt_reg &&
+            method < first_rt_reg + registers_per_rt * Regs::NumRenderTargets) {
+            const std::size_t rt_index = (method - first_rt_reg) / registers_per_rt;
+            dirty_flags.color_buffer.set(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)) {
+        if (method == MAXWELL3D_REG_INDEX(zeta_enable) ||
+            method == MAXWELL3D_REG_INDEX(zeta_width) ||
+            method == MAXWELL3D_REG_INDEX(zeta_height) ||
+            (method >= MAXWELL3D_REG_INDEX(zeta) &&
+             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) {
+        if (method >= MAXWELL3D_REG_INDEX(shader_config[0]) &&
+            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 <
-                MAXWELL3D_REG_INDEX(vertex_attrib_format) + regs.vertex_attrib_format.size()) {
+        if (method >= MAXWELL3D_REG_INDEX(vertex_attrib_format) &&
+            method < MAXWELL3D_REG_INDEX(vertex_attrib_format) + regs.vertex_attrib_format.size()) {
             dirty_flags.vertex_attrib_format = true;
         }
 
         // Vertex buffer
-        if (method_call.method >= MAXWELL3D_REG_INDEX(vertex_array) &&
-            method_call.method < MAXWELL3D_REG_INDEX(vertex_array) + 4 * 32) {
-            dirty_flags.vertex_array |=
-                1u << ((method_call.method - MAXWELL3D_REG_INDEX(vertex_array)) >> 2);
-        } else if (method_call.method >= MAXWELL3D_REG_INDEX(vertex_array_limit) &&
-                   method_call.method < MAXWELL3D_REG_INDEX(vertex_array_limit) + 2 * 32) {
-            dirty_flags.vertex_array |=
-                1u << ((method_call.method - MAXWELL3D_REG_INDEX(vertex_array_limit)) >> 1);
-        } else if (method_call.method >= MAXWELL3D_REG_INDEX(instanced_arrays) &&
-                   method_call.method < MAXWELL3D_REG_INDEX(instanced_arrays) + 32) {
-            dirty_flags.vertex_array |=
-                1u << (method_call.method - MAXWELL3D_REG_INDEX(instanced_arrays));
+        if (method >= MAXWELL3D_REG_INDEX(vertex_array) &&
+            method < MAXWELL3D_REG_INDEX(vertex_array) + 4 * 32) {
+            dirty_flags.vertex_array.set((method - MAXWELL3D_REG_INDEX(vertex_array)) >> 2);
+        } else if (method >= MAXWELL3D_REG_INDEX(vertex_array_limit) &&
+                   method < MAXWELL3D_REG_INDEX(vertex_array_limit) + 2 * 32) {
+            dirty_flags.vertex_array.set((method - MAXWELL3D_REG_INDEX(vertex_array_limit)) >> 1);
+        } else if (method >= MAXWELL3D_REG_INDEX(instanced_arrays) &&
+                   method < MAXWELL3D_REG_INDEX(instanced_arrays) + 32) {
+            dirty_flags.vertex_array.set(method - MAXWELL3D_REG_INDEX(instanced_arrays));
         }
     }
 
-    switch (method_call.method) {
+    switch (method) {
     case MAXWELL3D_REG_INDEX(macros.data): {
         ProcessMacroUpload(method_call.argument);
         break;
@@ -270,10 +270,9 @@ void Maxwell3D::ProcessMacroBind(u32 data) {
 }
 
 void Maxwell3D::ProcessQueryGet() {
-    GPUVAddr sequence_address = regs.query.QueryAddress();
+    const GPUVAddr sequence_address{regs.query.QueryAddress()};
     // Since the sequence address is given as a GPU VAddr, we have to convert it to an application
     // VAddr before writing.
-    std::optional<VAddr> address = memory_manager.GpuToCpuAddress(sequence_address);
 
     // TODO(Subv): Support the other query units.
     ASSERT_MSG(regs.query.query_get.unit == Regs::QueryUnit::Crop,
@@ -308,7 +307,7 @@ void Maxwell3D::ProcessQueryGet() {
             // Write the current query sequence to the sequence address.
             // TODO(Subv): Find out what happens if you use a long query type but mark it as a short
             // query.
-            Memory::Write32(*address, sequence);
+            memory_manager.Write32(sequence_address, sequence);
         } else {
             // Write the 128-bit result structure in long mode. Note: We emulate an infinitely fast
             // GPU, this command may actually take a while to complete in real hardware due to GPU
@@ -316,8 +315,8 @@ void Maxwell3D::ProcessQueryGet() {
             LongQueryResult query_result{};
             query_result.value = result;
             // TODO(Subv): Generate a real GPU timestamp and write it here instead of CoreTiming
-            query_result.timestamp = CoreTiming::GetTicks();
-            Memory::WriteBlock(*address, &query_result, sizeof(query_result));
+            query_result.timestamp = system.CoreTiming().GetTicks();
+            memory_manager.WriteBlock(sequence_address, &query_result, sizeof(query_result));
         }
         dirty_flags.OnMemoryWrite();
         break;
@@ -333,7 +332,7 @@ void Maxwell3D::DrawArrays() {
               regs.vertex_buffer.count);
     ASSERT_MSG(!(regs.index_array.count && regs.vertex_buffer.count), "Both indexed and direct?");
 
-    auto debug_context = Core::System::GetInstance().GetGPUDebugContext();
+    auto debug_context = system.GetGPUDebugContext();
 
     if (debug_context) {
         debug_context->OnEvent(Tegra::DebugContext::Event::IncomingPrimitiveBatch, nullptr);
@@ -386,16 +385,18 @@ void Maxwell3D::ProcessCBBind(Regs::ShaderStage stage) {
 
 void Maxwell3D::ProcessCBData(u32 value) {
     // Write the input value to the current const buffer at the current position.
-    GPUVAddr buffer_address = regs.const_buffer.BufferAddress();
+    const GPUVAddr buffer_address = regs.const_buffer.BufferAddress();
     ASSERT(buffer_address != 0);
 
     // Don't allow writing past the end of the buffer.
     ASSERT(regs.const_buffer.cb_pos + sizeof(u32) <= regs.const_buffer.cb_size);
 
-    std::optional<VAddr> address =
-        memory_manager.GpuToCpuAddress(buffer_address + regs.const_buffer.cb_pos);
+    const GPUVAddr address{buffer_address + regs.const_buffer.cb_pos};
+
+    u8* ptr{memory_manager.GetPointer(address)};
+    rasterizer.InvalidateRegion(ToCacheAddr(ptr), sizeof(u32));
+    memory_manager.Write32(address, value);
 
-    Memory::Write32(*address, value);
     dirty_flags.OnMemoryWrite();
 
     // Increment the current buffer position.
@@ -403,22 +404,19 @@ void Maxwell3D::ProcessCBData(u32 value) {
 }
 
 Texture::TICEntry Maxwell3D::GetTICEntry(u32 tic_index) const {
-    GPUVAddr tic_base_address = regs.tic.TICAddress();
-
-    GPUVAddr tic_address_gpu = tic_base_address + tic_index * sizeof(Texture::TICEntry);
-    std::optional<VAddr> tic_address_cpu = memory_manager.GpuToCpuAddress(tic_address_gpu);
+    const GPUVAddr tic_address_gpu{regs.tic.TICAddress() + tic_index * sizeof(Texture::TICEntry)};
 
     Texture::TICEntry tic_entry;
-    Memory::ReadBlock(*tic_address_cpu, &tic_entry, sizeof(Texture::TICEntry));
+    memory_manager.ReadBlock(tic_address_gpu, &tic_entry, sizeof(Texture::TICEntry));
 
     ASSERT_MSG(tic_entry.header_version == Texture::TICHeaderVersion::BlockLinear ||
                    tic_entry.header_version == Texture::TICHeaderVersion::Pitch,
                "TIC versions other than BlockLinear or Pitch are unimplemented");
 
-    auto r_type = tic_entry.r_type.Value();
-    auto g_type = tic_entry.g_type.Value();
-    auto b_type = tic_entry.b_type.Value();
-    auto a_type = tic_entry.a_type.Value();
+    const auto r_type = tic_entry.r_type.Value();
+    const auto g_type = tic_entry.g_type.Value();
+    const auto b_type = tic_entry.b_type.Value();
+    const auto a_type = tic_entry.a_type.Value();
 
     // TODO(Subv): Different data types for separate components are not supported
     ASSERT(r_type == g_type && r_type == b_type && r_type == a_type);
@@ -427,13 +425,10 @@ Texture::TICEntry Maxwell3D::GetTICEntry(u32 tic_index) const {
 }
 
 Texture::TSCEntry Maxwell3D::GetTSCEntry(u32 tsc_index) const {
-    GPUVAddr tsc_base_address = regs.tsc.TSCAddress();
-
-    GPUVAddr tsc_address_gpu = tsc_base_address + tsc_index * sizeof(Texture::TSCEntry);
-    std::optional<VAddr> tsc_address_cpu = memory_manager.GpuToCpuAddress(tsc_address_gpu);
+    const GPUVAddr tsc_address_gpu{regs.tsc.TSCAddress() + tsc_index * sizeof(Texture::TSCEntry)};
 
     Texture::TSCEntry tsc_entry;
-    Memory::ReadBlock(*tsc_address_cpu, &tsc_entry, sizeof(Texture::TSCEntry));
+    memory_manager.ReadBlock(tsc_address_gpu, &tsc_entry, sizeof(Texture::TSCEntry));
     return tsc_entry;
 }
 
@@ -452,8 +447,7 @@ std::vector<Texture::FullTextureInfo> Maxwell3D::GetStageTextures(Regs::ShaderSt
     for (GPUVAddr current_texture = tex_info_buffer.address + TextureInfoOffset;
          current_texture < tex_info_buffer_end; current_texture += sizeof(Texture::TextureHandle)) {
 
-        Texture::TextureHandle tex_handle{
-            Memory::Read32(*memory_manager.GpuToCpuAddress(current_texture))};
+        const Texture::TextureHandle tex_handle{memory_manager.Read32(current_texture)};
 
         Texture::FullTextureInfo tex_info{};
         // TODO(Subv): Use the shader to determine which textures are actually accessed.
@@ -462,23 +456,16 @@ std::vector<Texture::FullTextureInfo> Maxwell3D::GetStageTextures(Regs::ShaderSt
             sizeof(Texture::TextureHandle);
 
         // Load the TIC data.
-        if (tex_handle.tic_id != 0) {
-            tex_info.enabled = true;
-
-            auto tic_entry = GetTICEntry(tex_handle.tic_id);
-            // TODO(Subv): Workaround for BitField's move constructor being deleted.
-            std::memcpy(&tex_info.tic, &tic_entry, sizeof(tic_entry));
-        }
+        auto tic_entry = GetTICEntry(tex_handle.tic_id);
+        // TODO(Subv): Workaround for BitField's move constructor being deleted.
+        std::memcpy(&tex_info.tic, &tic_entry, sizeof(tic_entry));
 
         // Load the TSC data
-        if (tex_handle.tsc_id != 0) {
-            auto tsc_entry = GetTSCEntry(tex_handle.tsc_id);
-            // TODO(Subv): Workaround for BitField's move constructor being deleted.
-            std::memcpy(&tex_info.tsc, &tsc_entry, sizeof(tsc_entry));
-        }
+        auto tsc_entry = GetTSCEntry(tex_handle.tsc_id);
+        // TODO(Subv): Workaround for BitField's move constructor being deleted.
+        std::memcpy(&tex_info.tsc, &tsc_entry, sizeof(tsc_entry));
 
-        if (tex_info.enabled)
-            textures.push_back(tex_info);
+        textures.push_back(tex_info);
     }
 
     return textures;
@@ -490,31 +477,25 @@ Texture::FullTextureInfo Maxwell3D::GetStageTexture(Regs::ShaderStage stage,
     auto& tex_info_buffer = shader.const_buffers[regs.tex_cb_index];
     ASSERT(tex_info_buffer.enabled && tex_info_buffer.address != 0);
 
-    GPUVAddr tex_info_address = tex_info_buffer.address + offset * sizeof(Texture::TextureHandle);
+    const GPUVAddr tex_info_address =
+        tex_info_buffer.address + offset * sizeof(Texture::TextureHandle);
 
     ASSERT(tex_info_address < tex_info_buffer.address + tex_info_buffer.size);
 
-    std::optional<VAddr> tex_address_cpu = memory_manager.GpuToCpuAddress(tex_info_address);
-    Texture::TextureHandle tex_handle{Memory::Read32(*tex_address_cpu)};
+    const Texture::TextureHandle tex_handle{memory_manager.Read32(tex_info_address)};
 
     Texture::FullTextureInfo tex_info{};
     tex_info.index = static_cast<u32>(offset);
 
     // Load the TIC data.
-    if (tex_handle.tic_id != 0) {
-        tex_info.enabled = true;
-
-        auto tic_entry = GetTICEntry(tex_handle.tic_id);
-        // TODO(Subv): Workaround for BitField's move constructor being deleted.
-        std::memcpy(&tex_info.tic, &tic_entry, sizeof(tic_entry));
-    }
+    auto tic_entry = GetTICEntry(tex_handle.tic_id);
+    // TODO(Subv): Workaround for BitField's move constructor being deleted.
+    std::memcpy(&tex_info.tic, &tic_entry, sizeof(tic_entry));
 
     // Load the TSC data
-    if (tex_handle.tsc_id != 0) {
-        auto tsc_entry = GetTSCEntry(tex_handle.tsc_id);
-        // TODO(Subv): Workaround for BitField's move constructor being deleted.
-        std::memcpy(&tex_info.tsc, &tsc_entry, sizeof(tsc_entry));
-    }
+    auto tsc_entry = GetTSCEntry(tex_handle.tsc_id);
+    // TODO(Subv): Workaround for BitField's move constructor being deleted.
+    std::memcpy(&tex_info.tsc, &tsc_entry, sizeof(tsc_entry));
 
     return tex_info;
 }
diff --git a/src/video_core/engines/maxwell_3d.h b/src/video_core/engines/maxwell_3d.h
index 1f76aa670..7fbf1026e 100644
--- a/src/video_core/engines/maxwell_3d.h
+++ b/src/video_core/engines/maxwell_3d.h
@@ -5,8 +5,10 @@
 #pragma once
 
 #include <array>
+#include <bitset>
 #include <unordered_map>
 #include <vector>
+
 #include "common/assert.h"
 #include "common/bit_field.h"
 #include "common/common_funcs.h"
@@ -17,6 +19,10 @@
 #include "video_core/memory_manager.h"
 #include "video_core/textures/texture.h"
 
+namespace Core {
+class System;
+}
+
 namespace VideoCore {
 class RasterizerInterface;
 }
@@ -28,7 +34,8 @@ namespace Tegra::Engines {
 
 class Maxwell3D final {
 public:
-    explicit Maxwell3D(VideoCore::RasterizerInterface& rasterizer, MemoryManager& memory_manager);
+    explicit Maxwell3D(Core::System& system, VideoCore::RasterizerInterface& rasterizer,
+                       MemoryManager& memory_manager);
     ~Maxwell3D() = default;
 
     /// Register structure of the Maxwell3D engine.
@@ -498,7 +505,7 @@ public:
             f32 translate_z;
             INSERT_PADDING_WORDS(2);
 
-            MathUtil::Rectangle<s32> GetRect() const {
+            Common::Rectangle<s32> GetRect() const {
                 return {
                     GetX(),               // left
                     GetY() + GetHeight(), // top
@@ -1089,19 +1096,18 @@ public:
     MemoryManager& memory_manager;
 
     struct DirtyFlags {
-        u8 color_buffer = 0xFF;
-        bool zeta_buffer = true;
-
-        bool shaders = true;
+        std::bitset<8> color_buffer{0xFF};
+        std::bitset<32> vertex_array{0xFFFFFFFF};
 
         bool vertex_attrib_format = true;
-        u32 vertex_array = 0xFFFFFFFF;
+        bool zeta_buffer = true;
+        bool shaders = true;
 
         void OnMemoryWrite() {
-            color_buffer = 0xFF;
             zeta_buffer = true;
             shaders = true;
-            vertex_array = 0xFFFFFFFF;
+            color_buffer.set();
+            vertex_array.set();
         }
     };
 
@@ -1131,6 +1137,8 @@ public:
 private:
     void InitializeRegisterDefaults();
 
+    Core::System& system;
+
     VideoCore::RasterizerInterface& rasterizer;
 
     /// Start offsets of each macro in macro_memory
diff --git a/src/video_core/engines/maxwell_compute.cpp b/src/video_core/engines/maxwell_compute.cpp
deleted file mode 100644
index 656db6a61..000000000
--- a/src/video_core/engines/maxwell_compute.cpp
+++ /dev/null
@@ -1,28 +0,0 @@
-// Copyright 2018 yuzu Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#include "common/logging/log.h"
-#include "core/core.h"
-#include "video_core/engines/maxwell_compute.h"
-
-namespace Tegra::Engines {
-
-void MaxwellCompute::CallMethod(const GPU::MethodCall& method_call) {
-    ASSERT_MSG(method_call.method < Regs::NUM_REGS,
-               "Invalid MaxwellCompute register, increase the size of the Regs structure");
-
-    regs.reg_array[method_call.method] = method_call.argument;
-
-    switch (method_call.method) {
-    case MAXWELL_COMPUTE_REG_INDEX(compute): {
-        LOG_CRITICAL(HW_GPU, "Compute shaders are not implemented");
-        UNREACHABLE();
-        break;
-    }
-    default:
-        break;
-    }
-}
-
-} // namespace Tegra::Engines
diff --git a/src/video_core/engines/maxwell_dma.cpp b/src/video_core/engines/maxwell_dma.cpp
index 06462f570..a0ded4c25 100644
--- a/src/video_core/engines/maxwell_dma.cpp
+++ b/src/video_core/engines/maxwell_dma.cpp
@@ -2,17 +2,21 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#include "common/assert.h"
+#include "common/logging/log.h"
 #include "core/core.h"
 #include "core/memory.h"
 #include "video_core/engines/maxwell_3d.h"
 #include "video_core/engines/maxwell_dma.h"
 #include "video_core/rasterizer_interface.h"
+#include "video_core/renderer_base.h"
 #include "video_core/textures/decoders.h"
 
 namespace Tegra::Engines {
 
-MaxwellDMA::MaxwellDMA(VideoCore::RasterizerInterface& rasterizer, MemoryManager& memory_manager)
-    : memory_manager(memory_manager), rasterizer{rasterizer} {}
+MaxwellDMA::MaxwellDMA(Core::System& system, VideoCore::RasterizerInterface& rasterizer,
+                       MemoryManager& memory_manager)
+    : memory_manager(memory_manager), system{system}, rasterizer{rasterizer} {}
 
 void MaxwellDMA::CallMethod(const GPU::MethodCall& method_call) {
     ASSERT_MSG(method_call.method < Regs::NUM_REGS,
@@ -39,9 +43,6 @@ void MaxwellDMA::HandleCopy() {
     const GPUVAddr source = regs.src_address.Address();
     const GPUVAddr dest = regs.dst_address.Address();
 
-    const VAddr source_cpu = *memory_manager.GpuToCpuAddress(source);
-    const VAddr dest_cpu = *memory_manager.GpuToCpuAddress(dest);
-
     // TODO(Subv): Perform more research and implement all features of this engine.
     ASSERT(regs.exec.enable_swizzle == 0);
     ASSERT(regs.exec.query_mode == Regs::QueryMode::None);
@@ -57,14 +58,14 @@ void MaxwellDMA::HandleCopy() {
     }
 
     // All copies here update the main memory, so mark all rasterizer states as invalid.
-    Core::System::GetInstance().GPU().Maxwell3D().dirty_flags.OnMemoryWrite();
+    system.GPU().Maxwell3D().dirty_flags.OnMemoryWrite();
 
     if (regs.exec.is_dst_linear && regs.exec.is_src_linear) {
         // When the enable_2d bit is disabled, the copy is performed as if we were copying a 1D
         // buffer of length `x_count`, otherwise we copy a 2D image of dimensions (x_count,
         // y_count).
         if (!regs.exec.enable_2d) {
-            Memory::CopyBlock(dest_cpu, source_cpu, regs.x_count);
+            memory_manager.CopyBlock(dest, source, regs.x_count);
             return;
         }
 
@@ -73,9 +74,9 @@ void MaxwellDMA::HandleCopy() {
         // rectangle. There is no need to manually flush/invalidate the regions because
         // CopyBlock does that for us.
         for (u32 line = 0; line < regs.y_count; ++line) {
-            const VAddr source_line = source_cpu + line * regs.src_pitch;
-            const VAddr dest_line = dest_cpu + line * regs.dst_pitch;
-            Memory::CopyBlock(dest_line, source_line, regs.x_count);
+            const GPUVAddr source_line = source + line * regs.src_pitch;
+            const GPUVAddr dest_line = dest + line * regs.dst_pitch;
+            memory_manager.CopyBlock(dest_line, source_line, regs.x_count);
         }
         return;
     }
@@ -84,15 +85,18 @@ void MaxwellDMA::HandleCopy() {
 
     const std::size_t copy_size = regs.x_count * regs.y_count;
 
+    auto source_ptr{memory_manager.GetPointer(source)};
+    auto dst_ptr{memory_manager.GetPointer(dest)};
+
     const auto FlushAndInvalidate = [&](u32 src_size, u64 dst_size) {
         // TODO(Subv): For now, manually flush the regions until we implement GPU-accelerated
         // copying.
-        rasterizer.FlushRegion(source_cpu, src_size);
+        rasterizer.FlushRegion(ToCacheAddr(source_ptr), src_size);
 
         // We have to invalidate the destination region to evict any outdated surfaces from the
         // cache. We do this before actually writing the new data because the destination address
         // might contain a dirty surface that will have to be written back to memory.
-        rasterizer.InvalidateRegion(dest_cpu, dst_size);
+        rasterizer.InvalidateRegion(ToCacheAddr(dst_ptr), dst_size);
     };
 
     if (regs.exec.is_dst_linear && !regs.exec.is_src_linear) {
@@ -105,7 +109,7 @@ void MaxwellDMA::HandleCopy() {
                            copy_size * src_bytes_per_pixel);
 
         Texture::UnswizzleSubrect(regs.x_count, regs.y_count, regs.dst_pitch,
-                                  regs.src_params.size_x, src_bytes_per_pixel, source_cpu, dest_cpu,
+                                  regs.src_params.size_x, src_bytes_per_pixel, source_ptr, dst_ptr,
                                   regs.src_params.BlockHeight(), regs.src_params.pos_x,
                                   regs.src_params.pos_y);
     } else {
@@ -119,7 +123,7 @@ void MaxwellDMA::HandleCopy() {
 
         // If the input is linear and the output is tiled, swizzle the input and copy it over.
         Texture::SwizzleSubrect(regs.x_count, regs.y_count, regs.src_pitch, regs.dst_params.size_x,
-                                src_bpp, dest_cpu, source_cpu, regs.dst_params.BlockHeight());
+                                src_bpp, dst_ptr, source_ptr, regs.dst_params.BlockHeight());
     }
 }
 
diff --git a/src/video_core/engines/maxwell_dma.h b/src/video_core/engines/maxwell_dma.h
index 1f8cd65d2..34c369320 100644
--- a/src/video_core/engines/maxwell_dma.h
+++ b/src/video_core/engines/maxwell_dma.h
@@ -5,13 +5,17 @@
 #pragma once
 
 #include <array>
-#include "common/assert.h"
+#include <cstddef>
 #include "common/bit_field.h"
 #include "common/common_funcs.h"
 #include "common/common_types.h"
 #include "video_core/gpu.h"
 #include "video_core/memory_manager.h"
 
+namespace Core {
+class System;
+}
+
 namespace VideoCore {
 class RasterizerInterface;
 }
@@ -20,7 +24,8 @@ namespace Tegra::Engines {
 
 class MaxwellDMA final {
 public:
-    explicit MaxwellDMA(VideoCore::RasterizerInterface& rasterizer, MemoryManager& memory_manager);
+    explicit MaxwellDMA(Core::System& system, VideoCore::RasterizerInterface& rasterizer,
+                        MemoryManager& memory_manager);
     ~MaxwellDMA() = default;
 
     /// Write the value to the register identified by method.
@@ -137,6 +142,8 @@ public:
     MemoryManager& memory_manager;
 
 private:
+    Core::System& system;
+
     VideoCore::RasterizerInterface& rasterizer;
 
     /// Performs the copy from the source buffer to the destination buffer as configured in the
diff --git a/src/video_core/engines/shader_bytecode.h b/src/video_core/engines/shader_bytecode.h
index 269df9437..7f613370b 100644
--- a/src/video_core/engines/shader_bytecode.h
+++ b/src/video_core/engines/shader_bytecode.h
@@ -6,7 +6,6 @@
 
 #include <bitset>
 #include <optional>
-#include <string>
 #include <tuple>
 #include <vector>
 
@@ -186,7 +185,7 @@ enum class SubOp : u64 {
 };
 
 enum class F2iRoundingOp : u64 {
-    None = 0,
+    RoundEven = 0,
     Floor = 1,
     Ceil = 2,
     Trunc = 3,
@@ -325,11 +324,11 @@ enum class TextureQueryType : u64 {
 
 enum class TextureProcessMode : u64 {
     None = 0,
-    LZ = 1,  // Unknown, appears to be the same as none.
+    LZ = 1,  // Load LOD of zero.
     LB = 2,  // Load Bias.
-    LL = 3,  // Load LOD (LevelOfDetail)
-    LBA = 6, // Load Bias. The A is unknown, does not appear to differ with LB
-    LLA = 7  // Load LOD. The A is unknown, does not appear to differ with LL
+    LL = 3,  // Load LOD.
+    LBA = 6, // Load Bias. The A is unknown, does not appear to differ with LB.
+    LLA = 7  // Load LOD. The A is unknown, does not appear to differ with LL.
 };
 
 enum class TextureMiscMode : u64 {
@@ -376,9 +375,9 @@ enum class R2pMode : u64 {
 };
 
 enum class IpaInterpMode : u64 {
-    Linear = 0,
-    Perspective = 1,
-    Flat = 2,
+    Pass = 0,
+    Multiply = 1,
+    Constant = 2,
     Sc = 3,
 };
 
@@ -1446,6 +1445,7 @@ public:
         Flow,
         Synch,
         Memory,
+        Texture,
         FloatSet,
         FloatSetPredicate,
         IntegerSet,
@@ -1576,14 +1576,14 @@ private:
             INST("1110111101010---", Id::ST_L, Type::Memory, "ST_L"),
             INST("1110111011010---", Id::LDG, Type::Memory, "LDG"),
             INST("1110111011011---", Id::STG, Type::Memory, "STG"),
-            INST("110000----111---", Id::TEX, Type::Memory, "TEX"),
-            INST("1101111101001---", Id::TXQ, Type::Memory, "TXQ"),
-            INST("1101-00---------", Id::TEXS, Type::Memory, "TEXS"),
-            INST("1101101---------", Id::TLDS, Type::Memory, "TLDS"),
-            INST("110010----111---", Id::TLD4, Type::Memory, "TLD4"),
-            INST("1101111100------", Id::TLD4S, Type::Memory, "TLD4S"),
-            INST("110111110110----", Id::TMML_B, Type::Memory, "TMML_B"),
-            INST("1101111101011---", Id::TMML, Type::Memory, "TMML"),
+            INST("110000----111---", Id::TEX, Type::Texture, "TEX"),
+            INST("1101111101001---", Id::TXQ, Type::Texture, "TXQ"),
+            INST("1101-00---------", Id::TEXS, Type::Texture, "TEXS"),
+            INST("1101101---------", Id::TLDS, Type::Texture, "TLDS"),
+            INST("110010----111---", Id::TLD4, Type::Texture, "TLD4"),
+            INST("1101111100------", Id::TLD4S, Type::Texture, "TLD4S"),
+            INST("110111110110----", Id::TMML_B, Type::Texture, "TMML_B"),
+            INST("1101111101011---", Id::TMML, Type::Texture, "TMML"),
             INST("111000110000----", Id::EXIT, Type::Trivial, "EXIT"),
             INST("11100000--------", Id::IPA, Type::Trivial, "IPA"),
             INST("1111101111100---", Id::OUT_R, Type::Trivial, "OUT_R"),
diff --git a/src/video_core/engines/shader_header.h b/src/video_core/engines/shader_header.h
index cf2b76ff6..e86a7f04a 100644
--- a/src/video_core/engines/shader_header.h
+++ b/src/video_core/engines/shader_header.h
@@ -16,6 +16,13 @@ enum class OutputTopology : u32 {
     TriangleStrip = 7,
 };
 
+enum class AttributeUse : u8 {
+    Unused = 0,
+    Constant = 1,
+    Perspective = 2,
+    ScreenLinear = 3,
+};
+
 // Documentation in:
 // http://download.nvidia.com/open-gpu-doc/Shader-Program-Header/1/Shader-Program-Header.html#ImapTexture
 struct Header {
@@ -84,9 +91,15 @@ struct Header {
         } vtg;
 
         struct {
-            INSERT_PADDING_BYTES(3);  // ImapSystemValuesA
-            INSERT_PADDING_BYTES(1);  // ImapSystemValuesB
-            INSERT_PADDING_BYTES(32); // ImapGenericVector[32]
+            INSERT_PADDING_BYTES(3); // ImapSystemValuesA
+            INSERT_PADDING_BYTES(1); // ImapSystemValuesB
+            union {
+                BitField<0, 2, AttributeUse> x;
+                BitField<2, 2, AttributeUse> y;
+                BitField<4, 2, AttributeUse> w;
+                BitField<6, 2, AttributeUse> z;
+                u8 raw;
+            } imap_generic_vector[32];
             INSERT_PADDING_BYTES(2);  // ImapColor
             INSERT_PADDING_BYTES(2);  // ImapSystemValuesC
             INSERT_PADDING_BYTES(10); // ImapFixedFncTexture[10]
@@ -103,6 +116,28 @@ struct Header {
                 const u32 bit = render_target * 4 + component;
                 return omap.target & (1 << bit);
             }
+            AttributeUse GetAttributeIndexUse(u32 attribute, u32 index) const {
+                return static_cast<AttributeUse>(
+                    (imap_generic_vector[attribute].raw >> (index * 2)) & 0x03);
+            }
+            AttributeUse GetAttributeUse(u32 attribute) const {
+                AttributeUse result = AttributeUse::Unused;
+                for (u32 i = 0; i < 4; i++) {
+                    const auto index = GetAttributeIndexUse(attribute, i);
+                    if (index == AttributeUse::Unused) {
+                        continue;
+                    }
+                    if (result == AttributeUse::Unused || result == index) {
+                        result = index;
+                        continue;
+                    }
+                    LOG_CRITICAL(HW_GPU, "Generic Attribute Conflict in Interpolation Mode");
+                    if (index == AttributeUse::Perspective) {
+                        result = index;
+                    }
+                }
+                return result;
+            }
         } ps;
     };
 
diff --git a/src/video_core/gpu.cpp b/src/video_core/gpu.cpp
index d3d32a359..66c690494 100644
--- a/src/video_core/gpu.cpp
+++ b/src/video_core/gpu.cpp
@@ -3,21 +3,23 @@
 // Refer to the license.txt file included.
 
 #include "common/assert.h"
+#include "core/core.h"
 #include "core/core_timing.h"
 #include "core/memory.h"
 #include "video_core/engines/fermi_2d.h"
+#include "video_core/engines/kepler_compute.h"
 #include "video_core/engines/kepler_memory.h"
 #include "video_core/engines/maxwell_3d.h"
-#include "video_core/engines/maxwell_compute.h"
 #include "video_core/engines/maxwell_dma.h"
 #include "video_core/gpu.h"
-#include "video_core/rasterizer_interface.h"
+#include "video_core/renderer_base.h"
 
 namespace Tegra {
 
 u32 FramebufferConfig::BytesPerPixel(PixelFormat format) {
     switch (format) {
     case PixelFormat::ABGR8:
+    case PixelFormat::BGRA8:
         return 4;
     default:
         return 4;
@@ -26,14 +28,15 @@ u32 FramebufferConfig::BytesPerPixel(PixelFormat format) {
     UNREACHABLE();
 }
 
-GPU::GPU(VideoCore::RasterizerInterface& rasterizer) {
+GPU::GPU(Core::System& system, VideoCore::RendererBase& renderer) : renderer{renderer} {
+    auto& rasterizer{renderer.Rasterizer()};
     memory_manager = std::make_unique<Tegra::MemoryManager>();
     dma_pusher = std::make_unique<Tegra::DmaPusher>(*this);
-    maxwell_3d = std::make_unique<Engines::Maxwell3D>(rasterizer, *memory_manager);
+    maxwell_3d = std::make_unique<Engines::Maxwell3D>(system, rasterizer, *memory_manager);
     fermi_2d = std::make_unique<Engines::Fermi2D>(rasterizer, *memory_manager);
-    maxwell_compute = std::make_unique<Engines::MaxwellCompute>();
-    maxwell_dma = std::make_unique<Engines::MaxwellDMA>(rasterizer, *memory_manager);
-    kepler_memory = std::make_unique<Engines::KeplerMemory>(rasterizer, *memory_manager);
+    kepler_compute = std::make_unique<Engines::KeplerCompute>(*memory_manager);
+    maxwell_dma = std::make_unique<Engines::MaxwellDMA>(system, rasterizer, *memory_manager);
+    kepler_memory = std::make_unique<Engines::KeplerMemory>(system, rasterizer, *memory_manager);
 }
 
 GPU::~GPU() = default;
@@ -245,8 +248,8 @@ void GPU::CallEngineMethod(const MethodCall& method_call) {
     case EngineID::MAXWELL_B:
         maxwell_3d->CallMethod(method_call);
         break;
-    case EngineID::MAXWELL_COMPUTE_B:
-        maxwell_compute->CallMethod(method_call);
+    case EngineID::KEPLER_COMPUTE_B:
+        kepler_compute->CallMethod(method_call);
         break;
     case EngineID::MAXWELL_DMA_COPY_A:
         maxwell_dma->CallMethod(method_call);
@@ -271,7 +274,6 @@ void GPU::ProcessSemaphoreTriggerMethod() {
     const auto op =
         static_cast<GpuSemaphoreOperation>(regs.semaphore_trigger & semaphoreOperationMask);
     if (op == GpuSemaphoreOperation::WriteLong) {
-        auto address = memory_manager->GpuToCpuAddress(regs.smaphore_address.SmaphoreAddress());
         struct Block {
             u32 sequence;
             u32 zeros = 0;
@@ -282,12 +284,10 @@ void GPU::ProcessSemaphoreTriggerMethod() {
         block.sequence = regs.semaphore_sequence;
         // TODO(Kmather73): Generate a real GPU timestamp and write it here instead of
         // CoreTiming
-        block.timestamp = CoreTiming::GetTicks();
-        Memory::WriteBlock(*address, &block, sizeof(block));
+        block.timestamp = Core::System::GetInstance().CoreTiming().GetTicks();
+        memory_manager->WriteBlock(regs.smaphore_address.SmaphoreAddress(), &block, sizeof(block));
     } else {
-        const auto address =
-            memory_manager->GpuToCpuAddress(regs.smaphore_address.SmaphoreAddress());
-        const u32 word = Memory::Read32(*address);
+        const u32 word{memory_manager->Read32(regs.smaphore_address.SmaphoreAddress())};
         if ((op == GpuSemaphoreOperation::AcquireEqual && word == regs.semaphore_sequence) ||
             (op == GpuSemaphoreOperation::AcquireGequal &&
              static_cast<s32>(word - regs.semaphore_sequence) > 0) ||
@@ -314,13 +314,11 @@ void GPU::ProcessSemaphoreTriggerMethod() {
 }
 
 void GPU::ProcessSemaphoreRelease() {
-    const auto address = memory_manager->GpuToCpuAddress(regs.smaphore_address.SmaphoreAddress());
-    Memory::Write32(*address, regs.semaphore_release);
+    memory_manager->Write32(regs.smaphore_address.SmaphoreAddress(), regs.semaphore_release);
 }
 
 void GPU::ProcessSemaphoreAcquire() {
-    const auto address = memory_manager->GpuToCpuAddress(regs.smaphore_address.SmaphoreAddress());
-    const u32 word = Memory::Read32(*address);
+    const u32 word = memory_manager->Read32(regs.smaphore_address.SmaphoreAddress());
     const auto value = regs.semaphore_acquire;
     if (word != value) {
         regs.acquire_active = true;
diff --git a/src/video_core/gpu.h b/src/video_core/gpu.h
index fb8975811..a14b95c30 100644
--- a/src/video_core/gpu.h
+++ b/src/video_core/gpu.h
@@ -6,16 +6,24 @@
 
 #include <array>
 #include <memory>
-#include <vector>
 #include "common/common_types.h"
 #include "core/hle/service/nvflinger/buffer_queue.h"
 #include "video_core/dma_pusher.h"
 #include "video_core/memory_manager.h"
 
-namespace VideoCore {
-class RasterizerInterface;
+using CacheAddr = std::uintptr_t;
+inline CacheAddr ToCacheAddr(const void* host_ptr) {
+    return reinterpret_cast<CacheAddr>(host_ptr);
+}
+
+namespace Core {
+class System;
 }
 
+namespace VideoCore {
+class RendererBase;
+} // namespace VideoCore
+
 namespace Tegra {
 
 enum class RenderTargetFormat : u32 {
@@ -80,6 +88,7 @@ class DebugContext;
 struct FramebufferConfig {
     enum class PixelFormat : u32 {
         ABGR8 = 1,
+        BGRA8 = 5,
     };
 
     /**
@@ -96,29 +105,30 @@ struct FramebufferConfig {
 
     using TransformFlags = Service::NVFlinger::BufferQueue::BufferTransformFlags;
     TransformFlags transform_flags;
-    MathUtil::Rectangle<int> crop_rect;
+    Common::Rectangle<int> crop_rect;
 };
 
 namespace Engines {
 class Fermi2D;
 class Maxwell3D;
-class MaxwellCompute;
 class MaxwellDMA;
+class KeplerCompute;
 class KeplerMemory;
 } // namespace Engines
 
 enum class EngineID {
     FERMI_TWOD_A = 0x902D, // 2D Engine
     MAXWELL_B = 0xB197,    // 3D Engine
-    MAXWELL_COMPUTE_B = 0xB1C0,
+    KEPLER_COMPUTE_B = 0xB1C0,
     KEPLER_INLINE_TO_MEMORY_B = 0xA140,
     MAXWELL_DMA_COPY_A = 0xB0B5,
 };
 
-class GPU final {
+class GPU {
 public:
-    explicit GPU(VideoCore::RasterizerInterface& rasterizer);
-    ~GPU();
+    explicit GPU(Core::System& system, VideoCore::RendererBase& renderer);
+
+    virtual ~GPU();
 
     struct MethodCall {
         u32 method{};
@@ -196,8 +206,42 @@ public:
         };
     } regs{};
 
+    /// Push GPU command entries to be processed
+    virtual void PushGPUEntries(Tegra::CommandList&& entries) = 0;
+
+    /// Swap buffers (render frame)
+    virtual void SwapBuffers(
+        std::optional<std::reference_wrapper<const Tegra::FramebufferConfig>> framebuffer) = 0;
+
+    /// Notify rasterizer that any caches of the specified region should be flushed to Switch memory
+    virtual void FlushRegion(CacheAddr addr, u64 size) = 0;
+
+    /// Notify rasterizer that any caches of the specified region should be invalidated
+    virtual void InvalidateRegion(CacheAddr addr, u64 size) = 0;
+
+    /// Notify rasterizer that any caches of the specified region should be flushed and invalidated
+    virtual void FlushAndInvalidateRegion(CacheAddr addr, u64 size) = 0;
+
 private:
+    void ProcessBindMethod(const MethodCall& method_call);
+    void ProcessSemaphoreTriggerMethod();
+    void ProcessSemaphoreRelease();
+    void ProcessSemaphoreAcquire();
+
+    /// Calls a GPU puller method.
+    void CallPullerMethod(const MethodCall& method_call);
+
+    /// Calls a GPU engine method.
+    void CallEngineMethod(const MethodCall& method_call);
+
+    /// Determines where the method should be executed.
+    bool ExecuteMethodOnEngine(const MethodCall& method_call);
+
+protected:
     std::unique_ptr<Tegra::DmaPusher> dma_pusher;
+    VideoCore::RendererBase& renderer;
+
+private:
     std::unique_ptr<Tegra::MemoryManager> memory_manager;
 
     /// Mapping of command subchannels to their bound engine ids.
@@ -208,23 +252,11 @@ private:
     /// 2D engine
     std::unique_ptr<Engines::Fermi2D> fermi_2d;
     /// Compute engine
-    std::unique_ptr<Engines::MaxwellCompute> maxwell_compute;
+    std::unique_ptr<Engines::KeplerCompute> kepler_compute;
     /// DMA engine
     std::unique_ptr<Engines::MaxwellDMA> maxwell_dma;
     /// Inline memory engine
     std::unique_ptr<Engines::KeplerMemory> kepler_memory;
-
-    void ProcessBindMethod(const MethodCall& method_call);
-    void ProcessSemaphoreTriggerMethod();
-    void ProcessSemaphoreRelease();
-    void ProcessSemaphoreAcquire();
-
-    // Calls a GPU puller method.
-    void CallPullerMethod(const MethodCall& method_call);
-    // Calls a GPU engine method.
-    void CallEngineMethod(const MethodCall& method_call);
-    // Determines where the method should be executed.
-    bool ExecuteMethodOnEngine(const MethodCall& method_call);
 };
 
 #define ASSERT_REG_POSITION(field_name, position)                                                  \
diff --git a/src/video_core/gpu_asynch.cpp b/src/video_core/gpu_asynch.cpp
new file mode 100644
index 000000000..8b355cf7b
--- /dev/null
+++ b/src/video_core/gpu_asynch.cpp
@@ -0,0 +1,37 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "video_core/gpu_asynch.h"
+#include "video_core/gpu_thread.h"
+#include "video_core/renderer_base.h"
+
+namespace VideoCommon {
+
+GPUAsynch::GPUAsynch(Core::System& system, VideoCore::RendererBase& renderer)
+    : Tegra::GPU(system, renderer), gpu_thread{renderer, *dma_pusher} {}
+
+GPUAsynch::~GPUAsynch() = default;
+
+void GPUAsynch::PushGPUEntries(Tegra::CommandList&& entries) {
+    gpu_thread.SubmitList(std::move(entries));
+}
+
+void GPUAsynch::SwapBuffers(
+    std::optional<std::reference_wrapper<const Tegra::FramebufferConfig>> framebuffer) {
+    gpu_thread.SwapBuffers(std::move(framebuffer));
+}
+
+void GPUAsynch::FlushRegion(CacheAddr addr, u64 size) {
+    gpu_thread.FlushRegion(addr, size);
+}
+
+void GPUAsynch::InvalidateRegion(CacheAddr addr, u64 size) {
+    gpu_thread.InvalidateRegion(addr, size);
+}
+
+void GPUAsynch::FlushAndInvalidateRegion(CacheAddr addr, u64 size) {
+    gpu_thread.FlushAndInvalidateRegion(addr, size);
+}
+
+} // namespace VideoCommon
diff --git a/src/video_core/gpu_asynch.h b/src/video_core/gpu_asynch.h
new file mode 100644
index 000000000..1dcc61a6c
--- /dev/null
+++ b/src/video_core/gpu_asynch.h
@@ -0,0 +1,37 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "video_core/gpu.h"
+#include "video_core/gpu_thread.h"
+
+namespace VideoCore {
+class RendererBase;
+} // namespace VideoCore
+
+namespace VideoCommon {
+
+namespace GPUThread {
+class ThreadManager;
+} // namespace GPUThread
+
+/// Implementation of GPU interface that runs the GPU asynchronously
+class GPUAsynch : public Tegra::GPU {
+public:
+    explicit GPUAsynch(Core::System& system, VideoCore::RendererBase& renderer);
+    ~GPUAsynch() override;
+
+    void PushGPUEntries(Tegra::CommandList&& entries) override;
+    void SwapBuffers(
+        std::optional<std::reference_wrapper<const Tegra::FramebufferConfig>> framebuffer) override;
+    void FlushRegion(CacheAddr addr, u64 size) override;
+    void InvalidateRegion(CacheAddr addr, u64 size) override;
+    void FlushAndInvalidateRegion(CacheAddr addr, u64 size) override;
+
+private:
+    GPUThread::ThreadManager gpu_thread;
+};
+
+} // namespace VideoCommon
diff --git a/src/video_core/gpu_synch.cpp b/src/video_core/gpu_synch.cpp
new file mode 100644
index 000000000..2cfc900ed
--- /dev/null
+++ b/src/video_core/gpu_synch.cpp
@@ -0,0 +1,37 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "video_core/gpu_synch.h"
+#include "video_core/renderer_base.h"
+
+namespace VideoCommon {
+
+GPUSynch::GPUSynch(Core::System& system, VideoCore::RendererBase& renderer)
+    : Tegra::GPU(system, renderer) {}
+
+GPUSynch::~GPUSynch() = default;
+
+void GPUSynch::PushGPUEntries(Tegra::CommandList&& entries) {
+    dma_pusher->Push(std::move(entries));
+    dma_pusher->DispatchCalls();
+}
+
+void GPUSynch::SwapBuffers(
+    std::optional<std::reference_wrapper<const Tegra::FramebufferConfig>> framebuffer) {
+    renderer.SwapBuffers(std::move(framebuffer));
+}
+
+void GPUSynch::FlushRegion(CacheAddr addr, u64 size) {
+    renderer.Rasterizer().FlushRegion(addr, size);
+}
+
+void GPUSynch::InvalidateRegion(CacheAddr addr, u64 size) {
+    renderer.Rasterizer().InvalidateRegion(addr, size);
+}
+
+void GPUSynch::FlushAndInvalidateRegion(CacheAddr addr, u64 size) {
+    renderer.Rasterizer().FlushAndInvalidateRegion(addr, size);
+}
+
+} // namespace VideoCommon
diff --git a/src/video_core/gpu_synch.h b/src/video_core/gpu_synch.h
new file mode 100644
index 000000000..766b5631c
--- /dev/null
+++ b/src/video_core/gpu_synch.h
@@ -0,0 +1,29 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "video_core/gpu.h"
+
+namespace VideoCore {
+class RendererBase;
+} // namespace VideoCore
+
+namespace VideoCommon {
+
+/// Implementation of GPU interface that runs the GPU synchronously
+class GPUSynch : public Tegra::GPU {
+public:
+    explicit GPUSynch(Core::System& system, VideoCore::RendererBase& renderer);
+    ~GPUSynch() override;
+
+    void PushGPUEntries(Tegra::CommandList&& entries) override;
+    void SwapBuffers(
+        std::optional<std::reference_wrapper<const Tegra::FramebufferConfig>> framebuffer) override;
+    void FlushRegion(CacheAddr addr, u64 size) override;
+    void InvalidateRegion(CacheAddr addr, u64 size) override;
+    void FlushAndInvalidateRegion(CacheAddr addr, u64 size) override;
+};
+
+} // namespace VideoCommon
diff --git a/src/video_core/gpu_thread.cpp b/src/video_core/gpu_thread.cpp
new file mode 100644
index 000000000..086b2f625
--- /dev/null
+++ b/src/video_core/gpu_thread.cpp
@@ -0,0 +1,98 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/assert.h"
+#include "common/microprofile.h"
+#include "core/frontend/scope_acquire_window_context.h"
+#include "video_core/dma_pusher.h"
+#include "video_core/gpu.h"
+#include "video_core/gpu_thread.h"
+#include "video_core/renderer_base.h"
+
+namespace VideoCommon::GPUThread {
+
+/// Runs the GPU thread
+static void RunThread(VideoCore::RendererBase& renderer, Tegra::DmaPusher& dma_pusher,
+                      SynchState& state) {
+    MicroProfileOnThreadCreate("GpuThread");
+
+    // Wait for first GPU command before acquiring the window context
+    state.WaitForCommands();
+
+    // If emulation was stopped during disk shader loading, abort before trying to acquire context
+    if (!state.is_running) {
+        return;
+    }
+
+    Core::Frontend::ScopeAcquireWindowContext acquire_context{renderer.GetRenderWindow()};
+
+    CommandDataContainer next;
+    while (state.is_running) {
+        state.WaitForCommands();
+        while (!state.queue.Empty()) {
+            state.queue.Pop(next);
+            if (const auto submit_list = std::get_if<SubmitListCommand>(&next.data)) {
+                dma_pusher.Push(std::move(submit_list->entries));
+                dma_pusher.DispatchCalls();
+            } else if (const auto data = std::get_if<SwapBuffersCommand>(&next.data)) {
+                state.DecrementFramesCounter();
+                renderer.SwapBuffers(std::move(data->framebuffer));
+            } else if (const auto data = std::get_if<FlushRegionCommand>(&next.data)) {
+                renderer.Rasterizer().FlushRegion(data->addr, data->size);
+            } else if (const auto data = std::get_if<InvalidateRegionCommand>(&next.data)) {
+                renderer.Rasterizer().InvalidateRegion(data->addr, data->size);
+            } else if (const auto data = std::get_if<EndProcessingCommand>(&next.data)) {
+                return;
+            } else {
+                UNREACHABLE();
+            }
+        }
+    }
+}
+
+ThreadManager::ThreadManager(VideoCore::RendererBase& renderer, Tegra::DmaPusher& dma_pusher)
+    : renderer{renderer}, dma_pusher{dma_pusher}, thread{RunThread, std::ref(renderer),
+                                                         std::ref(dma_pusher), std::ref(state)} {}
+
+ThreadManager::~ThreadManager() {
+    // Notify GPU thread that a shutdown is pending
+    PushCommand(EndProcessingCommand());
+    thread.join();
+}
+
+void ThreadManager::SubmitList(Tegra::CommandList&& entries) {
+    PushCommand(SubmitListCommand(std::move(entries)));
+}
+
+void ThreadManager::SwapBuffers(
+    std::optional<std::reference_wrapper<const Tegra::FramebufferConfig>> framebuffer) {
+    state.IncrementFramesCounter();
+    PushCommand(SwapBuffersCommand(std::move(framebuffer)));
+    state.WaitForFrames();
+}
+
+void ThreadManager::FlushRegion(CacheAddr addr, u64 size) {
+    PushCommand(FlushRegionCommand(addr, size));
+}
+
+void ThreadManager::InvalidateRegion(CacheAddr addr, u64 size) {
+    if (state.queue.Empty()) {
+        // It's quicker to invalidate a single region on the CPU if the queue is already empty
+        renderer.Rasterizer().InvalidateRegion(addr, size);
+    } else {
+        PushCommand(InvalidateRegionCommand(addr, size));
+    }
+}
+
+void ThreadManager::FlushAndInvalidateRegion(CacheAddr addr, u64 size) {
+    // Skip flush on asynch mode, as FlushAndInvalidateRegion is not used for anything too important
+    InvalidateRegion(addr, size);
+}
+
+void ThreadManager::PushCommand(CommandData&& command_data) {
+    state.queue.Push(CommandDataContainer(std::move(command_data)));
+    state.SignalCommands();
+}
+
+} // namespace VideoCommon::GPUThread
diff --git a/src/video_core/gpu_thread.h b/src/video_core/gpu_thread.h
new file mode 100644
index 000000000..8cd7db1c6
--- /dev/null
+++ b/src/video_core/gpu_thread.h
@@ -0,0 +1,185 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <array>
+#include <atomic>
+#include <condition_variable>
+#include <memory>
+#include <mutex>
+#include <optional>
+#include <thread>
+#include <variant>
+
+#include "common/threadsafe_queue.h"
+#include "video_core/gpu.h"
+
+namespace Tegra {
+struct FramebufferConfig;
+class DmaPusher;
+} // namespace Tegra
+
+namespace VideoCore {
+class RendererBase;
+} // namespace VideoCore
+
+namespace VideoCommon::GPUThread {
+
+/// Command to signal to the GPU thread that processing has ended
+struct EndProcessingCommand final {};
+
+/// Command to signal to the GPU thread that a command list is ready for processing
+struct SubmitListCommand final {
+    explicit SubmitListCommand(Tegra::CommandList&& entries) : entries{std::move(entries)} {}
+
+    Tegra::CommandList entries;
+};
+
+/// Command to signal to the GPU thread that a swap buffers is pending
+struct SwapBuffersCommand final {
+    explicit SwapBuffersCommand(std::optional<const Tegra::FramebufferConfig> framebuffer)
+        : framebuffer{std::move(framebuffer)} {}
+
+    std::optional<Tegra::FramebufferConfig> framebuffer;
+};
+
+/// Command to signal to the GPU thread to flush a region
+struct FlushRegionCommand final {
+    explicit constexpr FlushRegionCommand(CacheAddr addr, u64 size) : addr{addr}, size{size} {}
+
+    CacheAddr addr;
+    u64 size;
+};
+
+/// Command to signal to the GPU thread to invalidate a region
+struct InvalidateRegionCommand final {
+    explicit constexpr InvalidateRegionCommand(CacheAddr addr, u64 size) : addr{addr}, size{size} {}
+
+    CacheAddr addr;
+    u64 size;
+};
+
+/// Command to signal to the GPU thread to flush and invalidate a region
+struct FlushAndInvalidateRegionCommand final {
+    explicit constexpr FlushAndInvalidateRegionCommand(CacheAddr addr, u64 size)
+        : addr{addr}, size{size} {}
+
+    CacheAddr addr;
+    u64 size;
+};
+
+using CommandData =
+    std::variant<EndProcessingCommand, SubmitListCommand, SwapBuffersCommand, FlushRegionCommand,
+                 InvalidateRegionCommand, FlushAndInvalidateRegionCommand>;
+
+struct CommandDataContainer {
+    CommandDataContainer() = default;
+
+    CommandDataContainer(CommandData&& data) : data{std::move(data)} {}
+
+    CommandDataContainer& operator=(const CommandDataContainer& t) {
+        data = std::move(t.data);
+        return *this;
+    }
+
+    CommandData data;
+};
+
+/// Struct used to synchronize the GPU thread
+struct SynchState final {
+    std::atomic_bool is_running{true};
+    std::atomic_int queued_frame_count{};
+    std::mutex frames_mutex;
+    std::mutex commands_mutex;
+    std::condition_variable commands_condition;
+    std::condition_variable frames_condition;
+
+    void IncrementFramesCounter() {
+        std::lock_guard<std::mutex> lock{frames_mutex};
+        ++queued_frame_count;
+    }
+
+    void DecrementFramesCounter() {
+        {
+            std::lock_guard<std::mutex> lock{frames_mutex};
+            --queued_frame_count;
+
+            if (queued_frame_count) {
+                return;
+            }
+        }
+        frames_condition.notify_one();
+    }
+
+    void WaitForFrames() {
+        {
+            std::lock_guard<std::mutex> lock{frames_mutex};
+            if (!queued_frame_count) {
+                return;
+            }
+        }
+
+        // Wait for the GPU to be idle (all commands to be executed)
+        {
+            std::unique_lock<std::mutex> lock{frames_mutex};
+            frames_condition.wait(lock, [this] { return !queued_frame_count; });
+        }
+    }
+
+    void SignalCommands() {
+        {
+            std::unique_lock<std::mutex> lock{commands_mutex};
+            if (queue.Empty()) {
+                return;
+            }
+        }
+
+        commands_condition.notify_one();
+    }
+
+    void WaitForCommands() {
+        std::unique_lock<std::mutex> lock{commands_mutex};
+        commands_condition.wait(lock, [this] { return !queue.Empty(); });
+    }
+
+    using CommandQueue = Common::SPSCQueue<CommandDataContainer>;
+    CommandQueue queue;
+};
+
+/// Class used to manage the GPU thread
+class ThreadManager final {
+public:
+    explicit ThreadManager(VideoCore::RendererBase& renderer, Tegra::DmaPusher& dma_pusher);
+    ~ThreadManager();
+
+    /// Push GPU command entries to be processed
+    void SubmitList(Tegra::CommandList&& entries);
+
+    /// Swap buffers (render frame)
+    void SwapBuffers(
+        std::optional<std::reference_wrapper<const Tegra::FramebufferConfig>> framebuffer);
+
+    /// Notify rasterizer that any caches of the specified region should be flushed to Switch memory
+    void FlushRegion(CacheAddr addr, u64 size);
+
+    /// Notify rasterizer that any caches of the specified region should be invalidated
+    void InvalidateRegion(CacheAddr addr, u64 size);
+
+    /// Notify rasterizer that any caches of the specified region should be flushed and invalidated
+    void FlushAndInvalidateRegion(CacheAddr addr, u64 size);
+
+private:
+    /// Pushes a command to be executed by the GPU thread
+    void PushCommand(CommandData&& command_data);
+
+private:
+    SynchState state;
+    VideoCore::RendererBase& renderer;
+    Tegra::DmaPusher& dma_pusher;
+    std::thread thread;
+    std::thread::id thread_id;
+};
+
+} // namespace VideoCommon::GPUThread
diff --git a/src/video_core/memory_manager.cpp b/src/video_core/memory_manager.cpp
index 47247f097..8e8f36f28 100644
--- a/src/video_core/memory_manager.cpp
+++ b/src/video_core/memory_manager.cpp
@@ -5,6 +5,7 @@
 #include "common/alignment.h"
 #include "common/assert.h"
 #include "common/logging/log.h"
+#include "core/memory.h"
 #include "video_core/memory_manager.h"
 
 namespace Tegra {
@@ -154,22 +155,59 @@ std::optional<VAddr> MemoryManager::GpuToCpuAddress(GPUVAddr gpu_addr) {
     const VAddr base_addr{PageSlot(gpu_addr)};
 
     if (base_addr == static_cast<u64>(PageStatus::Allocated) ||
-        base_addr == static_cast<u64>(PageStatus::Unmapped)) {
+        base_addr == static_cast<u64>(PageStatus::Unmapped) ||
+        base_addr == static_cast<u64>(PageStatus::Reserved)) {
         return {};
     }
 
     return base_addr + (gpu_addr & PAGE_MASK);
 }
 
-std::vector<GPUVAddr> MemoryManager::CpuToGpuAddress(VAddr cpu_addr) const {
-    std::vector<GPUVAddr> results;
-    for (const auto& region : mapped_regions) {
-        if (cpu_addr >= region.cpu_addr && cpu_addr < (region.cpu_addr + region.size)) {
-            const u64 offset{cpu_addr - region.cpu_addr};
-            results.push_back(region.gpu_addr + offset);
-        }
-    }
-    return results;
+u8 MemoryManager::Read8(GPUVAddr addr) {
+    return Memory::Read8(*GpuToCpuAddress(addr));
+}
+
+u16 MemoryManager::Read16(GPUVAddr addr) {
+    return Memory::Read16(*GpuToCpuAddress(addr));
+}
+
+u32 MemoryManager::Read32(GPUVAddr addr) {
+    return Memory::Read32(*GpuToCpuAddress(addr));
+}
+
+u64 MemoryManager::Read64(GPUVAddr addr) {
+    return Memory::Read64(*GpuToCpuAddress(addr));
+}
+
+void MemoryManager::Write8(GPUVAddr addr, u8 data) {
+    Memory::Write8(*GpuToCpuAddress(addr), data);
+}
+
+void MemoryManager::Write16(GPUVAddr addr, u16 data) {
+    Memory::Write16(*GpuToCpuAddress(addr), data);
+}
+
+void MemoryManager::Write32(GPUVAddr addr, u32 data) {
+    Memory::Write32(*GpuToCpuAddress(addr), data);
+}
+
+void MemoryManager::Write64(GPUVAddr addr, u64 data) {
+    Memory::Write64(*GpuToCpuAddress(addr), data);
+}
+
+u8* MemoryManager::GetPointer(GPUVAddr addr) {
+    return Memory::GetPointer(*GpuToCpuAddress(addr));
+}
+
+void MemoryManager::ReadBlock(GPUVAddr src_addr, void* dest_buffer, std::size_t size) {
+    std::memcpy(dest_buffer, GetPointer(src_addr), size);
+}
+void MemoryManager::WriteBlock(GPUVAddr dest_addr, const void* src_buffer, std::size_t size) {
+    std::memcpy(GetPointer(dest_addr), src_buffer, size);
+}
+
+void MemoryManager::CopyBlock(GPUVAddr dest_addr, GPUVAddr src_addr, std::size_t size) {
+    std::memcpy(GetPointer(dest_addr), GetPointer(src_addr), size);
 }
 
 VAddr& MemoryManager::PageSlot(GPUVAddr gpu_addr) {
diff --git a/src/video_core/memory_manager.h b/src/video_core/memory_manager.h
index fb03497ca..425e2f31c 100644
--- a/src/video_core/memory_manager.h
+++ b/src/video_core/memory_manager.h
@@ -27,12 +27,27 @@ public:
     GPUVAddr UnmapBuffer(GPUVAddr gpu_addr, u64 size);
     GPUVAddr GetRegionEnd(GPUVAddr region_start) const;
     std::optional<VAddr> GpuToCpuAddress(GPUVAddr gpu_addr);
-    std::vector<GPUVAddr> CpuToGpuAddress(VAddr cpu_addr) const;
 
     static constexpr u64 PAGE_BITS = 16;
     static constexpr u64 PAGE_SIZE = 1 << PAGE_BITS;
     static constexpr u64 PAGE_MASK = PAGE_SIZE - 1;
 
+    u8 Read8(GPUVAddr addr);
+    u16 Read16(GPUVAddr addr);
+    u32 Read32(GPUVAddr addr);
+    u64 Read64(GPUVAddr addr);
+
+    void Write8(GPUVAddr addr, u8 data);
+    void Write16(GPUVAddr addr, u16 data);
+    void Write32(GPUVAddr addr, u32 data);
+    void Write64(GPUVAddr addr, u64 data);
+
+    u8* GetPointer(GPUVAddr vaddr);
+
+    void ReadBlock(GPUVAddr src_addr, void* dest_buffer, std::size_t size);
+    void WriteBlock(GPUVAddr dest_addr, const void* src_buffer, std::size_t size);
+    void CopyBlock(VAddr dest_addr, VAddr src_addr, std::size_t size);
+
 private:
     enum class PageStatus : u64 {
         Unmapped = 0xFFFFFFFFFFFFFFFFULL,
diff --git a/src/video_core/morton.cpp b/src/video_core/morton.cpp
index b68f4fb13..3e91cbc83 100644
--- a/src/video_core/morton.cpp
+++ b/src/video_core/morton.cpp
@@ -6,7 +6,6 @@
 #include <cstring>
 #include "common/assert.h"
 #include "common/common_types.h"
-#include "core/memory.h"
 #include "video_core/morton.h"
 #include "video_core/surface.h"
 #include "video_core/textures/decoders.h"
@@ -16,12 +15,12 @@ namespace VideoCore {
 using Surface::GetBytesPerPixel;
 using Surface::PixelFormat;
 
-using MortonCopyFn = void (*)(u32, u32, u32, u32, u32, u32, u8*, std::size_t, VAddr);
+using MortonCopyFn = void (*)(u32, u32, u32, u32, u32, u32, u8*, u8*);
 using ConversionArray = std::array<MortonCopyFn, Surface::MaxPixelFormat>;
 
 template <bool morton_to_linear, PixelFormat format>
 static void MortonCopy(u32 stride, u32 block_height, u32 height, u32 block_depth, u32 depth,
-                       u32 tile_width_spacing, u8* buffer, std::size_t buffer_size, VAddr addr) {
+                       u32 tile_width_spacing, u8* buffer, u8* addr) {
     constexpr u32 bytes_per_pixel = GetBytesPerPixel(format);
 
     // With the BCn formats (DXT and DXN), each 4x4 tile is swizzled instead of just individual
@@ -34,150 +33,146 @@ static void MortonCopy(u32 stride, u32 block_height, u32 height, u32 block_depth
                                          stride, height, depth, block_height, block_depth,
                                          tile_width_spacing);
     } else {
-        Tegra::Texture::CopySwizzledData(
-            (stride + tile_size_x - 1) / tile_size_x, (height + tile_size_y - 1) / tile_size_y,
-            depth, bytes_per_pixel, bytes_per_pixel, Memory::GetPointer(addr), buffer, false,
-            block_height, block_depth, tile_width_spacing);
+        Tegra::Texture::CopySwizzledData((stride + tile_size_x - 1) / tile_size_x,
+                                         (height + tile_size_y - 1) / tile_size_y, depth,
+                                         bytes_per_pixel, bytes_per_pixel, addr, buffer, false,
+                                         block_height, block_depth, tile_width_spacing);
     }
 }
 
 static constexpr ConversionArray morton_to_linear_fns = {
-    // clang-format off
-        MortonCopy<true, PixelFormat::ABGR8U>,
-        MortonCopy<true, PixelFormat::ABGR8S>,
-        MortonCopy<true, PixelFormat::ABGR8UI>,
-        MortonCopy<true, PixelFormat::B5G6R5U>,
-        MortonCopy<true, PixelFormat::A2B10G10R10U>,
-        MortonCopy<true, PixelFormat::A1B5G5R5U>,
-        MortonCopy<true, PixelFormat::R8U>,
-        MortonCopy<true, PixelFormat::R8UI>,
-        MortonCopy<true, PixelFormat::RGBA16F>,
-        MortonCopy<true, PixelFormat::RGBA16U>,
-        MortonCopy<true, PixelFormat::RGBA16UI>,
-        MortonCopy<true, PixelFormat::R11FG11FB10F>,
-        MortonCopy<true, PixelFormat::RGBA32UI>,
-        MortonCopy<true, PixelFormat::DXT1>,
-        MortonCopy<true, PixelFormat::DXT23>,
-        MortonCopy<true, PixelFormat::DXT45>,
-        MortonCopy<true, PixelFormat::DXN1>,
-        MortonCopy<true, PixelFormat::DXN2UNORM>,
-        MortonCopy<true, PixelFormat::DXN2SNORM>,
-        MortonCopy<true, PixelFormat::BC7U>,
-        MortonCopy<true, PixelFormat::BC6H_UF16>,
-        MortonCopy<true, PixelFormat::BC6H_SF16>,
-        MortonCopy<true, PixelFormat::ASTC_2D_4X4>,
-        MortonCopy<true, PixelFormat::BGRA8>,
-        MortonCopy<true, PixelFormat::RGBA32F>,
-        MortonCopy<true, PixelFormat::RG32F>,
-        MortonCopy<true, PixelFormat::R32F>,
-        MortonCopy<true, PixelFormat::R16F>,
-        MortonCopy<true, PixelFormat::R16U>,
-        MortonCopy<true, PixelFormat::R16S>,
-        MortonCopy<true, PixelFormat::R16UI>,
-        MortonCopy<true, PixelFormat::R16I>,
-        MortonCopy<true, PixelFormat::RG16>,
-        MortonCopy<true, PixelFormat::RG16F>,
-        MortonCopy<true, PixelFormat::RG16UI>,
-        MortonCopy<true, PixelFormat::RG16I>,
-        MortonCopy<true, PixelFormat::RG16S>,
-        MortonCopy<true, PixelFormat::RGB32F>,
-        MortonCopy<true, PixelFormat::RGBA8_SRGB>,
-        MortonCopy<true, PixelFormat::RG8U>,
-        MortonCopy<true, PixelFormat::RG8S>,
-        MortonCopy<true, PixelFormat::RG32UI>,
-        MortonCopy<true, PixelFormat::R32UI>,
-        MortonCopy<true, PixelFormat::ASTC_2D_8X8>,
-        MortonCopy<true, PixelFormat::ASTC_2D_8X5>,
-        MortonCopy<true, PixelFormat::ASTC_2D_5X4>,
-        MortonCopy<true, PixelFormat::BGRA8_SRGB>,
-        MortonCopy<true, PixelFormat::DXT1_SRGB>,
-        MortonCopy<true, PixelFormat::DXT23_SRGB>,
-        MortonCopy<true, PixelFormat::DXT45_SRGB>,
-        MortonCopy<true, PixelFormat::BC7U_SRGB>,
-        MortonCopy<true, PixelFormat::ASTC_2D_4X4_SRGB>,
-        MortonCopy<true, PixelFormat::ASTC_2D_8X8_SRGB>,
-        MortonCopy<true, PixelFormat::ASTC_2D_8X5_SRGB>,
-        MortonCopy<true, PixelFormat::ASTC_2D_5X4_SRGB>,
-        MortonCopy<true, PixelFormat::ASTC_2D_5X5>,
-        MortonCopy<true, PixelFormat::ASTC_2D_5X5_SRGB>,
-        MortonCopy<true, PixelFormat::ASTC_2D_10X8>,
-        MortonCopy<true, PixelFormat::ASTC_2D_10X8_SRGB>,
-        MortonCopy<true, PixelFormat::Z32F>,
-        MortonCopy<true, PixelFormat::Z16>,
-        MortonCopy<true, PixelFormat::Z24S8>,
-        MortonCopy<true, PixelFormat::S8Z24>,
-        MortonCopy<true, PixelFormat::Z32FS8>,
-    // clang-format on
+    MortonCopy<true, PixelFormat::ABGR8U>,
+    MortonCopy<true, PixelFormat::ABGR8S>,
+    MortonCopy<true, PixelFormat::ABGR8UI>,
+    MortonCopy<true, PixelFormat::B5G6R5U>,
+    MortonCopy<true, PixelFormat::A2B10G10R10U>,
+    MortonCopy<true, PixelFormat::A1B5G5R5U>,
+    MortonCopy<true, PixelFormat::R8U>,
+    MortonCopy<true, PixelFormat::R8UI>,
+    MortonCopy<true, PixelFormat::RGBA16F>,
+    MortonCopy<true, PixelFormat::RGBA16U>,
+    MortonCopy<true, PixelFormat::RGBA16UI>,
+    MortonCopy<true, PixelFormat::R11FG11FB10F>,
+    MortonCopy<true, PixelFormat::RGBA32UI>,
+    MortonCopy<true, PixelFormat::DXT1>,
+    MortonCopy<true, PixelFormat::DXT23>,
+    MortonCopy<true, PixelFormat::DXT45>,
+    MortonCopy<true, PixelFormat::DXN1>,
+    MortonCopy<true, PixelFormat::DXN2UNORM>,
+    MortonCopy<true, PixelFormat::DXN2SNORM>,
+    MortonCopy<true, PixelFormat::BC7U>,
+    MortonCopy<true, PixelFormat::BC6H_UF16>,
+    MortonCopy<true, PixelFormat::BC6H_SF16>,
+    MortonCopy<true, PixelFormat::ASTC_2D_4X4>,
+    MortonCopy<true, PixelFormat::BGRA8>,
+    MortonCopy<true, PixelFormat::RGBA32F>,
+    MortonCopy<true, PixelFormat::RG32F>,
+    MortonCopy<true, PixelFormat::R32F>,
+    MortonCopy<true, PixelFormat::R16F>,
+    MortonCopy<true, PixelFormat::R16U>,
+    MortonCopy<true, PixelFormat::R16S>,
+    MortonCopy<true, PixelFormat::R16UI>,
+    MortonCopy<true, PixelFormat::R16I>,
+    MortonCopy<true, PixelFormat::RG16>,
+    MortonCopy<true, PixelFormat::RG16F>,
+    MortonCopy<true, PixelFormat::RG16UI>,
+    MortonCopy<true, PixelFormat::RG16I>,
+    MortonCopy<true, PixelFormat::RG16S>,
+    MortonCopy<true, PixelFormat::RGB32F>,
+    MortonCopy<true, PixelFormat::RGBA8_SRGB>,
+    MortonCopy<true, PixelFormat::RG8U>,
+    MortonCopy<true, PixelFormat::RG8S>,
+    MortonCopy<true, PixelFormat::RG32UI>,
+    MortonCopy<true, PixelFormat::R32UI>,
+    MortonCopy<true, PixelFormat::ASTC_2D_8X8>,
+    MortonCopy<true, PixelFormat::ASTC_2D_8X5>,
+    MortonCopy<true, PixelFormat::ASTC_2D_5X4>,
+    MortonCopy<true, PixelFormat::BGRA8_SRGB>,
+    MortonCopy<true, PixelFormat::DXT1_SRGB>,
+    MortonCopy<true, PixelFormat::DXT23_SRGB>,
+    MortonCopy<true, PixelFormat::DXT45_SRGB>,
+    MortonCopy<true, PixelFormat::BC7U_SRGB>,
+    MortonCopy<true, PixelFormat::ASTC_2D_4X4_SRGB>,
+    MortonCopy<true, PixelFormat::ASTC_2D_8X8_SRGB>,
+    MortonCopy<true, PixelFormat::ASTC_2D_8X5_SRGB>,
+    MortonCopy<true, PixelFormat::ASTC_2D_5X4_SRGB>,
+    MortonCopy<true, PixelFormat::ASTC_2D_5X5>,
+    MortonCopy<true, PixelFormat::ASTC_2D_5X5_SRGB>,
+    MortonCopy<true, PixelFormat::ASTC_2D_10X8>,
+    MortonCopy<true, PixelFormat::ASTC_2D_10X8_SRGB>,
+    MortonCopy<true, PixelFormat::Z32F>,
+    MortonCopy<true, PixelFormat::Z16>,
+    MortonCopy<true, PixelFormat::Z24S8>,
+    MortonCopy<true, PixelFormat::S8Z24>,
+    MortonCopy<true, PixelFormat::Z32FS8>,
 };
 
 static constexpr ConversionArray linear_to_morton_fns = {
-    // clang-format off
-        MortonCopy<false, PixelFormat::ABGR8U>,
-        MortonCopy<false, PixelFormat::ABGR8S>,
-        MortonCopy<false, PixelFormat::ABGR8UI>,
-        MortonCopy<false, PixelFormat::B5G6R5U>,
-        MortonCopy<false, PixelFormat::A2B10G10R10U>,
-        MortonCopy<false, PixelFormat::A1B5G5R5U>,
-        MortonCopy<false, PixelFormat::R8U>,
-        MortonCopy<false, PixelFormat::R8UI>,
-        MortonCopy<false, PixelFormat::RGBA16F>,
-        MortonCopy<false, PixelFormat::RGBA16U>,
-        MortonCopy<false, PixelFormat::RGBA16UI>,
-        MortonCopy<false, PixelFormat::R11FG11FB10F>,
-        MortonCopy<false, PixelFormat::RGBA32UI>,
-        MortonCopy<false, PixelFormat::DXT1>,
-        MortonCopy<false, PixelFormat::DXT23>,
-        MortonCopy<false, PixelFormat::DXT45>,
-        MortonCopy<false, PixelFormat::DXN1>,
-        MortonCopy<false, PixelFormat::DXN2UNORM>,
-        MortonCopy<false, PixelFormat::DXN2SNORM>,
-        MortonCopy<false, PixelFormat::BC7U>,
-        MortonCopy<false, PixelFormat::BC6H_UF16>,
-        MortonCopy<false, PixelFormat::BC6H_SF16>,
-        // TODO(Subv): Swizzling ASTC formats are not supported
-        nullptr,
-        MortonCopy<false, PixelFormat::BGRA8>,
-        MortonCopy<false, PixelFormat::RGBA32F>,
-        MortonCopy<false, PixelFormat::RG32F>,
-        MortonCopy<false, PixelFormat::R32F>,
-        MortonCopy<false, PixelFormat::R16F>,
-        MortonCopy<false, PixelFormat::R16U>,
-        MortonCopy<false, PixelFormat::R16S>,
-        MortonCopy<false, PixelFormat::R16UI>,
-        MortonCopy<false, PixelFormat::R16I>,
-        MortonCopy<false, PixelFormat::RG16>,
-        MortonCopy<false, PixelFormat::RG16F>,
-        MortonCopy<false, PixelFormat::RG16UI>,
-        MortonCopy<false, PixelFormat::RG16I>,
-        MortonCopy<false, PixelFormat::RG16S>,
-        MortonCopy<false, PixelFormat::RGB32F>,
-        MortonCopy<false, PixelFormat::RGBA8_SRGB>,
-        MortonCopy<false, PixelFormat::RG8U>,
-        MortonCopy<false, PixelFormat::RG8S>,
-        MortonCopy<false, PixelFormat::RG32UI>,
-        MortonCopy<false, PixelFormat::R32UI>,
-        nullptr,
-        nullptr,
-        nullptr,
-        MortonCopy<false, PixelFormat::BGRA8_SRGB>,
-        MortonCopy<false, PixelFormat::DXT1_SRGB>,
-        MortonCopy<false, PixelFormat::DXT23_SRGB>,
-        MortonCopy<false, PixelFormat::DXT45_SRGB>,
-        MortonCopy<false, PixelFormat::BC7U_SRGB>,
-        nullptr,
-        nullptr,
-        nullptr,
-        nullptr,
-        nullptr,
-        nullptr,
-        nullptr,
-        nullptr,
-        MortonCopy<false, PixelFormat::Z32F>,
-        MortonCopy<false, PixelFormat::Z16>,
-        MortonCopy<false, PixelFormat::Z24S8>,
-        MortonCopy<false, PixelFormat::S8Z24>,
-        MortonCopy<false, PixelFormat::Z32FS8>,
-    // clang-format on
+    MortonCopy<false, PixelFormat::ABGR8U>,
+    MortonCopy<false, PixelFormat::ABGR8S>,
+    MortonCopy<false, PixelFormat::ABGR8UI>,
+    MortonCopy<false, PixelFormat::B5G6R5U>,
+    MortonCopy<false, PixelFormat::A2B10G10R10U>,
+    MortonCopy<false, PixelFormat::A1B5G5R5U>,
+    MortonCopy<false, PixelFormat::R8U>,
+    MortonCopy<false, PixelFormat::R8UI>,
+    MortonCopy<false, PixelFormat::RGBA16F>,
+    MortonCopy<false, PixelFormat::RGBA16U>,
+    MortonCopy<false, PixelFormat::RGBA16UI>,
+    MortonCopy<false, PixelFormat::R11FG11FB10F>,
+    MortonCopy<false, PixelFormat::RGBA32UI>,
+    MortonCopy<false, PixelFormat::DXT1>,
+    MortonCopy<false, PixelFormat::DXT23>,
+    MortonCopy<false, PixelFormat::DXT45>,
+    MortonCopy<false, PixelFormat::DXN1>,
+    MortonCopy<false, PixelFormat::DXN2UNORM>,
+    MortonCopy<false, PixelFormat::DXN2SNORM>,
+    MortonCopy<false, PixelFormat::BC7U>,
+    MortonCopy<false, PixelFormat::BC6H_UF16>,
+    MortonCopy<false, PixelFormat::BC6H_SF16>,
+    // TODO(Subv): Swizzling ASTC formats are not supported
+    nullptr,
+    MortonCopy<false, PixelFormat::BGRA8>,
+    MortonCopy<false, PixelFormat::RGBA32F>,
+    MortonCopy<false, PixelFormat::RG32F>,
+    MortonCopy<false, PixelFormat::R32F>,
+    MortonCopy<false, PixelFormat::R16F>,
+    MortonCopy<false, PixelFormat::R16U>,
+    MortonCopy<false, PixelFormat::R16S>,
+    MortonCopy<false, PixelFormat::R16UI>,
+    MortonCopy<false, PixelFormat::R16I>,
+    MortonCopy<false, PixelFormat::RG16>,
+    MortonCopy<false, PixelFormat::RG16F>,
+    MortonCopy<false, PixelFormat::RG16UI>,
+    MortonCopy<false, PixelFormat::RG16I>,
+    MortonCopy<false, PixelFormat::RG16S>,
+    MortonCopy<false, PixelFormat::RGB32F>,
+    MortonCopy<false, PixelFormat::RGBA8_SRGB>,
+    MortonCopy<false, PixelFormat::RG8U>,
+    MortonCopy<false, PixelFormat::RG8S>,
+    MortonCopy<false, PixelFormat::RG32UI>,
+    MortonCopy<false, PixelFormat::R32UI>,
+    nullptr,
+    nullptr,
+    nullptr,
+    MortonCopy<false, PixelFormat::BGRA8_SRGB>,
+    MortonCopy<false, PixelFormat::DXT1_SRGB>,
+    MortonCopy<false, PixelFormat::DXT23_SRGB>,
+    MortonCopy<false, PixelFormat::DXT45_SRGB>,
+    MortonCopy<false, PixelFormat::BC7U_SRGB>,
+    nullptr,
+    nullptr,
+    nullptr,
+    nullptr,
+    nullptr,
+    nullptr,
+    nullptr,
+    nullptr,
+    MortonCopy<false, PixelFormat::Z32F>,
+    MortonCopy<false, PixelFormat::Z16>,
+    MortonCopy<false, PixelFormat::Z24S8>,
+    MortonCopy<false, PixelFormat::S8Z24>,
+    MortonCopy<false, PixelFormat::Z32FS8>,
 };
 
 static MortonCopyFn GetSwizzleFunction(MortonSwizzleMode mode, Surface::PixelFormat format) {
@@ -191,45 +186,6 @@ static MortonCopyFn GetSwizzleFunction(MortonSwizzleMode mode, Surface::PixelFor
     return morton_to_linear_fns[static_cast<std::size_t>(format)];
 }
 
-/// 8x8 Z-Order coordinate from 2D coordinates
-static u32 MortonInterleave(u32 x, u32 y) {
-    static const u32 xlut[] = {0x00, 0x01, 0x04, 0x05, 0x10, 0x11, 0x14, 0x15};
-    static const u32 ylut[] = {0x00, 0x02, 0x08, 0x0a, 0x20, 0x22, 0x28, 0x2a};
-    return xlut[x % 8] + ylut[y % 8];
-}
-
-/// Calculates the offset of the position of the pixel in Morton order
-static u32 GetMortonOffset(u32 x, u32 y, u32 bytes_per_pixel) {
-    // Images are split into 8x8 tiles. Each tile is composed of four 4x4 subtiles each
-    // of which is composed of four 2x2 subtiles each of which is composed of four texels.
-    // Each structure is embedded into the next-bigger one in a diagonal pattern, e.g.
-    // texels are laid out in a 2x2 subtile like this:
-    // 2 3
-    // 0 1
-    //
-    // The full 8x8 tile has the texels arranged like this:
-    //
-    // 42 43 46 47 58 59 62 63
-    // 40 41 44 45 56 57 60 61
-    // 34 35 38 39 50 51 54 55
-    // 32 33 36 37 48 49 52 53
-    // 10 11 14 15 26 27 30 31
-    // 08 09 12 13 24 25 28 29
-    // 02 03 06 07 18 19 22 23
-    // 00 01 04 05 16 17 20 21
-    //
-    // This pattern is what's called Z-order curve, or Morton order.
-
-    const unsigned int block_height = 8;
-    const unsigned int coarse_x = x & ~7;
-
-    u32 i = MortonInterleave(x, y);
-
-    const unsigned int offset = coarse_x * block_height;
-
-    return (i + offset) * bytes_per_pixel;
-}
-
 static u32 MortonInterleave128(u32 x, u32 y) {
     // 128x128 Z-Order coordinate from 2D coordinates
     static constexpr u32 xlut[] = {
@@ -325,14 +281,14 @@ static u32 GetMortonOffset128(u32 x, u32 y, u32 bytes_per_pixel) {
 
 void MortonSwizzle(MortonSwizzleMode mode, Surface::PixelFormat format, u32 stride,
                    u32 block_height, u32 height, u32 block_depth, u32 depth, u32 tile_width_spacing,
-                   u8* buffer, std::size_t buffer_size, VAddr addr) {
-
+                   u8* buffer, u8* addr) {
     GetSwizzleFunction(mode, format)(stride, block_height, height, block_depth, depth,
-                                     tile_width_spacing, buffer, buffer_size, addr);
+                                     tile_width_spacing, buffer, addr);
 }
 
-void MortonCopyPixels128(u32 width, u32 height, u32 bytes_per_pixel, u32 linear_bytes_per_pixel,
-                         u8* morton_data, u8* linear_data, bool morton_to_linear) {
+void MortonCopyPixels128(MortonSwizzleMode mode, u32 width, u32 height, u32 bytes_per_pixel,
+                         u32 linear_bytes_per_pixel, u8* morton_data, u8* linear_data) {
+    const bool morton_to_linear = mode == MortonSwizzleMode::MortonToLinear;
     u8* data_ptrs[2];
     for (u32 y = 0; y < height; ++y) {
         for (u32 x = 0; x < width; ++x) {
diff --git a/src/video_core/morton.h b/src/video_core/morton.h
index 065f59ce3..ee5b45555 100644
--- a/src/video_core/morton.h
+++ b/src/video_core/morton.h
@@ -13,9 +13,9 @@ enum class MortonSwizzleMode { MortonToLinear, LinearToMorton };
 
 void MortonSwizzle(MortonSwizzleMode mode, VideoCore::Surface::PixelFormat format, u32 stride,
                    u32 block_height, u32 height, u32 block_depth, u32 depth, u32 tile_width_spacing,
-                   u8* buffer, std::size_t buffer_size, VAddr addr);
+                   u8* buffer, u8* addr);
 
-void MortonCopyPixels128(u32 width, u32 height, u32 bytes_per_pixel, u32 linear_bytes_per_pixel,
-                         u8* morton_data, u8* linear_data, bool morton_to_linear);
+void MortonCopyPixels128(MortonSwizzleMode mode, u32 width, u32 height, u32 bytes_per_pixel,
+                         u32 linear_bytes_per_pixel, u8* morton_data, u8* linear_data);
 
 } // namespace VideoCore
diff --git a/src/video_core/rasterizer_cache.h b/src/video_core/rasterizer_cache.h
index bcf0c15a4..ecd9986a0 100644
--- a/src/video_core/rasterizer_cache.h
+++ b/src/video_core/rasterizer_cache.h
@@ -4,6 +4,7 @@
 
 #pragma once
 
+#include <mutex>
 #include <set>
 #include <unordered_map>
 
@@ -12,14 +13,26 @@
 
 #include "common/common_types.h"
 #include "core/settings.h"
+#include "video_core/gpu.h"
 #include "video_core/rasterizer_interface.h"
 
 class RasterizerCacheObject {
 public:
+    explicit RasterizerCacheObject(const u8* host_ptr)
+        : host_ptr{host_ptr}, cache_addr{ToCacheAddr(host_ptr)} {}
+
     virtual ~RasterizerCacheObject();
 
+    CacheAddr GetCacheAddr() const {
+        return cache_addr;
+    }
+
+    const u8* GetHostPtr() const {
+        return host_ptr;
+    }
+
     /// Gets the address of the shader in guest memory, required for cache management
-    virtual VAddr GetAddr() const = 0;
+    virtual VAddr GetCpuAddr() const = 0;
 
     /// Gets the size of the shader in guest memory, required for cache management
     virtual std::size_t GetSizeInBytes() const = 0;
@@ -58,6 +71,8 @@ private:
     bool is_registered{};      ///< Whether the object is currently registered with the cache
     bool is_dirty{};           ///< Whether the object is dirty (out of sync with guest memory)
     u64 last_modified_ticks{}; ///< When the object was last modified, used for in-order flushing
+    CacheAddr cache_addr{};    ///< Cache address memory, unique from emulated virtual address space
+    const u8* host_ptr{};      ///< Pointer to the memory backing this cached region
 };
 
 template <class T>
@@ -68,7 +83,9 @@ public:
     explicit RasterizerCache(VideoCore::RasterizerInterface& rasterizer) : rasterizer{rasterizer} {}
 
     /// Write any cached resources overlapping the specified region back to memory
-    void FlushRegion(Tegra::GPUVAddr addr, size_t size) {
+    void FlushRegion(CacheAddr addr, std::size_t size) {
+        std::lock_guard<std::recursive_mutex> lock{mutex};
+
         const auto& objects{GetSortedObjectsFromRegion(addr, size)};
         for (auto& object : objects) {
             FlushObject(object);
@@ -76,7 +93,9 @@ public:
     }
 
     /// Mark the specified region as being invalidated
-    void InvalidateRegion(VAddr addr, u64 size) {
+    void InvalidateRegion(CacheAddr addr, u64 size) {
+        std::lock_guard<std::recursive_mutex> lock{mutex};
+
         const auto& objects{GetSortedObjectsFromRegion(addr, size)};
         for (auto& object : objects) {
             if (!object->IsRegistered()) {
@@ -89,49 +108,70 @@ public:
 
     /// Invalidates everything in the cache
     void InvalidateAll() {
+        std::lock_guard<std::recursive_mutex> lock{mutex};
+
         while (interval_cache.begin() != interval_cache.end()) {
             Unregister(*interval_cache.begin()->second.begin());
         }
     }
 
 protected:
-    /// Tries to get an object from the cache with the specified address
-    T TryGet(VAddr addr) const {
+    /// Tries to get an object from the cache with the specified cache address
+    T TryGet(CacheAddr addr) const {
         const auto iter = map_cache.find(addr);
         if (iter != map_cache.end())
             return iter->second;
         return nullptr;
     }
 
+    T TryGet(const void* addr) const {
+        const auto iter = map_cache.find(ToCacheAddr(addr));
+        if (iter != map_cache.end())
+            return iter->second;
+        return nullptr;
+    }
+
     /// Register an object into the cache
     void Register(const T& object) {
+        std::lock_guard<std::recursive_mutex> lock{mutex};
+
         object->SetIsRegistered(true);
         interval_cache.add({GetInterval(object), ObjectSet{object}});
-        map_cache.insert({object->GetAddr(), object});
-        rasterizer.UpdatePagesCachedCount(object->GetAddr(), object->GetSizeInBytes(), 1);
+        map_cache.insert({object->GetCacheAddr(), object});
+        rasterizer.UpdatePagesCachedCount(object->GetCpuAddr(), object->GetSizeInBytes(), 1);
     }
 
     /// Unregisters an object from the cache
     void Unregister(const T& object) {
-        object->SetIsRegistered(false);
-        rasterizer.UpdatePagesCachedCount(object->GetAddr(), object->GetSizeInBytes(), -1);
-        // Only flush if use_accurate_gpu_emulation is enabled, as it incurs a performance hit
-        if (Settings::values.use_accurate_gpu_emulation) {
-            FlushObject(object);
-        }
+        std::lock_guard<std::recursive_mutex> lock{mutex};
 
+        object->SetIsRegistered(false);
+        rasterizer.UpdatePagesCachedCount(object->GetCpuAddr(), object->GetSizeInBytes(), -1);
         interval_cache.subtract({GetInterval(object), ObjectSet{object}});
-        map_cache.erase(object->GetAddr());
+        map_cache.erase(object->GetCacheAddr());
     }
 
     /// Returns a ticks counter used for tracking when cached objects were last modified
     u64 GetModifiedTicks() {
+        std::lock_guard<std::recursive_mutex> lock{mutex};
+
         return ++modified_ticks;
     }
 
+    /// Flushes the specified object, updating appropriate cache state as needed
+    void FlushObject(const T& object) {
+        std::lock_guard<std::recursive_mutex> lock{mutex};
+
+        if (!object->IsDirty()) {
+            return;
+        }
+        object->Flush();
+        object->MarkAsModified(false, *this);
+    }
+
 private:
     /// Returns a list of cached objects from the specified memory region, ordered by access time
-    std::vector<T> GetSortedObjectsFromRegion(VAddr addr, u64 size) {
+    std::vector<T> GetSortedObjectsFromRegion(CacheAddr addr, u64 size) {
         if (size == 0) {
             return {};
         }
@@ -154,27 +194,19 @@ private:
         return objects;
     }
 
-    /// Flushes the specified object, updating appropriate cache state as needed
-    void FlushObject(const T& object) {
-        if (!object->IsDirty()) {
-            return;
-        }
-        object->Flush();
-        object->MarkAsModified(false, *this);
-    }
-
     using ObjectSet = std::set<T>;
-    using ObjectCache = std::unordered_map<VAddr, T>;
-    using IntervalCache = boost::icl::interval_map<VAddr, ObjectSet>;
+    using ObjectCache = std::unordered_map<CacheAddr, T>;
+    using IntervalCache = boost::icl::interval_map<CacheAddr, ObjectSet>;
     using ObjectInterval = typename IntervalCache::interval_type;
 
     static auto GetInterval(const T& object) {
-        return ObjectInterval::right_open(object->GetAddr(),
-                                          object->GetAddr() + object->GetSizeInBytes());
+        return ObjectInterval::right_open(object->GetCacheAddr(),
+                                          object->GetCacheAddr() + object->GetSizeInBytes());
     }
 
     ObjectCache map_cache;
     IntervalCache interval_cache; ///< Cache of objects
     u64 modified_ticks{};         ///< Counter of cache state ticks, used for in-order flushing
     VideoCore::RasterizerInterface& rasterizer;
+    std::recursive_mutex mutex;
 };
diff --git a/src/video_core/rasterizer_interface.h b/src/video_core/rasterizer_interface.h
index 4c08bb148..76e292e87 100644
--- a/src/video_core/rasterizer_interface.h
+++ b/src/video_core/rasterizer_interface.h
@@ -4,6 +4,7 @@
 
 #pragma once
 
+#include <atomic>
 #include <functional>
 #include "common/common_types.h"
 #include "video_core/engines/fermi_2d.h"
@@ -34,18 +35,20 @@ public:
     virtual void FlushAll() = 0;
 
     /// Notify rasterizer that any caches of the specified region should be flushed to Switch memory
-    virtual void FlushRegion(VAddr addr, u64 size) = 0;
+    virtual void FlushRegion(CacheAddr addr, u64 size) = 0;
 
     /// Notify rasterizer that any caches of the specified region should be invalidated
-    virtual void InvalidateRegion(VAddr addr, u64 size) = 0;
+    virtual void InvalidateRegion(CacheAddr addr, u64 size) = 0;
 
     /// Notify rasterizer that any caches of the specified region should be flushed to Switch memory
     /// and invalidated
-    virtual void FlushAndInvalidateRegion(VAddr addr, u64 size) = 0;
+    virtual void FlushAndInvalidateRegion(CacheAddr addr, u64 size) = 0;
 
     /// Attempt to use a faster method to perform a surface copy
     virtual bool AccelerateSurfaceCopy(const Tegra::Engines::Fermi2D::Regs::Surface& src,
-                                       const Tegra::Engines::Fermi2D::Regs::Surface& dst) {
+                                       const Tegra::Engines::Fermi2D::Regs::Surface& dst,
+                                       const Common::Rectangle<u32>& src_rect,
+                                       const Common::Rectangle<u32>& dst_rect) {
         return false;
     }
 
@@ -60,6 +63,10 @@ public:
     }
 
     /// Increase/decrease the number of object in pages touching the specified region
-    virtual void UpdatePagesCachedCount(Tegra::GPUVAddr addr, u64 size, int delta) {}
+    virtual void UpdatePagesCachedCount(VAddr addr, u64 size, int delta) {}
+
+    /// Initialize disk cached resources for the game being emulated
+    virtual void LoadDiskResources(const std::atomic_bool& stop_loading = false,
+                                   const DiskResourceLoadCallback& callback = {}) {}
 };
 } // namespace VideoCore
diff --git a/src/video_core/renderer_base.cpp b/src/video_core/renderer_base.cpp
index 94223f45f..919d1f2d4 100644
--- a/src/video_core/renderer_base.cpp
+++ b/src/video_core/renderer_base.cpp
@@ -2,6 +2,7 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#include "common/logging/log.h"
 #include "core/frontend/emu_window.h"
 #include "core/settings.h"
 #include "video_core/renderer_base.h"
diff --git a/src/video_core/renderer_opengl/gl_buffer_cache.cpp b/src/video_core/renderer_opengl/gl_buffer_cache.cpp
index bd2b30e77..5048ed6ce 100644
--- a/src/video_core/renderer_opengl/gl_buffer_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_buffer_cache.cpp
@@ -13,23 +13,28 @@
 
 namespace OpenGL {
 
+CachedBufferEntry::CachedBufferEntry(VAddr cpu_addr, std::size_t size, GLintptr offset,
+                                     std::size_t alignment, u8* host_ptr)
+    : cpu_addr{cpu_addr}, size{size}, offset{offset}, alignment{alignment}, RasterizerCacheObject{
+                                                                                host_ptr} {}
+
 OGLBufferCache::OGLBufferCache(RasterizerOpenGL& rasterizer, std::size_t size)
     : RasterizerCache{rasterizer}, stream_buffer(size, true) {}
 
 GLintptr OGLBufferCache::UploadMemory(Tegra::GPUVAddr gpu_addr, std::size_t size,
                                       std::size_t alignment, bool cache) {
     auto& memory_manager = Core::System::GetInstance().GPU().MemoryManager();
-    const std::optional<VAddr> cpu_addr{memory_manager.GpuToCpuAddress(gpu_addr)};
 
     // Cache management is a big overhead, so only cache entries with a given size.
     // TODO: Figure out which size is the best for given games.
     cache &= size >= 2048;
 
+    const auto& host_ptr{memory_manager.GetPointer(gpu_addr)};
     if (cache) {
-        auto entry = TryGet(*cpu_addr);
+        auto entry = TryGet(host_ptr);
         if (entry) {
-            if (entry->size >= size && entry->alignment == alignment) {
-                return entry->offset;
+            if (entry->GetSize() >= size && entry->GetAlignment() == alignment) {
+                return entry->GetOffset();
             }
             Unregister(entry);
         }
@@ -38,17 +43,17 @@ GLintptr OGLBufferCache::UploadMemory(Tegra::GPUVAddr gpu_addr, std::size_t size
     AlignBuffer(alignment);
     const GLintptr uploaded_offset = buffer_offset;
 
-    Memory::ReadBlock(*cpu_addr, buffer_ptr, size);
+    if (!host_ptr) {
+        return uploaded_offset;
+    }
 
+    std::memcpy(buffer_ptr, host_ptr, size);
     buffer_ptr += size;
     buffer_offset += size;
 
     if (cache) {
-        auto entry = std::make_shared<CachedBufferEntry>();
-        entry->offset = uploaded_offset;
-        entry->size = size;
-        entry->alignment = alignment;
-        entry->addr = *cpu_addr;
+        auto entry = std::make_shared<CachedBufferEntry>(
+            *memory_manager.GpuToCpuAddress(gpu_addr), size, uploaded_offset, alignment, host_ptr);
         Register(entry);
     }
 
diff --git a/src/video_core/renderer_opengl/gl_buffer_cache.h b/src/video_core/renderer_opengl/gl_buffer_cache.h
index c11acfb79..1de1f84ae 100644
--- a/src/video_core/renderer_opengl/gl_buffer_cache.h
+++ b/src/video_core/renderer_opengl/gl_buffer_cache.h
@@ -17,22 +17,39 @@ namespace OpenGL {
 
 class RasterizerOpenGL;
 
-struct CachedBufferEntry final : public RasterizerCacheObject {
-    VAddr GetAddr() const override {
-        return addr;
+class CachedBufferEntry final : public RasterizerCacheObject {
+public:
+    explicit CachedBufferEntry(VAddr cpu_addr, std::size_t size, GLintptr offset,
+                               std::size_t alignment, u8* host_ptr);
+
+    VAddr GetCpuAddr() const override {
+        return cpu_addr;
     }
 
     std::size_t GetSizeInBytes() const override {
         return size;
     }
 
+    std::size_t GetSize() const {
+        return size;
+    }
+
+    GLintptr GetOffset() const {
+        return offset;
+    }
+
+    std::size_t GetAlignment() const {
+        return alignment;
+    }
+
     // We do not have to flush this cache as things in it are never modified by us.
     void Flush() override {}
 
-    VAddr addr;
-    std::size_t size;
-    GLintptr offset;
-    std::size_t alignment;
+private:
+    VAddr cpu_addr{};
+    std::size_t size{};
+    GLintptr offset{};
+    std::size_t alignment{};
 };
 
 class OGLBufferCache final : public RasterizerCache<std::shared_ptr<CachedBufferEntry>> {
diff --git a/src/video_core/renderer_opengl/gl_global_cache.cpp b/src/video_core/renderer_opengl/gl_global_cache.cpp
index c7f32feaa..c8dbcacbd 100644
--- a/src/video_core/renderer_opengl/gl_global_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_global_cache.cpp
@@ -7,7 +7,6 @@
 #include "common/assert.h"
 #include "common/logging/log.h"
 #include "core/core.h"
-#include "core/memory.h"
 #include "video_core/renderer_opengl/gl_global_cache.h"
 #include "video_core/renderer_opengl/gl_rasterizer.h"
 #include "video_core/renderer_opengl/gl_shader_decompiler.h"
@@ -15,12 +14,13 @@
 
 namespace OpenGL {
 
-CachedGlobalRegion::CachedGlobalRegion(VAddr addr, u32 size) : addr{addr}, size{size} {
+CachedGlobalRegion::CachedGlobalRegion(VAddr cpu_addr, u32 size, u8* host_ptr)
+    : cpu_addr{cpu_addr}, size{size}, RasterizerCacheObject{host_ptr} {
     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");
+    LabelGLObject(GL_BUFFER, buffer.handle, cpu_addr, "GlobalMemory");
 }
 
 void CachedGlobalRegion::Reload(u32 size_) {
@@ -35,10 +35,10 @@ void CachedGlobalRegion::Reload(u32 size_) {
 
     // TODO(Rodrigo): Get rid of Memory::GetPointer with a staging buffer
     glBindBuffer(GL_SHADER_STORAGE_BUFFER, buffer.handle);
-    glBufferData(GL_SHADER_STORAGE_BUFFER, size, Memory::GetPointer(addr), GL_DYNAMIC_DRAW);
+    glBufferData(GL_SHADER_STORAGE_BUFFER, size, GetHostPtr(), GL_DYNAMIC_DRAW);
 }
 
-GlobalRegion GlobalRegionCacheOpenGL::TryGetReservedGlobalRegion(VAddr addr, u32 size) const {
+GlobalRegion GlobalRegionCacheOpenGL::TryGetReservedGlobalRegion(CacheAddr addr, u32 size) const {
     const auto search{reserve.find(addr)};
     if (search == reserve.end()) {
         return {};
@@ -46,19 +46,22 @@ GlobalRegion GlobalRegionCacheOpenGL::TryGetReservedGlobalRegion(VAddr addr, u32
     return search->second;
 }
 
-GlobalRegion GlobalRegionCacheOpenGL::GetUncachedGlobalRegion(VAddr addr, u32 size) {
-    GlobalRegion region{TryGetReservedGlobalRegion(addr, size)};
+GlobalRegion GlobalRegionCacheOpenGL::GetUncachedGlobalRegion(Tegra::GPUVAddr addr, u32 size,
+                                                              u8* host_ptr) {
+    GlobalRegion region{TryGetReservedGlobalRegion(ToCacheAddr(host_ptr), size)};
     if (!region) {
         // No reserved surface available, create a new one and reserve it
-        region = std::make_shared<CachedGlobalRegion>(addr, size);
+        auto& memory_manager{Core::System::GetInstance().GPU().MemoryManager()};
+        const auto cpu_addr = *memory_manager.GpuToCpuAddress(addr);
+        region = std::make_shared<CachedGlobalRegion>(cpu_addr, size, host_ptr);
         ReserveGlobalRegion(region);
     }
     region->Reload(size);
     return region;
 }
 
-void GlobalRegionCacheOpenGL::ReserveGlobalRegion(const GlobalRegion& region) {
-    reserve[region->GetAddr()] = region;
+void GlobalRegionCacheOpenGL::ReserveGlobalRegion(GlobalRegion region) {
+    reserve.insert_or_assign(region->GetCacheAddr(), std::move(region));
 }
 
 GlobalRegionCacheOpenGL::GlobalRegionCacheOpenGL(RasterizerOpenGL& rasterizer)
@@ -69,22 +72,20 @@ GlobalRegion GlobalRegionCacheOpenGL::GetGlobalRegion(
     Tegra::Engines::Maxwell3D::Regs::ShaderStage stage) {
 
     auto& gpu{Core::System::GetInstance().GPU()};
-    const auto cbufs = gpu.Maxwell3D().state.shader_stages[static_cast<u64>(stage)];
-    const auto cbuf_addr = gpu.MemoryManager().GpuToCpuAddress(
-        cbufs.const_buffers[global_region.GetCbufIndex()].address + global_region.GetCbufOffset());
-    ASSERT(cbuf_addr);
-
-    const auto actual_addr_gpu = Memory::Read64(*cbuf_addr);
-    const auto size = Memory::Read32(*cbuf_addr + 8);
-    const auto actual_addr = gpu.MemoryManager().GpuToCpuAddress(actual_addr_gpu);
-    ASSERT(actual_addr);
+    auto& memory_manager{gpu.MemoryManager()};
+    const auto cbufs{gpu.Maxwell3D().state.shader_stages[static_cast<u64>(stage)]};
+    const auto addr{cbufs.const_buffers[global_region.GetCbufIndex()].address +
+                    global_region.GetCbufOffset()};
+    const auto actual_addr{memory_manager.Read64(addr)};
+    const auto size{memory_manager.Read32(addr + 8)};
 
     // Look up global region in the cache based on address
-    GlobalRegion region = TryGet(*actual_addr);
+    const auto& host_ptr{memory_manager.GetPointer(actual_addr)};
+    GlobalRegion region{TryGet(host_ptr)};
 
     if (!region) {
         // No global region found - create a new one
-        region = GetUncachedGlobalRegion(*actual_addr, size);
+        region = GetUncachedGlobalRegion(actual_addr, size, host_ptr);
         Register(region);
     }
 
diff --git a/src/video_core/renderer_opengl/gl_global_cache.h b/src/video_core/renderer_opengl/gl_global_cache.h
index 37830bb7c..a840491f7 100644
--- a/src/video_core/renderer_opengl/gl_global_cache.h
+++ b/src/video_core/renderer_opengl/gl_global_cache.h
@@ -27,15 +27,13 @@ using GlobalRegion = std::shared_ptr<CachedGlobalRegion>;
 
 class CachedGlobalRegion final : public RasterizerCacheObject {
 public:
-    explicit CachedGlobalRegion(VAddr addr, u32 size);
+    explicit CachedGlobalRegion(VAddr cpu_addr, u32 size, u8* host_ptr);
 
-    /// Gets the address of the shader in guest memory, required for cache management
-    VAddr GetAddr() const {
-        return addr;
+    VAddr GetCpuAddr() const override {
+        return cpu_addr;
     }
 
-    /// Gets the size of the shader in guest memory, required for cache management
-    std::size_t GetSizeInBytes() const {
+    std::size_t GetSizeInBytes() const override {
         return size;
     }
 
@@ -53,9 +51,8 @@ public:
     }
 
 private:
-    VAddr addr{};
+    VAddr cpu_addr{};
     u32 size{};
-
     OGLBuffer buffer;
 };
 
@@ -68,11 +65,11 @@ public:
                                  Tegra::Engines::Maxwell3D::Regs::ShaderStage stage);
 
 private:
-    GlobalRegion TryGetReservedGlobalRegion(VAddr addr, u32 size) const;
-    GlobalRegion GetUncachedGlobalRegion(VAddr addr, u32 size);
-    void ReserveGlobalRegion(const GlobalRegion& region);
+    GlobalRegion TryGetReservedGlobalRegion(CacheAddr addr, u32 size) const;
+    GlobalRegion GetUncachedGlobalRegion(Tegra::GPUVAddr addr, u32 size, u8* host_ptr);
+    void ReserveGlobalRegion(GlobalRegion region);
 
-    std::unordered_map<VAddr, GlobalRegion> reserve;
+    std::unordered_map<CacheAddr, GlobalRegion> reserve;
 };
 
 } // namespace OpenGL
diff --git a/src/video_core/renderer_opengl/gl_primitive_assembler.cpp b/src/video_core/renderer_opengl/gl_primitive_assembler.cpp
index d9ed08437..75d816795 100644
--- a/src/video_core/renderer_opengl/gl_primitive_assembler.cpp
+++ b/src/video_core/renderer_opengl/gl_primitive_assembler.cpp
@@ -46,8 +46,7 @@ GLintptr PrimitiveAssembler::MakeQuadIndexed(Tegra::GPUVAddr gpu_addr, std::size
     auto [dst_pointer, index_offset] = buffer_cache.ReserveMemory(map_size);
 
     auto& memory_manager = Core::System::GetInstance().GPU().MemoryManager();
-    const std::optional<VAddr> cpu_addr{memory_manager.GpuToCpuAddress(gpu_addr)};
-    const u8* source{Memory::GetPointer(*cpu_addr)};
+    const u8* source{memory_manager.GetPointer(gpu_addr)};
 
     for (u32 primitive = 0; primitive < count / 4; ++primitive) {
         for (std::size_t i = 0; i < TRIANGLES_PER_QUAD; ++i) {
@@ -62,4 +61,4 @@ GLintptr PrimitiveAssembler::MakeQuadIndexed(Tegra::GPUVAddr gpu_addr, std::size
     return index_offset;
 }
 
-} // namespace OpenGL
\ No newline at end of file
+} // namespace OpenGL
diff --git a/src/video_core/renderer_opengl/gl_rasterizer.cpp b/src/video_core/renderer_opengl/gl_rasterizer.cpp
index 9f7c837d6..198c54872 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer.cpp
+++ b/src/video_core/renderer_opengl/gl_rasterizer.cpp
@@ -22,6 +22,7 @@
 #include "core/settings.h"
 #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_gen.h"
 #include "video_core/renderer_opengl/maxwell_to_gl.h"
 #include "video_core/renderer_opengl/renderer_opengl.h"
@@ -99,9 +100,11 @@ 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), global_cache{*this} {
+RasterizerOpenGL::RasterizerOpenGL(Core::Frontend::EmuWindow& window, Core::System& system,
+                                   ScreenInfo& info)
+    : res_cache{*this}, shader_cache{*this, system}, global_cache{*this},
+      emu_window{window}, system{system}, screen_info{info},
+      buffer_cache(*this, STREAM_BUFFER_SIZE) {
     // Create sampler objects
     for (std::size_t i = 0; i < texture_samplers.size(); ++i) {
         texture_samplers[i].Create();
@@ -116,7 +119,7 @@ RasterizerOpenGL::RasterizerOpenGL(Core::Frontend::EmuWindow& window, ScreenInfo
 
     glGetIntegerv(GL_UNIFORM_BUFFER_OFFSET_ALIGNMENT, &uniform_buffer_alignment);
 
-    LOG_CRITICAL(Render_OpenGL, "Sync fixed function OpenGL state here!");
+    LOG_DEBUG(Render_OpenGL, "Sync fixed function OpenGL state here");
     CheckExtensions();
 }
 
@@ -136,7 +139,7 @@ void RasterizerOpenGL::CheckExtensions() {
 }
 
 GLuint RasterizerOpenGL::SetupVertexFormat() {
-    auto& gpu = Core::System::GetInstance().GPU().Maxwell3D();
+    auto& gpu = system.GPU().Maxwell3D();
     const auto& regs = gpu.regs;
 
     if (!gpu.dirty_flags.vertex_attrib_format) {
@@ -175,7 +178,7 @@ GLuint RasterizerOpenGL::SetupVertexFormat() {
                 continue;
 
             const auto& buffer = regs.vertex_array[attrib.buffer];
-            LOG_TRACE(HW_GPU,
+            LOG_TRACE(Render_OpenGL,
                       "vertex attrib {}, count={}, size={}, type={}, offset={}, normalize={}",
                       index, attrib.ComponentCount(), attrib.SizeString(), attrib.TypeString(),
                       attrib.offset.Value(), attrib.IsNormalized());
@@ -198,24 +201,24 @@ GLuint RasterizerOpenGL::SetupVertexFormat() {
     }
 
     // Rebinding the VAO invalidates the vertex buffer bindings.
-    gpu.dirty_flags.vertex_array = 0xFFFFFFFF;
+    gpu.dirty_flags.vertex_array.set();
 
     state.draw.vertex_array = vao_entry.handle;
     return vao_entry.handle;
 }
 
 void RasterizerOpenGL::SetupVertexBuffer(GLuint vao) {
-    auto& gpu = Core::System::GetInstance().GPU().Maxwell3D();
+    auto& gpu = system.GPU().Maxwell3D();
     const auto& regs = gpu.regs;
 
-    if (!gpu.dirty_flags.vertex_array)
+    if (gpu.dirty_flags.vertex_array.none())
         return;
 
     MICROPROFILE_SCOPE(OpenGL_VB);
 
     // Upload all guest vertex arrays sequentially to our buffer
     for (u32 index = 0; index < Maxwell::NumVertexArrays; ++index) {
-        if (~gpu.dirty_flags.vertex_array & (1u << index))
+        if (!gpu.dirty_flags.vertex_array[index])
             continue;
 
         const auto& vertex_array = regs.vertex_array[index];
@@ -242,11 +245,11 @@ void RasterizerOpenGL::SetupVertexBuffer(GLuint vao) {
         }
     }
 
-    gpu.dirty_flags.vertex_array = 0;
+    gpu.dirty_flags.vertex_array.reset();
 }
 
 DrawParameters RasterizerOpenGL::SetupDraw() {
-    const auto& gpu = Core::System::GetInstance().GPU().Maxwell3D();
+    const auto& gpu = system.GPU().Maxwell3D();
     const auto& regs = gpu.regs;
     const bool is_indexed = accelerate_draw == AccelDraw::Indexed;
 
@@ -295,7 +298,7 @@ DrawParameters RasterizerOpenGL::SetupDraw() {
 
 void RasterizerOpenGL::SetupShaders(GLenum primitive_mode) {
     MICROPROFILE_SCOPE(OpenGL_Shader);
-    auto& gpu = Core::System::GetInstance().GPU().Maxwell3D();
+    auto& gpu = system.GPU().Maxwell3D();
 
     BaseBindings base_bindings;
     std::array<bool, Maxwell::NumClipDistances> clip_distances{};
@@ -341,9 +344,8 @@ void RasterizerOpenGL::SetupShaders(GLenum primitive_mode) {
             shader_program_manager->UseProgrammableFragmentShader(program_handle);
             break;
         default:
-            LOG_CRITICAL(HW_GPU, "Unimplemented shader index={}, enable={}, offset=0x{:08X}", index,
-                         shader_config.enable.Value(), shader_config.offset);
-            UNREACHABLE();
+            UNIMPLEMENTED_MSG("Unimplemented shader index={}, enable={}, offset=0x{:08X}", index,
+                              shader_config.enable.Value(), shader_config.offset);
         }
 
         const auto stage_enum = static_cast<Maxwell::ShaderStage>(stage);
@@ -412,7 +414,7 @@ void RasterizerOpenGL::SetupCachedFramebuffer(const FramebufferCacheKey& fbkey,
 }
 
 std::size_t RasterizerOpenGL::CalculateVertexArraysSize() const {
-    const auto& regs = Core::System::GetInstance().GPU().Maxwell3D().regs;
+    const auto& regs = system.GPU().Maxwell3D().regs;
 
     std::size_t size = 0;
     for (u32 index = 0; index < Maxwell::NumVertexArrays; ++index) {
@@ -430,7 +432,7 @@ std::size_t RasterizerOpenGL::CalculateVertexArraysSize() const {
 }
 
 std::size_t RasterizerOpenGL::CalculateIndexBufferSize() const {
-    const auto& regs = Core::System::GetInstance().GPU().Maxwell3D().regs;
+    const auto& regs = system.GPU().Maxwell3D().regs;
 
     return static_cast<std::size_t>(regs.index_array.count) *
            static_cast<std::size_t>(regs.index_array.FormatSizeInBytes());
@@ -477,17 +479,22 @@ void RasterizerOpenGL::UpdatePagesCachedCount(VAddr addr, u64 size, int delta) {
         cached_pages.add({pages_interval, delta});
 }
 
+void RasterizerOpenGL::LoadDiskResources(const std::atomic_bool& stop_loading,
+                                         const VideoCore::DiskResourceLoadCallback& callback) {
+    shader_cache.LoadDiskCache(stop_loading, callback);
+}
+
 std::pair<bool, bool> RasterizerOpenGL::ConfigureFramebuffers(
     OpenGLState& current_state, bool using_color_fb, bool using_depth_fb, bool preserve_contents,
     std::optional<std::size_t> single_color_target) {
     MICROPROFILE_SCOPE(OpenGL_Framebuffer);
-    const auto& gpu = Core::System::GetInstance().GPU().Maxwell3D();
+    auto& gpu = system.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) {
+    if (fb_config_state == current_framebuffer_config_state &&
+        gpu.dirty_flags.color_buffer.none() && !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
@@ -575,7 +582,7 @@ void RasterizerOpenGL::Clear() {
     const auto prev_state{state};
     SCOPE_EXIT({ prev_state.Apply(); });
 
-    const auto& regs = Core::System::GetInstance().GPU().Maxwell3D().regs;
+    const auto& regs = system.GPU().Maxwell3D().regs;
     bool use_color{};
     bool use_depth{};
     bool use_stencil{};
@@ -666,7 +673,7 @@ void RasterizerOpenGL::DrawArrays() {
         return;
 
     MICROPROFILE_SCOPE(OpenGL_Drawing);
-    auto& gpu = Core::System::GetInstance().GPU().Maxwell3D();
+    auto& gpu = system.GPU().Maxwell3D();
     const auto& regs = gpu.regs;
 
     ConfigureFramebuffers(state);
@@ -714,10 +721,10 @@ void RasterizerOpenGL::DrawArrays() {
     // Add space for at least 18 constant buffers
     buffer_size += Maxwell::MaxConstBuffers * (MaxConstbufferSize + uniform_buffer_alignment);
 
-    bool invalidate = buffer_cache.Map(buffer_size);
+    const bool invalidate = buffer_cache.Map(buffer_size);
     if (invalidate) {
         // As all cached buffers are invalidated, we need to recheck their state.
-        gpu.dirty_flags.vertex_array = 0xFFFFFFFF;
+        gpu.dirty_flags.vertex_array.set();
     }
 
     const GLuint vao = SetupVertexFormat();
@@ -731,55 +738,45 @@ void RasterizerOpenGL::DrawArrays() {
     shader_program_manager->ApplyTo(state);
     state.Apply();
 
-    // Execute draw call
+    res_cache.SignalPreDrawCall();
     params.DispatchDraw();
-
-    // Disable scissor test
-    state.viewports[0].scissor.enabled = false;
+    res_cache.SignalPostDrawCall();
 
     accelerate_draw = AccelDraw::Disabled;
-
-    // Unbind textures for potential future use as framebuffer attachments
-    for (auto& texture_unit : state.texture_units) {
-        texture_unit.Unbind();
-    }
-    state.Apply();
 }
 
 void RasterizerOpenGL::FlushAll() {}
 
-void RasterizerOpenGL::FlushRegion(VAddr addr, u64 size) {
+void RasterizerOpenGL::FlushRegion(CacheAddr addr, u64 size) {
     MICROPROFILE_SCOPE(OpenGL_CacheManagement);
-
-    if (Settings::values.use_accurate_gpu_emulation) {
-        // Only flush if use_accurate_gpu_emulation is enabled, as it incurs a performance hit
-        res_cache.FlushRegion(addr, size);
+    if (!addr || !size) {
+        return;
     }
+    res_cache.FlushRegion(addr, size);
 }
 
-void RasterizerOpenGL::InvalidateRegion(VAddr addr, u64 size) {
+void RasterizerOpenGL::InvalidateRegion(CacheAddr addr, u64 size) {
     MICROPROFILE_SCOPE(OpenGL_CacheManagement);
+    if (!addr || !size) {
+        return;
+    }
     res_cache.InvalidateRegion(addr, size);
     shader_cache.InvalidateRegion(addr, size);
     global_cache.InvalidateRegion(addr, size);
     buffer_cache.InvalidateRegion(addr, size);
 }
 
-void RasterizerOpenGL::FlushAndInvalidateRegion(VAddr addr, u64 size) {
+void RasterizerOpenGL::FlushAndInvalidateRegion(CacheAddr addr, u64 size) {
     FlushRegion(addr, size);
     InvalidateRegion(addr, size);
 }
 
 bool RasterizerOpenGL::AccelerateSurfaceCopy(const Tegra::Engines::Fermi2D::Regs::Surface& src,
-                                             const Tegra::Engines::Fermi2D::Regs::Surface& dst) {
+                                             const Tegra::Engines::Fermi2D::Regs::Surface& dst,
+                                             const Common::Rectangle<u32>& src_rect,
+                                             const Common::Rectangle<u32>& dst_rect) {
     MICROPROFILE_SCOPE(OpenGL_Blits);
-
-    if (Settings::values.use_accurate_gpu_emulation) {
-        // Skip the accelerated copy and perform a slow but more accurate copy
-        return false;
-    }
-
-    res_cache.FermiCopySurface(src, dst);
+    res_cache.FermiCopySurface(src, dst, src_rect, dst_rect);
     return true;
 }
 
@@ -791,7 +788,7 @@ bool RasterizerOpenGL::AccelerateDisplay(const Tegra::FramebufferConfig& config,
 
     MICROPROFILE_SCOPE(OpenGL_CacheManagement);
 
-    const auto& surface{res_cache.TryFindFramebufferSurface(framebuffer_addr)};
+    const auto& surface{res_cache.TryFindFramebufferSurface(Memory::GetPointer(framebuffer_addr))};
     if (!surface) {
         return {};
     }
@@ -802,7 +799,10 @@ bool RasterizerOpenGL::AccelerateDisplay(const Tegra::FramebufferConfig& config,
         VideoCore::Surface::PixelFormatFromGPUPixelFormat(config.pixel_format)};
     ASSERT_MSG(params.width == config.width, "Framebuffer width is different");
     ASSERT_MSG(params.height == config.height, "Framebuffer height is different");
-    ASSERT_MSG(params.pixel_format == pixel_format, "Framebuffer pixel_format is different");
+
+    if (params.pixel_format != pixel_format) {
+        LOG_WARNING(Render_OpenGL, "Framebuffer pixel_format is different");
+    }
 
     screen_info.display_texture = surface->Texture().handle;
 
@@ -811,104 +811,87 @@ bool RasterizerOpenGL::AccelerateDisplay(const Tegra::FramebufferConfig& config,
 
 void RasterizerOpenGL::SamplerInfo::Create() {
     sampler.Create();
-    mag_filter = min_filter = Tegra::Texture::TextureFilter::Linear;
-    wrap_u = wrap_v = wrap_p = Tegra::Texture::WrapMode::Wrap;
-    uses_depth_compare = false;
+    mag_filter = Tegra::Texture::TextureFilter::Linear;
+    min_filter = Tegra::Texture::TextureFilter::Linear;
+    wrap_u = Tegra::Texture::WrapMode::Wrap;
+    wrap_v = Tegra::Texture::WrapMode::Wrap;
+    wrap_p = Tegra::Texture::WrapMode::Wrap;
+    use_depth_compare = false;
     depth_compare_func = Tegra::Texture::DepthCompareFunc::Never;
 
-    // default is GL_LINEAR_MIPMAP_LINEAR
+    // OpenGL's default is GL_LINEAR_MIPMAP_LINEAR
     glSamplerParameteri(sampler.handle, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
-    // Other attributes have correct defaults
     glSamplerParameteri(sampler.handle, GL_TEXTURE_COMPARE_FUNC, GL_NEVER);
+
+    // Other attributes have correct defaults
 }
 
 void RasterizerOpenGL::SamplerInfo::SyncWithConfig(const Tegra::Texture::TSCEntry& config) {
-    const GLuint s = sampler.handle;
+    const GLuint sampler_id = sampler.handle;
     if (mag_filter != config.mag_filter) {
         mag_filter = config.mag_filter;
         glSamplerParameteri(
-            s, GL_TEXTURE_MAG_FILTER,
+            sampler_id, GL_TEXTURE_MAG_FILTER,
             MaxwellToGL::TextureFilterMode(mag_filter, Tegra::Texture::TextureMipmapFilter::None));
     }
-    if (min_filter != config.min_filter || mip_filter != config.mip_filter) {
+    if (min_filter != config.min_filter || mipmap_filter != config.mipmap_filter) {
         min_filter = config.min_filter;
-        mip_filter = config.mip_filter;
-        glSamplerParameteri(s, GL_TEXTURE_MIN_FILTER,
-                            MaxwellToGL::TextureFilterMode(min_filter, mip_filter));
+        mipmap_filter = config.mipmap_filter;
+        glSamplerParameteri(sampler_id, GL_TEXTURE_MIN_FILTER,
+                            MaxwellToGL::TextureFilterMode(min_filter, mipmap_filter));
     }
 
     if (wrap_u != config.wrap_u) {
         wrap_u = config.wrap_u;
-        glSamplerParameteri(s, GL_TEXTURE_WRAP_S, MaxwellToGL::WrapMode(wrap_u));
+        glSamplerParameteri(sampler_id, GL_TEXTURE_WRAP_S, MaxwellToGL::WrapMode(wrap_u));
     }
     if (wrap_v != config.wrap_v) {
         wrap_v = config.wrap_v;
-        glSamplerParameteri(s, GL_TEXTURE_WRAP_T, MaxwellToGL::WrapMode(wrap_v));
+        glSamplerParameteri(sampler_id, GL_TEXTURE_WRAP_T, MaxwellToGL::WrapMode(wrap_v));
     }
     if (wrap_p != config.wrap_p) {
         wrap_p = config.wrap_p;
-        glSamplerParameteri(s, GL_TEXTURE_WRAP_R, MaxwellToGL::WrapMode(wrap_p));
+        glSamplerParameteri(sampler_id, GL_TEXTURE_WRAP_R, MaxwellToGL::WrapMode(wrap_p));
     }
 
-    if (uses_depth_compare != (config.depth_compare_enabled == 1)) {
-        uses_depth_compare = (config.depth_compare_enabled == 1);
-        if (uses_depth_compare) {
-            glSamplerParameteri(s, GL_TEXTURE_COMPARE_MODE, GL_COMPARE_REF_TO_TEXTURE);
-        } else {
-            glSamplerParameteri(s, GL_TEXTURE_COMPARE_MODE, GL_NONE);
-        }
+    if (const bool enabled = config.depth_compare_enabled == 1; use_depth_compare != enabled) {
+        use_depth_compare = enabled;
+        glSamplerParameteri(sampler_id, GL_TEXTURE_COMPARE_MODE,
+                            use_depth_compare ? GL_COMPARE_REF_TO_TEXTURE : GL_NONE);
     }
 
     if (depth_compare_func != config.depth_compare_func) {
         depth_compare_func = config.depth_compare_func;
-        glSamplerParameteri(s, GL_TEXTURE_COMPARE_FUNC,
+        glSamplerParameteri(sampler_id, GL_TEXTURE_COMPARE_FUNC,
                             MaxwellToGL::DepthCompareFunc(depth_compare_func));
     }
 
-    GLvec4 new_border_color;
-    if (config.srgb_conversion) {
-        new_border_color[0] = config.srgb_border_color_r / 255.0f;
-        new_border_color[1] = config.srgb_border_color_g / 255.0f;
-        new_border_color[2] = config.srgb_border_color_g / 255.0f;
-    } else {
-        new_border_color[0] = config.border_color_r;
-        new_border_color[1] = config.border_color_g;
-        new_border_color[2] = config.border_color_b;
-    }
-    new_border_color[3] = config.border_color_a;
-
-    if (border_color != new_border_color) {
+    if (const auto new_border_color = config.GetBorderColor(); border_color != new_border_color) {
         border_color = new_border_color;
-        glSamplerParameterfv(s, GL_TEXTURE_BORDER_COLOR, border_color.data());
+        glSamplerParameterfv(sampler_id, GL_TEXTURE_BORDER_COLOR, border_color.data());
     }
 
-    const float anisotropic_max = static_cast<float>(1 << config.max_anisotropy.Value());
-    if (anisotropic_max != max_anisotropic) {
-        max_anisotropic = anisotropic_max;
+    if (const float anisotropic = config.GetMaxAnisotropy(); max_anisotropic != anisotropic) {
+        max_anisotropic = anisotropic;
         if (GLAD_GL_ARB_texture_filter_anisotropic) {
-            glSamplerParameterf(s, GL_TEXTURE_MAX_ANISOTROPY, max_anisotropic);
+            glSamplerParameterf(sampler_id, GL_TEXTURE_MAX_ANISOTROPY, max_anisotropic);
         } else if (GLAD_GL_EXT_texture_filter_anisotropic) {
-            glSamplerParameterf(s, GL_TEXTURE_MAX_ANISOTROPY_EXT, max_anisotropic);
+            glSamplerParameterf(sampler_id, GL_TEXTURE_MAX_ANISOTROPY_EXT, max_anisotropic);
         }
     }
-    const float lod_min = static_cast<float>(config.min_lod_clamp.Value()) / 256.0f;
-    if (lod_min != min_lod) {
-        min_lod = lod_min;
-        glSamplerParameterf(s, GL_TEXTURE_MIN_LOD, min_lod);
-    }
 
-    const float lod_max = static_cast<float>(config.max_lod_clamp.Value()) / 256.0f;
-    if (lod_max != max_lod) {
-        max_lod = lod_max;
-        glSamplerParameterf(s, GL_TEXTURE_MAX_LOD, max_lod);
+    if (const float min = config.GetMinLod(); min_lod != min) {
+        min_lod = min;
+        glSamplerParameterf(sampler_id, GL_TEXTURE_MIN_LOD, min_lod);
+    }
+    if (const float max = config.GetMaxLod(); max_lod != max) {
+        max_lod = max;
+        glSamplerParameterf(sampler_id, GL_TEXTURE_MAX_LOD, max_lod);
     }
-    const u32 bias = config.mip_lod_bias.Value();
-    // Sign extend the 13-bit value.
-    constexpr u32 mask = 1U << (13 - 1);
-    const float bias_lod = static_cast<s32>((bias ^ mask) - mask) / 256.f;
-    if (lod_bias != bias_lod) {
-        lod_bias = bias_lod;
-        glSamplerParameterf(s, GL_TEXTURE_LOD_BIAS, lod_bias);
+
+    if (const float bias = config.GetLodBias(); lod_bias != bias) {
+        lod_bias = bias;
+        glSamplerParameterf(sampler_id, GL_TEXTURE_LOD_BIAS, lod_bias);
     }
 }
 
@@ -916,7 +899,7 @@ void RasterizerOpenGL::SetupConstBuffers(Tegra::Engines::Maxwell3D::Regs::Shader
                                          const Shader& shader, GLuint program_handle,
                                          BaseBindings base_bindings) {
     MICROPROFILE_SCOPE(OpenGL_UBO);
-    const auto& gpu = Core::System::GetInstance().GPU();
+    const auto& gpu = system.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_buffers;
@@ -948,8 +931,8 @@ void RasterizerOpenGL::SetupConstBuffers(Tegra::Engines::Maxwell3D::Regs::Shader
             size = buffer.size;
 
             if (size > MaxConstbufferSize) {
-                LOG_CRITICAL(HW_GPU, "indirect constbuffer size {} exceeds maximum {}", size,
-                             MaxConstbufferSize);
+                LOG_WARNING(Render_OpenGL, "Indirect constbuffer size {} exceeds maximum {}", size,
+                            MaxConstbufferSize);
                 size = MaxConstbufferSize;
             }
         } else {
@@ -995,7 +978,7 @@ void RasterizerOpenGL::SetupGlobalRegions(Tegra::Engines::Maxwell3D::Regs::Shade
 void RasterizerOpenGL::SetupTextures(Maxwell::ShaderStage stage, const Shader& shader,
                                      GLuint program_handle, BaseBindings base_bindings) {
     MICROPROFILE_SCOPE(OpenGL_Texture);
-    const auto& gpu = Core::System::GetInstance().GPU();
+    const auto& gpu = system.GPU();
     const auto& maxwell3d = gpu.Maxwell3D();
     const auto& entries = shader->GetShaderEntries().samplers;
 
@@ -1004,35 +987,25 @@ void RasterizerOpenGL::SetupTextures(Maxwell::ShaderStage stage, const Shader& s
 
     for (u32 bindpoint = 0; bindpoint < entries.size(); ++bindpoint) {
         const auto& entry = entries[bindpoint];
+        const auto texture = maxwell3d.GetStageTexture(stage, entry.GetOffset());
         const u32 current_bindpoint = base_bindings.sampler + bindpoint;
-        auto& unit = state.texture_units[current_bindpoint];
-
-        const auto texture = maxwell3d.GetStageTexture(entry.GetStage(), entry.GetOffset());
-        if (!texture.enabled) {
-            unit.texture = 0;
-            continue;
-        }
 
         texture_samplers[current_bindpoint].SyncWithConfig(texture.tsc);
 
-        Surface surface = res_cache.GetTextureSurface(texture, entry);
-        if (surface != nullptr) {
-            unit.texture =
-                entry.IsArray() ? surface->TextureLayer().handle : surface->Texture().handle;
-            unit.target = entry.IsArray() ? surface->TargetLayer() : surface->Target();
-            unit.swizzle.r = MaxwellToGL::SwizzleSource(texture.tic.x_source);
-            unit.swizzle.g = MaxwellToGL::SwizzleSource(texture.tic.y_source);
-            unit.swizzle.b = MaxwellToGL::SwizzleSource(texture.tic.z_source);
-            unit.swizzle.a = MaxwellToGL::SwizzleSource(texture.tic.w_source);
+        if (Surface surface = res_cache.GetTextureSurface(texture, entry); surface) {
+            state.texture_units[current_bindpoint].texture =
+                surface->Texture(entry.IsArray()).handle;
+            surface->UpdateSwizzle(texture.tic.x_source, texture.tic.y_source, texture.tic.z_source,
+                                   texture.tic.w_source);
         } else {
             // Can occur when texture addr is null or its memory is unmapped/invalid
-            unit.texture = 0;
+            state.texture_units[current_bindpoint].texture = 0;
         }
     }
 }
 
 void RasterizerOpenGL::SyncViewport(OpenGLState& current_state) {
-    const auto& regs = Core::System::GetInstance().GPU().Maxwell3D().regs;
+    const auto& regs = system.GPU().Maxwell3D().regs;
     const bool geometry_shaders_enabled =
         regs.IsShaderConfigEnabled(static_cast<size_t>(Maxwell::ShaderProgram::Geometry));
     const std::size_t viewport_count =
@@ -1040,7 +1013,7 @@ void RasterizerOpenGL::SyncViewport(OpenGLState& current_state) {
     for (std::size_t i = 0; i < viewport_count; i++) {
         auto& viewport = current_state.viewports[i];
         const auto& src = regs.viewports[i];
-        const MathUtil::Rectangle<s32> viewport_rect{regs.viewport_transform[i].GetRect()};
+        const Common::Rectangle<s32> viewport_rect{regs.viewport_transform[i].GetRect()};
         viewport.x = viewport_rect.left;
         viewport.y = viewport_rect.bottom;
         viewport.width = viewport_rect.GetWidth();
@@ -1055,7 +1028,7 @@ void RasterizerOpenGL::SyncViewport(OpenGLState& current_state) {
 void RasterizerOpenGL::SyncClipEnabled(
     const std::array<bool, Maxwell::Regs::NumClipDistances>& clip_mask) {
 
-    const auto& regs = Core::System::GetInstance().GPU().Maxwell3D().regs;
+    const auto& regs = system.GPU().Maxwell3D().regs;
     const std::array<bool, Maxwell::Regs::NumClipDistances> reg_state{
         regs.clip_distance_enabled.c0 != 0, regs.clip_distance_enabled.c1 != 0,
         regs.clip_distance_enabled.c2 != 0, regs.clip_distance_enabled.c3 != 0,
@@ -1072,7 +1045,7 @@ void RasterizerOpenGL::SyncClipCoef() {
 }
 
 void RasterizerOpenGL::SyncCullMode() {
-    const auto& regs = Core::System::GetInstance().GPU().Maxwell3D().regs;
+    const auto& regs = system.GPU().Maxwell3D().regs;
 
     state.cull.enabled = regs.cull.enabled != 0;
 
@@ -1096,14 +1069,14 @@ void RasterizerOpenGL::SyncCullMode() {
 }
 
 void RasterizerOpenGL::SyncPrimitiveRestart() {
-    const auto& regs = Core::System::GetInstance().GPU().Maxwell3D().regs;
+    const auto& regs = system.GPU().Maxwell3D().regs;
 
     state.primitive_restart.enabled = regs.primitive_restart.enabled;
     state.primitive_restart.index = regs.primitive_restart.index;
 }
 
 void RasterizerOpenGL::SyncDepthTestState() {
-    const auto& regs = Core::System::GetInstance().GPU().Maxwell3D().regs;
+    const auto& regs = system.GPU().Maxwell3D().regs;
 
     state.depth.test_enabled = regs.depth_test_enable != 0;
     state.depth.write_mask = regs.depth_write_enabled ? GL_TRUE : GL_FALSE;
@@ -1115,7 +1088,7 @@ void RasterizerOpenGL::SyncDepthTestState() {
 }
 
 void RasterizerOpenGL::SyncStencilTestState() {
-    const auto& regs = Core::System::GetInstance().GPU().Maxwell3D().regs;
+    const auto& regs = system.GPU().Maxwell3D().regs;
     state.stencil.test_enabled = regs.stencil_enable != 0;
 
     if (!regs.stencil_enable) {
@@ -1149,7 +1122,7 @@ void RasterizerOpenGL::SyncStencilTestState() {
 }
 
 void RasterizerOpenGL::SyncColorMask() {
-    const auto& regs = Core::System::GetInstance().GPU().Maxwell3D().regs;
+    const auto& regs = system.GPU().Maxwell3D().regs;
     const std::size_t count =
         regs.independent_blend_enable ? Tegra::Engines::Maxwell3D::Regs::NumRenderTargets : 1;
     for (std::size_t i = 0; i < count; i++) {
@@ -1163,18 +1136,18 @@ void RasterizerOpenGL::SyncColorMask() {
 }
 
 void RasterizerOpenGL::SyncMultiSampleState() {
-    const auto& regs = Core::System::GetInstance().GPU().Maxwell3D().regs;
+    const auto& regs = system.GPU().Maxwell3D().regs;
     state.multisample_control.alpha_to_coverage = regs.multisample_control.alpha_to_coverage != 0;
     state.multisample_control.alpha_to_one = regs.multisample_control.alpha_to_one != 0;
 }
 
 void RasterizerOpenGL::SyncFragmentColorClampState() {
-    const auto& regs = Core::System::GetInstance().GPU().Maxwell3D().regs;
+    const auto& regs = system.GPU().Maxwell3D().regs;
     state.fragment_color_clamp.enabled = regs.frag_color_clamp != 0;
 }
 
 void RasterizerOpenGL::SyncBlendState() {
-    const auto& regs = Core::System::GetInstance().GPU().Maxwell3D().regs;
+    const auto& regs = system.GPU().Maxwell3D().regs;
 
     state.blend_color.red = regs.blend_color.r;
     state.blend_color.green = regs.blend_color.g;
@@ -1216,7 +1189,7 @@ void RasterizerOpenGL::SyncBlendState() {
 }
 
 void RasterizerOpenGL::SyncLogicOpState() {
-    const auto& regs = Core::System::GetInstance().GPU().Maxwell3D().regs;
+    const auto& regs = system.GPU().Maxwell3D().regs;
 
     state.logic_op.enabled = regs.logic_op.enable != 0;
 
@@ -1230,7 +1203,7 @@ void RasterizerOpenGL::SyncLogicOpState() {
 }
 
 void RasterizerOpenGL::SyncScissorTest(OpenGLState& current_state) {
-    const auto& regs = Core::System::GetInstance().GPU().Maxwell3D().regs;
+    const auto& regs = system.GPU().Maxwell3D().regs;
     const bool geometry_shaders_enabled =
         regs.IsShaderConfigEnabled(static_cast<size_t>(Maxwell::ShaderProgram::Geometry));
     const std::size_t viewport_count =
@@ -1252,21 +1225,17 @@ void RasterizerOpenGL::SyncScissorTest(OpenGLState& current_state) {
 }
 
 void RasterizerOpenGL::SyncTransformFeedback() {
-    const auto& regs = Core::System::GetInstance().GPU().Maxwell3D().regs;
-
-    if (regs.tfb_enabled != 0) {
-        LOG_CRITICAL(Render_OpenGL, "Transform feedbacks are not implemented");
-        UNREACHABLE();
-    }
+    const auto& regs = system.GPU().Maxwell3D().regs;
+    UNIMPLEMENTED_IF_MSG(regs.tfb_enabled != 0, "Transform feedbacks are not implemented");
 }
 
 void RasterizerOpenGL::SyncPointState() {
-    const auto& regs = Core::System::GetInstance().GPU().Maxwell3D().regs;
+    const auto& regs = system.GPU().Maxwell3D().regs;
     state.point.size = regs.point_size;
 }
 
 void RasterizerOpenGL::SyncPolygonOffset() {
-    const auto& regs = Core::System::GetInstance().GPU().Maxwell3D().regs;
+    const auto& regs = system.GPU().Maxwell3D().regs;
     state.polygon_offset.fill_enable = regs.polygon_offset_fill_enable != 0;
     state.polygon_offset.line_enable = regs.polygon_offset_line_enable != 0;
     state.polygon_offset.point_enable = regs.polygon_offset_point_enable != 0;
@@ -1276,13 +1245,9 @@ void RasterizerOpenGL::SyncPolygonOffset() {
 }
 
 void RasterizerOpenGL::CheckAlphaTests() {
-    const auto& regs = Core::System::GetInstance().GPU().Maxwell3D().regs;
-
-    if (regs.alpha_test_enabled != 0 && regs.rt_control.count > 1) {
-        LOG_CRITICAL(Render_OpenGL, "Alpha Testing is enabled with Multiple Render Targets, "
-                                    "this behavior is undefined.");
-        UNREACHABLE();
-    }
+    const auto& regs = system.GPU().Maxwell3D().regs;
+    UNIMPLEMENTED_IF_MSG(regs.alpha_test_enabled != 0 && regs.rt_control.count > 1,
+                         "Alpha Testing is enabled with more than one rendertarget");
 }
 
 } // namespace OpenGL
diff --git a/src/video_core/renderer_opengl/gl_rasterizer.h b/src/video_core/renderer_opengl/gl_rasterizer.h
index 7f2bf0f8b..30f3e8acb 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer.h
+++ b/src/video_core/renderer_opengl/gl_rasterizer.h
@@ -5,6 +5,7 @@
 #pragma once
 
 #include <array>
+#include <atomic>
 #include <cstddef>
 #include <map>
 #include <memory>
@@ -33,6 +34,10 @@
 #include "video_core/renderer_opengl/gl_state.h"
 #include "video_core/renderer_opengl/gl_stream_buffer.h"
 
+namespace Core {
+class System;
+}
+
 namespace Core::Frontend {
 class EmuWindow;
 }
@@ -45,21 +50,26 @@ struct FramebufferCacheKey;
 
 class RasterizerOpenGL : public VideoCore::RasterizerInterface {
 public:
-    explicit RasterizerOpenGL(Core::Frontend::EmuWindow& renderer, ScreenInfo& info);
+    explicit RasterizerOpenGL(Core::Frontend::EmuWindow& window, Core::System& system,
+                              ScreenInfo& info);
     ~RasterizerOpenGL() override;
 
     void DrawArrays() override;
     void Clear() override;
     void FlushAll() override;
-    void FlushRegion(VAddr addr, u64 size) override;
-    void InvalidateRegion(VAddr addr, u64 size) override;
-    void FlushAndInvalidateRegion(VAddr addr, u64 size) override;
+    void FlushRegion(CacheAddr addr, u64 size) override;
+    void InvalidateRegion(CacheAddr addr, u64 size) override;
+    void FlushAndInvalidateRegion(CacheAddr addr, u64 size) override;
     bool AccelerateSurfaceCopy(const Tegra::Engines::Fermi2D::Regs::Surface& src,
-                               const Tegra::Engines::Fermi2D::Regs::Surface& dst) override;
+                               const Tegra::Engines::Fermi2D::Regs::Surface& dst,
+                               const Common::Rectangle<u32>& src_rect,
+                               const Common::Rectangle<u32>& dst_rect) override;
     bool AccelerateDisplay(const Tegra::FramebufferConfig& config, VAddr framebuffer_addr,
                            u32 pixel_stride) override;
     bool AccelerateDrawBatch(bool is_indexed) override;
-    void UpdatePagesCachedCount(Tegra::GPUVAddr addr, u64 size, int delta) override;
+    void UpdatePagesCachedCount(VAddr addr, u64 size, int delta) override;
+    void LoadDiskResources(const std::atomic_bool& stop_loading,
+                           const VideoCore::DiskResourceLoadCallback& callback) override;
 
     /// Maximum supported size that a constbuffer can have in bytes.
     static constexpr std::size_t MaxConstbufferSize = 0x10000;
@@ -84,11 +94,12 @@ private:
     private:
         Tegra::Texture::TextureFilter mag_filter = Tegra::Texture::TextureFilter::Nearest;
         Tegra::Texture::TextureFilter min_filter = Tegra::Texture::TextureFilter::Nearest;
-        Tegra::Texture::TextureMipmapFilter mip_filter = Tegra::Texture::TextureMipmapFilter::None;
+        Tegra::Texture::TextureMipmapFilter mipmap_filter =
+            Tegra::Texture::TextureMipmapFilter::None;
         Tegra::Texture::WrapMode wrap_u = Tegra::Texture::WrapMode::ClampToEdge;
         Tegra::Texture::WrapMode wrap_v = Tegra::Texture::WrapMode::ClampToEdge;
         Tegra::Texture::WrapMode wrap_p = Tegra::Texture::WrapMode::ClampToEdge;
-        bool uses_depth_compare = false;
+        bool use_depth_compare = false;
         Tegra::Texture::DepthCompareFunc depth_compare_func =
             Tegra::Texture::DepthCompareFunc::Always;
         GLvec4 border_color = {};
@@ -204,6 +215,7 @@ private:
     GlobalRegionCacheOpenGL global_cache;
 
     Core::Frontend::EmuWindow& emu_window;
+    Core::System& system;
 
     ScreenInfo& screen_info;
 
diff --git a/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp b/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp
index 50286432d..57329cd61 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp
@@ -3,6 +3,7 @@
 // Refer to the license.txt file included.
 
 #include <algorithm>
+#include <optional>
 #include <glad/glad.h>
 
 #include "common/alignment.h"
@@ -18,10 +19,9 @@
 #include "video_core/morton.h"
 #include "video_core/renderer_opengl/gl_rasterizer.h"
 #include "video_core/renderer_opengl/gl_rasterizer_cache.h"
-#include "video_core/renderer_opengl/gl_state.h"
 #include "video_core/renderer_opengl/utils.h"
 #include "video_core/surface.h"
-#include "video_core/textures/astc.h"
+#include "video_core/textures/convert.h"
 #include "video_core/textures/decoders.h"
 
 namespace OpenGL {
@@ -44,23 +44,22 @@ struct FormatTuple {
     bool compressed;
 };
 
-static void ApplyTextureDefaults(GLenum target, u32 max_mip_level) {
-    glTexParameteri(target, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
-    glTexParameteri(target, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
-    glTexParameteri(target, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
-    glTexParameteri(target, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
-    glTexParameteri(target, GL_TEXTURE_MAX_LEVEL, max_mip_level - 1);
+static void ApplyTextureDefaults(GLuint texture, u32 max_mip_level) {
+    glTextureParameteri(texture, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
+    glTextureParameteri(texture, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
+    glTextureParameteri(texture, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
+    glTextureParameteri(texture, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
+    glTextureParameteri(texture, GL_TEXTURE_MAX_LEVEL, max_mip_level - 1);
     if (max_mip_level == 1) {
-        glTexParameterf(target, GL_TEXTURE_LOD_BIAS, 1000.0);
+        glTextureParameterf(texture, GL_TEXTURE_LOD_BIAS, 1000.0);
     }
 }
 
 void SurfaceParams::InitCacheParameters(Tegra::GPUVAddr gpu_addr_) {
     auto& memory_manager{Core::System::GetInstance().GPU().MemoryManager()};
-    const auto cpu_addr{memory_manager.GpuToCpuAddress(gpu_addr_)};
 
-    addr = cpu_addr ? *cpu_addr : 0;
     gpu_addr = gpu_addr_;
+    host_ptr = memory_manager.GetPointer(gpu_addr_);
     size_in_bytes = SizeInBytesRaw();
 
     if (IsPixelFormatASTC(pixel_format)) {
@@ -126,6 +125,9 @@ std::size_t SurfaceParams::InnerMemorySize(bool force_gl, bool layer_only,
 
     params.width = Common::AlignUp(config.tic.Width(), GetCompressionFactor(params.pixel_format));
     params.height = Common::AlignUp(config.tic.Height(), GetCompressionFactor(params.pixel_format));
+    if (!params.is_tiled) {
+        params.pitch = config.tic.Pitch();
+    }
     params.unaligned_height = config.tic.Height();
     params.target = SurfaceTargetFromTextureType(config.tic.texture_type);
     params.identity = SurfaceClass::Uploaded;
@@ -192,7 +194,13 @@ std::size_t SurfaceParams::InnerMemorySize(bool force_gl, bool layer_only,
                              config.format == Tegra::RenderTargetFormat::RGBA8_SRGB;
     params.component_type = ComponentTypeFromRenderTarget(config.format);
     params.type = GetFormatType(params.pixel_format);
-    params.width = config.width;
+    if (params.is_tiled) {
+        params.width = config.width;
+    } else {
+        params.pitch = config.width;
+        const u32 bpp = params.GetFormatBpp() / 8;
+        params.width = params.pitch / bpp;
+    }
     params.height = config.height;
     params.unaligned_height = config.height;
     params.target = SurfaceTarget::Texture2D;
@@ -391,7 +399,28 @@ static const FormatTuple& GetFormatTuple(PixelFormat pixel_format, ComponentType
     return format;
 }
 
-MathUtil::Rectangle<u32> SurfaceParams::GetRect(u32 mip_level) const {
+/// Returns the discrepant array target
+constexpr GLenum GetArrayDiscrepantTarget(SurfaceTarget target) {
+    switch (target) {
+    case SurfaceTarget::Texture1D:
+        return GL_TEXTURE_1D_ARRAY;
+    case SurfaceTarget::Texture2D:
+        return GL_TEXTURE_2D_ARRAY;
+    case SurfaceTarget::Texture3D:
+        return GL_NONE;
+    case SurfaceTarget::Texture1DArray:
+        return GL_TEXTURE_1D;
+    case SurfaceTarget::Texture2DArray:
+        return GL_TEXTURE_2D;
+    case SurfaceTarget::TextureCubemap:
+        return GL_TEXTURE_CUBE_MAP_ARRAY;
+    case SurfaceTarget::TextureCubeArray:
+        return GL_TEXTURE_CUBE_MAP;
+    }
+    return GL_NONE;
+}
+
+Common::Rectangle<u32> SurfaceParams::GetRect(u32 mip_level) const {
     u32 actual_height{std::max(1U, unaligned_height >> mip_level)};
     if (IsPixelFormatASTC(pixel_format)) {
         // ASTC formats must stop at the ATSC block size boundary
@@ -415,8 +444,8 @@ void SwizzleFunc(const MortonSwizzleMode& mode, const SurfaceParams& params,
         for (u32 i = 0; i < params.depth; i++) {
             MortonSwizzle(mode, params.pixel_format, params.MipWidth(mip_level),
                           params.MipBlockHeight(mip_level), params.MipHeight(mip_level),
-                          params.MipBlockDepth(mip_level), params.tile_width_spacing, 1,
-                          gl_buffer.data() + offset_gl, gl_size, params.addr + offset);
+                          params.MipBlockDepth(mip_level), 1, params.tile_width_spacing,
+                          gl_buffer.data() + offset_gl, params.host_ptr + offset);
             offset += layer_size;
             offset_gl += gl_size;
         }
@@ -425,11 +454,12 @@ void SwizzleFunc(const MortonSwizzleMode& mode, const SurfaceParams& params,
         MortonSwizzle(mode, params.pixel_format, params.MipWidth(mip_level),
                       params.MipBlockHeight(mip_level), params.MipHeight(mip_level),
                       params.MipBlockDepth(mip_level), depth, params.tile_width_spacing,
-                      gl_buffer.data(), gl_buffer.size(), params.addr + offset);
+                      gl_buffer.data(), params.host_ptr + offset);
     }
 }
 
-static void FastCopySurface(const Surface& src_surface, const Surface& dst_surface) {
+void RasterizerCacheOpenGL::FastCopySurface(const Surface& src_surface,
+                                            const Surface& dst_surface) {
     const auto& src_params{src_surface->GetSurfaceParams()};
     const auto& dst_params{dst_surface->GetSurfaceParams()};
 
@@ -439,12 +469,15 @@ static void FastCopySurface(const Surface& src_surface, const Surface& dst_surfa
     glCopyImageSubData(src_surface->Texture().handle, SurfaceTargetToGL(src_params.target), 0, 0, 0,
                        0, dst_surface->Texture().handle, SurfaceTargetToGL(dst_params.target), 0, 0,
                        0, 0, width, height, 1);
+
+    dst_surface->MarkAsModified(true, *this);
 }
 
 MICROPROFILE_DEFINE(OpenGL_CopySurface, "OpenGL", "CopySurface", MP_RGB(128, 192, 64));
-static void CopySurface(const Surface& src_surface, const Surface& dst_surface,
-                        const GLuint copy_pbo_handle, const GLenum src_attachment = 0,
-                        const GLenum dst_attachment = 0, const std::size_t cubemap_face = 0) {
+void RasterizerCacheOpenGL::CopySurface(const Surface& src_surface, const Surface& dst_surface,
+                                        const GLuint copy_pbo_handle, const GLenum src_attachment,
+                                        const GLenum dst_attachment,
+                                        const std::size_t cubemap_face) {
     MICROPROFILE_SCOPE(OpenGL_CopySurface);
     ASSERT_MSG(dst_attachment == 0, "Unimplemented");
 
@@ -479,9 +512,9 @@ static void CopySurface(const Surface& src_surface, const Surface& dst_surface,
                               "reinterpretation but the texture is tiled.");
         }
         const std::size_t remaining_size = dst_params.size_in_bytes - src_params.size_in_bytes;
-
+        auto& memory_manager{Core::System::GetInstance().GPU().MemoryManager()};
         glBufferSubData(GL_PIXEL_PACK_BUFFER, src_params.size_in_bytes, remaining_size,
-                        Memory::GetPointer(dst_params.addr + src_params.size_in_bytes));
+                        memory_manager.GetPointer(dst_params.gpu_addr + src_params.size_in_bytes));
     }
 
     glBindBuffer(GL_PIXEL_PACK_BUFFER, 0);
@@ -524,62 +557,52 @@ static void CopySurface(const Surface& src_surface, const Surface& dst_surface,
         }
         glBindBuffer(GL_PIXEL_UNPACK_BUFFER, 0);
     }
+
+    dst_surface->MarkAsModified(true, *this);
 }
 
 CachedSurface::CachedSurface(const SurfaceParams& params)
-    : params(params), gl_target(SurfaceTargetToGL(params.target)),
-      cached_size_in_bytes(params.size_in_bytes) {
-    texture.Create();
-    const auto& rect{params.GetRect()};
-
-    // Keep track of previous texture bindings
-    OpenGLState cur_state = OpenGLState::GetCurState();
-    const auto& old_tex = cur_state.texture_units[0];
-    SCOPE_EXIT({
-        cur_state.texture_units[0] = old_tex;
-        cur_state.Apply();
-    });
-
-    cur_state.texture_units[0].texture = texture.handle;
-    cur_state.texture_units[0].target = SurfaceTargetToGL(params.target);
-    cur_state.Apply();
-    glActiveTexture(GL_TEXTURE0);
+    : params{params}, gl_target{SurfaceTargetToGL(params.target)},
+      cached_size_in_bytes{params.size_in_bytes}, RasterizerCacheObject{params.host_ptr} {
+    texture.Create(gl_target);
+
+    // TODO(Rodrigo): Using params.GetRect() returns a different size than using its Mip*(0)
+    // alternatives. This signals a bug on those functions.
+    const auto width = static_cast<GLsizei>(params.MipWidth(0));
+    const auto height = static_cast<GLsizei>(params.MipHeight(0));
+    memory_size = params.MemorySize();
+    reinterpreted = false;
 
     const auto& format_tuple = GetFormatTuple(params.pixel_format, params.component_type);
     gl_internal_format = format_tuple.internal_format;
-    gl_is_compressed = format_tuple.compressed;
 
-    if (!format_tuple.compressed) {
-        // Only pre-create the texture for non-compressed textures.
-        switch (params.target) {
-        case SurfaceTarget::Texture1D:
-            glTexStorage1D(SurfaceTargetToGL(params.target), params.max_mip_level,
-                           format_tuple.internal_format, rect.GetWidth());
-            break;
-        case SurfaceTarget::Texture2D:
-        case SurfaceTarget::TextureCubemap:
-            glTexStorage2D(SurfaceTargetToGL(params.target), params.max_mip_level,
-                           format_tuple.internal_format, rect.GetWidth(), rect.GetHeight());
-            break;
-        case SurfaceTarget::Texture3D:
-        case SurfaceTarget::Texture2DArray:
-        case SurfaceTarget::TextureCubeArray:
-            glTexStorage3D(SurfaceTargetToGL(params.target), params.max_mip_level,
-                           format_tuple.internal_format, rect.GetWidth(), rect.GetHeight(),
-                           params.depth);
-            break;
-        default:
-            LOG_CRITICAL(Render_OpenGL, "Unimplemented surface target={}",
-                         static_cast<u32>(params.target));
-            UNREACHABLE();
-            glTexStorage2D(GL_TEXTURE_2D, params.max_mip_level, format_tuple.internal_format,
-                           rect.GetWidth(), rect.GetHeight());
-        }
+    switch (params.target) {
+    case SurfaceTarget::Texture1D:
+        glTextureStorage1D(texture.handle, params.max_mip_level, format_tuple.internal_format,
+                           width);
+        break;
+    case SurfaceTarget::Texture2D:
+    case SurfaceTarget::TextureCubemap:
+        glTextureStorage2D(texture.handle, params.max_mip_level, format_tuple.internal_format,
+                           width, height);
+        break;
+    case SurfaceTarget::Texture3D:
+    case SurfaceTarget::Texture2DArray:
+    case SurfaceTarget::TextureCubeArray:
+        glTextureStorage3D(texture.handle, params.max_mip_level, format_tuple.internal_format,
+                           width, height, params.depth);
+        break;
+    default:
+        LOG_CRITICAL(Render_OpenGL, "Unimplemented surface target={}",
+                     static_cast<u32>(params.target));
+        UNREACHABLE();
+        glTextureStorage2D(texture.handle, params.max_mip_level, format_tuple.internal_format,
+                           width, height);
     }
 
-    ApplyTextureDefaults(SurfaceTargetToGL(params.target), params.max_mip_level);
+    ApplyTextureDefaults(texture.handle, params.max_mip_level);
 
-    OpenGL::LabelGLObject(GL_TEXTURE, texture.handle, params.addr, params.IdentityString());
+    OpenGL::LabelGLObject(GL_TEXTURE, texture.handle, params.gpu_addr, params.IdentityString());
 
     // Clamp size to mapped GPU memory region
     // TODO(bunnei): Super Mario Odyssey maps a 0x40000 byte region and then uses it for a 0x80000
@@ -592,103 +615,8 @@ CachedSurface::CachedSurface(const SurfaceParams& params)
         LOG_ERROR(HW_GPU, "Surface size {} exceeds region size {}", params.size_in_bytes, max_size);
         cached_size_in_bytes = max_size;
     }
-}
-
-static void ConvertS8Z24ToZ24S8(std::vector<u8>& data, u32 width, u32 height, bool reverse) {
-    union S8Z24 {
-        BitField<0, 24, u32> z24;
-        BitField<24, 8, u32> s8;
-    };
-    static_assert(sizeof(S8Z24) == 4, "S8Z24 is incorrect size");
-
-    union Z24S8 {
-        BitField<0, 8, u32> s8;
-        BitField<8, 24, u32> z24;
-    };
-    static_assert(sizeof(Z24S8) == 4, "Z24S8 is incorrect size");
-
-    S8Z24 s8z24_pixel{};
-    Z24S8 z24s8_pixel{};
-    constexpr auto bpp{GetBytesPerPixel(PixelFormat::S8Z24)};
-    for (std::size_t y = 0; y < height; ++y) {
-        for (std::size_t x = 0; x < width; ++x) {
-            const std::size_t offset{bpp * (y * width + x)};
-            if (reverse) {
-                std::memcpy(&z24s8_pixel, &data[offset], sizeof(Z24S8));
-                s8z24_pixel.s8.Assign(z24s8_pixel.s8);
-                s8z24_pixel.z24.Assign(z24s8_pixel.z24);
-                std::memcpy(&data[offset], &s8z24_pixel, sizeof(S8Z24));
-            } else {
-                std::memcpy(&s8z24_pixel, &data[offset], sizeof(S8Z24));
-                z24s8_pixel.s8.Assign(s8z24_pixel.s8);
-                z24s8_pixel.z24.Assign(s8z24_pixel.z24);
-                std::memcpy(&data[offset], &z24s8_pixel, sizeof(Z24S8));
-            }
-        }
-    }
-}
-
-/**
- * Helper function to perform software conversion (as needed) when loading a buffer from Switch
- * memory. This is for Maxwell pixel formats that cannot be represented as-is in OpenGL or with
- * typical desktop GPUs.
- */
-static void ConvertFormatAsNeeded_LoadGLBuffer(std::vector<u8>& data, PixelFormat pixel_format,
-                                               u32 width, u32 height, u32 depth) {
-    switch (pixel_format) {
-    case PixelFormat::ASTC_2D_4X4:
-    case PixelFormat::ASTC_2D_8X8:
-    case PixelFormat::ASTC_2D_8X5:
-    case PixelFormat::ASTC_2D_5X4:
-    case PixelFormat::ASTC_2D_5X5:
-    case PixelFormat::ASTC_2D_4X4_SRGB:
-    case PixelFormat::ASTC_2D_8X8_SRGB:
-    case PixelFormat::ASTC_2D_8X5_SRGB:
-    case PixelFormat::ASTC_2D_5X4_SRGB:
-    case PixelFormat::ASTC_2D_5X5_SRGB:
-    case PixelFormat::ASTC_2D_10X8:
-    case PixelFormat::ASTC_2D_10X8_SRGB: {
-        // Convert ASTC pixel formats to RGBA8, as most desktop GPUs do not support ASTC.
-        u32 block_width{};
-        u32 block_height{};
-        std::tie(block_width, block_height) = GetASTCBlockSize(pixel_format);
-        data =
-            Tegra::Texture::ASTC::Decompress(data, width, height, depth, block_width, block_height);
-        break;
-    }
-    case PixelFormat::S8Z24:
-        // Convert the S8Z24 depth format to Z24S8, as OpenGL does not support S8Z24.
-        ConvertS8Z24ToZ24S8(data, width, height, false);
-        break;
-    }
-}
 
-/**
- * Helper function to perform software conversion (as needed) when flushing a buffer from OpenGL to
- * Switch memory. This is for Maxwell pixel formats that cannot be represented as-is in OpenGL or
- * with typical desktop GPUs.
- */
-static void ConvertFormatAsNeeded_FlushGLBuffer(std::vector<u8>& data, PixelFormat pixel_format,
-                                                u32 width, u32 height) {
-    switch (pixel_format) {
-    case PixelFormat::ASTC_2D_4X4:
-    case PixelFormat::ASTC_2D_8X8:
-    case PixelFormat::ASTC_2D_4X4_SRGB:
-    case PixelFormat::ASTC_2D_8X8_SRGB:
-    case PixelFormat::ASTC_2D_5X5:
-    case PixelFormat::ASTC_2D_5X5_SRGB:
-    case PixelFormat::ASTC_2D_10X8:
-    case PixelFormat::ASTC_2D_10X8_SRGB: {
-        LOG_CRITICAL(HW_GPU, "Conversion of format {} after texture flushing is not implemented",
-                     static_cast<u32>(pixel_format));
-        UNREACHABLE();
-        break;
-    }
-    case PixelFormat::S8Z24:
-        // Convert the Z24S8 depth format to S8Z24, as OpenGL does not support S8Z24.
-        ConvertS8Z24ToZ24S8(data, width, height, true);
-        break;
-    }
+    cpu_addr = *memory_manager.GpuToCpuAddress(params.gpu_addr);
 }
 
 MICROPROFILE_DEFINE(OpenGL_SurfaceLoad, "OpenGL", "Surface Load", MP_RGB(128, 192, 64));
@@ -703,13 +631,31 @@ void CachedSurface::LoadGLBuffer() {
         for (u32 i = 0; i < params.max_mip_level; i++)
             SwizzleFunc(MortonSwizzleMode::MortonToLinear, params, gl_buffer[i], i);
     } else {
-        const auto texture_src_data{Memory::GetPointer(params.addr)};
-        const auto texture_src_data_end{texture_src_data + params.size_in_bytes_gl};
-        gl_buffer[0].assign(texture_src_data, texture_src_data_end);
+        const u32 bpp = params.GetFormatBpp() / 8;
+        const u32 copy_size = params.width * bpp;
+        if (params.pitch == copy_size) {
+            std::memcpy(gl_buffer[0].data(), params.host_ptr, params.size_in_bytes_gl);
+        } else {
+            const u8* start{params.host_ptr};
+            u8* write_to = gl_buffer[0].data();
+            for (u32 h = params.height; h > 0; h--) {
+                std::memcpy(write_to, start, copy_size);
+                start += params.pitch;
+                write_to += copy_size;
+            }
+        }
     }
     for (u32 i = 0; i < params.max_mip_level; i++) {
-        ConvertFormatAsNeeded_LoadGLBuffer(gl_buffer[i], params.pixel_format, params.MipWidth(i),
-                                           params.MipHeight(i), params.MipDepth(i));
+        const u32 width = params.MipWidth(i);
+        const u32 height = params.MipHeight(i);
+        const u32 depth = params.MipDepth(i);
+        if (VideoCore::Surface::IsPixelFormatASTC(params.pixel_format)) {
+            // Reserve size for RGBA8 conversion
+            constexpr std::size_t rgba_bpp = 4;
+            gl_buffer[i].resize(std::max(gl_buffer[i].size(), width * height * depth * rgba_bpp));
+        }
+        Tegra::Texture::ConvertFromGuestToHost(gl_buffer[i].data(), params.pixel_format, width,
+                                               height, depth, true, true);
     }
 }
 
@@ -732,17 +678,27 @@ void CachedSurface::FlushGLBuffer() {
     glGetTextureImage(texture.handle, 0, tuple.format, tuple.type,
                       static_cast<GLsizei>(gl_buffer[0].size()), gl_buffer[0].data());
     glPixelStorei(GL_PACK_ROW_LENGTH, 0);
-    ConvertFormatAsNeeded_FlushGLBuffer(gl_buffer[0], params.pixel_format, params.width,
-                                        params.height);
-    const u8* const texture_src_data = Memory::GetPointer(params.addr);
-    ASSERT(texture_src_data);
+    Tegra::Texture::ConvertFromHostToGuest(gl_buffer[0].data(), params.pixel_format, params.width,
+                                           params.height, params.depth, true, true);
     if (params.is_tiled) {
         ASSERT_MSG(params.block_width == 1, "Block width is defined as {} on texture type {}",
                    params.block_width, static_cast<u32>(params.target));
 
         SwizzleFunc(MortonSwizzleMode::LinearToMorton, params, gl_buffer[0], 0);
     } else {
-        std::memcpy(Memory::GetPointer(GetAddr()), gl_buffer[0].data(), GetSizeInBytes());
+        const u32 bpp = params.GetFormatBpp() / 8;
+        const u32 copy_size = params.width * bpp;
+        if (params.pitch == copy_size) {
+            std::memcpy(params.host_ptr, gl_buffer[0].data(), GetSizeInBytes());
+        } else {
+            u8* start{params.host_ptr};
+            const u8* read_to = gl_buffer[0].data();
+            for (u32 h = params.height; h > 0; h--) {
+                std::memcpy(start, read_to, copy_size);
+                start += params.pitch;
+                read_to += copy_size;
+            }
+        }
     }
 }
 
@@ -751,63 +707,50 @@ void CachedSurface::UploadGLMipmapTexture(u32 mip_map, GLuint read_fb_handle,
     const auto& rect{params.GetRect(mip_map)};
 
     // Load data from memory to the surface
-    const GLint x0 = static_cast<GLint>(rect.left);
-    const GLint y0 = static_cast<GLint>(rect.bottom);
-    std::size_t buffer_offset =
+    const auto x0 = static_cast<GLint>(rect.left);
+    const auto y0 = static_cast<GLint>(rect.bottom);
+    auto buffer_offset =
         static_cast<std::size_t>(static_cast<std::size_t>(y0) * params.MipWidth(mip_map) +
                                  static_cast<std::size_t>(x0)) *
         GetBytesPerPixel(params.pixel_format);
 
     const FormatTuple& tuple = GetFormatTuple(params.pixel_format, params.component_type);
-    const GLuint target_tex = texture.handle;
-    OpenGLState cur_state = OpenGLState::GetCurState();
-
-    const auto& old_tex = cur_state.texture_units[0];
-    SCOPE_EXIT({
-        cur_state.texture_units[0] = old_tex;
-        cur_state.Apply();
-    });
-    cur_state.texture_units[0].texture = target_tex;
-    cur_state.texture_units[0].target = SurfaceTargetToGL(params.target);
-    cur_state.Apply();
 
     // Ensure no bad interactions with GL_UNPACK_ALIGNMENT
     ASSERT(params.MipWidth(mip_map) * GetBytesPerPixel(params.pixel_format) % 4 == 0);
     glPixelStorei(GL_UNPACK_ROW_LENGTH, static_cast<GLint>(params.MipWidth(mip_map)));
 
-    GLsizei image_size = static_cast<GLsizei>(params.GetMipmapSizeGL(mip_map, false));
-    glActiveTexture(GL_TEXTURE0);
+    const auto image_size = static_cast<GLsizei>(params.GetMipmapSizeGL(mip_map, false));
     if (tuple.compressed) {
         switch (params.target) {
         case SurfaceTarget::Texture2D:
-            glCompressedTexImage2D(SurfaceTargetToGL(params.target), mip_map, tuple.internal_format,
-                                   static_cast<GLsizei>(params.MipWidth(mip_map)),
-                                   static_cast<GLsizei>(params.MipHeight(mip_map)), 0, image_size,
-                                   &gl_buffer[mip_map][buffer_offset]);
+            glCompressedTextureSubImage2D(
+                texture.handle, mip_map, 0, 0, static_cast<GLsizei>(params.MipWidth(mip_map)),
+                static_cast<GLsizei>(params.MipHeight(mip_map)), tuple.internal_format, image_size,
+                &gl_buffer[mip_map][buffer_offset]);
             break;
         case SurfaceTarget::Texture3D:
-            glCompressedTexImage3D(SurfaceTargetToGL(params.target), mip_map, tuple.internal_format,
-                                   static_cast<GLsizei>(params.MipWidth(mip_map)),
-                                   static_cast<GLsizei>(params.MipHeight(mip_map)),
-                                   static_cast<GLsizei>(params.MipDepth(mip_map)), 0, image_size,
-                                   &gl_buffer[mip_map][buffer_offset]);
+            glCompressedTextureSubImage3D(
+                texture.handle, mip_map, 0, 0, 0, static_cast<GLsizei>(params.MipWidth(mip_map)),
+                static_cast<GLsizei>(params.MipHeight(mip_map)),
+                static_cast<GLsizei>(params.MipDepth(mip_map)), tuple.internal_format, image_size,
+                &gl_buffer[mip_map][buffer_offset]);
             break;
         case SurfaceTarget::Texture2DArray:
         case SurfaceTarget::TextureCubeArray:
-            glCompressedTexImage3D(SurfaceTargetToGL(params.target), mip_map, tuple.internal_format,
-                                   static_cast<GLsizei>(params.MipWidth(mip_map)),
-                                   static_cast<GLsizei>(params.MipHeight(mip_map)),
-                                   static_cast<GLsizei>(params.depth), 0, image_size,
-                                   &gl_buffer[mip_map][buffer_offset]);
+            glCompressedTextureSubImage3D(
+                texture.handle, mip_map, 0, 0, 0, static_cast<GLsizei>(params.MipWidth(mip_map)),
+                static_cast<GLsizei>(params.MipHeight(mip_map)), static_cast<GLsizei>(params.depth),
+                tuple.internal_format, image_size, &gl_buffer[mip_map][buffer_offset]);
             break;
         case SurfaceTarget::TextureCubemap: {
-            GLsizei layer_size = static_cast<GLsizei>(params.LayerSizeGL(mip_map));
+            const auto layer_size = static_cast<GLsizei>(params.LayerSizeGL(mip_map));
             for (std::size_t face = 0; face < params.depth; ++face) {
-                glCompressedTexImage2D(static_cast<GLenum>(GL_TEXTURE_CUBE_MAP_POSITIVE_X + face),
-                                       mip_map, tuple.internal_format,
-                                       static_cast<GLsizei>(params.MipWidth(mip_map)),
-                                       static_cast<GLsizei>(params.MipHeight(mip_map)), 0,
-                                       layer_size, &gl_buffer[mip_map][buffer_offset]);
+                glCompressedTextureSubImage3D(
+                    texture.handle, mip_map, 0, 0, static_cast<GLint>(face),
+                    static_cast<GLsizei>(params.MipWidth(mip_map)),
+                    static_cast<GLsizei>(params.MipHeight(mip_map)), 1, tuple.internal_format,
+                    layer_size, &gl_buffer[mip_map][buffer_offset]);
                 buffer_offset += layer_size;
             }
             break;
@@ -816,46 +759,43 @@ void CachedSurface::UploadGLMipmapTexture(u32 mip_map, GLuint read_fb_handle,
             LOG_CRITICAL(Render_OpenGL, "Unimplemented surface target={}",
                          static_cast<u32>(params.target));
             UNREACHABLE();
-            glCompressedTexImage2D(GL_TEXTURE_2D, mip_map, tuple.internal_format,
-                                   static_cast<GLsizei>(params.MipWidth(mip_map)),
-                                   static_cast<GLsizei>(params.MipHeight(mip_map)), 0,
-                                   static_cast<GLsizei>(params.size_in_bytes_gl),
-                                   &gl_buffer[mip_map][buffer_offset]);
+            glCompressedTextureSubImage2D(
+                texture.handle, mip_map, 0, 0, static_cast<GLsizei>(params.MipWidth(mip_map)),
+                static_cast<GLsizei>(params.MipHeight(mip_map)), tuple.internal_format,
+                static_cast<GLsizei>(params.size_in_bytes_gl), &gl_buffer[mip_map][buffer_offset]);
         }
     } else {
-
         switch (params.target) {
         case SurfaceTarget::Texture1D:
-            glTexSubImage1D(SurfaceTargetToGL(params.target), mip_map, x0,
-                            static_cast<GLsizei>(rect.GetWidth()), tuple.format, tuple.type,
-                            &gl_buffer[mip_map][buffer_offset]);
+            glTextureSubImage1D(texture.handle, mip_map, x0, static_cast<GLsizei>(rect.GetWidth()),
+                                tuple.format, tuple.type, &gl_buffer[mip_map][buffer_offset]);
             break;
         case SurfaceTarget::Texture2D:
-            glTexSubImage2D(SurfaceTargetToGL(params.target), mip_map, x0, y0,
-                            static_cast<GLsizei>(rect.GetWidth()),
-                            static_cast<GLsizei>(rect.GetHeight()), tuple.format, tuple.type,
-                            &gl_buffer[mip_map][buffer_offset]);
+            glTextureSubImage2D(texture.handle, mip_map, x0, y0,
+                                static_cast<GLsizei>(rect.GetWidth()),
+                                static_cast<GLsizei>(rect.GetHeight()), tuple.format, tuple.type,
+                                &gl_buffer[mip_map][buffer_offset]);
             break;
         case SurfaceTarget::Texture3D:
-            glTexSubImage3D(SurfaceTargetToGL(params.target), mip_map, x0, y0, 0,
-                            static_cast<GLsizei>(rect.GetWidth()),
-                            static_cast<GLsizei>(rect.GetHeight()), params.MipDepth(mip_map),
-                            tuple.format, tuple.type, &gl_buffer[mip_map][buffer_offset]);
+            glTextureSubImage3D(texture.handle, mip_map, x0, y0, 0,
+                                static_cast<GLsizei>(rect.GetWidth()),
+                                static_cast<GLsizei>(rect.GetHeight()), params.MipDepth(mip_map),
+                                tuple.format, tuple.type, &gl_buffer[mip_map][buffer_offset]);
             break;
         case SurfaceTarget::Texture2DArray:
         case SurfaceTarget::TextureCubeArray:
-            glTexSubImage3D(SurfaceTargetToGL(params.target), mip_map, x0, y0, 0,
-                            static_cast<GLsizei>(rect.GetWidth()),
-                            static_cast<GLsizei>(rect.GetHeight()), params.depth, tuple.format,
-                            tuple.type, &gl_buffer[mip_map][buffer_offset]);
+            glTextureSubImage3D(texture.handle, mip_map, x0, y0, 0,
+                                static_cast<GLsizei>(rect.GetWidth()),
+                                static_cast<GLsizei>(rect.GetHeight()), params.depth, tuple.format,
+                                tuple.type, &gl_buffer[mip_map][buffer_offset]);
             break;
         case SurfaceTarget::TextureCubemap: {
             std::size_t start = buffer_offset;
             for (std::size_t face = 0; face < params.depth; ++face) {
-                glTexSubImage2D(static_cast<GLenum>(GL_TEXTURE_CUBE_MAP_POSITIVE_X + face), mip_map,
-                                x0, y0, static_cast<GLsizei>(rect.GetWidth()),
-                                static_cast<GLsizei>(rect.GetHeight()), tuple.format, tuple.type,
-                                &gl_buffer[mip_map][buffer_offset]);
+                glTextureSubImage3D(texture.handle, mip_map, x0, y0, static_cast<GLint>(face),
+                                    static_cast<GLsizei>(rect.GetWidth()),
+                                    static_cast<GLsizei>(rect.GetHeight()), 1, tuple.format,
+                                    tuple.type, &gl_buffer[mip_map][buffer_offset]);
                 buffer_offset += params.LayerSizeGL(mip_map);
             }
             break;
@@ -864,41 +804,33 @@ void CachedSurface::UploadGLMipmapTexture(u32 mip_map, GLuint read_fb_handle,
             LOG_CRITICAL(Render_OpenGL, "Unimplemented surface target={}",
                          static_cast<u32>(params.target));
             UNREACHABLE();
-            glTexSubImage2D(GL_TEXTURE_2D, mip_map, x0, y0, static_cast<GLsizei>(rect.GetWidth()),
-                            static_cast<GLsizei>(rect.GetHeight()), tuple.format, tuple.type,
-                            &gl_buffer[mip_map][buffer_offset]);
+            glTextureSubImage2D(texture.handle, mip_map, x0, y0,
+                                static_cast<GLsizei>(rect.GetWidth()),
+                                static_cast<GLsizei>(rect.GetHeight()), tuple.format, tuple.type,
+                                &gl_buffer[mip_map][buffer_offset]);
         }
     }
 
     glPixelStorei(GL_UNPACK_ROW_LENGTH, 0);
 }
 
-void CachedSurface::EnsureTextureView() {
-    if (texture_view.handle != 0)
+void CachedSurface::EnsureTextureDiscrepantView() {
+    if (discrepant_view.handle != 0)
         return;
-    // Compressed texture are not being created with immutable storage
-    UNIMPLEMENTED_IF(gl_is_compressed);
 
-    const GLenum target{TargetLayer()};
+    const GLenum target{GetArrayDiscrepantTarget(params.target)};
+    ASSERT(target != GL_NONE);
+
     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, min_level,
+    glGenTextures(1, &discrepant_view.handle);
+    glTextureView(discrepant_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];
-    SCOPE_EXIT({
-        cur_state.texture_units[0] = old_tex;
-        cur_state.Apply();
-    });
-    cur_state.texture_units[0].texture = texture_view.handle;
-    cur_state.texture_units[0].target = target;
-    cur_state.Apply();
-
-    ApplyTextureDefaults(target, params.max_mip_level);
+    ApplyTextureDefaults(discrepant_view.handle, params.max_mip_level);
+    glTextureParameteriv(discrepant_view.handle, GL_TEXTURE_SWIZZLE_RGBA,
+                         reinterpret_cast<const GLint*>(swizzle.data()));
 }
 
 MICROPROFILE_DEFINE(OpenGL_TextureUL, "OpenGL", "Texture Upload", MP_RGB(128, 192, 64));
@@ -909,6 +841,25 @@ void CachedSurface::UploadGLTexture(GLuint read_fb_handle, GLuint draw_fb_handle
         UploadGLMipmapTexture(i, read_fb_handle, draw_fb_handle);
 }
 
+void CachedSurface::UpdateSwizzle(Tegra::Texture::SwizzleSource swizzle_x,
+                                  Tegra::Texture::SwizzleSource swizzle_y,
+                                  Tegra::Texture::SwizzleSource swizzle_z,
+                                  Tegra::Texture::SwizzleSource swizzle_w) {
+    const GLenum new_x = MaxwellToGL::SwizzleSource(swizzle_x);
+    const GLenum new_y = MaxwellToGL::SwizzleSource(swizzle_y);
+    const GLenum new_z = MaxwellToGL::SwizzleSource(swizzle_z);
+    const GLenum new_w = MaxwellToGL::SwizzleSource(swizzle_w);
+    if (swizzle[0] == new_x && swizzle[1] == new_y && swizzle[2] == new_z && swizzle[3] == new_w) {
+        return;
+    }
+    swizzle = {new_x, new_y, new_z, new_w};
+    const auto swizzle_data = reinterpret_cast<const GLint*>(swizzle.data());
+    glTextureParameteriv(texture.handle, GL_TEXTURE_SWIZZLE_RGBA, swizzle_data);
+    if (discrepant_view.handle != 0) {
+        glTextureParameteriv(discrepant_view.handle, GL_TEXTURE_SWIZZLE_RGBA, swizzle_data);
+    }
+}
+
 RasterizerCacheOpenGL::RasterizerCacheOpenGL(RasterizerOpenGL& rasterizer)
     : RasterizerCache{rasterizer} {
     read_framebuffer.Create();
@@ -946,53 +897,59 @@ Surface RasterizerCacheOpenGL::GetColorBufferSurface(std::size_t index, bool pre
     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];
+    if (!gpu.dirty_flags.color_buffer[index]) {
+        return current_color_buffers[index];
     }
-    gpu.dirty_flags.color_buffer &= ~(1u << static_cast<u32>(index));
+    gpu.dirty_flags.color_buffer.reset(index);
 
     ASSERT(index < Tegra::Engines::Maxwell3D::Regs::NumRenderTargets);
 
     if (index >= regs.rt_control.count) {
-        return last_color_buffers[index] = {};
+        return current_color_buffers[index] = {};
     }
 
     if (regs.rt[index].Address() == 0 || regs.rt[index].format == Tegra::RenderTargetFormat::NONE) {
-        return last_color_buffers[index] = {};
+        return current_color_buffers[index] = {};
     }
 
     const SurfaceParams color_params{SurfaceParams::CreateForFramebuffer(index)};
 
-    return last_color_buffers[index] = GetSurface(color_params, preserve_contents);
+    return current_color_buffers[index] = GetSurface(color_params, preserve_contents);
 }
 
 void RasterizerCacheOpenGL::LoadSurface(const Surface& surface) {
     surface->LoadGLBuffer();
     surface->UploadGLTexture(read_framebuffer.handle, draw_framebuffer.handle);
     surface->MarkAsModified(false, *this);
+    surface->MarkForReload(false);
 }
 
 Surface RasterizerCacheOpenGL::GetSurface(const SurfaceParams& params, bool preserve_contents) {
-    if (params.addr == 0 || params.height * params.width == 0) {
+    if (params.gpu_addr == 0 || params.height * params.width == 0) {
         return {};
     }
 
     // Look up surface in the cache based on address
-    Surface surface{TryGet(params.addr)};
+    Surface surface{TryGet(params.host_ptr)};
     if (surface) {
         if (surface->GetSurfaceParams().IsCompatibleSurface(params)) {
-            // Use the cached surface as-is
+            // Use the cached surface as-is unless it's not synced with memory
+            if (surface->MustReload())
+                LoadSurface(surface);
             return surface;
         } else if (preserve_contents) {
             // If surface parameters changed and we care about keeping the previous data, recreate
             // the surface from the old one
             Surface new_surface{RecreateSurface(surface, params)};
-            Unregister(surface);
+            UnregisterSurface(surface);
             Register(new_surface);
+            if (new_surface->IsUploaded()) {
+                RegisterReinterpretSurface(new_surface);
+            }
             return new_surface;
         } else {
             // Delete the old surface before creating a new one to prevent collisions.
-            Unregister(surface);
+            UnregisterSurface(surface);
         }
     }
 
@@ -1022,14 +979,16 @@ void RasterizerCacheOpenGL::FastLayeredCopySurface(const Surface& src_surface,
                                                    const Surface& dst_surface) {
     const auto& init_params{src_surface->GetSurfaceParams()};
     const auto& dst_params{dst_surface->GetSurfaceParams()};
-    VAddr address = init_params.addr;
-    const std::size_t layer_size = dst_params.LayerMemorySize();
+    auto& memory_manager{Core::System::GetInstance().GPU().MemoryManager()};
+    Tegra::GPUVAddr address{init_params.gpu_addr};
+    const std::size_t layer_size{dst_params.LayerMemorySize()};
     for (u32 layer = 0; layer < dst_params.depth; layer++) {
         for (u32 mipmap = 0; mipmap < dst_params.max_mip_level; mipmap++) {
-            const VAddr sub_address = address + dst_params.GetMipmapLevelOffset(mipmap);
-            const Surface& copy = TryGet(sub_address);
-            if (!copy)
+            const Tegra::GPUVAddr sub_address{address + dst_params.GetMipmapLevelOffset(mipmap)};
+            const Surface& copy{TryGet(memory_manager.GetPointer(sub_address))};
+            if (!copy) {
                 continue;
+            }
             const auto& src_params{copy->GetSurfaceParams()};
             const u32 width{std::min(src_params.width, dst_params.MipWidth(mipmap))};
             const u32 height{std::min(src_params.height, dst_params.MipHeight(mipmap))};
@@ -1041,26 +1000,161 @@ void RasterizerCacheOpenGL::FastLayeredCopySurface(const Surface& src_surface,
         }
         address += layer_size;
     }
+
+    dst_surface->MarkAsModified(true, *this);
+}
+
+static bool BlitSurface(const Surface& src_surface, const Surface& dst_surface,
+                        const Common::Rectangle<u32>& src_rect,
+                        const Common::Rectangle<u32>& dst_rect, GLuint read_fb_handle,
+                        GLuint draw_fb_handle, GLenum src_attachment = 0, GLenum dst_attachment = 0,
+                        std::size_t cubemap_face = 0) {
+
+    const auto& src_params{src_surface->GetSurfaceParams()};
+    const auto& dst_params{dst_surface->GetSurfaceParams()};
+
+    OpenGLState prev_state{OpenGLState::GetCurState()};
+    SCOPE_EXIT({ prev_state.Apply(); });
+
+    OpenGLState state;
+    state.draw.read_framebuffer = read_fb_handle;
+    state.draw.draw_framebuffer = draw_fb_handle;
+    state.Apply();
+
+    u32 buffers{};
+
+    if (src_params.type == SurfaceType::ColorTexture) {
+        switch (src_params.target) {
+        case SurfaceTarget::Texture2D:
+            glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_COLOR_ATTACHMENT0 + src_attachment,
+                                   GL_TEXTURE_2D, src_surface->Texture().handle, 0);
+            glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D,
+                                   0, 0);
+            break;
+        case SurfaceTarget::TextureCubemap:
+            glFramebufferTexture2D(
+                GL_READ_FRAMEBUFFER, GL_COLOR_ATTACHMENT0 + src_attachment,
+                static_cast<GLenum>(GL_TEXTURE_CUBE_MAP_POSITIVE_X + cubemap_face),
+                src_surface->Texture().handle, 0);
+            glFramebufferTexture2D(
+                GL_READ_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT,
+                static_cast<GLenum>(GL_TEXTURE_CUBE_MAP_POSITIVE_X + cubemap_face), 0, 0);
+            break;
+        case SurfaceTarget::Texture2DArray:
+            glFramebufferTextureLayer(GL_READ_FRAMEBUFFER, GL_COLOR_ATTACHMENT0 + src_attachment,
+                                      src_surface->Texture().handle, 0, 0);
+            glFramebufferTextureLayer(GL_READ_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, 0, 0, 0);
+            break;
+        case SurfaceTarget::Texture3D:
+            glFramebufferTexture3D(GL_READ_FRAMEBUFFER, GL_COLOR_ATTACHMENT0 + src_attachment,
+                                   SurfaceTargetToGL(src_params.target),
+                                   src_surface->Texture().handle, 0, 0);
+            glFramebufferTexture3D(GL_READ_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT,
+                                   SurfaceTargetToGL(src_params.target), 0, 0, 0);
+            break;
+        default:
+            glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_COLOR_ATTACHMENT0 + src_attachment,
+                                   GL_TEXTURE_2D, src_surface->Texture().handle, 0);
+            glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D,
+                                   0, 0);
+            break;
+        }
+
+        switch (dst_params.target) {
+        case SurfaceTarget::Texture2D:
+            glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0 + dst_attachment,
+                                   GL_TEXTURE_2D, dst_surface->Texture().handle, 0);
+            glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D,
+                                   0, 0);
+            break;
+        case SurfaceTarget::TextureCubemap:
+            glFramebufferTexture2D(
+                GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0 + dst_attachment,
+                static_cast<GLenum>(GL_TEXTURE_CUBE_MAP_POSITIVE_X + cubemap_face),
+                dst_surface->Texture().handle, 0);
+            glFramebufferTexture2D(
+                GL_DRAW_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT,
+                static_cast<GLenum>(GL_TEXTURE_CUBE_MAP_POSITIVE_X + cubemap_face), 0, 0);
+            break;
+        case SurfaceTarget::Texture2DArray:
+            glFramebufferTextureLayer(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0 + dst_attachment,
+                                      dst_surface->Texture().handle, 0, 0);
+            glFramebufferTextureLayer(GL_DRAW_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, 0, 0, 0);
+            break;
+
+        case SurfaceTarget::Texture3D:
+            glFramebufferTexture3D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0 + dst_attachment,
+                                   SurfaceTargetToGL(dst_params.target),
+                                   dst_surface->Texture().handle, 0, 0);
+            glFramebufferTexture3D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT,
+                                   SurfaceTargetToGL(dst_params.target), 0, 0, 0);
+            break;
+        default:
+            glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0 + dst_attachment,
+                                   GL_TEXTURE_2D, dst_surface->Texture().handle, 0);
+            glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D,
+                                   0, 0);
+            break;
+        }
+
+        buffers = GL_COLOR_BUFFER_BIT;
+    } else if (src_params.type == SurfaceType::Depth) {
+        glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_COLOR_ATTACHMENT0 + src_attachment,
+                               GL_TEXTURE_2D, 0, 0);
+        glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, GL_TEXTURE_2D,
+                               src_surface->Texture().handle, 0);
+        glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_STENCIL_ATTACHMENT, GL_TEXTURE_2D, 0, 0);
+
+        glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0 + dst_attachment,
+                               GL_TEXTURE_2D, 0, 0);
+        glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, GL_TEXTURE_2D,
+                               dst_surface->Texture().handle, 0);
+        glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_STENCIL_ATTACHMENT, GL_TEXTURE_2D, 0, 0);
+
+        buffers = GL_DEPTH_BUFFER_BIT;
+    } else if (src_params.type == SurfaceType::DepthStencil) {
+        glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_COLOR_ATTACHMENT0 + src_attachment,
+                               GL_TEXTURE_2D, 0, 0);
+        glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D,
+                               src_surface->Texture().handle, 0);
+
+        glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0 + dst_attachment,
+                               GL_TEXTURE_2D, 0, 0);
+        glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D,
+                               dst_surface->Texture().handle, 0);
+
+        buffers = GL_DEPTH_BUFFER_BIT | GL_STENCIL_BUFFER_BIT;
+    }
+
+    glBlitFramebuffer(src_rect.left, src_rect.top, src_rect.right, src_rect.bottom, dst_rect.left,
+                      dst_rect.top, dst_rect.right, dst_rect.bottom, buffers,
+                      buffers == GL_COLOR_BUFFER_BIT ? GL_LINEAR : GL_NEAREST);
+
+    return true;
 }
 
 void RasterizerCacheOpenGL::FermiCopySurface(
     const Tegra::Engines::Fermi2D::Regs::Surface& src_config,
-    const Tegra::Engines::Fermi2D::Regs::Surface& dst_config) {
+    const Tegra::Engines::Fermi2D::Regs::Surface& dst_config,
+    const Common::Rectangle<u32>& src_rect, const Common::Rectangle<u32>& dst_rect) {
 
     const auto& src_params = SurfaceParams::CreateForFermiCopySurface(src_config);
     const auto& dst_params = SurfaceParams::CreateForFermiCopySurface(dst_config);
 
-    ASSERT(src_params.width == dst_params.width);
-    ASSERT(src_params.height == dst_params.height);
     ASSERT(src_params.pixel_format == dst_params.pixel_format);
     ASSERT(src_params.block_height == dst_params.block_height);
     ASSERT(src_params.is_tiled == dst_params.is_tiled);
     ASSERT(src_params.depth == dst_params.depth);
-    ASSERT(src_params.depth == 1); // Currently, FastCopySurface only works with 2D surfaces
     ASSERT(src_params.target == dst_params.target);
     ASSERT(src_params.rt.index == dst_params.rt.index);
 
-    FastCopySurface(GetSurface(src_params, true), GetSurface(dst_params, false));
+    auto src_surface = GetSurface(src_params, true);
+    auto dst_surface = GetSurface(dst_params, true);
+
+    BlitSurface(src_surface, dst_surface, src_rect, dst_rect, read_framebuffer.handle,
+                draw_framebuffer.handle);
+
+    dst_surface->MarkAsModified(true, *this);
 }
 
 void RasterizerCacheOpenGL::AccurateCopySurface(const Surface& src_surface,
@@ -1069,7 +1163,8 @@ void RasterizerCacheOpenGL::AccurateCopySurface(const Surface& src_surface,
     const auto& dst_params{dst_surface->GetSurfaceParams()};
 
     // Flush enough memory for both the source and destination surface
-    FlushRegion(src_params.addr, std::max(src_params.MemorySize(), dst_params.MemorySize()));
+    FlushRegion(ToCacheAddr(src_params.host_ptr),
+                std::max(src_params.MemorySize(), dst_params.MemorySize()));
 
     LoadSurface(dst_surface);
 }
@@ -1106,7 +1201,11 @@ Surface RasterizerCacheOpenGL::RecreateSurface(const Surface& old_surface,
     case SurfaceTarget::TextureCubemap:
     case SurfaceTarget::Texture2DArray:
     case SurfaceTarget::TextureCubeArray:
-        FastLayeredCopySurface(old_surface, new_surface);
+        if (old_params.pixel_format == new_params.pixel_format)
+            FastLayeredCopySurface(old_surface, new_surface);
+        else {
+            AccurateCopySurface(old_surface, new_surface);
+        }
         break;
     default:
         LOG_CRITICAL(Render_OpenGL, "Unimplemented surface target={}",
@@ -1117,8 +1216,8 @@ Surface RasterizerCacheOpenGL::RecreateSurface(const Surface& old_surface,
     return new_surface;
 }
 
-Surface RasterizerCacheOpenGL::TryFindFramebufferSurface(VAddr addr) const {
-    return TryGet(addr);
+Surface RasterizerCacheOpenGL::TryFindFramebufferSurface(const u8* host_ptr) const {
+    return TryGet(host_ptr);
 }
 
 void RasterizerCacheOpenGL::ReserveSurface(const Surface& surface) {
@@ -1135,4 +1234,109 @@ Surface RasterizerCacheOpenGL::TryGetReservedSurface(const SurfaceParams& params
     return {};
 }
 
+static std::optional<u32> TryFindBestMipMap(std::size_t memory, const SurfaceParams params,
+                                            u32 height) {
+    for (u32 i = 0; i < params.max_mip_level; i++) {
+        if (memory == params.GetMipmapSingleSize(i) && params.MipHeight(i) == height) {
+            return {i};
+        }
+    }
+    return {};
+}
+
+static std::optional<u32> TryFindBestLayer(Tegra::GPUVAddr addr, const SurfaceParams params,
+                                           u32 mipmap) {
+    const std::size_t size{params.LayerMemorySize()};
+    Tegra::GPUVAddr start{params.gpu_addr + params.GetMipmapLevelOffset(mipmap)};
+    for (u32 i = 0; i < params.depth; i++) {
+        if (start == addr) {
+            return {i};
+        }
+        start += size;
+    }
+    return {};
+}
+
+static bool LayerFitReinterpretSurface(RasterizerCacheOpenGL& cache, const Surface render_surface,
+                                       const Surface blitted_surface) {
+    const auto& dst_params = blitted_surface->GetSurfaceParams();
+    const auto& src_params = render_surface->GetSurfaceParams();
+    const std::size_t src_memory_size = src_params.size_in_bytes;
+    const std::optional<u32> level =
+        TryFindBestMipMap(src_memory_size, dst_params, src_params.height);
+    if (level.has_value()) {
+        if (src_params.width == dst_params.MipWidthGobAligned(*level) &&
+            src_params.height == dst_params.MipHeight(*level) &&
+            src_params.block_height >= dst_params.MipBlockHeight(*level)) {
+            const std::optional<u32> slot =
+                TryFindBestLayer(render_surface->GetSurfaceParams().gpu_addr, dst_params, *level);
+            if (slot.has_value()) {
+                glCopyImageSubData(render_surface->Texture().handle,
+                                   SurfaceTargetToGL(src_params.target), 0, 0, 0, 0,
+                                   blitted_surface->Texture().handle,
+                                   SurfaceTargetToGL(dst_params.target), *level, 0, 0, *slot,
+                                   dst_params.MipWidth(*level), dst_params.MipHeight(*level), 1);
+                blitted_surface->MarkAsModified(true, cache);
+                return true;
+            }
+        }
+    }
+    return false;
+}
+
+static bool IsReinterpretInvalid(const Surface render_surface, const Surface blitted_surface) {
+    const VAddr bound1 = blitted_surface->GetCpuAddr() + blitted_surface->GetMemorySize();
+    const VAddr bound2 = render_surface->GetCpuAddr() + render_surface->GetMemorySize();
+    if (bound2 > bound1)
+        return true;
+    const auto& dst_params = blitted_surface->GetSurfaceParams();
+    const auto& src_params = render_surface->GetSurfaceParams();
+    return (dst_params.component_type != src_params.component_type);
+}
+
+static bool IsReinterpretInvalidSecond(const Surface render_surface,
+                                       const Surface blitted_surface) {
+    const auto& dst_params = blitted_surface->GetSurfaceParams();
+    const auto& src_params = render_surface->GetSurfaceParams();
+    return (dst_params.height > src_params.height && dst_params.width > src_params.width);
+}
+
+bool RasterizerCacheOpenGL::PartialReinterpretSurface(Surface triggering_surface,
+                                                      Surface intersect) {
+    if (IsReinterpretInvalid(triggering_surface, intersect)) {
+        UnregisterSurface(intersect);
+        return false;
+    }
+    if (!LayerFitReinterpretSurface(*this, triggering_surface, intersect)) {
+        if (IsReinterpretInvalidSecond(triggering_surface, intersect)) {
+            UnregisterSurface(intersect);
+            return false;
+        }
+        FlushObject(intersect);
+        FlushObject(triggering_surface);
+        intersect->MarkForReload(true);
+    }
+    return true;
+}
+
+void RasterizerCacheOpenGL::SignalPreDrawCall() {
+    if (texception && GLAD_GL_ARB_texture_barrier) {
+        glTextureBarrier();
+    }
+    texception = false;
+}
+
+void RasterizerCacheOpenGL::SignalPostDrawCall() {
+    for (u32 i = 0; i < Maxwell::NumRenderTargets; i++) {
+        if (current_color_buffers[i] != nullptr) {
+            Surface intersect =
+                CollideOnReinterpretedSurface(current_color_buffers[i]->GetCacheAddr());
+            if (intersect != nullptr) {
+                PartialReinterpretSurface(current_color_buffers[i], intersect);
+                texception = true;
+            }
+        }
+    }
+}
+
 } // namespace OpenGL
diff --git a/src/video_core/renderer_opengl/gl_rasterizer_cache.h b/src/video_core/renderer_opengl/gl_rasterizer_cache.h
index 8d7d6722c..9366f47f2 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer_cache.h
+++ b/src/video_core/renderer_opengl/gl_rasterizer_cache.h
@@ -8,6 +8,7 @@
 #include <map>
 #include <memory>
 #include <string>
+#include <unordered_set>
 #include <vector>
 
 #include "common/alignment.h"
@@ -27,15 +28,15 @@ namespace OpenGL {
 
 class CachedSurface;
 using Surface = std::shared_ptr<CachedSurface>;
-using SurfaceSurfaceRect_Tuple = std::tuple<Surface, Surface, MathUtil::Rectangle<u32>>;
+using SurfaceSurfaceRect_Tuple = std::tuple<Surface, Surface, Common::Rectangle<u32>>;
 
 using SurfaceTarget = VideoCore::Surface::SurfaceTarget;
 using SurfaceType = VideoCore::Surface::SurfaceType;
 using PixelFormat = VideoCore::Surface::PixelFormat;
 using ComponentType = VideoCore::Surface::ComponentType;
+using Maxwell = Tegra::Engines::Maxwell3D::Regs;
 
 struct SurfaceParams {
-
     enum class SurfaceClass {
         Uploaded,
         RenderTarget,
@@ -71,7 +72,7 @@ struct SurfaceParams {
     }
 
     /// Returns the rectangle corresponding to this surface
-    MathUtil::Rectangle<u32> GetRect(u32 mip_level = 0) const;
+    Common::Rectangle<u32> GetRect(u32 mip_level = 0) const;
 
     /// Returns the total size of this surface in bytes, adjusted for compression
     std::size_t SizeInBytesRaw(bool ignore_tiled = false) const {
@@ -140,10 +141,18 @@ struct SurfaceParams {
         return offset;
     }
 
+    std::size_t GetMipmapSingleSize(u32 mip_level) const {
+        return InnerMipmapMemorySize(mip_level, false, is_layered);
+    }
+
     u32 MipWidth(u32 mip_level) const {
         return std::max(1U, width >> mip_level);
     }
 
+    u32 MipWidthGobAligned(u32 mip_level) const {
+        return Common::AlignUp(std::max(1U, width >> mip_level), 64U * 8U / GetFormatBpp());
+    }
+
     u32 MipHeight(u32 mip_level) const {
         return std::max(1U, height >> mip_level);
     }
@@ -168,20 +177,27 @@ struct SurfaceParams {
     }
 
     u32 MipBlockDepth(u32 mip_level) const {
-        if (mip_level == 0)
+        if (mip_level == 0) {
             return block_depth;
-        if (is_layered)
+        }
+
+        if (is_layered) {
             return 1;
-        u32 depth = MipDepth(mip_level);
+        }
+
+        const u32 mip_depth = MipDepth(mip_level);
         u32 bd = 32;
-        while (bd > 1 && depth * 2 <= bd) {
+        while (bd > 1 && mip_depth * 2 <= bd) {
             bd >>= 1;
         }
+
         if (bd == 32) {
-            u32 bh = MipBlockHeight(mip_level);
-            if (bh >= 4)
+            const u32 bh = MipBlockHeight(mip_level);
+            if (bh >= 4) {
                 return 16;
+            }
         }
+
         return bd;
     }
 
@@ -272,6 +288,7 @@ struct SurfaceParams {
     u32 height;
     u32 depth;
     u32 unaligned_height;
+    u32 pitch;
     SurfaceTarget target;
     SurfaceClass identity;
     u32 max_mip_level;
@@ -279,7 +296,7 @@ struct SurfaceParams {
     bool is_array;
     bool srgb_conversion;
     // Parameters used for caching
-    VAddr addr;
+    u8* host_ptr;
     Tegra::GPUVAddr gpu_addr;
     std::size_t size_in_bytes;
     std::size_t size_in_bytes_gl;
@@ -328,16 +345,20 @@ class RasterizerOpenGL;
 
 class CachedSurface final : public RasterizerCacheObject {
 public:
-    CachedSurface(const SurfaceParams& params);
+    explicit CachedSurface(const SurfaceParams& params);
 
-    VAddr GetAddr() const override {
-        return params.addr;
+    VAddr GetCpuAddr() const override {
+        return cpu_addr;
     }
 
     std::size_t GetSizeInBytes() const override {
         return cached_size_in_bytes;
     }
 
+    std::size_t GetMemorySize() const {
+        return memory_size;
+    }
+
     void Flush() override {
         FlushGLBuffer();
     }
@@ -346,31 +367,19 @@ public:
         return texture;
     }
 
-    const OGLTexture& TextureLayer() {
-        if (params.is_array) {
-            return Texture();
+    const OGLTexture& Texture(bool as_array) {
+        if (params.is_array == as_array) {
+            return texture;
+        } else {
+            EnsureTextureDiscrepantView();
+            return discrepant_view;
         }
-        EnsureTextureView();
-        return texture_view;
     }
 
     GLenum Target() const {
         return gl_target;
     }
 
-    GLenum TargetLayer() const {
-        using VideoCore::Surface::SurfaceTarget;
-        switch (params.target) {
-        case SurfaceTarget::Texture1D:
-            return GL_TEXTURE_1D_ARRAY;
-        case SurfaceTarget::Texture2D:
-            return GL_TEXTURE_2D_ARRAY;
-        case SurfaceTarget::TextureCubemap:
-            return GL_TEXTURE_CUBE_MAP_ARRAY;
-        }
-        return Target();
-    }
-
     const SurfaceParams& GetSurfaceParams() const {
         return params;
     }
@@ -382,19 +391,48 @@ public:
     // Upload data in gl_buffer to this surface's texture
     void UploadGLTexture(GLuint read_fb_handle, GLuint draw_fb_handle);
 
+    void UpdateSwizzle(Tegra::Texture::SwizzleSource swizzle_x,
+                       Tegra::Texture::SwizzleSource swizzle_y,
+                       Tegra::Texture::SwizzleSource swizzle_z,
+                       Tegra::Texture::SwizzleSource swizzle_w);
+
+    void MarkReinterpreted() {
+        reinterpreted = true;
+    }
+
+    bool IsReinterpreted() const {
+        return reinterpreted;
+    }
+
+    void MarkForReload(bool reload) {
+        must_reload = reload;
+    }
+
+    bool MustReload() const {
+        return must_reload;
+    }
+
+    bool IsUploaded() const {
+        return params.identity == SurfaceParams::SurfaceClass::Uploaded;
+    }
+
 private:
     void UploadGLMipmapTexture(u32 mip_map, GLuint read_fb_handle, GLuint draw_fb_handle);
 
-    void EnsureTextureView();
+    void EnsureTextureDiscrepantView();
 
     OGLTexture texture;
-    OGLTexture texture_view;
+    OGLTexture discrepant_view;
     std::vector<std::vector<u8>> gl_buffer;
     SurfaceParams params{};
     GLenum gl_target{};
     GLenum gl_internal_format{};
-    bool gl_is_compressed{};
     std::size_t cached_size_in_bytes{};
+    std::array<GLenum, 4> swizzle{GL_RED, GL_GREEN, GL_BLUE, GL_ALPHA};
+    std::size_t memory_size;
+    bool reinterpreted = false;
+    bool must_reload = false;
+    VAddr cpu_addr{};
 };
 
 class RasterizerCacheOpenGL final : public RasterizerCache<Surface> {
@@ -412,11 +450,16 @@ public:
     Surface GetColorBufferSurface(std::size_t index, bool preserve_contents);
 
     /// Tries to find a framebuffer using on the provided CPU address
-    Surface TryFindFramebufferSurface(VAddr addr) const;
+    Surface TryFindFramebufferSurface(const u8* host_ptr) const;
 
     /// Copies the contents of one surface to another
     void FermiCopySurface(const Tegra::Engines::Fermi2D::Regs::Surface& src_config,
-                          const Tegra::Engines::Fermi2D::Regs::Surface& dst_config);
+                          const Tegra::Engines::Fermi2D::Regs::Surface& dst_config,
+                          const Common::Rectangle<u32>& src_rect,
+                          const Common::Rectangle<u32>& dst_rect);
+
+    void SignalPreDrawCall();
+    void SignalPostDrawCall();
 
 private:
     void LoadSurface(const Surface& surface);
@@ -434,9 +477,17 @@ private:
     /// Tries to get a reserved surface for the specified parameters
     Surface TryGetReservedSurface(const SurfaceParams& params);
 
+    // Partialy reinterpret a surface based on a triggering_surface that collides with it.
+    // returns true if the reinterpret was successful, false in case it was not.
+    bool PartialReinterpretSurface(Surface triggering_surface, Surface intersect);
+
     /// Performs a slow but accurate surface copy, flushing to RAM and reinterpreting the data
     void AccurateCopySurface(const Surface& src_surface, const Surface& dst_surface);
     void FastLayeredCopySurface(const Surface& src_surface, const Surface& dst_surface);
+    void FastCopySurface(const Surface& src_surface, const Surface& dst_surface);
+    void CopySurface(const Surface& src_surface, const Surface& dst_surface,
+                     const GLuint copy_pbo_handle, const GLenum src_attachment = 0,
+                     const GLenum dst_attachment = 0, const std::size_t cubemap_face = 0);
 
     /// The surface reserve is a "backup" cache, this is where we put unique surfaces that have
     /// previously been used. This is to prevent surfaces from being constantly created and
@@ -446,12 +497,50 @@ private:
     OGLFramebuffer read_framebuffer;
     OGLFramebuffer draw_framebuffer;
 
+    bool texception = false;
+
     /// 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;
+    std::array<Surface, Maxwell::NumRenderTargets> last_color_buffers;
+    std::array<Surface, Maxwell::NumRenderTargets> current_color_buffers;
     Surface last_depth_buffer;
+
+    using SurfaceIntervalCache = boost::icl::interval_map<CacheAddr, Surface>;
+    using SurfaceInterval = typename SurfaceIntervalCache::interval_type;
+
+    static auto GetReinterpretInterval(const Surface& object) {
+        return SurfaceInterval::right_open(object->GetCacheAddr() + 1,
+                                           object->GetCacheAddr() + object->GetMemorySize() - 1);
+    }
+
+    // Reinterpreted surfaces are very fragil as the game may keep rendering into them.
+    SurfaceIntervalCache reinterpreted_surfaces;
+
+    void RegisterReinterpretSurface(Surface reinterpret_surface) {
+        auto interval = GetReinterpretInterval(reinterpret_surface);
+        reinterpreted_surfaces.insert({interval, reinterpret_surface});
+        reinterpret_surface->MarkReinterpreted();
+    }
+
+    Surface CollideOnReinterpretedSurface(CacheAddr addr) const {
+        const SurfaceInterval interval{addr};
+        for (auto& pair :
+             boost::make_iterator_range(reinterpreted_surfaces.equal_range(interval))) {
+            return pair.second;
+        }
+        return nullptr;
+    }
+
+    /// Unregisters an object from the cache
+    void UnregisterSurface(const Surface& object) {
+        if (object->IsReinterpreted()) {
+            auto interval = GetReinterpretInterval(object);
+            reinterpreted_surfaces.erase(interval);
+        }
+        Unregister(object);
+    }
 };
 
 } // namespace OpenGL
diff --git a/src/video_core/renderer_opengl/gl_resource_manager.cpp b/src/video_core/renderer_opengl/gl_resource_manager.cpp
index 1da744158..bfe666a73 100644
--- a/src/video_core/renderer_opengl/gl_resource_manager.cpp
+++ b/src/video_core/renderer_opengl/gl_resource_manager.cpp
@@ -15,12 +15,12 @@ MICROPROFILE_DEFINE(OpenGL_ResourceDeletion, "OpenGL", "Resource Deletion", MP_R
 
 namespace OpenGL {
 
-void OGLTexture::Create() {
+void OGLTexture::Create(GLenum target) {
     if (handle != 0)
         return;
 
     MICROPROFILE_SCOPE(OpenGL_ResourceCreation);
-    glGenTextures(1, &handle);
+    glCreateTextures(target, 1, &handle);
 }
 
 void OGLTexture::Release() {
@@ -71,7 +71,8 @@ void OGLShader::Release() {
 }
 
 void OGLProgram::CreateFromSource(const char* vert_shader, const char* geo_shader,
-                                  const char* frag_shader, bool separable_program) {
+                                  const char* frag_shader, bool separable_program,
+                                  bool hint_retrievable) {
     OGLShader vert, geo, frag;
     if (vert_shader)
         vert.Create(vert_shader, GL_VERTEX_SHADER);
@@ -81,7 +82,7 @@ void OGLProgram::CreateFromSource(const char* vert_shader, const char* geo_shade
         frag.Create(frag_shader, GL_FRAGMENT_SHADER);
 
     MICROPROFILE_SCOPE(OpenGL_ResourceCreation);
-    Create(separable_program, vert.handle, geo.handle, frag.handle);
+    Create(separable_program, hint_retrievable, vert.handle, geo.handle, frag.handle);
 }
 
 void OGLProgram::Release() {
diff --git a/src/video_core/renderer_opengl/gl_resource_manager.h b/src/video_core/renderer_opengl/gl_resource_manager.h
index e33f1e973..fbb93ee49 100644
--- a/src/video_core/renderer_opengl/gl_resource_manager.h
+++ b/src/video_core/renderer_opengl/gl_resource_manager.h
@@ -28,7 +28,7 @@ public:
     }
 
     /// Creates a new internal OpenGL resource and stores the handle
-    void Create();
+    void Create(GLenum target);
 
     /// Deletes the internal OpenGL resource
     void Release();
@@ -101,15 +101,15 @@ public:
     }
 
     template <typename... T>
-    void Create(bool separable_program, T... shaders) {
+    void Create(bool separable_program, bool hint_retrievable, T... shaders) {
         if (handle != 0)
             return;
-        handle = GLShader::LoadProgram(separable_program, shaders...);
+        handle = GLShader::LoadProgram(separable_program, hint_retrievable, shaders...);
     }
 
     /// Creates a new internal OpenGL resource and stores the handle
     void CreateFromSource(const char* vert_shader, const char* geo_shader, const char* frag_shader,
-                          bool separable_program = false);
+                          bool separable_program = false, bool hint_retrievable = false);
 
     /// Deletes the internal OpenGL resource
     void Release();
diff --git a/src/video_core/renderer_opengl/gl_shader_cache.cpp b/src/video_core/renderer_opengl/gl_shader_cache.cpp
index 90eda7814..1ed740877 100644
--- a/src/video_core/renderer_opengl/gl_shader_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_cache.cpp
@@ -11,6 +11,7 @@
 #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_disk_cache.h"
 #include "video_core/renderer_opengl/gl_shader_manager.h"
 #include "video_core/renderer_opengl/utils.h"
 #include "video_core/shader/shader_ir.h"
@@ -19,18 +20,28 @@ namespace OpenGL {
 
 using VideoCommon::Shader::ProgramCode;
 
+// One UBO is always reserved for emulation values
+constexpr u32 RESERVED_UBOS = 1;
+
+struct UnspecializedShader {
+    std::string code;
+    GLShader::ShaderEntries entries;
+    Maxwell::ShaderProgram program_type;
+};
+
+namespace {
+
 /// Gets the address for the specified shader stage program
-static VAddr GetShaderAddress(Maxwell::ShaderProgram program) {
-    const auto& gpu = Core::System::GetInstance().GPU().Maxwell3D();
-    const auto& shader_config = gpu.regs.shader_config[static_cast<std::size_t>(program)];
-    return *gpu.memory_manager.GpuToCpuAddress(gpu.regs.code_address.CodeAddress() +
-                                               shader_config.offset);
+Tegra::GPUVAddr GetShaderAddress(Maxwell::ShaderProgram program) {
+    const auto& gpu{Core::System::GetInstance().GPU().Maxwell3D()};
+    const auto& shader_config{gpu.regs.shader_config[static_cast<std::size_t>(program)]};
+    return gpu.regs.code_address.CodeAddress() + shader_config.offset;
 }
 
 /// Gets the shader program code from memory for the specified address
-static ProgramCode GetShaderCode(VAddr addr) {
+ProgramCode GetShaderCode(const u8* host_ptr) {
     ProgramCode program_code(VideoCommon::Shader::MAX_PROGRAM_LENGTH);
-    Memory::ReadBlock(addr, program_code.data(), program_code.size() * sizeof(u64));
+    std::memcpy(program_code.data(), host_ptr, program_code.size() * sizeof(u64));
     return program_code;
 }
 
@@ -49,38 +60,198 @@ constexpr GLenum GetShaderType(Maxwell::ShaderProgram program_type) {
     }
 }
 
-CachedShader::CachedShader(VAddr addr, Maxwell::ShaderProgram program_type)
-    : addr{addr}, program_type{program_type}, setup{GetShaderCode(addr)} {
+/// Gets if the current instruction offset is a scheduler instruction
+constexpr bool IsSchedInstruction(std::size_t offset, std::size_t main_offset) {
+    // Sched instructions appear once every 4 instructions.
+    constexpr std::size_t SchedPeriod = 4;
+    const std::size_t absolute_offset = offset - main_offset;
+    return (absolute_offset % SchedPeriod) == 0;
+}
 
-    GLShader::ProgramResult program_result;
+/// Describes primitive behavior on geometry shaders
+constexpr std::tuple<const char*, const char*, u32> GetPrimitiveDescription(GLenum primitive_mode) {
+    switch (primitive_mode) {
+    case GL_POINTS:
+        return {"points", "Points", 1};
+    case GL_LINES:
+    case GL_LINE_STRIP:
+        return {"lines", "Lines", 2};
+    case GL_LINES_ADJACENCY:
+    case GL_LINE_STRIP_ADJACENCY:
+        return {"lines_adjacency", "LinesAdj", 4};
+    case GL_TRIANGLES:
+    case GL_TRIANGLE_STRIP:
+    case GL_TRIANGLE_FAN:
+        return {"triangles", "Triangles", 3};
+    case GL_TRIANGLES_ADJACENCY:
+    case GL_TRIANGLE_STRIP_ADJACENCY:
+        return {"triangles_adjacency", "TrianglesAdj", 6};
+    default:
+        return {"points", "Invalid", 1};
+    }
+}
 
-    switch (program_type) {
-    case Maxwell::ShaderProgram::VertexA:
+/// Calculates the size of a program stream
+std::size_t CalculateProgramSize(const GLShader::ProgramCode& program) {
+    constexpr std::size_t start_offset = 10;
+    std::size_t offset = start_offset;
+    std::size_t size = start_offset * sizeof(u64);
+    while (offset < program.size()) {
+        const u64 instruction = program[offset];
+        if (!IsSchedInstruction(offset, start_offset)) {
+            if (instruction == 0 || (instruction >> 52) == 0x50b) {
+                // End on Maxwell's "nop" instruction
+                break;
+            }
+        }
+        size += sizeof(u64);
+        offset++;
+    }
+    // The last instruction is included in the program size
+    return std::min(size + sizeof(u64), program.size() * sizeof(u64));
+}
+
+/// Hashes one (or two) program streams
+u64 GetUniqueIdentifier(Maxwell::ShaderProgram program_type, const ProgramCode& code,
+                        const ProgramCode& code_b) {
+    u64 unique_identifier =
+        Common::CityHash64(reinterpret_cast<const char*>(code.data()), CalculateProgramSize(code));
+    if (program_type != Maxwell::ShaderProgram::VertexA) {
+        return unique_identifier;
+    }
+    // VertexA programs include two programs
+
+    std::size_t seed = 0;
+    boost::hash_combine(seed, unique_identifier);
+
+    const u64 identifier_b = Common::CityHash64(reinterpret_cast<const char*>(code_b.data()),
+                                                CalculateProgramSize(code_b));
+    boost::hash_combine(seed, identifier_b);
+    return static_cast<u64>(seed);
+}
+
+/// Creates an unspecialized program from code streams
+GLShader::ProgramResult CreateProgram(Maxwell::ShaderProgram program_type, ProgramCode program_code,
+                                      ProgramCode program_code_b) {
+    GLShader::ShaderSetup setup(program_code);
+    if (program_type == Maxwell::ShaderProgram::VertexA) {
         // VertexB is always enabled, so when VertexA is enabled, we have two vertex shaders.
         // Conventional HW does not support this, so we combine VertexA and VertexB into one
         // stage here.
-        setup.SetProgramB(GetShaderCode(GetShaderAddress(Maxwell::ShaderProgram::VertexB)));
+        setup.SetProgramB(program_code_b);
+    }
+    setup.program.unique_identifier =
+        GetUniqueIdentifier(program_type, program_code, program_code_b);
+
+    switch (program_type) {
+    case Maxwell::ShaderProgram::VertexA:
     case Maxwell::ShaderProgram::VertexB:
-        CalculateProperties();
-        program_result = GLShader::GenerateVertexShader(setup);
-        break;
+        return GLShader::GenerateVertexShader(setup);
     case Maxwell::ShaderProgram::Geometry:
-        CalculateProperties();
-        program_result = GLShader::GenerateGeometryShader(setup);
-        break;
+        return GLShader::GenerateGeometryShader(setup);
     case Maxwell::ShaderProgram::Fragment:
-        CalculateProperties();
-        program_result = GLShader::GenerateFragmentShader(setup);
-        break;
+        return GLShader::GenerateFragmentShader(setup);
     default:
         LOG_CRITICAL(HW_GPU, "Unimplemented program_type={}", static_cast<u32>(program_type));
         UNREACHABLE();
+        return {};
+    }
+}
+
+CachedProgram SpecializeShader(const std::string& code, const GLShader::ShaderEntries& entries,
+                               Maxwell::ShaderProgram program_type, BaseBindings base_bindings,
+                               GLenum primitive_mode, bool hint_retrievable = false) {
+    std::string source = "#version 430 core\n";
+    source += fmt::format("#define EMULATION_UBO_BINDING {}\n", base_bindings.cbuf++);
+
+    for (const auto& cbuf : entries.const_buffers) {
+        source +=
+            fmt::format("#define CBUF_BINDING_{} {}\n", cbuf.GetIndex(), base_bindings.cbuf++);
+    }
+    for (const auto& gmem : entries.global_memory_entries) {
+        source += fmt::format("#define GMEM_BINDING_{}_{} {}\n", gmem.GetCbufIndex(),
+                              gmem.GetCbufOffset(), base_bindings.gmem++);
+    }
+    for (const auto& sampler : entries.samplers) {
+        source += fmt::format("#define SAMPLER_BINDING_{} {}\n", sampler.GetIndex(),
+                              base_bindings.sampler++);
+    }
+
+    if (program_type == Maxwell::ShaderProgram::Geometry) {
+        const auto [glsl_topology, debug_name, max_vertices] =
+            GetPrimitiveDescription(primitive_mode);
+
+        source += "layout (" + std::string(glsl_topology) + ") in;\n";
+        source += "#define MAX_VERTEX_INPUT " + std::to_string(max_vertices) + '\n';
+    }
+
+    source += code;
+
+    OGLShader shader;
+    shader.Create(source.c_str(), GetShaderType(program_type));
+
+    auto program = std::make_shared<OGLProgram>();
+    program->Create(true, hint_retrievable, shader.handle);
+    return program;
+}
+
+std::set<GLenum> GetSupportedFormats() {
+    std::set<GLenum> supported_formats;
+
+    GLint num_formats{};
+    glGetIntegerv(GL_NUM_PROGRAM_BINARY_FORMATS, &num_formats);
+
+    std::vector<GLint> formats(num_formats);
+    glGetIntegerv(GL_PROGRAM_BINARY_FORMATS, formats.data());
+
+    for (const GLint format : formats)
+        supported_formats.insert(static_cast<GLenum>(format));
+    return supported_formats;
+}
+
+} // namespace
+
+CachedShader::CachedShader(VAddr cpu_addr, u64 unique_identifier,
+                           Maxwell::ShaderProgram program_type, ShaderDiskCacheOpenGL& disk_cache,
+                           const PrecompiledPrograms& precompiled_programs,
+                           ProgramCode&& program_code, ProgramCode&& program_code_b, u8* host_ptr)
+    : host_ptr{host_ptr}, cpu_addr{cpu_addr}, unique_identifier{unique_identifier},
+      program_type{program_type}, disk_cache{disk_cache},
+      precompiled_programs{precompiled_programs}, RasterizerCacheObject{host_ptr} {
+
+    const std::size_t code_size = CalculateProgramSize(program_code);
+    const std::size_t code_size_b =
+        program_code_b.empty() ? 0 : CalculateProgramSize(program_code_b);
+
+    GLShader::ProgramResult program_result =
+        CreateProgram(program_type, program_code, program_code_b);
+    if (program_result.first.empty()) {
+        // TODO(Rodrigo): Unimplemented shader stages hit here, avoid using these for now
         return;
     }
 
     code = program_result.first;
     entries = program_result.second;
     shader_length = entries.shader_length;
+
+    const ShaderDiskCacheRaw raw(unique_identifier, program_type,
+                                 static_cast<u32>(code_size / sizeof(u64)),
+                                 static_cast<u32>(code_size_b / sizeof(u64)),
+                                 std::move(program_code), std::move(program_code_b));
+    disk_cache.SaveRaw(raw);
+}
+
+CachedShader::CachedShader(VAddr cpu_addr, u64 unique_identifier,
+                           Maxwell::ShaderProgram program_type, ShaderDiskCacheOpenGL& disk_cache,
+                           const PrecompiledPrograms& precompiled_programs,
+                           GLShader::ProgramResult result, u8* host_ptr)
+    : cpu_addr{cpu_addr}, unique_identifier{unique_identifier}, program_type{program_type},
+      disk_cache{disk_cache}, precompiled_programs{precompiled_programs}, RasterizerCacheObject{
+                                                                              host_ptr} {
+
+    code = std::move(result.first);
+    entries = result.second;
+    shader_length = entries.shader_length;
 }
 
 std::tuple<GLuint, BaseBindings> CachedShader::GetProgramHandle(GLenum primitive_mode,
@@ -92,150 +263,255 @@ std::tuple<GLuint, BaseBindings> CachedShader::GetProgramHandle(GLenum primitive
         const auto [entry, is_cache_miss] = programs.try_emplace(base_bindings);
         auto& program = entry->second;
         if (is_cache_miss) {
-            std::string source = AllocateBindings(base_bindings);
-            source += code;
+            program = TryLoadProgram(primitive_mode, base_bindings);
+            if (!program) {
+                program =
+                    SpecializeShader(code, entries, program_type, base_bindings, primitive_mode);
+                disk_cache.SaveUsage(GetUsage(primitive_mode, base_bindings));
+            }
 
-            OGLShader shader;
-            shader.Create(source.c_str(), GetShaderType(program_type));
-            program.Create(true, shader.handle);
-            LabelGLObject(GL_PROGRAM, program.handle, addr);
+            LabelGLObject(GL_PROGRAM, program->handle, cpu_addr);
         }
 
-        handle = program.handle;
+        handle = program->handle;
     }
 
-    // Add const buffer and samplers offset reserved by this shader. One UBO binding is reserved for
-    // emulation values
-    base_bindings.cbuf += static_cast<u32>(entries.const_buffers.size()) + 1;
+    base_bindings.cbuf += static_cast<u32>(entries.const_buffers.size()) + RESERVED_UBOS;
     base_bindings.gmem += static_cast<u32>(entries.global_memory_entries.size());
     base_bindings.sampler += static_cast<u32>(entries.samplers.size());
 
     return {handle, base_bindings};
 }
 
-std::string CachedShader::AllocateBindings(BaseBindings base_bindings) {
-    std::string code = "#version 430 core\n";
-    code += fmt::format("#define EMULATION_UBO_BINDING {}\n", base_bindings.cbuf++);
-
-    for (const auto& cbuf : entries.const_buffers) {
-        code += fmt::format("#define CBUF_BINDING_{} {}\n", cbuf.GetIndex(), base_bindings.cbuf++);
-    }
-
-    for (const auto& gmem : entries.global_memory_entries) {
-        code += fmt::format("#define GMEM_BINDING_{}_{} {}\n", gmem.GetCbufIndex(),
-                            gmem.GetCbufOffset(), base_bindings.gmem++);
-    }
-
-    for (const auto& sampler : entries.samplers) {
-        code += fmt::format("#define SAMPLER_BINDING_{} {}\n", sampler.GetIndex(),
-                            base_bindings.sampler++);
-    }
-
-    return code;
-}
-
 GLuint CachedShader::GetGeometryShader(GLenum primitive_mode, BaseBindings base_bindings) {
     const auto [entry, is_cache_miss] = geometry_programs.try_emplace(base_bindings);
     auto& programs = entry->second;
 
     switch (primitive_mode) {
     case GL_POINTS:
-        return LazyGeometryProgram(programs.points, base_bindings, "points", 1, "ShaderPoints");
+        return LazyGeometryProgram(programs.points, base_bindings, primitive_mode);
     case GL_LINES:
     case GL_LINE_STRIP:
-        return LazyGeometryProgram(programs.lines, base_bindings, "lines", 2, "ShaderLines");
+        return LazyGeometryProgram(programs.lines, base_bindings, primitive_mode);
     case GL_LINES_ADJACENCY:
     case GL_LINE_STRIP_ADJACENCY:
-        return LazyGeometryProgram(programs.lines_adjacency, base_bindings, "lines_adjacency", 4,
-                                   "ShaderLinesAdjacency");
+        return LazyGeometryProgram(programs.lines_adjacency, base_bindings, primitive_mode);
     case GL_TRIANGLES:
     case GL_TRIANGLE_STRIP:
     case GL_TRIANGLE_FAN:
-        return LazyGeometryProgram(programs.triangles, base_bindings, "triangles", 3,
-                                   "ShaderTriangles");
+        return LazyGeometryProgram(programs.triangles, base_bindings, primitive_mode);
     case GL_TRIANGLES_ADJACENCY:
     case GL_TRIANGLE_STRIP_ADJACENCY:
-        return LazyGeometryProgram(programs.triangles_adjacency, base_bindings,
-                                   "triangles_adjacency", 6, "ShaderTrianglesAdjacency");
+        return LazyGeometryProgram(programs.triangles_adjacency, base_bindings, primitive_mode);
     default:
         UNREACHABLE_MSG("Unknown primitive mode.");
-        return LazyGeometryProgram(programs.points, base_bindings, "points", 1, "ShaderPoints");
+        return LazyGeometryProgram(programs.points, base_bindings, primitive_mode);
     }
 }
 
-GLuint CachedShader::LazyGeometryProgram(OGLProgram& target_program, BaseBindings base_bindings,
-                                         const std::string& glsl_topology, u32 max_vertices,
-                                         const std::string& debug_name) {
-    if (target_program.handle != 0) {
-        return target_program.handle;
+GLuint CachedShader::LazyGeometryProgram(CachedProgram& target_program, BaseBindings base_bindings,
+                                         GLenum primitive_mode) {
+    if (target_program) {
+        return target_program->handle;
+    }
+    const auto [glsl_name, debug_name, vertices] = GetPrimitiveDescription(primitive_mode);
+    target_program = TryLoadProgram(primitive_mode, base_bindings);
+    if (!target_program) {
+        target_program =
+            SpecializeShader(code, entries, program_type, base_bindings, primitive_mode);
+        disk_cache.SaveUsage(GetUsage(primitive_mode, base_bindings));
     }
-    std::string source = AllocateBindings(base_bindings);
-    source += "layout (" + glsl_topology + ") in;\n";
-    source += "#define MAX_VERTEX_INPUT " + std::to_string(max_vertices) + '\n';
-    source += code;
 
-    OGLShader shader;
-    shader.Create(source.c_str(), GL_GEOMETRY_SHADER);
-    target_program.Create(true, shader.handle);
-    LabelGLObject(GL_PROGRAM, target_program.handle, addr, debug_name);
-    return target_program.handle;
+    LabelGLObject(GL_PROGRAM, target_program->handle, cpu_addr, debug_name);
+
+    return target_program->handle;
 };
 
-static bool IsSchedInstruction(std::size_t offset, std::size_t main_offset) {
-    // sched instructions appear once every 4 instructions.
-    static constexpr std::size_t SchedPeriod = 4;
-    const std::size_t absolute_offset = offset - main_offset;
-    return (absolute_offset % SchedPeriod) == 0;
+CachedProgram CachedShader::TryLoadProgram(GLenum primitive_mode,
+                                           BaseBindings base_bindings) const {
+    const auto found = precompiled_programs.find(GetUsage(primitive_mode, base_bindings));
+    if (found == precompiled_programs.end()) {
+        return {};
+    }
+    return found->second;
 }
 
-static std::size_t CalculateProgramSize(const GLShader::ProgramCode& program) {
-    constexpr std::size_t start_offset = 10;
-    std::size_t offset = start_offset;
-    std::size_t size = start_offset * sizeof(u64);
-    while (offset < program.size()) {
-        const u64 inst = program[offset];
-        if (!IsSchedInstruction(offset, start_offset)) {
-            if (inst == 0 || (inst >> 52) == 0x50b) {
-                break;
+ShaderDiskCacheUsage CachedShader::GetUsage(GLenum primitive_mode,
+                                            BaseBindings base_bindings) const {
+    return {unique_identifier, base_bindings, primitive_mode};
+}
+
+ShaderCacheOpenGL::ShaderCacheOpenGL(RasterizerOpenGL& rasterizer, Core::System& system)
+    : RasterizerCache{rasterizer}, disk_cache{system} {}
+
+void ShaderCacheOpenGL::LoadDiskCache(const std::atomic_bool& stop_loading,
+                                      const VideoCore::DiskResourceLoadCallback& callback) {
+    const auto transferable = disk_cache.LoadTransferable();
+    if (!transferable) {
+        return;
+    }
+    const auto [raws, usages] = *transferable;
+
+    auto [decompiled, dumps] = disk_cache.LoadPrecompiled();
+
+    const auto supported_formats{GetSupportedFormats()};
+    const auto unspecialized{
+        GenerateUnspecializedShaders(stop_loading, callback, raws, decompiled)};
+    if (stop_loading)
+        return;
+
+    // Build shaders
+    if (callback)
+        callback(VideoCore::LoadCallbackStage::Build, 0, usages.size());
+    for (std::size_t i = 0; i < usages.size(); ++i) {
+        if (stop_loading)
+            return;
+
+        const auto& usage{usages[i]};
+        LOG_INFO(Render_OpenGL, "Building shader {:016x} ({} of {})", usage.unique_identifier,
+                 i + 1, usages.size());
+
+        const auto& unspec{unspecialized.at(usage.unique_identifier)};
+        const auto dump_it = dumps.find(usage);
+
+        CachedProgram shader;
+        if (dump_it != dumps.end()) {
+            // If the shader is dumped, attempt to load it with
+            shader = GeneratePrecompiledProgram(dump_it->second, supported_formats);
+            if (!shader) {
+                // Invalidate the precompiled cache if a shader dumped shader was rejected
+                disk_cache.InvalidatePrecompiled();
+                dumps.clear();
             }
         }
-        size += sizeof(inst);
-        offset++;
+        if (!shader) {
+            shader = SpecializeShader(unspec.code, unspec.entries, unspec.program_type,
+                                      usage.bindings, usage.primitive, true);
+        }
+        precompiled_programs.insert({usage, std::move(shader)});
+
+        if (callback)
+            callback(VideoCore::LoadCallbackStage::Build, i + 1, usages.size());
+    }
+
+    // TODO(Rodrigo): Do state tracking for transferable shaders and do a dummy draw before
+    // precompiling them
+
+    for (std::size_t i = 0; i < usages.size(); ++i) {
+        const auto& usage{usages[i]};
+        if (dumps.find(usage) == dumps.end()) {
+            const auto& program = precompiled_programs.at(usage);
+            disk_cache.SaveDump(usage, program->handle);
+        }
     }
-    return size;
 }
 
-void CachedShader::CalculateProperties() {
-    setup.program.real_size = CalculateProgramSize(setup.program.code);
-    setup.program.real_size_b = 0;
-    setup.program.unique_identifier = Common::CityHash64(
-        reinterpret_cast<const char*>(setup.program.code.data()), setup.program.real_size);
-    if (program_type == Maxwell::ShaderProgram::VertexA) {
-        std::size_t seed = 0;
-        boost::hash_combine(seed, setup.program.unique_identifier);
-        setup.program.real_size_b = CalculateProgramSize(setup.program.code_b);
-        const u64 identifier_b = Common::CityHash64(
-            reinterpret_cast<const char*>(setup.program.code_b.data()), setup.program.real_size_b);
-        boost::hash_combine(seed, identifier_b);
-        setup.program.unique_identifier = static_cast<u64>(seed);
+CachedProgram ShaderCacheOpenGL::GeneratePrecompiledProgram(
+    const ShaderDiskCacheDump& dump, const std::set<GLenum>& supported_formats) {
+
+    if (supported_formats.find(dump.binary_format) == supported_formats.end()) {
+        LOG_INFO(Render_OpenGL, "Precompiled cache entry with unsupported format - removing");
+        return {};
     }
+
+    CachedProgram shader = std::make_shared<OGLProgram>();
+    shader->handle = glCreateProgram();
+    glProgramParameteri(shader->handle, GL_PROGRAM_SEPARABLE, GL_TRUE);
+    glProgramBinary(shader->handle, dump.binary_format, dump.binary.data(),
+                    static_cast<GLsizei>(dump.binary.size()));
+
+    GLint link_status{};
+    glGetProgramiv(shader->handle, GL_LINK_STATUS, &link_status);
+    if (link_status == GL_FALSE) {
+        LOG_INFO(Render_OpenGL, "Precompiled cache rejected by the driver - removing");
+        return {};
+    }
+
+    return shader;
 }
 
-ShaderCacheOpenGL::ShaderCacheOpenGL(RasterizerOpenGL& rasterizer) : RasterizerCache{rasterizer} {}
+std::unordered_map<u64, UnspecializedShader> ShaderCacheOpenGL::GenerateUnspecializedShaders(
+    const std::atomic_bool& stop_loading, const VideoCore::DiskResourceLoadCallback& callback,
+    const std::vector<ShaderDiskCacheRaw>& raws,
+    const std::unordered_map<u64, ShaderDiskCacheDecompiled>& decompiled) {
+    std::unordered_map<u64, UnspecializedShader> unspecialized;
+
+    if (callback)
+        callback(VideoCore::LoadCallbackStage::Decompile, 0, raws.size());
+
+    for (std::size_t i = 0; i < raws.size(); ++i) {
+        if (stop_loading)
+            return {};
+
+        const auto& raw{raws[i]};
+        const u64 unique_identifier = raw.GetUniqueIdentifier();
+        const u64 calculated_hash =
+            GetUniqueIdentifier(raw.GetProgramType(), raw.GetProgramCode(), raw.GetProgramCodeB());
+        if (unique_identifier != calculated_hash) {
+            LOG_ERROR(
+                Render_OpenGL,
+                "Invalid hash in entry={:016x} (obtained hash={:016x}) - removing shader cache",
+                raw.GetUniqueIdentifier(), calculated_hash);
+            disk_cache.InvalidateTransferable();
+            return {};
+        }
+
+        GLShader::ProgramResult result;
+        if (const auto it = decompiled.find(unique_identifier); it != decompiled.end()) {
+            // If it's stored in the precompiled file, avoid decompiling it here
+            const auto& stored_decompiled{it->second};
+            result = {stored_decompiled.code, stored_decompiled.entries};
+        } else {
+            // Otherwise decompile the shader at boot and save the result to the decompiled file
+            result =
+                CreateProgram(raw.GetProgramType(), raw.GetProgramCode(), raw.GetProgramCodeB());
+            disk_cache.SaveDecompiled(unique_identifier, result.first, result.second);
+        }
+
+        precompiled_shaders.insert({unique_identifier, result});
+
+        unspecialized.insert(
+            {raw.GetUniqueIdentifier(),
+             {std::move(result.first), std::move(result.second), raw.GetProgramType()}});
+
+        if (callback)
+            callback(VideoCore::LoadCallbackStage::Decompile, i, raws.size());
+    }
+    return unspecialized;
+}
 
 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)};
+    auto& memory_manager{Core::System::GetInstance().GPU().MemoryManager()};
+    const Tegra::GPUVAddr program_addr{GetShaderAddress(program)};
 
     // Look up shader in the cache based on address
-    Shader shader{TryGet(program_addr)};
+    const auto& host_ptr{memory_manager.GetPointer(program_addr)};
+    Shader shader{TryGet(host_ptr)};
 
     if (!shader) {
         // No shader found - create a new one
-        shader = std::make_shared<CachedShader>(program_addr, program);
+        ProgramCode program_code{GetShaderCode(host_ptr)};
+        ProgramCode program_code_b;
+        if (program == Maxwell::ShaderProgram::VertexA) {
+            program_code_b = GetShaderCode(
+                memory_manager.GetPointer(GetShaderAddress(Maxwell::ShaderProgram::VertexB)));
+        }
+        const u64 unique_identifier = GetUniqueIdentifier(program, program_code, program_code_b);
+        const VAddr cpu_addr{*memory_manager.GpuToCpuAddress(program_addr)};
+        const auto found = precompiled_shaders.find(unique_identifier);
+        if (found != precompiled_shaders.end()) {
+            shader =
+                std::make_shared<CachedShader>(cpu_addr, unique_identifier, program, disk_cache,
+                                               precompiled_programs, found->second, host_ptr);
+        } else {
+            shader = std::make_shared<CachedShader>(
+                cpu_addr, unique_identifier, program, disk_cache, precompiled_programs,
+                std::move(program_code), std::move(program_code_b), host_ptr);
+        }
         Register(shader);
     }
 
diff --git a/src/video_core/renderer_opengl/gl_shader_cache.h b/src/video_core/renderer_opengl/gl_shader_cache.h
index 904d15dd0..fd1c85115 100644
--- a/src/video_core/renderer_opengl/gl_shader_cache.h
+++ b/src/video_core/renderer_opengl/gl_shader_cache.h
@@ -5,43 +5,52 @@
 #pragma once
 
 #include <array>
-#include <map>
 #include <memory>
+#include <set>
 #include <tuple>
+#include <unordered_map>
 
 #include <glad/glad.h>
 
 #include "common/assert.h"
 #include "common/common_types.h"
 #include "video_core/rasterizer_cache.h"
+#include "video_core/renderer_base.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_disk_cache.h"
 #include "video_core/renderer_opengl/gl_shader_gen.h"
 
+namespace Core {
+class System;
+} // namespace Core
+
 namespace OpenGL {
 
 class CachedShader;
 class RasterizerOpenGL;
+struct UnspecializedShader;
 
 using Shader = std::shared_ptr<CachedShader>;
+using CachedProgram = std::shared_ptr<OGLProgram>;
 using Maxwell = Tegra::Engines::Maxwell3D::Regs;
-
-struct BaseBindings {
-    u32 cbuf{};
-    u32 gmem{};
-    u32 sampler{};
-
-    bool operator<(const BaseBindings& rhs) const {
-        return std::tie(cbuf, gmem, sampler) < std::tie(rhs.cbuf, rhs.gmem, rhs.sampler);
-    }
-};
+using PrecompiledPrograms = std::unordered_map<ShaderDiskCacheUsage, CachedProgram>;
+using PrecompiledShaders = std::unordered_map<u64, GLShader::ProgramResult>;
 
 class CachedShader final : public RasterizerCacheObject {
 public:
-    CachedShader(VAddr addr, Maxwell::ShaderProgram program_type);
-
-    VAddr GetAddr() const override {
-        return addr;
+    explicit CachedShader(VAddr cpu_addr, u64 unique_identifier,
+                          Maxwell::ShaderProgram program_type, ShaderDiskCacheOpenGL& disk_cache,
+                          const PrecompiledPrograms& precompiled_programs,
+                          ProgramCode&& program_code, ProgramCode&& program_code_b, u8* host_ptr);
+
+    explicit CachedShader(VAddr cpu_addr, u64 unique_identifier,
+                          Maxwell::ShaderProgram program_type, ShaderDiskCacheOpenGL& disk_cache,
+                          const PrecompiledPrograms& precompiled_programs,
+                          GLShader::ProgramResult result, u8* host_ptr);
+
+    VAddr GetCpuAddr() const override {
+        return cpu_addr;
     }
 
     std::size_t GetSizeInBytes() const override {
@@ -65,49 +74,68 @@ private:
     // declared by the hardware. Workaround this issue by generating a different shader per input
     // topology class.
     struct GeometryPrograms {
-        OGLProgram points;
-        OGLProgram lines;
-        OGLProgram lines_adjacency;
-        OGLProgram triangles;
-        OGLProgram triangles_adjacency;
+        CachedProgram points;
+        CachedProgram lines;
+        CachedProgram lines_adjacency;
+        CachedProgram triangles;
+        CachedProgram triangles_adjacency;
     };
 
-    std::string AllocateBindings(BaseBindings base_bindings);
-
     GLuint GetGeometryShader(GLenum primitive_mode, BaseBindings base_bindings);
 
     /// Generates a geometry shader or returns one that already exists.
-    GLuint LazyGeometryProgram(OGLProgram& target_program, BaseBindings base_bindings,
-                               const std::string& glsl_topology, u32 max_vertices,
-                               const std::string& debug_name);
+    GLuint LazyGeometryProgram(CachedProgram& target_program, BaseBindings base_bindings,
+                               GLenum primitive_mode);
 
-    void CalculateProperties();
+    CachedProgram TryLoadProgram(GLenum primitive_mode, BaseBindings base_bindings) const;
 
-    VAddr addr{};
-    std::size_t shader_length{};
+    ShaderDiskCacheUsage GetUsage(GLenum primitive_mode, BaseBindings base_bindings) const;
+
+    u8* host_ptr{};
+    VAddr cpu_addr{};
+    u64 unique_identifier{};
     Maxwell::ShaderProgram program_type{};
-    GLShader::ShaderSetup setup;
+    ShaderDiskCacheOpenGL& disk_cache;
+    const PrecompiledPrograms& precompiled_programs;
+
+    std::size_t shader_length{};
     GLShader::ShaderEntries entries;
 
     std::string code;
 
-    std::map<BaseBindings, OGLProgram> programs;
-    std::map<BaseBindings, GeometryPrograms> geometry_programs;
+    std::unordered_map<BaseBindings, CachedProgram> programs;
+    std::unordered_map<BaseBindings, GeometryPrograms> geometry_programs;
 
-    std::map<u32, GLuint> cbuf_resource_cache;
-    std::map<u32, GLuint> gmem_resource_cache;
-    std::map<u32, GLint> uniform_cache;
+    std::unordered_map<u32, GLuint> cbuf_resource_cache;
+    std::unordered_map<u32, GLuint> gmem_resource_cache;
+    std::unordered_map<u32, GLint> uniform_cache;
 };
 
 class ShaderCacheOpenGL final : public RasterizerCache<Shader> {
 public:
-    explicit ShaderCacheOpenGL(RasterizerOpenGL& rasterizer);
+    explicit ShaderCacheOpenGL(RasterizerOpenGL& rasterizer, Core::System& system);
+
+    /// Loads disk cache for the current game
+    void LoadDiskCache(const std::atomic_bool& stop_loading,
+                       const VideoCore::DiskResourceLoadCallback& callback);
 
     /// Gets the current specified shader stage program
     Shader GetStageProgram(Maxwell::ShaderProgram program);
 
 private:
+    std::unordered_map<u64, UnspecializedShader> GenerateUnspecializedShaders(
+        const std::atomic_bool& stop_loading, const VideoCore::DiskResourceLoadCallback& callback,
+        const std::vector<ShaderDiskCacheRaw>& raws,
+        const std::unordered_map<u64, ShaderDiskCacheDecompiled>& decompiled);
+
+    CachedProgram GeneratePrecompiledProgram(const ShaderDiskCacheDump& dump,
+                                             const std::set<GLenum>& supported_formats);
+
     std::array<Shader, Maxwell::MaxShaderProgram> last_shaders;
+
+    ShaderDiskCacheOpenGL disk_cache;
+    PrecompiledShaders precompiled_shaders;
+    PrecompiledPrograms precompiled_programs;
 };
 
 } // namespace OpenGL
diff --git a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
index 36035d0d2..11d1169f0 100644
--- a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
@@ -5,7 +5,9 @@
 #include <array>
 #include <string>
 #include <string_view>
+#include <utility>
 #include <variant>
+#include <vector>
 
 #include <fmt/format.h>
 
@@ -20,6 +22,7 @@
 namespace OpenGL::GLShader {
 
 using Tegra::Shader::Attribute;
+using Tegra::Shader::AttributeUse;
 using Tegra::Shader::Header;
 using Tegra::Shader::IpaInterpMode;
 using Tegra::Shader::IpaMode;
@@ -171,7 +174,7 @@ public:
             code.AddLine(fmt::format("case 0x{:x}u: {{", address));
             ++code.scope;
 
-            VisitBasicBlock(bb);
+            VisitBlock(bb);
 
             --code.scope;
             code.AddLine('}');
@@ -193,15 +196,14 @@ public:
     ShaderEntries GetShaderEntries() const {
         ShaderEntries entries;
         for (const auto& cbuf : ir.GetConstantBuffers()) {
-            entries.const_buffers.emplace_back(cbuf.second, stage, GetConstBufferBlock(cbuf.first),
+            entries.const_buffers.emplace_back(cbuf.second.GetMaxOffset(), cbuf.second.IsIndirect(),
                                                cbuf.first);
         }
         for (const auto& sampler : ir.GetSamplers()) {
-            entries.samplers.emplace_back(sampler, stage, GetSampler(sampler));
+            entries.samplers.emplace_back(sampler);
         }
         for (const auto& gmem : ir.GetGlobalMemoryBases()) {
-            entries.global_memory_entries.emplace_back(gmem.cbuf_index, gmem.cbuf_offset, stage,
-                                                       GetGlobalMemoryBlock(gmem));
+            entries.global_memory_entries.emplace_back(gmem.cbuf_index, gmem.cbuf_offset);
         }
         entries.clip_distances = ir.GetClipDistances();
         entries.shader_length = ir.GetLength();
@@ -289,34 +291,22 @@ private:
         code.AddNewLine();
     }
 
-    std::string GetInputFlags(const IpaMode& input_mode) {
-        const IpaSampleMode sample_mode = input_mode.sampling_mode;
-        const IpaInterpMode interp_mode = input_mode.interpolation_mode;
+    std::string GetInputFlags(AttributeUse attribute) {
         std::string out;
 
-        switch (interp_mode) {
-        case IpaInterpMode::Flat:
+        switch (attribute) {
+        case AttributeUse::Constant:
             out += "flat ";
             break;
-        case IpaInterpMode::Linear:
+        case AttributeUse::ScreenLinear:
             out += "noperspective ";
             break;
-        case IpaInterpMode::Perspective:
+        case AttributeUse::Perspective:
             // Default, Smooth
             break;
         default:
-            UNIMPLEMENTED_MSG("Unhandled IPA interp mode: {}", static_cast<u32>(interp_mode));
-        }
-        switch (sample_mode) {
-        case IpaSampleMode::Centroid:
-            // It can be implemented with the "centroid " keyword in GLSL
-            UNIMPLEMENTED_MSG("Unimplemented IPA sampler mode centroid");
-            break;
-        case IpaSampleMode::Default:
-            // Default, n/a
-            break;
-        default:
-            UNIMPLEMENTED_MSG("Unimplemented IPA sampler mode: {}", static_cast<u32>(sample_mode));
+            LOG_CRITICAL(HW_GPU, "Unused attribute being fetched");
+            UNREACHABLE();
         }
         return out;
     }
@@ -325,16 +315,11 @@ private:
         const auto& attributes = ir.GetInputAttributes();
         for (const auto element : attributes) {
             const Attribute::Index index = element.first;
-            const IpaMode& input_mode = *element.second.begin();
             if (index < Attribute::Index::Attribute_0 || index > Attribute::Index::Attribute_31) {
                 // Skip when it's not a generic attribute
                 continue;
             }
 
-            ASSERT(element.second.size() > 0);
-            UNIMPLEMENTED_IF_MSG(element.second.size() > 1,
-                                 "Multiple input flag modes are not supported in GLSL");
-
             // TODO(bunnei): Use proper number of elements for these
             u32 idx = static_cast<u32>(index) - static_cast<u32>(Attribute::Index::Attribute_0);
             if (stage != ShaderStage::Vertex) {
@@ -346,8 +331,14 @@ private:
             if (stage == ShaderStage::Geometry) {
                 attr = "gs_" + attr + "[]";
             }
-            code.AddLine("layout (location = " + std::to_string(idx) + ") " +
-                         GetInputFlags(input_mode) + "in vec4 " + attr + ';');
+            std::string suffix;
+            if (stage == ShaderStage::Fragment) {
+                const auto input_mode =
+                    header.ps.GetAttributeUse(idx - GENERIC_VARYING_START_LOCATION);
+                suffix = GetInputFlags(input_mode);
+            }
+            code.AddLine("layout (location = " + std::to_string(idx) + ") " + suffix + "in vec4 " +
+                         attr + ';');
         }
         if (!attributes.empty())
             code.AddNewLine();
@@ -424,7 +415,7 @@ private:
             code.AddNewLine();
     }
 
-    void VisitBasicBlock(const BasicBlock& bb) {
+    void VisitBlock(const NodeBlock& bb) {
         for (const Node node : bb) {
             if (const std::string expr = Visit(node); !expr.empty()) {
                 code.AddLine(expr);
@@ -576,7 +567,7 @@ private:
             code.AddLine("if (" + Visit(conditional->GetCondition()) + ") {");
             ++code.scope;
 
-            VisitBasicBlock(conditional->GetCode());
+            VisitBlock(conditional->GetCode());
 
             --code.scope;
             code.AddLine('}');
@@ -617,17 +608,8 @@ private:
 
     std::string VisitOperand(Operation operation, std::size_t operand_index, Type type) {
         std::string value = VisitOperand(operation, operand_index);
-
         switch (type) {
-        case Type::Bool:
-        case Type::Bool2:
-        case Type::Float:
-            return value;
-        case Type::Int:
-            return "ftoi(" + value + ')';
-        case Type::Uint:
-            return "ftou(" + value + ')';
-        case Type::HalfFloat:
+        case Type::HalfFloat: {
             const auto half_meta = std::get_if<MetaHalfArithmetic>(&operation.GetMeta());
             if (!half_meta) {
                 value = "toHalf2(" + value + ')';
@@ -644,6 +626,26 @@ private:
                 return "vec2(toHalf2(" + value + ")[1])";
             }
         }
+        default:
+            return CastOperand(value, type);
+        }
+    }
+
+    std::string CastOperand(const std::string& value, Type type) const {
+        switch (type) {
+        case Type::Bool:
+        case Type::Bool2:
+        case Type::Float:
+            return value;
+        case Type::Int:
+            return "ftoi(" + value + ')';
+        case Type::Uint:
+            return "ftou(" + value + ')';
+        case Type::HalfFloat:
+            // Can't be handled as a stand-alone value
+            UNREACHABLE();
+            return value;
+        }
         UNREACHABLE();
         return value;
     }
@@ -651,6 +653,7 @@ private:
     std::string BitwiseCastResult(std::string value, Type type, bool needs_parenthesis = false) {
         switch (type) {
         case Type::Bool:
+        case Type::Bool2:
         case Type::Float:
             if (needs_parenthesis) {
                 return '(' + value + ')';
@@ -716,51 +719,68 @@ private:
     }
 
     std::string GenerateTexture(Operation operation, const std::string& func,
-                                bool is_extra_int = false) {
+                                const std::vector<std::pair<Type, Node>>& extras) {
         constexpr std::array<const char*, 4> coord_constructors = {"float", "vec2", "vec3", "vec4"};
 
         const auto meta = std::get_if<MetaTexture>(&operation.GetMeta());
-        const auto count = static_cast<u32>(operation.GetOperandsCount());
         ASSERT(meta);
 
+        const std::size_t count = operation.GetOperandsCount();
+        const bool has_array = meta->sampler.IsArray();
+        const bool has_shadow = meta->sampler.IsShadow();
+
         std::string expr = func;
         expr += '(';
         expr += GetSampler(meta->sampler);
         expr += ", ";
 
-        expr += coord_constructors[meta->coords_count - 1];
+        expr += coord_constructors.at(count + (has_array ? 1 : 0) + (has_shadow ? 1 : 0) - 1);
         expr += '(';
-        for (u32 i = 0; i < count; ++i) {
-            const bool is_extra = i >= meta->coords_count;
-            const bool is_array = i == meta->array_index;
-
-            std::string operand = [&]() {
-                if (is_extra && is_extra_int) {
-                    if (const auto immediate = std::get_if<ImmediateNode>(operation[i])) {
-                        return std::to_string(static_cast<s32>(immediate->GetValue()));
-                    } else {
-                        return "ftoi(" + Visit(operation[i]) + ')';
-                    }
-                } else {
-                    return Visit(operation[i]);
-                }
-            }();
-            if (is_array) {
-                ASSERT(!is_extra);
-                operand = "float(ftoi(" + operand + "))";
-            }
+        for (std::size_t i = 0; i < count; ++i) {
+            expr += Visit(operation[i]);
 
-            expr += operand;
+            const std::size_t next = i + 1;
+            if (next < count)
+                expr += ", ";
+        }
+        if (has_array) {
+            expr += ", float(ftoi(" + Visit(meta->array) + "))";
+        }
+        if (has_shadow) {
+            expr += ", " + Visit(meta->depth_compare);
+        }
+        expr += ')';
 
-            if (i + 1 == meta->coords_count) {
-                expr += ')';
+        for (const auto& extra_pair : extras) {
+            const auto [type, operand] = extra_pair;
+            if (operand == nullptr) {
+                continue;
+            }
+            expr += ", ";
+
+            switch (type) {
+            case Type::Int:
+                if (const auto immediate = std::get_if<ImmediateNode>(operand)) {
+                    // Inline the string as an immediate integer in GLSL (some extra arguments are
+                    // required to be constant)
+                    expr += std::to_string(static_cast<s32>(immediate->GetValue()));
+                } else {
+                    expr += "ftoi(" + Visit(operand) + ')';
+                }
+                break;
+            case Type::Float:
+                expr += Visit(operand);
+                break;
+            default: {
+                const auto type_int = static_cast<u32>(type);
+                UNIMPLEMENTED_MSG("Unimplemented extra type={}", type_int);
+                expr += '0';
+                break;
             }
-            if (i + 1 < count) {
-                expr += ", ";
             }
         }
-        expr += ')';
-        return expr;
+
+        return expr + ')';
     }
 
     std::string Assign(Operation operation) {
@@ -1135,37 +1155,38 @@ private:
                                   Type::HalfFloat);
     }
 
-    std::string F4Texture(Operation operation) {
+    std::string Texture(Operation operation) {
         const auto meta = std::get_if<MetaTexture>(&operation.GetMeta());
         ASSERT(meta);
 
-        std::string expr = GenerateTexture(operation, "texture");
+        std::string expr = GenerateTexture(operation, "texture", {{Type::Float, meta->bias}});
         if (meta->sampler.IsShadow()) {
             expr = "vec4(" + expr + ')';
         }
         return expr + GetSwizzle(meta->element);
     }
 
-    std::string F4TextureLod(Operation operation) {
+    std::string TextureLod(Operation operation) {
         const auto meta = std::get_if<MetaTexture>(&operation.GetMeta());
         ASSERT(meta);
 
-        std::string expr = GenerateTexture(operation, "textureLod");
+        std::string expr = GenerateTexture(operation, "textureLod", {{Type::Float, meta->lod}});
         if (meta->sampler.IsShadow()) {
             expr = "vec4(" + expr + ')';
         }
         return expr + GetSwizzle(meta->element);
     }
 
-    std::string F4TextureGather(Operation operation) {
+    std::string TextureGather(Operation operation) {
         const auto meta = std::get_if<MetaTexture>(&operation.GetMeta());
         ASSERT(meta);
 
-        return GenerateTexture(operation, "textureGather", !meta->sampler.IsShadow()) +
+        const auto type = meta->sampler.IsShadow() ? Type::Float : Type::Int;
+        return GenerateTexture(operation, "textureGather", {{type, meta->component}}) +
                GetSwizzle(meta->element);
     }
 
-    std::string F4TextureQueryDimensions(Operation operation) {
+    std::string TextureQueryDimensions(Operation operation) {
         const auto meta = std::get_if<MetaTexture>(&operation.GetMeta());
         ASSERT(meta);
 
@@ -1185,40 +1206,44 @@ private:
         return "0";
     }
 
-    std::string F4TextureQueryLod(Operation operation) {
+    std::string TextureQueryLod(Operation operation) {
         const auto meta = std::get_if<MetaTexture>(&operation.GetMeta());
         ASSERT(meta);
 
         if (meta->element < 2) {
-            return "itof(int((" + GenerateTexture(operation, "textureQueryLod") + " * vec2(256))" +
-                   GetSwizzle(meta->element) + "))";
+            return "itof(int((" + GenerateTexture(operation, "textureQueryLod", {}) +
+                   " * vec2(256))" + GetSwizzle(meta->element) + "))";
         }
         return "0";
     }
 
-    std::string F4TexelFetch(Operation operation) {
+    std::string TexelFetch(Operation operation) {
         constexpr std::array<const char*, 4> constructors = {"int", "ivec2", "ivec3", "ivec4"};
         const auto meta = std::get_if<MetaTexture>(&operation.GetMeta());
-        const auto count = static_cast<u32>(operation.GetOperandsCount());
         ASSERT(meta);
+        UNIMPLEMENTED_IF(meta->sampler.IsArray());
+        const std::size_t count = operation.GetOperandsCount();
 
         std::string expr = "texelFetch(";
         expr += GetSampler(meta->sampler);
         expr += ", ";
 
-        expr += constructors[meta->coords_count - 1];
+        expr += constructors.at(operation.GetOperandsCount() - 1);
         expr += '(';
-        for (u32 i = 0; i < count; ++i) {
+        for (std::size_t i = 0; i < count; ++i) {
             expr += VisitOperand(operation, i, Type::Int);
-
-            if (i + 1 == meta->coords_count) {
+            const std::size_t next = i + 1;
+            if (next == count)
                 expr += ')';
-            }
-            if (i + 1 < count) {
+            else if (next < count)
                 expr += ", ";
-            }
+        }
+        if (meta->lod) {
+            expr += ", ";
+            expr += CastOperand(Visit(meta->lod), Type::Int);
         }
         expr += ')';
+
         return expr + GetSwizzle(meta->element);
     }
 
@@ -1455,12 +1480,12 @@ private:
         &GLSLDecompiler::Logical2HNotEqual,
         &GLSLDecompiler::Logical2HGreaterEqual,
 
-        &GLSLDecompiler::F4Texture,
-        &GLSLDecompiler::F4TextureLod,
-        &GLSLDecompiler::F4TextureGather,
-        &GLSLDecompiler::F4TextureQueryDimensions,
-        &GLSLDecompiler::F4TextureQueryLod,
-        &GLSLDecompiler::F4TexelFetch,
+        &GLSLDecompiler::Texture,
+        &GLSLDecompiler::TextureLod,
+        &GLSLDecompiler::TextureGather,
+        &GLSLDecompiler::TextureQueryDimensions,
+        &GLSLDecompiler::TextureQueryLod,
+        &GLSLDecompiler::TexelFetch,
 
         &GLSLDecompiler::Branch,
         &GLSLDecompiler::PushFlowStack,
@@ -1563,4 +1588,4 @@ ProgramResult Decompile(const ShaderIR& ir, Maxwell::ShaderStage stage, const st
     return {decompiler.GetResult(), decompiler.GetShaderEntries()};
 }
 
-} // namespace OpenGL::GLShader
\ No newline at end of file
+} // namespace OpenGL::GLShader
diff --git a/src/video_core/renderer_opengl/gl_shader_decompiler.h b/src/video_core/renderer_opengl/gl_shader_decompiler.h
index 0856a1361..72aca4938 100644
--- a/src/video_core/renderer_opengl/gl_shader_decompiler.h
+++ b/src/video_core/renderer_opengl/gl_shader_decompiler.h
@@ -5,6 +5,7 @@
 #pragma once
 
 #include <array>
+#include <set>
 #include <string>
 #include <utility>
 #include <vector>
@@ -18,56 +19,29 @@ class ShaderIR;
 
 namespace OpenGL::GLShader {
 
+struct ShaderEntries;
+
 using Maxwell = Tegra::Engines::Maxwell3D::Regs;
+using ProgramResult = std::pair<std::string, ShaderEntries>;
+using SamplerEntry = VideoCommon::Shader::Sampler;
 
 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;
-    }
+    explicit ConstBufferEntry(u32 max_offset, bool is_indirect, u32 index)
+        : VideoCommon::Shader::ConstBuffer{max_offset, is_indirect}, index{index} {}
 
     u32 GetIndex() const {
         return 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;
-    }
-
-private:
-    std::string name;
-    Maxwell::ShaderStage stage{};
-};
-
 class GlobalMemoryEntry {
 public:
-    explicit GlobalMemoryEntry(u32 cbuf_index, u32 cbuf_offset, Maxwell::ShaderStage stage,
-                               std::string name)
-        : cbuf_index{cbuf_index}, cbuf_offset{cbuf_offset}, stage{stage}, name{std::move(name)} {}
+    explicit GlobalMemoryEntry(u32 cbuf_index, u32 cbuf_offset)
+        : cbuf_index{cbuf_index}, cbuf_offset{cbuf_offset} {}
 
     u32 GetCbufIndex() const {
         return cbuf_index;
@@ -77,19 +51,9 @@ public:
         return cbuf_offset;
     }
 
-    const std::string& GetName() const {
-        return name;
-    }
-
-    Maxwell::ShaderStage GetStage() const {
-        return stage;
-    }
-
 private:
     u32 cbuf_index{};
     u32 cbuf_offset{};
-    Maxwell::ShaderStage stage{};
-    std::string name;
 };
 
 struct ShaderEntries {
@@ -100,8 +64,6 @@ struct ShaderEntries {
     std::size_t shader_length{};
 };
 
-using ProgramResult = std::pair<std::string, ShaderEntries>;
-
 std::string GetCommonDeclarations();
 
 ProgramResult Decompile(const VideoCommon::Shader::ShaderIR& ir, Maxwell::ShaderStage stage,
diff --git a/src/video_core/renderer_opengl/gl_shader_disk_cache.cpp b/src/video_core/renderer_opengl/gl_shader_disk_cache.cpp
new file mode 100644
index 000000000..82fc4d44b
--- /dev/null
+++ b/src/video_core/renderer_opengl/gl_shader_disk_cache.cpp
@@ -0,0 +1,654 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <cstring>
+#include <fmt/format.h>
+#include <lz4.h>
+
+#include "common/assert.h"
+#include "common/common_paths.h"
+#include "common/common_types.h"
+#include "common/file_util.h"
+#include "common/logging/log.h"
+#include "common/scm_rev.h"
+
+#include "core/core.h"
+#include "core/hle/kernel/process.h"
+#include "core/settings.h"
+
+#include "video_core/renderer_opengl/gl_shader_cache.h"
+#include "video_core/renderer_opengl/gl_shader_disk_cache.h"
+
+namespace OpenGL {
+
+using ShaderCacheVersionHash = std::array<u8, 64>;
+
+enum class TransferableEntryKind : u32 {
+    Raw,
+    Usage,
+};
+
+enum class PrecompiledEntryKind : u32 {
+    Decompiled,
+    Dump,
+};
+
+constexpr u32 NativeVersion = 1;
+
+// Making sure sizes doesn't change by accident
+static_assert(sizeof(BaseBindings) == 12);
+static_assert(sizeof(ShaderDiskCacheUsage) == 24);
+
+namespace {
+
+ShaderCacheVersionHash GetShaderCacheVersionHash() {
+    ShaderCacheVersionHash hash{};
+    const std::size_t length = std::min(std::strlen(Common::g_shader_cache_version), hash.size());
+    std::memcpy(hash.data(), Common::g_shader_cache_version, length);
+    return hash;
+}
+
+template <typename T>
+std::vector<u8> CompressData(const T* source, std::size_t source_size) {
+    if (source_size > LZ4_MAX_INPUT_SIZE) {
+        // Source size exceeds LZ4 maximum input size
+        return {};
+    }
+    const auto source_size_int = static_cast<int>(source_size);
+    const int max_compressed_size = LZ4_compressBound(source_size_int);
+    std::vector<u8> compressed(max_compressed_size);
+    const int compressed_size = LZ4_compress_default(reinterpret_cast<const char*>(source),
+                                                     reinterpret_cast<char*>(compressed.data()),
+                                                     source_size_int, max_compressed_size);
+    if (compressed_size <= 0) {
+        // Compression failed
+        return {};
+    }
+    compressed.resize(compressed_size);
+    return compressed;
+}
+
+std::vector<u8> DecompressData(const std::vector<u8>& compressed, std::size_t uncompressed_size) {
+    std::vector<u8> uncompressed(uncompressed_size);
+    const int size_check = LZ4_decompress_safe(reinterpret_cast<const char*>(compressed.data()),
+                                               reinterpret_cast<char*>(uncompressed.data()),
+                                               static_cast<int>(compressed.size()),
+                                               static_cast<int>(uncompressed.size()));
+    if (static_cast<int>(uncompressed_size) != size_check) {
+        // Decompression failed
+        return {};
+    }
+    return uncompressed;
+}
+
+} // namespace
+
+ShaderDiskCacheRaw::ShaderDiskCacheRaw(u64 unique_identifier, Maxwell::ShaderProgram program_type,
+                                       u32 program_code_size, u32 program_code_size_b,
+                                       ProgramCode program_code, ProgramCode program_code_b)
+    : unique_identifier{unique_identifier}, program_type{program_type},
+      program_code_size{program_code_size}, program_code_size_b{program_code_size_b},
+      program_code{std::move(program_code)}, program_code_b{std::move(program_code_b)} {}
+
+ShaderDiskCacheRaw::ShaderDiskCacheRaw() = default;
+
+ShaderDiskCacheRaw::~ShaderDiskCacheRaw() = default;
+
+bool ShaderDiskCacheRaw::Load(FileUtil::IOFile& file) {
+    if (file.ReadBytes(&unique_identifier, sizeof(u64)) != sizeof(u64) ||
+        file.ReadBytes(&program_type, sizeof(u32)) != sizeof(u32)) {
+        return false;
+    }
+    u32 program_code_size{};
+    u32 program_code_size_b{};
+    if (file.ReadBytes(&program_code_size, sizeof(u32)) != sizeof(u32) ||
+        file.ReadBytes(&program_code_size_b, sizeof(u32)) != sizeof(u32)) {
+        return false;
+    }
+
+    program_code.resize(program_code_size);
+    program_code_b.resize(program_code_size_b);
+
+    if (file.ReadArray(program_code.data(), program_code_size) != program_code_size)
+        return false;
+
+    if (HasProgramA() &&
+        file.ReadArray(program_code_b.data(), program_code_size_b) != program_code_size_b) {
+        return false;
+    }
+    return true;
+}
+
+bool ShaderDiskCacheRaw::Save(FileUtil::IOFile& file) const {
+    if (file.WriteObject(unique_identifier) != 1 ||
+        file.WriteObject(static_cast<u32>(program_type)) != 1 ||
+        file.WriteObject(program_code_size) != 1 || file.WriteObject(program_code_size_b) != 1) {
+        return false;
+    }
+
+    if (file.WriteArray(program_code.data(), program_code_size) != program_code_size)
+        return false;
+
+    if (HasProgramA() &&
+        file.WriteArray(program_code_b.data(), program_code_size_b) != program_code_size_b) {
+        return false;
+    }
+    return true;
+}
+
+ShaderDiskCacheOpenGL::ShaderDiskCacheOpenGL(Core::System& system) : system{system} {}
+
+std::optional<std::pair<std::vector<ShaderDiskCacheRaw>, std::vector<ShaderDiskCacheUsage>>>
+ShaderDiskCacheOpenGL::LoadTransferable() {
+    // Skip games without title id
+    const bool has_title_id = system.CurrentProcess()->GetTitleID() != 0;
+    if (!Settings::values.use_disk_shader_cache || !has_title_id)
+        return {};
+    tried_to_load = true;
+
+    FileUtil::IOFile file(GetTransferablePath(), "rb");
+    if (!file.IsOpen()) {
+        LOG_INFO(Render_OpenGL, "No transferable shader cache found for game with title id={}",
+                 GetTitleID());
+        return {};
+    }
+
+    u32 version{};
+    if (file.ReadBytes(&version, sizeof(version)) != sizeof(version)) {
+        LOG_ERROR(Render_OpenGL,
+                  "Failed to get transferable cache version for title id={} - skipping",
+                  GetTitleID());
+        return {};
+    }
+
+    if (version < NativeVersion) {
+        LOG_INFO(Render_OpenGL, "Transferable shader cache is old - removing");
+        file.Close();
+        InvalidateTransferable();
+        return {};
+    }
+    if (version > NativeVersion) {
+        LOG_WARNING(Render_OpenGL, "Transferable shader cache was generated with a newer version "
+                                   "of the emulator - skipping");
+        return {};
+    }
+
+    // Version is valid, load the shaders
+    std::vector<ShaderDiskCacheRaw> raws;
+    std::vector<ShaderDiskCacheUsage> usages;
+    while (file.Tell() < file.GetSize()) {
+        TransferableEntryKind kind{};
+        if (file.ReadBytes(&kind, sizeof(u32)) != sizeof(u32)) {
+            LOG_ERROR(Render_OpenGL, "Failed to read transferable file - skipping");
+            return {};
+        }
+
+        switch (kind) {
+        case TransferableEntryKind::Raw: {
+            ShaderDiskCacheRaw entry;
+            if (!entry.Load(file)) {
+                LOG_ERROR(Render_OpenGL, "Failed to load transferable raw entry - skipping");
+                return {};
+            }
+            transferable.insert({entry.GetUniqueIdentifier(), {}});
+            raws.push_back(std::move(entry));
+            break;
+        }
+        case TransferableEntryKind::Usage: {
+            ShaderDiskCacheUsage usage{};
+            if (file.ReadBytes(&usage, sizeof(usage)) != sizeof(usage)) {
+                LOG_ERROR(Render_OpenGL, "Failed to load transferable usage entry - skipping");
+                return {};
+            }
+            usages.push_back(std::move(usage));
+            break;
+        }
+        default:
+            LOG_ERROR(Render_OpenGL, "Unknown transferable shader cache entry kind={} - skipping",
+                      static_cast<u32>(kind));
+            return {};
+        }
+    }
+    return {{raws, usages}};
+}
+
+std::pair<std::unordered_map<u64, ShaderDiskCacheDecompiled>,
+          std::unordered_map<ShaderDiskCacheUsage, ShaderDiskCacheDump>>
+ShaderDiskCacheOpenGL::LoadPrecompiled() {
+    if (!IsUsable())
+        return {};
+
+    FileUtil::IOFile file(GetPrecompiledPath(), "rb");
+    if (!file.IsOpen()) {
+        LOG_INFO(Render_OpenGL, "No precompiled shader cache found for game with title id={}",
+                 GetTitleID());
+        return {};
+    }
+
+    const auto result = LoadPrecompiledFile(file);
+    if (!result) {
+        LOG_INFO(Render_OpenGL,
+                 "Failed to load precompiled cache for game with title id={} - removing",
+                 GetTitleID());
+        file.Close();
+        InvalidatePrecompiled();
+        return {};
+    }
+    return *result;
+}
+
+std::optional<std::pair<std::unordered_map<u64, ShaderDiskCacheDecompiled>,
+                        std::unordered_map<ShaderDiskCacheUsage, ShaderDiskCacheDump>>>
+ShaderDiskCacheOpenGL::LoadPrecompiledFile(FileUtil::IOFile& file) {
+    ShaderCacheVersionHash file_hash{};
+    if (file.ReadArray(file_hash.data(), file_hash.size()) != file_hash.size()) {
+        return {};
+    }
+    if (GetShaderCacheVersionHash() != file_hash) {
+        LOG_INFO(Render_OpenGL, "Precompiled cache is from another version of the emulator");
+        return {};
+    }
+
+    std::unordered_map<u64, ShaderDiskCacheDecompiled> decompiled;
+    std::unordered_map<ShaderDiskCacheUsage, ShaderDiskCacheDump> dumps;
+    while (file.Tell() < file.GetSize()) {
+        PrecompiledEntryKind kind{};
+        if (file.ReadBytes(&kind, sizeof(u32)) != sizeof(u32)) {
+            return {};
+        }
+
+        switch (kind) {
+        case PrecompiledEntryKind::Decompiled: {
+            u64 unique_identifier{};
+            if (file.ReadBytes(&unique_identifier, sizeof(u64)) != sizeof(u64))
+                return {};
+
+            const auto entry = LoadDecompiledEntry(file);
+            if (!entry)
+                return {};
+            decompiled.insert({unique_identifier, std::move(*entry)});
+            break;
+        }
+        case PrecompiledEntryKind::Dump: {
+            ShaderDiskCacheUsage usage;
+            if (file.ReadBytes(&usage, sizeof(usage)) != sizeof(usage))
+                return {};
+
+            ShaderDiskCacheDump dump;
+            if (file.ReadBytes(&dump.binary_format, sizeof(u32)) != sizeof(u32))
+                return {};
+
+            u32 binary_length{};
+            u32 compressed_size{};
+            if (file.ReadBytes(&binary_length, sizeof(u32)) != sizeof(u32) ||
+                file.ReadBytes(&compressed_size, sizeof(u32)) != sizeof(u32)) {
+                return {};
+            }
+
+            std::vector<u8> compressed_binary(compressed_size);
+            if (file.ReadArray(compressed_binary.data(), compressed_binary.size()) !=
+                compressed_binary.size()) {
+                return {};
+            }
+
+            dump.binary = DecompressData(compressed_binary, binary_length);
+            if (dump.binary.empty()) {
+                return {};
+            }
+
+            dumps.insert({usage, dump});
+            break;
+        }
+        default:
+            return {};
+        }
+    }
+    return {{decompiled, dumps}};
+}
+
+std::optional<ShaderDiskCacheDecompiled> ShaderDiskCacheOpenGL::LoadDecompiledEntry(
+    FileUtil::IOFile& file) {
+    u32 code_size{};
+    u32 compressed_code_size{};
+    if (file.ReadBytes(&code_size, sizeof(u32)) != sizeof(u32) ||
+        file.ReadBytes(&compressed_code_size, sizeof(u32)) != sizeof(u32)) {
+        return {};
+    }
+
+    std::vector<u8> compressed_code(compressed_code_size);
+    if (file.ReadArray(compressed_code.data(), compressed_code.size()) != compressed_code.size()) {
+        return {};
+    }
+
+    const std::vector<u8> code = DecompressData(compressed_code, code_size);
+    if (code.empty()) {
+        return {};
+    }
+    ShaderDiskCacheDecompiled entry;
+    entry.code = std::string(reinterpret_cast<const char*>(code.data()), code_size);
+
+    u32 const_buffers_count{};
+    if (file.ReadBytes(&const_buffers_count, sizeof(u32)) != sizeof(u32))
+        return {};
+    for (u32 i = 0; i < const_buffers_count; ++i) {
+        u32 max_offset{};
+        u32 index{};
+        u8 is_indirect{};
+        if (file.ReadBytes(&max_offset, sizeof(u32)) != sizeof(u32) ||
+            file.ReadBytes(&index, sizeof(u32)) != sizeof(u32) ||
+            file.ReadBytes(&is_indirect, sizeof(u8)) != sizeof(u8)) {
+            return {};
+        }
+        entry.entries.const_buffers.emplace_back(max_offset, is_indirect != 0, index);
+    }
+
+    u32 samplers_count{};
+    if (file.ReadBytes(&samplers_count, sizeof(u32)) != sizeof(u32))
+        return {};
+    for (u32 i = 0; i < samplers_count; ++i) {
+        u64 offset{};
+        u64 index{};
+        u32 type{};
+        u8 is_array{};
+        u8 is_shadow{};
+        if (file.ReadBytes(&offset, sizeof(u64)) != sizeof(u64) ||
+            file.ReadBytes(&index, sizeof(u64)) != sizeof(u64) ||
+            file.ReadBytes(&type, sizeof(u32)) != sizeof(u32) ||
+            file.ReadBytes(&is_array, sizeof(u8)) != sizeof(u8) ||
+            file.ReadBytes(&is_shadow, sizeof(u8)) != sizeof(u8)) {
+            return {};
+        }
+        entry.entries.samplers.emplace_back(
+            static_cast<std::size_t>(offset), static_cast<std::size_t>(index),
+            static_cast<Tegra::Shader::TextureType>(type), is_array != 0, is_shadow != 0);
+    }
+
+    u32 global_memory_count{};
+    if (file.ReadBytes(&global_memory_count, sizeof(u32)) != sizeof(u32))
+        return {};
+    for (u32 i = 0; i < global_memory_count; ++i) {
+        u32 cbuf_index{};
+        u32 cbuf_offset{};
+        if (file.ReadBytes(&cbuf_index, sizeof(u32)) != sizeof(u32) ||
+            file.ReadBytes(&cbuf_offset, sizeof(u32)) != sizeof(u32)) {
+            return {};
+        }
+        entry.entries.global_memory_entries.emplace_back(cbuf_index, cbuf_offset);
+    }
+
+    for (auto& clip_distance : entry.entries.clip_distances) {
+        u8 clip_distance_raw{};
+        if (file.ReadBytes(&clip_distance_raw, sizeof(u8)) != sizeof(u8))
+            return {};
+        clip_distance = clip_distance_raw != 0;
+    }
+
+    u64 shader_length{};
+    if (file.ReadBytes(&shader_length, sizeof(u64)) != sizeof(u64))
+        return {};
+    entry.entries.shader_length = static_cast<std::size_t>(shader_length);
+
+    return entry;
+}
+
+bool ShaderDiskCacheOpenGL::SaveDecompiledFile(FileUtil::IOFile& file, u64 unique_identifier,
+                                               const std::string& code,
+                                               const std::vector<u8>& compressed_code,
+                                               const GLShader::ShaderEntries& entries) {
+    if (file.WriteObject(static_cast<u32>(PrecompiledEntryKind::Decompiled)) != 1 ||
+        file.WriteObject(unique_identifier) != 1 ||
+        file.WriteObject(static_cast<u32>(code.size())) != 1 ||
+        file.WriteObject(static_cast<u32>(compressed_code.size())) != 1 ||
+        file.WriteArray(compressed_code.data(), compressed_code.size()) != compressed_code.size()) {
+        return false;
+    }
+
+    if (file.WriteObject(static_cast<u32>(entries.const_buffers.size())) != 1)
+        return false;
+    for (const auto& cbuf : entries.const_buffers) {
+        if (file.WriteObject(static_cast<u32>(cbuf.GetMaxOffset())) != 1 ||
+            file.WriteObject(static_cast<u32>(cbuf.GetIndex())) != 1 ||
+            file.WriteObject(static_cast<u8>(cbuf.IsIndirect() ? 1 : 0)) != 1) {
+            return false;
+        }
+    }
+
+    if (file.WriteObject(static_cast<u32>(entries.samplers.size())) != 1)
+        return false;
+    for (const auto& sampler : entries.samplers) {
+        if (file.WriteObject(static_cast<u64>(sampler.GetOffset())) != 1 ||
+            file.WriteObject(static_cast<u64>(sampler.GetIndex())) != 1 ||
+            file.WriteObject(static_cast<u32>(sampler.GetType())) != 1 ||
+            file.WriteObject(static_cast<u8>(sampler.IsArray() ? 1 : 0)) != 1 ||
+            file.WriteObject(static_cast<u8>(sampler.IsShadow() ? 1 : 0)) != 1) {
+            return false;
+        }
+    }
+
+    if (file.WriteObject(static_cast<u32>(entries.global_memory_entries.size())) != 1)
+        return false;
+    for (const auto& gmem : entries.global_memory_entries) {
+        if (file.WriteObject(static_cast<u32>(gmem.GetCbufIndex())) != 1 ||
+            file.WriteObject(static_cast<u32>(gmem.GetCbufOffset())) != 1) {
+            return false;
+        }
+    }
+
+    for (const bool clip_distance : entries.clip_distances) {
+        if (file.WriteObject(static_cast<u8>(clip_distance ? 1 : 0)) != 1)
+            return false;
+    }
+
+    return file.WriteObject(static_cast<u64>(entries.shader_length)) == 1;
+}
+
+void ShaderDiskCacheOpenGL::InvalidateTransferable() const {
+    if (!FileUtil::Delete(GetTransferablePath())) {
+        LOG_ERROR(Render_OpenGL, "Failed to invalidate transferable file={}",
+                  GetTransferablePath());
+    }
+    InvalidatePrecompiled();
+}
+
+void ShaderDiskCacheOpenGL::InvalidatePrecompiled() const {
+    if (!FileUtil::Delete(GetPrecompiledPath())) {
+        LOG_ERROR(Render_OpenGL, "Failed to invalidate precompiled file={}", GetPrecompiledPath());
+    }
+}
+
+void ShaderDiskCacheOpenGL::SaveRaw(const ShaderDiskCacheRaw& entry) {
+    if (!IsUsable())
+        return;
+
+    const u64 id = entry.GetUniqueIdentifier();
+    if (transferable.find(id) != transferable.end()) {
+        // The shader already exists
+        return;
+    }
+
+    FileUtil::IOFile file = AppendTransferableFile();
+    if (!file.IsOpen())
+        return;
+    if (file.WriteObject(TransferableEntryKind::Raw) != 1 || !entry.Save(file)) {
+        LOG_ERROR(Render_OpenGL, "Failed to save raw transferable cache entry - removing");
+        file.Close();
+        InvalidateTransferable();
+        return;
+    }
+    transferable.insert({id, {}});
+}
+
+void ShaderDiskCacheOpenGL::SaveUsage(const ShaderDiskCacheUsage& usage) {
+    if (!IsUsable())
+        return;
+
+    const auto it = transferable.find(usage.unique_identifier);
+    ASSERT_MSG(it != transferable.end(), "Saving shader usage without storing raw previously");
+
+    auto& usages{it->second};
+    ASSERT(usages.find(usage) == usages.end());
+    usages.insert(usage);
+
+    FileUtil::IOFile file = AppendTransferableFile();
+    if (!file.IsOpen())
+        return;
+
+    if (file.WriteObject(TransferableEntryKind::Usage) != 1 || file.WriteObject(usage) != 1) {
+        LOG_ERROR(Render_OpenGL, "Failed to save usage transferable cache entry - removing");
+        file.Close();
+        InvalidateTransferable();
+        return;
+    }
+}
+
+void ShaderDiskCacheOpenGL::SaveDecompiled(u64 unique_identifier, const std::string& code,
+                                           const GLShader::ShaderEntries& entries) {
+    if (!IsUsable())
+        return;
+
+    const std::vector<u8> compressed_code{CompressData(code.data(), code.size())};
+    if (compressed_code.empty()) {
+        LOG_ERROR(Render_OpenGL, "Failed to compress GLSL code - skipping shader {:016x}",
+                  unique_identifier);
+        return;
+    }
+
+    FileUtil::IOFile file = AppendPrecompiledFile();
+    if (!file.IsOpen())
+        return;
+
+    if (!SaveDecompiledFile(file, unique_identifier, code, compressed_code, entries)) {
+        LOG_ERROR(Render_OpenGL,
+                  "Failed to save decompiled entry to the precompiled file - removing");
+        file.Close();
+        InvalidatePrecompiled();
+    }
+}
+
+void ShaderDiskCacheOpenGL::SaveDump(const ShaderDiskCacheUsage& usage, GLuint program) {
+    if (!IsUsable())
+        return;
+
+    GLint binary_length{};
+    glGetProgramiv(program, GL_PROGRAM_BINARY_LENGTH, &binary_length);
+
+    GLenum binary_format{};
+    std::vector<u8> binary(binary_length);
+    glGetProgramBinary(program, binary_length, nullptr, &binary_format, binary.data());
+
+    const std::vector<u8> compressed_binary = CompressData(binary.data(), binary.size());
+    if (compressed_binary.empty()) {
+        LOG_ERROR(Render_OpenGL, "Failed to compress binary program in shader={:016x}",
+                  usage.unique_identifier);
+        return;
+    }
+
+    FileUtil::IOFile file = AppendPrecompiledFile();
+    if (!file.IsOpen())
+        return;
+
+    if (file.WriteObject(static_cast<u32>(PrecompiledEntryKind::Dump)) != 1 ||
+        file.WriteObject(usage) != 1 || file.WriteObject(static_cast<u32>(binary_format)) != 1 ||
+        file.WriteObject(static_cast<u32>(binary_length)) != 1 ||
+        file.WriteObject(static_cast<u32>(compressed_binary.size())) != 1 ||
+        file.WriteArray(compressed_binary.data(), compressed_binary.size()) !=
+            compressed_binary.size()) {
+        LOG_ERROR(Render_OpenGL, "Failed to save binary program file in shader={:016x} - removing",
+                  usage.unique_identifier);
+        file.Close();
+        InvalidatePrecompiled();
+        return;
+    }
+}
+
+bool ShaderDiskCacheOpenGL::IsUsable() const {
+    return tried_to_load && Settings::values.use_disk_shader_cache;
+}
+
+FileUtil::IOFile ShaderDiskCacheOpenGL::AppendTransferableFile() const {
+    if (!EnsureDirectories())
+        return {};
+
+    const auto transferable_path{GetTransferablePath()};
+    const bool existed = FileUtil::Exists(transferable_path);
+
+    FileUtil::IOFile file(transferable_path, "ab");
+    if (!file.IsOpen()) {
+        LOG_ERROR(Render_OpenGL, "Failed to open transferable cache in path={}", transferable_path);
+        return {};
+    }
+    if (!existed || file.GetSize() == 0) {
+        // If the file didn't exist, write its version
+        if (file.WriteObject(NativeVersion) != 1) {
+            LOG_ERROR(Render_OpenGL, "Failed to write transferable cache version in path={}",
+                      transferable_path);
+            return {};
+        }
+    }
+    return file;
+}
+
+FileUtil::IOFile ShaderDiskCacheOpenGL::AppendPrecompiledFile() const {
+    if (!EnsureDirectories())
+        return {};
+
+    const auto precompiled_path{GetPrecompiledPath()};
+    const bool existed = FileUtil::Exists(precompiled_path);
+
+    FileUtil::IOFile file(precompiled_path, "ab");
+    if (!file.IsOpen()) {
+        LOG_ERROR(Render_OpenGL, "Failed to open precompiled cache in path={}", precompiled_path);
+        return {};
+    }
+
+    if (!existed || file.GetSize() == 0) {
+        const auto hash{GetShaderCacheVersionHash()};
+        if (file.WriteArray(hash.data(), hash.size()) != hash.size()) {
+            LOG_ERROR(Render_OpenGL, "Failed to write precompiled cache version hash in path={}",
+                      precompiled_path);
+            return {};
+        }
+    }
+    return file;
+}
+
+bool ShaderDiskCacheOpenGL::EnsureDirectories() const {
+    const auto CreateDir = [](const std::string& dir) {
+        if (!FileUtil::CreateDir(dir)) {
+            LOG_ERROR(Render_OpenGL, "Failed to create directory={}", dir);
+            return false;
+        }
+        return true;
+    };
+
+    return CreateDir(FileUtil::GetUserPath(FileUtil::UserPath::ShaderDir)) &&
+           CreateDir(GetBaseDir()) && CreateDir(GetTransferableDir()) &&
+           CreateDir(GetPrecompiledDir());
+}
+
+std::string ShaderDiskCacheOpenGL::GetTransferablePath() const {
+    return FileUtil::SanitizePath(GetTransferableDir() + DIR_SEP_CHR + GetTitleID() + ".bin");
+}
+
+std::string ShaderDiskCacheOpenGL::GetPrecompiledPath() const {
+    return FileUtil::SanitizePath(GetPrecompiledDir() + DIR_SEP_CHR + GetTitleID() + ".bin");
+}
+
+std::string ShaderDiskCacheOpenGL::GetTransferableDir() const {
+    return GetBaseDir() + DIR_SEP "transferable";
+}
+
+std::string ShaderDiskCacheOpenGL::GetPrecompiledDir() const {
+    return GetBaseDir() + DIR_SEP "precompiled";
+}
+
+std::string ShaderDiskCacheOpenGL::GetBaseDir() const {
+    return FileUtil::GetUserPath(FileUtil::UserPath::ShaderDir) + DIR_SEP "opengl";
+}
+
+std::string ShaderDiskCacheOpenGL::GetTitleID() const {
+    return fmt::format("{:016X}", system.CurrentProcess()->GetTitleID());
+}
+
+} // namespace OpenGL
diff --git a/src/video_core/renderer_opengl/gl_shader_disk_cache.h b/src/video_core/renderer_opengl/gl_shader_disk_cache.h
new file mode 100644
index 000000000..6be0c0547
--- /dev/null
+++ b/src/video_core/renderer_opengl/gl_shader_disk_cache.h
@@ -0,0 +1,245 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <optional>
+#include <string>
+#include <tuple>
+#include <unordered_map>
+#include <unordered_set>
+#include <utility>
+#include <vector>
+
+#include <glad/glad.h>
+
+#include "common/assert.h"
+#include "common/common_types.h"
+#include "video_core/engines/maxwell_3d.h"
+#include "video_core/renderer_opengl/gl_shader_gen.h"
+
+namespace Core {
+class System;
+}
+
+namespace FileUtil {
+class IOFile;
+}
+
+namespace OpenGL {
+
+using ProgramCode = std::vector<u64>;
+using Maxwell = Tegra::Engines::Maxwell3D::Regs;
+
+/// Allocated bindings used by an OpenGL shader program
+struct BaseBindings {
+    u32 cbuf{};
+    u32 gmem{};
+    u32 sampler{};
+
+    bool operator==(const BaseBindings& rhs) const {
+        return std::tie(cbuf, gmem, sampler) == std::tie(rhs.cbuf, rhs.gmem, rhs.sampler);
+    }
+
+    bool operator!=(const BaseBindings& rhs) const {
+        return !operator==(rhs);
+    }
+};
+
+/// Describes how a shader is used
+struct ShaderDiskCacheUsage {
+    u64 unique_identifier{};
+    BaseBindings bindings;
+    GLenum primitive{};
+
+    bool operator==(const ShaderDiskCacheUsage& rhs) const {
+        return std::tie(unique_identifier, bindings, primitive) ==
+               std::tie(rhs.unique_identifier, rhs.bindings, rhs.primitive);
+    }
+
+    bool operator!=(const ShaderDiskCacheUsage& rhs) const {
+        return !operator==(rhs);
+    }
+};
+
+} // namespace OpenGL
+
+namespace std {
+
+template <>
+struct hash<OpenGL::BaseBindings> {
+    std::size_t operator()(const OpenGL::BaseBindings& bindings) const {
+        return bindings.cbuf | bindings.gmem << 8 | bindings.sampler << 16;
+    }
+};
+
+template <>
+struct hash<OpenGL::ShaderDiskCacheUsage> {
+    std::size_t operator()(const OpenGL::ShaderDiskCacheUsage& usage) const {
+        return static_cast<std::size_t>(usage.unique_identifier) ^
+               std::hash<OpenGL::BaseBindings>()(usage.bindings) ^ usage.primitive << 16;
+    }
+};
+
+} // namespace std
+
+namespace OpenGL {
+
+/// Describes a shader how it's used by the guest GPU
+class ShaderDiskCacheRaw {
+public:
+    explicit ShaderDiskCacheRaw(u64 unique_identifier, Maxwell::ShaderProgram program_type,
+                                u32 program_code_size, u32 program_code_size_b,
+                                ProgramCode program_code, ProgramCode program_code_b);
+    ShaderDiskCacheRaw();
+    ~ShaderDiskCacheRaw();
+
+    bool Load(FileUtil::IOFile& file);
+
+    bool Save(FileUtil::IOFile& file) const;
+
+    u64 GetUniqueIdentifier() const {
+        return unique_identifier;
+    }
+
+    bool HasProgramA() const {
+        return program_type == Maxwell::ShaderProgram::VertexA;
+    }
+
+    Maxwell::ShaderProgram GetProgramType() const {
+        return program_type;
+    }
+
+    Maxwell::ShaderStage GetProgramStage() const {
+        switch (program_type) {
+        case Maxwell::ShaderProgram::VertexA:
+        case Maxwell::ShaderProgram::VertexB:
+            return Maxwell::ShaderStage::Vertex;
+        case Maxwell::ShaderProgram::TesselationControl:
+            return Maxwell::ShaderStage::TesselationControl;
+        case Maxwell::ShaderProgram::TesselationEval:
+            return Maxwell::ShaderStage::TesselationEval;
+        case Maxwell::ShaderProgram::Geometry:
+            return Maxwell::ShaderStage::Geometry;
+        case Maxwell::ShaderProgram::Fragment:
+            return Maxwell::ShaderStage::Fragment;
+        }
+        UNREACHABLE();
+    }
+
+    const ProgramCode& GetProgramCode() const {
+        return program_code;
+    }
+
+    const ProgramCode& GetProgramCodeB() const {
+        return program_code_b;
+    }
+
+private:
+    u64 unique_identifier{};
+    Maxwell::ShaderProgram program_type{};
+    u32 program_code_size{};
+    u32 program_code_size_b{};
+
+    ProgramCode program_code;
+    ProgramCode program_code_b;
+};
+
+/// Contains decompiled data from a shader
+struct ShaderDiskCacheDecompiled {
+    std::string code;
+    GLShader::ShaderEntries entries;
+};
+
+/// Contains an OpenGL dumped binary program
+struct ShaderDiskCacheDump {
+    GLenum binary_format;
+    std::vector<u8> binary;
+};
+
+class ShaderDiskCacheOpenGL {
+public:
+    explicit ShaderDiskCacheOpenGL(Core::System& system);
+
+    /// Loads transferable cache. If file has a old version or on failure, it deletes the file.
+    std::optional<std::pair<std::vector<ShaderDiskCacheRaw>, std::vector<ShaderDiskCacheUsage>>>
+    LoadTransferable();
+
+    /// Loads current game's precompiled cache. Invalidates on failure.
+    std::pair<std::unordered_map<u64, ShaderDiskCacheDecompiled>,
+              std::unordered_map<ShaderDiskCacheUsage, ShaderDiskCacheDump>>
+    LoadPrecompiled();
+
+    /// Removes the transferable (and precompiled) cache file.
+    void InvalidateTransferable() const;
+
+    /// Removes the precompiled cache file.
+    void InvalidatePrecompiled() const;
+
+    /// Saves a raw dump to the transferable file. Checks for collisions.
+    void SaveRaw(const ShaderDiskCacheRaw& entry);
+
+    /// Saves shader usage to the transferable file. Does not check for collisions.
+    void SaveUsage(const ShaderDiskCacheUsage& usage);
+
+    /// Saves a decompiled entry to the precompiled file. Does not check for collisions.
+    void SaveDecompiled(u64 unique_identifier, const std::string& code,
+                        const GLShader::ShaderEntries& entries);
+
+    /// Saves a dump entry to the precompiled file. Does not check for collisions.
+    void SaveDump(const ShaderDiskCacheUsage& usage, GLuint program);
+
+private:
+    /// Loads the transferable cache. Returns empty on failure.
+    std::optional<std::pair<std::unordered_map<u64, ShaderDiskCacheDecompiled>,
+                            std::unordered_map<ShaderDiskCacheUsage, ShaderDiskCacheDump>>>
+    LoadPrecompiledFile(FileUtil::IOFile& file);
+
+    /// Loads a decompiled cache entry from the passed file. Returns empty on failure.
+    std::optional<ShaderDiskCacheDecompiled> LoadDecompiledEntry(FileUtil::IOFile& file);
+
+    /// Saves a decompiled entry to the passed file. Returns true on success.
+    bool SaveDecompiledFile(FileUtil::IOFile& file, u64 unique_identifier, const std::string& code,
+                            const std::vector<u8>& compressed_code,
+                            const GLShader::ShaderEntries& entries);
+
+    /// Returns if the cache can be used
+    bool IsUsable() const;
+
+    /// Opens current game's transferable file and write it's header if it doesn't exist
+    FileUtil::IOFile AppendTransferableFile() const;
+
+    /// Opens current game's precompiled file and write it's header if it doesn't exist
+    FileUtil::IOFile AppendPrecompiledFile() const;
+
+    /// Create shader disk cache directories. Returns true on success.
+    bool EnsureDirectories() const;
+
+    /// Gets current game's transferable file path
+    std::string GetTransferablePath() const;
+
+    /// Gets current game's precompiled file path
+    std::string GetPrecompiledPath() const;
+
+    /// Get user's transferable directory path
+    std::string GetTransferableDir() const;
+
+    /// Get user's precompiled directory path
+    std::string GetPrecompiledDir() const;
+
+    /// Get user's shader directory path
+    std::string GetBaseDir() const;
+
+    /// Get current game's title id
+    std::string GetTitleID() const;
+
+    // Copre system
+    Core::System& system;
+    // Stored transferable shaders
+    std::map<u64, std::unordered_set<ShaderDiskCacheUsage>> transferable;
+    // The cache has been loaded at boot
+    bool tried_to_load{};
+};
+
+} // namespace OpenGL
\ No newline at end of file
diff --git a/src/video_core/renderer_opengl/gl_shader_gen.cpp b/src/video_core/renderer_opengl/gl_shader_gen.cpp
index 04e1db911..7d96649af 100644
--- a/src/video_core/renderer_opengl/gl_shader_gen.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_gen.cpp
@@ -124,7 +124,7 @@ layout (location = 5) out vec4 FragColor5;
 layout (location = 6) out vec4 FragColor6;
 layout (location = 7) out vec4 FragColor7;
 
-layout (location = 0) in vec4 position;
+layout (location = 0) in noperspective vec4 position;
 
 layout (std140, binding = EMULATION_UBO_BINDING) uniform fs_config {
     vec4 viewport_flip;
@@ -172,4 +172,4 @@ void main() {
     return {out, program.second};
 }
 
-} // namespace OpenGL::GLShader
\ No newline at end of file
+} // namespace OpenGL::GLShader
diff --git a/src/video_core/renderer_opengl/gl_shader_gen.h b/src/video_core/renderer_opengl/gl_shader_gen.h
index ac5e6917b..fba8e681b 100644
--- a/src/video_core/renderer_opengl/gl_shader_gen.h
+++ b/src/video_core/renderer_opengl/gl_shader_gen.h
@@ -26,12 +26,10 @@ struct ShaderSetup {
         ProgramCode code;
         ProgramCode code_b; // Used for dual vertex shaders
         u64 unique_identifier;
-        std::size_t real_size;
-        std::size_t real_size_b;
     } program;
 
     /// Used in scenarios where we have a dual vertex shaders
-    void SetProgramB(ProgramCode&& program_b) {
+    void SetProgramB(ProgramCode program_b) {
         program.code_b = std::move(program_b);
         has_program_b = true;
     }
diff --git a/src/video_core/renderer_opengl/gl_shader_util.h b/src/video_core/renderer_opengl/gl_shader_util.h
index 285594f50..03b7548c2 100644
--- a/src/video_core/renderer_opengl/gl_shader_util.h
+++ b/src/video_core/renderer_opengl/gl_shader_util.h
@@ -47,7 +47,7 @@ GLuint LoadShader(const char* source, GLenum type);
  * @returns Handle of the newly created OpenGL program object
  */
 template <typename... T>
-GLuint LoadProgram(bool separable_program, T... shaders) {
+GLuint LoadProgram(bool separable_program, bool hint_retrievable, T... shaders) {
     // Link the program
     LOG_DEBUG(Render_OpenGL, "Linking program...");
 
@@ -58,6 +58,9 @@ GLuint LoadProgram(bool separable_program, T... shaders) {
     if (separable_program) {
         glProgramParameteri(program_id, GL_PROGRAM_SEPARABLE, GL_TRUE);
     }
+    if (hint_retrievable) {
+        glProgramParameteri(program_id, GL_PROGRAM_BINARY_RETRIEVABLE_HINT, GL_TRUE);
+    }
 
     glLinkProgram(program_id);
 
diff --git a/src/video_core/renderer_opengl/gl_state.cpp b/src/video_core/renderer_opengl/gl_state.cpp
index b7ba59350..9419326a3 100644
--- a/src/video_core/renderer_opengl/gl_state.cpp
+++ b/src/video_core/renderer_opengl/gl_state.cpp
@@ -11,7 +11,9 @@
 namespace OpenGL {
 
 OpenGLState OpenGLState::cur_state;
+
 bool OpenGLState::s_rgb_used;
+
 OpenGLState::OpenGLState() {
     // These all match default OpenGL values
     geometry_shaders.enabled = false;
@@ -112,7 +114,6 @@ void OpenGLState::ApplyDefaultState() {
 }
 
 void OpenGLState::ApplySRgb() const {
-    // sRGB
     if (framebuffer_srgb.enabled != cur_state.framebuffer_srgb.enabled) {
         if (framebuffer_srgb.enabled) {
             // Track if sRGB is used
@@ -125,23 +126,20 @@ void OpenGLState::ApplySRgb() const {
 }
 
 void OpenGLState::ApplyCulling() const {
-    // Culling
-    const bool cull_changed = cull.enabled != cur_state.cull.enabled;
-    if (cull_changed) {
+    if (cull.enabled != cur_state.cull.enabled) {
         if (cull.enabled) {
             glEnable(GL_CULL_FACE);
         } else {
             glDisable(GL_CULL_FACE);
         }
     }
-    if (cull.enabled) {
-        if (cull_changed || cull.mode != cur_state.cull.mode) {
-            glCullFace(cull.mode);
-        }
 
-        if (cull_changed || cull.front_face != cur_state.cull.front_face) {
-            glFrontFace(cull.front_face);
-        }
+    if (cull.mode != cur_state.cull.mode) {
+        glCullFace(cull.mode);
+    }
+
+    if (cull.front_face != cur_state.cull.front_face) {
+        glFrontFace(cull.front_face);
     }
 }
 
@@ -172,72 +170,63 @@ void OpenGLState::ApplyColorMask() const {
 }
 
 void OpenGLState::ApplyDepth() const {
-    // Depth test
-    const bool depth_test_changed = depth.test_enabled != cur_state.depth.test_enabled;
-    if (depth_test_changed) {
+    if (depth.test_enabled != cur_state.depth.test_enabled) {
         if (depth.test_enabled) {
             glEnable(GL_DEPTH_TEST);
         } else {
             glDisable(GL_DEPTH_TEST);
         }
     }
-    if (depth.test_enabled &&
-        (depth_test_changed || depth.test_func != cur_state.depth.test_func)) {
+
+    if (depth.test_func != cur_state.depth.test_func) {
         glDepthFunc(depth.test_func);
     }
-    // Depth mask
+
     if (depth.write_mask != cur_state.depth.write_mask) {
         glDepthMask(depth.write_mask);
     }
 }
 
 void OpenGLState::ApplyPrimitiveRestart() const {
-    const bool primitive_restart_changed =
-        primitive_restart.enabled != cur_state.primitive_restart.enabled;
-    if (primitive_restart_changed) {
+    if (primitive_restart.enabled != cur_state.primitive_restart.enabled) {
         if (primitive_restart.enabled) {
             glEnable(GL_PRIMITIVE_RESTART);
         } else {
             glDisable(GL_PRIMITIVE_RESTART);
         }
     }
-    if (primitive_restart_changed ||
-        (primitive_restart.enabled &&
-         primitive_restart.index != cur_state.primitive_restart.index)) {
+
+    if (primitive_restart.index != cur_state.primitive_restart.index) {
         glPrimitiveRestartIndex(primitive_restart.index);
     }
 }
 
 void OpenGLState::ApplyStencilTest() const {
-    const bool stencil_test_changed = stencil.test_enabled != cur_state.stencil.test_enabled;
-    if (stencil_test_changed) {
+    if (stencil.test_enabled != cur_state.stencil.test_enabled) {
         if (stencil.test_enabled) {
             glEnable(GL_STENCIL_TEST);
         } else {
             glDisable(GL_STENCIL_TEST);
         }
     }
-    if (stencil.test_enabled) {
-        auto config_stencil = [stencil_test_changed](GLenum face, const auto& config,
-                                                     const auto& prev_config) {
-            if (stencil_test_changed || config.test_func != prev_config.test_func ||
-                config.test_ref != prev_config.test_ref ||
-                config.test_mask != prev_config.test_mask) {
-                glStencilFuncSeparate(face, config.test_func, config.test_ref, config.test_mask);
-            }
-            if (stencil_test_changed || config.action_depth_fail != prev_config.action_depth_fail ||
-                config.action_depth_pass != prev_config.action_depth_pass ||
-                config.action_stencil_fail != prev_config.action_stencil_fail) {
-                glStencilOpSeparate(face, config.action_stencil_fail, config.action_depth_fail,
-                                    config.action_depth_pass);
-            }
-            if (config.write_mask != prev_config.write_mask) {
-                glStencilMaskSeparate(face, config.write_mask);
-            }
-        };
-        config_stencil(GL_FRONT, stencil.front, cur_state.stencil.front);
-        config_stencil(GL_BACK, stencil.back, cur_state.stencil.back);
-    }
+
+    const auto ConfigStencil = [](GLenum face, const auto& config, const auto& prev_config) {
+        if (config.test_func != prev_config.test_func || config.test_ref != prev_config.test_ref ||
+            config.test_mask != prev_config.test_mask) {
+            glStencilFuncSeparate(face, config.test_func, config.test_ref, config.test_mask);
+        }
+        if (config.action_depth_fail != prev_config.action_depth_fail ||
+            config.action_depth_pass != prev_config.action_depth_pass ||
+            config.action_stencil_fail != prev_config.action_stencil_fail) {
+            glStencilOpSeparate(face, config.action_stencil_fail, config.action_depth_fail,
+                                config.action_depth_pass);
+        }
+        if (config.write_mask != prev_config.write_mask) {
+            glStencilMaskSeparate(face, config.write_mask);
+        }
+    };
+    ConfigStencil(GL_FRONT, stencil.front, cur_state.stencil.front);
+    ConfigStencil(GL_BACK, stencil.back, cur_state.stencil.back);
 }
 // Viewport does not affects glClearBuffer so emulate viewport using scissor test
 void OpenGLState::EmulateViewportWithScissor() {
@@ -278,19 +267,18 @@ void OpenGLState::ApplyViewport() const {
                 updated.depth_range_far != current.depth_range_far) {
                 glDepthRangeIndexed(i, updated.depth_range_near, updated.depth_range_far);
             }
-            const bool scissor_changed = updated.scissor.enabled != current.scissor.enabled;
-            if (scissor_changed) {
+
+            if (updated.scissor.enabled != current.scissor.enabled) {
                 if (updated.scissor.enabled) {
                     glEnablei(GL_SCISSOR_TEST, i);
                 } else {
                     glDisablei(GL_SCISSOR_TEST, i);
                 }
             }
-            if (updated.scissor.enabled &&
-                (scissor_changed || updated.scissor.x != current.scissor.x ||
-                 updated.scissor.y != current.scissor.y ||
-                 updated.scissor.width != current.scissor.width ||
-                 updated.scissor.height != current.scissor.height)) {
+
+            if (updated.scissor.x != current.scissor.x || updated.scissor.y != current.scissor.y ||
+                updated.scissor.width != current.scissor.width ||
+                updated.scissor.height != current.scissor.height) {
                 glScissorIndexed(i, updated.scissor.x, updated.scissor.y, updated.scissor.width,
                                  updated.scissor.height);
             }
@@ -302,22 +290,23 @@ void OpenGLState::ApplyViewport() const {
             updated.height != current.height) {
             glViewport(updated.x, updated.y, updated.width, updated.height);
         }
+
         if (updated.depth_range_near != current.depth_range_near ||
             updated.depth_range_far != current.depth_range_far) {
             glDepthRange(updated.depth_range_near, updated.depth_range_far);
         }
-        const bool scissor_changed = updated.scissor.enabled != current.scissor.enabled;
-        if (scissor_changed) {
+
+        if (updated.scissor.enabled != current.scissor.enabled) {
             if (updated.scissor.enabled) {
                 glEnable(GL_SCISSOR_TEST);
             } else {
                 glDisable(GL_SCISSOR_TEST);
             }
         }
-        if (updated.scissor.enabled && (scissor_changed || updated.scissor.x != current.scissor.x ||
-                                        updated.scissor.y != current.scissor.y ||
-                                        updated.scissor.width != current.scissor.width ||
-                                        updated.scissor.height != current.scissor.height)) {
+
+        if (updated.scissor.x != current.scissor.x || updated.scissor.y != current.scissor.y ||
+            updated.scissor.width != current.scissor.width ||
+            updated.scissor.height != current.scissor.height) {
             glScissor(updated.scissor.x, updated.scissor.y, updated.scissor.width,
                       updated.scissor.height);
         }
@@ -327,8 +316,7 @@ void OpenGLState::ApplyViewport() const {
 void OpenGLState::ApplyGlobalBlending() const {
     const Blend& current = cur_state.blend[0];
     const Blend& updated = blend[0];
-    const bool blend_changed = updated.enabled != current.enabled;
-    if (blend_changed) {
+    if (updated.enabled != current.enabled) {
         if (updated.enabled) {
             glEnable(GL_BLEND);
         } else {
@@ -338,15 +326,14 @@ void OpenGLState::ApplyGlobalBlending() const {
     if (!updated.enabled) {
         return;
     }
-    if (blend_changed || updated.src_rgb_func != current.src_rgb_func ||
+    if (updated.src_rgb_func != current.src_rgb_func ||
         updated.dst_rgb_func != current.dst_rgb_func || updated.src_a_func != current.src_a_func ||
         updated.dst_a_func != current.dst_a_func) {
         glBlendFuncSeparate(updated.src_rgb_func, updated.dst_rgb_func, updated.src_a_func,
                             updated.dst_a_func);
     }
 
-    if (blend_changed || updated.rgb_equation != current.rgb_equation ||
-        updated.a_equation != current.a_equation) {
+    if (updated.rgb_equation != current.rgb_equation || updated.a_equation != current.a_equation) {
         glBlendEquationSeparate(updated.rgb_equation, updated.a_equation);
     }
 }
@@ -354,26 +341,22 @@ void OpenGLState::ApplyGlobalBlending() const {
 void OpenGLState::ApplyTargetBlending(std::size_t target, bool force) const {
     const Blend& updated = blend[target];
     const Blend& current = cur_state.blend[target];
-    const bool blend_changed = updated.enabled != current.enabled || force;
-    if (blend_changed) {
+    if (updated.enabled != current.enabled || force) {
         if (updated.enabled) {
             glEnablei(GL_BLEND, static_cast<GLuint>(target));
         } else {
             glDisablei(GL_BLEND, static_cast<GLuint>(target));
         }
     }
-    if (!updated.enabled) {
-        return;
-    }
-    if (blend_changed || updated.src_rgb_func != current.src_rgb_func ||
+
+    if (updated.src_rgb_func != current.src_rgb_func ||
         updated.dst_rgb_func != current.dst_rgb_func || updated.src_a_func != current.src_a_func ||
         updated.dst_a_func != current.dst_a_func) {
         glBlendFuncSeparatei(static_cast<GLuint>(target), updated.src_rgb_func,
                              updated.dst_rgb_func, updated.src_a_func, updated.dst_a_func);
     }
 
-    if (blend_changed || updated.rgb_equation != current.rgb_equation ||
-        updated.a_equation != current.a_equation) {
+    if (updated.rgb_equation != current.rgb_equation || updated.a_equation != current.a_equation) {
         glBlendEquationSeparatei(static_cast<GLuint>(target), updated.rgb_equation,
                                  updated.a_equation);
     }
@@ -397,8 +380,7 @@ void OpenGLState::ApplyBlending() const {
 }
 
 void OpenGLState::ApplyLogicOp() const {
-    const bool logic_op_changed = logic_op.enabled != cur_state.logic_op.enabled;
-    if (logic_op_changed) {
+    if (logic_op.enabled != cur_state.logic_op.enabled) {
         if (logic_op.enabled) {
             glEnable(GL_COLOR_LOGIC_OP);
         } else {
@@ -406,14 +388,12 @@ void OpenGLState::ApplyLogicOp() const {
         }
     }
 
-    if (logic_op.enabled &&
-        (logic_op_changed || logic_op.operation != cur_state.logic_op.operation)) {
+    if (logic_op.operation != cur_state.logic_op.operation) {
         glLogicOp(logic_op.operation);
     }
 }
 
 void OpenGLState::ApplyPolygonOffset() const {
-
     const bool fill_enable_changed =
         polygon_offset.fill_enable != cur_state.polygon_offset.fill_enable;
     const bool line_enable_changed =
@@ -448,9 +428,7 @@ void OpenGLState::ApplyPolygonOffset() const {
         }
     }
 
-    if ((polygon_offset.fill_enable || polygon_offset.line_enable || polygon_offset.point_enable) &&
-        (factor_changed || units_changed || clamp_changed)) {
-
+    if (factor_changed || units_changed || clamp_changed) {
         if (GLAD_GL_EXT_polygon_offset_clamp && polygon_offset.clamp != 0) {
             glPolygonOffsetClamp(polygon_offset.factor, polygon_offset.units, polygon_offset.clamp);
         } else {
@@ -462,29 +440,35 @@ void OpenGLState::ApplyPolygonOffset() const {
 }
 
 void OpenGLState::ApplyTextures() const {
+    bool has_delta{};
+    std::size_t first{};
+    std::size_t last{};
+    std::array<GLuint, Tegra::Engines::Maxwell3D::Regs::NumTextureSamplers> textures;
+
     for (std::size_t i = 0; i < std::size(texture_units); ++i) {
         const auto& texture_unit = texture_units[i];
         const auto& cur_state_texture_unit = cur_state.texture_units[i];
+        textures[i] = texture_unit.texture;
 
-        if (texture_unit.texture != cur_state_texture_unit.texture) {
-            glActiveTexture(TextureUnits::MaxwellTexture(static_cast<int>(i)).Enum());
-            glBindTexture(texture_unit.target, texture_unit.texture);
-        }
-        // Update the texture swizzle
-        if (texture_unit.swizzle.r != cur_state_texture_unit.swizzle.r ||
-            texture_unit.swizzle.g != cur_state_texture_unit.swizzle.g ||
-            texture_unit.swizzle.b != cur_state_texture_unit.swizzle.b ||
-            texture_unit.swizzle.a != cur_state_texture_unit.swizzle.a) {
-            std::array<GLint, 4> mask = {texture_unit.swizzle.r, texture_unit.swizzle.g,
-                                         texture_unit.swizzle.b, texture_unit.swizzle.a};
-            glTexParameteriv(texture_unit.target, GL_TEXTURE_SWIZZLE_RGBA, mask.data());
+        if (textures[i] != cur_state_texture_unit.texture) {
+            if (!has_delta) {
+                first = i;
+                has_delta = true;
+            }
+            last = i;
         }
     }
+
+    if (has_delta) {
+        glBindTextures(static_cast<GLuint>(first), static_cast<GLsizei>(last - first + 1),
+                       textures.data() + first);
+    }
 }
 
 void OpenGLState::ApplySamplers() const {
     bool has_delta{};
-    std::size_t first{}, last{};
+    std::size_t first{};
+    std::size_t last{};
     std::array<GLuint, Tegra::Engines::Maxwell3D::Regs::NumTextureSamplers> samplers;
     for (std::size_t i = 0; i < std::size(samplers); ++i) {
         samplers[i] = texture_units[i].sampler;
@@ -498,7 +482,7 @@ void OpenGLState::ApplySamplers() const {
     }
     if (has_delta) {
         glBindSamplers(static_cast<GLuint>(first), static_cast<GLsizei>(last - first + 1),
-                       samplers.data());
+                       samplers.data() + first);
     }
 }
 
@@ -522,9 +506,9 @@ void OpenGLState::ApplyDepthClamp() const {
         depth_clamp.near_plane == cur_state.depth_clamp.near_plane) {
         return;
     }
-    if (depth_clamp.far_plane != depth_clamp.near_plane) {
-        UNIMPLEMENTED_MSG("Unimplemented Depth Clamp Separation!");
-    }
+    UNIMPLEMENTED_IF_MSG(depth_clamp.far_plane != depth_clamp.near_plane,
+                         "Unimplemented Depth Clamp Separation!");
+
     if (depth_clamp.far_plane || depth_clamp.near_plane) {
         glEnable(GL_DEPTH_CLAMP);
     } else {
diff --git a/src/video_core/renderer_opengl/gl_state.h b/src/video_core/renderer_opengl/gl_state.h
index a5a7c0920..9e1eda5b1 100644
--- a/src/video_core/renderer_opengl/gl_state.h
+++ b/src/video_core/renderer_opengl/gl_state.h
@@ -126,26 +126,14 @@ public:
     struct TextureUnit {
         GLuint texture; // GL_TEXTURE_BINDING_2D
         GLuint sampler; // GL_SAMPLER_BINDING
-        GLenum target;
-        struct {
-            GLint r; // GL_TEXTURE_SWIZZLE_R
-            GLint g; // GL_TEXTURE_SWIZZLE_G
-            GLint b; // GL_TEXTURE_SWIZZLE_B
-            GLint a; // GL_TEXTURE_SWIZZLE_A
-        } swizzle;
 
         void Unbind() {
             texture = 0;
-            swizzle.r = GL_RED;
-            swizzle.g = GL_GREEN;
-            swizzle.b = GL_BLUE;
-            swizzle.a = GL_ALPHA;
         }
 
         void Reset() {
             Unbind();
             sampler = 0;
-            target = GL_TEXTURE_2D;
         }
     };
     std::array<TextureUnit, Tegra::Engines::Maxwell3D::Regs::NumTextureSamplers> texture_units;
diff --git a/src/video_core/renderer_opengl/renderer_opengl.cpp b/src/video_core/renderer_opengl/renderer_opengl.cpp
index e37b65b38..5e3d862c6 100644
--- a/src/video_core/renderer_opengl/renderer_opengl.cpp
+++ b/src/video_core/renderer_opengl/renderer_opengl.cpp
@@ -98,8 +98,8 @@ static std::array<GLfloat, 3 * 2> MakeOrthographicMatrix(const float width, cons
     return matrix;
 }
 
-RendererOpenGL::RendererOpenGL(Core::Frontend::EmuWindow& window)
-    : VideoCore::RendererBase{window} {}
+RendererOpenGL::RendererOpenGL(Core::Frontend::EmuWindow& window, Core::System& system)
+    : VideoCore::RendererBase{window}, system{system} {}
 
 RendererOpenGL::~RendererOpenGL() = default;
 
@@ -107,7 +107,7 @@ RendererOpenGL::~RendererOpenGL() = default;
 void RendererOpenGL::SwapBuffers(
     std::optional<std::reference_wrapper<const Tegra::FramebufferConfig>> framebuffer) {
 
-    Core::System::GetInstance().GetPerfStats().EndSystemFrame();
+    system.GetPerfStats().EndSystemFrame();
 
     // Maintain the rasterizer's state as a priority
     OpenGLState prev_state = OpenGLState::GetCurState();
@@ -137,8 +137,8 @@ void RendererOpenGL::SwapBuffers(
 
     render_window.PollEvents();
 
-    Core::System::GetInstance().FrameLimiter().DoFrameLimiting(CoreTiming::GetGlobalTimeUs());
-    Core::System::GetInstance().GetPerfStats().BeginSystemFrame();
+    system.FrameLimiter().DoFrameLimiting(system.CoreTiming().GetGlobalTimeUs());
+    system.GetPerfStats().BeginSystemFrame();
 
     // Restore the rasterizer state
     prev_state.Apply();
@@ -164,17 +164,14 @@ void RendererOpenGL::LoadFBToScreenInfo(const Tegra::FramebufferConfig& framebuf
         // Reset the screen info's display texture to its own permanent texture
         screen_info.display_texture = screen_info.texture.resource.handle;
 
-        Memory::RasterizerFlushVirtualRegion(framebuffer_addr, size_in_bytes,
-                                             Memory::FlushMode::Flush);
+        rasterizer->FlushRegion(ToCacheAddr(Memory::GetPointer(framebuffer_addr)), size_in_bytes);
 
-        VideoCore::MortonCopyPixels128(framebuffer.width, framebuffer.height, bytes_per_pixel, 4,
-                                       Memory::GetPointer(framebuffer_addr),
-                                       gl_framebuffer_data.data(), true);
+        constexpr u32 linear_bpp = 4;
+        VideoCore::MortonCopyPixels128(VideoCore::MortonSwizzleMode::MortonToLinear,
+                                       framebuffer.width, framebuffer.height, bytes_per_pixel,
+                                       linear_bpp, Memory::GetPointer(framebuffer_addr),
+                                       gl_framebuffer_data.data());
 
-        state.texture_units[0].texture = screen_info.texture.resource.handle;
-        state.Apply();
-
-        glActiveTexture(GL_TEXTURE0);
         glPixelStorei(GL_UNPACK_ROW_LENGTH, static_cast<GLint>(framebuffer.stride));
 
         // Update existing texture
@@ -182,14 +179,11 @@ void RendererOpenGL::LoadFBToScreenInfo(const Tegra::FramebufferConfig& framebuf
         //       they differ from the LCD resolution.
         // TODO: Applications could theoretically crash yuzu here by specifying too large
         //       framebuffer sizes. We should make sure that this cannot happen.
-        glTexSubImage2D(GL_TEXTURE_2D, 0, 0, 0, framebuffer.width, framebuffer.height,
-                        screen_info.texture.gl_format, screen_info.texture.gl_type,
-                        gl_framebuffer_data.data());
+        glTextureSubImage2D(screen_info.texture.resource.handle, 0, 0, 0, framebuffer.width,
+                            framebuffer.height, screen_info.texture.gl_format,
+                            screen_info.texture.gl_type, gl_framebuffer_data.data());
 
         glPixelStorei(GL_UNPACK_ROW_LENGTH, 0);
-
-        state.texture_units[0].texture = 0;
-        state.Apply();
     }
 }
 
@@ -199,17 +193,8 @@ void RendererOpenGL::LoadFBToScreenInfo(const Tegra::FramebufferConfig& framebuf
  */
 void RendererOpenGL::LoadColorToActiveGLTexture(u8 color_r, u8 color_g, u8 color_b, u8 color_a,
                                                 const TextureInfo& texture) {
-    state.texture_units[0].texture = texture.resource.handle;
-    state.Apply();
-
-    glActiveTexture(GL_TEXTURE0);
-    u8 framebuffer_data[4] = {color_a, color_b, color_g, color_r};
-
-    // Update existing texture
-    glTexImage2D(GL_TEXTURE_2D, 0, GL_RGBA, 1, 1, 0, GL_RGBA, GL_UNSIGNED_BYTE, framebuffer_data);
-
-    state.texture_units[0].texture = 0;
-    state.Apply();
+    const u8 framebuffer_data[4] = {color_a, color_b, color_g, color_r};
+    glClearTexImage(texture.resource.handle, 0, GL_RGBA, GL_UNSIGNED_BYTE, framebuffer_data);
 }
 
 /**
@@ -249,55 +234,57 @@ void RendererOpenGL::InitOpenGLObjects() {
                               sizeof(ScreenRectVertex));
 
     // Allocate textures for the screen
-    screen_info.texture.resource.Create();
+    screen_info.texture.resource.Create(GL_TEXTURE_2D);
 
-    // Allocation of storage is deferred until the first frame, when we
-    // know the framebuffer size.
-
-    state.texture_units[0].texture = screen_info.texture.resource.handle;
-    state.Apply();
-
-    glActiveTexture(GL_TEXTURE0);
-    glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAX_LEVEL, 0);
-    glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
-    glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
-    glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
-    glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
+    const GLuint texture = screen_info.texture.resource.handle;
+    glTextureStorage2D(texture, 1, GL_RGBA8, 1, 1);
 
     screen_info.display_texture = screen_info.texture.resource.handle;
 
-    state.texture_units[0].texture = 0;
-    state.Apply();
-
     // Clear screen to black
     LoadColorToActiveGLTexture(0, 0, 0, 0, screen_info.texture);
 }
 
+void RendererOpenGL::AddTelemetryFields() {
+    const char* const gl_version{reinterpret_cast<char const*>(glGetString(GL_VERSION))};
+    const char* const gpu_vendor{reinterpret_cast<char const*>(glGetString(GL_VENDOR))};
+    const char* const gpu_model{reinterpret_cast<char const*>(glGetString(GL_RENDERER))};
+
+    LOG_INFO(Render_OpenGL, "GL_VERSION: {}", gl_version);
+    LOG_INFO(Render_OpenGL, "GL_VENDOR: {}", gpu_vendor);
+    LOG_INFO(Render_OpenGL, "GL_RENDERER: {}", gpu_model);
+
+    auto& telemetry_session = system.TelemetrySession();
+    telemetry_session.AddField(Telemetry::FieldType::UserSystem, "GPU_Vendor", gpu_vendor);
+    telemetry_session.AddField(Telemetry::FieldType::UserSystem, "GPU_Model", gpu_model);
+    telemetry_session.AddField(Telemetry::FieldType::UserSystem, "GPU_OpenGL_Version", gl_version);
+}
+
 void RendererOpenGL::CreateRasterizer() {
     if (rasterizer) {
         return;
     }
     // Initialize sRGB Usage
     OpenGLState::ClearsRGBUsed();
-    rasterizer = std::make_unique<RasterizerOpenGL>(render_window, screen_info);
+    rasterizer = std::make_unique<RasterizerOpenGL>(render_window, system, screen_info);
 }
 
 void RendererOpenGL::ConfigureFramebufferTexture(TextureInfo& texture,
                                                  const Tegra::FramebufferConfig& framebuffer) {
-
     texture.width = framebuffer.width;
     texture.height = framebuffer.height;
+    texture.pixel_format = framebuffer.pixel_format;
 
     GLint internal_format;
     switch (framebuffer.pixel_format) {
     case Tegra::FramebufferConfig::PixelFormat::ABGR8:
-        internal_format = GL_RGBA;
+        internal_format = GL_RGBA8;
         texture.gl_format = GL_RGBA;
         texture.gl_type = GL_UNSIGNED_INT_8_8_8_8_REV;
         gl_framebuffer_data.resize(texture.width * texture.height * 4);
         break;
     default:
-        internal_format = GL_RGBA;
+        internal_format = GL_RGBA8;
         texture.gl_format = GL_RGBA;
         texture.gl_type = GL_UNSIGNED_INT_8_8_8_8_REV;
         gl_framebuffer_data.resize(texture.width * texture.height * 4);
@@ -306,15 +293,9 @@ void RendererOpenGL::ConfigureFramebufferTexture(TextureInfo& texture,
         UNREACHABLE();
     }
 
-    state.texture_units[0].texture = texture.resource.handle;
-    state.Apply();
-
-    glActiveTexture(GL_TEXTURE0);
-    glTexImage2D(GL_TEXTURE_2D, 0, internal_format, texture.width, texture.height, 0,
-                 texture.gl_format, texture.gl_type, nullptr);
-
-    state.texture_units[0].texture = 0;
-    state.Apply();
+    texture.resource.Release();
+    texture.resource.Create(GL_TEXTURE_2D);
+    glTextureStorage2D(texture.resource.handle, 1, internal_format, texture.width, texture.height);
 }
 
 void RendererOpenGL::DrawScreenTriangles(const ScreenInfo& screen_info, float x, float y, float w,
@@ -356,7 +337,6 @@ 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
     state.framebuffer_srgb.enabled = OpenGLState::GetsRGBUsed();
@@ -417,7 +397,8 @@ void RendererOpenGL::CaptureScreenshot() {
     GLuint renderbuffer;
     glGenRenderbuffers(1, &renderbuffer);
     glBindRenderbuffer(GL_RENDERBUFFER, renderbuffer);
-    glRenderbufferStorage(GL_RENDERBUFFER, GL_RGB8, layout.width, layout.height);
+    glRenderbufferStorage(GL_RENDERBUFFER, state.GetsRGBUsed() ? GL_SRGB8 : GL_RGB8, layout.width,
+                          layout.height);
     glFramebufferRenderbuffer(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_RENDERBUFFER, renderbuffer);
 
     DrawScreen(layout);
@@ -501,17 +482,7 @@ bool RendererOpenGL::Init() {
         glDebugMessageCallback(DebugHandler, nullptr);
     }
 
-    const char* gl_version{reinterpret_cast<char const*>(glGetString(GL_VERSION))};
-    const char* gpu_vendor{reinterpret_cast<char const*>(glGetString(GL_VENDOR))};
-    const char* gpu_model{reinterpret_cast<char const*>(glGetString(GL_RENDERER))};
-
-    LOG_INFO(Render_OpenGL, "GL_VERSION: {}", gl_version);
-    LOG_INFO(Render_OpenGL, "GL_VENDOR: {}", gpu_vendor);
-    LOG_INFO(Render_OpenGL, "GL_RENDERER: {}", gpu_model);
-
-    Core::Telemetry().AddField(Telemetry::FieldType::UserSystem, "GPU_Vendor", gpu_vendor);
-    Core::Telemetry().AddField(Telemetry::FieldType::UserSystem, "GPU_Model", gpu_model);
-    Core::Telemetry().AddField(Telemetry::FieldType::UserSystem, "GPU_OpenGL_Version", gl_version);
+    AddTelemetryFields();
 
     if (!GLAD_GL_VERSION_4_3) {
         return false;
diff --git a/src/video_core/renderer_opengl/renderer_opengl.h b/src/video_core/renderer_opengl/renderer_opengl.h
index 1665018db..6cbf9d2cb 100644
--- a/src/video_core/renderer_opengl/renderer_opengl.h
+++ b/src/video_core/renderer_opengl/renderer_opengl.h
@@ -12,6 +12,10 @@
 #include "video_core/renderer_opengl/gl_resource_manager.h"
 #include "video_core/renderer_opengl/gl_state.h"
 
+namespace Core {
+class System;
+}
+
 namespace Core::Frontend {
 class EmuWindow;
 }
@@ -35,13 +39,13 @@ struct TextureInfo {
 /// Structure used for storing information about the display target for the Switch screen
 struct ScreenInfo {
     GLuint display_texture;
-    const MathUtil::Rectangle<float> display_texcoords{0.0f, 0.0f, 1.0f, 1.0f};
+    const Common::Rectangle<float> display_texcoords{0.0f, 0.0f, 1.0f, 1.0f};
     TextureInfo texture;
 };
 
 class RendererOpenGL : public VideoCore::RendererBase {
 public:
-    explicit RendererOpenGL(Core::Frontend::EmuWindow& window);
+    explicit RendererOpenGL(Core::Frontend::EmuWindow& window, Core::System& system);
     ~RendererOpenGL() override;
 
     /// Swap buffers (render frame)
@@ -56,6 +60,7 @@ public:
 
 private:
     void InitOpenGLObjects();
+    void AddTelemetryFields();
     void CreateRasterizer();
 
     void ConfigureFramebufferTexture(TextureInfo& texture,
@@ -72,6 +77,8 @@ private:
     void LoadColorToActiveGLTexture(u8 color_r, u8 color_g, u8 color_b, u8 color_a,
                                     const TextureInfo& texture);
 
+    Core::System& system;
+
     OpenGLState state;
 
     // OpenGL object IDs
@@ -96,7 +103,7 @@ private:
 
     /// Used for transforming the framebuffer orientation
     Tegra::FramebufferConfig::TransformFlags framebuffer_transform_flags;
-    MathUtil::Rectangle<int> framebuffer_crop_rect;
+    Common::Rectangle<int> framebuffer_crop_rect;
 };
 
 } // namespace OpenGL
diff --git a/src/video_core/renderer_vulkan/declarations.h b/src/video_core/renderer_vulkan/declarations.h
new file mode 100644
index 000000000..ba25b5bc7
--- /dev/null
+++ b/src/video_core/renderer_vulkan/declarations.h
@@ -0,0 +1,45 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <vulkan/vulkan.hpp>
+
+namespace Vulkan {
+
+// vulkan.hpp unique handlers use DispatchLoaderStatic
+template <typename T>
+using UniqueHandle = vk::UniqueHandle<T, vk::DispatchLoaderDynamic>;
+
+using UniqueAccelerationStructureNV = UniqueHandle<vk::AccelerationStructureNV>;
+using UniqueBuffer = UniqueHandle<vk::Buffer>;
+using UniqueBufferView = UniqueHandle<vk::BufferView>;
+using UniqueCommandBuffer = UniqueHandle<vk::CommandBuffer>;
+using UniqueCommandPool = UniqueHandle<vk::CommandPool>;
+using UniqueDescriptorPool = UniqueHandle<vk::DescriptorPool>;
+using UniqueDescriptorSet = UniqueHandle<vk::DescriptorSet>;
+using UniqueDescriptorSetLayout = UniqueHandle<vk::DescriptorSetLayout>;
+using UniqueDescriptorUpdateTemplate = UniqueHandle<vk::DescriptorUpdateTemplate>;
+using UniqueDevice = UniqueHandle<vk::Device>;
+using UniqueDeviceMemory = UniqueHandle<vk::DeviceMemory>;
+using UniqueEvent = UniqueHandle<vk::Event>;
+using UniqueFence = UniqueHandle<vk::Fence>;
+using UniqueFramebuffer = UniqueHandle<vk::Framebuffer>;
+using UniqueImage = UniqueHandle<vk::Image>;
+using UniqueImageView = UniqueHandle<vk::ImageView>;
+using UniqueIndirectCommandsLayoutNVX = UniqueHandle<vk::IndirectCommandsLayoutNVX>;
+using UniqueObjectTableNVX = UniqueHandle<vk::ObjectTableNVX>;
+using UniquePipeline = UniqueHandle<vk::Pipeline>;
+using UniquePipelineCache = UniqueHandle<vk::PipelineCache>;
+using UniquePipelineLayout = UniqueHandle<vk::PipelineLayout>;
+using UniqueQueryPool = UniqueHandle<vk::QueryPool>;
+using UniqueRenderPass = UniqueHandle<vk::RenderPass>;
+using UniqueSampler = UniqueHandle<vk::Sampler>;
+using UniqueSamplerYcbcrConversion = UniqueHandle<vk::SamplerYcbcrConversion>;
+using UniqueSemaphore = UniqueHandle<vk::Semaphore>;
+using UniqueShaderModule = UniqueHandle<vk::ShaderModule>;
+using UniqueSwapchainKHR = UniqueHandle<vk::SwapchainKHR>;
+using UniqueValidationCacheEXT = UniqueHandle<vk::ValidationCacheEXT>;
+
+} // namespace Vulkan
diff --git a/src/video_core/renderer_vulkan/maxwell_to_vk.cpp b/src/video_core/renderer_vulkan/maxwell_to_vk.cpp
new file mode 100644
index 000000000..34bf26ff2
--- /dev/null
+++ b/src/video_core/renderer_vulkan/maxwell_to_vk.cpp
@@ -0,0 +1,483 @@
+// Copyright 2019 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 "common/logging/log.h"
+#include "video_core/engines/maxwell_3d.h"
+#include "video_core/renderer_vulkan/declarations.h"
+#include "video_core/renderer_vulkan/maxwell_to_vk.h"
+#include "video_core/renderer_vulkan/vk_device.h"
+#include "video_core/surface.h"
+
+namespace Vulkan::MaxwellToVK {
+
+namespace Sampler {
+
+vk::Filter Filter(Tegra::Texture::TextureFilter filter) {
+    switch (filter) {
+    case Tegra::Texture::TextureFilter::Linear:
+        return vk::Filter::eLinear;
+    case Tegra::Texture::TextureFilter::Nearest:
+        return vk::Filter::eNearest;
+    }
+    UNIMPLEMENTED_MSG("Unimplemented sampler filter={}", static_cast<u32>(filter));
+    return {};
+}
+
+vk::SamplerMipmapMode MipmapMode(Tegra::Texture::TextureMipmapFilter mipmap_filter) {
+    switch (mipmap_filter) {
+    case Tegra::Texture::TextureMipmapFilter::None:
+        // TODO(Rodrigo): None seems to be mapped to OpenGL's mag and min filters without mipmapping
+        // (e.g. GL_NEAREST and GL_LINEAR). Vulkan doesn't have such a thing, find out if we have to
+        // use an image view with a single mipmap level to emulate this.
+        return vk::SamplerMipmapMode::eLinear;
+    case Tegra::Texture::TextureMipmapFilter::Linear:
+        return vk::SamplerMipmapMode::eLinear;
+    case Tegra::Texture::TextureMipmapFilter::Nearest:
+        return vk::SamplerMipmapMode::eNearest;
+    }
+    UNIMPLEMENTED_MSG("Unimplemented sampler mipmap mode={}", static_cast<u32>(mipmap_filter));
+    return {};
+}
+
+vk::SamplerAddressMode WrapMode(Tegra::Texture::WrapMode wrap_mode) {
+    switch (wrap_mode) {
+    case Tegra::Texture::WrapMode::Wrap:
+        return vk::SamplerAddressMode::eRepeat;
+    case Tegra::Texture::WrapMode::Mirror:
+        return vk::SamplerAddressMode::eMirroredRepeat;
+    case Tegra::Texture::WrapMode::ClampToEdge:
+        return vk::SamplerAddressMode::eClampToEdge;
+    case Tegra::Texture::WrapMode::Border:
+        return vk::SamplerAddressMode::eClampToBorder;
+    case Tegra::Texture::WrapMode::ClampOGL:
+        // TODO(Rodrigo): GL_CLAMP was removed as of OpenGL 3.1, to implement GL_CLAMP, we can use
+        // eClampToBorder to get the border color of the texture, and then sample the edge to
+        // manually mix them. However the shader part of this is not yet implemented.
+        return vk::SamplerAddressMode::eClampToBorder;
+    case Tegra::Texture::WrapMode::MirrorOnceClampToEdge:
+        return vk::SamplerAddressMode::eMirrorClampToEdge;
+    case Tegra::Texture::WrapMode::MirrorOnceBorder:
+        UNIMPLEMENTED();
+        return vk::SamplerAddressMode::eMirrorClampToEdge;
+    }
+    UNIMPLEMENTED_MSG("Unimplemented wrap mode={}", static_cast<u32>(wrap_mode));
+    return {};
+}
+
+vk::CompareOp DepthCompareFunction(Tegra::Texture::DepthCompareFunc depth_compare_func) {
+    switch (depth_compare_func) {
+    case Tegra::Texture::DepthCompareFunc::Never:
+        return vk::CompareOp::eNever;
+    case Tegra::Texture::DepthCompareFunc::Less:
+        return vk::CompareOp::eLess;
+    case Tegra::Texture::DepthCompareFunc::LessEqual:
+        return vk::CompareOp::eLessOrEqual;
+    case Tegra::Texture::DepthCompareFunc::Equal:
+        return vk::CompareOp::eEqual;
+    case Tegra::Texture::DepthCompareFunc::NotEqual:
+        return vk::CompareOp::eNotEqual;
+    case Tegra::Texture::DepthCompareFunc::Greater:
+        return vk::CompareOp::eGreater;
+    case Tegra::Texture::DepthCompareFunc::GreaterEqual:
+        return vk::CompareOp::eGreaterOrEqual;
+    case Tegra::Texture::DepthCompareFunc::Always:
+        return vk::CompareOp::eAlways;
+    }
+    UNIMPLEMENTED_MSG("Unimplemented sampler depth compare function={}",
+                      static_cast<u32>(depth_compare_func));
+    return {};
+}
+
+} // namespace Sampler
+
+struct FormatTuple {
+    vk::Format format;            ///< Vulkan format
+    ComponentType component_type; ///< Abstracted component type
+    bool attachable;              ///< True when this format can be used as an attachment
+};
+
+static constexpr std::array<FormatTuple, VideoCore::Surface::MaxPixelFormat> tex_format_tuples = {{
+    {vk::Format::eA8B8G8R8UnormPack32, ComponentType::UNorm, true},    // ABGR8U
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // ABGR8S
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // ABGR8UI
+    {vk::Format::eB5G6R5UnormPack16, ComponentType::UNorm, false},     // B5G6R5U
+    {vk::Format::eA2B10G10R10UnormPack32, ComponentType::UNorm, true}, // A2B10G10R10U
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // A1B5G5R5U
+    {vk::Format::eR8Unorm, ComponentType::UNorm, true},                // R8U
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // R8UI
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // RGBA16F
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // RGBA16U
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // RGBA16UI
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // R11FG11FB10F
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // RGBA32UI
+    {vk::Format::eBc1RgbaUnormBlock, ComponentType::UNorm, false},     // DXT1
+    {vk::Format::eBc2UnormBlock, ComponentType::UNorm, false},         // DXT23
+    {vk::Format::eBc3UnormBlock, ComponentType::UNorm, false},         // DXT45
+    {vk::Format::eBc4UnormBlock, ComponentType::UNorm, false},         // DXN1
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // DXN2UNORM
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // DXN2SNORM
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // BC7U
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // BC6H_UF16
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // BC6H_SF16
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // ASTC_2D_4X4
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // BGRA8
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // RGBA32F
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // RG32F
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // R32F
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // R16F
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // R16U
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // R16S
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // R16UI
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // R16I
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // RG16
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // RG16F
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // RG16UI
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // RG16I
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // RG16S
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // RGB32F
+    {vk::Format::eA8B8G8R8SrgbPack32, ComponentType::UNorm, true},     // RGBA8_SRGB
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // RG8U
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // RG8S
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // RG32UI
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // R32UI
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // ASTC_2D_8X8
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // ASTC_2D_8X5
+    {vk::Format::eUndefined, ComponentType::Invalid, false},           // ASTC_2D_5X4
+
+    // Compressed sRGB formats
+    {vk::Format::eUndefined, ComponentType::Invalid, false}, // BGRA8_SRGB
+    {vk::Format::eUndefined, ComponentType::Invalid, false}, // DXT1_SRGB
+    {vk::Format::eUndefined, ComponentType::Invalid, false}, // DXT23_SRGB
+    {vk::Format::eUndefined, ComponentType::Invalid, false}, // DXT45_SRGB
+    {vk::Format::eUndefined, ComponentType::Invalid, false}, // BC7U_SRGB
+    {vk::Format::eUndefined, ComponentType::Invalid, false}, // ASTC_2D_4X4_SRGB
+    {vk::Format::eUndefined, ComponentType::Invalid, false}, // ASTC_2D_8X8_SRGB
+    {vk::Format::eUndefined, ComponentType::Invalid, false}, // ASTC_2D_8X5_SRGB
+    {vk::Format::eUndefined, ComponentType::Invalid, false}, // ASTC_2D_5X4_SRGB
+    {vk::Format::eUndefined, ComponentType::Invalid, false}, // ASTC_2D_5X5
+    {vk::Format::eUndefined, ComponentType::Invalid, false}, // ASTC_2D_5X5_SRGB
+    {vk::Format::eUndefined, ComponentType::Invalid, false}, // ASTC_2D_10X8
+    {vk::Format::eUndefined, ComponentType::Invalid, false}, // ASTC_2D_10X8_SRGB
+
+    // Depth formats
+    {vk::Format::eD32Sfloat, ComponentType::Float, true}, // Z32F
+    {vk::Format::eD16Unorm, ComponentType::UNorm, true},  // Z16
+
+    // DepthStencil formats
+    {vk::Format::eD24UnormS8Uint, ComponentType::UNorm, true}, // Z24S8
+    {vk::Format::eD24UnormS8Uint, ComponentType::UNorm, true}, // S8Z24 (emulated)
+    {vk::Format::eUndefined, ComponentType::Invalid, false},   // Z32FS8
+}};
+
+static constexpr bool IsZetaFormat(PixelFormat pixel_format) {
+    return pixel_format >= PixelFormat::MaxColorFormat &&
+           pixel_format < PixelFormat::MaxDepthStencilFormat;
+}
+
+std::pair<vk::Format, bool> SurfaceFormat(const VKDevice& device, FormatType format_type,
+                                          PixelFormat pixel_format, ComponentType component_type) {
+    ASSERT(static_cast<std::size_t>(pixel_format) < tex_format_tuples.size());
+
+    const auto tuple = tex_format_tuples[static_cast<u32>(pixel_format)];
+    UNIMPLEMENTED_IF_MSG(tuple.format == vk::Format::eUndefined,
+                         "Unimplemented texture format with pixel format={} and component type={}",
+                         static_cast<u32>(pixel_format), static_cast<u32>(component_type));
+    ASSERT_MSG(component_type == tuple.component_type, "Component type mismatch");
+
+    auto usage = vk::FormatFeatureFlagBits::eSampledImage |
+                 vk::FormatFeatureFlagBits::eTransferDst | vk::FormatFeatureFlagBits::eTransferSrc;
+    if (tuple.attachable) {
+        usage |= IsZetaFormat(pixel_format) ? vk::FormatFeatureFlagBits::eDepthStencilAttachment
+                                            : vk::FormatFeatureFlagBits::eColorAttachment;
+    }
+    return {device.GetSupportedFormat(tuple.format, usage, format_type), tuple.attachable};
+}
+
+vk::ShaderStageFlagBits ShaderStage(Maxwell::ShaderStage stage) {
+    switch (stage) {
+    case Maxwell::ShaderStage::Vertex:
+        return vk::ShaderStageFlagBits::eVertex;
+    case Maxwell::ShaderStage::TesselationControl:
+        return vk::ShaderStageFlagBits::eTessellationControl;
+    case Maxwell::ShaderStage::TesselationEval:
+        return vk::ShaderStageFlagBits::eTessellationEvaluation;
+    case Maxwell::ShaderStage::Geometry:
+        return vk::ShaderStageFlagBits::eGeometry;
+    case Maxwell::ShaderStage::Fragment:
+        return vk::ShaderStageFlagBits::eFragment;
+    }
+    UNIMPLEMENTED_MSG("Unimplemented shader stage={}", static_cast<u32>(stage));
+    return {};
+}
+
+vk::PrimitiveTopology PrimitiveTopology(Maxwell::PrimitiveTopology topology) {
+    switch (topology) {
+    case Maxwell::PrimitiveTopology::Points:
+        return vk::PrimitiveTopology::ePointList;
+    case Maxwell::PrimitiveTopology::Lines:
+        return vk::PrimitiveTopology::eLineList;
+    case Maxwell::PrimitiveTopology::LineStrip:
+        return vk::PrimitiveTopology::eLineStrip;
+    case Maxwell::PrimitiveTopology::Triangles:
+        return vk::PrimitiveTopology::eTriangleList;
+    case Maxwell::PrimitiveTopology::TriangleStrip:
+        return vk::PrimitiveTopology::eTriangleStrip;
+    }
+    UNIMPLEMENTED_MSG("Unimplemented topology={}", static_cast<u32>(topology));
+    return {};
+}
+
+vk::Format VertexFormat(Maxwell::VertexAttribute::Type type, Maxwell::VertexAttribute::Size size) {
+    switch (type) {
+    case Maxwell::VertexAttribute::Type::SignedNorm:
+        break;
+    case Maxwell::VertexAttribute::Type::UnsignedNorm:
+        switch (size) {
+        case Maxwell::VertexAttribute::Size::Size_8_8_8_8:
+            return vk::Format::eR8G8B8A8Unorm;
+        default:
+            break;
+        }
+        break;
+    case Maxwell::VertexAttribute::Type::SignedInt:
+        break;
+    case Maxwell::VertexAttribute::Type::UnsignedInt:
+        switch (size) {
+        case Maxwell::VertexAttribute::Size::Size_32:
+            return vk::Format::eR32Uint;
+        default:
+            break;
+        }
+    case Maxwell::VertexAttribute::Type::UnsignedScaled:
+    case Maxwell::VertexAttribute::Type::SignedScaled:
+        break;
+    case Maxwell::VertexAttribute::Type::Float:
+        switch (size) {
+        case Maxwell::VertexAttribute::Size::Size_32_32_32_32:
+            return vk::Format::eR32G32B32A32Sfloat;
+        case Maxwell::VertexAttribute::Size::Size_32_32_32:
+            return vk::Format::eR32G32B32Sfloat;
+        case Maxwell::VertexAttribute::Size::Size_32_32:
+            return vk::Format::eR32G32Sfloat;
+        case Maxwell::VertexAttribute::Size::Size_32:
+            return vk::Format::eR32Sfloat;
+        default:
+            break;
+        }
+        break;
+    }
+    UNIMPLEMENTED_MSG("Unimplemented vertex format of type={} and size={}", static_cast<u32>(type),
+                      static_cast<u32>(size));
+    return {};
+}
+
+vk::CompareOp ComparisonOp(Maxwell::ComparisonOp comparison) {
+    switch (comparison) {
+    case Maxwell::ComparisonOp::Never:
+    case Maxwell::ComparisonOp::NeverOld:
+        return vk::CompareOp::eNever;
+    case Maxwell::ComparisonOp::Less:
+    case Maxwell::ComparisonOp::LessOld:
+        return vk::CompareOp::eLess;
+    case Maxwell::ComparisonOp::Equal:
+    case Maxwell::ComparisonOp::EqualOld:
+        return vk::CompareOp::eEqual;
+    case Maxwell::ComparisonOp::LessEqual:
+    case Maxwell::ComparisonOp::LessEqualOld:
+        return vk::CompareOp::eLessOrEqual;
+    case Maxwell::ComparisonOp::Greater:
+    case Maxwell::ComparisonOp::GreaterOld:
+        return vk::CompareOp::eGreater;
+    case Maxwell::ComparisonOp::NotEqual:
+    case Maxwell::ComparisonOp::NotEqualOld:
+        return vk::CompareOp::eNotEqual;
+    case Maxwell::ComparisonOp::GreaterEqual:
+    case Maxwell::ComparisonOp::GreaterEqualOld:
+        return vk::CompareOp::eGreaterOrEqual;
+    case Maxwell::ComparisonOp::Always:
+    case Maxwell::ComparisonOp::AlwaysOld:
+        return vk::CompareOp::eAlways;
+    }
+    UNIMPLEMENTED_MSG("Unimplemented comparison op={}", static_cast<u32>(comparison));
+    return {};
+}
+
+vk::IndexType IndexFormat(Maxwell::IndexFormat index_format) {
+    switch (index_format) {
+    case Maxwell::IndexFormat::UnsignedByte:
+        UNIMPLEMENTED_MSG("Vulkan does not support native u8 index format");
+        return vk::IndexType::eUint16;
+    case Maxwell::IndexFormat::UnsignedShort:
+        return vk::IndexType::eUint16;
+    case Maxwell::IndexFormat::UnsignedInt:
+        return vk::IndexType::eUint32;
+    }
+    UNIMPLEMENTED_MSG("Unimplemented index_format={}", static_cast<u32>(index_format));
+    return {};
+}
+
+vk::StencilOp StencilOp(Maxwell::StencilOp stencil_op) {
+    switch (stencil_op) {
+    case Maxwell::StencilOp::Keep:
+    case Maxwell::StencilOp::KeepOGL:
+        return vk::StencilOp::eKeep;
+    case Maxwell::StencilOp::Zero:
+    case Maxwell::StencilOp::ZeroOGL:
+        return vk::StencilOp::eZero;
+    case Maxwell::StencilOp::Replace:
+    case Maxwell::StencilOp::ReplaceOGL:
+        return vk::StencilOp::eReplace;
+    case Maxwell::StencilOp::Incr:
+    case Maxwell::StencilOp::IncrOGL:
+        return vk::StencilOp::eIncrementAndClamp;
+    case Maxwell::StencilOp::Decr:
+    case Maxwell::StencilOp::DecrOGL:
+        return vk::StencilOp::eDecrementAndClamp;
+    case Maxwell::StencilOp::Invert:
+    case Maxwell::StencilOp::InvertOGL:
+        return vk::StencilOp::eInvert;
+    case Maxwell::StencilOp::IncrWrap:
+    case Maxwell::StencilOp::IncrWrapOGL:
+        return vk::StencilOp::eIncrementAndWrap;
+    case Maxwell::StencilOp::DecrWrap:
+    case Maxwell::StencilOp::DecrWrapOGL:
+        return vk::StencilOp::eDecrementAndWrap;
+    }
+    UNIMPLEMENTED_MSG("Unimplemented stencil op={}", static_cast<u32>(stencil_op));
+    return {};
+}
+
+vk::BlendOp BlendEquation(Maxwell::Blend::Equation equation) {
+    switch (equation) {
+    case Maxwell::Blend::Equation::Add:
+    case Maxwell::Blend::Equation::AddGL:
+        return vk::BlendOp::eAdd;
+    case Maxwell::Blend::Equation::Subtract:
+    case Maxwell::Blend::Equation::SubtractGL:
+        return vk::BlendOp::eSubtract;
+    case Maxwell::Blend::Equation::ReverseSubtract:
+    case Maxwell::Blend::Equation::ReverseSubtractGL:
+        return vk::BlendOp::eReverseSubtract;
+    case Maxwell::Blend::Equation::Min:
+    case Maxwell::Blend::Equation::MinGL:
+        return vk::BlendOp::eMin;
+    case Maxwell::Blend::Equation::Max:
+    case Maxwell::Blend::Equation::MaxGL:
+        return vk::BlendOp::eMax;
+    }
+    UNIMPLEMENTED_MSG("Unimplemented blend equation={}", static_cast<u32>(equation));
+    return {};
+}
+
+vk::BlendFactor BlendFactor(Maxwell::Blend::Factor factor) {
+    switch (factor) {
+    case Maxwell::Blend::Factor::Zero:
+    case Maxwell::Blend::Factor::ZeroGL:
+        return vk::BlendFactor::eZero;
+    case Maxwell::Blend::Factor::One:
+    case Maxwell::Blend::Factor::OneGL:
+        return vk::BlendFactor::eOne;
+    case Maxwell::Blend::Factor::SourceColor:
+    case Maxwell::Blend::Factor::SourceColorGL:
+        return vk::BlendFactor::eSrcColor;
+    case Maxwell::Blend::Factor::OneMinusSourceColor:
+    case Maxwell::Blend::Factor::OneMinusSourceColorGL:
+        return vk::BlendFactor::eOneMinusSrcColor;
+    case Maxwell::Blend::Factor::SourceAlpha:
+    case Maxwell::Blend::Factor::SourceAlphaGL:
+        return vk::BlendFactor::eSrcAlpha;
+    case Maxwell::Blend::Factor::OneMinusSourceAlpha:
+    case Maxwell::Blend::Factor::OneMinusSourceAlphaGL:
+        return vk::BlendFactor::eOneMinusSrcAlpha;
+    case Maxwell::Blend::Factor::DestAlpha:
+    case Maxwell::Blend::Factor::DestAlphaGL:
+        return vk::BlendFactor::eDstAlpha;
+    case Maxwell::Blend::Factor::OneMinusDestAlpha:
+    case Maxwell::Blend::Factor::OneMinusDestAlphaGL:
+        return vk::BlendFactor::eOneMinusDstAlpha;
+    case Maxwell::Blend::Factor::DestColor:
+    case Maxwell::Blend::Factor::DestColorGL:
+        return vk::BlendFactor::eDstColor;
+    case Maxwell::Blend::Factor::OneMinusDestColor:
+    case Maxwell::Blend::Factor::OneMinusDestColorGL:
+        return vk::BlendFactor::eOneMinusDstColor;
+    case Maxwell::Blend::Factor::SourceAlphaSaturate:
+    case Maxwell::Blend::Factor::SourceAlphaSaturateGL:
+        return vk::BlendFactor::eSrcAlphaSaturate;
+    case Maxwell::Blend::Factor::Source1Color:
+    case Maxwell::Blend::Factor::Source1ColorGL:
+        return vk::BlendFactor::eSrc1Color;
+    case Maxwell::Blend::Factor::OneMinusSource1Color:
+    case Maxwell::Blend::Factor::OneMinusSource1ColorGL:
+        return vk::BlendFactor::eOneMinusSrc1Color;
+    case Maxwell::Blend::Factor::Source1Alpha:
+    case Maxwell::Blend::Factor::Source1AlphaGL:
+        return vk::BlendFactor::eSrc1Alpha;
+    case Maxwell::Blend::Factor::OneMinusSource1Alpha:
+    case Maxwell::Blend::Factor::OneMinusSource1AlphaGL:
+        return vk::BlendFactor::eOneMinusSrc1Alpha;
+    case Maxwell::Blend::Factor::ConstantColor:
+    case Maxwell::Blend::Factor::ConstantColorGL:
+        return vk::BlendFactor::eConstantColor;
+    case Maxwell::Blend::Factor::OneMinusConstantColor:
+    case Maxwell::Blend::Factor::OneMinusConstantColorGL:
+        return vk::BlendFactor::eOneMinusConstantColor;
+    case Maxwell::Blend::Factor::ConstantAlpha:
+    case Maxwell::Blend::Factor::ConstantAlphaGL:
+        return vk::BlendFactor::eConstantAlpha;
+    case Maxwell::Blend::Factor::OneMinusConstantAlpha:
+    case Maxwell::Blend::Factor::OneMinusConstantAlphaGL:
+        return vk::BlendFactor::eOneMinusConstantAlpha;
+    }
+    UNIMPLEMENTED_MSG("Unimplemented blend factor={}", static_cast<u32>(factor));
+    return {};
+}
+
+vk::FrontFace FrontFace(Maxwell::Cull::FrontFace front_face) {
+    switch (front_face) {
+    case Maxwell::Cull::FrontFace::ClockWise:
+        return vk::FrontFace::eClockwise;
+    case Maxwell::Cull::FrontFace::CounterClockWise:
+        return vk::FrontFace::eCounterClockwise;
+    }
+    UNIMPLEMENTED_MSG("Unimplemented front face={}", static_cast<u32>(front_face));
+    return {};
+}
+
+vk::CullModeFlags CullFace(Maxwell::Cull::CullFace cull_face) {
+    switch (cull_face) {
+    case Maxwell::Cull::CullFace::Front:
+        return vk::CullModeFlagBits::eFront;
+    case Maxwell::Cull::CullFace::Back:
+        return vk::CullModeFlagBits::eBack;
+    case Maxwell::Cull::CullFace::FrontAndBack:
+        return vk::CullModeFlagBits::eFrontAndBack;
+    }
+    UNIMPLEMENTED_MSG("Unimplemented cull face={}", static_cast<u32>(cull_face));
+    return {};
+}
+
+vk::ComponentSwizzle SwizzleSource(Tegra::Texture::SwizzleSource swizzle) {
+    switch (swizzle) {
+    case Tegra::Texture::SwizzleSource::Zero:
+        return vk::ComponentSwizzle::eZero;
+    case Tegra::Texture::SwizzleSource::R:
+        return vk::ComponentSwizzle::eR;
+    case Tegra::Texture::SwizzleSource::G:
+        return vk::ComponentSwizzle::eG;
+    case Tegra::Texture::SwizzleSource::B:
+        return vk::ComponentSwizzle::eB;
+    case Tegra::Texture::SwizzleSource::A:
+        return vk::ComponentSwizzle::eA;
+    case Tegra::Texture::SwizzleSource::OneInt:
+    case Tegra::Texture::SwizzleSource::OneFloat:
+        return vk::ComponentSwizzle::eOne;
+    }
+    UNIMPLEMENTED_MSG("Unimplemented swizzle source={}", static_cast<u32>(swizzle));
+    return {};
+}
+
+} // namespace Vulkan::MaxwellToVK
diff --git a/src/video_core/renderer_vulkan/maxwell_to_vk.h b/src/video_core/renderer_vulkan/maxwell_to_vk.h
new file mode 100644
index 000000000..4cadc0721
--- /dev/null
+++ b/src/video_core/renderer_vulkan/maxwell_to_vk.h
@@ -0,0 +1,58 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <utility>
+#include "common/common_types.h"
+#include "video_core/engines/maxwell_3d.h"
+#include "video_core/renderer_vulkan/declarations.h"
+#include "video_core/renderer_vulkan/vk_device.h"
+#include "video_core/surface.h"
+#include "video_core/textures/texture.h"
+
+namespace Vulkan::MaxwellToVK {
+
+using Maxwell = Tegra::Engines::Maxwell3D::Regs;
+using PixelFormat = VideoCore::Surface::PixelFormat;
+using ComponentType = VideoCore::Surface::ComponentType;
+
+namespace Sampler {
+
+vk::Filter Filter(Tegra::Texture::TextureFilter filter);
+
+vk::SamplerMipmapMode MipmapMode(Tegra::Texture::TextureMipmapFilter mipmap_filter);
+
+vk::SamplerAddressMode WrapMode(Tegra::Texture::WrapMode wrap_mode);
+
+vk::CompareOp DepthCompareFunction(Tegra::Texture::DepthCompareFunc depth_compare_func);
+
+} // namespace Sampler
+
+std::pair<vk::Format, bool> SurfaceFormat(const VKDevice& device, FormatType format_type,
+                                          PixelFormat pixel_format, ComponentType component_type);
+
+vk::ShaderStageFlagBits ShaderStage(Maxwell::ShaderStage stage);
+
+vk::PrimitiveTopology PrimitiveTopology(Maxwell::PrimitiveTopology topology);
+
+vk::Format VertexFormat(Maxwell::VertexAttribute::Type type, Maxwell::VertexAttribute::Size size);
+
+vk::CompareOp ComparisonOp(Maxwell::ComparisonOp comparison);
+
+vk::IndexType IndexFormat(Maxwell::IndexFormat index_format);
+
+vk::StencilOp StencilOp(Maxwell::StencilOp stencil_op);
+
+vk::BlendOp BlendEquation(Maxwell::Blend::Equation equation);
+
+vk::BlendFactor BlendFactor(Maxwell::Blend::Factor factor);
+
+vk::FrontFace FrontFace(Maxwell::Cull::FrontFace front_face);
+
+vk::CullModeFlags CullFace(Maxwell::Cull::CullFace cull_face);
+
+vk::ComponentSwizzle SwizzleSource(Tegra::Texture::SwizzleSource swizzle);
+
+} // namespace Vulkan::MaxwellToVK
diff --git a/src/video_core/renderer_vulkan/vk_buffer_cache.cpp b/src/video_core/renderer_vulkan/vk_buffer_cache.cpp
new file mode 100644
index 000000000..95eab3fec
--- /dev/null
+++ b/src/video_core/renderer_vulkan/vk_buffer_cache.cpp
@@ -0,0 +1,123 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <cstring>
+#include <memory>
+#include <optional>
+#include <tuple>
+
+#include "common/alignment.h"
+#include "common/assert.h"
+#include "core/memory.h"
+#include "video_core/renderer_vulkan/declarations.h"
+#include "video_core/renderer_vulkan/vk_buffer_cache.h"
+#include "video_core/renderer_vulkan/vk_scheduler.h"
+#include "video_core/renderer_vulkan/vk_stream_buffer.h"
+
+namespace Vulkan {
+
+CachedBufferEntry::CachedBufferEntry(VAddr cpu_addr, std::size_t size, u64 offset,
+                                     std::size_t alignment, u8* host_ptr)
+    : cpu_addr{cpu_addr}, size{size}, offset{offset}, alignment{alignment}, RasterizerCacheObject{
+                                                                                host_ptr} {}
+
+VKBufferCache::VKBufferCache(Tegra::MemoryManager& tegra_memory_manager,
+                             VideoCore::RasterizerInterface& rasterizer, const VKDevice& device,
+                             VKMemoryManager& memory_manager, VKScheduler& scheduler, u64 size)
+    : RasterizerCache{rasterizer}, tegra_memory_manager{tegra_memory_manager} {
+    const auto usage = vk::BufferUsageFlagBits::eVertexBuffer |
+                       vk::BufferUsageFlagBits::eIndexBuffer |
+                       vk::BufferUsageFlagBits::eUniformBuffer;
+    const auto access = vk::AccessFlagBits::eVertexAttributeRead | vk::AccessFlagBits::eIndexRead |
+                        vk::AccessFlagBits::eUniformRead;
+    stream_buffer =
+        std::make_unique<VKStreamBuffer>(device, memory_manager, scheduler, size, usage, access,
+                                         vk::PipelineStageFlagBits::eAllCommands);
+    buffer_handle = stream_buffer->GetBuffer();
+}
+
+VKBufferCache::~VKBufferCache() = default;
+
+u64 VKBufferCache::UploadMemory(Tegra::GPUVAddr gpu_addr, std::size_t size, u64 alignment,
+                                bool cache) {
+    const auto cpu_addr{tegra_memory_manager.GpuToCpuAddress(gpu_addr)};
+    ASSERT_MSG(cpu_addr, "Invalid GPU address");
+
+    // Cache management is a big overhead, so only cache entries with a given size.
+    // TODO: Figure out which size is the best for given games.
+    cache &= size >= 2048;
+
+    const auto& host_ptr{Memory::GetPointer(*cpu_addr)};
+    if (cache) {
+        auto entry = TryGet(host_ptr);
+        if (entry) {
+            if (entry->GetSize() >= size && entry->GetAlignment() == alignment) {
+                return entry->GetOffset();
+            }
+            Unregister(entry);
+        }
+    }
+
+    AlignBuffer(alignment);
+    const u64 uploaded_offset = buffer_offset;
+
+    if (!host_ptr) {
+        return uploaded_offset;
+    }
+
+    std::memcpy(buffer_ptr, host_ptr, size);
+    buffer_ptr += size;
+    buffer_offset += size;
+
+    if (cache) {
+        auto entry = std::make_shared<CachedBufferEntry>(*cpu_addr, size, uploaded_offset,
+                                                         alignment, host_ptr);
+        Register(entry);
+    }
+
+    return uploaded_offset;
+}
+
+u64 VKBufferCache::UploadHostMemory(const u8* raw_pointer, std::size_t size, u64 alignment) {
+    AlignBuffer(alignment);
+    std::memcpy(buffer_ptr, raw_pointer, size);
+    const u64 uploaded_offset = buffer_offset;
+
+    buffer_ptr += size;
+    buffer_offset += size;
+    return uploaded_offset;
+}
+
+std::tuple<u8*, u64> VKBufferCache::ReserveMemory(std::size_t size, u64 alignment) {
+    AlignBuffer(alignment);
+    u8* const uploaded_ptr = buffer_ptr;
+    const u64 uploaded_offset = buffer_offset;
+
+    buffer_ptr += size;
+    buffer_offset += size;
+    return {uploaded_ptr, uploaded_offset};
+}
+
+void VKBufferCache::Reserve(std::size_t max_size) {
+    bool invalidate;
+    std::tie(buffer_ptr, buffer_offset_base, invalidate) = stream_buffer->Reserve(max_size);
+    buffer_offset = buffer_offset_base;
+
+    if (invalidate) {
+        InvalidateAll();
+    }
+}
+
+VKExecutionContext VKBufferCache::Send(VKExecutionContext exctx) {
+    return stream_buffer->Send(exctx, buffer_offset - buffer_offset_base);
+}
+
+void VKBufferCache::AlignBuffer(std::size_t alignment) {
+    // Align the offset, not the mapped pointer
+    const u64 offset_aligned = Common::AlignUp(buffer_offset, alignment);
+    buffer_ptr += offset_aligned - buffer_offset;
+    buffer_offset = offset_aligned;
+}
+
+} // namespace Vulkan
diff --git a/src/video_core/renderer_vulkan/vk_buffer_cache.h b/src/video_core/renderer_vulkan/vk_buffer_cache.h
new file mode 100644
index 000000000..8b415744b
--- /dev/null
+++ b/src/video_core/renderer_vulkan/vk_buffer_cache.h
@@ -0,0 +1,104 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <memory>
+#include <tuple>
+
+#include "common/common_types.h"
+#include "video_core/gpu.h"
+#include "video_core/rasterizer_cache.h"
+#include "video_core/renderer_vulkan/declarations.h"
+#include "video_core/renderer_vulkan/vk_scheduler.h"
+
+namespace Tegra {
+class MemoryManager;
+}
+
+namespace Vulkan {
+
+class VKDevice;
+class VKFence;
+class VKMemoryManager;
+class VKStreamBuffer;
+
+class CachedBufferEntry final : public RasterizerCacheObject {
+public:
+    explicit CachedBufferEntry(VAddr cpu_addr, std::size_t size, u64 offset, std::size_t alignment,
+                               u8* host_ptr);
+
+    VAddr GetCpuAddr() const override {
+        return cpu_addr;
+    }
+
+    std::size_t GetSizeInBytes() const override {
+        return size;
+    }
+
+    std::size_t GetSize() const {
+        return size;
+    }
+
+    u64 GetOffset() const {
+        return offset;
+    }
+
+    std::size_t GetAlignment() const {
+        return alignment;
+    }
+
+    // We do not have to flush this cache as things in it are never modified by us.
+    void Flush() override {}
+
+private:
+    VAddr cpu_addr{};
+    std::size_t size{};
+    u64 offset{};
+    std::size_t alignment{};
+};
+
+class VKBufferCache final : public RasterizerCache<std::shared_ptr<CachedBufferEntry>> {
+public:
+    explicit VKBufferCache(Tegra::MemoryManager& tegra_memory_manager,
+                           VideoCore::RasterizerInterface& rasterizer, const VKDevice& device,
+                           VKMemoryManager& memory_manager, VKScheduler& scheduler, u64 size);
+    ~VKBufferCache();
+
+    /// Uploads data from a guest GPU address. Returns host's buffer offset where it's been
+    /// allocated.
+    u64 UploadMemory(Tegra::GPUVAddr gpu_addr, std::size_t size, u64 alignment = 4,
+                     bool cache = true);
+
+    /// Uploads from a host memory. Returns host's buffer offset where it's been allocated.
+    u64 UploadHostMemory(const u8* raw_pointer, std::size_t size, u64 alignment = 4);
+
+    /// Reserves memory to be used by host's CPU. Returns mapped address and offset.
+    std::tuple<u8*, u64> ReserveMemory(std::size_t size, u64 alignment = 4);
+
+    /// Reserves a region of memory to be used in subsequent upload/reserve operations.
+    void Reserve(std::size_t max_size);
+
+    /// Ensures that the set data is sent to the device.
+    [[nodiscard]] VKExecutionContext Send(VKExecutionContext exctx);
+
+    /// Returns the buffer cache handle.
+    vk::Buffer GetBuffer() const {
+        return buffer_handle;
+    }
+
+private:
+    void AlignBuffer(std::size_t alignment);
+
+    Tegra::MemoryManager& tegra_memory_manager;
+
+    std::unique_ptr<VKStreamBuffer> stream_buffer;
+    vk::Buffer buffer_handle;
+
+    u8* buffer_ptr = nullptr;
+    u64 buffer_offset = 0;
+    u64 buffer_offset_base = 0;
+};
+
+} // namespace Vulkan
diff --git a/src/video_core/renderer_vulkan/vk_device.cpp b/src/video_core/renderer_vulkan/vk_device.cpp
new file mode 100644
index 000000000..00242ecbe
--- /dev/null
+++ b/src/video_core/renderer_vulkan/vk_device.cpp
@@ -0,0 +1,238 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <map>
+#include <optional>
+#include <set>
+#include <vector>
+#include "common/assert.h"
+#include "video_core/renderer_vulkan/declarations.h"
+#include "video_core/renderer_vulkan/vk_device.h"
+
+namespace Vulkan {
+
+namespace Alternatives {
+
+constexpr std::array<vk::Format, 3> Depth24UnormS8Uint = {
+    vk::Format::eD32SfloatS8Uint, vk::Format::eD16UnormS8Uint, {}};
+constexpr std::array<vk::Format, 3> Depth16UnormS8Uint = {
+    vk::Format::eD24UnormS8Uint, vk::Format::eD32SfloatS8Uint, {}};
+
+} // namespace Alternatives
+
+constexpr const vk::Format* GetFormatAlternatives(vk::Format format) {
+    switch (format) {
+    case vk::Format::eD24UnormS8Uint:
+        return Alternatives::Depth24UnormS8Uint.data();
+    case vk::Format::eD16UnormS8Uint:
+        return Alternatives::Depth16UnormS8Uint.data();
+    default:
+        return nullptr;
+    }
+}
+
+constexpr vk::FormatFeatureFlags GetFormatFeatures(vk::FormatProperties properties,
+                                                   FormatType format_type) {
+    switch (format_type) {
+    case FormatType::Linear:
+        return properties.linearTilingFeatures;
+    case FormatType::Optimal:
+        return properties.optimalTilingFeatures;
+    case FormatType::Buffer:
+        return properties.bufferFeatures;
+    default:
+        return {};
+    }
+}
+
+VKDevice::VKDevice(const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDevice physical,
+                   vk::SurfaceKHR surface)
+    : physical{physical}, format_properties{GetFormatProperties(dldi, physical)} {
+    SetupFamilies(dldi, surface);
+    SetupProperties(dldi);
+}
+
+VKDevice::~VKDevice() = default;
+
+bool VKDevice::Create(const vk::DispatchLoaderDynamic& dldi, vk::Instance instance) {
+    const auto queue_cis = GetDeviceQueueCreateInfos();
+    vk::PhysicalDeviceFeatures device_features{};
+
+    const std::vector<const char*> extensions = {VK_KHR_SWAPCHAIN_EXTENSION_NAME};
+    const vk::DeviceCreateInfo device_ci({}, static_cast<u32>(queue_cis.size()), queue_cis.data(),
+                                         0, nullptr, static_cast<u32>(extensions.size()),
+                                         extensions.data(), &device_features);
+    vk::Device dummy_logical;
+    if (physical.createDevice(&device_ci, nullptr, &dummy_logical, dldi) != vk::Result::eSuccess) {
+        LOG_CRITICAL(Render_Vulkan, "Logical device failed to be created!");
+        return false;
+    }
+
+    dld.init(instance, dldi.vkGetInstanceProcAddr, dummy_logical, dldi.vkGetDeviceProcAddr);
+    logical = UniqueDevice(
+        dummy_logical, vk::ObjectDestroy<vk::NoParent, vk::DispatchLoaderDynamic>(nullptr, dld));
+
+    graphics_queue = logical->getQueue(graphics_family, 0, dld);
+    present_queue = logical->getQueue(present_family, 0, dld);
+    return true;
+}
+
+vk::Format VKDevice::GetSupportedFormat(vk::Format wanted_format,
+                                        vk::FormatFeatureFlags wanted_usage,
+                                        FormatType format_type) const {
+    if (IsFormatSupported(wanted_format, wanted_usage, format_type)) {
+        return wanted_format;
+    }
+    // The wanted format is not supported by hardware, search for alternatives
+    const vk::Format* alternatives = GetFormatAlternatives(wanted_format);
+    if (alternatives == nullptr) {
+        LOG_CRITICAL(Render_Vulkan,
+                     "Format={} with usage={} and type={} has no defined alternatives and host "
+                     "hardware does not support it",
+                     static_cast<u32>(wanted_format), static_cast<u32>(wanted_usage),
+                     static_cast<u32>(format_type));
+        UNREACHABLE();
+        return wanted_format;
+    }
+
+    std::size_t i = 0;
+    for (vk::Format alternative = alternatives[0]; alternative != vk::Format{};
+         alternative = alternatives[++i]) {
+        if (!IsFormatSupported(alternative, wanted_usage, format_type))
+            continue;
+        LOG_WARNING(Render_Vulkan,
+                    "Emulating format={} with alternative format={} with usage={} and type={}",
+                    static_cast<u32>(wanted_format), static_cast<u32>(alternative),
+                    static_cast<u32>(wanted_usage), static_cast<u32>(format_type));
+        return alternative;
+    }
+
+    // No alternatives found, panic
+    LOG_CRITICAL(Render_Vulkan,
+                 "Format={} with usage={} and type={} is not supported by the host hardware and "
+                 "doesn't support any of the alternatives",
+                 static_cast<u32>(wanted_format), static_cast<u32>(wanted_usage),
+                 static_cast<u32>(format_type));
+    UNREACHABLE();
+    return wanted_format;
+}
+
+bool VKDevice::IsFormatSupported(vk::Format wanted_format, vk::FormatFeatureFlags wanted_usage,
+                                 FormatType format_type) const {
+    const auto it = format_properties.find(wanted_format);
+    if (it == format_properties.end()) {
+        LOG_CRITICAL(Render_Vulkan, "Unimplemented format query={}", vk::to_string(wanted_format));
+        UNREACHABLE();
+        return true;
+    }
+    const vk::FormatFeatureFlags supported_usage = GetFormatFeatures(it->second, format_type);
+    return (supported_usage & wanted_usage) == wanted_usage;
+}
+
+bool VKDevice::IsSuitable(const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDevice physical,
+                          vk::SurfaceKHR surface) {
+    const std::string swapchain_extension = VK_KHR_SWAPCHAIN_EXTENSION_NAME;
+
+    bool has_swapchain{};
+    for (const auto& prop : physical.enumerateDeviceExtensionProperties(nullptr, dldi)) {
+        has_swapchain |= prop.extensionName == swapchain_extension;
+    }
+    if (!has_swapchain) {
+        // The device doesn't support creating swapchains.
+        return false;
+    }
+
+    bool has_graphics{}, has_present{};
+    const auto queue_family_properties = physical.getQueueFamilyProperties(dldi);
+    for (u32 i = 0; i < static_cast<u32>(queue_family_properties.size()); ++i) {
+        const auto& family = queue_family_properties[i];
+        if (family.queueCount == 0)
+            continue;
+
+        has_graphics |=
+            (family.queueFlags & vk::QueueFlagBits::eGraphics) != static_cast<vk::QueueFlagBits>(0);
+        has_present |= physical.getSurfaceSupportKHR(i, surface, dldi) != 0;
+    }
+    if (!has_graphics || !has_present) {
+        // The device doesn't have a graphics and present queue.
+        return false;
+    }
+
+    // TODO(Rodrigo): Check if the device matches all requeriments.
+    const vk::PhysicalDeviceProperties props = physical.getProperties(dldi);
+    if (props.limits.maxUniformBufferRange < 65536) {
+        return false;
+    }
+
+    // Device is suitable.
+    return true;
+}
+
+void VKDevice::SetupFamilies(const vk::DispatchLoaderDynamic& dldi, vk::SurfaceKHR surface) {
+    std::optional<u32> graphics_family_, present_family_;
+
+    const auto queue_family_properties = physical.getQueueFamilyProperties(dldi);
+    for (u32 i = 0; i < static_cast<u32>(queue_family_properties.size()); ++i) {
+        if (graphics_family_ && present_family_)
+            break;
+
+        const auto& queue_family = queue_family_properties[i];
+        if (queue_family.queueCount == 0)
+            continue;
+
+        if (queue_family.queueFlags & vk::QueueFlagBits::eGraphics)
+            graphics_family_ = i;
+        if (physical.getSurfaceSupportKHR(i, surface, dldi))
+            present_family_ = i;
+    }
+    ASSERT(graphics_family_ && present_family_);
+
+    graphics_family = *graphics_family_;
+    present_family = *present_family_;
+}
+
+void VKDevice::SetupProperties(const vk::DispatchLoaderDynamic& dldi) {
+    const vk::PhysicalDeviceProperties props = physical.getProperties(dldi);
+    device_type = props.deviceType;
+    uniform_buffer_alignment = static_cast<u64>(props.limits.minUniformBufferOffsetAlignment);
+}
+
+std::vector<vk::DeviceQueueCreateInfo> VKDevice::GetDeviceQueueCreateInfos() const {
+    static const float QUEUE_PRIORITY = 1.f;
+
+    std::set<u32> unique_queue_families = {graphics_family, present_family};
+    std::vector<vk::DeviceQueueCreateInfo> queue_cis;
+
+    for (u32 queue_family : unique_queue_families)
+        queue_cis.push_back({{}, queue_family, 1, &QUEUE_PRIORITY});
+
+    return queue_cis;
+}
+
+std::map<vk::Format, vk::FormatProperties> VKDevice::GetFormatProperties(
+    const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDevice physical) {
+    std::map<vk::Format, vk::FormatProperties> format_properties;
+
+    const auto AddFormatQuery = [&format_properties, &dldi, physical](vk::Format format) {
+        format_properties.emplace(format, physical.getFormatProperties(format, dldi));
+    };
+    AddFormatQuery(vk::Format::eA8B8G8R8UnormPack32);
+    AddFormatQuery(vk::Format::eB5G6R5UnormPack16);
+    AddFormatQuery(vk::Format::eA2B10G10R10UnormPack32);
+    AddFormatQuery(vk::Format::eR8G8B8A8Srgb);
+    AddFormatQuery(vk::Format::eR8Unorm);
+    AddFormatQuery(vk::Format::eD32Sfloat);
+    AddFormatQuery(vk::Format::eD16Unorm);
+    AddFormatQuery(vk::Format::eD16UnormS8Uint);
+    AddFormatQuery(vk::Format::eD24UnormS8Uint);
+    AddFormatQuery(vk::Format::eD32SfloatS8Uint);
+    AddFormatQuery(vk::Format::eBc1RgbaUnormBlock);
+    AddFormatQuery(vk::Format::eBc2UnormBlock);
+    AddFormatQuery(vk::Format::eBc3UnormBlock);
+    AddFormatQuery(vk::Format::eBc4UnormBlock);
+
+    return format_properties;
+}
+
+} // namespace Vulkan
diff --git a/src/video_core/renderer_vulkan/vk_device.h b/src/video_core/renderer_vulkan/vk_device.h
new file mode 100644
index 000000000..e87c7a508
--- /dev/null
+++ b/src/video_core/renderer_vulkan/vk_device.h
@@ -0,0 +1,116 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <map>
+#include <vector>
+#include "common/common_types.h"
+#include "video_core/renderer_vulkan/declarations.h"
+
+namespace Vulkan {
+
+/// Format usage descriptor
+enum class FormatType { Linear, Optimal, Buffer };
+
+/// Handles data specific to a physical device.
+class VKDevice final {
+public:
+    explicit VKDevice(const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDevice physical,
+                      vk::SurfaceKHR surface);
+    ~VKDevice();
+
+    /// Initializes the device. Returns true on success.
+    bool Create(const vk::DispatchLoaderDynamic& dldi, vk::Instance instance);
+
+    /**
+     * Returns a format supported by the device for the passed requeriments.
+     * @param wanted_format The ideal format to be returned. It may not be the returned format.
+     * @param wanted_usage The usage that must be fulfilled even if the format is not supported.
+     * @param format_type Format type usage.
+     * @returns A format supported by the device.
+     */
+    vk::Format GetSupportedFormat(vk::Format wanted_format, vk::FormatFeatureFlags wanted_usage,
+                                  FormatType format_type) const;
+
+    /// Returns the dispatch loader with direct function pointers of the device
+    const vk::DispatchLoaderDynamic& GetDispatchLoader() const {
+        return dld;
+    }
+
+    /// Returns the logical device
+    vk::Device GetLogical() const {
+        return logical.get();
+    }
+
+    /// Returns the physical device.
+    vk::PhysicalDevice GetPhysical() const {
+        return physical;
+    }
+
+    /// Returns the main graphics queue.
+    vk::Queue GetGraphicsQueue() const {
+        return graphics_queue;
+    }
+
+    /// Returns the main present queue.
+    vk::Queue GetPresentQueue() const {
+        return present_queue;
+    }
+
+    /// Returns main graphics queue family index.
+    u32 GetGraphicsFamily() const {
+        return graphics_family;
+    }
+
+    /// Returns main present queue family index.
+    u32 GetPresentFamily() const {
+        return present_family;
+    }
+
+    /// Returns if the device is integrated with the host CPU
+    bool IsIntegrated() const {
+        return device_type == vk::PhysicalDeviceType::eIntegratedGpu;
+    }
+
+    /// Returns uniform buffer alignment requeriment
+    u64 GetUniformBufferAlignment() const {
+        return uniform_buffer_alignment;
+    }
+
+    /// Checks if the physical device is suitable.
+    static bool IsSuitable(const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDevice physical,
+                           vk::SurfaceKHR surface);
+
+private:
+    /// Sets up queue families.
+    void SetupFamilies(const vk::DispatchLoaderDynamic& dldi, vk::SurfaceKHR surface);
+
+    /// Sets up device properties.
+    void SetupProperties(const vk::DispatchLoaderDynamic& dldi);
+
+    /// Returns a list of queue initialization descriptors.
+    std::vector<vk::DeviceQueueCreateInfo> GetDeviceQueueCreateInfos() const;
+
+    /// Returns true if a format is supported.
+    bool IsFormatSupported(vk::Format wanted_format, vk::FormatFeatureFlags wanted_usage,
+                           FormatType format_type) const;
+
+    /// Returns the device properties for Vulkan formats.
+    static std::map<vk::Format, vk::FormatProperties> GetFormatProperties(
+        const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDevice physical);
+
+    const vk::PhysicalDevice physical;  ///< Physical device
+    vk::DispatchLoaderDynamic dld;      ///< Device function pointers
+    UniqueDevice logical;               ///< Logical device
+    vk::Queue graphics_queue;           ///< Main graphics queue
+    vk::Queue present_queue;            ///< Main present queue
+    u32 graphics_family{};              ///< Main graphics queue family index
+    u32 present_family{};               ///< Main present queue family index
+    vk::PhysicalDeviceType device_type; ///< Physical device type
+    u64 uniform_buffer_alignment{};     ///< Uniform buffer alignment requeriment
+    std::map<vk::Format, vk::FormatProperties> format_properties; ///< Format properties dictionary
+};
+
+} // namespace Vulkan
diff --git a/src/video_core/renderer_vulkan/vk_memory_manager.cpp b/src/video_core/renderer_vulkan/vk_memory_manager.cpp
new file mode 100644
index 000000000..0451babbf
--- /dev/null
+++ b/src/video_core/renderer_vulkan/vk_memory_manager.cpp
@@ -0,0 +1,252 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <algorithm>
+#include <optional>
+#include <tuple>
+#include <vector>
+#include "common/alignment.h"
+#include "common/assert.h"
+#include "common/common_types.h"
+#include "common/logging/log.h"
+#include "video_core/renderer_vulkan/declarations.h"
+#include "video_core/renderer_vulkan/vk_device.h"
+#include "video_core/renderer_vulkan/vk_memory_manager.h"
+
+namespace Vulkan {
+
+// TODO(Rodrigo): Fine tune this number
+constexpr u64 ALLOC_CHUNK_SIZE = 64 * 1024 * 1024;
+
+class VKMemoryAllocation final {
+public:
+    explicit VKMemoryAllocation(const VKDevice& device, vk::DeviceMemory memory,
+                                vk::MemoryPropertyFlags properties, u64 alloc_size, u32 type)
+        : device{device}, memory{memory}, properties{properties}, alloc_size{alloc_size},
+          shifted_type{ShiftType(type)}, is_mappable{properties &
+                                                     vk::MemoryPropertyFlagBits::eHostVisible} {
+        if (is_mappable) {
+            const auto dev = device.GetLogical();
+            const auto& dld = device.GetDispatchLoader();
+            base_address = static_cast<u8*>(dev.mapMemory(memory, 0, alloc_size, {}, dld));
+        }
+    }
+
+    ~VKMemoryAllocation() {
+        const auto dev = device.GetLogical();
+        const auto& dld = device.GetDispatchLoader();
+        if (is_mappable)
+            dev.unmapMemory(memory, dld);
+        dev.free(memory, nullptr, dld);
+    }
+
+    VKMemoryCommit Commit(vk::DeviceSize commit_size, vk::DeviceSize alignment) {
+        auto found = TryFindFreeSection(free_iterator, alloc_size, static_cast<u64>(commit_size),
+                                        static_cast<u64>(alignment));
+        if (!found) {
+            found = TryFindFreeSection(0, free_iterator, static_cast<u64>(commit_size),
+                                       static_cast<u64>(alignment));
+            if (!found) {
+                // Signal out of memory, it'll try to do more allocations.
+                return nullptr;
+            }
+        }
+        u8* address = is_mappable ? base_address + *found : nullptr;
+        auto commit = std::make_unique<VKMemoryCommitImpl>(this, memory, address, *found,
+                                                           *found + commit_size);
+        commits.push_back(commit.get());
+
+        // Last commit's address is highly probable to be free.
+        free_iterator = *found + commit_size;
+
+        return commit;
+    }
+
+    void Free(const VKMemoryCommitImpl* commit) {
+        ASSERT(commit);
+        const auto it =
+            std::find_if(commits.begin(), commits.end(),
+                         [&](const auto& stored_commit) { return stored_commit == commit; });
+        if (it == commits.end()) {
+            LOG_CRITICAL(Render_Vulkan, "Freeing unallocated commit!");
+            UNREACHABLE();
+            return;
+        }
+        commits.erase(it);
+    }
+
+    /// Returns whether this allocation is compatible with the arguments.
+    bool IsCompatible(vk::MemoryPropertyFlags wanted_properties, u32 type_mask) const {
+        return (wanted_properties & properties) != vk::MemoryPropertyFlagBits(0) &&
+               (type_mask & shifted_type) != 0;
+    }
+
+private:
+    static constexpr u32 ShiftType(u32 type) {
+        return 1U << type;
+    }
+
+    /// A memory allocator, it may return a free region between "start" and "end" with the solicited
+    /// requeriments.
+    std::optional<u64> TryFindFreeSection(u64 start, u64 end, u64 size, u64 alignment) const {
+        u64 iterator = start;
+        while (iterator + size < end) {
+            const u64 try_left = Common::AlignUp(iterator, alignment);
+            const u64 try_right = try_left + size;
+
+            bool overlap = false;
+            for (const auto& commit : commits) {
+                const auto [commit_left, commit_right] = commit->interval;
+                if (try_left < commit_right && commit_left < try_right) {
+                    // There's an overlap, continue the search where the overlapping commit ends.
+                    iterator = commit_right;
+                    overlap = true;
+                    break;
+                }
+            }
+            if (!overlap) {
+                // A free address has been found.
+                return try_left;
+            }
+        }
+        // No free regions where found, return an empty optional.
+        return std::nullopt;
+    }
+
+    const VKDevice& device;                   ///< Vulkan device.
+    const vk::DeviceMemory memory;            ///< Vulkan memory allocation handler.
+    const vk::MemoryPropertyFlags properties; ///< Vulkan properties.
+    const u64 alloc_size;                     ///< Size of this allocation.
+    const u32 shifted_type;                   ///< Stored Vulkan type of this allocation, shifted.
+    const bool is_mappable;                   ///< Whether the allocation is mappable.
+
+    /// Base address of the mapped pointer.
+    u8* base_address{};
+
+    /// Hints where the next free region is likely going to be.
+    u64 free_iterator{};
+
+    /// Stores all commits done from this allocation.
+    std::vector<const VKMemoryCommitImpl*> commits;
+};
+
+VKMemoryManager::VKMemoryManager(const VKDevice& device)
+    : device{device}, props{device.GetPhysical().getMemoryProperties(device.GetDispatchLoader())},
+      is_memory_unified{GetMemoryUnified(props)} {}
+
+VKMemoryManager::~VKMemoryManager() = default;
+
+VKMemoryCommit VKMemoryManager::Commit(const vk::MemoryRequirements& reqs, bool host_visible) {
+    ASSERT(reqs.size < ALLOC_CHUNK_SIZE);
+
+    // When a host visible commit is asked, search for host visible and coherent, otherwise search
+    // for a fast device local type.
+    const vk::MemoryPropertyFlags wanted_properties =
+        host_visible
+            ? vk::MemoryPropertyFlagBits::eHostVisible | vk::MemoryPropertyFlagBits::eHostCoherent
+            : vk::MemoryPropertyFlagBits::eDeviceLocal;
+
+    const auto TryCommit = [&]() -> VKMemoryCommit {
+        for (auto& alloc : allocs) {
+            if (!alloc->IsCompatible(wanted_properties, reqs.memoryTypeBits))
+                continue;
+
+            if (auto commit = alloc->Commit(reqs.size, reqs.alignment); commit) {
+                return commit;
+            }
+        }
+        return {};
+    };
+
+    if (auto commit = TryCommit(); commit) {
+        return commit;
+    }
+
+    // Commit has failed, allocate more memory.
+    if (!AllocMemory(wanted_properties, reqs.memoryTypeBits, ALLOC_CHUNK_SIZE)) {
+        // TODO(Rodrigo): Try to use host memory.
+        LOG_CRITICAL(Render_Vulkan, "Ran out of memory!");
+        UNREACHABLE();
+    }
+
+    // Commit again, this time it won't fail since there's a fresh allocation above. If it does,
+    // there's a bug.
+    auto commit = TryCommit();
+    ASSERT(commit);
+    return commit;
+}
+
+VKMemoryCommit VKMemoryManager::Commit(vk::Buffer buffer, bool host_visible) {
+    const auto dev = device.GetLogical();
+    const auto& dld = device.GetDispatchLoader();
+    const auto requeriments = dev.getBufferMemoryRequirements(buffer, dld);
+    auto commit = Commit(requeriments, host_visible);
+    dev.bindBufferMemory(buffer, commit->GetMemory(), commit->GetOffset(), dld);
+    return commit;
+}
+
+VKMemoryCommit VKMemoryManager::Commit(vk::Image image, bool host_visible) {
+    const auto dev = device.GetLogical();
+    const auto& dld = device.GetDispatchLoader();
+    const auto requeriments = dev.getImageMemoryRequirements(image, dld);
+    auto commit = Commit(requeriments, host_visible);
+    dev.bindImageMemory(image, commit->GetMemory(), commit->GetOffset(), dld);
+    return commit;
+}
+
+bool VKMemoryManager::AllocMemory(vk::MemoryPropertyFlags wanted_properties, u32 type_mask,
+                                  u64 size) {
+    const u32 type = [&]() {
+        for (u32 type_index = 0; type_index < props.memoryTypeCount; ++type_index) {
+            const auto flags = props.memoryTypes[type_index].propertyFlags;
+            if ((type_mask & (1U << type_index)) && (flags & wanted_properties)) {
+                // The type matches in type and in the wanted properties.
+                return type_index;
+            }
+        }
+        LOG_CRITICAL(Render_Vulkan, "Couldn't find a compatible memory type!");
+        UNREACHABLE();
+        return 0u;
+    }();
+
+    const auto dev = device.GetLogical();
+    const auto& dld = device.GetDispatchLoader();
+
+    // Try to allocate found type.
+    const vk::MemoryAllocateInfo memory_ai(size, type);
+    vk::DeviceMemory memory;
+    if (const vk::Result res = dev.allocateMemory(&memory_ai, nullptr, &memory, dld);
+        res != vk::Result::eSuccess) {
+        LOG_CRITICAL(Render_Vulkan, "Device allocation failed with code {}!", vk::to_string(res));
+        return false;
+    }
+    allocs.push_back(
+        std::make_unique<VKMemoryAllocation>(device, memory, wanted_properties, size, type));
+    return true;
+}
+
+/*static*/ bool VKMemoryManager::GetMemoryUnified(const vk::PhysicalDeviceMemoryProperties& props) {
+    for (u32 heap_index = 0; heap_index < props.memoryHeapCount; ++heap_index) {
+        if (!(props.memoryHeaps[heap_index].flags & vk::MemoryHeapFlagBits::eDeviceLocal)) {
+            // Memory is considered unified when heaps are device local only.
+            return false;
+        }
+    }
+    return true;
+}
+
+VKMemoryCommitImpl::VKMemoryCommitImpl(VKMemoryAllocation* allocation, vk::DeviceMemory memory,
+                                       u8* data, u64 begin, u64 end)
+    : interval(std::make_pair(begin, end)), memory{memory}, allocation{allocation}, data{data} {}
+
+VKMemoryCommitImpl::~VKMemoryCommitImpl() {
+    allocation->Free(this);
+}
+
+u8* VKMemoryCommitImpl::GetData() const {
+    ASSERT_MSG(data != nullptr, "Trying to access an unmapped commit.");
+    return data;
+}
+
+} // namespace Vulkan
diff --git a/src/video_core/renderer_vulkan/vk_memory_manager.h b/src/video_core/renderer_vulkan/vk_memory_manager.h
new file mode 100644
index 000000000..073597b35
--- /dev/null
+++ b/src/video_core/renderer_vulkan/vk_memory_manager.h
@@ -0,0 +1,87 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <memory>
+#include <utility>
+#include <vector>
+#include "common/common_types.h"
+#include "video_core/renderer_vulkan/declarations.h"
+
+namespace Vulkan {
+
+class VKDevice;
+class VKMemoryAllocation;
+class VKMemoryCommitImpl;
+
+using VKMemoryCommit = std::unique_ptr<VKMemoryCommitImpl>;
+
+class VKMemoryManager final {
+public:
+    explicit VKMemoryManager(const VKDevice& device);
+    ~VKMemoryManager();
+
+    /**
+     * Commits a memory with the specified requeriments.
+     * @param reqs Requeriments returned from a Vulkan call.
+     * @param host_visible Signals the allocator that it *must* use host visible and coherent
+     * memory. When passing false, it will try to allocate device local memory.
+     * @returns A memory commit.
+     */
+    VKMemoryCommit Commit(const vk::MemoryRequirements& reqs, bool host_visible);
+
+    /// Commits memory required by the buffer and binds it.
+    VKMemoryCommit Commit(vk::Buffer buffer, bool host_visible);
+
+    /// Commits memory required by the image and binds it.
+    VKMemoryCommit Commit(vk::Image image, bool host_visible);
+
+    /// Returns true if the memory allocations are done always in host visible and coherent memory.
+    bool IsMemoryUnified() const {
+        return is_memory_unified;
+    }
+
+private:
+    /// Allocates a chunk of memory.
+    bool AllocMemory(vk::MemoryPropertyFlags wanted_properties, u32 type_mask, u64 size);
+
+    /// Returns true if the device uses an unified memory model.
+    static bool GetMemoryUnified(const vk::PhysicalDeviceMemoryProperties& props);
+
+    const VKDevice& device;                                  ///< Device handler.
+    const vk::PhysicalDeviceMemoryProperties props;          ///< Physical device properties.
+    const bool is_memory_unified;                            ///< True if memory model is unified.
+    std::vector<std::unique_ptr<VKMemoryAllocation>> allocs; ///< Current allocations.
+};
+
+class VKMemoryCommitImpl final {
+    friend VKMemoryAllocation;
+
+public:
+    explicit VKMemoryCommitImpl(VKMemoryAllocation* allocation, vk::DeviceMemory memory, u8* data,
+                                u64 begin, u64 end);
+    ~VKMemoryCommitImpl();
+
+    /// Returns the writeable memory map. The commit has to be mappable.
+    u8* GetData() const;
+
+    /// Returns the Vulkan memory handler.
+    vk::DeviceMemory GetMemory() const {
+        return memory;
+    }
+
+    /// Returns the start position of the commit relative to the allocation.
+    vk::DeviceSize GetOffset() const {
+        return static_cast<vk::DeviceSize>(interval.first);
+    }
+
+private:
+    std::pair<u64, u64> interval{};   ///< Interval where the commit exists.
+    vk::DeviceMemory memory;          ///< Vulkan device memory handler.
+    VKMemoryAllocation* allocation{}; ///< Pointer to the large memory allocation.
+    u8* data{}; ///< Pointer to the host mapped memory, it has the commit offset included.
+};
+
+} // namespace Vulkan
diff --git a/src/video_core/renderer_vulkan/vk_resource_manager.cpp b/src/video_core/renderer_vulkan/vk_resource_manager.cpp
new file mode 100644
index 000000000..a1e117443
--- /dev/null
+++ b/src/video_core/renderer_vulkan/vk_resource_manager.cpp
@@ -0,0 +1,285 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <algorithm>
+#include <optional>
+#include "common/assert.h"
+#include "common/logging/log.h"
+#include "video_core/renderer_vulkan/declarations.h"
+#include "video_core/renderer_vulkan/vk_device.h"
+#include "video_core/renderer_vulkan/vk_resource_manager.h"
+
+namespace Vulkan {
+
+// TODO(Rodrigo): Fine tune these numbers.
+constexpr std::size_t COMMAND_BUFFER_POOL_SIZE = 0x1000;
+constexpr std::size_t FENCES_GROW_STEP = 0x40;
+
+class CommandBufferPool final : public VKFencedPool {
+public:
+    CommandBufferPool(const VKDevice& device)
+        : VKFencedPool(COMMAND_BUFFER_POOL_SIZE), device{device} {}
+
+    void Allocate(std::size_t begin, std::size_t end) {
+        const auto dev = device.GetLogical();
+        const auto& dld = device.GetDispatchLoader();
+        const u32 graphics_family = device.GetGraphicsFamily();
+
+        auto pool = std::make_unique<Pool>();
+
+        // Command buffers are going to be commited, recorded, executed every single usage cycle.
+        // They are also going to be reseted when commited.
+        const auto pool_flags = vk::CommandPoolCreateFlagBits::eTransient |
+                                vk::CommandPoolCreateFlagBits::eResetCommandBuffer;
+        const vk::CommandPoolCreateInfo cmdbuf_pool_ci(pool_flags, graphics_family);
+        pool->handle = dev.createCommandPoolUnique(cmdbuf_pool_ci, nullptr, dld);
+
+        const vk::CommandBufferAllocateInfo cmdbuf_ai(*pool->handle,
+                                                      vk::CommandBufferLevel::ePrimary,
+                                                      static_cast<u32>(COMMAND_BUFFER_POOL_SIZE));
+        pool->cmdbufs =
+            dev.allocateCommandBuffersUnique<std::allocator<UniqueCommandBuffer>>(cmdbuf_ai, dld);
+
+        pools.push_back(std::move(pool));
+    }
+
+    vk::CommandBuffer Commit(VKFence& fence) {
+        const std::size_t index = CommitResource(fence);
+        const auto pool_index = index / COMMAND_BUFFER_POOL_SIZE;
+        const auto sub_index = index % COMMAND_BUFFER_POOL_SIZE;
+        return *pools[pool_index]->cmdbufs[sub_index];
+    }
+
+private:
+    struct Pool {
+        UniqueCommandPool handle;
+        std::vector<UniqueCommandBuffer> cmdbufs;
+    };
+
+    const VKDevice& device;
+
+    std::vector<std::unique_ptr<Pool>> pools;
+};
+
+VKResource::VKResource() = default;
+
+VKResource::~VKResource() = default;
+
+VKFence::VKFence(const VKDevice& device, UniqueFence handle)
+    : device{device}, handle{std::move(handle)} {}
+
+VKFence::~VKFence() = default;
+
+void VKFence::Wait() {
+    const auto dev = device.GetLogical();
+    const auto& dld = device.GetDispatchLoader();
+    dev.waitForFences({*handle}, true, std::numeric_limits<u64>::max(), dld);
+}
+
+void VKFence::Release() {
+    is_owned = false;
+}
+
+void VKFence::Commit() {
+    is_owned = true;
+    is_used = true;
+}
+
+bool VKFence::Tick(bool gpu_wait, bool owner_wait) {
+    if (!is_used) {
+        // If a fence is not used it's always free.
+        return true;
+    }
+    if (is_owned && !owner_wait) {
+        // The fence is still being owned (Release has not been called) and ownership wait has
+        // not been asked.
+        return false;
+    }
+
+    const auto dev = device.GetLogical();
+    const auto& dld = device.GetDispatchLoader();
+    if (gpu_wait) {
+        // Wait for the fence if it has been requested.
+        dev.waitForFences({*handle}, true, std::numeric_limits<u64>::max(), dld);
+    } else {
+        if (dev.getFenceStatus(*handle, dld) != vk::Result::eSuccess) {
+            // Vulkan fence is not ready, not much it can do here
+            return false;
+        }
+    }
+
+    // Broadcast resources their free state.
+    for (auto* resource : protected_resources) {
+        resource->OnFenceRemoval(this);
+    }
+    protected_resources.clear();
+
+    // Prepare fence for reusage.
+    dev.resetFences({*handle}, dld);
+    is_used = false;
+    return true;
+}
+
+void VKFence::Protect(VKResource* resource) {
+    protected_resources.push_back(resource);
+}
+
+void VKFence::Unprotect(VKResource* resource) {
+    const auto it = std::find(protected_resources.begin(), protected_resources.end(), resource);
+    ASSERT(it != protected_resources.end());
+
+    resource->OnFenceRemoval(this);
+    protected_resources.erase(it);
+}
+
+VKFenceWatch::VKFenceWatch() = default;
+
+VKFenceWatch::~VKFenceWatch() {
+    if (fence) {
+        fence->Unprotect(this);
+    }
+}
+
+void VKFenceWatch::Wait() {
+    if (fence == nullptr) {
+        return;
+    }
+    fence->Wait();
+    fence->Unprotect(this);
+}
+
+void VKFenceWatch::Watch(VKFence& new_fence) {
+    Wait();
+    fence = &new_fence;
+    fence->Protect(this);
+}
+
+bool VKFenceWatch::TryWatch(VKFence& new_fence) {
+    if (fence) {
+        return false;
+    }
+    fence = &new_fence;
+    fence->Protect(this);
+    return true;
+}
+
+void VKFenceWatch::OnFenceRemoval(VKFence* signaling_fence) {
+    ASSERT_MSG(signaling_fence == fence, "Removing the wrong fence");
+    fence = nullptr;
+}
+
+VKFencedPool::VKFencedPool(std::size_t grow_step) : grow_step{grow_step} {}
+
+VKFencedPool::~VKFencedPool() = default;
+
+std::size_t VKFencedPool::CommitResource(VKFence& fence) {
+    const auto Search = [&](std::size_t begin, std::size_t end) -> std::optional<std::size_t> {
+        for (std::size_t iterator = begin; iterator < end; ++iterator) {
+            if (watches[iterator]->TryWatch(fence)) {
+                // The resource is now being watched, a free resource was successfully found.
+                return iterator;
+            }
+        }
+        return {};
+    };
+    // Try to find a free resource from the hinted position to the end.
+    auto found = Search(free_iterator, watches.size());
+    if (!found) {
+        // Search from beginning to the hinted position.
+        found = Search(0, free_iterator);
+        if (!found) {
+            // Both searches failed, the pool is full; handle it.
+            const std::size_t free_resource = ManageOverflow();
+
+            // Watch will wait for the resource to be free.
+            watches[free_resource]->Watch(fence);
+            found = free_resource;
+        }
+    }
+    // Free iterator is hinted to the resource after the one that's been commited.
+    free_iterator = (*found + 1) % watches.size();
+    return *found;
+}
+
+std::size_t VKFencedPool::ManageOverflow() {
+    const std::size_t old_capacity = watches.size();
+    Grow();
+
+    // The last entry is guaranted to be free, since it's the first element of the freshly
+    // allocated resources.
+    return old_capacity;
+}
+
+void VKFencedPool::Grow() {
+    const std::size_t old_capacity = watches.size();
+    watches.resize(old_capacity + grow_step);
+    std::generate(watches.begin() + old_capacity, watches.end(),
+                  []() { return std::make_unique<VKFenceWatch>(); });
+    Allocate(old_capacity, old_capacity + grow_step);
+}
+
+VKResourceManager::VKResourceManager(const VKDevice& device) : device{device} {
+    GrowFences(FENCES_GROW_STEP);
+    command_buffer_pool = std::make_unique<CommandBufferPool>(device);
+}
+
+VKResourceManager::~VKResourceManager() = default;
+
+VKFence& VKResourceManager::CommitFence() {
+    const auto StepFences = [&](bool gpu_wait, bool owner_wait) -> VKFence* {
+        const auto Tick = [=](auto& fence) { return fence->Tick(gpu_wait, owner_wait); };
+        const auto hinted = fences.begin() + fences_iterator;
+
+        auto it = std::find_if(hinted, fences.end(), Tick);
+        if (it == fences.end()) {
+            it = std::find_if(fences.begin(), hinted, Tick);
+            if (it == hinted) {
+                return nullptr;
+            }
+        }
+        fences_iterator = std::distance(fences.begin(), it) + 1;
+        if (fences_iterator >= fences.size())
+            fences_iterator = 0;
+
+        auto& fence = *it;
+        fence->Commit();
+        return fence.get();
+    };
+
+    VKFence* found_fence = StepFences(false, false);
+    if (!found_fence) {
+        // Try again, this time waiting.
+        found_fence = StepFences(true, false);
+
+        if (!found_fence) {
+            // Allocate new fences and try again.
+            LOG_INFO(Render_Vulkan, "Allocating new fences {} -> {}", fences.size(),
+                     fences.size() + FENCES_GROW_STEP);
+
+            GrowFences(FENCES_GROW_STEP);
+            found_fence = StepFences(true, false);
+            ASSERT(found_fence != nullptr);
+        }
+    }
+    return *found_fence;
+}
+
+vk::CommandBuffer VKResourceManager::CommitCommandBuffer(VKFence& fence) {
+    return command_buffer_pool->Commit(fence);
+}
+
+void VKResourceManager::GrowFences(std::size_t new_fences_count) {
+    const auto dev = device.GetLogical();
+    const auto& dld = device.GetDispatchLoader();
+    const vk::FenceCreateInfo fence_ci;
+
+    const std::size_t previous_size = fences.size();
+    fences.resize(previous_size + new_fences_count);
+
+    std::generate(fences.begin() + previous_size, fences.end(), [&]() {
+        return std::make_unique<VKFence>(device, dev.createFenceUnique(fence_ci, nullptr, dld));
+    });
+}
+
+} // namespace Vulkan
diff --git a/src/video_core/renderer_vulkan/vk_resource_manager.h b/src/video_core/renderer_vulkan/vk_resource_manager.h
new file mode 100644
index 000000000..5bfe4cead
--- /dev/null
+++ b/src/video_core/renderer_vulkan/vk_resource_manager.h
@@ -0,0 +1,180 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <cstddef>
+#include <memory>
+#include <vector>
+#include "video_core/renderer_vulkan/declarations.h"
+
+namespace Vulkan {
+
+class VKDevice;
+class VKFence;
+class VKResourceManager;
+
+class CommandBufferPool;
+
+/// Interface for a Vulkan resource
+class VKResource {
+public:
+    explicit VKResource();
+    virtual ~VKResource();
+
+    /**
+     * Signals the object that an owning fence has been signaled.
+     * @param signaling_fence Fence that signals its usage end.
+     */
+    virtual void OnFenceRemoval(VKFence* signaling_fence) = 0;
+};
+
+/**
+ * Fences take ownership of objects, protecting them from GPU-side or driver-side concurrent access.
+ * They must be commited from the resource manager. Their usage flow is: commit the fence from the
+ * resource manager, protect resources with it and use them, send the fence to an execution queue
+ * and Wait for it if needed and then call Release. Used resources will automatically be signaled
+ * when they are free to be reused.
+ * @brief Protects resources for concurrent usage and signals its release.
+ */
+class VKFence {
+    friend class VKResourceManager;
+
+public:
+    explicit VKFence(const VKDevice& device, UniqueFence handle);
+    ~VKFence();
+
+    /**
+     * Waits for the fence to be signaled.
+     * @warning You must have ownership of the fence and it has to be previously sent to a queue to
+     * call this function.
+     */
+    void Wait();
+
+    /**
+     * Releases ownership of the fence. Pass after it has been sent to an execution queue.
+     * Unmanaged usage of the fence after the call will result in undefined behavior because it may
+     * be being used for something else.
+     */
+    void Release();
+
+    /// Protects a resource with this fence.
+    void Protect(VKResource* resource);
+
+    /// Removes protection for a resource.
+    void Unprotect(VKResource* resource);
+
+    /// Retreives the fence.
+    operator vk::Fence() const {
+        return *handle;
+    }
+
+private:
+    /// Take ownership of the fence.
+    void Commit();
+
+    /**
+     * Updates the fence status.
+     * @warning Waiting for the owner might soft lock the execution.
+     * @param gpu_wait Wait for the fence to be signaled by the driver.
+     * @param owner_wait Wait for the owner to signal its freedom.
+     * @returns True if the fence is free. Waiting for gpu and owner will always return true.
+     */
+    bool Tick(bool gpu_wait, bool owner_wait);
+
+    const VKDevice& device;                       ///< Device handler
+    UniqueFence handle;                           ///< Vulkan fence
+    std::vector<VKResource*> protected_resources; ///< List of resources protected by this fence
+    bool is_owned = false; ///< The fence has been commited but not released yet.
+    bool is_used = false;  ///< The fence has been commited but it has not been checked to be free.
+};
+
+/**
+ * A fence watch is used to keep track of the usage of a fence and protect a resource or set of
+ * resources without having to inherit VKResource from their handlers.
+ */
+class VKFenceWatch final : public VKResource {
+public:
+    explicit VKFenceWatch();
+    ~VKFenceWatch();
+
+    /// Waits for the fence to be released.
+    void Wait();
+
+    /**
+     * Waits for a previous fence and watches a new one.
+     * @param new_fence New fence to wait to.
+     */
+    void Watch(VKFence& new_fence);
+
+    /**
+     * Checks if it's currently being watched and starts watching it if it's available.
+     * @returns True if a watch has started, false if it's being watched.
+     */
+    bool TryWatch(VKFence& new_fence);
+
+    void OnFenceRemoval(VKFence* signaling_fence) override;
+
+private:
+    VKFence* fence{}; ///< Fence watching this resource. nullptr when the watch is free.
+};
+
+/**
+ * Handles a pool of resources protected by fences. Manages resource overflow allocating more
+ * resources.
+ */
+class VKFencedPool {
+public:
+    explicit VKFencedPool(std::size_t grow_step);
+    virtual ~VKFencedPool();
+
+protected:
+    /**
+     * Commits a free resource and protects it with a fence. It may allocate new resources.
+     * @param fence Fence that protects the commited resource.
+     * @returns Index of the resource commited.
+     */
+    std::size_t CommitResource(VKFence& fence);
+
+    /// Called when a chunk of resources have to be allocated.
+    virtual void Allocate(std::size_t begin, std::size_t end) = 0;
+
+private:
+    /// Manages pool overflow allocating new resources.
+    std::size_t ManageOverflow();
+
+    /// Allocates a new page of resources.
+    void Grow();
+
+    std::size_t grow_step = 0;     ///< Number of new resources created after an overflow
+    std::size_t free_iterator = 0; ///< Hint to where the next free resources is likely to be found
+    std::vector<std::unique_ptr<VKFenceWatch>> watches; ///< Set of watched resources
+};
+
+/**
+ * The resource manager handles all resources that can be protected with a fence avoiding
+ * driver-side or GPU-side concurrent usage. Usage is documented in VKFence.
+ */
+class VKResourceManager final {
+public:
+    explicit VKResourceManager(const VKDevice& device);
+    ~VKResourceManager();
+
+    /// Commits a fence. It has to be sent to a queue and released.
+    VKFence& CommitFence();
+
+    /// Commits an unused command buffer and protects it with a fence.
+    vk::CommandBuffer CommitCommandBuffer(VKFence& fence);
+
+private:
+    /// Allocates new fences.
+    void GrowFences(std::size_t new_fences_count);
+
+    const VKDevice& device;          ///< Device handler.
+    std::size_t fences_iterator = 0; ///< Index where a free fence is likely to be found.
+    std::vector<std::unique_ptr<VKFence>> fences;           ///< Pool of fences.
+    std::unique_ptr<CommandBufferPool> command_buffer_pool; ///< Pool of command buffers.
+};
+
+} // namespace Vulkan
diff --git a/src/video_core/renderer_vulkan/vk_sampler_cache.cpp b/src/video_core/renderer_vulkan/vk_sampler_cache.cpp
new file mode 100644
index 000000000..ed3178f09
--- /dev/null
+++ b/src/video_core/renderer_vulkan/vk_sampler_cache.cpp
@@ -0,0 +1,81 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <cstring>
+#include <optional>
+#include <unordered_map>
+
+#include "common/assert.h"
+#include "common/cityhash.h"
+#include "video_core/renderer_vulkan/declarations.h"
+#include "video_core/renderer_vulkan/maxwell_to_vk.h"
+#include "video_core/renderer_vulkan/vk_sampler_cache.h"
+#include "video_core/textures/texture.h"
+
+namespace Vulkan {
+
+static std::optional<vk::BorderColor> TryConvertBorderColor(std::array<float, 4> color) {
+    // TODO(Rodrigo): Manage integer border colors
+    if (color == std::array<float, 4>{0, 0, 0, 0}) {
+        return vk::BorderColor::eFloatTransparentBlack;
+    } else if (color == std::array<float, 4>{0, 0, 0, 1}) {
+        return vk::BorderColor::eFloatOpaqueBlack;
+    } else if (color == std::array<float, 4>{1, 1, 1, 1}) {
+        return vk::BorderColor::eFloatOpaqueWhite;
+    } else {
+        return {};
+    }
+}
+
+std::size_t SamplerCacheKey::Hash() const {
+    static_assert(sizeof(raw) % sizeof(u64) == 0);
+    return static_cast<std::size_t>(
+        Common::CityHash64(reinterpret_cast<const char*>(raw.data()), sizeof(raw) / sizeof(u64)));
+}
+
+bool SamplerCacheKey::operator==(const SamplerCacheKey& rhs) const {
+    return raw == rhs.raw;
+}
+
+VKSamplerCache::VKSamplerCache(const VKDevice& device) : device{device} {}
+
+VKSamplerCache::~VKSamplerCache() = default;
+
+vk::Sampler VKSamplerCache::GetSampler(const Tegra::Texture::TSCEntry& tsc) {
+    const auto [entry, is_cache_miss] = cache.try_emplace(SamplerCacheKey{tsc});
+    auto& sampler = entry->second;
+    if (is_cache_miss) {
+        sampler = CreateSampler(tsc);
+    }
+    return *sampler;
+}
+
+UniqueSampler VKSamplerCache::CreateSampler(const Tegra::Texture::TSCEntry& tsc) {
+    const float max_anisotropy = tsc.GetMaxAnisotropy();
+    const bool has_anisotropy = max_anisotropy > 1.0f;
+
+    const auto border_color = tsc.GetBorderColor();
+    const auto vk_border_color = TryConvertBorderColor(border_color);
+    UNIMPLEMENTED_IF_MSG(!vk_border_color, "Unimplemented border color {} {} {} {}",
+                         border_color[0], border_color[1], border_color[2], border_color[3]);
+
+    constexpr bool unnormalized_coords = false;
+
+    const vk::SamplerCreateInfo sampler_ci(
+        {}, MaxwellToVK::Sampler::Filter(tsc.mag_filter),
+        MaxwellToVK::Sampler::Filter(tsc.min_filter),
+        MaxwellToVK::Sampler::MipmapMode(tsc.mipmap_filter),
+        MaxwellToVK::Sampler::WrapMode(tsc.wrap_u), MaxwellToVK::Sampler::WrapMode(tsc.wrap_v),
+        MaxwellToVK::Sampler::WrapMode(tsc.wrap_p), tsc.GetLodBias(), has_anisotropy,
+        max_anisotropy, tsc.depth_compare_enabled,
+        MaxwellToVK::Sampler::DepthCompareFunction(tsc.depth_compare_func), tsc.GetMinLod(),
+        tsc.GetMaxLod(), vk_border_color.value_or(vk::BorderColor::eFloatTransparentBlack),
+        unnormalized_coords);
+
+    const auto& dld = device.GetDispatchLoader();
+    const auto dev = device.GetLogical();
+    return dev.createSamplerUnique(sampler_ci, nullptr, dld);
+}
+
+} // namespace Vulkan
diff --git a/src/video_core/renderer_vulkan/vk_sampler_cache.h b/src/video_core/renderer_vulkan/vk_sampler_cache.h
new file mode 100644
index 000000000..c6394dc87
--- /dev/null
+++ b/src/video_core/renderer_vulkan/vk_sampler_cache.h
@@ -0,0 +1,56 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <unordered_map>
+
+#include "common/common_types.h"
+#include "video_core/renderer_vulkan/declarations.h"
+#include "video_core/textures/texture.h"
+
+namespace Vulkan {
+
+class VKDevice;
+
+struct SamplerCacheKey final : public Tegra::Texture::TSCEntry {
+    std::size_t Hash() const;
+
+    bool operator==(const SamplerCacheKey& rhs) const;
+
+    bool operator!=(const SamplerCacheKey& rhs) const {
+        return !operator==(rhs);
+    }
+};
+
+} // namespace Vulkan
+
+namespace std {
+
+template <>
+struct hash<Vulkan::SamplerCacheKey> {
+    std::size_t operator()(const Vulkan::SamplerCacheKey& k) const noexcept {
+        return k.Hash();
+    }
+};
+
+} // namespace std
+
+namespace Vulkan {
+
+class VKSamplerCache {
+public:
+    explicit VKSamplerCache(const VKDevice& device);
+    ~VKSamplerCache();
+
+    vk::Sampler GetSampler(const Tegra::Texture::TSCEntry& tsc);
+
+private:
+    UniqueSampler CreateSampler(const Tegra::Texture::TSCEntry& tsc);
+
+    const VKDevice& device;
+    std::unordered_map<SamplerCacheKey, UniqueSampler> cache;
+};
+
+} // namespace Vulkan
diff --git a/src/video_core/renderer_vulkan/vk_scheduler.cpp b/src/video_core/renderer_vulkan/vk_scheduler.cpp
new file mode 100644
index 000000000..f1fea1871
--- /dev/null
+++ b/src/video_core/renderer_vulkan/vk_scheduler.cpp
@@ -0,0 +1,60 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/assert.h"
+#include "common/logging/log.h"
+#include "video_core/renderer_vulkan/declarations.h"
+#include "video_core/renderer_vulkan/vk_device.h"
+#include "video_core/renderer_vulkan/vk_resource_manager.h"
+#include "video_core/renderer_vulkan/vk_scheduler.h"
+
+namespace Vulkan {
+
+VKScheduler::VKScheduler(const VKDevice& device, VKResourceManager& resource_manager)
+    : device{device}, resource_manager{resource_manager} {
+    next_fence = &resource_manager.CommitFence();
+    AllocateNewContext();
+}
+
+VKScheduler::~VKScheduler() = default;
+
+VKExecutionContext VKScheduler::GetExecutionContext() const {
+    return VKExecutionContext(current_fence, current_cmdbuf);
+}
+
+VKExecutionContext VKScheduler::Flush(vk::Semaphore semaphore) {
+    SubmitExecution(semaphore);
+    current_fence->Release();
+    AllocateNewContext();
+    return GetExecutionContext();
+}
+
+VKExecutionContext VKScheduler::Finish(vk::Semaphore semaphore) {
+    SubmitExecution(semaphore);
+    current_fence->Wait();
+    current_fence->Release();
+    AllocateNewContext();
+    return GetExecutionContext();
+}
+
+void VKScheduler::SubmitExecution(vk::Semaphore semaphore) {
+    const auto& dld = device.GetDispatchLoader();
+    current_cmdbuf.end(dld);
+
+    const auto queue = device.GetGraphicsQueue();
+    const vk::SubmitInfo submit_info(0, nullptr, nullptr, 1, &current_cmdbuf, semaphore ? 1u : 0u,
+                                     &semaphore);
+    queue.submit({submit_info}, *current_fence, dld);
+}
+
+void VKScheduler::AllocateNewContext() {
+    current_fence = next_fence;
+    current_cmdbuf = resource_manager.CommitCommandBuffer(*current_fence);
+    next_fence = &resource_manager.CommitFence();
+
+    const auto& dld = device.GetDispatchLoader();
+    current_cmdbuf.begin({vk::CommandBufferUsageFlagBits::eOneTimeSubmit}, dld);
+}
+
+} // namespace Vulkan
diff --git a/src/video_core/renderer_vulkan/vk_scheduler.h b/src/video_core/renderer_vulkan/vk_scheduler.h
new file mode 100644
index 000000000..cfaf5376f
--- /dev/null
+++ b/src/video_core/renderer_vulkan/vk_scheduler.h
@@ -0,0 +1,69 @@
+// 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"
+#include "video_core/renderer_vulkan/declarations.h"
+
+namespace Vulkan {
+
+class VKDevice;
+class VKExecutionContext;
+class VKFence;
+class VKResourceManager;
+
+/// The scheduler abstracts command buffer and fence management with an interface that's able to do
+/// OpenGL-like operations on Vulkan command buffers.
+class VKScheduler {
+public:
+    explicit VKScheduler(const VKDevice& device, VKResourceManager& resource_manager);
+    ~VKScheduler();
+
+    /// Gets the current execution context.
+    [[nodiscard]] VKExecutionContext GetExecutionContext() const;
+
+    /// Sends the current execution context to the GPU. It invalidates the current execution context
+    /// and returns a new one.
+    VKExecutionContext Flush(vk::Semaphore semaphore = nullptr);
+
+    /// Sends the current execution context to the GPU and waits for it to complete. It invalidates
+    /// the current execution context and returns a new one.
+    VKExecutionContext Finish(vk::Semaphore semaphore = nullptr);
+
+private:
+    void SubmitExecution(vk::Semaphore semaphore);
+
+    void AllocateNewContext();
+
+    const VKDevice& device;
+    VKResourceManager& resource_manager;
+    vk::CommandBuffer current_cmdbuf;
+    VKFence* current_fence = nullptr;
+    VKFence* next_fence = nullptr;
+};
+
+class VKExecutionContext {
+    friend class VKScheduler;
+
+public:
+    VKExecutionContext() = default;
+
+    VKFence& GetFence() const {
+        return *fence;
+    }
+
+    vk::CommandBuffer GetCommandBuffer() const {
+        return cmdbuf;
+    }
+
+private:
+    explicit VKExecutionContext(VKFence* fence, vk::CommandBuffer cmdbuf)
+        : fence{fence}, cmdbuf{cmdbuf} {}
+
+    VKFence* fence{};
+    vk::CommandBuffer cmdbuf;
+};
+
+} // namespace Vulkan
diff --git a/src/video_core/renderer_vulkan/vk_stream_buffer.cpp b/src/video_core/renderer_vulkan/vk_stream_buffer.cpp
new file mode 100644
index 000000000..58ffa42f2
--- /dev/null
+++ b/src/video_core/renderer_vulkan/vk_stream_buffer.cpp
@@ -0,0 +1,90 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <algorithm>
+#include <memory>
+#include <optional>
+#include <vector>
+
+#include "common/assert.h"
+#include "video_core/renderer_vulkan/declarations.h"
+#include "video_core/renderer_vulkan/vk_device.h"
+#include "video_core/renderer_vulkan/vk_memory_manager.h"
+#include "video_core/renderer_vulkan/vk_resource_manager.h"
+#include "video_core/renderer_vulkan/vk_scheduler.h"
+#include "video_core/renderer_vulkan/vk_stream_buffer.h"
+
+namespace Vulkan {
+
+constexpr u64 WATCHES_INITIAL_RESERVE = 0x4000;
+constexpr u64 WATCHES_RESERVE_CHUNK = 0x1000;
+
+VKStreamBuffer::VKStreamBuffer(const VKDevice& device, VKMemoryManager& memory_manager,
+                               VKScheduler& scheduler, u64 size, vk::BufferUsageFlags usage,
+                               vk::AccessFlags access, vk::PipelineStageFlags pipeline_stage)
+    : device{device}, scheduler{scheduler}, buffer_size{size}, access{access}, pipeline_stage{
+                                                                                   pipeline_stage} {
+    CreateBuffers(memory_manager, usage);
+    ReserveWatches(WATCHES_INITIAL_RESERVE);
+}
+
+VKStreamBuffer::~VKStreamBuffer() = default;
+
+std::tuple<u8*, u64, bool> VKStreamBuffer::Reserve(u64 size) {
+    ASSERT(size <= buffer_size);
+    mapped_size = size;
+
+    if (offset + size > buffer_size) {
+        // The buffer would overflow, save the amount of used buffers, signal an invalidation and
+        // reset the state.
+        invalidation_mark = used_watches;
+        used_watches = 0;
+        offset = 0;
+    }
+
+    return {mapped_pointer + offset, offset, invalidation_mark.has_value()};
+}
+
+VKExecutionContext VKStreamBuffer::Send(VKExecutionContext exctx, u64 size) {
+    ASSERT_MSG(size <= mapped_size, "Reserved size is too small");
+
+    if (invalidation_mark) {
+        // TODO(Rodrigo): Find a better way to invalidate than waiting for all watches to finish.
+        exctx = scheduler.Flush();
+        std::for_each(watches.begin(), watches.begin() + *invalidation_mark,
+                      [&](auto& resource) { resource->Wait(); });
+        invalidation_mark = std::nullopt;
+    }
+
+    if (used_watches + 1 >= watches.size()) {
+        // Ensure that there are enough watches.
+        ReserveWatches(WATCHES_RESERVE_CHUNK);
+    }
+    // Add a watch for this allocation.
+    watches[used_watches++]->Watch(exctx.GetFence());
+
+    offset += size;
+
+    return exctx;
+}
+
+void VKStreamBuffer::CreateBuffers(VKMemoryManager& memory_manager, vk::BufferUsageFlags usage) {
+    const vk::BufferCreateInfo buffer_ci({}, buffer_size, usage, vk::SharingMode::eExclusive, 0,
+                                         nullptr);
+
+    const auto dev = device.GetLogical();
+    const auto& dld = device.GetDispatchLoader();
+    buffer = dev.createBufferUnique(buffer_ci, nullptr, dld);
+    commit = memory_manager.Commit(*buffer, true);
+    mapped_pointer = commit->GetData();
+}
+
+void VKStreamBuffer::ReserveWatches(std::size_t grow_size) {
+    const std::size_t previous_size = watches.size();
+    watches.resize(previous_size + grow_size);
+    std::generate(watches.begin() + previous_size, watches.end(),
+                  []() { return std::make_unique<VKFenceWatch>(); });
+}
+
+} // namespace Vulkan
diff --git a/src/video_core/renderer_vulkan/vk_stream_buffer.h b/src/video_core/renderer_vulkan/vk_stream_buffer.h
new file mode 100644
index 000000000..69d036ccd
--- /dev/null
+++ b/src/video_core/renderer_vulkan/vk_stream_buffer.h
@@ -0,0 +1,72 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <memory>
+#include <optional>
+#include <tuple>
+#include <vector>
+
+#include "common/common_types.h"
+#include "video_core/renderer_vulkan/declarations.h"
+#include "video_core/renderer_vulkan/vk_memory_manager.h"
+
+namespace Vulkan {
+
+class VKDevice;
+class VKFence;
+class VKFenceWatch;
+class VKResourceManager;
+class VKScheduler;
+
+class VKStreamBuffer {
+public:
+    explicit VKStreamBuffer(const VKDevice& device, VKMemoryManager& memory_manager,
+                            VKScheduler& scheduler, u64 size, vk::BufferUsageFlags usage,
+                            vk::AccessFlags access, vk::PipelineStageFlags pipeline_stage);
+    ~VKStreamBuffer();
+
+    /**
+     * Reserves a region of memory from the stream buffer.
+     * @param size Size to reserve.
+     * @returns A tuple in the following order: Raw memory pointer (with offset added), buffer
+     * offset and a boolean that's true when buffer has been invalidated.
+     */
+    std::tuple<u8*, u64, bool> Reserve(u64 size);
+
+    /// Ensures that "size" bytes of memory are available to the GPU, potentially recording a copy.
+    [[nodiscard]] VKExecutionContext Send(VKExecutionContext exctx, u64 size);
+
+    vk::Buffer GetBuffer() const {
+        return *buffer;
+    }
+
+private:
+    /// Creates Vulkan buffer handles committing the required the required memory.
+    void CreateBuffers(VKMemoryManager& memory_manager, vk::BufferUsageFlags usage);
+
+    /// Increases the amount of watches available.
+    void ReserveWatches(std::size_t grow_size);
+
+    const VKDevice& device;                      ///< Vulkan device manager.
+    VKScheduler& scheduler;                      ///< Command scheduler.
+    const u64 buffer_size;                       ///< Total size of the stream buffer.
+    const vk::AccessFlags access;                ///< Access usage of this stream buffer.
+    const vk::PipelineStageFlags pipeline_stage; ///< Pipeline usage of this stream buffer.
+
+    UniqueBuffer buffer;   ///< Mapped buffer.
+    VKMemoryCommit commit; ///< Memory commit.
+    u8* mapped_pointer{};  ///< Pointer to the host visible commit
+
+    u64 offset{};      ///< Buffer iterator.
+    u64 mapped_size{}; ///< Size reserved for the current copy.
+
+    std::vector<std::unique_ptr<VKFenceWatch>> watches; ///< Total watches
+    std::size_t used_watches{}; ///< Count of watches, reset on invalidation.
+    std::optional<std::size_t>
+        invalidation_mark{}; ///< Number of watches used in the current invalidation.
+};
+
+} // namespace Vulkan
diff --git a/src/video_core/shader/decode.cpp b/src/video_core/shader/decode.cpp
index 812983a99..e4c438792 100644
--- a/src/video_core/shader/decode.cpp
+++ b/src/video_core/shader/decode.cpp
@@ -121,15 +121,15 @@ ExitMethod ShaderIR::Scan(u32 begin, u32 end, std::set<u32>& labels) {
     return exit_method = ExitMethod::AlwaysReturn;
 }
 
-BasicBlock ShaderIR::DecodeRange(u32 begin, u32 end) {
-    BasicBlock basic_block;
+NodeBlock ShaderIR::DecodeRange(u32 begin, u32 end) {
+    NodeBlock 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) {
+u32 ShaderIR::DecodeInstr(NodeBlock& bb, u32 pc) {
     // Ignore sched instructions when generating code.
     if (IsSchedInstruction(pc, main_offset)) {
         return pc + 1;
@@ -151,39 +151,39 @@ u32 ShaderIR::DecodeInstr(BasicBlock& bb, u32 pc) {
     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},
-        };
+    static const std::map<OpCode::Type, u32 (ShaderIR::*)(NodeBlock&, 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::Texture, &ShaderIR::DecodeTexture},
+        {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);
+        pc = (this->*decoder->second)(tmp_block, pc);
     } else {
-        pc = DecodeOther(tmp_block, bb, pc);
+        pc = DecodeOther(tmp_block, pc);
     }
 
     // Some instructions (like SSY) don't have a predicate field, they are always unconditionally
@@ -192,11 +192,14 @@ u32 ShaderIR::DecodeInstr(BasicBlock& bb, u32 pc) {
     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)));
+        const Node conditional =
+            Conditional(GetPredicate(pred_index, instr.negate_pred != 0), std::move(tmp_block));
+        global_code.push_back(conditional);
+        bb.push_back(conditional);
     } else {
         for (auto& node : tmp_block) {
-            bb.push_back(std::move(node));
+            global_code.push_back(node);
+            bb.push_back(node);
         }
     }
 
diff --git a/src/video_core/shader/decode/arithmetic.cpp b/src/video_core/shader/decode/arithmetic.cpp
index 51b8d55d4..3190e2d7c 100644
--- a/src/video_core/shader/decode/arithmetic.cpp
+++ b/src/video_core/shader/decode/arithmetic.cpp
@@ -13,7 +13,7 @@ using Tegra::Shader::Instruction;
 using Tegra::Shader::OpCode;
 using Tegra::Shader::SubOp;
 
-u32 ShaderIR::DecodeArithmetic(BasicBlock& bb, const BasicBlock& code, u32 pc) {
+u32 ShaderIR::DecodeArithmetic(NodeBlock& bb, u32 pc) {
     const Instruction instr = {program_code[pc]};
     const auto opcode = OpCode::Decode(instr);
 
diff --git a/src/video_core/shader/decode/arithmetic_half.cpp b/src/video_core/shader/decode/arithmetic_half.cpp
index 37eef2bf2..baee89107 100644
--- a/src/video_core/shader/decode/arithmetic_half.cpp
+++ b/src/video_core/shader/decode/arithmetic_half.cpp
@@ -12,7 +12,7 @@ namespace VideoCommon::Shader {
 using Tegra::Shader::Instruction;
 using Tegra::Shader::OpCode;
 
-u32 ShaderIR::DecodeArithmeticHalf(BasicBlock& bb, const BasicBlock& code, u32 pc) {
+u32 ShaderIR::DecodeArithmeticHalf(NodeBlock& bb, u32 pc) {
     const Instruction instr = {program_code[pc]};
     const auto opcode = OpCode::Decode(instr);
 
diff --git a/src/video_core/shader/decode/arithmetic_half_immediate.cpp b/src/video_core/shader/decode/arithmetic_half_immediate.cpp
index 7b4f7d284..c2164ba50 100644
--- a/src/video_core/shader/decode/arithmetic_half_immediate.cpp
+++ b/src/video_core/shader/decode/arithmetic_half_immediate.cpp
@@ -12,7 +12,7 @@ namespace VideoCommon::Shader {
 using Tegra::Shader::Instruction;
 using Tegra::Shader::OpCode;
 
-u32 ShaderIR::DecodeArithmeticHalfImmediate(BasicBlock& bb, const BasicBlock& code, u32 pc) {
+u32 ShaderIR::DecodeArithmeticHalfImmediate(NodeBlock& bb, u32 pc) {
     const Instruction instr = {program_code[pc]};
     const auto opcode = OpCode::Decode(instr);
 
diff --git a/src/video_core/shader/decode/arithmetic_immediate.cpp b/src/video_core/shader/decode/arithmetic_immediate.cpp
index 4fd3db54e..0d139c0d2 100644
--- a/src/video_core/shader/decode/arithmetic_immediate.cpp
+++ b/src/video_core/shader/decode/arithmetic_immediate.cpp
@@ -12,7 +12,7 @@ namespace VideoCommon::Shader {
 using Tegra::Shader::Instruction;
 using Tegra::Shader::OpCode;
 
-u32 ShaderIR::DecodeArithmeticImmediate(BasicBlock& bb, const BasicBlock& code, u32 pc) {
+u32 ShaderIR::DecodeArithmeticImmediate(NodeBlock& bb, u32 pc) {
     const Instruction instr = {program_code[pc]};
     const auto opcode = OpCode::Decode(instr);
 
diff --git a/src/video_core/shader/decode/arithmetic_integer.cpp b/src/video_core/shader/decode/arithmetic_integer.cpp
index cc9a76a19..9fd4b273e 100644
--- a/src/video_core/shader/decode/arithmetic_integer.cpp
+++ b/src/video_core/shader/decode/arithmetic_integer.cpp
@@ -15,7 +15,7 @@ using Tegra::Shader::OpCode;
 using Tegra::Shader::Pred;
 using Tegra::Shader::Register;
 
-u32 ShaderIR::DecodeArithmeticInteger(BasicBlock& bb, const BasicBlock& code, u32 pc) {
+u32 ShaderIR::DecodeArithmeticInteger(NodeBlock& bb, u32 pc) {
     const Instruction instr = {program_code[pc]};
     const auto opcode = OpCode::Decode(instr);
 
@@ -41,7 +41,7 @@ u32 ShaderIR::DecodeArithmeticInteger(BasicBlock& bb, const BasicBlock& code, u3
 
         const Node value = Operation(OperationCode::IAdd, PRECISE, op_a, op_b);
 
-        SetInternalFlagsFromInteger(bb, value, instr.op_32.generates_cc);
+        SetInternalFlagsFromInteger(bb, value, instr.generates_cc);
         SetRegister(bb, instr.gpr0, value);
         break;
     }
@@ -242,7 +242,7 @@ u32 ShaderIR::DecodeArithmeticInteger(BasicBlock& bb, const BasicBlock& code, u3
     return pc;
 }
 
-void ShaderIR::WriteLop3Instruction(BasicBlock& bb, Register dest, Node op_a, Node op_b, Node op_c,
+void ShaderIR::WriteLop3Instruction(NodeBlock& 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);
@@ -284,4 +284,4 @@ void ShaderIR::WriteLop3Instruction(BasicBlock& bb, Register dest, Node op_a, No
     SetRegister(bb, dest, value);
 }
 
-} // namespace VideoCommon::Shader
\ No newline at end of file
+} // namespace VideoCommon::Shader
diff --git a/src/video_core/shader/decode/arithmetic_integer_immediate.cpp b/src/video_core/shader/decode/arithmetic_integer_immediate.cpp
index b26a6e473..3ed5ccc5a 100644
--- a/src/video_core/shader/decode/arithmetic_integer_immediate.cpp
+++ b/src/video_core/shader/decode/arithmetic_integer_immediate.cpp
@@ -16,7 +16,7 @@ using Tegra::Shader::Pred;
 using Tegra::Shader::PredicateResultMode;
 using Tegra::Shader::Register;
 
-u32 ShaderIR::DecodeArithmeticIntegerImmediate(BasicBlock& bb, const BasicBlock& code, u32 pc) {
+u32 ShaderIR::DecodeArithmeticIntegerImmediate(NodeBlock& bb, u32 pc) {
     const Instruction instr = {program_code[pc]};
     const auto opcode = OpCode::Decode(instr);
 
@@ -54,9 +54,9 @@ u32 ShaderIR::DecodeArithmeticIntegerImmediate(BasicBlock& bb, const BasicBlock&
     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) {
+void ShaderIR::WriteLogicOperation(NodeBlock& 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:
diff --git a/src/video_core/shader/decode/bfe.cpp b/src/video_core/shader/decode/bfe.cpp
index 0734141b0..6a95dc928 100644
--- a/src/video_core/shader/decode/bfe.cpp
+++ b/src/video_core/shader/decode/bfe.cpp
@@ -12,7 +12,7 @@ namespace VideoCommon::Shader {
 using Tegra::Shader::Instruction;
 using Tegra::Shader::OpCode;
 
-u32 ShaderIR::DecodeBfe(BasicBlock& bb, const BasicBlock& code, u32 pc) {
+u32 ShaderIR::DecodeBfe(NodeBlock& bb, u32 pc) {
     const Instruction instr = {program_code[pc]};
     const auto opcode = OpCode::Decode(instr);
 
diff --git a/src/video_core/shader/decode/bfi.cpp b/src/video_core/shader/decode/bfi.cpp
index 942d6729d..601d66f1f 100644
--- a/src/video_core/shader/decode/bfi.cpp
+++ b/src/video_core/shader/decode/bfi.cpp
@@ -12,7 +12,7 @@ namespace VideoCommon::Shader {
 using Tegra::Shader::Instruction;
 using Tegra::Shader::OpCode;
 
-u32 ShaderIR::DecodeBfi(BasicBlock& bb, const BasicBlock& code, u32 pc) {
+u32 ShaderIR::DecodeBfi(NodeBlock& bb, u32 pc) {
     const Instruction instr = {program_code[pc]};
     const auto opcode = OpCode::Decode(instr);
 
diff --git a/src/video_core/shader/decode/conversion.cpp b/src/video_core/shader/decode/conversion.cpp
index 728a393a1..55a6fbbf2 100644
--- a/src/video_core/shader/decode/conversion.cpp
+++ b/src/video_core/shader/decode/conversion.cpp
@@ -13,7 +13,7 @@ using Tegra::Shader::Instruction;
 using Tegra::Shader::OpCode;
 using Tegra::Shader::Register;
 
-u32 ShaderIR::DecodeConversion(BasicBlock& bb, const BasicBlock& code, u32 pc) {
+u32 ShaderIR::DecodeConversion(NodeBlock& bb, u32 pc) {
     const Instruction instr = {program_code[pc]};
     const auto opcode = OpCode::Decode(instr);
 
@@ -118,8 +118,8 @@ u32 ShaderIR::DecodeConversion(BasicBlock& bb, const BasicBlock& code, u32 pc) {
 
         value = [&]() {
             switch (instr.conversion.f2i.rounding) {
-            case Tegra::Shader::F2iRoundingOp::None:
-                return value;
+            case Tegra::Shader::F2iRoundingOp::RoundEven:
+                return Operation(OperationCode::FRoundEven, PRECISE, value);
             case Tegra::Shader::F2iRoundingOp::Floor:
                 return Operation(OperationCode::FFloor, PRECISE, value);
             case Tegra::Shader::F2iRoundingOp::Ceil:
@@ -146,4 +146,4 @@ u32 ShaderIR::DecodeConversion(BasicBlock& bb, const BasicBlock& code, u32 pc) {
     return pc;
 }
 
-} // namespace VideoCommon::Shader
\ No newline at end of file
+} // namespace VideoCommon::Shader
diff --git a/src/video_core/shader/decode/ffma.cpp b/src/video_core/shader/decode/ffma.cpp
index 52f39d3ff..0559cc8de 100644
--- a/src/video_core/shader/decode/ffma.cpp
+++ b/src/video_core/shader/decode/ffma.cpp
@@ -12,7 +12,7 @@ namespace VideoCommon::Shader {
 using Tegra::Shader::Instruction;
 using Tegra::Shader::OpCode;
 
-u32 ShaderIR::DecodeFfma(BasicBlock& bb, const BasicBlock& code, u32 pc) {
+u32 ShaderIR::DecodeFfma(NodeBlock& bb, u32 pc) {
     const Instruction instr = {program_code[pc]};
     const auto opcode = OpCode::Decode(instr);
 
diff --git a/src/video_core/shader/decode/float_set.cpp b/src/video_core/shader/decode/float_set.cpp
index 9f9da2278..1bd6755dd 100644
--- a/src/video_core/shader/decode/float_set.cpp
+++ b/src/video_core/shader/decode/float_set.cpp
@@ -12,7 +12,7 @@ namespace VideoCommon::Shader {
 using Tegra::Shader::Instruction;
 using Tegra::Shader::OpCode;
 
-u32 ShaderIR::DecodeFloatSet(BasicBlock& bb, const BasicBlock& code, u32 pc) {
+u32 ShaderIR::DecodeFloatSet(NodeBlock& bb, u32 pc) {
     const Instruction instr = {program_code[pc]};
     const auto opcode = OpCode::Decode(instr);
 
diff --git a/src/video_core/shader/decode/float_set_predicate.cpp b/src/video_core/shader/decode/float_set_predicate.cpp
index dd3aef6f2..9285b8d05 100644
--- a/src/video_core/shader/decode/float_set_predicate.cpp
+++ b/src/video_core/shader/decode/float_set_predicate.cpp
@@ -13,7 +13,7 @@ using Tegra::Shader::Instruction;
 using Tegra::Shader::OpCode;
 using Tegra::Shader::Pred;
 
-u32 ShaderIR::DecodeFloatSetPredicate(BasicBlock& bb, const BasicBlock& code, u32 pc) {
+u32 ShaderIR::DecodeFloatSetPredicate(NodeBlock& bb, u32 pc) {
     const Instruction instr = {program_code[pc]};
     const auto opcode = OpCode::Decode(instr);
 
diff --git a/src/video_core/shader/decode/half_set.cpp b/src/video_core/shader/decode/half_set.cpp
index dfd7cb98f..748368555 100644
--- a/src/video_core/shader/decode/half_set.cpp
+++ b/src/video_core/shader/decode/half_set.cpp
@@ -14,7 +14,7 @@ namespace VideoCommon::Shader {
 using Tegra::Shader::Instruction;
 using Tegra::Shader::OpCode;
 
-u32 ShaderIR::DecodeHalfSet(BasicBlock& bb, const BasicBlock& code, u32 pc) {
+u32 ShaderIR::DecodeHalfSet(NodeBlock& bb, u32 pc) {
     const Instruction instr = {program_code[pc]};
     const auto opcode = OpCode::Decode(instr);
 
diff --git a/src/video_core/shader/decode/half_set_predicate.cpp b/src/video_core/shader/decode/half_set_predicate.cpp
index 53c44ae5a..e68512692 100644
--- a/src/video_core/shader/decode/half_set_predicate.cpp
+++ b/src/video_core/shader/decode/half_set_predicate.cpp
@@ -13,7 +13,7 @@ using Tegra::Shader::Instruction;
 using Tegra::Shader::OpCode;
 using Tegra::Shader::Pred;
 
-u32 ShaderIR::DecodeHalfSetPredicate(BasicBlock& bb, const BasicBlock& code, u32 pc) {
+u32 ShaderIR::DecodeHalfSetPredicate(NodeBlock& bb, u32 pc) {
     const Instruction instr = {program_code[pc]};
     const auto opcode = OpCode::Decode(instr);
 
diff --git a/src/video_core/shader/decode/hfma2.cpp b/src/video_core/shader/decode/hfma2.cpp
index 43a0a9e10..7a07c5ec6 100644
--- a/src/video_core/shader/decode/hfma2.cpp
+++ b/src/video_core/shader/decode/hfma2.cpp
@@ -16,7 +16,7 @@ using Tegra::Shader::HalfType;
 using Tegra::Shader::Instruction;
 using Tegra::Shader::OpCode;
 
-u32 ShaderIR::DecodeHfma2(BasicBlock& bb, const BasicBlock& code, u32 pc) {
+u32 ShaderIR::DecodeHfma2(NodeBlock& bb, u32 pc) {
     const Instruction instr = {program_code[pc]};
     const auto opcode = OpCode::Decode(instr);
 
diff --git a/src/video_core/shader/decode/integer_set.cpp b/src/video_core/shader/decode/integer_set.cpp
index 16eb3985f..a3bf17eba 100644
--- a/src/video_core/shader/decode/integer_set.cpp
+++ b/src/video_core/shader/decode/integer_set.cpp
@@ -12,7 +12,7 @@ namespace VideoCommon::Shader {
 using Tegra::Shader::Instruction;
 using Tegra::Shader::OpCode;
 
-u32 ShaderIR::DecodeIntegerSet(BasicBlock& bb, const BasicBlock& code, u32 pc) {
+u32 ShaderIR::DecodeIntegerSet(NodeBlock& bb, u32 pc) {
     const Instruction instr = {program_code[pc]};
     const auto opcode = OpCode::Decode(instr);
 
diff --git a/src/video_core/shader/decode/integer_set_predicate.cpp b/src/video_core/shader/decode/integer_set_predicate.cpp
index daf97174b..aad836d24 100644
--- a/src/video_core/shader/decode/integer_set_predicate.cpp
+++ b/src/video_core/shader/decode/integer_set_predicate.cpp
@@ -13,7 +13,7 @@ using Tegra::Shader::Instruction;
 using Tegra::Shader::OpCode;
 using Tegra::Shader::Pred;
 
-u32 ShaderIR::DecodeIntegerSetPredicate(BasicBlock& bb, const BasicBlock& code, u32 pc) {
+u32 ShaderIR::DecodeIntegerSetPredicate(NodeBlock& bb, u32 pc) {
     const Instruction instr = {program_code[pc]};
     const auto opcode = OpCode::Decode(instr);
 
diff --git a/src/video_core/shader/decode/memory.cpp b/src/video_core/shader/decode/memory.cpp
index 3dd26da20..ea3c71eed 100644
--- a/src/video_core/shader/decode/memory.cpp
+++ b/src/video_core/shader/decode/memory.cpp
@@ -17,26 +17,8 @@ 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) {
+u32 ShaderIR::DecodeMemory(NodeBlock& bb, u32 pc) {
     const Instruction instr = {program_code[pc]};
     const auto opcode = OpCode::Decode(instr);
 
@@ -48,7 +30,7 @@ u32 ShaderIR::DecodeMemory(BasicBlock& bb, const BasicBlock& code, u32 pc) {
         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::IpaMode input_mode{Tegra::Shader::IpaInterpMode::Pass,
                                           Tegra::Shader::IpaSampleMode::Default};
 
         u64 next_element = instr.attribute.fmt20.element;
@@ -160,7 +142,8 @@ u32 ShaderIR::DecodeMemory(BasicBlock& bb, const BasicBlock& code, u32 pc) {
         }();
 
         const Node addr_register = GetRegister(instr.gpr8);
-        const Node base_address = TrackCbuf(addr_register, code, static_cast<s64>(code.size()));
+        const Node base_address =
+            TrackCbuf(addr_register, global_code, static_cast<s64>(global_code.size()));
         const auto cbuf = std::get_if<CbufNode>(base_address);
         ASSERT(cbuf != nullptr);
         const auto cbuf_offset_imm = std::get_if<ImmediateNode>(cbuf->GetOffset());
@@ -246,197 +229,6 @@ u32 ShaderIR::DecodeMemory(BasicBlock& bb, const BasicBlock& code, u32 pc) {
         }
         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);
-
-        u32 indexer = 0;
-        switch (instr.txq.query_type) {
-        case Tegra::Shader::TextureQueryType::Dimension: {
-            for (u32 element = 0; element < 4; ++element) {
-                if (instr.txq.IsComponentEnabled(element)) {
-                    MetaTexture meta{sampler, element};
-                    const Node value = Operation(OperationCode::F4TextureQueryDimensions,
-                                                 std::move(meta), GetRegister(instr.gpr8));
-                    SetTemporal(bb, indexer++, value);
-                }
-            }
-            for (u32 i = 0; i < indexer; ++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());
     }
@@ -444,328 +236,4 @@ u32 ShaderIR::DecodeMemory(BasicBlock& bb, const BasicBlock& code, u32 pc) {
     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
diff --git a/src/video_core/shader/decode/other.cpp b/src/video_core/shader/decode/other.cpp
index c1e5f4efb..d750a2936 100644
--- a/src/video_core/shader/decode/other.cpp
+++ b/src/video_core/shader/decode/other.cpp
@@ -14,7 +14,7 @@ using Tegra::Shader::Instruction;
 using Tegra::Shader::OpCode;
 using Tegra::Shader::Register;
 
-u32 ShaderIR::DecodeOther(BasicBlock& bb, const BasicBlock& code, u32 pc) {
+u32 ShaderIR::DecodeOther(NodeBlock& bb, u32 pc) {
     const Instruction instr = {program_code[pc]};
     const auto opcode = OpCode::Decode(instr);
 
@@ -135,7 +135,18 @@ u32 ShaderIR::DecodeOther(BasicBlock& bb, const BasicBlock& code, u32 pc) {
                                                 instr.ipa.sample_mode.Value()};
 
         const Node attr = GetInputAttribute(attribute.index, attribute.element, input_mode);
-        const Node value = GetSaturatedFloat(attr, instr.ipa.saturate);
+        Node value = attr;
+        const Tegra::Shader::Attribute::Index index = attribute.index.Value();
+        if (index >= Tegra::Shader::Attribute::Index::Attribute_0 &&
+            index <= Tegra::Shader::Attribute::Index::Attribute_31) {
+            // TODO(Blinkhawk): There are cases where a perspective attribute use PASS.
+            // In theory by setting them as perspective, OpenGL does the perspective correction.
+            // A way must figured to reverse the last step of it.
+            if (input_mode.interpolation_mode == Tegra::Shader::IpaInterpMode::Multiply) {
+                value = Operation(OperationCode::FMul, PRECISE, value, GetRegister(instr.gpr20));
+            }
+        }
+        value = GetSaturatedFloat(value, instr.ipa.saturate);
 
         SetRegister(bb, instr.gpr0, value);
         break;
@@ -175,4 +186,4 @@ u32 ShaderIR::DecodeOther(BasicBlock& bb, const BasicBlock& code, u32 pc) {
     return pc;
 }
 
-} // namespace VideoCommon::Shader
\ No newline at end of file
+} // namespace VideoCommon::Shader
diff --git a/src/video_core/shader/decode/predicate_set_predicate.cpp b/src/video_core/shader/decode/predicate_set_predicate.cpp
index 1717f0653..83c61680e 100644
--- a/src/video_core/shader/decode/predicate_set_predicate.cpp
+++ b/src/video_core/shader/decode/predicate_set_predicate.cpp
@@ -13,7 +13,7 @@ using Tegra::Shader::Instruction;
 using Tegra::Shader::OpCode;
 using Tegra::Shader::Pred;
 
-u32 ShaderIR::DecodePredicateSetPredicate(BasicBlock& bb, const BasicBlock& code, u32 pc) {
+u32 ShaderIR::DecodePredicateSetPredicate(NodeBlock& bb, u32 pc) {
     const Instruction instr = {program_code[pc]};
     const auto opcode = OpCode::Decode(instr);
 
diff --git a/src/video_core/shader/decode/predicate_set_register.cpp b/src/video_core/shader/decode/predicate_set_register.cpp
index 8bd15fb00..d0495995d 100644
--- a/src/video_core/shader/decode/predicate_set_register.cpp
+++ b/src/video_core/shader/decode/predicate_set_register.cpp
@@ -12,7 +12,7 @@ namespace VideoCommon::Shader {
 using Tegra::Shader::Instruction;
 using Tegra::Shader::OpCode;
 
-u32 ShaderIR::DecodePredicateSetRegister(BasicBlock& bb, const BasicBlock& code, u32 pc) {
+u32 ShaderIR::DecodePredicateSetRegister(NodeBlock& bb, u32 pc) {
     const Instruction instr = {program_code[pc]};
     const auto opcode = OpCode::Decode(instr);
 
diff --git a/src/video_core/shader/decode/register_set_predicate.cpp b/src/video_core/shader/decode/register_set_predicate.cpp
index bdb4424a6..f070e8912 100644
--- a/src/video_core/shader/decode/register_set_predicate.cpp
+++ b/src/video_core/shader/decode/register_set_predicate.cpp
@@ -12,7 +12,7 @@ namespace VideoCommon::Shader {
 using Tegra::Shader::Instruction;
 using Tegra::Shader::OpCode;
 
-u32 ShaderIR::DecodeRegisterSetPredicate(BasicBlock& bb, const BasicBlock& code, u32 pc) {
+u32 ShaderIR::DecodeRegisterSetPredicate(NodeBlock& bb, u32 pc) {
     const Instruction instr = {program_code[pc]};
     const auto opcode = OpCode::Decode(instr);
 
diff --git a/src/video_core/shader/decode/shift.cpp b/src/video_core/shader/decode/shift.cpp
index 6623f8ff9..951e85f44 100644
--- a/src/video_core/shader/decode/shift.cpp
+++ b/src/video_core/shader/decode/shift.cpp
@@ -12,7 +12,7 @@ namespace VideoCommon::Shader {
 using Tegra::Shader::Instruction;
 using Tegra::Shader::OpCode;
 
-u32 ShaderIR::DecodeShift(BasicBlock& bb, const BasicBlock& code, u32 pc) {
+u32 ShaderIR::DecodeShift(NodeBlock& bb, u32 pc) {
     const Instruction instr = {program_code[pc]};
     const auto opcode = OpCode::Decode(instr);
 
diff --git a/src/video_core/shader/decode/texture.cpp b/src/video_core/shader/decode/texture.cpp
new file mode 100644
index 000000000..a99ae19bf
--- /dev/null
+++ b/src/video_core/shader/decode/texture.cpp
@@ -0,0 +1,534 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <algorithm>
+#include <vector>
+#include <fmt/format.h>
+
+#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;
+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::DecodeTexture(NodeBlock& bb, u32 pc) {
+    const Instruction instr = {program_code[pc]};
+    const auto opcode = OpCode::Decode(instr);
+
+    switch (opcode->get().GetId()) {
+    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);
+
+        // TODO(Subv): Figure out how the sampler type is encoded in the TLD4S instruction.
+        std::vector<Node> coords;
+        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 Node component = 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 coords_copy = coords;
+            MetaTexture meta{sampler, {}, {}, {}, {}, component, element};
+            values[element] = Operation(OperationCode::TextureGather, meta, std::move(coords_copy));
+        }
+
+        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);
+
+        u32 indexer = 0;
+        switch (instr.txq.query_type) {
+        case Tegra::Shader::TextureQueryType::Dimension: {
+            for (u32 element = 0; element < 4; ++element) {
+                if (!instr.txq.IsComponentEnabled(element)) {
+                    continue;
+                }
+                MetaTexture meta{sampler, {}, {}, {}, {}, {}, element};
+                const Node value =
+                    Operation(OperationCode::TextureQueryDimensions, meta, GetRegister(instr.gpr8));
+                SetTemporal(bb, indexer++, value);
+            }
+            for (u32 i = 0; i < indexer; ++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{sampler, {}, {}, {}, {}, {}, element};
+            const Node value = Operation(OperationCode::TextureQueryLod, meta, 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, "TLDS.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(NodeBlock& 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(NodeBlock& 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(NodeBlock& 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, std::vector<Node> coords,
+                               Node array, Node depth_compare, u32 bias_offset) {
+    const bool is_array = array;
+    const bool is_shadow = depth_compare;
+
+    UNIMPLEMENTED_IF_MSG((texture_type == TextureType::Texture3D && (is_array || is_shadow)) ||
+                             (texture_type == TextureType::TextureCube && is_array && is_shadow),
+                         "This method is not supported.");
+
+    const auto& sampler = GetSampler(instr.sampler, texture_type, is_array, is_shadow);
+
+    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 && is_shadow) ||
+          (texture_type == Tegra::Shader::TextureType::TextureCube && is_array && is_shadow));
+
+    const OperationCode read_method =
+        (lod_needed && gl_lod_supported) ? OperationCode::TextureLod : OperationCode::Texture;
+
+    UNIMPLEMENTED_IF(process_mode != TextureProcessMode::None && !gl_lod_supported);
+
+    Node bias = {};
+    Node lod = {};
+    if (process_mode != TextureProcessMode::None && gl_lod_supported) {
+        switch (process_mode) {
+        case TextureProcessMode::LZ:
+            lod = Immediate(0.0f);
+            break;
+        case TextureProcessMode::LB:
+            // 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
+            bias = GetRegister(instr.gpr20.Value() + bias_offset);
+            break;
+        case TextureProcessMode::LL:
+            lod = GetRegister(instr.gpr20.Value() + bias_offset);
+            break;
+        default:
+            UNIMPLEMENTED_MSG("Unimplemented process mode={}", static_cast<u32>(process_mode));
+            break;
+        }
+    }
+
+    Node4 values;
+    for (u32 element = 0; element < values.size(); ++element) {
+        auto copy_coords = coords;
+        MetaTexture meta{sampler, array, depth_compare, bias, lod, {}, element};
+        values[element] = Operation(read_method, meta, std::move(copy_coords));
+    }
+
+    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));
+    }
+
+    const Node array = is_array ? GetRegister(array_register) : nullptr;
+
+    Node dc{};
+    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);
+        dc = GetRegister(depth_register);
+    }
+
+    return GetTextureCode(instr, texture_type, process_mode, coords, array, dc, 0);
+}
+
+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;
+    const u32 bias_offset = coord_count > 2 ? 1 : 0;
+
+    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));
+    }
+
+    const Node array = is_array ? GetRegister(array_register) : nullptr;
+
+    Node dc{};
+    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);
+        dc = GetRegister(depth_register);
+    }
+
+    return GetTextureCode(instr, texture_type, process_mode, coords, array, dc, bias_offset);
+}
+
+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));
+
+    const auto& sampler = GetSampler(instr.sampler, texture_type, is_array, depth_compare);
+
+    Node4 values;
+    for (u32 element = 0; element < values.size(); ++element) {
+        auto coords_copy = coords;
+        MetaTexture meta{sampler, GetRegister(array_register), {}, {}, {}, {}, element};
+        values[element] = Operation(OperationCode::TextureGather, meta, std::move(coords_copy));
+    }
+
+    return values;
+}
+
+Node4 ShaderIR::GetTldsCode(Instruction instr, TextureType texture_type, bool is_array) {
+    const std::size_t type_coord_count = GetCoordCount(texture_type);
+    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));
+    }
+
+    const Node array = is_array ? GetRegister(array_register) : nullptr;
+    // When lod is used always is in gpr20
+    const Node lod = lod_enabled ? GetRegister(instr.gpr20) : Immediate(0);
+
+    const auto& sampler = GetSampler(instr.sampler, texture_type, is_array, false);
+
+    Node4 values;
+    for (u32 element = 0; element < values.size(); ++element) {
+        auto coords_copy = coords;
+        MetaTexture meta{sampler, array, {}, {}, lod, {}, element};
+        values[element] = Operation(OperationCode::TexelFetch, meta, std::move(coords_copy));
+    }
+    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
\ No newline at end of file
diff --git a/src/video_core/shader/decode/video.cpp b/src/video_core/shader/decode/video.cpp
index c3432356d..956c01d9b 100644
--- a/src/video_core/shader/decode/video.cpp
+++ b/src/video_core/shader/decode/video.cpp
@@ -15,7 +15,7 @@ using Tegra::Shader::Pred;
 using Tegra::Shader::VideoType;
 using Tegra::Shader::VmadShr;
 
-u32 ShaderIR::DecodeVideo(BasicBlock& bb, const BasicBlock& code, u32 pc) {
+u32 ShaderIR::DecodeVideo(NodeBlock& bb, u32 pc) {
     const Instruction instr = {program_code[pc]};
     const auto opcode = OpCode::Decode(instr);
 
diff --git a/src/video_core/shader/decode/xmad.cpp b/src/video_core/shader/decode/xmad.cpp
index 9cb864500..c34843307 100644
--- a/src/video_core/shader/decode/xmad.cpp
+++ b/src/video_core/shader/decode/xmad.cpp
@@ -12,7 +12,7 @@ namespace VideoCommon::Shader {
 using Tegra::Shader::Instruction;
 using Tegra::Shader::OpCode;
 
-u32 ShaderIR::DecodeXmad(BasicBlock& bb, const BasicBlock& code, u32 pc) {
+u32 ShaderIR::DecodeXmad(NodeBlock& bb, u32 pc) {
     const Instruction instr = {program_code[pc]};
     const auto opcode = OpCode::Decode(instr);
 
diff --git a/src/video_core/shader/shader_ir.cpp b/src/video_core/shader/shader_ir.cpp
index d7747103e..ac5112d78 100644
--- a/src/video_core/shader/shader_ir.cpp
+++ b/src/video_core/shader/shader_ir.cpp
@@ -337,27 +337,27 @@ Node ShaderIR::GetConditionCode(Tegra::Shader::ConditionCode cc) {
     }
 }
 
-void ShaderIR::SetRegister(BasicBlock& bb, Register dest, Node src) {
+void ShaderIR::SetRegister(NodeBlock& bb, Register dest, Node src) {
     bb.push_back(Operation(OperationCode::Assign, GetRegister(dest), src));
 }
 
-void ShaderIR::SetPredicate(BasicBlock& bb, u64 dest, Node src) {
+void ShaderIR::SetPredicate(NodeBlock& bb, u64 dest, Node src) {
     bb.push_back(Operation(OperationCode::LogicalAssign, GetPredicate(dest), src));
 }
 
-void ShaderIR::SetInternalFlag(BasicBlock& bb, InternalFlag flag, Node value) {
+void ShaderIR::SetInternalFlag(NodeBlock& bb, InternalFlag flag, Node value) {
     bb.push_back(Operation(OperationCode::LogicalAssign, GetInternalFlag(flag), value));
 }
 
-void ShaderIR::SetLocalMemory(BasicBlock& bb, Node address, Node value) {
+void ShaderIR::SetLocalMemory(NodeBlock& bb, Node address, Node value) {
     bb.push_back(Operation(OperationCode::Assign, GetLocalMemory(address), value));
 }
 
-void ShaderIR::SetTemporal(BasicBlock& bb, u32 id, Node value) {
+void ShaderIR::SetTemporal(NodeBlock& bb, u32 id, Node value) {
     SetRegister(bb, Register::ZeroIndex + 1 + id, value);
 }
 
-void ShaderIR::SetInternalFlagsFromFloat(BasicBlock& bb, Node value, bool sets_cc) {
+void ShaderIR::SetInternalFlagsFromFloat(NodeBlock& bb, Node value, bool sets_cc) {
     if (!sets_cc) {
         return;
     }
@@ -366,7 +366,7 @@ void ShaderIR::SetInternalFlagsFromFloat(BasicBlock& bb, Node value, bool sets_c
     LOG_WARNING(HW_GPU, "Condition codes implementation is incomplete");
 }
 
-void ShaderIR::SetInternalFlagsFromInteger(BasicBlock& bb, Node value, bool sets_cc) {
+void ShaderIR::SetInternalFlagsFromInteger(NodeBlock& bb, Node value, bool sets_cc) {
     if (!sets_cc) {
         return;
     }
diff --git a/src/video_core/shader/shader_ir.h b/src/video_core/shader/shader_ir.h
index 6e42e3dfb..5bc3a3900 100644
--- a/src/video_core/shader/shader_ir.h
+++ b/src/video_core/shader/shader_ir.h
@@ -39,7 +39,7 @@ using NodeData =
                  PredicateNode, AbufNode, CbufNode, LmemNode, GmemNode, CommentNode>;
 using Node = const NodeData*;
 using Node4 = std::array<Node, 4>;
-using BasicBlock = std::vector<Node>;
+using NodeBlock = std::vector<Node>;
 
 constexpr u32 MAX_PROGRAM_LENGTH = 0x1000;
 
@@ -156,12 +156,12 @@ enum class OperationCode {
     Logical2HNotEqual,     /// (MetaHalfArithmetic, f16vec2 a, f16vec2) -> bool2
     Logical2HGreaterEqual, /// (MetaHalfArithmetic, f16vec2 a, f16vec2) -> bool2
 
-    F4Texture,                /// (MetaTexture, float[N] coords, float[M] params) -> float4
-    F4TextureLod,             /// (MetaTexture, float[N] coords, float[M] params) -> float4
-    F4TextureGather,          /// (MetaTexture, float[N] coords, float[M] params) -> float4
-    F4TextureQueryDimensions, /// (MetaTexture, float a) -> float4
-    F4TextureQueryLod,        /// (MetaTexture, float[N] coords) -> float4
-    F4TexelFetch,             /// (MetaTexture, int[N], int) -> float4
+    Texture,                /// (MetaTexture, float[N] coords) -> float4
+    TextureLod,             /// (MetaTexture, float[N] coords) -> float4
+    TextureGather,          /// (MetaTexture, float[N] coords) -> float4
+    TextureQueryDimensions, /// (MetaTexture, float a) -> float4
+    TextureQueryLod,        /// (MetaTexture, float[N] coords) -> float4
+    TexelFetch,             /// (MetaTexture, int[N], int) -> float4
 
     Branch,        /// (uint branch_target) -> void
     PushFlowStack, /// (uint branch_target) -> void
@@ -236,6 +236,11 @@ private:
 
 class ConstBuffer {
 public:
+    explicit ConstBuffer(u32 max_offset, bool is_indirect)
+        : max_offset{max_offset}, is_indirect{is_indirect} {}
+
+    ConstBuffer() = default;
+
     void MarkAsUsed(u64 offset) {
         max_offset = std::max(max_offset, static_cast<u32>(offset));
     }
@@ -252,6 +257,10 @@ public:
         return max_offset + sizeof(float);
     }
 
+    u32 GetMaxOffset() const {
+        return max_offset;
+    }
+
 private:
     u32 max_offset{};
     bool is_indirect{};
@@ -279,9 +288,12 @@ struct MetaHalfArithmetic {
 
 struct MetaTexture {
     const Sampler& sampler;
+    Node array{};
+    Node depth_compare{};
+    Node bias{};
+    Node lod{};
+    Node component{};
     u32 element{};
-    u32 coords_count{};
-    std::optional<u32> array_index;
 };
 
 constexpr MetaArithmetic PRECISE = {true};
@@ -530,7 +542,7 @@ public:
         Decode();
     }
 
-    const std::map<u32, BasicBlock>& GetBasicBlocks() const {
+    const std::map<u32, NodeBlock>& GetBasicBlocks() const {
         return basic_blocks;
     }
 
@@ -581,7 +593,7 @@ private:
 
     ExitMethod Scan(u32 begin, u32 end, std::set<u32>& labels);
 
-    BasicBlock DecodeRange(u32 begin, u32 end);
+    NodeBlock DecodeRange(u32 begin, u32 end);
 
     /**
      * Decodes a single instruction from Tegra to IR.
@@ -589,33 +601,34 @@ private:
      * @param pc Program counter. Offset to decode.
      * @return Next address to decode.
      */
-    u32 DecodeInstr(BasicBlock& bb, u32 pc);
-
-    u32 DecodeArithmetic(BasicBlock& bb, const BasicBlock& code, u32 pc);
-    u32 DecodeArithmeticImmediate(BasicBlock& bb, const BasicBlock& code, u32 pc);
-    u32 DecodeBfe(BasicBlock& bb, const BasicBlock& code, u32 pc);
-    u32 DecodeBfi(BasicBlock& bb, const BasicBlock& code, u32 pc);
-    u32 DecodeShift(BasicBlock& bb, const BasicBlock& code, u32 pc);
-    u32 DecodeArithmeticInteger(BasicBlock& bb, const BasicBlock& code, u32 pc);
-    u32 DecodeArithmeticIntegerImmediate(BasicBlock& bb, const BasicBlock& code, u32 pc);
-    u32 DecodeArithmeticHalf(BasicBlock& bb, const BasicBlock& code, u32 pc);
-    u32 DecodeArithmeticHalfImmediate(BasicBlock& bb, const BasicBlock& code, u32 pc);
-    u32 DecodeFfma(BasicBlock& bb, const BasicBlock& code, u32 pc);
-    u32 DecodeHfma2(BasicBlock& bb, const BasicBlock& code, u32 pc);
-    u32 DecodeConversion(BasicBlock& bb, const BasicBlock& code, u32 pc);
-    u32 DecodeMemory(BasicBlock& bb, const BasicBlock& code, u32 pc);
-    u32 DecodeFloatSetPredicate(BasicBlock& bb, const BasicBlock& code, u32 pc);
-    u32 DecodeIntegerSetPredicate(BasicBlock& bb, const BasicBlock& code, u32 pc);
-    u32 DecodeHalfSetPredicate(BasicBlock& bb, const BasicBlock& code, u32 pc);
-    u32 DecodePredicateSetRegister(BasicBlock& bb, const BasicBlock& code, u32 pc);
-    u32 DecodePredicateSetPredicate(BasicBlock& bb, const BasicBlock& code, u32 pc);
-    u32 DecodeRegisterSetPredicate(BasicBlock& bb, const BasicBlock& code, u32 pc);
-    u32 DecodeFloatSet(BasicBlock& bb, const BasicBlock& code, u32 pc);
-    u32 DecodeIntegerSet(BasicBlock& bb, const BasicBlock& code, u32 pc);
-    u32 DecodeHalfSet(BasicBlock& bb, const BasicBlock& code, u32 pc);
-    u32 DecodeVideo(BasicBlock& bb, const BasicBlock& code, u32 pc);
-    u32 DecodeXmad(BasicBlock& bb, const BasicBlock& code, u32 pc);
-    u32 DecodeOther(BasicBlock& bb, const BasicBlock& code, u32 pc);
+    u32 DecodeInstr(NodeBlock& bb, u32 pc);
+
+    u32 DecodeArithmetic(NodeBlock& bb, u32 pc);
+    u32 DecodeArithmeticImmediate(NodeBlock& bb, u32 pc);
+    u32 DecodeBfe(NodeBlock& bb, u32 pc);
+    u32 DecodeBfi(NodeBlock& bb, u32 pc);
+    u32 DecodeShift(NodeBlock& bb, u32 pc);
+    u32 DecodeArithmeticInteger(NodeBlock& bb, u32 pc);
+    u32 DecodeArithmeticIntegerImmediate(NodeBlock& bb, u32 pc);
+    u32 DecodeArithmeticHalf(NodeBlock& bb, u32 pc);
+    u32 DecodeArithmeticHalfImmediate(NodeBlock& bb, u32 pc);
+    u32 DecodeFfma(NodeBlock& bb, u32 pc);
+    u32 DecodeHfma2(NodeBlock& bb, u32 pc);
+    u32 DecodeConversion(NodeBlock& bb, u32 pc);
+    u32 DecodeMemory(NodeBlock& bb, u32 pc);
+    u32 DecodeTexture(NodeBlock& bb, u32 pc);
+    u32 DecodeFloatSetPredicate(NodeBlock& bb, u32 pc);
+    u32 DecodeIntegerSetPredicate(NodeBlock& bb, u32 pc);
+    u32 DecodeHalfSetPredicate(NodeBlock& bb, u32 pc);
+    u32 DecodePredicateSetRegister(NodeBlock& bb, u32 pc);
+    u32 DecodePredicateSetPredicate(NodeBlock& bb, u32 pc);
+    u32 DecodeRegisterSetPredicate(NodeBlock& bb, u32 pc);
+    u32 DecodeFloatSet(NodeBlock& bb, u32 pc);
+    u32 DecodeIntegerSet(NodeBlock& bb, u32 pc);
+    u32 DecodeHalfSet(NodeBlock& bb, u32 pc);
+    u32 DecodeVideo(NodeBlock& bb, u32 pc);
+    u32 DecodeXmad(NodeBlock& bb, u32 pc);
+    u32 DecodeOther(NodeBlock& bb, u32 pc);
 
     /// Internalizes node's data and returns a managed pointer to a clone of that node
     Node StoreNode(NodeData&& node_data);
@@ -664,20 +677,20 @@ private:
     Node GetTemporal(u32 id);
 
     /// Sets a register. src value must be a number-evaluated node.
-    void SetRegister(BasicBlock& bb, Tegra::Shader::Register dest, Node src);
+    void SetRegister(NodeBlock& bb, Tegra::Shader::Register dest, Node src);
     /// Sets a predicate. src value must be a bool-evaluated node
-    void SetPredicate(BasicBlock& bb, u64 dest, Node src);
+    void SetPredicate(NodeBlock& bb, u64 dest, Node src);
     /// Sets an internal flag. src value must be a bool-evaluated node
-    void SetInternalFlag(BasicBlock& bb, InternalFlag flag, Node value);
+    void SetInternalFlag(NodeBlock& bb, InternalFlag flag, Node value);
     /// Sets a local memory address. address and value must be a number-evaluated node
-    void SetLocalMemory(BasicBlock& bb, Node address, Node value);
+    void SetLocalMemory(NodeBlock& bb, Node address, Node value);
     /// Sets a temporal. Internally it uses a post-RZ register
-    void SetTemporal(BasicBlock& bb, u32 id, Node value);
+    void SetTemporal(NodeBlock& bb, u32 id, Node value);
 
     /// Sets internal flags from a float
-    void SetInternalFlagsFromFloat(BasicBlock& bb, Node value, bool sets_cc = true);
+    void SetInternalFlagsFromFloat(NodeBlock& bb, Node value, bool sets_cc = true);
     /// Sets internal flags from an integer
-    void SetInternalFlagsFromInteger(BasicBlock& bb, Node value, bool sets_cc = true);
+    void SetInternalFlagsFromInteger(NodeBlock& bb, Node value, bool sets_cc = true);
 
     /// Conditionally absolute/negated float. Absolute is applied first
     Node GetOperandAbsNegFloat(Node value, bool absolute, bool negate);
@@ -718,12 +731,12 @@ private:
     /// Extracts a sequence of bits from a node
     Node BitfieldExtract(Node value, u32 offset, u32 bits);
 
-    void WriteTexInstructionFloat(BasicBlock& bb, Tegra::Shader::Instruction instr,
+    void WriteTexInstructionFloat(NodeBlock& bb, Tegra::Shader::Instruction instr,
                                   const Node4& components);
 
-    void WriteTexsInstructionFloat(BasicBlock& bb, Tegra::Shader::Instruction instr,
+    void WriteTexsInstructionFloat(NodeBlock& bb, Tegra::Shader::Instruction instr,
                                    const Node4& components);
-    void WriteTexsInstructionHalfFloat(BasicBlock& bb, Tegra::Shader::Instruction instr,
+    void WriteTexsInstructionHalfFloat(NodeBlock& bb, Tegra::Shader::Instruction instr,
                                        const Node4& components);
 
     Node4 GetTexCode(Tegra::Shader::Instruction instr, Tegra::Shader::TextureType texture_type,
@@ -745,23 +758,22 @@ private:
         bool lod_bias_enabled, std::size_t max_coords, std::size_t max_inputs);
 
     Node4 GetTextureCode(Tegra::Shader::Instruction instr, Tegra::Shader::TextureType texture_type,
-                         Tegra::Shader::TextureProcessMode process_mode, bool depth_compare,
-                         bool is_array, std::size_t array_offset, std::size_t bias_offset,
-                         std::vector<Node>&& coords);
+                         Tegra::Shader::TextureProcessMode process_mode, std::vector<Node> coords,
+                         Node array, Node depth_compare, u32 bias_offset);
 
     Node GetVideoOperand(Node op, bool is_chunk, bool is_signed, Tegra::Shader::VideoType type,
                          u64 byte_height);
 
-    void WriteLogicOperation(BasicBlock& bb, Tegra::Shader::Register dest,
+    void WriteLogicOperation(NodeBlock& bb, Tegra::Shader::Register dest,
                              Tegra::Shader::LogicOperation logic_op, Node op_a, Node op_b,
                              Tegra::Shader::PredicateResultMode predicate_mode,
                              Tegra::Shader::Pred predicate, bool sets_cc);
-    void WriteLop3Instruction(BasicBlock& bb, Tegra::Shader::Register dest, Node op_a, Node op_b,
+    void WriteLop3Instruction(NodeBlock& bb, Tegra::Shader::Register dest, Node op_a, Node op_b,
                               Node op_c, Node imm_lut, bool sets_cc);
 
-    Node TrackCbuf(Node tracked, const BasicBlock& code, s64 cursor);
+    Node TrackCbuf(Node tracked, const NodeBlock& code, s64 cursor);
 
-    std::pair<Node, s64> TrackRegister(const GprNode* tracked, const BasicBlock& code, s64 cursor);
+    std::pair<Node, s64> TrackRegister(const GprNode* tracked, const NodeBlock& code, s64 cursor);
 
     template <typename... T>
     Node Operation(OperationCode code, const T*... operands) {
@@ -803,7 +815,8 @@ private:
     u32 coverage_end{};
     std::map<std::pair<u32, u32>, ExitMethod> exit_method_map;
 
-    std::map<u32, BasicBlock> basic_blocks;
+    std::map<u32, NodeBlock> basic_blocks;
+    NodeBlock global_code;
 
     std::vector<std::unique_ptr<NodeData>> stored_nodes;
 
diff --git a/src/video_core/shader/track.cpp b/src/video_core/shader/track.cpp
index d6d29ee9f..33b071747 100644
--- a/src/video_core/shader/track.cpp
+++ b/src/video_core/shader/track.cpp
@@ -11,7 +11,7 @@
 namespace VideoCommon::Shader {
 
 namespace {
-std::pair<Node, s64> FindOperation(const BasicBlock& code, s64 cursor,
+std::pair<Node, s64> FindOperation(const NodeBlock& code, s64 cursor,
                                    OperationCode operation_code) {
     for (; cursor >= 0; --cursor) {
         const Node node = code[cursor];
@@ -19,12 +19,19 @@ std::pair<Node, s64> FindOperation(const BasicBlock& code, s64 cursor,
             if (operation->GetCode() == operation_code)
                 return {node, cursor};
         }
+        if (const auto conditional = std::get_if<ConditionalNode>(node)) {
+            const auto& conditional_code = conditional->GetCode();
+            const auto [found, internal_cursor] = FindOperation(
+                conditional_code, static_cast<s64>(conditional_code.size() - 1), operation_code);
+            if (found)
+                return {found, cursor};
+        }
     }
     return {};
 }
 } // namespace
 
-Node ShaderIR::TrackCbuf(Node tracked, const BasicBlock& code, s64 cursor) {
+Node ShaderIR::TrackCbuf(Node tracked, const NodeBlock& code, s64 cursor) {
     if (const auto cbuf = std::get_if<CbufNode>(tracked)) {
         // Cbuf found, but it has to be immediate
         return std::holds_alternative<ImmediateNode>(*cbuf->GetOffset()) ? tracked : nullptr;
@@ -50,10 +57,14 @@ Node ShaderIR::TrackCbuf(Node tracked, const BasicBlock& code, s64 cursor) {
         }
         return nullptr;
     }
+    if (const auto conditional = std::get_if<ConditionalNode>(tracked)) {
+        const auto& conditional_code = conditional->GetCode();
+        return TrackCbuf(tracked, conditional_code, static_cast<s64>(conditional_code.size()));
+    }
     return nullptr;
 }
 
-std::pair<Node, s64> ShaderIR::TrackRegister(const GprNode* tracked, const BasicBlock& code,
+std::pair<Node, s64> ShaderIR::TrackRegister(const GprNode* tracked, const NodeBlock& code,
                                              s64 cursor) {
     for (; cursor >= 0; --cursor) {
         const auto [found_node, new_cursor] = FindOperation(code, cursor, OperationCode::Assign);
diff --git a/src/video_core/surface.cpp b/src/video_core/surface.cpp
index 2f6612a35..a7ac26d71 100644
--- a/src/video_core/surface.cpp
+++ b/src/video_core/surface.cpp
@@ -89,8 +89,6 @@ PixelFormat PixelFormatFromDepthFormat(Tegra::DepthFormat format) {
 
 PixelFormat PixelFormatFromRenderTargetFormat(Tegra::RenderTargetFormat format) {
     switch (format) {
-        // TODO (Hexagon12): Converting SRGBA to RGBA is a hack and doesn't completely correct the
-        // gamma.
     case Tegra::RenderTargetFormat::RGBA8_SRGB:
         return PixelFormat::RGBA8_SRGB;
     case Tegra::RenderTargetFormat::RGBA8_UNORM:
@@ -426,6 +424,8 @@ PixelFormat PixelFormatFromGPUPixelFormat(Tegra::FramebufferConfig::PixelFormat
     switch (format) {
     case Tegra::FramebufferConfig::PixelFormat::ABGR8:
         return PixelFormat::ABGR8U;
+    case Tegra::FramebufferConfig::PixelFormat::BGRA8:
+        return PixelFormat::BGRA8;
     default:
         LOG_CRITICAL(HW_GPU, "Unimplemented format={}", static_cast<u32>(format));
         UNREACHABLE();
diff --git a/src/video_core/textures/astc.cpp b/src/video_core/textures/astc.cpp
index bc50a4876..b508d64e9 100644
--- a/src/video_core/textures/astc.cpp
+++ b/src/video_core/textures/astc.cpp
@@ -23,28 +23,12 @@
 
 #include "video_core/textures/astc.h"
 
-class BitStream {
+class InputBitStream {
 public:
-    explicit BitStream(unsigned char* ptr, int nBits = 0, int start_offset = 0)
+    explicit InputBitStream(const unsigned char* ptr, int nBits = 0, int start_offset = 0)
         : m_NumBits(nBits), m_CurByte(ptr), m_NextBit(start_offset % 8) {}
 
-    ~BitStream() = default;
-
-    int GetBitsWritten() const {
-        return m_BitsWritten;
-    }
-
-    void WriteBitsR(unsigned int val, unsigned int nBits) {
-        for (unsigned int i = 0; i < nBits; i++) {
-            WriteBit((val >> (nBits - i - 1)) & 1);
-        }
-    }
-
-    void WriteBits(unsigned int val, unsigned int nBits) {
-        for (unsigned int i = 0; i < nBits; i++) {
-            WriteBit((val >> i) & 1);
-        }
-    }
+    ~InputBitStream() = default;
 
     int GetBitsRead() const {
         return m_BitsRead;
@@ -71,6 +55,38 @@ public:
     }
 
 private:
+    const int m_NumBits;
+    const unsigned char* m_CurByte;
+    int m_NextBit = 0;
+    int m_BitsRead = 0;
+
+    bool done = false;
+};
+
+class OutputBitStream {
+public:
+    explicit OutputBitStream(unsigned char* ptr, int nBits = 0, int start_offset = 0)
+        : m_NumBits(nBits), m_CurByte(ptr), m_NextBit(start_offset % 8) {}
+
+    ~OutputBitStream() = default;
+
+    int GetBitsWritten() const {
+        return m_BitsWritten;
+    }
+
+    void WriteBitsR(unsigned int val, unsigned int nBits) {
+        for (unsigned int i = 0; i < nBits; i++) {
+            WriteBit((val >> (nBits - i - 1)) & 1);
+        }
+    }
+
+    void WriteBits(unsigned int val, unsigned int nBits) {
+        for (unsigned int i = 0; i < nBits; i++) {
+            WriteBit((val >> i) & 1);
+        }
+    }
+
+private:
     void WriteBit(int b) {
 
         if (done)
@@ -238,8 +254,8 @@ public:
     // Fills result with the values that are encoded in the given
     // bitstream. We must know beforehand what the maximum possible
     // value is, and how many values we're decoding.
-    static void DecodeIntegerSequence(std::vector<IntegerEncodedValue>& result, BitStream& bits,
-                                      uint32_t maxRange, uint32_t nValues) {
+    static void DecodeIntegerSequence(std::vector<IntegerEncodedValue>& result,
+                                      InputBitStream& bits, uint32_t maxRange, uint32_t nValues) {
         // Determine encoding parameters
         IntegerEncodedValue val = IntegerEncodedValue::CreateEncoding(maxRange);
 
@@ -267,7 +283,7 @@ public:
     }
 
 private:
-    static void DecodeTritBlock(BitStream& bits, std::vector<IntegerEncodedValue>& result,
+    static void DecodeTritBlock(InputBitStream& bits, std::vector<IntegerEncodedValue>& result,
                                 uint32_t nBitsPerValue) {
         // Implement the algorithm in section C.2.12
         uint32_t m[5];
@@ -327,7 +343,7 @@ private:
         }
     }
 
-    static void DecodeQuintBlock(BitStream& bits, std::vector<IntegerEncodedValue>& result,
+    static void DecodeQuintBlock(InputBitStream& bits, std::vector<IntegerEncodedValue>& result,
                                  uint32_t nBitsPerValue) {
         // Implement the algorithm in section C.2.12
         uint32_t m[3];
@@ -406,7 +422,7 @@ struct TexelWeightParams {
     }
 };
 
-static TexelWeightParams DecodeBlockInfo(BitStream& strm) {
+static TexelWeightParams DecodeBlockInfo(InputBitStream& strm) {
     TexelWeightParams params;
 
     // Read the entire block mode all at once
@@ -605,7 +621,7 @@ static TexelWeightParams DecodeBlockInfo(BitStream& strm) {
     return params;
 }
 
-static void FillVoidExtentLDR(BitStream& strm, uint32_t* const outBuf, uint32_t blockWidth,
+static void FillVoidExtentLDR(InputBitStream& strm, uint32_t* const outBuf, uint32_t blockWidth,
                               uint32_t blockHeight) {
     // Don't actually care about the void extent, just read the bits...
     for (int i = 0; i < 4; ++i) {
@@ -821,7 +837,7 @@ static void DecodeColorValues(uint32_t* out, uint8_t* data, const uint32_t* mode
 
     // We now have enough to decode our integer sequence.
     std::vector<IntegerEncodedValue> decodedColorValues;
-    BitStream colorStream(data);
+    InputBitStream colorStream(data);
     IntegerEncodedValue::DecodeIntegerSequence(decodedColorValues, colorStream, range, nValues);
 
     // Once we have the decoded values, we need to dequantize them to the 0-255 range
@@ -1365,9 +1381,9 @@ static void ComputeEndpoints(Pixel& ep1, Pixel& ep2, const uint32_t*& colorValue
 #undef READ_INT_VALUES
 }
 
-static void DecompressBlock(uint8_t inBuf[16], const uint32_t blockWidth,
+static void DecompressBlock(const uint8_t inBuf[16], const uint32_t blockWidth,
                             const uint32_t blockHeight, uint32_t* outBuf) {
-    BitStream strm(inBuf);
+    InputBitStream strm(inBuf);
     TexelWeightParams weightParams = DecodeBlockInfo(strm);
 
     // Was there an error?
@@ -1421,7 +1437,7 @@ static void DecompressBlock(uint8_t inBuf[16], const uint32_t blockWidth,
     // Define color data.
     uint8_t colorEndpointData[16];
     memset(colorEndpointData, 0, sizeof(colorEndpointData));
-    BitStream colorEndpointStream(colorEndpointData, 16 * 8, 0);
+    OutputBitStream colorEndpointStream(colorEndpointData, 16 * 8, 0);
 
     // Read extra config data...
     uint32_t baseCEM = 0;
@@ -1549,7 +1565,7 @@ static void DecompressBlock(uint8_t inBuf[16], const uint32_t blockWidth,
     memset(texelWeightData + clearByteStart, 0, 16 - clearByteStart);
 
     std::vector<IntegerEncodedValue> texelWeightValues;
-    BitStream weightStream(texelWeightData);
+    InputBitStream weightStream(texelWeightData);
 
     IntegerEncodedValue::DecodeIntegerSequence(texelWeightValues, weightStream,
                                                weightParams.m_MaxWeight,
@@ -1597,7 +1613,7 @@ static void DecompressBlock(uint8_t inBuf[16], const uint32_t blockWidth,
 
 namespace Tegra::Texture::ASTC {
 
-std::vector<uint8_t> Decompress(std::vector<uint8_t>& data, uint32_t width, uint32_t height,
+std::vector<uint8_t> Decompress(const uint8_t* data, uint32_t width, uint32_t height,
                                 uint32_t depth, uint32_t block_width, uint32_t block_height) {
     uint32_t blockIdx = 0;
     std::vector<uint8_t> outData(height * width * depth * 4);
@@ -1605,7 +1621,7 @@ std::vector<uint8_t> Decompress(std::vector<uint8_t>& data, uint32_t width, uint
         for (uint32_t j = 0; j < height; j += block_height) {
             for (uint32_t i = 0; i < width; i += block_width) {
 
-                uint8_t* blockPtr = data.data() + blockIdx * 16;
+                const uint8_t* blockPtr = data + blockIdx * 16;
 
                 // Blocks can be at most 12x12
                 uint32_t uncompData[144];
diff --git a/src/video_core/textures/astc.h b/src/video_core/textures/astc.h
index d419dd025..991cdba72 100644
--- a/src/video_core/textures/astc.h
+++ b/src/video_core/textures/astc.h
@@ -9,7 +9,7 @@
 
 namespace Tegra::Texture::ASTC {
 
-std::vector<uint8_t> Decompress(std::vector<uint8_t>& data, uint32_t width, uint32_t height,
+std::vector<uint8_t> Decompress(const uint8_t* data, uint32_t width, uint32_t height,
                                 uint32_t depth, uint32_t block_width, uint32_t block_height);
 
 } // namespace Tegra::Texture::ASTC
diff --git a/src/video_core/textures/convert.cpp b/src/video_core/textures/convert.cpp
new file mode 100644
index 000000000..5e439f036
--- /dev/null
+++ b/src/video_core/textures/convert.cpp
@@ -0,0 +1,92 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <algorithm>
+#include <cstring>
+#include <tuple>
+#include <vector>
+
+#include "common/assert.h"
+#include "common/common_types.h"
+#include "common/logging/log.h"
+#include "video_core/textures/astc.h"
+#include "video_core/textures/convert.h"
+
+namespace Tegra::Texture {
+
+using VideoCore::Surface::PixelFormat;
+
+template <bool reverse>
+void SwapS8Z24ToZ24S8(u8* data, u32 width, u32 height) {
+    union S8Z24 {
+        BitField<0, 24, u32> z24;
+        BitField<24, 8, u32> s8;
+    };
+    static_assert(sizeof(S8Z24) == 4, "S8Z24 is incorrect size");
+
+    union Z24S8 {
+        BitField<0, 8, u32> s8;
+        BitField<8, 24, u32> z24;
+    };
+    static_assert(sizeof(Z24S8) == 4, "Z24S8 is incorrect size");
+
+    S8Z24 s8z24_pixel{};
+    Z24S8 z24s8_pixel{};
+    constexpr auto bpp{
+        VideoCore::Surface::GetBytesPerPixel(VideoCore::Surface::PixelFormat::S8Z24)};
+    for (std::size_t y = 0; y < height; ++y) {
+        for (std::size_t x = 0; x < width; ++x) {
+            const std::size_t offset{bpp * (y * width + x)};
+            if constexpr (reverse) {
+                std::memcpy(&z24s8_pixel, &data[offset], sizeof(Z24S8));
+                s8z24_pixel.s8.Assign(z24s8_pixel.s8);
+                s8z24_pixel.z24.Assign(z24s8_pixel.z24);
+                std::memcpy(&data[offset], &s8z24_pixel, sizeof(S8Z24));
+            } else {
+                std::memcpy(&s8z24_pixel, &data[offset], sizeof(S8Z24));
+                z24s8_pixel.s8.Assign(s8z24_pixel.s8);
+                z24s8_pixel.z24.Assign(s8z24_pixel.z24);
+                std::memcpy(&data[offset], &z24s8_pixel, sizeof(Z24S8));
+            }
+        }
+    }
+}
+
+static void ConvertS8Z24ToZ24S8(u8* data, u32 width, u32 height) {
+    SwapS8Z24ToZ24S8<false>(data, width, height);
+}
+
+static void ConvertZ24S8ToS8Z24(u8* data, u32 width, u32 height) {
+    SwapS8Z24ToZ24S8<true>(data, width, height);
+}
+
+void ConvertFromGuestToHost(u8* data, PixelFormat pixel_format, u32 width, u32 height, u32 depth,
+                            bool convert_astc, bool convert_s8z24) {
+    if (convert_astc && IsPixelFormatASTC(pixel_format)) {
+        // Convert ASTC pixel formats to RGBA8, as most desktop GPUs do not support ASTC.
+        u32 block_width{};
+        u32 block_height{};
+        std::tie(block_width, block_height) = GetASTCBlockSize(pixel_format);
+        const std::vector<u8> rgba8_data =
+            Tegra::Texture::ASTC::Decompress(data, width, height, depth, block_width, block_height);
+        std::copy(rgba8_data.begin(), rgba8_data.end(), data);
+
+    } else if (convert_s8z24 && pixel_format == PixelFormat::S8Z24) {
+        Tegra::Texture::ConvertS8Z24ToZ24S8(data, width, height);
+    }
+}
+
+void ConvertFromHostToGuest(u8* data, PixelFormat pixel_format, u32 width, u32 height, u32 depth,
+                            bool convert_astc, bool convert_s8z24) {
+    if (convert_astc && IsPixelFormatASTC(pixel_format)) {
+        LOG_CRITICAL(HW_GPU, "Conversion of format {} after texture flushing is not implemented",
+                     static_cast<u32>(pixel_format));
+        UNREACHABLE();
+
+    } else if (convert_s8z24 && pixel_format == PixelFormat::S8Z24) {
+        Tegra::Texture::ConvertZ24S8ToS8Z24(data, width, height);
+    }
+}
+
+} // namespace Tegra::Texture
\ No newline at end of file
diff --git a/src/video_core/textures/convert.h b/src/video_core/textures/convert.h
new file mode 100644
index 000000000..07cd8b5da
--- /dev/null
+++ b/src/video_core/textures/convert.h
@@ -0,0 +1,18 @@
+// 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"
+#include "video_core/surface.h"
+
+namespace Tegra::Texture {
+
+void ConvertFromGuestToHost(u8* data, VideoCore::Surface::PixelFormat pixel_format, u32 width,
+                            u32 height, u32 depth, bool convert_astc, bool convert_s8z24);
+
+void ConvertFromHostToGuest(u8* data, VideoCore::Surface::PixelFormat pixel_format, u32 width,
+                            u32 height, u32 depth, bool convert_astc, bool convert_s8z24);
+
+} // namespace Tegra::Texture
\ No newline at end of file
diff --git a/src/video_core/textures/decoders.cpp b/src/video_core/textures/decoders.cpp
index 5db75de22..995d0e068 100644
--- a/src/video_core/textures/decoders.cpp
+++ b/src/video_core/textures/decoders.cpp
@@ -6,7 +6,6 @@
 #include <cstring>
 #include "common/alignment.h"
 #include "common/assert.h"
-#include "core/memory.h"
 #include "video_core/gpu.h"
 #include "video_core/textures/decoders.h"
 #include "video_core/textures/texture.h"
@@ -103,8 +102,8 @@ void FastProcessBlock(u8* const swizzled_data, u8* const unswizzled_data, const
                 const u32 swizzle_offset{y_address + table[(xb / fast_swizzle_align) % 4]};
                 const u32 out_x = xb * out_bytes_per_pixel / bytes_per_pixel;
                 const u32 pixel_index{out_x + pixel_base};
-                data_ptrs[unswizzle] = swizzled_data + swizzle_offset;
-                data_ptrs[!unswizzle] = unswizzled_data + pixel_index;
+                data_ptrs[unswizzle ? 1 : 0] = swizzled_data + swizzle_offset;
+                data_ptrs[unswizzle ? 0 : 1] = unswizzled_data + pixel_index;
                 std::memcpy(data_ptrs[0], data_ptrs[1], fast_swizzle_align);
             }
             pixel_base += stride_x;
@@ -154,7 +153,7 @@ void SwizzledData(u8* const swizzled_data, u8* const unswizzled_data, const bool
             for (u32 xb = 0; xb < blocks_on_x; xb++) {
                 const u32 x_start = xb * block_x_elements;
                 const u32 x_end = std::min(width, x_start + block_x_elements);
-                if (fast) {
+                if constexpr (fast) {
                     FastProcessBlock(swizzled_data, unswizzled_data, unswizzle, x_start, y_start,
                                      z_start, x_end, y_end, z_end, tile_offset, xy_block_size,
                                      layer_z, stride_x, bytes_per_pixel, out_bytes_per_pixel);
@@ -230,18 +229,18 @@ u32 BytesPerPixel(TextureFormat format) {
     }
 }
 
-void UnswizzleTexture(u8* const unswizzled_data, VAddr address, u32 tile_size_x, u32 tile_size_y,
+void UnswizzleTexture(u8* const unswizzled_data, u8* address, u32 tile_size_x, u32 tile_size_y,
                       u32 bytes_per_pixel, u32 width, u32 height, u32 depth, u32 block_height,
                       u32 block_depth, u32 width_spacing) {
     CopySwizzledData((width + tile_size_x - 1) / tile_size_x,
                      (height + tile_size_y - 1) / tile_size_y, depth, bytes_per_pixel,
-                     bytes_per_pixel, Memory::GetPointer(address), unswizzled_data, true,
-                     block_height, block_depth, width_spacing);
+                     bytes_per_pixel, address, unswizzled_data, true, block_height, block_depth,
+                     width_spacing);
 }
 
-std::vector<u8> UnswizzleTexture(VAddr address, u32 tile_size_x, u32 tile_size_y,
-                                 u32 bytes_per_pixel, u32 width, u32 height, u32 depth,
-                                 u32 block_height, u32 block_depth, u32 width_spacing) {
+std::vector<u8> UnswizzleTexture(u8* address, u32 tile_size_x, u32 tile_size_y, u32 bytes_per_pixel,
+                                 u32 width, u32 height, u32 depth, u32 block_height,
+                                 u32 block_depth, u32 width_spacing) {
     std::vector<u8> unswizzled_data(width * height * depth * bytes_per_pixel);
     UnswizzleTexture(unswizzled_data.data(), address, tile_size_x, tile_size_y, bytes_per_pixel,
                      width, height, depth, block_height, block_depth, width_spacing);
@@ -249,8 +248,7 @@ std::vector<u8> UnswizzleTexture(VAddr address, u32 tile_size_x, u32 tile_size_y
 }
 
 void SwizzleSubrect(u32 subrect_width, u32 subrect_height, u32 source_pitch, u32 swizzled_width,
-                    u32 bytes_per_pixel, VAddr swizzled_data, VAddr unswizzled_data,
-                    u32 block_height) {
+                    u32 bytes_per_pixel, u8* swizzled_data, u8* unswizzled_data, u32 block_height) {
     const u32 image_width_in_gobs{(swizzled_width * bytes_per_pixel + (gob_size_x - 1)) /
                                   gob_size_x};
     for (u32 line = 0; line < subrect_height; ++line) {
@@ -262,17 +260,17 @@ void SwizzleSubrect(u32 subrect_width, u32 subrect_height, u32 source_pitch, u32
             const u32 gob_address =
                 gob_address_y + (x * bytes_per_pixel / gob_size_x) * gob_size * block_height;
             const u32 swizzled_offset = gob_address + table[(x * bytes_per_pixel) % gob_size_x];
-            const VAddr source_line = unswizzled_data + line * source_pitch + x * bytes_per_pixel;
-            const VAddr dest_addr = swizzled_data + swizzled_offset;
+            u8* source_line = unswizzled_data + line * source_pitch + x * bytes_per_pixel;
+            u8* dest_addr = swizzled_data + swizzled_offset;
 
-            Memory::CopyBlock(dest_addr, source_line, bytes_per_pixel);
+            std::memcpy(dest_addr, source_line, bytes_per_pixel);
         }
     }
 }
 
 void UnswizzleSubrect(u32 subrect_width, u32 subrect_height, u32 dest_pitch, u32 swizzled_width,
-                      u32 bytes_per_pixel, VAddr swizzled_data, VAddr unswizzled_data,
-                      u32 block_height, u32 offset_x, u32 offset_y) {
+                      u32 bytes_per_pixel, u8* swizzled_data, u8* unswizzled_data, u32 block_height,
+                      u32 offset_x, u32 offset_y) {
     for (u32 line = 0; line < subrect_height; ++line) {
         const u32 y2 = line + offset_y;
         const u32 gob_address_y = (y2 / (gob_size_y * block_height)) * gob_size * block_height +
@@ -282,10 +280,10 @@ void UnswizzleSubrect(u32 subrect_width, u32 subrect_height, u32 dest_pitch, u32
             const u32 x2 = (x + offset_x) * bytes_per_pixel;
             const u32 gob_address = gob_address_y + (x2 / gob_size_x) * gob_size * block_height;
             const u32 swizzled_offset = gob_address + table[x2 % gob_size_x];
-            const VAddr dest_line = unswizzled_data + line * dest_pitch + x * bytes_per_pixel;
-            const VAddr source_addr = swizzled_data + swizzled_offset;
+            u8* dest_line = unswizzled_data + line * dest_pitch + x * bytes_per_pixel;
+            u8* source_addr = swizzled_data + swizzled_offset;
 
-            Memory::CopyBlock(dest_line, source_addr, bytes_per_pixel);
+            std::memcpy(dest_line, source_addr, bytes_per_pixel);
         }
     }
 }
diff --git a/src/video_core/textures/decoders.h b/src/video_core/textures/decoders.h
index 85b7e9f7b..e078fa274 100644
--- a/src/video_core/textures/decoders.h
+++ b/src/video_core/textures/decoders.h
@@ -16,18 +16,15 @@ inline std::size_t GetGOBSize() {
     return 512;
 }
 
-/**
- * Unswizzles a swizzled texture without changing its format.
- */
-void UnswizzleTexture(u8* unswizzled_data, VAddr address, u32 tile_size_x, u32 tile_size_y,
+/// Unswizzles a swizzled texture without changing its format.
+void UnswizzleTexture(u8* unswizzled_data, u8* address, u32 tile_size_x, u32 tile_size_y,
                       u32 bytes_per_pixel, u32 width, u32 height, u32 depth,
                       u32 block_height = TICEntry::DefaultBlockHeight,
                       u32 block_depth = TICEntry::DefaultBlockHeight, u32 width_spacing = 0);
-/**
- * Unswizzles a swizzled texture without changing its format.
- */
-std::vector<u8> UnswizzleTexture(VAddr address, u32 tile_size_x, u32 tile_size_y,
-                                 u32 bytes_per_pixel, u32 width, u32 height, u32 depth,
+
+/// Unswizzles a swizzled texture without changing its format.
+std::vector<u8> UnswizzleTexture(u8* address, u32 tile_size_x, u32 tile_size_y, u32 bytes_per_pixel,
+                                 u32 width, u32 height, u32 depth,
                                  u32 block_height = TICEntry::DefaultBlockHeight,
                                  u32 block_depth = TICEntry::DefaultBlockHeight,
                                  u32 width_spacing = 0);
@@ -37,25 +34,21 @@ void CopySwizzledData(u32 width, u32 height, u32 depth, u32 bytes_per_pixel,
                       u32 out_bytes_per_pixel, u8* swizzled_data, u8* unswizzled_data,
                       bool unswizzle, u32 block_height, u32 block_depth, u32 width_spacing);
 
-/**
- * Decodes an unswizzled texture into a A8R8G8B8 texture.
- */
+/// Decodes an unswizzled texture into a A8R8G8B8 texture.
 std::vector<u8> DecodeTexture(const std::vector<u8>& texture_data, TextureFormat format, u32 width,
                               u32 height);
 
-/**
- * This function calculates the correct size of a texture depending if it's tiled or not.
- */
+/// This function calculates the correct size of a texture depending if it's tiled or not.
 std::size_t CalculateSize(bool tiled, u32 bytes_per_pixel, u32 width, u32 height, u32 depth,
                           u32 block_height, u32 block_depth);
 
 /// Copies an untiled subrectangle into a tiled surface.
 void SwizzleSubrect(u32 subrect_width, u32 subrect_height, u32 source_pitch, u32 swizzled_width,
-                    u32 bytes_per_pixel, VAddr swizzled_data, VAddr unswizzled_data,
-                    u32 block_height);
+                    u32 bytes_per_pixel, u8* swizzled_data, u8* unswizzled_data, u32 block_height);
+
 /// Copies a tiled subrectangle into a linear surface.
 void UnswizzleSubrect(u32 subrect_width, u32 subrect_height, u32 dest_pitch, u32 swizzled_width,
-                      u32 bytes_per_pixel, VAddr swizzled_data, VAddr unswizzled_data,
-                      u32 block_height, u32 offset_x, u32 offset_y);
+                      u32 bytes_per_pixel, u8* swizzled_data, u8* unswizzled_data, u32 block_height,
+                      u32 offset_x, u32 offset_y);
 
 } // namespace Tegra::Texture
diff --git a/src/video_core/textures/texture.h b/src/video_core/textures/texture.h
index e7c78bee2..93ecc6e31 100644
--- a/src/video_core/textures/texture.h
+++ b/src/video_core/textures/texture.h
@@ -4,6 +4,7 @@
 
 #pragma once
 
+#include <array>
 #include "common/assert.h"
 #include "common/bit_field.h"
 #include "common/common_funcs.h"
@@ -182,7 +183,7 @@ struct TICEntry {
     };
     union {
         BitField<0, 16, u32> height_minus_1;
-        BitField<16, 15, u32> depth_minus_1;
+        BitField<16, 14, u32> depth_minus_1;
     };
     union {
         BitField<6, 13, u32> mip_lod_bias;
@@ -282,34 +283,62 @@ enum class TextureMipmapFilter : u32 {
 
 struct TSCEntry {
     union {
-        BitField<0, 3, WrapMode> wrap_u;
-        BitField<3, 3, WrapMode> wrap_v;
-        BitField<6, 3, WrapMode> wrap_p;
-        BitField<9, 1, u32> depth_compare_enabled;
-        BitField<10, 3, DepthCompareFunc> depth_compare_func;
-        BitField<13, 1, u32> srgb_conversion;
-        BitField<20, 3, u32> max_anisotropy;
+        struct {
+            union {
+                BitField<0, 3, WrapMode> wrap_u;
+                BitField<3, 3, WrapMode> wrap_v;
+                BitField<6, 3, WrapMode> wrap_p;
+                BitField<9, 1, u32> depth_compare_enabled;
+                BitField<10, 3, DepthCompareFunc> depth_compare_func;
+                BitField<13, 1, u32> srgb_conversion;
+                BitField<20, 3, u32> max_anisotropy;
+            };
+            union {
+                BitField<0, 2, TextureFilter> mag_filter;
+                BitField<4, 2, TextureFilter> min_filter;
+                BitField<6, 2, TextureMipmapFilter> mipmap_filter;
+                BitField<9, 1, u32> cubemap_interface_filtering;
+                BitField<12, 13, u32> mip_lod_bias;
+            };
+            union {
+                BitField<0, 12, u32> min_lod_clamp;
+                BitField<12, 12, u32> max_lod_clamp;
+                BitField<24, 8, u32> srgb_border_color_r;
+            };
+            union {
+                BitField<12, 8, u32> srgb_border_color_g;
+                BitField<20, 8, u32> srgb_border_color_b;
+            };
+            std::array<f32, 4> border_color;
+        };
+        std::array<u8, 0x20> raw;
     };
-    union {
-        BitField<0, 2, TextureFilter> mag_filter;
-        BitField<4, 2, TextureFilter> min_filter;
-        BitField<6, 2, TextureMipmapFilter> mip_filter;
-        BitField<9, 1, u32> cubemap_interface_filtering;
-        BitField<12, 13, u32> mip_lod_bias;
-    };
-    union {
-        BitField<0, 12, u32> min_lod_clamp;
-        BitField<12, 12, u32> max_lod_clamp;
-        BitField<24, 8, u32> srgb_border_color_r;
-    };
-    union {
-        BitField<12, 8, u32> srgb_border_color_g;
-        BitField<20, 8, u32> srgb_border_color_b;
-    };
-    float border_color_r;
-    float border_color_g;
-    float border_color_b;
-    float border_color_a;
+
+    float GetMaxAnisotropy() const {
+        return static_cast<float>(1U << max_anisotropy);
+    }
+
+    float GetMinLod() const {
+        return static_cast<float>(min_lod_clamp) / 256.0f;
+    }
+
+    float GetMaxLod() const {
+        return static_cast<float>(max_lod_clamp) / 256.0f;
+    }
+
+    float GetLodBias() const {
+        // Sign extend the 13-bit value.
+        constexpr u32 mask = 1U << (13 - 1);
+        return static_cast<s32>((mip_lod_bias ^ mask) - mask) / 256.0f;
+    }
+
+    std::array<float, 4> GetBorderColor() const {
+        if (srgb_conversion) {
+            return {srgb_border_color_r / 255.0f, srgb_border_color_g / 255.0f,
+                    srgb_border_color_b / 255.0f, border_color[3]};
+        }
+        return border_color;
+    }
 };
 static_assert(sizeof(TSCEntry) == 0x20, "TSCEntry has wrong size");
 
@@ -317,7 +346,6 @@ struct FullTextureInfo {
     u32 index;
     TICEntry tic;
     TSCEntry tsc;
-    bool enabled;
 };
 
 /// Returns the number of bytes per pixel of the input texture format.
diff --git a/src/video_core/video_core.cpp b/src/video_core/video_core.cpp
index 0b8ccdd44..cb82ecf3f 100644
--- a/src/video_core/video_core.cpp
+++ b/src/video_core/video_core.cpp
@@ -11,8 +11,9 @@
 
 namespace VideoCore {
 
-std::unique_ptr<RendererBase> CreateRenderer(Core::Frontend::EmuWindow& emu_window) {
-    return std::make_unique<OpenGL::RendererOpenGL>(emu_window);
+std::unique_ptr<RendererBase> CreateRenderer(Core::Frontend::EmuWindow& emu_window,
+                                             Core::System& system) {
+    return std::make_unique<OpenGL::RendererOpenGL>(emu_window, system);
 }
 
 u16 GetResolutionScaleFactor(const RendererBase& renderer) {
diff --git a/src/video_core/video_core.h b/src/video_core/video_core.h
index 5b373bcb1..3c583f195 100644
--- a/src/video_core/video_core.h
+++ b/src/video_core/video_core.h
@@ -6,6 +6,10 @@
 
 #include <memory>
 
+namespace Core {
+class System;
+}
+
 namespace Core::Frontend {
 class EmuWindow;
 }
@@ -20,7 +24,8 @@ class RendererBase;
  * @note The returned renderer instance is simply allocated. Its Init()
  *       function still needs to be called to fully complete its setup.
  */
-std::unique_ptr<RendererBase> CreateRenderer(Core::Frontend::EmuWindow& emu_window);
+std::unique_ptr<RendererBase> CreateRenderer(Core::Frontend::EmuWindow& emu_window,
+                                             Core::System& system);
 
 u16 GetResolutionScaleFactor(const RendererBase& renderer);
 
diff --git a/src/web_service/verify_login.h b/src/web_service/verify_login.h
index 39db32dbb..821b345d7 100644
--- a/src/web_service/verify_login.h
+++ b/src/web_service/verify_login.h
@@ -4,8 +4,6 @@
 
 #pragma once
 
-#include <functional>
-#include <future>
 #include <string>
 
 namespace WebService {
diff --git a/src/web_service/web_backend.cpp b/src/web_service/web_backend.cpp
index b7737b615..40da1a4e2 100644
--- a/src/web_service/web_backend.cpp
+++ b/src/web_service/web_backend.cpp
@@ -10,7 +10,6 @@
 #include "common/common_types.h"
 #include "common/logging/log.h"
 #include "common/web_result.h"
-#include "core/settings.h"
 #include "web_service/web_backend.h"
 
 namespace WebService {
diff --git a/src/yuzu/applets/web_browser.cpp b/src/yuzu/applets/web_browser.cpp
index 6a9138d53..979b9ec14 100644
--- a/src/yuzu/applets/web_browser.cpp
+++ b/src/yuzu/applets/web_browser.cpp
@@ -56,6 +56,8 @@ constexpr char NX_SHIM_INJECT_SCRIPT[] = R"(
     window.nx.endApplet = function() {
         applet_done = true;
     };
+
+    window.onkeypress = function(e) { if (e.keyCode === 13) { applet_done = true; } };
 )";
 
 QString GetNXShimInjectionScript() {
diff --git a/src/yuzu/bootmanager.cpp b/src/yuzu/bootmanager.cpp
index f74cb693a..05ad19e1d 100644
--- a/src/yuzu/bootmanager.cpp
+++ b/src/yuzu/bootmanager.cpp
@@ -20,16 +20,25 @@
 EmuThread::EmuThread(GRenderWindow* render_window) : render_window(render_window) {}
 
 void EmuThread::run() {
-    if (!Settings::values.use_multi_core) {
-        // Single core mode must acquire OpenGL context for entire emulation session
-        render_window->MakeCurrent();
-    }
+    render_window->MakeCurrent();
 
     MicroProfileOnThreadCreate("EmuThread");
 
-    stop_run = false;
+    emit LoadProgress(VideoCore::LoadCallbackStage::Prepare, 0, 0);
+
+    Core::System::GetInstance().Renderer().Rasterizer().LoadDiskResources(
+        stop_run, [this](VideoCore::LoadCallbackStage stage, std::size_t value, std::size_t total) {
+            emit LoadProgress(stage, value, total);
+        });
+
+    emit LoadProgress(VideoCore::LoadCallbackStage::Complete, 0, 0);
+
+    if (Settings::values.use_asynchronous_gpu_emulation) {
+        // Release OpenGL context for the GPU thread
+        render_window->DoneCurrent();
+    }
 
-    // holds whether the cpu was running during the last iteration,
+    // Holds whether the cpu was running during the last iteration,
     // so that the DebugModeLeft signal can be emitted before the
     // next execution step
     bool was_active = false;
@@ -112,7 +121,6 @@ GRenderWindow::GRenderWindow(QWidget* parent, EmuThread* emu_thread)
     setAttribute(Qt::WA_AcceptTouchEvents);
 
     InputCommon::Init();
-    InputCommon::StartJoystickEventHandler();
     connect(this, &GRenderWindow::FirstFrameDisplayed, static_cast<GMainWindow*>(parent),
             &GMainWindow::OnLoadComplete);
 }
diff --git a/src/yuzu/bootmanager.h b/src/yuzu/bootmanager.h
index d1f37e503..7226e690e 100644
--- a/src/yuzu/bootmanager.h
+++ b/src/yuzu/bootmanager.h
@@ -22,6 +22,10 @@ class GGLWidgetInternal;
 class GMainWindow;
 class GRenderWindow;
 
+namespace VideoCore {
+enum class LoadCallbackStage;
+}
+
 class EmuThread : public QThread {
     Q_OBJECT
 
@@ -75,7 +79,7 @@ public:
 private:
     bool exec_step = false;
     bool running = false;
-    std::atomic<bool> stop_run{false};
+    std::atomic_bool stop_run{false};
     std::mutex running_mutex;
     std::condition_variable running_cv;
 
@@ -101,6 +105,8 @@ signals:
     void DebugModeLeft();
 
     void ErrorThrown(Core::System::ResultStatus, std::string);
+
+    void LoadProgress(VideoCore::LoadCallbackStage stage, std::size_t value, std::size_t total);
 };
 
 class GRenderWindow : public QWidget, public Core::Frontend::EmuWindow {
diff --git a/src/yuzu/compatdb.cpp b/src/yuzu/compatdb.cpp
index 5f0896f84..c8b0a5ec0 100644
--- a/src/yuzu/compatdb.cpp
+++ b/src/yuzu/compatdb.cpp
@@ -53,15 +53,15 @@ void CompatDB::Submit() {
     case CompatDBPage::Final:
         back();
         LOG_DEBUG(Frontend, "Compatibility Rating: {}", compatibility->checkedId());
-        Core::Telemetry().AddField(Telemetry::FieldType::UserFeedback, "Compatibility",
-                                   compatibility->checkedId());
+        Core::System::GetInstance().TelemetrySession().AddField(
+            Telemetry::FieldType::UserFeedback, "Compatibility", compatibility->checkedId());
 
         button(NextButton)->setEnabled(false);
         button(NextButton)->setText(tr("Submitting"));
         button(QWizard::CancelButton)->setVisible(false);
 
         testcase_watcher.setFuture(QtConcurrent::run(
-            [this]() { return Core::System::GetInstance().TelemetrySession().SubmitTestcase(); }));
+            [] { return Core::System::GetInstance().TelemetrySession().SubmitTestcase(); }));
         break;
     default:
         LOG_ERROR(Frontend, "Unexpected page: {}", currentId());
diff --git a/src/yuzu/configuration/config.cpp b/src/yuzu/configuration/config.cpp
index ddf4cf552..4650f96a3 100644
--- a/src/yuzu/configuration/config.cpp
+++ b/src/yuzu/configuration/config.cpp
@@ -209,7 +209,7 @@ void Config::ReadPlayerValues() {
     for (std::size_t p = 0; p < Settings::values.players.size(); ++p) {
         auto& player = Settings::values.players[p];
 
-        player.connected = qt_config->value(QString("player_%1_connected").arg(p), false).toBool();
+        player.connected = ReadSetting(QString("player_%1_connected").arg(p), false).toBool();
 
         player.type = static_cast<Settings::ControllerType>(
             qt_config
@@ -269,7 +269,7 @@ void Config::ReadPlayerValues() {
 }
 
 void Config::ReadDebugValues() {
-    Settings::values.debug_pad_enabled = qt_config->value("debug_pad_enabled", false).toBool();
+    Settings::values.debug_pad_enabled = ReadSetting("debug_pad_enabled", false).toBool();
     for (int i = 0; i < Settings::NativeButton::NumButtons; ++i) {
         std::string default_param = InputCommon::GenerateKeyboardParam(default_buttons[i]);
         Settings::values.debug_pad_buttons[i] =
@@ -298,7 +298,7 @@ void Config::ReadDebugValues() {
 }
 
 void Config::ReadKeyboardValues() {
-    Settings::values.keyboard_enabled = qt_config->value("keyboard_enabled", false).toBool();
+    Settings::values.keyboard_enabled = ReadSetting("keyboard_enabled", false).toBool();
 
     std::transform(default_keyboard_keys.begin(), default_keyboard_keys.end(),
                    Settings::values.keyboard_keys.begin(), InputCommon::GenerateKeyboardParam);
@@ -311,7 +311,7 @@ void Config::ReadKeyboardValues() {
 }
 
 void Config::ReadMouseValues() {
-    Settings::values.mouse_enabled = qt_config->value("mouse_enabled", false).toBool();
+    Settings::values.mouse_enabled = ReadSetting("mouse_enabled", false).toBool();
 
     for (int i = 0; i < Settings::NativeMouseButton::NumMouseButtons; ++i) {
         std::string default_param = InputCommon::GenerateKeyboardParam(default_mouse_buttons[i]);
@@ -327,16 +327,14 @@ void Config::ReadMouseValues() {
 }
 
 void Config::ReadTouchscreenValues() {
-    Settings::values.touchscreen.enabled = qt_config->value("touchscreen_enabled", true).toBool();
+    Settings::values.touchscreen.enabled = ReadSetting("touchscreen_enabled", true).toBool();
     Settings::values.touchscreen.device =
-        qt_config->value("touchscreen_device", "engine:emu_window").toString().toStdString();
-
-    Settings::values.touchscreen.finger = qt_config->value("touchscreen_finger", 0).toUInt();
-    Settings::values.touchscreen.rotation_angle = qt_config->value("touchscreen_angle", 0).toUInt();
-    Settings::values.touchscreen.diameter_x =
-        qt_config->value("touchscreen_diameter_x", 15).toUInt();
-    Settings::values.touchscreen.diameter_y =
-        qt_config->value("touchscreen_diameter_y", 15).toUInt();
+        ReadSetting("touchscreen_device", "engine:emu_window").toString().toStdString();
+
+    Settings::values.touchscreen.finger = ReadSetting("touchscreen_finger", 0).toUInt();
+    Settings::values.touchscreen.rotation_angle = ReadSetting("touchscreen_angle", 0).toUInt();
+    Settings::values.touchscreen.diameter_x = ReadSetting("touchscreen_diameter_x", 15).toUInt();
+    Settings::values.touchscreen.diameter_y = ReadSetting("touchscreen_diameter_y", 15).toUInt();
     qt_config->endGroup();
 }
 
@@ -357,38 +355,41 @@ void Config::ReadValues() {
     ReadTouchscreenValues();
 
     Settings::values.motion_device =
-        qt_config->value("motion_device", "engine:motion_emu,update_period:100,sensitivity:0.01")
+        ReadSetting("motion_device", "engine:motion_emu,update_period:100,sensitivity:0.01")
             .toString()
             .toStdString();
 
     qt_config->beginGroup("Core");
-    Settings::values.use_cpu_jit = qt_config->value("use_cpu_jit", true).toBool();
-    Settings::values.use_multi_core = qt_config->value("use_multi_core", false).toBool();
+    Settings::values.use_cpu_jit = ReadSetting("use_cpu_jit", true).toBool();
+    Settings::values.use_multi_core = ReadSetting("use_multi_core", false).toBool();
     qt_config->endGroup();
 
     qt_config->beginGroup("Renderer");
-    Settings::values.resolution_factor = qt_config->value("resolution_factor", 1.0).toFloat();
-    Settings::values.use_frame_limit = qt_config->value("use_frame_limit", true).toBool();
-    Settings::values.frame_limit = qt_config->value("frame_limit", 100).toInt();
+    Settings::values.resolution_factor = ReadSetting("resolution_factor", 1.0).toFloat();
+    Settings::values.use_frame_limit = ReadSetting("use_frame_limit", true).toBool();
+    Settings::values.frame_limit = ReadSetting("frame_limit", 100).toInt();
+    Settings::values.use_disk_shader_cache = ReadSetting("use_disk_shader_cache", true).toBool();
     Settings::values.use_accurate_gpu_emulation =
-        qt_config->value("use_accurate_gpu_emulation", false).toBool();
+        ReadSetting("use_accurate_gpu_emulation", false).toBool();
+    Settings::values.use_asynchronous_gpu_emulation =
+        ReadSetting("use_asynchronous_gpu_emulation", false).toBool();
 
-    Settings::values.bg_red = qt_config->value("bg_red", 0.0).toFloat();
-    Settings::values.bg_green = qt_config->value("bg_green", 0.0).toFloat();
-    Settings::values.bg_blue = qt_config->value("bg_blue", 0.0).toFloat();
+    Settings::values.bg_red = ReadSetting("bg_red", 0.0).toFloat();
+    Settings::values.bg_green = ReadSetting("bg_green", 0.0).toFloat();
+    Settings::values.bg_blue = ReadSetting("bg_blue", 0.0).toFloat();
     qt_config->endGroup();
 
     qt_config->beginGroup("Audio");
-    Settings::values.sink_id = qt_config->value("output_engine", "auto").toString().toStdString();
+    Settings::values.sink_id = ReadSetting("output_engine", "auto").toString().toStdString();
     Settings::values.enable_audio_stretching =
-        qt_config->value("enable_audio_stretching", true).toBool();
+        ReadSetting("enable_audio_stretching", true).toBool();
     Settings::values.audio_device_id =
-        qt_config->value("output_device", "auto").toString().toStdString();
-    Settings::values.volume = qt_config->value("volume", 1).toFloat();
+        ReadSetting("output_device", "auto").toString().toStdString();
+    Settings::values.volume = ReadSetting("volume", 1).toFloat();
     qt_config->endGroup();
 
     qt_config->beginGroup("Data Storage");
-    Settings::values.use_virtual_sd = qt_config->value("use_virtual_sd", true).toBool();
+    Settings::values.use_virtual_sd = ReadSetting("use_virtual_sd", true).toBool();
     FileUtil::GetUserPath(
         FileUtil::UserPath::NANDDir,
         qt_config
@@ -406,30 +407,30 @@ void Config::ReadValues() {
     qt_config->endGroup();
 
     qt_config->beginGroup("Core");
-    Settings::values.use_cpu_jit = qt_config->value("use_cpu_jit", true).toBool();
-    Settings::values.use_multi_core = qt_config->value("use_multi_core", false).toBool();
+    Settings::values.use_cpu_jit = ReadSetting("use_cpu_jit", true).toBool();
+    Settings::values.use_multi_core = ReadSetting("use_multi_core", false).toBool();
     qt_config->endGroup();
 
     qt_config->beginGroup("System");
-    Settings::values.use_docked_mode = qt_config->value("use_docked_mode", false).toBool();
-    Settings::values.enable_nfc = qt_config->value("enable_nfc", true).toBool();
+    Settings::values.use_docked_mode = ReadSetting("use_docked_mode", false).toBool();
+    Settings::values.enable_nfc = ReadSetting("enable_nfc", true).toBool();
 
-    Settings::values.current_user = std::clamp<int>(qt_config->value("current_user", 0).toInt(), 0,
-                                                    Service::Account::MAX_USERS - 1);
+    Settings::values.current_user =
+        std::clamp<int>(ReadSetting("current_user", 0).toInt(), 0, Service::Account::MAX_USERS - 1);
 
-    Settings::values.language_index = qt_config->value("language_index", 1).toInt();
+    Settings::values.language_index = ReadSetting("language_index", 1).toInt();
 
-    const auto rng_seed_enabled = qt_config->value("rng_seed_enabled", false).toBool();
+    const auto rng_seed_enabled = ReadSetting("rng_seed_enabled", false).toBool();
     if (rng_seed_enabled) {
-        Settings::values.rng_seed = qt_config->value("rng_seed", 0).toULongLong();
+        Settings::values.rng_seed = ReadSetting("rng_seed", 0).toULongLong();
     } else {
         Settings::values.rng_seed = std::nullopt;
     }
 
-    const auto custom_rtc_enabled = qt_config->value("custom_rtc_enabled", false).toBool();
+    const auto custom_rtc_enabled = ReadSetting("custom_rtc_enabled", false).toBool();
     if (custom_rtc_enabled) {
         Settings::values.custom_rtc =
-            std::chrono::seconds(qt_config->value("custom_rtc", 0).toULongLong());
+            std::chrono::seconds(ReadSetting("custom_rtc", 0).toULongLong());
     } else {
         Settings::values.custom_rtc = std::nullopt;
     }
@@ -437,35 +438,35 @@ void Config::ReadValues() {
     qt_config->endGroup();
 
     qt_config->beginGroup("Miscellaneous");
-    Settings::values.log_filter = qt_config->value("log_filter", "*:Info").toString().toStdString();
-    Settings::values.use_dev_keys = qt_config->value("use_dev_keys", false).toBool();
+    Settings::values.log_filter = ReadSetting("log_filter", "*:Info").toString().toStdString();
+    Settings::values.use_dev_keys = ReadSetting("use_dev_keys", false).toBool();
     qt_config->endGroup();
 
     qt_config->beginGroup("Debugging");
-    Settings::values.use_gdbstub = qt_config->value("use_gdbstub", false).toBool();
-    Settings::values.gdbstub_port = qt_config->value("gdbstub_port", 24689).toInt();
-    Settings::values.program_args = qt_config->value("program_args", "").toString().toStdString();
-    Settings::values.dump_exefs = qt_config->value("dump_exefs", false).toBool();
-    Settings::values.dump_nso = qt_config->value("dump_nso", false).toBool();
+    Settings::values.use_gdbstub = ReadSetting("use_gdbstub", false).toBool();
+    Settings::values.gdbstub_port = ReadSetting("gdbstub_port", 24689).toInt();
+    Settings::values.program_args = ReadSetting("program_args", "").toString().toStdString();
+    Settings::values.dump_exefs = ReadSetting("dump_exefs", false).toBool();
+    Settings::values.dump_nso = ReadSetting("dump_nso", false).toBool();
     qt_config->endGroup();
 
     qt_config->beginGroup("WebService");
-    Settings::values.enable_telemetry = qt_config->value("enable_telemetry", true).toBool();
+    Settings::values.enable_telemetry = ReadSetting("enable_telemetry", true).toBool();
     Settings::values.web_api_url =
-        qt_config->value("web_api_url", "https://api.yuzu-emu.org").toString().toStdString();
-    Settings::values.yuzu_username = qt_config->value("yuzu_username").toString().toStdString();
-    Settings::values.yuzu_token = qt_config->value("yuzu_token").toString().toStdString();
+        ReadSetting("web_api_url", "https://api.yuzu-emu.org").toString().toStdString();
+    Settings::values.yuzu_username = ReadSetting("yuzu_username").toString().toStdString();
+    Settings::values.yuzu_token = ReadSetting("yuzu_token").toString().toStdString();
     qt_config->endGroup();
 
     const auto size = qt_config->beginReadArray("DisabledAddOns");
     for (int i = 0; i < size; ++i) {
         qt_config->setArrayIndex(i);
-        const auto title_id = qt_config->value("title_id", 0).toULongLong();
+        const auto title_id = ReadSetting("title_id", 0).toULongLong();
         std::vector<std::string> out;
         const auto d_size = qt_config->beginReadArray("disabled");
         for (int j = 0; j < d_size; ++j) {
             qt_config->setArrayIndex(j);
-            out.push_back(qt_config->value("d", "").toString().toStdString());
+            out.push_back(ReadSetting("d", "").toString().toStdString());
         }
         qt_config->endArray();
         Settings::values.disabled_addons.insert_or_assign(title_id, out);
@@ -473,41 +474,38 @@ void Config::ReadValues() {
     qt_config->endArray();
 
     qt_config->beginGroup("UI");
-    UISettings::values.theme = qt_config->value("theme", UISettings::themes[0].second).toString();
+    UISettings::values.theme = ReadSetting("theme", UISettings::themes[0].second).toString();
     UISettings::values.enable_discord_presence =
-        qt_config->value("enable_discord_presence", true).toBool();
+        ReadSetting("enable_discord_presence", true).toBool();
     UISettings::values.screenshot_resolution_factor =
-        static_cast<u16>(qt_config->value("screenshot_resolution_factor", 0).toUInt());
-    UISettings::values.select_user_on_boot =
-        qt_config->value("select_user_on_boot", false).toBool();
+        static_cast<u16>(ReadSetting("screenshot_resolution_factor", 0).toUInt());
+    UISettings::values.select_user_on_boot = ReadSetting("select_user_on_boot", false).toBool();
 
     qt_config->beginGroup("UIGameList");
-    UISettings::values.show_unknown = qt_config->value("show_unknown", true).toBool();
-    UISettings::values.show_add_ons = qt_config->value("show_add_ons", true).toBool();
-    UISettings::values.icon_size = qt_config->value("icon_size", 64).toUInt();
-    UISettings::values.row_1_text_id = qt_config->value("row_1_text_id", 3).toUInt();
-    UISettings::values.row_2_text_id = qt_config->value("row_2_text_id", 2).toUInt();
+    UISettings::values.show_unknown = ReadSetting("show_unknown", true).toBool();
+    UISettings::values.show_add_ons = ReadSetting("show_add_ons", true).toBool();
+    UISettings::values.icon_size = ReadSetting("icon_size", 64).toUInt();
+    UISettings::values.row_1_text_id = ReadSetting("row_1_text_id", 3).toUInt();
+    UISettings::values.row_2_text_id = ReadSetting("row_2_text_id", 2).toUInt();
     qt_config->endGroup();
 
     qt_config->beginGroup("UILayout");
-    UISettings::values.geometry = qt_config->value("geometry").toByteArray();
-    UISettings::values.state = qt_config->value("state").toByteArray();
-    UISettings::values.renderwindow_geometry =
-        qt_config->value("geometryRenderWindow").toByteArray();
-    UISettings::values.gamelist_header_state =
-        qt_config->value("gameListHeaderState").toByteArray();
+    UISettings::values.geometry = ReadSetting("geometry").toByteArray();
+    UISettings::values.state = ReadSetting("state").toByteArray();
+    UISettings::values.renderwindow_geometry = ReadSetting("geometryRenderWindow").toByteArray();
+    UISettings::values.gamelist_header_state = ReadSetting("gameListHeaderState").toByteArray();
     UISettings::values.microprofile_geometry =
-        qt_config->value("microProfileDialogGeometry").toByteArray();
+        ReadSetting("microProfileDialogGeometry").toByteArray();
     UISettings::values.microprofile_visible =
-        qt_config->value("microProfileDialogVisible", false).toBool();
+        ReadSetting("microProfileDialogVisible", false).toBool();
     qt_config->endGroup();
 
     qt_config->beginGroup("Paths");
-    UISettings::values.roms_path = qt_config->value("romsPath").toString();
-    UISettings::values.symbols_path = qt_config->value("symbolsPath").toString();
-    UISettings::values.gamedir = qt_config->value("gameListRootDir", ".").toString();
-    UISettings::values.gamedir_deepscan = qt_config->value("gameListDeepScan", false).toBool();
-    UISettings::values.recent_files = qt_config->value("recentFiles").toStringList();
+    UISettings::values.roms_path = ReadSetting("romsPath").toString();
+    UISettings::values.symbols_path = ReadSetting("symbolsPath").toString();
+    UISettings::values.gamedir = ReadSetting("gameListRootDir", ".").toString();
+    UISettings::values.gamedir_deepscan = ReadSetting("gameListDeepScan", false).toBool();
+    UISettings::values.recent_files = ReadSetting("recentFiles").toStringList();
     qt_config->endGroup();
 
     qt_config->beginGroup("Shortcuts");
@@ -520,8 +518,8 @@ void Config::ReadValues() {
             qt_config->beginGroup(hotkey);
             UISettings::values.shortcuts.emplace_back(UISettings::Shortcut(
                 group + "/" + hotkey,
-                UISettings::ContextualShortcut(qt_config->value("KeySeq").toString(),
-                                               qt_config->value("Context").toInt())));
+                UISettings::ContextualShortcut(ReadSetting("KeySeq").toString(),
+                                               ReadSetting("Context").toInt())));
             qt_config->endGroup();
         }
 
@@ -529,16 +527,16 @@ void Config::ReadValues() {
     }
     qt_config->endGroup();
 
-    UISettings::values.single_window_mode = qt_config->value("singleWindowMode", true).toBool();
-    UISettings::values.fullscreen = qt_config->value("fullscreen", false).toBool();
-    UISettings::values.display_titlebar = qt_config->value("displayTitleBars", true).toBool();
-    UISettings::values.show_filter_bar = qt_config->value("showFilterBar", true).toBool();
-    UISettings::values.show_status_bar = qt_config->value("showStatusBar", true).toBool();
-    UISettings::values.confirm_before_closing = qt_config->value("confirmClose", true).toBool();
-    UISettings::values.first_start = qt_config->value("firstStart", true).toBool();
-    UISettings::values.callout_flags = qt_config->value("calloutFlags", 0).toUInt();
-    UISettings::values.show_console = qt_config->value("showConsole", false).toBool();
-    UISettings::values.profile_index = qt_config->value("profileIndex", 0).toUInt();
+    UISettings::values.single_window_mode = ReadSetting("singleWindowMode", true).toBool();
+    UISettings::values.fullscreen = ReadSetting("fullscreen", false).toBool();
+    UISettings::values.display_titlebar = ReadSetting("displayTitleBars", true).toBool();
+    UISettings::values.show_filter_bar = ReadSetting("showFilterBar", true).toBool();
+    UISettings::values.show_status_bar = ReadSetting("showStatusBar", true).toBool();
+    UISettings::values.confirm_before_closing = ReadSetting("confirmClose", true).toBool();
+    UISettings::values.first_start = ReadSetting("firstStart", true).toBool();
+    UISettings::values.callout_flags = ReadSetting("calloutFlags", 0).toUInt();
+    UISettings::values.show_console = ReadSetting("showConsole", false).toBool();
+    UISettings::values.profile_index = ReadSetting("profileIndex", 0).toUInt();
 
     ApplyDefaultProfileIfInputInvalid();
 
@@ -549,62 +547,79 @@ void Config::SavePlayerValues() {
     for (std::size_t p = 0; p < Settings::values.players.size(); ++p) {
         const auto& player = Settings::values.players[p];
 
-        qt_config->setValue(QString("player_%1_connected").arg(p), player.connected);
-        qt_config->setValue(QString("player_%1_type").arg(p), static_cast<u8>(player.type));
+        WriteSetting(QString("player_%1_connected").arg(p), player.connected, false);
+        WriteSetting(QString("player_%1_type").arg(p), static_cast<u8>(player.type),
+                     static_cast<u8>(Settings::ControllerType::DualJoycon));
 
-        qt_config->setValue(QString("player_%1_body_color_left").arg(p), player.body_color_left);
-        qt_config->setValue(QString("player_%1_body_color_right").arg(p), player.body_color_right);
-        qt_config->setValue(QString("player_%1_button_color_left").arg(p),
-                            player.button_color_left);
-        qt_config->setValue(QString("player_%1_button_color_right").arg(p),
-                            player.button_color_right);
+        WriteSetting(QString("player_%1_body_color_left").arg(p), player.body_color_left,
+                     Settings::JOYCON_BODY_NEON_BLUE);
+        WriteSetting(QString("player_%1_body_color_right").arg(p), player.body_color_right,
+                     Settings::JOYCON_BODY_NEON_RED);
+        WriteSetting(QString("player_%1_button_color_left").arg(p), player.button_color_left,
+                     Settings::JOYCON_BUTTONS_NEON_BLUE);
+        WriteSetting(QString("player_%1_button_color_right").arg(p), player.button_color_right,
+                     Settings::JOYCON_BUTTONS_NEON_RED);
 
         for (int i = 0; i < Settings::NativeButton::NumButtons; ++i) {
-            qt_config->setValue(QString("player_%1_").arg(p) +
-                                    QString::fromStdString(Settings::NativeButton::mapping[i]),
-                                QString::fromStdString(player.buttons[i]));
+            std::string default_param = InputCommon::GenerateKeyboardParam(default_buttons[i]);
+            WriteSetting(QString("player_%1_").arg(p) +
+                             QString::fromStdString(Settings::NativeButton::mapping[i]),
+                         QString::fromStdString(player.buttons[i]),
+                         QString::fromStdString(default_param));
         }
         for (int i = 0; i < Settings::NativeAnalog::NumAnalogs; ++i) {
-            qt_config->setValue(QString("player_%1_").arg(p) +
-                                    QString::fromStdString(Settings::NativeAnalog::mapping[i]),
-                                QString::fromStdString(player.analogs[i]));
+            std::string default_param = InputCommon::GenerateAnalogParamFromKeys(
+                default_analogs[i][0], default_analogs[i][1], default_analogs[i][2],
+                default_analogs[i][3], default_analogs[i][4], 0.5f);
+            WriteSetting(QString("player_%1_").arg(p) +
+                             QString::fromStdString(Settings::NativeAnalog::mapping[i]),
+                         QString::fromStdString(player.analogs[i]),
+                         QString::fromStdString(default_param));
         }
     }
 }
 
 void Config::SaveDebugValues() {
-    qt_config->setValue("debug_pad_enabled", Settings::values.debug_pad_enabled);
+    WriteSetting("debug_pad_enabled", Settings::values.debug_pad_enabled, false);
     for (int i = 0; i < Settings::NativeButton::NumButtons; ++i) {
-        qt_config->setValue(QString("debug_pad_") +
-                                QString::fromStdString(Settings::NativeButton::mapping[i]),
-                            QString::fromStdString(Settings::values.debug_pad_buttons[i]));
+        std::string default_param = InputCommon::GenerateKeyboardParam(default_buttons[i]);
+        WriteSetting(QString("debug_pad_") +
+                         QString::fromStdString(Settings::NativeButton::mapping[i]),
+                     QString::fromStdString(Settings::values.debug_pad_buttons[i]),
+                     QString::fromStdString(default_param));
     }
     for (int i = 0; i < Settings::NativeAnalog::NumAnalogs; ++i) {
-        qt_config->setValue(QString("debug_pad_") +
-                                QString::fromStdString(Settings::NativeAnalog::mapping[i]),
-                            QString::fromStdString(Settings::values.debug_pad_analogs[i]));
+        std::string default_param = InputCommon::GenerateAnalogParamFromKeys(
+            default_analogs[i][0], default_analogs[i][1], default_analogs[i][2],
+            default_analogs[i][3], default_analogs[i][4], 0.5f);
+        WriteSetting(QString("debug_pad_") +
+                         QString::fromStdString(Settings::NativeAnalog::mapping[i]),
+                     QString::fromStdString(Settings::values.debug_pad_analogs[i]),
+                     QString::fromStdString(default_param));
     }
 }
 
 void Config::SaveMouseValues() {
-    qt_config->setValue("mouse_enabled", Settings::values.mouse_enabled);
+    WriteSetting("mouse_enabled", Settings::values.mouse_enabled, false);
 
     for (int i = 0; i < Settings::NativeMouseButton::NumMouseButtons; ++i) {
-        qt_config->setValue(QString("mouse_") +
-                                QString::fromStdString(Settings::NativeMouseButton::mapping[i]),
-                            QString::fromStdString(Settings::values.mouse_buttons[i]));
+        std::string default_param = InputCommon::GenerateKeyboardParam(default_mouse_buttons[i]);
+        WriteSetting(QString("mouse_") +
+                         QString::fromStdString(Settings::NativeMouseButton::mapping[i]),
+                     QString::fromStdString(Settings::values.mouse_buttons[i]),
+                     QString::fromStdString(default_param));
     }
 }
 
 void Config::SaveTouchscreenValues() {
-    qt_config->setValue("touchscreen_enabled", Settings::values.touchscreen.enabled);
-    qt_config->setValue("touchscreen_device",
-                        QString::fromStdString(Settings::values.touchscreen.device));
-
-    qt_config->setValue("touchscreen_finger", Settings::values.touchscreen.finger);
-    qt_config->setValue("touchscreen_angle", Settings::values.touchscreen.rotation_angle);
-    qt_config->setValue("touchscreen_diameter_x", Settings::values.touchscreen.diameter_x);
-    qt_config->setValue("touchscreen_diameter_y", Settings::values.touchscreen.diameter_y);
+    WriteSetting("touchscreen_enabled", Settings::values.touchscreen.enabled, true);
+    WriteSetting("touchscreen_device", QString::fromStdString(Settings::values.touchscreen.device),
+                 "engine:emu_window");
+
+    WriteSetting("touchscreen_finger", Settings::values.touchscreen.finger, 0);
+    WriteSetting("touchscreen_angle", Settings::values.touchscreen.rotation_angle, 0);
+    WriteSetting("touchscreen_diameter_x", Settings::values.touchscreen.diameter_x, 15);
+    WriteSetting("touchscreen_diameter_y", Settings::values.touchscreen.diameter_y, 15);
 }
 
 void Config::SaveValues() {
@@ -615,88 +630,96 @@ void Config::SaveValues() {
     SaveMouseValues();
     SaveTouchscreenValues();
 
-    qt_config->setValue("motion_device", QString::fromStdString(Settings::values.motion_device));
-    qt_config->setValue("keyboard_enabled", Settings::values.keyboard_enabled);
+    WriteSetting("motion_device", QString::fromStdString(Settings::values.motion_device),
+                 "engine:motion_emu,update_period:100,sensitivity:0.01");
+    WriteSetting("keyboard_enabled", Settings::values.keyboard_enabled, false);
 
     qt_config->endGroup();
 
     qt_config->beginGroup("Core");
-    qt_config->setValue("use_cpu_jit", Settings::values.use_cpu_jit);
-    qt_config->setValue("use_multi_core", Settings::values.use_multi_core);
+    WriteSetting("use_cpu_jit", Settings::values.use_cpu_jit, true);
+    WriteSetting("use_multi_core", Settings::values.use_multi_core, false);
     qt_config->endGroup();
 
     qt_config->beginGroup("Renderer");
-    qt_config->setValue("resolution_factor", (double)Settings::values.resolution_factor);
-    qt_config->setValue("use_frame_limit", Settings::values.use_frame_limit);
-    qt_config->setValue("frame_limit", Settings::values.frame_limit);
-    qt_config->setValue("use_accurate_gpu_emulation", Settings::values.use_accurate_gpu_emulation);
+    WriteSetting("resolution_factor", (double)Settings::values.resolution_factor, 1.0);
+    WriteSetting("use_frame_limit", Settings::values.use_frame_limit, true);
+    WriteSetting("frame_limit", Settings::values.frame_limit, 100);
+    WriteSetting("use_disk_shader_cache", Settings::values.use_disk_shader_cache, true);
+    WriteSetting("use_accurate_gpu_emulation", Settings::values.use_accurate_gpu_emulation, false);
+    WriteSetting("use_asynchronous_gpu_emulation", Settings::values.use_asynchronous_gpu_emulation,
+                 false);
 
     // Cast to double because Qt's written float values are not human-readable
-    qt_config->setValue("bg_red", (double)Settings::values.bg_red);
-    qt_config->setValue("bg_green", (double)Settings::values.bg_green);
-    qt_config->setValue("bg_blue", (double)Settings::values.bg_blue);
+    WriteSetting("bg_red", (double)Settings::values.bg_red, 0.0);
+    WriteSetting("bg_green", (double)Settings::values.bg_green, 0.0);
+    WriteSetting("bg_blue", (double)Settings::values.bg_blue, 0.0);
     qt_config->endGroup();
 
     qt_config->beginGroup("Audio");
-    qt_config->setValue("output_engine", QString::fromStdString(Settings::values.sink_id));
-    qt_config->setValue("enable_audio_stretching", Settings::values.enable_audio_stretching);
-    qt_config->setValue("output_device", QString::fromStdString(Settings::values.audio_device_id));
-    qt_config->setValue("volume", Settings::values.volume);
+    WriteSetting("output_engine", QString::fromStdString(Settings::values.sink_id), "auto");
+    WriteSetting("enable_audio_stretching", Settings::values.enable_audio_stretching, true);
+    WriteSetting("output_device", QString::fromStdString(Settings::values.audio_device_id), "auto");
+    WriteSetting("volume", Settings::values.volume, 1.0f);
     qt_config->endGroup();
 
     qt_config->beginGroup("Data Storage");
-    qt_config->setValue("use_virtual_sd", Settings::values.use_virtual_sd);
-    qt_config->setValue("nand_directory",
-                        QString::fromStdString(FileUtil::GetUserPath(FileUtil::UserPath::NANDDir)));
-    qt_config->setValue("sdmc_directory",
-                        QString::fromStdString(FileUtil::GetUserPath(FileUtil::UserPath::SDMCDir)));
+    WriteSetting("use_virtual_sd", Settings::values.use_virtual_sd, true);
+    WriteSetting("nand_directory",
+                 QString::fromStdString(FileUtil::GetUserPath(FileUtil::UserPath::NANDDir)),
+                 QString::fromStdString(FileUtil::GetUserPath(FileUtil::UserPath::NANDDir)));
+    WriteSetting("sdmc_directory",
+                 QString::fromStdString(FileUtil::GetUserPath(FileUtil::UserPath::SDMCDir)),
+                 QString::fromStdString(FileUtil::GetUserPath(FileUtil::UserPath::SDMCDir)));
     qt_config->endGroup();
 
     qt_config->beginGroup("System");
-    qt_config->setValue("use_docked_mode", Settings::values.use_docked_mode);
-    qt_config->setValue("enable_nfc", Settings::values.enable_nfc);
-    qt_config->setValue("current_user", Settings::values.current_user);
-    qt_config->setValue("language_index", Settings::values.language_index);
+    WriteSetting("use_docked_mode", Settings::values.use_docked_mode, false);
+    WriteSetting("enable_nfc", Settings::values.enable_nfc, true);
+    WriteSetting("current_user", Settings::values.current_user, 0);
+    WriteSetting("language_index", Settings::values.language_index, 1);
 
-    qt_config->setValue("rng_seed_enabled", Settings::values.rng_seed.has_value());
-    qt_config->setValue("rng_seed", Settings::values.rng_seed.value_or(0));
+    WriteSetting("rng_seed_enabled", Settings::values.rng_seed.has_value(), false);
+    WriteSetting("rng_seed", Settings::values.rng_seed.value_or(0), 0);
 
-    qt_config->setValue("custom_rtc_enabled", Settings::values.custom_rtc.has_value());
-    qt_config->setValue("custom_rtc",
-                        QVariant::fromValue<long long>(
-                            Settings::values.custom_rtc.value_or(std::chrono::seconds{}).count()));
+    WriteSetting("custom_rtc_enabled", Settings::values.custom_rtc.has_value(), false);
+    WriteSetting("custom_rtc",
+                 QVariant::fromValue<long long>(
+                     Settings::values.custom_rtc.value_or(std::chrono::seconds{}).count()),
+                 0);
 
     qt_config->endGroup();
 
     qt_config->beginGroup("Miscellaneous");
-    qt_config->setValue("log_filter", QString::fromStdString(Settings::values.log_filter));
-    qt_config->setValue("use_dev_keys", Settings::values.use_dev_keys);
+    WriteSetting("log_filter", QString::fromStdString(Settings::values.log_filter), "*:Info");
+    WriteSetting("use_dev_keys", Settings::values.use_dev_keys, false);
     qt_config->endGroup();
 
     qt_config->beginGroup("Debugging");
-    qt_config->setValue("use_gdbstub", Settings::values.use_gdbstub);
-    qt_config->setValue("gdbstub_port", Settings::values.gdbstub_port);
-    qt_config->setValue("program_args", QString::fromStdString(Settings::values.program_args));
-    qt_config->setValue("dump_exefs", Settings::values.dump_exefs);
-    qt_config->setValue("dump_nso", Settings::values.dump_nso);
+    WriteSetting("use_gdbstub", Settings::values.use_gdbstub, false);
+    WriteSetting("gdbstub_port", Settings::values.gdbstub_port, 24689);
+    WriteSetting("program_args", QString::fromStdString(Settings::values.program_args), "");
+    WriteSetting("dump_exefs", Settings::values.dump_exefs, false);
+    WriteSetting("dump_nso", Settings::values.dump_nso, false);
     qt_config->endGroup();
 
     qt_config->beginGroup("WebService");
-    qt_config->setValue("enable_telemetry", Settings::values.enable_telemetry);
-    qt_config->setValue("web_api_url", QString::fromStdString(Settings::values.web_api_url));
-    qt_config->setValue("yuzu_username", QString::fromStdString(Settings::values.yuzu_username));
-    qt_config->setValue("yuzu_token", QString::fromStdString(Settings::values.yuzu_token));
+    WriteSetting("enable_telemetry", Settings::values.enable_telemetry, true);
+    WriteSetting("web_api_url", QString::fromStdString(Settings::values.web_api_url),
+                 "https://api.yuzu-emu.org");
+    WriteSetting("yuzu_username", QString::fromStdString(Settings::values.yuzu_username));
+    WriteSetting("yuzu_token", QString::fromStdString(Settings::values.yuzu_token));
     qt_config->endGroup();
 
     qt_config->beginWriteArray("DisabledAddOns");
     int i = 0;
     for (const auto& elem : Settings::values.disabled_addons) {
         qt_config->setArrayIndex(i);
-        qt_config->setValue("title_id", QVariant::fromValue<u64>(elem.first));
+        WriteSetting("title_id", QVariant::fromValue<u64>(elem.first), 0);
         qt_config->beginWriteArray("disabled");
         for (std::size_t j = 0; j < elem.second.size(); ++j) {
             qt_config->setArrayIndex(static_cast<int>(j));
-            qt_config->setValue("d", QString::fromStdString(elem.second[j]));
+            WriteSetting("d", QString::fromStdString(elem.second[j]), "");
         }
         qt_config->endArray();
         ++i;
@@ -704,60 +727,86 @@ void Config::SaveValues() {
     qt_config->endArray();
 
     qt_config->beginGroup("UI");
-    qt_config->setValue("theme", UISettings::values.theme);
-    qt_config->setValue("enable_discord_presence", UISettings::values.enable_discord_presence);
-    qt_config->setValue("screenshot_resolution_factor",
-                        UISettings::values.screenshot_resolution_factor);
-    qt_config->setValue("select_user_on_boot", UISettings::values.select_user_on_boot);
+    WriteSetting("theme", UISettings::values.theme, UISettings::themes[0].second);
+    WriteSetting("enable_discord_presence", UISettings::values.enable_discord_presence, true);
+    WriteSetting("screenshot_resolution_factor", UISettings::values.screenshot_resolution_factor,
+                 0);
+    WriteSetting("select_user_on_boot", UISettings::values.select_user_on_boot, false);
 
     qt_config->beginGroup("UIGameList");
-    qt_config->setValue("show_unknown", UISettings::values.show_unknown);
-    qt_config->setValue("show_add_ons", UISettings::values.show_add_ons);
-    qt_config->setValue("icon_size", UISettings::values.icon_size);
-    qt_config->setValue("row_1_text_id", UISettings::values.row_1_text_id);
-    qt_config->setValue("row_2_text_id", UISettings::values.row_2_text_id);
+    WriteSetting("show_unknown", UISettings::values.show_unknown, true);
+    WriteSetting("show_add_ons", UISettings::values.show_add_ons, true);
+    WriteSetting("icon_size", UISettings::values.icon_size, 64);
+    WriteSetting("row_1_text_id", UISettings::values.row_1_text_id, 3);
+    WriteSetting("row_2_text_id", UISettings::values.row_2_text_id, 2);
     qt_config->endGroup();
 
     qt_config->beginGroup("UILayout");
-    qt_config->setValue("geometry", UISettings::values.geometry);
-    qt_config->setValue("state", UISettings::values.state);
-    qt_config->setValue("geometryRenderWindow", UISettings::values.renderwindow_geometry);
-    qt_config->setValue("gameListHeaderState", UISettings::values.gamelist_header_state);
-    qt_config->setValue("microProfileDialogGeometry", UISettings::values.microprofile_geometry);
-    qt_config->setValue("microProfileDialogVisible", UISettings::values.microprofile_visible);
+    WriteSetting("geometry", UISettings::values.geometry);
+    WriteSetting("state", UISettings::values.state);
+    WriteSetting("geometryRenderWindow", UISettings::values.renderwindow_geometry);
+    WriteSetting("gameListHeaderState", UISettings::values.gamelist_header_state);
+    WriteSetting("microProfileDialogGeometry", UISettings::values.microprofile_geometry);
+    WriteSetting("microProfileDialogVisible", UISettings::values.microprofile_visible, false);
     qt_config->endGroup();
 
     qt_config->beginGroup("Paths");
-    qt_config->setValue("romsPath", UISettings::values.roms_path);
-    qt_config->setValue("symbolsPath", UISettings::values.symbols_path);
-    qt_config->setValue("screenshotPath", UISettings::values.screenshot_path);
-    qt_config->setValue("gameListRootDir", UISettings::values.gamedir);
-    qt_config->setValue("gameListDeepScan", UISettings::values.gamedir_deepscan);
-    qt_config->setValue("recentFiles", UISettings::values.recent_files);
+    WriteSetting("romsPath", UISettings::values.roms_path);
+    WriteSetting("symbolsPath", UISettings::values.symbols_path);
+    WriteSetting("screenshotPath", UISettings::values.screenshot_path);
+    WriteSetting("gameListRootDir", UISettings::values.gamedir, ".");
+    WriteSetting("gameListDeepScan", UISettings::values.gamedir_deepscan, false);
+    WriteSetting("recentFiles", UISettings::values.recent_files);
     qt_config->endGroup();
 
     qt_config->beginGroup("Shortcuts");
     for (auto shortcut : UISettings::values.shortcuts) {
-        qt_config->setValue(shortcut.first + "/KeySeq", shortcut.second.first);
-        qt_config->setValue(shortcut.first + "/Context", shortcut.second.second);
+        WriteSetting(shortcut.first + "/KeySeq", shortcut.second.first);
+        WriteSetting(shortcut.first + "/Context", shortcut.second.second);
     }
     qt_config->endGroup();
 
-    qt_config->setValue("singleWindowMode", UISettings::values.single_window_mode);
-    qt_config->setValue("fullscreen", UISettings::values.fullscreen);
-    qt_config->setValue("displayTitleBars", UISettings::values.display_titlebar);
-    qt_config->setValue("showFilterBar", UISettings::values.show_filter_bar);
-    qt_config->setValue("showStatusBar", UISettings::values.show_status_bar);
-    qt_config->setValue("confirmClose", UISettings::values.confirm_before_closing);
-    qt_config->setValue("firstStart", UISettings::values.first_start);
-    qt_config->setValue("calloutFlags", UISettings::values.callout_flags);
-    qt_config->setValue("showConsole", UISettings::values.show_console);
-    qt_config->setValue("profileIndex", UISettings::values.profile_index);
+    WriteSetting("singleWindowMode", UISettings::values.single_window_mode, true);
+    WriteSetting("fullscreen", UISettings::values.fullscreen, false);
+    WriteSetting("displayTitleBars", UISettings::values.display_titlebar, true);
+    WriteSetting("showFilterBar", UISettings::values.show_filter_bar, true);
+    WriteSetting("showStatusBar", UISettings::values.show_status_bar, true);
+    WriteSetting("confirmClose", UISettings::values.confirm_before_closing, true);
+    WriteSetting("firstStart", UISettings::values.first_start, true);
+    WriteSetting("calloutFlags", UISettings::values.callout_flags, 0);
+    WriteSetting("showConsole", UISettings::values.show_console, false);
+    WriteSetting("profileIndex", UISettings::values.profile_index, 0);
     qt_config->endGroup();
 }
 
+QVariant Config::ReadSetting(const QString& name) const {
+    return qt_config->value(name);
+}
+
+QVariant Config::ReadSetting(const QString& name, const QVariant& default_value) const {
+    QVariant result;
+    if (qt_config->value(name + "/default", false).toBool()) {
+        result = default_value;
+    } else {
+        result = qt_config->value(name, default_value);
+    }
+    return result;
+}
+
+void Config::WriteSetting(const QString& name, const QVariant& value) {
+    qt_config->setValue(name, value);
+}
+
+void Config::WriteSetting(const QString& name, const QVariant& value,
+                          const QVariant& default_value) {
+    qt_config->setValue(name + "/default", value == default_value);
+    qt_config->setValue(name, value);
+}
+
 void Config::Reload() {
     ReadValues();
+    // To apply default value changes
+    SaveValues();
     Settings::Apply();
 }
 
diff --git a/src/yuzu/configuration/config.h b/src/yuzu/configuration/config.h
index e73ad19bb..f4185db18 100644
--- a/src/yuzu/configuration/config.h
+++ b/src/yuzu/configuration/config.h
@@ -42,6 +42,11 @@ private:
     void SaveMouseValues();
     void SaveTouchscreenValues();
 
+    QVariant ReadSetting(const QString& name) const;
+    QVariant ReadSetting(const QString& name, const QVariant& default_value) const;
+    void WriteSetting(const QString& name, const QVariant& value);
+    void WriteSetting(const QString& name, const QVariant& value, const QVariant& default_value);
+
     std::unique_ptr<QSettings> qt_config;
     std::string qt_config_loc;
 };
diff --git a/src/yuzu/configuration/configure_graphics.cpp b/src/yuzu/configuration/configure_graphics.cpp
index 8290b4384..dd1d67488 100644
--- a/src/yuzu/configuration/configure_graphics.cpp
+++ b/src/yuzu/configuration/configure_graphics.cpp
@@ -62,9 +62,7 @@ ConfigureGraphics::ConfigureGraphics(QWidget* parent)
         const QColor new_bg_color = QColorDialog::getColor(bg_color);
         if (!new_bg_color.isValid())
             return;
-        bg_color = new_bg_color;
-        ui->bg_button->setStyleSheet(
-            QString("QPushButton { background-color: %1 }").arg(bg_color.name()));
+        UpdateBackgroundColorButton(new_bg_color);
     });
 }
 
@@ -75,11 +73,12 @@ void ConfigureGraphics::setConfiguration() {
         static_cast<int>(FromResolutionFactor(Settings::values.resolution_factor)));
     ui->toggle_frame_limit->setChecked(Settings::values.use_frame_limit);
     ui->frame_limit->setValue(Settings::values.frame_limit);
+    ui->use_disk_shader_cache->setChecked(Settings::values.use_disk_shader_cache);
     ui->use_accurate_gpu_emulation->setChecked(Settings::values.use_accurate_gpu_emulation);
-    bg_color = QColor::fromRgbF(Settings::values.bg_red, Settings::values.bg_green,
-                                Settings::values.bg_blue);
-    ui->bg_button->setStyleSheet(
-        QString("QPushButton { background-color: %1 }").arg(bg_color.name()));
+    ui->use_asynchronous_gpu_emulation->setEnabled(!Core::System::GetInstance().IsPoweredOn());
+    ui->use_asynchronous_gpu_emulation->setChecked(Settings::values.use_asynchronous_gpu_emulation);
+    UpdateBackgroundColorButton(QColor::fromRgbF(Settings::values.bg_red, Settings::values.bg_green,
+                                                 Settings::values.bg_blue));
 }
 
 void ConfigureGraphics::applyConfiguration() {
@@ -87,8 +86,21 @@ void ConfigureGraphics::applyConfiguration() {
         ToResolutionFactor(static_cast<Resolution>(ui->resolution_factor_combobox->currentIndex()));
     Settings::values.use_frame_limit = ui->toggle_frame_limit->isChecked();
     Settings::values.frame_limit = ui->frame_limit->value();
+    Settings::values.use_disk_shader_cache = ui->use_disk_shader_cache->isChecked();
     Settings::values.use_accurate_gpu_emulation = ui->use_accurate_gpu_emulation->isChecked();
+    Settings::values.use_asynchronous_gpu_emulation =
+        ui->use_asynchronous_gpu_emulation->isChecked();
     Settings::values.bg_red = static_cast<float>(bg_color.redF());
     Settings::values.bg_green = static_cast<float>(bg_color.greenF());
     Settings::values.bg_blue = static_cast<float>(bg_color.blueF());
 }
+
+void ConfigureGraphics::UpdateBackgroundColorButton(QColor color) {
+    bg_color = color;
+
+    QPixmap pixmap(ui->bg_button->size());
+    pixmap.fill(bg_color);
+
+    const QIcon color_icon(pixmap);
+    ui->bg_button->setIcon(color_icon);
+}
diff --git a/src/yuzu/configuration/configure_graphics.h b/src/yuzu/configuration/configure_graphics.h
index d6ffc6fde..f2799822d 100644
--- a/src/yuzu/configuration/configure_graphics.h
+++ b/src/yuzu/configuration/configure_graphics.h
@@ -23,6 +23,8 @@ public:
 private:
     void setConfiguration();
 
+    void UpdateBackgroundColorButton(QColor color);
+
     std::unique_ptr<Ui::ConfigureGraphics> ui;
     QColor bg_color;
 };
diff --git a/src/yuzu/configuration/configure_graphics.ui b/src/yuzu/configuration/configure_graphics.ui
index e278cdd05..c6767e0ca 100644
--- a/src/yuzu/configuration/configure_graphics.ui
+++ b/src/yuzu/configuration/configure_graphics.ui
@@ -50,6 +50,13 @@
          </layout>
         </item>
         <item>
+         <widget class="QCheckBox" name="use_disk_shader_cache">
+          <property name="text">
+           <string>Use disk shader cache</string>
+          </property>
+         </widget>
+        </item>
+        <item>
          <widget class="QCheckBox" name="use_accurate_gpu_emulation">
           <property name="text">
            <string>Use accurate GPU emulation (slow)</string>
@@ -57,6 +64,13 @@
          </widget>
         </item>
         <item>
+         <widget class="QCheckBox" name="use_asynchronous_gpu_emulation">
+          <property name="text">
+           <string>Use asynchronous GPU emulation</string>
+          </property>
+         </widget>
+        </item>
+        <item>
          <layout class="QHBoxLayout" name="horizontalLayout">
           <item>
            <widget class="QLabel" name="label">
diff --git a/src/yuzu/debugger/graphics/graphics_surface.cpp b/src/yuzu/debugger/graphics/graphics_surface.cpp
index 209798521..29f01dfb2 100644
--- a/src/yuzu/debugger/graphics/graphics_surface.cpp
+++ b/src/yuzu/debugger/graphics/graphics_surface.cpp
@@ -383,13 +383,12 @@ void GraphicsSurfaceWidget::OnUpdate() {
     // TODO: Implement a good way to visualize alpha components!
 
     QImage decoded_image(surface_width, surface_height, QImage::Format_ARGB32);
-    std::optional<VAddr> address = gpu.MemoryManager().GpuToCpuAddress(surface_address);
 
     // TODO(bunnei): Will not work with BCn formats that swizzle 4x4 tiles.
     // Needs to be fixed if we plan to use this feature more, otherwise we may remove it.
     auto unswizzled_data = Tegra::Texture::UnswizzleTexture(
-        *address, 1, 1, Tegra::Texture::BytesPerPixel(surface_format), surface_width,
-        surface_height, 1U);
+        gpu.MemoryManager().GetPointer(surface_address), 1, 1,
+        Tegra::Texture::BytesPerPixel(surface_format), surface_width, surface_height, 1U);
 
     auto texture_data = Tegra::Texture::DecodeTexture(unswizzled_data, surface_format,
                                                       surface_width, surface_height);
@@ -398,7 +397,7 @@ void GraphicsSurfaceWidget::OnUpdate() {
 
     for (unsigned int y = 0; y < surface_height; ++y) {
         for (unsigned int x = 0; x < surface_width; ++x) {
-            Math::Vec4<u8> color;
+            Common::Vec4<u8> color;
             color[0] = texture_data[x + y * surface_width + 0];
             color[1] = texture_data[x + y * surface_width + 1];
             color[2] = texture_data[x + y * surface_width + 2];
diff --git a/src/yuzu/debugger/wait_tree.cpp b/src/yuzu/debugger/wait_tree.cpp
index f50225d5f..06ad74ffe 100644
--- a/src/yuzu/debugger/wait_tree.cpp
+++ b/src/yuzu/debugger/wait_tree.cpp
@@ -81,9 +81,8 @@ QString WaitTreeText::GetText() const {
     return text;
 }
 
-WaitTreeMutexInfo::WaitTreeMutexInfo(VAddr mutex_address) : mutex_address(mutex_address) {
-    const auto& handle_table = Core::CurrentProcess()->GetHandleTable();
-
+WaitTreeMutexInfo::WaitTreeMutexInfo(VAddr mutex_address, const Kernel::HandleTable& handle_table)
+    : mutex_address(mutex_address) {
     mutex_value = Memory::Read32(mutex_address);
     owner_handle = static_cast<Kernel::Handle>(mutex_value & Kernel::Mutex::MutexOwnerMask);
     owner = handle_table.Get<Kernel::Thread>(owner_handle);
@@ -316,7 +315,8 @@ std::vector<std::unique_ptr<WaitTreeItem>> WaitTreeThread::GetChildren() const {
 
     const VAddr mutex_wait_address = thread.GetMutexWaitAddress();
     if (mutex_wait_address != 0) {
-        list.push_back(std::make_unique<WaitTreeMutexInfo>(mutex_wait_address));
+        const auto& handle_table = thread.GetOwnerProcess()->GetHandleTable();
+        list.push_back(std::make_unique<WaitTreeMutexInfo>(mutex_wait_address, handle_table));
     } else {
         list.push_back(std::make_unique<WaitTreeText>(tr("not waiting for mutex")));
     }
diff --git a/src/yuzu/debugger/wait_tree.h b/src/yuzu/debugger/wait_tree.h
index 365c3dbfe..62886609d 100644
--- a/src/yuzu/debugger/wait_tree.h
+++ b/src/yuzu/debugger/wait_tree.h
@@ -17,6 +17,7 @@
 class EmuThread;
 
 namespace Kernel {
+class HandleTable;
 class ReadableEvent;
 class WaitObject;
 class Thread;
@@ -72,7 +73,7 @@ public:
 class WaitTreeMutexInfo : public WaitTreeExpandableItem {
     Q_OBJECT
 public:
-    explicit WaitTreeMutexInfo(VAddr mutex_address);
+    explicit WaitTreeMutexInfo(VAddr mutex_address, const Kernel::HandleTable& handle_table);
     ~WaitTreeMutexInfo() override;
 
     QString GetText() const override;
diff --git a/src/yuzu/loading_screen.cpp b/src/yuzu/loading_screen.cpp
index 907aac4f1..86f6d0165 100644
--- a/src/yuzu/loading_screen.cpp
+++ b/src/yuzu/loading_screen.cpp
@@ -43,6 +43,7 @@ QProgressBar {
 }
 QProgressBar::chunk {
   background-color: #0ab9e6;
+  width: 1px;
 })";
 
 constexpr const char PROGRESSBAR_STYLE_BUILD[] = R"(
@@ -53,7 +54,8 @@ QProgressBar {
   padding: 2px;
 }
 QProgressBar::chunk {
- background-color: #ff3c28;
+  background-color: #ff3c28;
+  width: 1px;
 })";
 
 constexpr const char PROGRESSBAR_STYLE_COMPLETE[] = R"(
diff --git a/src/yuzu/loading_screen.ui b/src/yuzu/loading_screen.ui
index a67d273fd..820b47536 100644
--- a/src/yuzu/loading_screen.ui
+++ b/src/yuzu/loading_screen.ui
@@ -132,7 +132,7 @@ border-radius: 15px;
 font: 75 15pt &quot;Arial&quot;;</string>
           </property>
           <property name="text">
-           <string>Stage 1 of 2. Estimate Time 5m 4s</string>
+           <string>Estimated Time 5m 4s</string>
           </property>
          </widget>
         </item>
@@ -146,6 +146,9 @@ font: 75 15pt &quot;Arial&quot;;</string>
         <property name="text">
          <string/>
         </property>
+        <property name="alignment">
+         <set>Qt::AlignCenter</set>
+        </property>
         <property name="margin">
          <number>30</number>
         </property>
diff --git a/src/yuzu/main.cpp b/src/yuzu/main.cpp
index ab403b3ac..41ba3c4c6 100644
--- a/src/yuzu/main.cpp
+++ b/src/yuzu/main.cpp
@@ -11,6 +11,7 @@
 #include "applets/profile_select.h"
 #include "applets/software_keyboard.h"
 #include "applets/web_browser.h"
+#include "configuration/configure_input.h"
 #include "configuration/configure_per_general.h"
 #include "core/file_sys/vfs.h"
 #include "core/file_sys/vfs_real.h"
@@ -339,6 +340,11 @@ void GMainWindow::WebBrowserOpenPage(std::string_view filename, std::string_view
                 .arg(QString::fromStdString(std::to_string(key_code))));
     };
 
+    QMessageBox::information(
+        this, tr("Exit"),
+        tr("To exit the web application, use the game provided controls to select exit, select the "
+           "'Exit Web Applet' option in the menu bar, or press the 'Enter' key."));
+
     bool running_exit_check = false;
     while (!finished) {
         QApplication::processEvents();
@@ -522,6 +528,7 @@ void GMainWindow::InitializeHotkeys() {
                                    Qt::ApplicationShortcut);
     hotkey_registry.RegisterHotkey("Main Window", "Capture Screenshot",
                                    QKeySequence(QKeySequence::Print));
+    hotkey_registry.RegisterHotkey("Main Window", "Change Docked Mode", QKeySequence(Qt::Key_F10));
 
     hotkey_registry.LoadHotkeys();
 
@@ -561,7 +568,10 @@ void GMainWindow::InitializeHotkeys() {
                 Settings::values.use_frame_limit = !Settings::values.use_frame_limit;
                 UpdateStatusBar();
             });
-    constexpr u16 SPEED_LIMIT_STEP = 5;
+    // TODO: Remove this comment/static whenever the next major release of
+    // MSVC occurs and we make it a requirement (see:
+    // https://developercommunity.visualstudio.com/content/problem/93922/constexprs-are-trying-to-be-captured-in-lambda-fun.html)
+    static constexpr u16 SPEED_LIMIT_STEP = 5;
     connect(hotkey_registry.GetHotkey("Main Window", "Increase Speed Limit", this),
             &QShortcut::activated, this, [&] {
                 if (Settings::values.frame_limit < 9999 - SPEED_LIMIT_STEP) {
@@ -588,6 +598,12 @@ void GMainWindow::InitializeHotkeys() {
                     OnCaptureScreenshot();
                 }
             });
+    connect(hotkey_registry.GetHotkey("Main Window", "Change Docked Mode", this),
+            &QShortcut::activated, this, [&] {
+                Settings::values.use_docked_mode = !Settings::values.use_docked_mode;
+                OnDockedModeChanged(!Settings::values.use_docked_mode,
+                                    Settings::values.use_docked_mode);
+            });
 }
 
 void GMainWindow::SetDefaultUIGeometry() {
@@ -846,7 +862,7 @@ bool GMainWindow::LoadROM(const QString& filename) {
     }
     game_path = filename;
 
-    Core::Telemetry().AddField(Telemetry::FieldType::App, "Frontend", "Qt");
+    system.TelemetrySession().AddField(Telemetry::FieldType::App, "Frontend", "Qt");
     return true;
 }
 
@@ -887,6 +903,9 @@ void GMainWindow::BootGame(const QString& filename) {
     connect(emu_thread.get(), &EmuThread::DebugModeLeft, waitTreeWidget,
             &WaitTreeWidget::OnDebugModeLeft, Qt::BlockingQueuedConnection);
 
+    connect(emu_thread.get(), &EmuThread::LoadProgress, loading_screen,
+            &LoadingScreen::OnLoadProgress, Qt::QueuedConnection);
+
     // Update the GUI
     if (ui.action_Single_Window_Mode->isChecked()) {
         game_list->hide();
@@ -1682,12 +1701,16 @@ void GMainWindow::OnToggleFilterBar() {
 
 void GMainWindow::OnCaptureScreenshot() {
     OnPauseGame();
-    const QString path =
-        QFileDialog::getSaveFileName(this, tr("Capture Screenshot"),
-                                     UISettings::values.screenshot_path, tr("PNG Image (*.png)"));
-    if (!path.isEmpty()) {
-        UISettings::values.screenshot_path = QFileInfo(path).path();
-        render_window->CaptureScreenshot(UISettings::values.screenshot_resolution_factor, path);
+    QFileDialog png_dialog(this, tr("Capture Screenshot"), UISettings::values.screenshot_path,
+                           tr("PNG Image (*.png)"));
+    png_dialog.setAcceptMode(QFileDialog::AcceptSave);
+    png_dialog.setDefaultSuffix("png");
+    if (png_dialog.exec()) {
+        const QString path = png_dialog.selectedFiles().first();
+        if (!path.isEmpty()) {
+            UISettings::values.screenshot_path = QFileInfo(path).path();
+            render_window->CaptureScreenshot(UISettings::values.screenshot_resolution_factor, path);
+        }
     }
     OnStartGame();
 }
diff --git a/src/yuzu_cmd/config.cpp b/src/yuzu_cmd/config.cpp
index 7a77f76e8..32e78049c 100644
--- a/src/yuzu_cmd/config.cpp
+++ b/src/yuzu_cmd/config.cpp
@@ -346,23 +346,28 @@ void Config::ReadValues() {
 
     // Renderer
     Settings::values.resolution_factor =
-        (float)sdl2_config->GetReal("Renderer", "resolution_factor", 1.0);
+        static_cast<float>(sdl2_config->GetReal("Renderer", "resolution_factor", 1.0));
     Settings::values.use_frame_limit = sdl2_config->GetBoolean("Renderer", "use_frame_limit", true);
     Settings::values.frame_limit =
         static_cast<u16>(sdl2_config->GetInteger("Renderer", "frame_limit", 100));
+    Settings::values.use_disk_shader_cache =
+        sdl2_config->GetBoolean("Renderer", "use_disk_shader_cache", false);
     Settings::values.use_accurate_gpu_emulation =
         sdl2_config->GetBoolean("Renderer", "use_accurate_gpu_emulation", false);
+    Settings::values.use_asynchronous_gpu_emulation =
+        sdl2_config->GetBoolean("Renderer", "use_asynchronous_gpu_emulation", false);
 
-    Settings::values.bg_red = (float)sdl2_config->GetReal("Renderer", "bg_red", 0.0);
-    Settings::values.bg_green = (float)sdl2_config->GetReal("Renderer", "bg_green", 0.0);
-    Settings::values.bg_blue = (float)sdl2_config->GetReal("Renderer", "bg_blue", 0.0);
+    Settings::values.bg_red = static_cast<float>(sdl2_config->GetReal("Renderer", "bg_red", 0.0));
+    Settings::values.bg_green =
+        static_cast<float>(sdl2_config->GetReal("Renderer", "bg_green", 0.0));
+    Settings::values.bg_blue = static_cast<float>(sdl2_config->GetReal("Renderer", "bg_blue", 0.0));
 
     // Audio
     Settings::values.sink_id = sdl2_config->Get("Audio", "output_engine", "auto");
     Settings::values.enable_audio_stretching =
         sdl2_config->GetBoolean("Audio", "enable_audio_stretching", true);
     Settings::values.audio_device_id = sdl2_config->Get("Audio", "output_device", "auto");
-    Settings::values.volume = sdl2_config->GetReal("Audio", "volume", 1);
+    Settings::values.volume = static_cast<float>(sdl2_config->GetReal("Audio", "volume", 1));
 
     Settings::values.language_index = sdl2_config->GetInteger("System", "language_index", 1);
 
diff --git a/src/yuzu_cmd/default_ini.h b/src/yuzu_cmd/default_ini.h
index ba51a4a51..6538af098 100644
--- a/src/yuzu_cmd/default_ini.h
+++ b/src/yuzu_cmd/default_ini.h
@@ -110,10 +110,18 @@ use_frame_limit =
 # 1 - 9999: Speed limit as a percentage of target game speed. 100 (default)
 frame_limit =
 
+# Whether to use disk based shader cache
+# 0 (default): Off, 1 : On
+use_disk_shader_cache =
+
 # Whether to use accurate GPU emulation
 # 0 (default): Off (fast), 1 : On (slow)
 use_accurate_gpu_emulation =
 
+# Whether to use asynchronous GPU emulation
+# 0 : Off (slow), 1 (default): On (fast)
+use_asynchronous_gpu_emulation =
+
 # The clear color for the renderer. What shows up on the sides of the bottom screen.
 # Must be in range of 0.0-1.0. Defaults to 1.0 for all.
 bg_red =
diff --git a/src/yuzu_cmd/emu_window/emu_window_sdl2.cpp b/src/yuzu_cmd/emu_window/emu_window_sdl2.cpp
index 7df8eff53..de7a26e14 100644
--- a/src/yuzu_cmd/emu_window/emu_window_sdl2.cpp
+++ b/src/yuzu_cmd/emu_window/emu_window_sdl2.cpp
@@ -135,16 +135,16 @@ bool EmuWindow_SDL2::SupportsRequiredGLExtensions() {
 }
 
 EmuWindow_SDL2::EmuWindow_SDL2(bool fullscreen) {
-    InputCommon::Init();
-
-    SDL_SetMainReady();
-
     // Initialize the window
     if (SDL_Init(SDL_INIT_VIDEO | SDL_INIT_JOYSTICK) < 0) {
         LOG_CRITICAL(Frontend, "Failed to initialize SDL2! Exiting...");
         exit(1);
     }
 
+    InputCommon::Init();
+
+    SDL_SetMainReady();
+
     SDL_GL_SetAttribute(SDL_GL_CONTEXT_MAJOR_VERSION, 4);
     SDL_GL_SetAttribute(SDL_GL_CONTEXT_MINOR_VERSION, 3);
     SDL_GL_SetAttribute(SDL_GL_CONTEXT_PROFILE_MASK, SDL_GL_CONTEXT_PROFILE_CORE);
@@ -201,11 +201,9 @@ EmuWindow_SDL2::EmuWindow_SDL2(bool fullscreen) {
 }
 
 EmuWindow_SDL2::~EmuWindow_SDL2() {
-    InputCommon::SDL::CloseSDLJoysticks();
+    InputCommon::Shutdown();
     SDL_GL_DeleteContext(gl_context);
     SDL_Quit();
-
-    InputCommon::Shutdown();
 }
 
 void EmuWindow_SDL2::SwapBuffers() {
@@ -262,7 +260,6 @@ void EmuWindow_SDL2::PollEvents() {
             is_open = false;
             break;
         default:
-            InputCommon::SDL::HandleGameControllerEvent(event);
             break;
         }
     }
diff --git a/src/yuzu_cmd/yuzu.cpp b/src/yuzu_cmd/yuzu.cpp
index 806127b12..245f25847 100644
--- a/src/yuzu_cmd/yuzu.cpp
+++ b/src/yuzu_cmd/yuzu.cpp
@@ -28,6 +28,7 @@
 #include "core/loader/loader.h"
 #include "core/settings.h"
 #include "core/telemetry_session.h"
+#include "video_core/renderer_base.h"
 #include "yuzu_cmd/config.h"
 #include "yuzu_cmd/emu_window/emu_window_sdl2.h"
 
@@ -113,9 +114,9 @@ int main(int argc, char** argv) {
     };
 
     while (optind < argc) {
-        char arg = getopt_long(argc, argv, "g:fhvp::", long_options, &option_index);
+        int arg = getopt_long(argc, argv, "g:fhvp::", long_options, &option_index);
         if (arg != -1) {
-            switch (arg) {
+            switch (static_cast<char>(arg)) {
             case 'g':
                 errno = 0;
                 gdb_port = strtoul(optarg, &endarg, 0);
@@ -215,7 +216,9 @@ int main(int argc, char** argv) {
         }
     }
 
-    Core::Telemetry().AddField(Telemetry::FieldType::App, "Frontend", "SDL");
+    system.TelemetrySession().AddField(Telemetry::FieldType::App, "Frontend", "SDL");
+
+    system.Renderer().Rasterizer().LoadDiskResources();
 
     while (emu_window->IsOpen()) {
         system.RunLoop();